1321 files changed, 49769 insertions, 29554 deletions

diff --git a/.git-blame-ignore-revs b/.git-blame-ignore-revs
index 01569c6..ea789ab 100644
--- a/.git-blame-ignore-revs
+++ b/.git-blame-ignore-revs
@@ -106,3 +106,40 @@ d33bf2e9df578ff7e44fd22504d6ad5a122b7ee6
 ce46adb8b7ce645353eccaedf31ed9765dab77bb
 68070f908bb7ac5f0b5fa9722caa504ecf723f6b
 5213c57cb1f0d78aad9a253b7f6a2b62ff4c7859
+
+# [mlir] Update create method
+9e7834cadf48292b5d127d6d98f9e6d565ed5d9a
+284a5c2c0b97edddf255ea210f939203ad3d09f2
+c090ed53fb73f59cf221f5610430af8047758117
+fcbcfe44cff00101a6a98a73971398eb8dd87710
+258daf539583b80e0217d1d87941412d65cf16aa
+c610b244937ed847b0275ccb038c0f2d36310b4a
+b58ad3650f2195117f484d551ffbada27e7d1e14
+258d04c810ab10f101324cbf1fe3c7be65eb1938
+a6bf40d1c6cf010b3ad90bf7f410983453f4deb2
+dcfc853c51aecf6538182378c016f8e1604e7e97
+3f74334c38120bbdefac012d478dfce8e4eb0906
+a636b7bfdd1d8304b78e8b42ec900a21736d4afb
+75aa7065dcf653de7870758cd502a7c714f4bcd7
+2f5312563fd5cb2e355ec49109f3e63875337c7c
+967626b842551ecd997c0d10eb68c3015b63a3d7
+588845defd09359a8b87db339b563af848cf45a7
+b0434925c98c9a8906afea60a1304c870b1f574a
+8fff238b2c363b036ce9e7bf7abab3acafc87ab2
+38976a03cd367b27437e0d1e81c0ccaee2777b47
+eaa67a3cf041009ae33a45159d0465262c3af5dc
+b0312be6aa664e4cb9abec6d080e971493093d05
+2736fbd8324bf21a130c8abd4bd0e7d3aa840ac1
+4ae9fdca8af095afd91705f8dd143e93b304b6fb
+f904cdd6c3049e605d24ed17680e80e7133908a0
+972ac59c9af4ad47af0b3542ae936b3470727e5f
+7b787965431e666858fdf66db25ee5a129833927
+c3823af156b517d926a56e3d0d585e2a15720e96
+dce6679cf5cbbdaffb9c2b51dc762c5c6689ea78
+9844ba6d9740206129b52633c555f767eaa45581
+5547c6cd03ddddd405a09e51624e1f19955a85b1
+a3a007ad5fa20abc90ead4e1030b481bf109b4cf
+46f6df0848ea04449c6179ecdedc404ee5b5cf11
+b7e332d3f59f567b1999fbcc660d7837cba8e406
+6056f942abe83b05406df8b04e95ec37a3d160b5
+906295b8a31c8dac5aa845864c0bca9f02f86184
diff --git a/.github/CODEOWNERS b/.github/CODEOWNERS
index dd7e177..23e13fd 100644
--- a/.github/CODEOWNERS
+++ b/.github/CODEOWNERS
@@ -120,8 +120,9 @@
 /mlir/**/Index* @mogball
 
 # MLIR Python Bindings
-/mlir/test/python/ @ftynse @makslevental @stellaraccident
-/mlir/python/ @ftynse @makslevental @stellaraccident
+/mlir/test/python/ @ftynse @makslevental @stellaraccident @rolfmorel
+/mlir/python/ @ftynse @makslevental @stellaraccident @rolfmorel
+/mlir/lib/Bindings/Python @makslevental @rolfmorel
 
 # MLIR Mem2Reg/SROA
 /mlir/**/Transforms/Mem2Reg.* @moxinilian
diff --git a/.github/workflows/containers/github-action-ci/Dockerfile b/.github/workflows/containers/github-action-ci/Dockerfile
index a6ee62307..df39071 100644
--- a/.github/workflows/containers/github-action-ci/Dockerfile
+++ b/.github/workflows/containers/github-action-ci/Dockerfile
@@ -58,10 +58,8 @@ RUN apt-get update && \
     python3-psutil \
     sudo \
     # These are needed by the premerge pipeline. Pip is used to install
-    # dependent python packages and ccache is used for build caching. File and
-    # tzdata are used for tests.
+    # dependent python packages. File and tzdata are used for tests.
     python3-pip \
-    ccache \
     file \
     tzdata && \
     apt-get clean && \
diff --git a/.github/workflows/libcxx-build-containers.yml b/.github/workflows/libcxx-build-containers.yml
index 564a7934..f432e3d 100644
--- a/.github/workflows/libcxx-build-containers.yml
+++ b/.github/workflows/libcxx-build-containers.yml
@@ -18,8 +18,6 @@ on:
       - 'libcxx/utils/ci/**'
       - '.github/workflows/libcxx-build-containers.yml'
   pull_request:
-    branches:
-      - main
     paths:
       - 'libcxx/utils/ci/**'
       - '.github/workflows/libcxx-build-containers.yml'
diff --git a/.github/workflows/release-asset-audit.yml b/.github/workflows/release-asset-audit.yml
index 8112d8a..7a1f232 100644
--- a/.github/workflows/release-asset-audit.yml
+++ b/.github/workflows/release-asset-audit.yml
@@ -22,7 +22,12 @@ jobs:
     runs-on: ubuntu-24.04
     if: github.repository == 'llvm/llvm-project'
     steps:
-      - uses: actions/checkout@a5ac7e51b41094c92402da3b24376905380afc29 #v4.1.6
+      - name: Checkout LLVM
+        uses: actions/checkout@11bd71901bbe5b1630ceea73d27597364c9af683 # v4.2.2
+        with:
+          sparse-checkout: |
+            .github/workflows/release-asset-audit.py
+            llvm/utils/git/requirements.txt
       - name: "Run Audit Script"
         env:
           GITHUB_TOKEN: ${{ github.token }}
diff --git a/bolt/test/X86/debug-fission-single-convert.s b/bolt/test/X86/debug-fission-single-convert.s
index 02c92902..ea05ccd 100644
--- a/bolt/test/X86/debug-fission-single-convert.s
+++ b/bolt/test/X86/debug-fission-single-convert.s
@@ -14,14 +14,15 @@
 # RUN:   -nostartfiles \
 # RUN:   -Wl,--script=%p/Inputs/debug-fission-script.txt \
 # RUN:   %t.o -o %t.exe
+# RUN: mkdir -p %t.dwarf-output
 # RUN: llvm-bolt %t.exe \
 # RUN:   --reorder-blocks=reverse \
 # RUN:   --update-debug-sections \
-# RUN:   --dwarf-output-path=%T \
+# RUN:   --dwarf-output-path=%t.dwarf-output \
 # RUN:   --always-convert-to-ranges=true \
 # RUN:   -o %t.bolt.1.exe 2>&1 | FileCheck %s
 # RUN: llvm-dwarfdump --show-form --verbose --debug-ranges %t.bolt.1.exe &> %tAddrIndexTest
-# RUN: not llvm-dwarfdump --show-form --verbose --debug-info %T/debug-fission-simple-convert.dwo0.dwo >> %tAddrIndexTest
+# RUN: not llvm-dwarfdump --show-form --verbose --debug-info %t.dwarf-output/debug-fission-simple-convert.dwo0.dwo >> %tAddrIndexTest
 # RUN: cat %tAddrIndexTest | FileCheck %s --check-prefix=CHECK-DWO-DWO
 # RUN: llvm-dwarfdump --show-form --verbose   --debug-addr  %t.bolt.1.exe | FileCheck %s --check-prefix=CHECK-ADDR-SEC
 
diff --git a/bolt/test/X86/debug-fission-single.s b/bolt/test/X86/debug-fission-single.s
index 1aa502f..7ff53df 100644
--- a/bolt/test/X86/debug-fission-single.s
+++ b/bolt/test/X86/debug-fission-single.s
@@ -14,13 +14,14 @@
 # RUN:   -nostartfiles \
 # RUN:   -Wl,--script=%p/Inputs/debug-fission-script.txt \
 # RUN:   %t.o -o %t.exe
+# RUN: mkdir -p %t.dwarf-output
 # RUN: llvm-bolt %t.exe \
 # RUN:   --reorder-blocks=reverse \
 # RUN:   --update-debug-sections \
-# RUN:   --dwarf-output-path=%T \
+# RUN:   --dwarf-output-path=%t.dwarf-output \
 # RUN:   -o %t.bolt.1.exe 2>&1 | FileCheck %s
 # RUN: llvm-dwarfdump --show-form --verbose --debug-ranges %t.bolt.1.exe &> %tAddrIndexTest
-# RUN: llvm-dwarfdump --show-form --verbose --debug-info %T/debug-fission-simple.dwo0.dwo >> %tAddrIndexTest
+# RUN: llvm-dwarfdump --show-form --verbose --debug-info %t.dwarf-output/debug-fission-simple.dwo0.dwo >> %tAddrIndexTest
 # RUN: cat %tAddrIndexTest | FileCheck %s --check-prefix=CHECK-DWO-DWO
 # RUN: llvm-dwarfdump --show-form --verbose   --debug-addr  %t.bolt.1.exe | FileCheck %s --check-prefix=CHECK-ADDR-SEC
 
diff --git a/bolt/test/X86/inlined-function-mixed.test b/bolt/test/X86/inlined-function-mixed.test
index 9f6ef39..4fc1594 100644
--- a/bolt/test/X86/inlined-function-mixed.test
+++ b/bolt/test/X86/inlined-function-mixed.test
@@ -1,9 +1,9 @@
 ## Make sure inlining from a unit with debug info into unit without
 ## debug info does not cause a crash.
 
-RUN: %clangxx %cxxflags %S/Inputs/inlined.cpp -c -o %T/inlined.o
-RUN: %clangxx %cxxflags %S/Inputs/inlinee.cpp -c -o %T/inlinee.o -g
-RUN: %clangxx %cxxflags %T/inlined.o %T/inlinee.o -o %t
+RUN: %clangxx %cxxflags %S/Inputs/inlined.cpp -c -o %t.inlined.o
+RUN: %clangxx %cxxflags %S/Inputs/inlinee.cpp -c -o %t.inlinee.o -g
+RUN: %clangxx %cxxflags %t.inlined.o %t.inlinee.o -o %t
 
 RUN: llvm-bolt %t -o %t.bolt --update-debug-sections --reorder-blocks=reverse \
 RUN:   --inline-small-functions --force-inline=main | FileCheck %s
diff --git a/clang-tools-extra/clang-doc/JSONGenerator.cpp b/clang-tools-extra/clang-doc/JSONGenerator.cpp
index 92a4117..599b381 100644
--- a/clang-tools-extra/clang-doc/JSONGenerator.cpp
+++ b/clang-tools-extra/clang-doc/JSONGenerator.cpp
@@ -103,6 +103,18 @@ static json::Value extractTextComments(Object *ParagraphComment) {
   return *ParagraphComment->get("Children");
 }
 
+static json::Value extractVerbatimComments(json::Array VerbatimLines) {
+  json::Value TextArray = json::Array();
+  auto &TextArrayRef = *TextArray.getAsArray();
+  for (auto &Line : VerbatimLines)
+    TextArrayRef.push_back(*Line.getAsObject()
+                                ->get("VerbatimBlockLineComment")
+                                ->getAsObject()
+                                ->get("Text"));
+
+  return TextArray;
+}
+
 static Object serializeComment(const CommentInfo &I, Object &Description) {
   // taken from PR #142273
   Object Obj = Object();
@@ -126,6 +138,8 @@ static Object serializeComment(const CommentInfo &I, Object &Description) {
     auto TextCommentsArray = extractTextComments(CARef.front().getAsObject());
     if (I.Name == "brief")
       insertComment(Description, TextCommentsArray, "BriefComments");
+    else if (I.Name == "return")
+      insertComment(Description, TextCommentsArray, "ReturnComments");
     return Obj;
   }
 
@@ -147,19 +161,19 @@ static Object serializeComment(const CommentInfo &I, Object &Description) {
     Child.insert({"ParamName", I.ParamName});
     Child.insert({"Direction", I.Direction});
     Child.insert({"Explicit", I.Explicit});
-    Child.insert({"Children", ChildArr});
-    Obj.insert({commentKindToString(I.Kind), ChildVal});
+    auto TextCommentsArray = extractTextComments(CARef.front().getAsObject());
+    Child.insert({"Children", TextCommentsArray});
+    if (I.Kind == CommentKind::CK_ParamCommandComment)
+      insertComment(Description, ChildVal, "ParamComments");
     return Obj;
   }
 
   case CommentKind::CK_VerbatimBlockComment: {
-    Child.insert({"Text", I.Text});
-    if (!I.CloseName.empty())
-      Child.insert({"CloseName", I.CloseName});
-    Child.insert({"Children", ChildArr});
-    if (I.CloseName == "endcode")
-      insertComment(Description, ChildVal, "CodeComments");
-    else if (I.CloseName == "endverbatim")
+    if (I.CloseName == "endcode") {
+      // We don't support \code language specification
+      auto TextCommentsArray = extractVerbatimComments(CARef);
+      insertComment(Description, TextCommentsArray, "CodeComments");
+    } else if (I.CloseName == "endverbatim")
       insertComment(Description, ChildVal, "VerbatimComments");
     return Obj;
   }
diff --git a/clang-tools-extra/clang-doc/assets/clang-doc-mustache.css b/clang-tools-extra/clang-doc/assets/clang-doc-mustache.css
index a885a36..e555ee7 100644
--- a/clang-tools-extra/clang-doc/assets/clang-doc-mustache.css
+++ b/clang-tools-extra/clang-doc/assets/clang-doc-mustache.css
@@ -469,3 +469,7 @@ a, a:visited, a:hover, a:active {
     text-decoration: none;
     color: inherit;
 }
+
+.code-block {
+  white-space: pre-line;
+}
diff --git a/clang-tools-extra/clang-doc/assets/comment-template.mustache b/clang-tools-extra/clang-doc/assets/comment-template.mustache
index f2edb1b..4e38e5f 100644
--- a/clang-tools-extra/clang-doc/assets/comment-template.mustache
+++ b/clang-tools-extra/clang-doc/assets/comment-template.mustache
@@ -24,6 +24,36 @@
     {{>Comments}}
     {{/Children}}
 {{/ParagraphComment}}
+{{#HasParamComments}}
+    <h3>Parameters</h3>
+    {{#ParamComments}}
+    <div>
+        <b>{{ParamName}}</b> {{#Explicit}}{{Direction}}{{/Explicit}} {{#Children}}{{>Comments}}{{/Children}}
+    </div> 
+    {{/ParamComments}}
+{{/HasParamComments}}
+{{#HasReturnComments}}
+    <h3>Returns</h3>
+    {{#ReturnComments}}
+        {{#.}}
+        <p>{{TextComment}}</p>
+        {{/.}}
+    {{/ReturnComments}}
+{{/HasReturnComments}}
+{{#HasCodeComments}}
+    <h3>Code</h3> 
+    {{#CodeComments}}
+    <div>
+        <pre class="code-block">
+            <code>
+            {{#.}}
+            {{.}}
+            {{/.}}
+            </code>
+        </pre>
+    </div>
+    {{/CodeComments}}
+{{/HasCodeComments}}
 {{#BlockCommandComment}}
     <div class="block-command-comment__command">
         <div class="block-command-command">
diff --git a/clang-tools-extra/clang-doc/assets/function-template.mustache b/clang-tools-extra/clang-doc/assets/function-template.mustache
index 6683afa..2510a4d 100644
--- a/clang-tools-extra/clang-doc/assets/function-template.mustache
+++ b/clang-tools-extra/clang-doc/assets/function-template.mustache
@@ -14,10 +14,10 @@
             </code>
         </pre>
         {{! Function Comments }}
-        {{#FunctionComments}}
+        {{#Description}}
         <div>
             {{>Comments}}
         </div>
-        {{/FunctionComments}}
+        {{/Description}}
     </div>
 </div>
diff --git a/clang-tools-extra/clang-tidy/llvm/CMakeLists.txt b/clang-tools-extra/clang-tidy/llvm/CMakeLists.txt
index 4f1da43..41386cd 100644
--- a/clang-tools-extra/clang-tidy/llvm/CMakeLists.txt
+++ b/clang-tools-extra/clang-tidy/llvm/CMakeLists.txt
@@ -17,7 +17,6 @@ add_clang_library(clangTidyLLVMModule STATIC
   clangTidy
   clangTidyReadabilityModule
   clangTidyUtils
-  clangTransformer
 
   DEPENDS
   omp_gen
@@ -31,4 +30,5 @@ clang_target_link_libraries(clangTidyLLVMModule
   clangBasic
   clangLex
   clangTooling
+  clangTransformer
   )
diff --git a/clang-tools-extra/clang-tidy/modernize/UseDesignatedInitializersCheck.cpp b/clang-tools-extra/clang-tidy/modernize/UseDesignatedInitializersCheck.cpp
index 7ea9676..bb8fb240 100644
--- a/clang-tools-extra/clang-tidy/modernize/UseDesignatedInitializersCheck.cpp
+++ b/clang-tools-extra/clang-tidy/modernize/UseDesignatedInitializersCheck.cpp
@@ -121,10 +121,11 @@ void UseDesignatedInitializersCheck::registerMatchers(MatchFinder *Finder) {
       hasAnyBase(hasType(cxxRecordDecl(has(fieldDecl()))));
   Finder->addMatcher(
       initListExpr(
-          hasType(cxxRecordDecl(
-                      RestrictToPODTypes ? isPOD() : isAggregate(),
-                      unless(anyOf(HasBaseWithFields, hasName("::std::array"))))
-                      .bind("type")),
+          hasType(hasUnqualifiedDesugaredType(recordType(hasDeclaration(
+              cxxRecordDecl(
+                  RestrictToPODTypes ? isPOD() : isAggregate(),
+                  unless(anyOf(HasBaseWithFields, hasName("::std::array"))))
+                  .bind("type"))))),
           IgnoreSingleElementAggregates ? hasMoreThanOneElement() : anything(),
           unless(isFullyDesignated()))
           .bind("init"),
@@ -155,7 +156,7 @@ void UseDesignatedInitializersCheck::check(
       DiagnosticBuilder Diag =
           diag(InitList->getLBraceLoc(),
                "use designated initializer list to initialize %0");
-      Diag << Type << InitList->getSourceRange();
+      Diag << InitList->getType() << InitList->getSourceRange();
       for (const Stmt *InitExpr : *SyntacticInitList) {
         const auto Designator = Designators[InitExpr->getBeginLoc()];
         if (Designator && !Designator->empty())
diff --git a/clang-tools-extra/clang-tidy/modernize/UseUsingCheck.cpp b/clang-tools-extra/clang-tidy/modernize/UseUsingCheck.cpp
index 936a906..e307339 100644
--- a/clang-tools-extra/clang-tidy/modernize/UseUsingCheck.cpp
+++ b/clang-tools-extra/clang-tidy/modernize/UseUsingCheck.cpp
@@ -92,7 +92,7 @@ void UseUsingCheck::check(const MatchFinder::MatchResult &Result) {
   const LangOptions &LO = getLangOpts();
 
   // Match CXXRecordDecl only to store the range of the last non-implicit full
-  // declaration, to later check whether it's within the typdef itself.
+  // declaration, to later check whether it's within the typedef itself.
   const auto *MatchedTagDecl = Result.Nodes.getNodeAs<TagDecl>(TagDeclName);
   if (MatchedTagDecl) {
     // It is not sufficient to just track the last TagDecl that we've seen,
@@ -152,7 +152,7 @@ void UseUsingCheck::check(const MatchFinder::MatchResult &Result) {
     StringRef ExtraReference = "";
     if (MainTypeEndLoc.isValid() && TypeRange.fullyContains(MainTypeEndLoc)) {
       // Each type introduced in a typedef can specify being a reference or
-      // pointer type seperately, so we need to sigure out if the new using-decl
+      // pointer type separately, so we need to figure out if the new using-decl
       // needs to be to a reference or pointer as well.
       const SourceLocation Tok = utils::lexer::findPreviousAnyTokenKind(
           MatchedDecl->getLocation(), SM, LO, tok::TokenKind::star,
diff --git a/clang-tools-extra/clangd/refactor/tweaks/CMakeLists.txt b/clang-tools-extra/clangd/refactor/tweaks/CMakeLists.txt
index 59475b0..1d6e380 100644
--- a/clang-tools-extra/clangd/refactor/tweaks/CMakeLists.txt
+++ b/clang-tools-extra/clangd/refactor/tweaks/CMakeLists.txt
@@ -14,9 +14,9 @@ set(LLVM_LINK_COMPONENTS
 add_clang_library(clangDaemonTweaks OBJECT
   AddUsing.cpp
   AnnotateHighlightings.cpp
-  DumpAST.cpp
   DefineInline.cpp
   DefineOutline.cpp
+  DumpAST.cpp
   ExpandDeducedType.cpp
   ExpandMacro.cpp
   ExtractFunction.cpp
@@ -24,6 +24,7 @@ add_clang_library(clangDaemonTweaks OBJECT
   MemberwiseConstructor.cpp
   ObjCLocalizeStringLiteral.cpp
   ObjCMemberwiseInitializer.cpp
+  OverridePureVirtuals.cpp
   PopulateSwitch.cpp
   RawStringLiteral.cpp
   RemoveUsingNamespace.cpp
diff --git a/clang-tools-extra/clangd/refactor/tweaks/OverridePureVirtuals.cpp b/clang-tools-extra/clangd/refactor/tweaks/OverridePureVirtuals.cpp
new file mode 100644
index 0000000..16febec
--- /dev/null
+++ b/clang-tools-extra/clangd/refactor/tweaks/OverridePureVirtuals.cpp
@@ -0,0 +1,374 @@
+//===--- OverridePureVirtuals.cpp --------------------------------*- C++-*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// Tweak to automatically generate stubs for pure virtual methods inherited from
+// base classes.
+//
+// Purpose:
+//   - Simplifies making a derived class concrete by automating the creation of
+//     required method overrides from abstract bases.
+//
+// Tweak Summary:
+//
+// 1. Activation Conditions (prepare):
+//    - The tweak activates when the cursor is over a C++ class definition.
+//    - The class must be abstract (it, or its base classes, have unimplemented
+//      pure virtual functions).
+//    - It must also inherit from at least one other abstract class.
+//
+// 2. Identifying Missing Methods:
+//    - The tweak scans the inheritance hierarchy of the current class.
+//    - It identifies all unique pure virtual methods from base classes
+//      that are not yet implemented or overridden.
+//    - These missing methods are then grouped by their original access
+//      specifier (e.g., public, protected).
+//
+// 3. Code Generation and Insertion:
+//    - For each group of missing methods, stubs are inserted.
+//    - If an access specifier section (like `public:`) exists, stubs are
+//      inserted there; otherwise, a new section is created and appended.
+//    - Each generated stub includes the `override` keyword, a `// TODO:`
+//      comment, and a `static_assert(false, ...)` to force a compile-time
+//      error if the method remains unimplemented.
+//    - The base method's signature is adjusted (e.g., `virtual` and `= 0`
+//      are removed for the override).
+//
+// 4. Code Action Provided:
+//    - A single code action titled "Override pure virtual methods" is offered.
+//    - Applying this action results in a single source file modification
+//      containing all the generated method stubs.
+//
+// Example:
+//
+//  class Base {
+//  public:
+//    virtual void publicMethod() = 0;
+//  protected:
+//    virtual auto privateMethod() const -> int = 0;
+//  };
+//
+// Before:
+//                              // cursor here
+//  class Derived : public Base {}^;
+//
+// After:
+//
+// class Derived : public Base {
+//  public:
+//   void publicMethod() override {
+//     // TODO: Implement this pure virtual method.
+//     static_assert(false, "Method `publicMethod` is not implemented.");
+//   }
+//
+//  protected:
+//   auto privateMethod() const -> int override {
+//     // TODO: Implement this pure virtual method.
+//     static_assert(false, "Method `privateMethod` is not implemented.");
+//   }
+// };
+//
+//===----------------------------------------------------------------------===//
+
+#include "refactor/Tweak.h"
+#include "support/Token.h"
+
+#include "clang/AST/ASTContext.h"
+#include "clang/AST/DeclCXX.h"
+#include "clang/AST/Type.h"
+#include "clang/AST/TypeLoc.h"
+#include "clang/Basic/LLVM.h"
+#include "clang/Basic/SourceLocation.h"
+#include "clang/Tooling/Core/Replacement.h"
+#include "llvm/ADT/DenseSet.h"
+#include "llvm/Support/FormatVariadic.h"
+#include <string>
+
+namespace clang {
+namespace clangd {
+namespace {
+
+// This function removes the "virtual" and the "= 0" at the end;
+// e.g.:
+//   "virtual void foo(int var = 0) = 0"  // input.
+//   "void foo(int var = 0)"              // output.
+std::string removePureVirtualSyntax(const std::string &MethodDecl,
+                                    const LangOptions &LangOpts) {
+  assert(!MethodDecl.empty());
+
+  TokenStream TS = lex(MethodDecl, LangOpts);
+
+  std::string DeclString;
+  for (const clangd::Token &Tk : TS.tokens()) {
+    if (Tk.Kind == clang::tok::raw_identifier && Tk.text() == "virtual")
+      continue;
+
+    // If the ending two tokens are "= 0", we break here and we already have the
+    // method's string without the pure virtual syntax.
+    const auto &Next = Tk.next();
+    if (Next.next().Kind == tok::eof && Tk.Kind == clang::tok::equal &&
+        Next.text() == "0")
+      break;
+
+    DeclString += Tk.text();
+    if (Tk.Kind != tok::l_paren && Next.Kind != tok::comma &&
+        Next.Kind != tok::r_paren && Next.Kind != tok::l_paren)
+      DeclString += ' ';
+  }
+  // Trim the last whitespace.
+  if (DeclString.back() == ' ')
+    DeclString.pop_back();
+
+  return DeclString;
+}
+
+class OverridePureVirtuals final : public Tweak {
+public:
+  const char *id() const final; // defined by REGISTER_TWEAK.
+  bool prepare(const Selection &Sel) override;
+  Expected<Effect> apply(const Selection &Sel) override;
+  std::string title() const override { return "Override pure virtual methods"; }
+  llvm::StringLiteral kind() const override {
+    return CodeAction::QUICKFIX_KIND;
+  }
+
+private:
+  // Stores the CXXRecordDecl of the class being modified.
+  const CXXRecordDecl *CurrentDeclDef = nullptr;
+  // Stores pure virtual methods that need overriding, grouped by their original
+  // access specifier.
+  llvm::MapVector<AccessSpecifier, llvm::SmallVector<const CXXMethodDecl *>>
+      MissingMethodsByAccess;
+  // Stores the source locations of existing access specifiers in CurrentDecl.
+  llvm::MapVector<AccessSpecifier, SourceLocation> AccessSpecifierLocations;
+  // Helper function to gather information before applying the tweak.
+  void collectMissingPureVirtuals();
+};
+
+REGISTER_TWEAK(OverridePureVirtuals)
+
+// Function to get all unique pure virtual methods from the entire
+// base class hierarchy of CurrentDeclDef.
+llvm::SmallVector<const clang::CXXMethodDecl *>
+getAllUniquePureVirtualsFromBaseHierarchy(
+    const clang::CXXRecordDecl *CurrentDeclDef) {
+  llvm::SmallVector<const clang::CXXMethodDecl *> AllPureVirtualsInHierarchy;
+  llvm::DenseSet<const clang::CXXMethodDecl *> CanonicalPureVirtualsSeen;
+
+  if (!CurrentDeclDef || !CurrentDeclDef->getDefinition())
+    return AllPureVirtualsInHierarchy;
+
+  const clang::CXXRecordDecl *Def = CurrentDeclDef->getDefinition();
+
+  Def->forallBases([&](const clang::CXXRecordDecl *BaseDefinition) {
+    for (const clang::CXXMethodDecl *Method : BaseDefinition->methods()) {
+      if (Method->isPureVirtual() &&
+          CanonicalPureVirtualsSeen.insert(Method->getCanonicalDecl()).second)
+        AllPureVirtualsInHierarchy.emplace_back(Method);
+    }
+    // Continue iterating through all bases.
+    return true;
+  });
+
+  return AllPureVirtualsInHierarchy;
+}
+
+// Gets canonical declarations of methods already overridden or implemented in
+// class D.
+llvm::SetVector<const CXXMethodDecl *>
+getImplementedOrOverriddenCanonicals(const CXXRecordDecl *D) {
+  llvm::SetVector<const CXXMethodDecl *> ImplementedSet;
+  for (const CXXMethodDecl *M : D->methods()) {
+    // If M provides an implementation for any virtual method it overrides.
+    // A method is an "implementation" if it's virtual and not pure.
+    // Or if it directly overrides a base method.
+    for (const CXXMethodDecl *OverriddenM : M->overridden_methods())
+      ImplementedSet.insert(OverriddenM->getCanonicalDecl());
+  }
+  return ImplementedSet;
+}
+
+// Get the location of every colon of the `AccessSpecifier`.
+llvm::MapVector<AccessSpecifier, SourceLocation>
+getSpecifierLocations(const CXXRecordDecl *D) {
+  llvm::MapVector<AccessSpecifier, SourceLocation> Locs;
+  for (auto *DeclNode : D->decls()) {
+    if (const auto *ASD = llvm::dyn_cast<AccessSpecDecl>(DeclNode))
+      Locs[ASD->getAccess()] = ASD->getColonLoc();
+  }
+  return Locs;
+}
+
+bool hasAbstractBaseAncestor(const clang::CXXRecordDecl *CurrentDecl) {
+  assert(CurrentDecl && CurrentDecl->getDefinition());
+
+  return llvm::any_of(
+      CurrentDecl->getDefinition()->bases(), [](CXXBaseSpecifier BaseSpec) {
+        const auto *D = BaseSpec.getType()->getAsCXXRecordDecl();
+        const auto *Def = D ? D->getDefinition() : nullptr;
+        return Def && Def->isAbstract();
+      });
+}
+
+// The tweak is available if the selection is over an abstract C++ class
+// definition that also inherits from at least one other abstract class.
+bool OverridePureVirtuals::prepare(const Selection &Sel) {
+  const SelectionTree::Node *Node = Sel.ASTSelection.commonAncestor();
+  if (!Node)
+    return false;
+
+  // Make sure we have a definition.
+  CurrentDeclDef = Node->ASTNode.get<CXXRecordDecl>();
+  if (!CurrentDeclDef || !CurrentDeclDef->getDefinition())
+    return false;
+
+  // From now on, we should work with the definition.
+  CurrentDeclDef = CurrentDeclDef->getDefinition();
+
+  // Only offer for abstract classes with abstract bases.
+  return CurrentDeclDef->isAbstract() &&
+         hasAbstractBaseAncestor(CurrentDeclDef);
+}
+
+// Collects all pure virtual methods from base classes that `CurrentDeclDef` has
+// not yet overridden, grouped by their original access specifier.
+//
+// Results are stored in `MissingMethodsByAccess` and `AccessSpecifierLocations`
+// is also populated.
+void OverridePureVirtuals::collectMissingPureVirtuals() {
+  if (!CurrentDeclDef)
+    return;
+
+  AccessSpecifierLocations = getSpecifierLocations(CurrentDeclDef);
+  MissingMethodsByAccess.clear();
+
+  // Get all unique pure virtual methods from the entire base class hierarchy.
+  llvm::SmallVector<const CXXMethodDecl *> AllPureVirtualsInHierarchy =
+      getAllUniquePureVirtualsFromBaseHierarchy(CurrentDeclDef);
+
+  // Get methods already implemented or overridden in CurrentDecl.
+  const auto ImplementedOrOverriddenSet =
+      getImplementedOrOverriddenCanonicals(CurrentDeclDef);
+
+  // Filter AllPureVirtualsInHierarchy to find those not in
+  // ImplementedOrOverriddenSet, which needs to be overriden.
+  for (const CXXMethodDecl *BaseMethod : AllPureVirtualsInHierarchy) {
+    bool AlreadyHandled = ImplementedOrOverriddenSet.contains(BaseMethod);
+    if (!AlreadyHandled)
+      MissingMethodsByAccess[BaseMethod->getAccess()].emplace_back(BaseMethod);
+  }
+}
+
+std::string generateOverrideString(const CXXMethodDecl *Method,
+                                   const LangOptions &LangOpts) {
+  std::string MethodDecl;
+  auto OS = llvm::raw_string_ostream(MethodDecl);
+  Method->print(OS);
+
+  return llvm::formatv(
+             "\n  {0} override {{\n"
+             "    // TODO: Implement this pure virtual method.\n"
+             "    static_assert(false, \"Method `{1}` is not implemented.\");\n"
+             "  }",
+             removePureVirtualSyntax(MethodDecl, LangOpts), Method->getName())
+      .str();
+}
+
+// Free function to generate the string for a group of method overrides.
+std::string generateOverridesStringForGroup(
+    llvm::SmallVector<const CXXMethodDecl *> Methods,
+    const LangOptions &LangOpts) {
+  llvm::SmallVector<std::string> MethodsString;
+  MethodsString.reserve(Methods.size());
+
+  for (const CXXMethodDecl *Method : Methods) {
+    MethodsString.emplace_back(generateOverrideString(Method, LangOpts));
+  }
+
+  return llvm::join(MethodsString, "\n") + '\n';
+}
+
+Expected<Tweak::Effect> OverridePureVirtuals::apply(const Selection &Sel) {
+  // The correctness of this tweak heavily relies on the accurate population of
+  // these members.
+  collectMissingPureVirtuals();
+  // The `prepare` should prevent this. If the prepare identifies an abstract
+  // method, then is must have missing methods.
+  assert(!MissingMethodsByAccess.empty());
+
+  const auto &SM = Sel.AST->getSourceManager();
+  const auto &LangOpts = Sel.AST->getLangOpts();
+
+  tooling::Replacements EditReplacements;
+  // Stores text for new access specifier sections that are not already present
+  // in the class.
+  // Example:
+  //  public:    // ...
+  //  protected: // ...
+  std::string NewSectionsToAppendText;
+
+  for (const auto &[AS, Methods] : MissingMethodsByAccess) {
+    assert(!Methods.empty());
+
+    std::string MethodsGroupString =
+        generateOverridesStringForGroup(Methods, LangOpts);
+
+    auto *ExistingSpecLocIter = AccessSpecifierLocations.find(AS);
+    bool ASExists = ExistingSpecLocIter != AccessSpecifierLocations.end();
+    if (ASExists) {
+      // Access specifier section already exists in the class.
+      // Get location immediately *after* the colon.
+      SourceLocation InsertLoc =
+          ExistingSpecLocIter->second.getLocWithOffset(1);
+
+      // Create a replacement to insert the method declarations.
+      // The replacement is at InsertLoc, has length 0 (insertion), and uses
+      // InsertionText.
+      std::string InsertionText = MethodsGroupString;
+      tooling::Replacement Rep(SM, InsertLoc, 0, InsertionText);
+      if (auto Err = EditReplacements.add(Rep))
+        return llvm::Expected<Tweak::Effect>(std::move(Err));
+    } else {
+      // Access specifier section does not exist in the class.
+      // These methods will be grouped into NewSectionsToAppendText and added
+      // towards the end of the class definition.
+      NewSectionsToAppendText +=
+          getAccessSpelling(AS).str() + ':' + MethodsGroupString;
+    }
+  }
+
+  // After processing all access specifiers, add any newly created sections
+  // (stored in NewSectionsToAppendText) to the end of the class.
+  if (!NewSectionsToAppendText.empty()) {
+    // AppendLoc is the SourceLocation of the closing brace '}' of the class.
+    // The replacement will insert text *before* this closing brace.
+    SourceLocation AppendLoc = CurrentDeclDef->getBraceRange().getEnd();
+    std::string FinalAppendText = std::move(NewSectionsToAppendText);
+
+    if (!CurrentDeclDef->decls_empty() || !EditReplacements.empty()) {
+      FinalAppendText = '\n' + FinalAppendText;
+    }
+
+    // Create a replacement to append the new sections.
+    tooling::Replacement Rep(SM, AppendLoc, 0, FinalAppendText);
+    if (auto Err = EditReplacements.add(Rep))
+      return llvm::Expected<Tweak::Effect>(std::move(Err));
+  }
+
+  if (EditReplacements.empty()) {
+    return llvm::make_error<llvm::StringError>(
+        "No changes to apply (internal error or no methods generated).",
+        llvm::inconvertibleErrorCode());
+  }
+
+  // Return the collected replacements as the effect of this tweak.
+  return Effect::mainFileEdit(SM, EditReplacements);
+}
+
+} // namespace
+} // namespace clangd
+} // namespace clang
diff --git a/clang-tools-extra/clangd/unittests/CMakeLists.txt b/clang-tools-extra/clangd/unittests/CMakeLists.txt
index dffdcd5..d425070 100644
--- a/clang-tools-extra/clangd/unittests/CMakeLists.txt
+++ b/clang-tools-extra/clangd/unittests/CMakeLists.txt
@@ -131,6 +131,7 @@ add_unittest(ClangdUnitTests ClangdTests
   tweaks/MemberwiseConstructorTests.cpp
   tweaks/ObjCLocalizeStringLiteralTests.cpp
   tweaks/ObjCMemberwiseInitializerTests.cpp
+  tweaks/OverridePureVirtualsTests.cpp
   tweaks/PopulateSwitchTests.cpp
   tweaks/RawStringLiteralTests.cpp
   tweaks/RemoveUsingNamespaceTests.cpp
diff --git a/clang-tools-extra/clangd/unittests/tweaks/OverridePureVirtualsTests.cpp b/clang-tools-extra/clangd/unittests/tweaks/OverridePureVirtualsTests.cpp
new file mode 100644
index 0000000..b7dcbee
--- /dev/null
+++ b/clang-tools-extra/clangd/unittests/tweaks/OverridePureVirtualsTests.cpp
@@ -0,0 +1,720 @@
+//===-- OverridePureVirtualsTests.cpp ---------------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "TweakTesting.h"
+#include "gtest/gtest.h"
+
+namespace clang {
+namespace clangd {
+namespace {
+
+class OverridePureVirtualsTests : public TweakTest {
+protected:
+  OverridePureVirtualsTests() : TweakTest("OverridePureVirtuals") {}
+};
+
+TEST_F(OverridePureVirtualsTests, MinimalUnavailable) {
+  EXPECT_UNAVAILABLE("class ^C {};");
+}
+
+TEST_F(OverridePureVirtualsTests, MinimalAvailable) {
+  EXPECT_AVAILABLE(R"cpp(
+class B { public: virtual void Foo() = 0; };
+class ^C : public B {};
+)cpp");
+}
+
+TEST_F(OverridePureVirtualsTests, UnavailableWhenOverriden) {
+  EXPECT_UNAVAILABLE(
+      R"cpp(
+class B {
+public:
+  virtual void foo() = 0;
+};
+
+class ^D : public B {
+public:
+  void foo() override;
+};
+)cpp");
+}
+
+TEST_F(OverridePureVirtualsTests, AvailabilityNoOverride) {
+  EXPECT_AVAILABLE(R"cpp(
+class Base {
+public:
+virtual ~Base() = default;
+virtual void F1() = 0;
+virtual void F2() = 0;
+};
+
+class ^Derived : public Base {
+public:
+};
+
+)cpp");
+}
+
+TEST_F(OverridePureVirtualsTests, AvailabilityPartiallyOverridden) {
+  EXPECT_AVAILABLE(R"cpp(
+class Base {
+public:
+virtual ~Base() = default;
+virtual void F1() = 0;
+virtual void F2() = 0;
+};
+
+class ^Derived : public Base {
+public:
+void F1() override;
+};
+)cpp");
+}
+
+TEST_F(OverridePureVirtualsTests, EmptyDerivedClass) {
+  const char *Before = R"cpp(
+class Base {
+public:
+virtual ~Base() = default;
+virtual void F1() = 0;
+virtual void F2(int P1, const int &P2) = 0;
+};
+
+class ^Derived : public Base {};
+)cpp";
+  const auto *Expected = R"cpp(
+class Base {
+public:
+virtual ~Base() = default;
+virtual void F1() = 0;
+virtual void F2(int P1, const int &P2) = 0;
+};
+
+class Derived : public Base {
+public:
+  void F1() override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `F1` is not implemented.");
+  }
+
+  void F2(int P1, const int & P2) override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `F2` is not implemented.");
+  }
+};
+)cpp";
+  auto Applied = apply(Before);
+  EXPECT_EQ(Expected, Applied) << "Applied result:\n" << Applied;
+}
+
+TEST_F(OverridePureVirtualsTests, SingleBaseClassPartiallyImplemented) {
+  auto Applied = apply(
+      R"cpp(
+class Base {
+public:
+virtual ~Base() = default;
+virtual void F1() = 0;
+virtual void F2() = 0;
+};
+
+class ^Derived : public Base {
+public:
+  void F1() override;
+};
+)cpp");
+
+  const auto *Expected = R"cpp(
+class Base {
+public:
+virtual ~Base() = default;
+virtual void F1() = 0;
+virtual void F2() = 0;
+};
+
+class Derived : public Base {
+public:
+  void F2() override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `F2` is not implemented.");
+  }
+
+  void F1() override;
+};
+)cpp";
+  EXPECT_EQ(Applied, Expected) << "Applied result:\n" << Applied;
+}
+
+TEST_F(OverridePureVirtualsTests, MultipleDirectBaseClasses) {
+  const char *Before = R"cpp(
+class Base1 {
+public:
+  virtual void func1() = 0;
+};
+class Base2 {
+protected:
+  virtual bool func2(char c) const = 0;
+};
+
+class ^Derived : public Base1, public Base2 {};
+)cpp";
+  const auto *Expected = R"cpp(
+class Base1 {
+public:
+  virtual void func1() = 0;
+};
+class Base2 {
+protected:
+  virtual bool func2(char c) const = 0;
+};
+
+class Derived : public Base1, public Base2 {
+public:
+  void func1() override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `func1` is not implemented.");
+  }
+protected:
+  bool func2(char c) const override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `func2` is not implemented.");
+  }
+};
+)cpp";
+  auto Applied = apply(Before);
+  EXPECT_EQ(Expected, Applied) << "Applied result:\n" << Applied;
+}
+
+TEST_F(OverridePureVirtualsTests, UnnamedParametersInBase) {
+  const char *Before = R"cpp(
+struct S {};
+class Base {
+public:
+  virtual void func(int, const S&, char*) = 0;
+};
+
+class ^Derived : public Base {};
+)cpp";
+
+  const auto *Expected = R"cpp(
+struct S {};
+class Base {
+public:
+  virtual void func(int, const S&, char*) = 0;
+};
+
+class Derived : public Base {
+public:
+  void func(int, const S &, char *) override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `func` is not implemented.");
+  }
+};
+)cpp";
+  auto Applied = apply(Before);
+  EXPECT_EQ(Expected, Applied) << "Applied result:\n" << Applied;
+}
+
+TEST_F(OverridePureVirtualsTests, DiamondInheritance) {
+  const char *Before = R"cpp(
+class Top {
+public:
+  virtual ~Top() = default;
+  virtual void diamond_func() = 0;
+};
+class Left : virtual public Top {};
+class Right : virtual public Top {};
+class ^Bottom : public Left, public Right {};
+)cpp";
+  const auto *Expected = R"cpp(
+class Top {
+public:
+  virtual ~Top() = default;
+  virtual void diamond_func() = 0;
+};
+class Left : virtual public Top {};
+class Right : virtual public Top {};
+class Bottom : public Left, public Right {
+public:
+  void diamond_func() override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `diamond_func` is not implemented.");
+  }
+};
+)cpp";
+  auto Applied = apply(Before);
+  EXPECT_EQ(Expected, Applied) << "Applied result:\n" << Applied;
+}
+
+TEST_F(OverridePureVirtualsTests, MixedAccessSpecifiers) {
+  const char *Before = R"cpp(
+class Base {
+public:
+  virtual void pub_func() = 0;
+  virtual void pub_func2(char) const = 0;
+protected:
+  virtual int prot_func(int x) const = 0;
+};
+
+class ^Derived : public Base {
+  int member; // Existing member
+public:
+  Derived(int m) : member(m) {}
+};
+)cpp";
+  const auto *Expected = R"cpp(
+class Base {
+public:
+  virtual void pub_func() = 0;
+  virtual void pub_func2(char) const = 0;
+protected:
+  virtual int prot_func(int x) const = 0;
+};
+
+class Derived : public Base {
+  int member; // Existing member
+public:
+  void pub_func() override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `pub_func` is not implemented.");
+  }
+
+  void pub_func2(char) const override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `pub_func2` is not implemented.");
+  }
+
+  Derived(int m) : member(m) {}
+
+protected:
+  int prot_func(int x) const override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `prot_func` is not implemented.");
+  }
+};
+)cpp";
+  auto Applied = apply(Before);
+  EXPECT_EQ(Expected, Applied) << "Applied result:\n" << Applied;
+}
+
+TEST_F(OverridePureVirtualsTests, OutOfOrderMixedAccessSpecifiers) {
+  const char *Before = R"cpp(
+class Base {
+public:
+  virtual void pub_func() = 0;
+  virtual void pub_func2(char) const = 0;
+protected:
+  virtual int prot_func(int x) const = 0;
+};
+
+class ^Derived : public Base {
+  int member; // Existing member
+protected:
+  void foo();
+public:
+  Derived(int m) : member(m) {}
+};
+)cpp";
+  const auto *Expected = R"cpp(
+class Base {
+public:
+  virtual void pub_func() = 0;
+  virtual void pub_func2(char) const = 0;
+protected:
+  virtual int prot_func(int x) const = 0;
+};
+
+class Derived : public Base {
+  int member; // Existing member
+protected:
+  int prot_func(int x) const override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `prot_func` is not implemented.");
+  }
+
+  void foo();
+public:
+  void pub_func() override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `pub_func` is not implemented.");
+  }
+
+  void pub_func2(char) const override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `pub_func2` is not implemented.");
+  }
+
+  Derived(int m) : member(m) {}
+};
+)cpp";
+  auto Applied = apply(Before);
+  EXPECT_EQ(Expected, Applied) << "Applied result:\n" << Applied;
+}
+
+TEST_F(OverridePureVirtualsTests, MultiAccessSpecifiersOverride) {
+  constexpr auto Before = R"cpp(
+class Base {
+public:
+  virtual void foo() = 0;
+protected:
+  virtual void bar() = 0;
+};
+
+class ^Derived : public Base {};
+)cpp";
+
+  constexpr auto Expected = R"cpp(
+class Base {
+public:
+  virtual void foo() = 0;
+protected:
+  virtual void bar() = 0;
+};
+
+class Derived : public Base {
+public:
+  void foo() override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `foo` is not implemented.");
+  }
+protected:
+  void bar() override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `bar` is not implemented.");
+  }
+};
+)cpp";
+  auto Applied = apply(Before);
+  EXPECT_EQ(Expected, Applied) << "Applied result:\n" << Applied;
+}
+
+TEST_F(OverridePureVirtualsTests, AccessSpecifierAlreadyExisting) {
+  const char *Before = R"cpp(
+class Base {
+public:
+  virtual void func1() = 0;
+};
+
+class ^Derived : public Base {
+public:
+};
+)cpp";
+
+  const auto *Expected = R"cpp(
+class Base {
+public:
+  virtual void func1() = 0;
+};
+
+class Derived : public Base {
+public:
+  void func1() override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `func1` is not implemented.");
+  }
+
+};
+)cpp";
+  auto Applied = apply(Before);
+  EXPECT_EQ(Expected, Applied) << "Applied result:\n" << Applied;
+}
+
+TEST_F(OverridePureVirtualsTests, ConstexprSpecifier) {
+  ExtraArgs.push_back("-std=c++20");
+
+  constexpr auto Before = R"cpp(
+class B {
+public:
+  constexpr virtual int getValue() const = 0;
+};
+
+class ^D : public B {};
+)cpp";
+
+  constexpr auto Expected = R"cpp(
+class B {
+public:
+  constexpr virtual int getValue() const = 0;
+};
+
+class D : public B {
+public:
+  constexpr int getValue() const override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `getValue` is not implemented.");
+  }
+};
+)cpp";
+  auto Applied = apply(Before);
+  EXPECT_EQ(Expected, Applied) << "Applied result:\n" << Applied;
+}
+
+TEST_F(OverridePureVirtualsTests, ConstevalSpecifier) {
+  ExtraArgs.push_back("-std=c++20");
+
+  constexpr auto Before = R"cpp(
+class B {
+public:
+  virtual consteval float calculate() = 0;
+};
+
+class ^D : public B {};
+)cpp";
+
+  constexpr auto Expected = R"cpp(
+class B {
+public:
+  virtual consteval float calculate() = 0;
+};
+
+class D : public B {
+public:
+  consteval float calculate() override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `calculate` is not implemented.");
+  }
+};
+)cpp";
+  auto Applied = apply(Before);
+  EXPECT_EQ(Expected, Applied) << "Applied result:\n" << Applied;
+}
+
+TEST_F(OverridePureVirtualsTests, LValueRefQualifier) {
+  constexpr auto Before = R"cpp(
+class B {
+public:
+  virtual void process() & = 0;
+};
+
+class ^D : public B {};
+)cpp";
+
+  constexpr auto Expected = R"cpp(
+class B {
+public:
+  virtual void process() & = 0;
+};
+
+class D : public B {
+public:
+  void process() & override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `process` is not implemented.");
+  }
+};
+)cpp";
+  auto Applied = apply(Before);
+  EXPECT_EQ(Expected, Applied) << "Applied result:\n" << Applied;
+}
+
+TEST_F(OverridePureVirtualsTests, RValueRefQualifier) {
+  constexpr auto Before = R"cpp(
+class B {
+public:
+  virtual bool isValid() && = 0;
+};
+
+class ^D : public B {};
+)cpp";
+
+  constexpr auto Expected = R"cpp(
+class B {
+public:
+  virtual bool isValid() && = 0;
+};
+
+class D : public B {
+public:
+  bool isValid() && override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `isValid` is not implemented.");
+  }
+};
+)cpp";
+  auto Applied = apply(Before);
+  EXPECT_EQ(Expected, Applied) << "Applied result:\n" << Applied;
+}
+
+TEST_F(OverridePureVirtualsTests, SimpleTrailingReturnType) {
+  constexpr auto Before = R"cpp(
+class B {
+public:
+  virtual auto getStatus() -> bool = 0;
+};
+
+class ^D : public B {};
+)cpp";
+
+  constexpr auto Expected = R"cpp(
+class B {
+public:
+  virtual auto getStatus() -> bool = 0;
+};
+
+class D : public B {
+public:
+  auto getStatus() -> bool override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `getStatus` is not implemented.");
+  }
+};
+)cpp";
+  auto Applied = apply(Before);
+  EXPECT_EQ(Expected, Applied) << "Applied result:\n" << Applied;
+}
+
+TEST_F(OverridePureVirtualsTests, ConstexprLValueRefAndTrailingReturn) {
+  ExtraArgs.push_back("-std=c++20");
+
+  constexpr auto Before = R"cpp(
+class B {
+public:
+  constexpr virtual auto getData() & -> const char * = 0;
+};
+
+class ^D : public B {};
+)cpp";
+
+  constexpr auto Expected = R"cpp(
+class B {
+public:
+  constexpr virtual auto getData() & -> const char * = 0;
+};
+
+class D : public B {
+public:
+  constexpr auto getData() & -> const char * override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `getData` is not implemented.");
+  }
+};
+)cpp";
+  auto Applied = apply(Before);
+  EXPECT_EQ(Expected, Applied) << "Applied result:\n" << Applied;
+}
+
+TEST_F(OverridePureVirtualsTests, ConstevalRValueRefAndTrailingReturn) {
+  ExtraArgs.push_back("-std=c++20");
+
+  constexpr auto Before = R"cpp(
+class B {
+public:
+  virtual consteval auto foo() && -> double = 0;
+};
+
+class ^D : public B {};
+)cpp";
+
+  constexpr auto Expected = R"cpp(
+class B {
+public:
+  virtual consteval auto foo() && -> double = 0;
+};
+
+class D : public B {
+public:
+  consteval auto foo() && -> double override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `foo` is not implemented.");
+  }
+};
+)cpp";
+  auto Applied = apply(Before);
+  EXPECT_EQ(Expected, Applied) << "Applied result:\n" << Applied;
+}
+
+TEST_F(OverridePureVirtualsTests, CombinedFeaturesWithTrailingReturnTypes) {
+  ExtraArgs.push_back("-std=c++20");
+
+  constexpr auto Before = R"cpp(
+class B {
+public:
+  virtual auto f1() & -> int = 0;
+  constexpr virtual auto f2() && -> int = 0;
+  virtual consteval auto f3() -> int = 0;
+  virtual auto f4() const & -> char = 0;
+  constexpr virtual auto f5() const && -> bool = 0;
+};
+
+class ^D : public B {};
+)cpp";
+
+  constexpr auto Expected = R"cpp(
+class B {
+public:
+  virtual auto f1() & -> int = 0;
+  constexpr virtual auto f2() && -> int = 0;
+  virtual consteval auto f3() -> int = 0;
+  virtual auto f4() const & -> char = 0;
+  constexpr virtual auto f5() const && -> bool = 0;
+};
+
+class D : public B {
+public:
+  auto f1() & -> int override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `f1` is not implemented.");
+  }
+
+  constexpr auto f2() && -> int override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `f2` is not implemented.");
+  }
+
+  consteval auto f3() -> int override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `f3` is not implemented.");
+  }
+
+  auto f4() const & -> char override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `f4` is not implemented.");
+  }
+
+  constexpr auto f5() const && -> bool override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `f5` is not implemented.");
+  }
+};
+)cpp";
+  auto Applied = apply(Before);
+  EXPECT_EQ(Expected, Applied) << "Applied result:\n" << Applied;
+}
+
+TEST_F(OverridePureVirtualsTests, DefaultParameters) {
+  ExtraArgs.push_back("-std=c++20");
+
+  constexpr auto Before = R"cpp(
+class B {
+public:
+  virtual void foo(int var = 0) = 0;
+};
+
+class ^D : public B {};
+)cpp";
+
+  constexpr auto Expected = R"cpp(
+class B {
+public:
+  virtual void foo(int var = 0) = 0;
+};
+
+class D : public B {
+public:
+  void foo(int var = 0) override {
+    // TODO: Implement this pure virtual method.
+    static_assert(false, "Method `foo` is not implemented.");
+  }
+};
+)cpp";
+  auto Applied = apply(Before);
+  EXPECT_EQ(Expected, Applied) << "Applied result:\n" << Applied;
+}
+
+} // namespace
+} // namespace clangd
+} // namespace clang
diff --git a/clang-tools-extra/docs/ReleaseNotes.rst b/clang-tools-extra/docs/ReleaseNotes.rst
index 3ea1c51..dc78278 100644
--- a/clang-tools-extra/docs/ReleaseNotes.rst
+++ b/clang-tools-extra/docs/ReleaseNotes.rst
@@ -67,6 +67,14 @@ Code completion
 Code actions
 ^^^^^^^^^^^^
 
+- New ``Override pure virtual methods`` code action. When invoked on a class
+  definition, this action automatically generates C++ ``override`` declarations
+  for all pure virtual methods inherited from its base classes that have not yet
+  been implemented. The generated method stubs prompts the user for the actual
+  implementation. The overrides are intelligently grouped under their original
+  access specifiers (e.g., ``public``, ``protected``), creating new access
+  specifier blocks if necessary.
+
 Signature help
 ^^^^^^^^^^^^^^
 
@@ -128,6 +136,10 @@ Changes in existing checks
 - Improved :doc:`misc-header-include-cycle
   <clang-tidy/checks/misc/header-include-cycle>` check performance.
 
+- Improved :doc:`modernize-use-designated-initializers
+  <clang-tidy/checks/modernize/use-designated-initializers>` check to
+  suggest using designated initializers for aliased aggregate types.
+
 - Improved :doc:`modernize-use-std-format
   <clang-tidy/checks/modernize/use-std-format>` check to correctly match
   when the format string is converted to a different type by an implicit
diff --git a/clang-tools-extra/test/clang-doc/Inputs/basic-project/include/Circle.h b/clang-tools-extra/test/clang-doc/Inputs/basic-project/include/Circle.h
index 7bee3ff..74bffcd 100644
--- a/clang-tools-extra/test/clang-doc/Inputs/basic-project/include/Circle.h
+++ b/clang-tools-extra/test/clang-doc/Inputs/basic-project/include/Circle.h
@@ -26,6 +26,10 @@ public:
     /**
      * @brief Calculates the perimeter of the circle.
      *
+     * @code
+     * Circle circle(5.0);
+     * double perimeter = circle.perimeter();
+     * @endcode
      * @return double The perimeter of the circle.
      */
     double perimeter() const override;
diff --git a/clang-tools-extra/test/clang-doc/basic-project.mustache.test b/clang-tools-extra/test/clang-doc/basic-project.mustache.test
index 4fb38e2..e2d9da6 100644
--- a/clang-tools-extra/test/clang-doc/basic-project.mustache.test
+++ b/clang-tools-extra/test/clang-doc/basic-project.mustache.test
@@ -83,6 +83,19 @@ HTML-SHAPE:             <code class="language-cpp code-clang-doc">
 HTML-SHAPE: double area ()
 HTML-SHAPE:             </code>
 HTML-SHAPE:         </pre>
+HTML-SHAPE:        <div>
+HTML-SHAPE:                        <div>
+HTML-SHAPE:                            <p> Calculates the area of the shape.</p>
+HTML-SHAPE:                        </div>
+HTML-SHAPE:                            <div>
+HTML-SHAPE:                        <p></p>
+HTML-SHAPE:                    </div>
+HTML-SHAPE:                        <div>
+HTML-SHAPE:                        <p></p>
+HTML-SHAPE:                    </div>
+HTML-SHAPE:                    <h3>Returns</h3>
+HTML-SHAPE:                        <p> double The area of the shape.</p>
+HTML-SHAPE:                        </div>
 HTML-SHAPE:     </div>
 HTML-SHAPE: </div>
 HTML-SHAPE: <div class="delimiter-container">
@@ -92,6 +105,19 @@ HTML-SHAPE:             <code class="language-cpp code-clang-doc">
 HTML-SHAPE: double perimeter ()
 HTML-SHAPE:             </code>
 HTML-SHAPE:         </pre>
+HTML-SHAPE:        <div>
+HTML-SHAPE:                        <div>
+HTML-SHAPE:                            <p> Calculates the perimeter of the shape.</p>
+HTML-SHAPE:                        </div>
+HTML-SHAPE:                            <div>
+HTML-SHAPE:                        <p></p>
+HTML-SHAPE:                    </div>
+HTML-SHAPE:                        <div>
+HTML-SHAPE:                        <p></p>
+HTML-SHAPE:                    </div>
+HTML-SHAPE:                    <h3>Returns</h3>
+HTML-SHAPE:                        <p> double The perimeter of the shape.</p>
+HTML-SHAPE:                        </div>
 HTML-SHAPE:     </div>
 HTML-SHAPE: </div>
 HTML-SHAPE: <div class="delimiter-container">
@@ -101,6 +127,14 @@ HTML-SHAPE:             <code class="language-cpp code-clang-doc">
 HTML-SHAPE: void ~Shape ()
 HTML-SHAPE:             </code>
 HTML-SHAPE:         </pre>
+HTML-SHAPE:        <div>
+HTML-SHAPE:                        <div>
+HTML-SHAPE:                            <p> Virtual destructor.</p>
+HTML-SHAPE:                        </div>
+HTML-SHAPE:                            <div>
+HTML-SHAPE:                        <p></p>
+HTML-SHAPE:                    </div>
+HTML-SHAPE:                        </div>
 HTML-SHAPE:     </div>
 HTML-SHAPE: </div>
 HTML-SHAPE: </div>
@@ -220,6 +254,36 @@ HTML-CALC:             <code class="language-cpp code-clang-doc">
 HTML-CALC: int add (int a, int b)
 HTML-CALC:             </code>
 HTML-CALC:         </pre>
+HTML-CALC:        <div>
+HTML-CALC:                        <div>
+HTML-CALC:                            <p> Adds two integers.</p>
+HTML-CALC:                        </div>
+HTML-CALC:                            <div>
+HTML-CALC:                        <p></p>
+HTML-CALC:                    </div>
+HTML-CALC:                        <div>
+HTML-CALC:                        <p></p>
+HTML-CALC:                    </div>
+HTML-CALC:                    <h3>Parameters</h3>
+HTML-CALC:                    <div>
+HTML-CALC:                        <b>a</b>      <div>
+HTML-CALC:                        <p> First integer.</p>
+HTML-CALC:                    </div>
+HTML-CALC:                    <div>
+HTML-CALC:                        <p></p>
+HTML-CALC:                    </div>
+HTML-CALC:                    </div>
+HTML-CALC:                    <div>
+HTML-CALC:                        <b>b</b>      <div>
+HTML-CALC:                        <p> Second integer.</p>
+HTML-CALC:                    </div>
+HTML-CALC:                    <div>
+HTML-CALC:                        <p></p>
+HTML-CALC:                    </div>
+HTML-CALC:                    </div>
+HTML-CALC:                    <h3>Returns</h3>
+HTML-CALC:                        <p> int The sum of a and b.</p>
+HTML-CALC:                        </div>
 HTML-CALC:     </div>
 HTML-CALC: </div>
 HTML-CALC: <div class="delimiter-container">
@@ -229,6 +293,19 @@ HTML-CALC:             <code class="language-cpp code-clang-doc">
 HTML-CALC: int subtract (int a, int b)
 HTML-CALC:             </code>
 HTML-CALC:         </pre>
+HTML-CALC:        <div>
+HTML-CALC:                        <div>
+HTML-CALC:                            <p> Subtracts the second integer from the first.</p>
+HTML-CALC:                        </div>
+HTML-CALC:                            <div>
+HTML-CALC:                        <p></p>
+HTML-CALC:                    </div>
+HTML-CALC:                        <div>
+HTML-CALC:                        <p></p>
+HTML-CALC:                    </div>
+HTML-CALC:                    <h3>Returns</h3>
+HTML-CALC:                        <p> int The result of a - b.</p>
+HTML-CALC:                        </div>
 HTML-CALC:     </div>
 HTML-CALC: </div>
 HTML-CALC: <div class="delimiter-container">
@@ -238,6 +315,36 @@ HTML-CALC:             <code class="language-cpp code-clang-doc">
 HTML-CALC: int multiply (int a, int b)
 HTML-CALC:             </code>
 HTML-CALC:         </pre>
+HTML-CALC:        <div>
+HTML-CALC:                        <div>
+HTML-CALC:                            <p> Multiplies two integers.</p>
+HTML-CALC:                        </div>
+HTML-CALC:                            <div>
+HTML-CALC:                        <p></p>
+HTML-CALC:                    </div>
+HTML-CALC:                        <div>
+HTML-CALC:                        <p></p>
+HTML-CALC:                    </div>
+HTML-CALC:                    <h3>Parameters</h3>
+HTML-CALC:                    <div>
+HTML-CALC:                        <b>a</b>      <div>
+HTML-CALC:                        <p> First integer.</p>
+HTML-CALC:                    </div>
+HTML-CALC:                    <div>
+HTML-CALC:                        <p></p>
+HTML-CALC:                    </div>
+HTML-CALC:                    </div>
+HTML-CALC:                    <div>
+HTML-CALC:                        <b>b</b>      <div>
+HTML-CALC:                        <p> Second integer.</p>
+HTML-CALC:                    </div>
+HTML-CALC:                    <div>
+HTML-CALC:                        <p></p>
+HTML-CALC:                    </div>
+HTML-CALC:                    </div>
+HTML-CALC:                    <h3>Returns</h3>
+HTML-CALC:                        <p> int The product of a and b.</p>
+HTML-CALC:                        </div>
 HTML-CALC:     </div>
 HTML-CALC: </div>
 HTML-CALC: <div class="delimiter-container">
@@ -247,6 +354,37 @@ HTML-CALC:             <code class="language-cpp code-clang-doc">
 HTML-CALC: double divide (int a, int b)
 HTML-CALC:             </code>
 HTML-CALC:         </pre>
+HTML-CALC:        <div>
+HTML-CALC:                        <div>
+HTML-CALC:                            <p> Divides the first integer by the second.</p>
+HTML-CALC:                        </div>
+HTML-CALC:                            <div>
+HTML-CALC:                        <p></p>
+HTML-CALC:                    </div>
+HTML-CALC:                        <div>
+HTML-CALC:                        <p></p>
+HTML-CALC:                    </div>
+HTML-CALC:                    <h3>Parameters</h3>
+HTML-CALC:                    <div>
+HTML-CALC:                        <b>a</b>      <div>
+HTML-CALC:                        <p> First integer.</p>
+HTML-CALC:                    </div>
+HTML-CALC:                    <div>
+HTML-CALC:                        <p></p>
+HTML-CALC:                    </div>
+HTML-CALC:                    </div>
+HTML-CALC:                    <div>
+HTML-CALC:                        <b>b</b>      <div>
+HTML-CALC:                        <p> Second integer.</p>
+HTML-CALC:                    </div>
+HTML-CALC:                    <div>
+HTML-CALC:                        <p></p>
+HTML-CALC:                    </div>
+HTML-CALC:                    </div>
+HTML-CALC:                    <h3>Returns</h3>
+HTML-CALC:                        <p> double The result of a / b.</p>
+HTML-CALC:                        <p></p>
+HTML-CALC:                        </div>
 HTML-CALC:     </div>
 HTML-CALC: </div>
 HTML-CALC: <div class="delimiter-container">
@@ -256,6 +394,36 @@ HTML-CALC:             <code class="language-cpp code-clang-doc">
 HTML-CALC: int mod (int a, int b)
 HTML-CALC:             </code>
 HTML-CALC:         </pre>
+HTML-CALC:        <div>
+HTML-CALC:                        <div>
+HTML-CALC:                            <p> Performs the mod operation on integers.</p>
+HTML-CALC:                        </div>
+HTML-CALC:                            <div>
+HTML-CALC:                        <p></p>
+HTML-CALC:                    </div>
+HTML-CALC:                        <div>
+HTML-CALC:                        <p></p>
+HTML-CALC:                    </div>
+HTML-CALC:                    <h3>Parameters</h3>
+HTML-CALC:                    <div>
+HTML-CALC:                        <b>a</b>      <div>
+HTML-CALC:                        <p> First integer.</p>
+HTML-CALC:                    </div>
+HTML-CALC:                    <div>
+HTML-CALC:                        <p></p>
+HTML-CALC:                    </div>
+HTML-CALC:                    </div>
+HTML-CALC:                    <div>
+HTML-CALC:                        <b>b</b>      <div>
+HTML-CALC:                        <p> Second integer.</p>
+HTML-CALC:                    </div>
+HTML-CALC:                    <div>
+HTML-CALC:                        <p></p>
+HTML-CALC:                    </div>
+HTML-CALC:                    </div>
+HTML-CALC:                    <h3>Returns</h3>
+HTML-CALC:                        <p> The result of a % b.</p>
+HTML-CALC:                        </div>
 HTML-CALC:     </div>
 HTML-CALC: </div>
 HTML-CALC: </div>
@@ -344,6 +512,31 @@ HTML-RECTANGLE:             <code class="language-cpp code-clang-doc">
 HTML-RECTANGLE: void Rectangle (double width, double height)
 HTML-RECTANGLE:             </code>
 HTML-RECTANGLE:         </pre>
+HTML-RECTANGLE:        <div>
+HTML-RECTANGLE:                        <div>
+HTML-RECTANGLE:                            <p> Constructs a new Rectangle object.</p>
+HTML-RECTANGLE:                        </div>
+HTML-RECTANGLE:                            <div>
+HTML-RECTANGLE:                        <p></p>
+HTML-RECTANGLE:                    </div>
+HTML-RECTANGLE:                        <div>
+HTML-RECTANGLE:                        <p></p>
+HTML-RECTANGLE:                    </div>
+HTML-RECTANGLE:                    <h3>Parameters</h3>
+HTML-RECTANGLE:                    <div>
+HTML-RECTANGLE:                        <b>width</b>      <div>
+HTML-RECTANGLE:                        <p> Width of the rectangle.</p>
+HTML-RECTANGLE:                    </div>
+HTML-RECTANGLE:                    <div>
+HTML-RECTANGLE:                        <p></p>
+HTML-RECTANGLE:                    </div>
+HTML-RECTANGLE:                    </div>
+HTML-RECTANGLE:                    <div>
+HTML-RECTANGLE:                        <b>height</b>      <div>
+HTML-RECTANGLE:                        <p> Height of the rectangle.</p>
+HTML-RECTANGLE:                    </div>
+HTML-RECTANGLE:                    </div>
+HTML-RECTANGLE:                        </div>
 HTML-RECTANGLE:     </div>
 HTML-RECTANGLE: </div>
 HTML-RECTANGLE: <div class="delimiter-container">
@@ -353,6 +546,19 @@ HTML-RECTANGLE:             <code class="language-cpp code-clang-doc">
 HTML-RECTANGLE: double area ()
 HTML-RECTANGLE:             </code>
 HTML-RECTANGLE:         </pre>
+HTML-RECTANGLE:        <div>
+HTML-RECTANGLE:                        <div>
+HTML-RECTANGLE:                            <p> Calculates the area of the rectangle.</p>
+HTML-RECTANGLE:                        </div>
+HTML-RECTANGLE:                            <div>
+HTML-RECTANGLE:                        <p></p>
+HTML-RECTANGLE:                    </div>
+HTML-RECTANGLE:                        <div>
+HTML-RECTANGLE:                        <p></p>
+HTML-RECTANGLE:                    </div>
+HTML-RECTANGLE:                    <h3>Returns</h3>
+HTML-RECTANGLE:                        <p> double The area of the rectangle.</p>
+HTML-RECTANGLE:                        </div>
 HTML-RECTANGLE:     </div>
 HTML-RECTANGLE: </div>
 HTML-RECTANGLE: <div class="delimiter-container">
@@ -362,6 +568,19 @@ HTML-RECTANGLE:             <code class="language-cpp code-clang-doc">
 HTML-RECTANGLE: double perimeter ()
 HTML-RECTANGLE:             </code>
 HTML-RECTANGLE:         </pre>
+HTML-RECTANGLE:        <div>
+HTML-RECTANGLE:                        <div>
+HTML-RECTANGLE:                            <p> Calculates the perimeter of the rectangle.</p>
+HTML-RECTANGLE:                        </div>
+HTML-RECTANGLE:                            <div>
+HTML-RECTANGLE:                        <p></p>
+HTML-RECTANGLE:                    </div>
+HTML-RECTANGLE:                        <div>
+HTML-RECTANGLE:                        <p></p>
+HTML-RECTANGLE:                    </div>
+HTML-RECTANGLE:                    <h3>Returns</h3>
+HTML-RECTANGLE:                        <p> double The perimeter of the rectangle.</p>
+HTML-RECTANGLE:                        </div>
 HTML-RECTANGLE:     </div>
 HTML-RECTANGLE: </div>
 HTML-RECTANGLE: </div>
@@ -450,6 +669,23 @@ HTML-CIRCLE:             <code class="language-cpp code-clang-doc">
 HTML-CIRCLE: void Circle (double radius)
 HTML-CIRCLE:             </code>
 HTML-CIRCLE:         </pre>
+HTML-CIRCLE:        <div>
+HTML-CIRCLE:                        <div>
+HTML-CIRCLE:                            <p> Constructs a new Circle object.</p>
+HTML-CIRCLE:                        </div>
+HTML-CIRCLE:                            <div>
+HTML-CIRCLE:                        <p></p>
+HTML-CIRCLE:                    </div>
+HTML-CIRCLE:                        <div>
+HTML-CIRCLE:                        <p></p>
+HTML-CIRCLE:                    </div>
+HTML-CIRCLE:                    <h3>Parameters</h3>
+HTML-CIRCLE:                    <div>
+HTML-CIRCLE:                        <b>radius</b>      <div>
+HTML-CIRCLE:                        <p> Radius of the circle.</p>
+HTML-CIRCLE:                    </div>
+HTML-CIRCLE:                    </div>
+HTML-CIRCLE:                        </div>
 HTML-CIRCLE:     </div>
 HTML-CIRCLE: </div>
 HTML-CIRCLE: <div class="delimiter-container">
@@ -459,6 +695,19 @@ HTML-CIRCLE:             <code class="language-cpp code-clang-doc">
 HTML-CIRCLE: double area ()
 HTML-CIRCLE:             </code>
 HTML-CIRCLE:         </pre>
+HTML-CIRCLE:        <div>
+HTML-CIRCLE:                        <div>
+HTML-CIRCLE:                            <p> Calculates the area of the circle.</p>
+HTML-CIRCLE:                        </div>
+HTML-CIRCLE:                            <div>
+HTML-CIRCLE:                        <p></p>
+HTML-CIRCLE:                    </div>
+HTML-CIRCLE:                        <div>
+HTML-CIRCLE:                        <p></p>
+HTML-CIRCLE:                    </div>
+HTML-CIRCLE:                    <h3>Returns</h3>
+HTML-CIRCLE:                        <p> double The area of the circle.</p>
+HTML-CIRCLE:                        </div>
 HTML-CIRCLE:     </div>
 HTML-CIRCLE: </div>
 HTML-CIRCLE: <div class="delimiter-container">
@@ -468,6 +717,28 @@ HTML-CIRCLE:             <code class="language-cpp code-clang-doc">
 HTML-CIRCLE: double perimeter ()
 HTML-CIRCLE:             </code>
 HTML-CIRCLE:         </pre>
+HTML-CIRCLE:        <div>
+HTML-CIRCLE:                        <div>
+HTML-CIRCLE:                            <p> Calculates the perimeter of the circle.</p>
+HTML-CIRCLE:                        </div>
+HTML-CIRCLE:                            <div>
+HTML-CIRCLE:                        <p></p>
+HTML-CIRCLE:                    </div>
+HTML-CIRCLE:                        <div>
+HTML-CIRCLE:                        <p></p>
+HTML-CIRCLE:                    </div>
+HTML-CIRCLE:                    <h3>Returns</h3>
+HTML-CIRCLE:                        <p> double The perimeter of the circle.</p>
+HTML-CIRCLE:                    <h3>Code</h3> 
+HTML-CIRCLE:                    <div>
+HTML-CIRCLE:                        <pre class="code-block">
+HTML-CIRCLE:                            <code>
+HTML-CIRCLE:                            Circle circle(5.0);
+HTML-CIRCLE:                            double perimeter = circle.perimeter();
+HTML-CIRCLE:                            </code>
+HTML-CIRCLE:                        </pre>
+HTML-CIRCLE:                    </div>
+HTML-CIRCLE:                        </div>
 HTML-CIRCLE:     </div>
 HTML-CIRCLE: </div>
 HTML-CIRCLE: </div>
diff --git a/clang-tools-extra/test/clang-tidy/checkers/modernize/use-designated-initializers.cpp b/clang-tools-extra/test/clang-tidy/checkers/modernize/use-designated-initializers.cpp
index cdadeed..88e0636 100644
--- a/clang-tools-extra/test/clang-tidy/checkers/modernize/use-designated-initializers.cpp
+++ b/clang-tools-extra/test/clang-tidy/checkers/modernize/use-designated-initializers.cpp
@@ -224,3 +224,55 @@ std::array a{1,2,3};
 std::array<int,2> b{10, 11};
 using array = std::array<int, 2>;
 array c{10, 11};
+
+struct S16 {
+    int a;
+    int b;
+};
+
+using S17 = S16;
+
+S17 s171{1, 2};
+// CHECK-MESSAGES: :[[@LINE-1]]:9: warning: use designated initializer list to initialize 'S17' (aka 'S16') [modernize-use-designated-initializers]
+// CHECK-MESSAGES: :[[@LINE-9]]:1: note: aggregate type is defined here
+// CHECK-FIXES: S17 s171{.a=1, .b=2};
+// CHECK-MESSAGES-POD: :[[@LINE-4]]:9: warning: use designated initializer list to initialize 'S17' (aka 'S16') [modernize-use-designated-initializers]
+// CHECK-MESSAGES-POD: :[[@LINE-12]]:1: note: aggregate type is defined here
+// CHECK-FIXES-POD: S17 s171{.a=1, .b=2};
+
+S17 s172{.a=1, 2};
+// CHECK-MESSAGES: :[[@LINE-1]]:16: warning: use designated init expression to initialize field 'b' [modernize-use-designated-initializers]
+// CHECK-MESSAGES-POD: :[[@LINE-2]]:16: warning: use designated init expression to initialize field 'b' [modernize-use-designated-initializers]
+// CHECK-FIXES: S17 s172{.a=1, .b=2};
+
+S17 s173{.a=1, .b=2}; // no issue
+
+typedef S16 S18;
+
+S18 s181{1, 2};
+// CHECK-MESSAGES: :[[@LINE-1]]:9: warning: use designated initializer list to initialize 'S18' (aka 'S16') [modernize-use-designated-initializers]
+// CHECK-MESSAGES-POD: :[[@LINE-2]]:9: warning: use designated initializer list to initialize 'S18' (aka 'S16') [modernize-use-designated-initializers]
+
+S18 s182{1, .b=2};
+// CHECK-MESSAGES: :[[@LINE-1]]:10: warning: use designated init expression to initialize field 'a' [modernize-use-designated-initializers]
+// CHECK-MESSAGES-POD: :[[@LINE-2]]:10: warning: use designated init expression to initialize field 'a' [modernize-use-designated-initializers]
+// CHECK-FIXES: S18 s182{.a=1, .b=2};
+
+S18 s183{.a=1, .b=2}; // no issue
+
+struct S19 {
+    int i;
+    S17 s17;
+    S18 s18;
+};
+
+S19 s191{1, {2, .b=3}, {4, 5}};
+// CHECK-MESSAGES: :[[@LINE-1]]:9: warning: use designated initializer list to initialize 'S19' [modernize-use-designated-initializers]
+// CHECK-MESSAGES: :[[@LINE-8]]:1: note: aggregate type is defined here
+// CHECK-MESSAGES: :[[@LINE-3]]:14: warning: use designated init expression to initialize field 'a' [modernize-use-designated-initializers]
+// CHECK-MESSAGES: :[[@LINE-4]]:24: warning: use designated initializer list to initialize 'S18' (aka 'S16') [modernize-use-designated-initializers]
+// CHECK-MESSAGES-POD: :[[@LINE-5]]:9: warning: use designated initializer list to initialize 'S19' [modernize-use-designated-initializers]
+// CHECK-MESSAGES-POD: :[[@LINE-12]]:1: note: aggregate type is defined here
+// CHECK-MESSAGES-POD: :[[@LINE-7]]:14: warning: use designated init expression to initialize field 'a' [modernize-use-designated-initializers]
+// CHECK-MESSAGES-POD: :[[@LINE-8]]:24: warning: use designated initializer list to initialize 'S18' (aka 'S16') [modernize-use-designated-initializers]
+// CHECK-FIXES: S19 s191{.i=1, .s17={.a=2, .b=3}, .s18={.a=4, .b=5}};
diff --git a/clang/docs/ClangFormatStyleOptions.rst b/clang/docs/ClangFormatStyleOptions.rst
index 7f970f6..02986a9 100644
--- a/clang/docs/ClangFormatStyleOptions.rst
+++ b/clang/docs/ClangFormatStyleOptions.rst
@@ -1912,7 +1912,7 @@ the configuration (without a prefix: ``Auto``).
 
   * ``SFS_InlineOnly`` (in configuration: ``InlineOnly``)
     Only merge functions defined inside a class. Same as ``inline``,
-    except it does not implies ``empty``: i.e. top level empty functions
+    except it does not imply ``empty``: i.e. top level empty functions
     are not merged either.
 
     .. code-block:: c++
diff --git a/clang/docs/OpenMPSupport.rst b/clang/docs/OpenMPSupport.rst
index fc44f4c..58cd10a 100644
--- a/clang/docs/OpenMPSupport.rst
+++ b/clang/docs/OpenMPSupport.rst
@@ -1,500 +1,506 @@
-.. raw:: html
-
-  <style type="text/css">
-    .none { background-color: #FFCCCC }
-    .part { background-color: #FFFF99 }
-    .good { background-color: #CCFF99 }
-  </style>
-
-.. role:: none
-.. role:: part
-.. role:: good
-
-.. contents::
-   :local:
-
-==============
-OpenMP Support
-==============
-
-Clang fully supports OpenMP 4.5, almost all of 5.0 and most of 5.1/2.
-Clang supports offloading to X86_64, AArch64, PPC64[LE], NVIDIA GPUs (all models) and AMD GPUs (all models).
-
-In addition, the LLVM OpenMP runtime `libomp` supports the OpenMP Tools
-Interface (OMPT) on x86, x86_64, AArch64, and PPC64 on Linux, Windows, and macOS.
-OMPT is also supported for NVIDIA and AMD GPUs.
-
-For the list of supported features from OpenMP 5.0 and 5.1
-see `OpenMP implementation details`_ and `OpenMP 51 implementation details`_.
-
-General improvements
-====================
-- New collapse clause scheme to avoid expensive remainder operations.
-  Compute loop index variables after collapsing a loop nest via the
-  collapse clause by replacing the expensive remainder operation with
-  multiplications and additions.
-
-- When using the collapse clause on a loop nest the default behavior
-  is to automatically extend the representation of the loop counter to
-  64 bits for the cases where the sizes of the collapsed loops are not
-  known at compile time. To prevent this conservative choice and use
-  at most 32 bits, compile your program with the
-  `-fopenmp-optimistic-collapse`.
-
-
-GPU devices support
-===================
-
-Data-sharing modes
-------------------
-
-Clang supports two data-sharing models for Cuda devices: `Generic` and `Cuda`
-modes. The default mode is `Generic`. `Cuda` mode can give an additional
-performance and can be activated using the `-fopenmp-cuda-mode` flag. In
-`Generic` mode all local variables that can be shared in the parallel regions
-are stored in the global memory. In `Cuda` mode local variables are not shared
-between the threads and it is user responsibility to share the required data
-between the threads in the parallel regions. Often, the optimizer is able to
-reduce the cost of `Generic` mode to the level of `Cuda` mode, but the flag,
-as well as other assumption flags, can be used for tuning.
-
-Features not supported or with limited support for Cuda devices
----------------------------------------------------------------
-
-- Cancellation constructs are not supported.
-
-- Doacross loop nest is not supported.
-
-- User-defined reductions are supported only for trivial types.
-
-- Nested parallelism: inner parallel regions are executed sequentially.
-
-- Debug information for OpenMP target regions is supported, but sometimes it may
-  be required to manually specify the address class of the inspected variables.
-  In some cases the local variables are actually allocated in the global memory,
-  but the debug info may be not aware of it.
-
-
-.. _OpenMP implementation details:
-
-OpenMP 5.0 Implementation Details
-=================================
-
-The following table provides a quick overview over various OpenMP 5.0 features
-and their implementation status. Please post on the
-`Discourse forums (Runtimes - OpenMP category)`_ for more
-information or if you want to help with the
-implementation.
-
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-|Category                      | Feature                                                      | Status                   | Reviews                                                               |
-+==============================+==============================================================+==========================+=======================================================================+
-| loop                         | support != in the canonical loop form                        | :good:`done`             | D54441                                                                |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| loop                         | #pragma omp loop (directive)                                 | :part:`partial`          | D145823 (combined forms)                                              |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| loop                         | #pragma omp loop bind                                        | :part:`worked on`        | D144634 (needs review)                                                |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| loop                         | collapse imperfectly nested loop                             | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| loop                         | collapse non-rectangular nested loop                         | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| loop                         | C++ range-base for loop                                      | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| loop                         | clause: if for SIMD directives                               | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| loop                         | inclusive scan (matching C++17 PSTL)                         | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| memory management            | memory allocators                                            | :good:`done`             | r341687,r357929                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| memory management            | allocate directive and allocate clause                       | :good:`done`             | r355614,r335952                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| OMPD                         | OMPD interfaces                                              | :good:`done`             | https://reviews.llvm.org/D99914   (Supports only HOST(CPU) and Linux  |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| OMPT                         | OMPT interfaces (callback support)                           | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| thread affinity              | thread affinity                                              | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| task                         | taskloop reduction                                           | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| task                         | task affinity                                                | :part:`not upstream`     | https://github.com/jklinkenberg/openmp/tree/task-affinity             |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| task                         | clause: depend on the taskwait construct                     | :good:`done`             | D113540 (regular codegen only)                                        |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| task                         | depend objects and detachable tasks                          | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| task                         | mutexinoutset dependence-type for tasks                      | :good:`done`             | D53380,D57576                                                         |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| task                         | combined taskloop constructs                                 | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| task                         | master taskloop                                              | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| task                         | parallel master taskloop                                     | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| task                         | master taskloop simd                                         | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| task                         | parallel master taskloop simd                                | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| SIMD                         | atomic and simd constructs inside SIMD code                  | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| SIMD                         | SIMD nontemporal                                             | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | infer target functions from initializers                     | :part:`worked on`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | infer target variables from initializers                     | :good:`done`             | D146418                                                               |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | OMP_TARGET_OFFLOAD environment variable                      | :good:`done`             | D50522                                                                |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | support full 'defaultmap' functionality                      | :good:`done`             | D69204                                                                |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | device specific functions                                    | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | clause: device_type                                          | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | clause: extended device                                      | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | clause: uses_allocators clause                               | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | clause: in_reduction                                         | :part:`worked on`        | r308768                                                               |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | omp_get_device_num()                                         | :good:`done`             | D54342,D128347                                                        |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | structure mapping of references                              | :none:`unclaimed`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | nested target declare                                        | :good:`done`             | D51378                                                                |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | implicitly map 'this' (this[:1])                             | :good:`done`             | D55982                                                                |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | allow access to the reference count (omp_target_is_present)  | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | requires directive                                           | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | clause: unified_shared_memory                                | :good:`done`             | D52625,D52359                                                         |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | clause: unified_address                                      | :part:`partial`          |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | clause: reverse_offload                                      | :part:`partial`          | D52780,D155003                                                        |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | clause: atomic_default_mem_order                             | :good:`done`             | D53513                                                                |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | clause: dynamic_allocators                                   | :part:`unclaimed parts`  | D53079                                                                |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | user-defined mappers                                         | :good:`done`             | D56326,D58638,D58523,D58074,D60972,D59474                             |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | map array-section with implicit mapper                       | :good:`done`             |  https://github.com/llvm/llvm-project/pull/101101                     |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | mapping lambda expression                                    | :good:`done`             | D51107                                                                |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | clause: use_device_addr for target data                      | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | support close modifier on map clause                         | :good:`done`             | D55719,D55892                                                         |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | teams construct on the host device                           | :good:`done`             | r371553                                                               |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | support non-contiguous array sections for target update      | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | pointer attachment                                           | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| atomic                       | hints for the atomic construct                               | :good:`done`             | D51233                                                                |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| base language                | C11 support                                                  | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| base language                | C++11/14/17 support                                          | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| base language                | lambda support                                               | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | array shaping                                                | :good:`done`             | D74144                                                                |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | library shutdown (omp_pause_resource[_all])                  | :good:`done`             | D55078                                                                |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | metadirectives                                               | :part:`mostly done`      | D91944, https://github.com/llvm/llvm-project/pull/128640              |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | conditional modifier for lastprivate clause                  | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | iterator and multidependences                                | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | depobj directive and depobj dependency kind                  | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | user-defined function variants                               | :good:`done`.            | D67294, D64095, D71847, D71830, D109635                               |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | pointer/reference to pointer based array reductions          | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | prevent new type definitions in clauses                      | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| memory model                 | memory model update (seq_cst, acq_rel, release, acquire,...) | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-
-
-.. _OpenMP 51 implementation details:
-
-OpenMP 5.1 Implementation Details
-=================================
-
-The following table provides a quick overview over various OpenMP 5.1 features
-and their implementation status.
-Please post on the
-`Discourse forums (Runtimes - OpenMP category)`_ for more
-information or if you want to help with the
-implementation.
-
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-|Category                      | Feature                                                      | Status                   | Reviews                                                               |
-+==============================+==============================================================+==========================+=======================================================================+
-| atomic                       | 'compare' clause on atomic construct                         | :good:`done`             | D120290, D120007, D118632, D120200, D116261, D118547, D116637         |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| atomic                       | 'fail' clause on atomic construct                            | :part:`worked on`        | D123235 (in progress)                                                 |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| base language                | C++ attribute specifier syntax                               | :good:`done`             | D105648                                                               |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | 'present' map type modifier                                  | :good:`done`             | D83061, D83062, D84422                                                |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | 'present' motion modifier                                    | :good:`done`             | D84711, D84712                                                        |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | 'present' in defaultmap clause                               | :good:`done`             | D92427                                                                |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | map clause reordering based on 'present' modifier            | :none:`unclaimed`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | device-specific environment variables                        | :none:`unclaimed`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | omp_target_is_accessible routine                             | :none:`unclaimed`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | omp_get_mapped_ptr routine                                   | :good:`done`             | D141545                                                               |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | new async target memory copy routines                        | :good:`done`             | D136103                                                               |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | thread_limit clause on target construct                      | :part:`partial`          | D141540 (offload), D152054 (host, in progress)                        |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | has_device_addr clause on target construct                   | :none:`unclaimed`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | iterators in map clause or motion clauses                    | :none:`unclaimed`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | indirect clause on declare target directive                  | :none:`unclaimed`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | allow virtual functions calls for mapped object on device    | :part:`partial`          |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | interop construct                                            | :part:`partial`          | parsing/sema done: D98558, D98834, D98815                             |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| device                       | assorted routines for querying interoperable properties      | :part:`partial`          | D106674                                                               |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| loop                         | Loop tiling transformation                                   | :good:`done`             | D76342                                                                |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| loop                         | Loop unrolling transformation                                | :good:`done`             | D99459                                                                |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| loop                         | 'reproducible'/'unconstrained' modifiers in 'order' clause   | :part:`partial`          | D127855                                                               |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| memory management            | alignment for allocate directive and clause                  | :good:`done`             | D115683                                                               |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| memory management            | 'allocator' modifier for allocate clause                     | :good:`done`             | https://github.com/llvm/llvm-project/pull/114883                      |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| memory management            | 'align' modifier for allocate clause                         | :good:`done`             | https://github.com/llvm/llvm-project/pull/121814                      |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| memory management            | new memory management routines                               | :none:`unclaimed`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| memory management            | changes to omp_alloctrait_key enum                           | :none:`unclaimed`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| memory model                 | seq_cst clause on flush construct                            | :good:`done`             | https://github.com/llvm/llvm-project/pull/114072                      |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | 'omp_all_memory' keyword and use in 'depend' clause          | :good:`done`             | D125828, D126321                                                      |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | error directive                                              | :good:`done`             | D139166                                                               |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | scope construct                                              | :good:`done`             | D157933, https://github.com/llvm/llvm-project/pull/109197             |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | routines for controlling and querying team regions           | :part:`partial`          | D95003 (libomp only)                                                  |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | changes to ompt_scope_endpoint_t enum                        | :none:`unclaimed`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | omp_display_env routine                                      | :good:`done`             | D74956                                                                |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | extended OMP_PLACES syntax                                   | :none:`unclaimed`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | OMP_NUM_TEAMS and OMP_TEAMS_THREAD_LIMIT env vars            | :good:`done`             | D138769                                                               |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | 'target_device' selector in context specifier                | :none:`worked on`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | begin/end declare variant                                    | :good:`done`             | D71179                                                                |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | dispatch construct and function variant argument adjustment  | :part:`worked on`        | D99537, D99679                                                        |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | assumes directives                                           | :part:`worked on`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | assume directive                                             | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | nothing directive                                            | :good:`done`             | D123286                                                               |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | masked construct and related combined constructs             | :good:`done`             | D99995, D100514, PR-121741(parallel_masked_taskloop)                  |
-|                              |                                                              |                          | PR-121746(parallel_masked_task_loop_simd),PR-121914(masked_taskloop)  |
-|                              |                                                              |                          | PR-121916(masked_taskloop_simd)                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| misc                         | default(firstprivate) & default(private)                     | :good:`done`             | D75591 (firstprivate), D125912 (private)                              |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| other                        | deprecating master construct                                 | :none:`unclaimed`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| OMPT                         | new barrier types added to ompt_sync_region_t enum           | :none:`unclaimed`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| OMPT                         | async data transfers added to ompt_target_data_op_t enum     | :none:`unclaimed`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| OMPT                         | new barrier state values added to ompt_state_t enum          | :none:`unclaimed`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| OMPT                         | new 'emi' callbacks for external monitoring interfaces       | :good:`done`             |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| OMPT                         | device tracing interface                                     | :none:`unclaimed`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| task                         | 'strict' modifier for taskloop construct                     | :none:`unclaimed`        |                                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| task                         | inoutset in depend clause                                    | :good:`done`             | D97085, D118383                                                       |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-| task                         | nowait clause on taskwait                                    | :part:`partial`          | parsing/sema done: D131830, D141531                                   |
-+------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+
-
-
-.. _OpenMP 6.0 implementation details:
-
-OpenMP 6.0 Implementation Details
-=================================
-
-The following table provides a quick overview over various OpenMP 6.0 features
-and their implementation status. Please post on the
-`Discourse forums (Runtimes - OpenMP category)`_ for more
-information or if you want to help with the
-implementation.
-
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-|Feature                                                      | C/C++ Status              |  Fortran Status           | Reviews                                                                  |
-+=============================================================+===========================+===========================+==========================================================================+
-| free-agent threads                                          | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| threadset clause                                            | :`worked on`              | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Recording of task graphs                                    | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Parallel inductions                                         | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| init_complete for scan directive                            | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Loop transformation constructs                              | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| loop stripe transformation                                  | :good:`done`              | https://github.com/llvm/llvm-project/pull/119891                                                     |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| work distribute construct                                   | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| task_iteration                                              | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| memscope clause for atomic and flush                        | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| transparent clause (hull tasks)                             | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| rule-based compound directives                              | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| C23, C++23                                                  | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Fortran 2023                                                | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| decl attribute for declarative directives                   | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| C attribute syntax                                          | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| pure directives in DO CONCURRENT                            | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Optional argument for all clauses                           | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Function references for locator list items                  | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| All clauses accept directive name modifier                  | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Extensions to depobj construct                              | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Extensions to atomic construct                              | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+ 
-| Private reductions                                          | :good:`mostly`            | :none:`unclaimed`         | Parse/Sema:https://github.com/llvm/llvm-project/pull/129938              |
-|                                                             |                           |                           | Codegen: https://github.com/llvm/llvm-project/pull/134709                |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Self maps                                                   | :part:`partial`           | :none:`unclaimed`         | parsing/sema done: https://github.com/llvm/llvm-project/pull/129888      |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Release map type for declare mapper                         | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Extensions to interop construct                             | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| no_openmp_constructs                                        | :good:`done`              | :none:`unclaimed`         | https://github.com/llvm/llvm-project/pull/125933                         |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| safe_sync and progress with identifier and API              | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| OpenMP directives in concurrent loop regions                | :good:`done`              | :none:`unclaimed`         | https://github.com/llvm/llvm-project/pull/125621                         |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| atomics constructs on concurrent loop regions               | :good:`done`              | :none:`unclaimed`         | https://github.com/llvm/llvm-project/pull/125621                         |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Loop construct with DO CONCURRENT                           | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| device_type clause for target construct                     | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| nowait for ancestor target directives                       | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| New API for devices' num_teams/thread_limit                 | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Host and device environment variables                       | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| num_threads ICV and clause accepts list                     | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Numeric names for environment variables                     | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Increment between places for OMP_PLACES                     | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| OMP_AVAILABLE_DEVICES envirable                             | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Traits for default device envirable                         | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Optionally omit array length expression                     | :good:`done`              | :none:`unclaimed`         | https://github.com/llvm/llvm-project/pull/148048                         |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Canonical loop sequences                                    | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Clarifications to Fortran map semantics                     | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| default clause at target construct                          | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| ref count update use_device_{ptr, addr}                     | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Clarifications to implicit reductions                       | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| ref modifier for map clauses                                | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| map-type modifiers in arbitrary position                    | :good:`done`              | :none:`unclaimed`         | https://github.com/llvm/llvm-project/pull/90499                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Lift nesting restriction on concurrent loop                 | :good:`done`              | :none:`unclaimed`         | https://github.com/llvm/llvm-project/pull/125621                         |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| priority clause for target constructs                       | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| changes to target_data construct                            | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| Non-const do_not_sync for nowait/nogroup                    | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-| need_device_addr modifier for adjust_args clause            | :part:`partial`           | :none:`unclaimed`         | Parsing/Sema: https://github.com/llvm/llvm-project/pull/143442           |
-|                                                             |                           |                           |               https://github.com/llvm/llvm-project/pull/149586           |
-+-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+
-
-OpenMP Extensions
-=================
-
-The following table provides a quick overview over various OpenMP
-extensions and their implementation status.  These extensions are not
-currently defined by any standard, so links to associated LLVM
-documentation are provided.  As these extensions mature, they will be
-considered for standardization. Please post on the
-`Discourse forums (Runtimes - OpenMP category)`_ to provide feedback.
-
-+------------------------------+-----------------------------------------------------------------------------------+--------------------------+--------------------------------------------------------+
-|Category                      | Feature                                                                           | Status                   | Reviews                                                |
-+==============================+===================================================================================+==========================+========================================================+
-| atomic extension             | `'atomic' strictly nested within 'teams'                                          | :good:`prototyped`       | D126323                                                |
-|                              | <https://openmp.llvm.org/docs/openacc/OpenMPExtensions.html#atomicWithinTeams>`_  |                          |                                                        |
-+------------------------------+-----------------------------------------------------------------------------------+--------------------------+--------------------------------------------------------+
-| device extension             | `'ompx_hold' map type modifier                                                    | :good:`prototyped`       | D106509, D106510                                       |
-|                              | <https://openmp.llvm.org/docs/openacc/OpenMPExtensions.html#ompx-hold>`_          |                          |                                                        |
-+------------------------------+-----------------------------------------------------------------------------------+--------------------------+--------------------------------------------------------+
-| device extension             | `'ompx_bare' clause on 'target teams' construct                                   | :good:`prototyped`       | #66844, #70612                                         |
-|                              | <https://www.osti.gov/servlets/purl/2205717>`_                                    |                          |                                                        |
-+------------------------------+-----------------------------------------------------------------------------------+--------------------------+--------------------------------------------------------+
-| device extension             | Multi-dim 'num_teams' and 'thread_limit' clause on 'target teams ompx_bare'       | :good:`partial`          | #99732, #101407, #102715                               |
-|                              | construct                                                                         |                          |                                                        |
-+------------------------------+-----------------------------------------------------------------------------------+--------------------------+--------------------------------------------------------+
-
-.. _Discourse forums (Runtimes - OpenMP category): https://discourse.llvm.org/c/runtimes/openmp/35
+.. raw:: html

+

+  <style type="text/css">

+    .none { background-color: #FFCCCC }

+    .part { background-color: #FFFF99 }

+    .good { background-color: #CCFF99 }

+  </style>

+

+.. role:: none

+.. role:: part

+.. role:: good

+

+.. contents::

+   :local:

+

+==============

+OpenMP Support

+==============

+

+Clang fully supports OpenMP 4.5, almost all of 5.0 and most of 5.1/2.

+Clang supports offloading to X86_64, AArch64, PPC64[LE], NVIDIA GPUs (all models) and AMD GPUs (all models).

+

+In addition, the LLVM OpenMP runtime `libomp` supports the OpenMP Tools

+Interface (OMPT) on x86, x86_64, AArch64, and PPC64 on Linux, Windows, and macOS.

+OMPT is also supported for NVIDIA and AMD GPUs.

+

+For the list of supported features from OpenMP 5.0 and 5.1

+see `OpenMP implementation details`_ and `OpenMP 51 implementation details`_.

+

+General improvements

+====================

+- New collapse clause scheme to avoid expensive remainder operations.

+  Compute loop index variables after collapsing a loop nest via the

+  collapse clause by replacing the expensive remainder operation with

+  multiplications and additions.

+

+- When using the collapse clause on a loop nest the default behavior

+  is to automatically extend the representation of the loop counter to

+  64 bits for the cases where the sizes of the collapsed loops are not

+  known at compile time. To prevent this conservative choice and use

+  at most 32 bits, compile your program with the

+  `-fopenmp-optimistic-collapse`.

+

+

+GPU devices support

+===================

+

+Data-sharing modes

+------------------

+

+Clang supports two data-sharing models for Cuda devices: `Generic` and `Cuda`

+modes. The default mode is `Generic`. `Cuda` mode can give an additional

+performance and can be activated using the `-fopenmp-cuda-mode` flag. In

+`Generic` mode all local variables that can be shared in the parallel regions

+are stored in the global memory. In `Cuda` mode local variables are not shared

+between the threads and it is user responsibility to share the required data

+between the threads in the parallel regions. Often, the optimizer is able to

+reduce the cost of `Generic` mode to the level of `Cuda` mode, but the flag,

+as well as other assumption flags, can be used for tuning.

+

+Features not supported or with limited support for Cuda devices

+---------------------------------------------------------------

+

+- Cancellation constructs are not supported.

+

+- Doacross loop nest is not supported.

+

+- User-defined reductions are supported only for trivial types.

+

+- Nested parallelism: inner parallel regions are executed sequentially.

+

+- Debug information for OpenMP target regions is supported, but sometimes it may

+  be required to manually specify the address class of the inspected variables.

+  In some cases the local variables are actually allocated in the global memory,

+  but the debug info may be not aware of it.

+

+

+.. _OpenMP implementation details:

+

+OpenMP 5.0 Implementation Details

+=================================

+

+The following table provides a quick overview over various OpenMP 5.0 features

+and their implementation status. Please post on the

+`Discourse forums (Runtimes - OpenMP category)`_ for more

+information or if you want to help with the

+implementation.

+

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+|Category                      | Feature                                                      | Status                   | Reviews                                                               |

++==============================+==============================================================+==========================+=======================================================================+

+| loop                         | support != in the canonical loop form                        | :good:`done`             | D54441                                                                |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| loop                         | #pragma omp loop (directive)                                 | :part:`partial`          | D145823 (combined forms)                                              |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| loop                         | #pragma omp loop bind                                        | :part:`worked on`        | D144634 (needs review)                                                |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| loop                         | collapse imperfectly nested loop                             | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| loop                         | collapse non-rectangular nested loop                         | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| loop                         | C++ range-base for loop                                      | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| loop                         | clause: if for SIMD directives                               | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| loop                         | inclusive scan (matching C++17 PSTL)                         | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| memory management            | memory allocators                                            | :good:`done`             | r341687,r357929                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| memory management            | allocate directive and allocate clause                       | :good:`done`             | r355614,r335952                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| OMPD                         | OMPD interfaces                                              | :good:`done`             | https://reviews.llvm.org/D99914   (Supports only HOST(CPU) and Linux  |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| OMPT                         | OMPT interfaces (callback support)                           | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| thread affinity              | thread affinity                                              | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| task                         | taskloop reduction                                           | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| task                         | task affinity                                                | :part:`not upstream`     | https://github.com/jklinkenberg/openmp/tree/task-affinity             |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| task                         | clause: depend on the taskwait construct                     | :good:`done`             | D113540 (regular codegen only)                                        |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| task                         | depend objects and detachable tasks                          | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| task                         | mutexinoutset dependence-type for tasks                      | :good:`done`             | D53380,D57576                                                         |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| task                         | combined taskloop constructs                                 | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| task                         | master taskloop                                              | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| task                         | parallel master taskloop                                     | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| task                         | master taskloop simd                                         | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| task                         | parallel master taskloop simd                                | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| SIMD                         | atomic and simd constructs inside SIMD code                  | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| SIMD                         | SIMD nontemporal                                             | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | infer target functions from initializers                     | :part:`worked on`        |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | infer target variables from initializers                     | :good:`done`             | D146418                                                               |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | OMP_TARGET_OFFLOAD environment variable                      | :good:`done`             | D50522                                                                |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | support full 'defaultmap' functionality                      | :good:`done`             | D69204                                                                |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | device specific functions                                    | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | clause: device_type                                          | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | clause: extended device                                      | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | clause: uses_allocators clause                               | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | clause: in_reduction                                         | :part:`worked on`        | r308768                                                               |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | omp_get_device_num()                                         | :good:`done`             | D54342,D128347                                                        |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | structure mapping of references                              | :none:`unclaimed`        |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | nested target declare                                        | :good:`done`             | D51378                                                                |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | implicitly map 'this' (this[:1])                             | :good:`done`             | D55982                                                                |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | allow access to the reference count (omp_target_is_present)  | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | requires directive                                           | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | clause: unified_shared_memory                                | :good:`done`             | D52625,D52359                                                         |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | clause: unified_address                                      | :part:`partial`          |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | clause: reverse_offload                                      | :part:`partial`          | D52780,D155003                                                        |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | clause: atomic_default_mem_order                             | :good:`done`             | D53513                                                                |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | clause: dynamic_allocators                                   | :part:`unclaimed parts`  | D53079                                                                |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | user-defined mappers                                         | :good:`done`             | D56326,D58638,D58523,D58074,D60972,D59474                             |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | map array-section with implicit mapper                       | :good:`done`             |  https://github.com/llvm/llvm-project/pull/101101                     |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | mapping lambda expression                                    | :good:`done`             | D51107                                                                |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | clause: use_device_addr for target data                      | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | support close modifier on map clause                         | :good:`done`             | D55719,D55892                                                         |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | teams construct on the host device                           | :good:`done`             | r371553                                                               |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | support non-contiguous array sections for target update      | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | pointer attachment                                           | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| atomic                       | hints for the atomic construct                               | :good:`done`             | D51233                                                                |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| base language                | C11 support                                                  | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| base language                | C++11/14/17 support                                          | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| base language                | lambda support                                               | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | array shaping                                                | :good:`done`             | D74144                                                                |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | library shutdown (omp_pause_resource[_all])                  | :good:`done`             | D55078                                                                |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | metadirectives                                               | :part:`mostly done`      | D91944, https://github.com/llvm/llvm-project/pull/128640              |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | conditional modifier for lastprivate clause                  | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | iterator and multidependences                                | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | depobj directive and depobj dependency kind                  | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | user-defined function variants                               | :good:`done`.            | D67294, D64095, D71847, D71830, D109635                               |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | pointer/reference to pointer based array reductions          | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | prevent new type definitions in clauses                      | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| memory model                 | memory model update (seq_cst, acq_rel, release, acquire,...) | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+

+

+.. _OpenMP 51 implementation details:

+

+OpenMP 5.1 Implementation Details

+=================================

+

+The following table provides a quick overview over various OpenMP 5.1 features

+and their implementation status.

+Please post on the

+`Discourse forums (Runtimes - OpenMP category)`_ for more

+information or if you want to help with the

+implementation.

+

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+|Category                      | Feature                                                      | Status                   | Reviews                                                               |

++==============================+==============================================================+==========================+=======================================================================+

+| atomic                       | 'compare' clause on atomic construct                         | :good:`done`             | D120290, D120007, D118632, D120200, D116261, D118547, D116637         |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| atomic                       | 'fail' clause on atomic construct                            | :part:`worked on`        | D123235 (in progress)                                                 |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| base language                | C++ attribute specifier syntax                               | :good:`done`             | D105648                                                               |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | 'present' map type modifier                                  | :good:`done`             | D83061, D83062, D84422                                                |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | 'present' motion modifier                                    | :good:`done`             | D84711, D84712                                                        |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | 'present' in defaultmap clause                               | :good:`done`             | D92427                                                                |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | map clause reordering based on 'present' modifier            | :none:`unclaimed`        |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | device-specific environment variables                        | :none:`unclaimed`        |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | omp_target_is_accessible routine                             | :none:`unclaimed`        |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | omp_get_mapped_ptr routine                                   | :good:`done`             | D141545                                                               |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | new async target memory copy routines                        | :good:`done`             | D136103                                                               |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | thread_limit clause on target construct                      | :part:`partial`          | D141540 (offload), D152054 (host, in progress)                        |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | has_device_addr clause on target construct                   | :none:`unclaimed`        |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | iterators in map clause or motion clauses                    | :none:`unclaimed`        |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | indirect clause on declare target directive                  | :part:`In Progress`      |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | allow virtual functions calls for mapped object on device    | :part:`partial`          |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | interop construct                                            | :part:`partial`          | parsing/sema done: D98558, D98834, D98815                             |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| device                       | assorted routines for querying interoperable properties      | :part:`partial`          | D106674                                                               |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| loop                         | Loop tiling transformation                                   | :good:`done`             | D76342                                                                |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| loop                         | Loop unrolling transformation                                | :good:`done`             | D99459                                                                |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| loop                         | 'reproducible'/'unconstrained' modifiers in 'order' clause   | :part:`partial`          | D127855                                                               |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| memory management            | alignment for allocate directive and clause                  | :good:`done`             | D115683                                                               |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| memory management            | 'allocator' modifier for allocate clause                     | :good:`done`             | https://github.com/llvm/llvm-project/pull/114883                      |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| memory management            | 'align' modifier for allocate clause                         | :good:`done`             | https://github.com/llvm/llvm-project/pull/121814                      |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| memory management            | new memory management routines                               | :none:`unclaimed`        |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| memory management            | changes to omp_alloctrait_key enum                           | :none:`unclaimed`        |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| memory model                 | seq_cst clause on flush construct                            | :good:`done`             | https://github.com/llvm/llvm-project/pull/114072                      |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | 'omp_all_memory' keyword and use in 'depend' clause          | :good:`done`             | D125828, D126321                                                      |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | error directive                                              | :good:`done`             | D139166                                                               |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | scope construct                                              | :good:`done`             | D157933, https://github.com/llvm/llvm-project/pull/109197             |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | routines for controlling and querying team regions           | :part:`partial`          | D95003 (libomp only)                                                  |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | changes to ompt_scope_endpoint_t enum                        | :none:`unclaimed`        |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | omp_display_env routine                                      | :good:`done`             | D74956                                                                |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | extended OMP_PLACES syntax                                   | :none:`unclaimed`        |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | OMP_NUM_TEAMS and OMP_TEAMS_THREAD_LIMIT env vars            | :good:`done`             | D138769                                                               |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | 'target_device' selector in context specifier                | :none:`worked on`        |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | begin/end declare variant                                    | :good:`done`             | D71179                                                                |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | dispatch construct and function variant argument adjustment  | :part:`worked on`        | D99537, D99679                                                        |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | assumes directives                                           | :part:`worked on`        |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | assume directive                                             | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | nothing directive                                            | :good:`done`             | D123286                                                               |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | masked construct and related combined constructs             | :good:`done`             | D99995, D100514, PR-121741(parallel_masked_taskloop)                  |

+|                              |                                                              |                          | PR-121746(parallel_masked_task_loop_simd),PR-121914(masked_taskloop)  |

+|                              |                                                              |                          | PR-121916(masked_taskloop_simd)                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| misc                         | default(firstprivate) & default(private)                     | :good:`done`             | D75591 (firstprivate), D125912 (private)                              |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| other                        | deprecating master construct                                 | :none:`unclaimed`        |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| OMPT                         | new barrier types added to ompt_sync_region_t enum           | :none:`unclaimed`        |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| OMPT                         | async data transfers added to ompt_target_data_op_t enum     | :none:`unclaimed`        |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| OMPT                         | new barrier state values added to ompt_state_t enum          | :none:`unclaimed`        |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| OMPT                         | new 'emi' callbacks for external monitoring interfaces       | :good:`done`             |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| OMPT                         | device tracing interface                                     | :none:`unclaimed`        |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| task                         | 'strict' modifier for taskloop construct                     | :none:`unclaimed`        |                                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| task                         | inoutset in depend clause                                    | :good:`done`             | D97085, D118383                                                       |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+| task                         | nowait clause on taskwait                                    | :part:`partial`          | parsing/sema done: D131830, D141531                                   |

++------------------------------+--------------------------------------------------------------+--------------------------+-----------------------------------------------------------------------+

+

+

+.. _OpenMP 6.0 implementation details:

+

+OpenMP 6.0 Implementation Details

+=================================

+

+The following table provides a quick overview over various OpenMP 6.0 features

+and their implementation status. Please post on the

+`Discourse forums (Runtimes - OpenMP category)`_ for more

+information or if you want to help with the

+implementation.

+

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+|Feature                                                      | C/C++ Status              |  Fortran Status           | Reviews                                                                  |

++=============================================================+===========================+===========================+==========================================================================+

+| free-agent threads                                          | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| threadset clause                                            | :`worked on`              | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Recording of task graphs                                    | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Parallel inductions                                         | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| init_complete for scan directive                            | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Loop transformation constructs                              | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| loop stripe transformation                                  | :good:`done`              | https://github.com/llvm/llvm-project/pull/119891                                                     |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| work distribute construct                                   | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| task_iteration                                              | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| memscope clause for atomic and flush                        | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| transparent clause (hull tasks)                             | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| rule-based compound directives                              | :none:`unclaimed`         | :part:`In Progress`       | Testing for Fortran missing                                              |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| C23, C++23                                                  | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Fortran 2023                                                | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| decl attribute for declarative directives                   | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| C attribute syntax                                          | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| pure directives in DO CONCURRENT                            | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Optional argument for all clauses                           | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Function references for locator list items                  | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| All clauses accept directive name modifier                  | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Extensions to depobj construct                              | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Extensions to atomic construct                              | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+ 

+| Private reductions                                          | :good:`mostly`            | :none:`unclaimed`         | Parse/Sema:https://github.com/llvm/llvm-project/pull/129938              |

+|                                                             |                           |                           | Codegen: https://github.com/llvm/llvm-project/pull/134709                |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Self maps                                                   | :part:`partial`           | :none:`unclaimed`         | parsing/sema done: https://github.com/llvm/llvm-project/pull/129888      |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Release map type for declare mapper                         | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Extensions to interop construct                             | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| no_openmp_constructs                                        | :good:`done`              | :none:`unclaimed`         | https://github.com/llvm/llvm-project/pull/125933                         |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| safe_sync and progress with identifier and API              | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| OpenMP directives in concurrent loop regions                | :good:`done`              | :none:`unclaimed`         | https://github.com/llvm/llvm-project/pull/125621                         |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| atomics constructs on concurrent loop regions               | :good:`done`              | :none:`unclaimed`         | https://github.com/llvm/llvm-project/pull/125621                         |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Loop construct with DO CONCURRENT                           | :none:`unclaimed`         | :part:`In Progress`       |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| device_type clause for target construct                     | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| nowait for ancestor target directives                       | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| New API for devices' num_teams/thread_limit                 | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Host and device environment variables                       | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| num_threads ICV and clause accepts list                     | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Numeric names for environment variables                     | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Increment between places for OMP_PLACES                     | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| OMP_AVAILABLE_DEVICES envirable                             | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Traits for default device envirable                         | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Optionally omit array length expression                     | :good:`done`              | :none:`unclaimed`         | https://github.com/llvm/llvm-project/pull/148048                         |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Canonical loop sequences                                    | :none:`unclaimed`         | :part:`In Progress`       |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Clarifications to Fortran map semantics                     | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| default clause at target construct                          | :part:`In Progress`         | :none:`unclaimed`         |                                                                        |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| ref count update use_device_{ptr, addr}                     | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Clarifications to implicit reductions                       | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| ref modifier for map clauses                                | :part:`In Progress`       | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| map-type modifiers in arbitrary position                    | :good:`done`              | :none:`unclaimed`         | https://github.com/llvm/llvm-project/pull/90499                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Lift nesting restriction on concurrent loop                 | :good:`done`              | :none:`unclaimed`         | https://github.com/llvm/llvm-project/pull/125621                         |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| priority clause for target constructs                       | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| changes to target_data construct                            | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Non-const do_not_sync for nowait/nogroup                    | :none:`unclaimed`         | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| need_device_addr modifier for adjust_args clause            | :part:`partial`           | :none:`unclaimed`         | Parsing/Sema: https://github.com/llvm/llvm-project/pull/143442           |

+|                                                             |                           |                           |               https://github.com/llvm/llvm-project/pull/149586           |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Prescriptive num_threads                                    | :part:`In Progress`       | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Message and severity clauses                                | :part:`In Progress`       | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+| Local clause on declare target                              | :part:`In Progress`       | :none:`unclaimed`         |                                                                          |

++-------------------------------------------------------------+---------------------------+---------------------------+--------------------------------------------------------------------------+

+

+OpenMP Extensions

+=================

+

+The following table provides a quick overview over various OpenMP

+extensions and their implementation status.  These extensions are not

+currently defined by any standard, so links to associated LLVM

+documentation are provided.  As these extensions mature, they will be

+considered for standardization. Please post on the

+`Discourse forums (Runtimes - OpenMP category)`_ to provide feedback.

+

++------------------------------+-----------------------------------------------------------------------------------+--------------------------+--------------------------------------------------------+

+|Category                      | Feature                                                                           | Status                   | Reviews                                                |

++==============================+===================================================================================+==========================+========================================================+

+| atomic extension             | `'atomic' strictly nested within 'teams'                                          | :good:`prototyped`       | D126323                                                |

+|                              | <https://openmp.llvm.org/docs/openacc/OpenMPExtensions.html#atomicWithinTeams>`_  |                          |                                                        |

++------------------------------+-----------------------------------------------------------------------------------+--------------------------+--------------------------------------------------------+

+| device extension             | `'ompx_hold' map type modifier                                                    | :good:`prototyped`       | D106509, D106510                                       |

+|                              | <https://openmp.llvm.org/docs/openacc/OpenMPExtensions.html#ompx-hold>`_          |                          |                                                        |

++------------------------------+-----------------------------------------------------------------------------------+--------------------------+--------------------------------------------------------+

+| device extension             | `'ompx_bare' clause on 'target teams' construct                                   | :good:`prototyped`       | #66844, #70612                                         |

+|                              | <https://www.osti.gov/servlets/purl/2205717>`_                                    |                          |                                                        |

++------------------------------+-----------------------------------------------------------------------------------+--------------------------+--------------------------------------------------------+

+| device extension             | Multi-dim 'num_teams' and 'thread_limit' clause on 'target teams ompx_bare'       | :good:`partial`          | #99732, #101407, #102715                               |

+|                              | construct                                                                         |                          |                                                        |

++------------------------------+-----------------------------------------------------------------------------------+--------------------------+--------------------------------------------------------+

+

+.. _Discourse forums (Runtimes - OpenMP category): https://discourse.llvm.org/c/runtimes/openmp/35

diff --git a/clang/docs/ReleaseNotes.rst b/clang/docs/ReleaseNotes.rst
index ec51ffd..95a0bbf 100644
--- a/clang/docs/ReleaseNotes.rst
+++ b/clang/docs/ReleaseNotes.rst
@@ -89,9 +89,18 @@ Non-comprehensive list of changes in this release
 -------------------------------------------------
 - Added ``__builtin_elementwise_minnumnum`` and ``__builtin_elementwise_maxnumnum``.
 
+- Trapping UBSan (e.g. ``-fsanitize-trap=undefined``) now emits a string describing the reason for 
+  trapping into the generated debug info. This feature allows debuggers (e.g. LLDB) to display 
+  the reason for trapping if the trap is reached. The string is currently encoded in the debug 
+  info as an artificial frame that claims to be inlined at the trap location. The function used 
+  for the artificial frame is an artificial function whose name encodes the reason for trapping. 
+  The encoding used is currently the same as ``__builtin_verbose_trap`` but might change in the future. 
+  This feature is enabled by default but can be disabled by compiling with 
+  ``-fno-sanitize-annotate-debug-info-traps``.
 
 New Compiler Flags
 ------------------
+- New option ``-fno-sanitize-annotate-debug-info-traps`` added to disable emitting trap reasons into the debug info when compiling with trapping UBSan (e.g. ``-fsanitize-trap=undefined``).
 
 Deprecated Compiler Flags
 -------------------------
diff --git a/clang/include/clang/AST/ASTContext.h b/clang/include/clang/AST/ASTContext.h
index 17cbfb2..0273109 100644
--- a/clang/include/clang/AST/ASTContext.h
+++ b/clang/include/clang/AST/ASTContext.h
@@ -238,9 +238,9 @@ class ASTContext : public RefCountedBase<ASTContext> {
   mutable llvm::FoldingSet<ElaboratedType> ElaboratedTypes{
       GeneralTypesLog2InitSize};
   mutable llvm::FoldingSet<DependentNameType> DependentNameTypes;
-  mutable llvm::ContextualFoldingSet<DependentTemplateSpecializationType,
-                                     ASTContext&>
-    DependentTemplateSpecializationTypes;
+  mutable llvm::DenseMap<llvm::FoldingSetNodeID,
+                         DependentTemplateSpecializationType *>
+      DependentTemplateSpecializationTypes;
   mutable llvm::FoldingSet<PackExpansionType> PackExpansionTypes;
   mutable llvm::FoldingSet<ObjCObjectTypeImpl> ObjCObjectTypes;
   mutable llvm::FoldingSet<ObjCObjectPointerType> ObjCObjectPointerTypes;
diff --git a/clang/include/clang/AST/Type.h b/clang/include/clang/AST/Type.h
index 764e9d50..98810fb 100644
--- a/clang/include/clang/AST/Type.h
+++ b/clang/include/clang/AST/Type.h
@@ -7276,8 +7276,7 @@ public:
 /// Represents a template specialization type whose template cannot be
 /// resolved, e.g.
 ///   A<T>::template B<T>
-class DependentTemplateSpecializationType : public TypeWithKeyword,
-                                            public llvm::FoldingSetNode {
+class DependentTemplateSpecializationType : public TypeWithKeyword {
   friend class ASTContext; // ASTContext creates these
 
   DependentTemplateStorage Name;
diff --git a/clang/include/clang/Basic/BuiltinsWebAssembly.def b/clang/include/clang/Basic/BuiltinsWebAssembly.def
index e2afcc0..d31b726 100644
--- a/clang/include/clang/Basic/BuiltinsWebAssembly.def
+++ b/clang/include/clang/Basic/BuiltinsWebAssembly.def
@@ -199,6 +199,12 @@ TARGET_BUILTIN(__builtin_wasm_ref_is_null_extern, "ii", "nct", "reference-types"
 // return type.
 TARGET_BUILTIN(__builtin_wasm_ref_null_func, "i", "nct", "reference-types")
 
+// Check if the runtime type of a function pointer matches its static type. Used
+// to avoid "function signature mismatch" traps. Takes a function pointer, uses
+// table.get to look up the pointer in __indirect_function_table and then
+// ref.test to test the type.
+TARGET_BUILTIN(__builtin_wasm_test_function_pointer_signature, "i.", "nct", "gc")
+
 // Table builtins
 TARGET_BUILTIN(__builtin_wasm_table_set,  "viii", "t", "reference-types")
 TARGET_BUILTIN(__builtin_wasm_table_get,  "iii", "t", "reference-types")
diff --git a/clang/include/clang/Basic/CodeGenOptions.def b/clang/include/clang/Basic/CodeGenOptions.def
index e137738..423b696 100644
--- a/clang/include/clang/Basic/CodeGenOptions.def
+++ b/clang/include/clang/Basic/CodeGenOptions.def
@@ -307,6 +307,7 @@ CODEGENOPT(SanitizeBinaryMetadataAtomics, 1, 0, Benign) ///< Emit PCs for atomic
 CODEGENOPT(SanitizeBinaryMetadataUAR, 1, 0, Benign) ///< Emit PCs for start of functions
                                                     ///< that are subject for use-after-return checking.
 CODEGENOPT(SanitizeStats     , 1, 0, Benign) ///< Collect statistics for sanitizers.
+CODEGENOPT(SanitizeDebugTrapReasons, 1, 1 , Benign) ///< Enable UBSan trapping messages
 CODEGENOPT(SimplifyLibCalls  , 1, 1, Benign) ///< Set when -fbuiltin is enabled.
 CODEGENOPT(SoftFloat         , 1, 0, Benign) ///< -soft-float.
 CODEGENOPT(SpeculativeLoadHardening, 1, 0, Benign) ///< Enable speculative load hardening.
diff --git a/clang/include/clang/Basic/DiagnosticSemaKinds.td b/clang/include/clang/Basic/DiagnosticSemaKinds.td
index 4a21321..da4216d 100644
--- a/clang/include/clang/Basic/DiagnosticSemaKinds.td
+++ b/clang/include/clang/Basic/DiagnosticSemaKinds.td
@@ -7575,6 +7575,8 @@ def err_typecheck_illegal_increment_decrement : Error<
   "cannot %select{decrement|increment}1 value of type %0">;
 def err_typecheck_expect_int : Error<
   "used type %0 where integer is required">;
+def err_typecheck_expect_function_pointer
+    : Error<"used type %0 where function pointer is required">;
 def err_typecheck_expect_hlsl_resource : Error<
   "used type %0 where __hlsl_resource_t is required">;
 def err_typecheck_arithmetic_incomplete_or_sizeless_type : Error<
@@ -12389,6 +12391,13 @@ def err_invalid_module_name : Error<"%0 is an invalid name for a module">;
 def err_extern_def_in_header_unit : Error<
   "non-inline external definitions are not permitted in C++ header units">;
 
+def warn_exposure : Warning <
+  "TU local entity %0 is exposed">,
+  InGroup<DiagGroup<"TU-local-entity-exposure">>;
+def warn_reference_tu_local_entity_in_other_tu : Warning <
+  "instantiation of %0 triggers reference to TU-local entity %1 from other TU '%2'">,
+  InGroup<DiagGroup<"reference-tu-local-entity-in-other-tu">>;
+
 def warn_experimental_header_unit : Warning<
   "the implementation of header units is in an experimental phase">,
   InGroup<DiagGroup<"experimental-header-units">>;
@@ -13202,6 +13211,10 @@ def err_wasm_builtin_arg_must_match_table_element_type : Error <
   "%ordinal0 argument must match the element type of the WebAssembly table in the %ordinal1 argument">;
 def err_wasm_builtin_arg_must_be_integer_type : Error <
   "%ordinal0 argument must be an integer">;
+def err_wasm_builtin_test_fp_sig_cannot_include_reference_type
+    : Error<"not supported for "
+            "function pointers with a reference type %select{return "
+            "value|parameter}0">;
 
 // OpenACC diagnostics.
 def warn_acc_routine_unimplemented
@@ -13257,6 +13270,11 @@ def err_acc_not_a_var_ref
 def err_acc_not_a_var_ref_use_device_declare
     : Error<"OpenACC variable %select{in 'use_device' clause|on 'declare' "
             "construct}0 is not a valid variable name or array name">;
+def ext_acc_array_section_use_device_declare
+    : Extension<
+          "sub-array as a variable %select{in 'use_device' clause|on "
+          "'declare' construct}0 is not a valid variable name or array name">,
+      InGroup<DiagGroup<"openacc-extension">>;
 def err_acc_not_a_var_ref_cache
     : Error<"OpenACC variable in 'cache' directive is not a valid sub-array or "
             "array element">;
@@ -13318,8 +13336,9 @@ def err_acc_reduction_num_gangs_conflict
             "appear on a '%2' construct "
             "with a '%3' clause%select{ with more than 1 argument|}0">;
 def err_acc_reduction_type
-    : Error<"OpenACC 'reduction' variable must be of scalar type, sub-array, or a "
-            "composite of scalar types;%select{| sub-array base}1 type is %0">;
+    : Error<"OpenACC 'reduction' variable must be of scalar type, aggregate, "
+            "sub-array, or a composite of scalar types;%select{| sub-array "
+            "base}1 type is %0">;
 def err_acc_reduction_composite_type
     : Error<"OpenACC 'reduction' variable must be a composite of scalar types; "
             "%1 %select{is not a class or struct|is incomplete|is not an "
diff --git a/clang/include/clang/CIR/Dialect/IR/CIRAttrs.td b/clang/include/clang/CIR/Dialect/IR/CIRAttrs.td
index 29d8aea..9961544 100644
--- a/clang/include/clang/CIR/Dialect/IR/CIRAttrs.td
+++ b/clang/include/clang/CIR/Dialect/IR/CIRAttrs.td
@@ -147,6 +147,18 @@ def UndefAttr : CIR_TypedAttr<"Undef", "undef"> {
 }
 
 //===----------------------------------------------------------------------===//
+// PoisonAttr
+//===----------------------------------------------------------------------===//
+
+def CIR_PoisonAttr : CIR_TypedAttr<"Poison", "poison"> {
+  let summary = "Represent a typed poison constant";
+  let description = [{
+    The PoisonAttr represents a typed poison constant, corresponding to LLVM's
+    notion of poison.
+  }];
+}
+
+//===----------------------------------------------------------------------===//
 // IntegerAttr
 //===----------------------------------------------------------------------===//
 
diff --git a/clang/include/clang/CIR/Dialect/IR/CIROps.td b/clang/include/clang/CIR/Dialect/IR/CIROps.td
index 32bb900..e2ddbd1 100644
--- a/clang/include/clang/CIR/Dialect/IR/CIROps.td
+++ b/clang/include/clang/CIR/Dialect/IR/CIROps.td
@@ -2817,6 +2817,8 @@ class CIR_BitOpBase<string mnemonic, TypeConstraint operandTy>
   let assemblyFormat = [{
     $input `:` type($result) attr-dict
   }];
+
+  let hasFolder = 1;
 }
 
 class CIR_BitZeroCountOpBase<string mnemonic, TypeConstraint operandTy>
@@ -3024,6 +3026,8 @@ def CIR_RotateOp : CIR_Op<"rotate", [Pure, SameOperandsAndResultType]> {
     bool isRotateLeft() { return getRotateLeft(); }
     bool isRotateRight() { return !getRotateLeft(); }
   }];
+
+  let hasFolder = 1;
 }
 
 //===----------------------------------------------------------------------===//
diff --git a/clang/include/clang/Driver/Options.td b/clang/include/clang/Driver/Options.td
index 916400e..e2ab046 100644
--- a/clang/include/clang/Driver/Options.td
+++ b/clang/include/clang/Driver/Options.td
@@ -2597,6 +2597,16 @@ def fsanitize_undefined_trap_on_error
 def fno_sanitize_undefined_trap_on_error
     : Flag<["-"], "fno-sanitize-undefined-trap-on-error">, Group<f_clang_Group>,
       Alias<fno_sanitize_trap_EQ>, AliasArgs<["undefined"]>;
+defm sanitize_debug_trap_reasons
+    : BoolFOption<
+          "sanitize-debug-trap-reasons",
+          CodeGenOpts<"SanitizeDebugTrapReasons">, DefaultTrue,
+          PosFlag<SetTrue, [], [ClangOption, CC1Option],
+                  "Annotate trap blocks in debug info with UBSan trap reasons">,
+          NegFlag<SetFalse, [], [ClangOption, CC1Option],
+                  "Do not annotate trap blocks in debug info with UBSan trap "
+                  "reasons">>;
+
 defm sanitize_minimal_runtime : BoolOption<"f", "sanitize-minimal-runtime",
   CodeGenOpts<"SanitizeMinimalRuntime">, DefaultFalse,
   PosFlag<SetTrue>,
diff --git a/clang/include/clang/Format/Format.h b/clang/include/clang/Format/Format.h
index a54ab19..31582a4 100644
--- a/clang/include/clang/Format/Format.h
+++ b/clang/include/clang/Format/Format.h
@@ -831,7 +831,7 @@ struct FormatStyle {
     /// Never merge functions into a single line.
     SFS_None,
     /// Only merge functions defined inside a class. Same as ``inline``,
-    /// except it does not implies ``empty``: i.e. top level empty functions
+    /// except it does not imply ``empty``: i.e. top level empty functions
     /// are not merged either.
     /// \code
     ///   class Foo {
diff --git a/clang/include/clang/Sema/Sema.h b/clang/include/clang/Sema/Sema.h
index 73eb730..423dcf9 100644
--- a/clang/include/clang/Sema/Sema.h
+++ b/clang/include/clang/Sema/Sema.h
@@ -9945,6 +9945,20 @@ private:
 
   VisibleModuleSet VisibleModules;
 
+  /// Whether we had imported any named modules.
+  bool HadImportedNamedModules = false;
+  /// The set of instantiations we need to check if they references TU-local
+  /// entity from TUs. This only makes sense if we imported any named modules.
+  llvm::SmallVector<std::pair<FunctionDecl *, SourceLocation>>
+      PendingCheckReferenceForTULocal;
+  /// Implement [basic.link]p18, which requires that we can't use TU-local
+  /// entities from other TUs (ignoring header units).
+  void checkReferenceToTULocalFromOtherTU(FunctionDecl *FD,
+                                          SourceLocation PointOfInstantiation);
+  /// Implement [basic.link]p17, which diagnose for non TU local exposure in
+  /// module interface or module partition.
+  void checkExposure(const TranslationUnitDecl *TU);
+
   ///@}
 
   //
diff --git a/clang/include/clang/Sema/SemaWasm.h b/clang/include/clang/Sema/SemaWasm.h
index 2123e07..8c0639f 100644
--- a/clang/include/clang/Sema/SemaWasm.h
+++ b/clang/include/clang/Sema/SemaWasm.h
@@ -37,6 +37,7 @@ public:
   bool BuiltinWasmTableGrow(CallExpr *TheCall);
   bool BuiltinWasmTableFill(CallExpr *TheCall);
   bool BuiltinWasmTableCopy(CallExpr *TheCall);
+  bool BuiltinWasmTestFunctionPointerSignature(CallExpr *TheCall);
 
   WebAssemblyImportNameAttr *
   mergeImportNameAttr(Decl *D, const WebAssemblyImportNameAttr &AL);
diff --git a/clang/include/clang/StaticAnalyzer/Checkers/Checkers.td b/clang/include/clang/StaticAnalyzer/Checkers/Checkers.td
index 38584c9..6225977 100644
--- a/clang/include/clang/StaticAnalyzer/Checkers/Checkers.td
+++ b/clang/include/clang/StaticAnalyzer/Checkers/Checkers.td
@@ -206,21 +206,15 @@ def CallAndMessageChecker : Checker<"CallAndMessage">,
   Documentation<HasDocumentation>,
   Dependencies<[CallAndMessageModeling]>;
 
-def DereferenceModeling : Checker<"DereferenceModeling">,
-  HelpText<"General support for dereference related checkers">,
-  Documentation<NotDocumented>,
-  Hidden;
-
 def FixedAddressDereferenceChecker
     : Checker<"FixedAddressDereference">,
       HelpText<"Check for dereferences of fixed addresses">,
-      Documentation<HasDocumentation>,
-      Dependencies<[DereferenceModeling]>;
+      Documentation<HasDocumentation>;
 
-def NullDereferenceChecker : Checker<"NullDereference">,
-  HelpText<"Check for dereferences of null pointers">,
-  Documentation<HasDocumentation>,
-  Dependencies<[DereferenceModeling]>;
+def NullDereferenceChecker
+    : Checker<"NullDereference">,
+      HelpText<"Check for dereferences of null pointers">,
+      Documentation<HasDocumentation>;
 
 def NonNullParamChecker : Checker<"NonNullParamChecker">,
   HelpText<"Check for null pointers passed as arguments to a function whose "
diff --git a/clang/lib/AST/ASTContext.cpp b/clang/lib/AST/ASTContext.cpp
index 6b6275f..16cf114 100644
--- a/clang/lib/AST/ASTContext.cpp
+++ b/clang/lib/AST/ASTContext.cpp
@@ -940,7 +940,6 @@ ASTContext::ASTContext(LangOptions &LOpts, SourceManager &SM,
       FunctionProtoTypes(this_(), FunctionProtoTypesLog2InitSize),
       DependentTypeOfExprTypes(this_()), DependentDecltypeTypes(this_()),
       DependentPackIndexingTypes(this_()), TemplateSpecializationTypes(this_()),
-      DependentTemplateSpecializationTypes(this_()),
       DependentBitIntTypes(this_()), SubstTemplateTemplateParmPacks(this_()),
       DeducedTemplates(this_()), ArrayParameterTypes(this_()),
       CanonTemplateTemplateParms(this_()), SourceMgr(SM), LangOpts(LOpts),
@@ -5979,10 +5978,9 @@ QualType ASTContext::getDependentTemplateSpecializationType(
   llvm::FoldingSetNodeID ID;
   DependentTemplateSpecializationType::Profile(ID, *this, Keyword, Name, Args);
 
-  void *InsertPos = nullptr;
-  if (auto *T = DependentTemplateSpecializationTypes.FindNodeOrInsertPos(
-          ID, InsertPos))
-    return QualType(T, 0);
+  if (auto const T_iter = DependentTemplateSpecializationTypes.find(ID);
+      T_iter != DependentTemplateSpecializationTypes.end())
+    return QualType(T_iter->getSecond(), 0);
 
   NestedNameSpecifier *NNS = Name.getQualifier();
 
@@ -6001,11 +5999,6 @@ QualType ASTContext::getDependentTemplateSpecializationType(
           CanonKeyword, {CanonNNS, Name.getName(), /*HasTemplateKeyword=*/true},
           CanonArgs,
           /*IsCanonical=*/true);
-      // Find the insert position again.
-      [[maybe_unused]] auto *Nothing =
-          DependentTemplateSpecializationTypes.FindNodeOrInsertPos(ID,
-                                                                   InsertPos);
-      assert(!Nothing && "canonical type broken");
     }
   } else {
     assert(Keyword == getCanonicalElaboratedTypeKeyword(Keyword));
@@ -6021,8 +6014,13 @@ QualType ASTContext::getDependentTemplateSpecializationType(
                        alignof(DependentTemplateSpecializationType));
   auto *T =
       new (Mem) DependentTemplateSpecializationType(Keyword, Name, Args, Canon);
+#ifndef NDEBUG
+  llvm::FoldingSetNodeID InsertedID;
+  T->Profile(InsertedID, *this);
+  assert(InsertedID == ID && "ID does not match");
+#endif
   Types.push_back(T);
-  DependentTemplateSpecializationTypes.InsertNode(T, InsertPos);
+  DependentTemplateSpecializationTypes.try_emplace(ID, T);
   return QualType(T, 0);
 }
 
diff --git a/clang/lib/AST/OSLog.cpp b/clang/lib/AST/OSLog.cpp
index b777d4d..91f8410 100644
--- a/clang/lib/AST/OSLog.cpp
+++ b/clang/lib/AST/OSLog.cpp
@@ -1,4 +1,16 @@
-// TODO: header template
+//===--- OSLog.cpp - OS log format string analysis ------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+///
+/// \file
+/// This file implements analysis functions for OS log format strings and
+/// buffer layout computation for __builtin_os_log_format and related builtins.
+///
+//===----------------------------------------------------------------------===//
 
 #include "clang/AST/OSLog.h"
 #include "clang/AST/Attr.h"
@@ -137,8 +149,8 @@ public:
     for (auto &Data : ArgsData) {
       if (!Data.MaskType.empty()) {
         CharUnits Size = CharUnits::fromQuantity(8);
-        Layout.Items.emplace_back(OSLogBufferItem::MaskKind, nullptr,
-                                  Size, 0, Data.MaskType);
+        Layout.Items.emplace_back(OSLogBufferItem::MaskKind, nullptr, Size, 0,
+                                  Data.MaskType);
       }
 
       if (Data.FieldWidth) {
diff --git a/clang/lib/Basic/Targets/WebAssembly.cpp b/clang/lib/Basic/Targets/WebAssembly.cpp
index af25d25..e362350e 100644
--- a/clang/lib/Basic/Targets/WebAssembly.cpp
+++ b/clang/lib/Basic/Targets/WebAssembly.cpp
@@ -64,6 +64,7 @@ bool WebAssemblyTargetInfo::hasFeature(StringRef Feature) const {
       .Case("mutable-globals", HasMutableGlobals)
       .Case("nontrapping-fptoint", HasNontrappingFPToInt)
       .Case("reference-types", HasReferenceTypes)
+      .Case("gc", HasGC)
       .Case("relaxed-simd", SIMDLevel >= RelaxedSIMD)
       .Case("sign-ext", HasSignExt)
       .Case("simd128", SIMDLevel >= SIMD128)
@@ -106,6 +107,8 @@ void WebAssemblyTargetInfo::getTargetDefines(const LangOptions &Opts,
     Builder.defineMacro("__wasm_nontrapping_fptoint__");
   if (HasReferenceTypes)
     Builder.defineMacro("__wasm_reference_types__");
+  if (HasGC)
+    Builder.defineMacro("__wasm_gc__");
   if (SIMDLevel >= RelaxedSIMD)
     Builder.defineMacro("__wasm_relaxed_simd__");
   if (HasSignExt)
@@ -307,6 +310,14 @@ bool WebAssemblyTargetInfo::handleTargetFeatures(
       HasReferenceTypes = false;
       continue;
     }
+    if (Feature == "+gc") {
+      HasGC = true;
+      continue;
+    }
+    if (Feature == "-gc") {
+      HasGC = false;
+      continue;
+    }
     if (Feature == "+relaxed-simd") {
       SIMDLevel = std::max(SIMDLevel, RelaxedSIMD);
       continue;
@@ -353,6 +364,11 @@ bool WebAssemblyTargetInfo::handleTargetFeatures(
     return false;
   }
 
+  // gc implies reference-types
+  if (HasGC) {
+    HasReferenceTypes = true;
+  }
+
   // bulk-memory-opt is a subset of bulk-memory.
   if (HasBulkMemory) {
     HasBulkMemoryOpt = true;
diff --git a/clang/lib/Basic/Targets/WebAssembly.h b/clang/lib/Basic/Targets/WebAssembly.h
index 57b366c..c47c8cc 100644
--- a/clang/lib/Basic/Targets/WebAssembly.h
+++ b/clang/lib/Basic/Targets/WebAssembly.h
@@ -69,6 +69,7 @@ class LLVM_LIBRARY_VISIBILITY WebAssemblyTargetInfo : public TargetInfo {
   bool HasMutableGlobals = false;
   bool HasNontrappingFPToInt = false;
   bool HasReferenceTypes = false;
+  bool HasGC = false;
   bool HasSignExt = false;
   bool HasTailCall = false;
   bool HasWideArithmetic = false;
diff --git a/clang/lib/CIR/CodeGen/CIRGenExprAggregate.cpp b/clang/lib/CIR/CodeGen/CIRGenExprAggregate.cpp
index 0d12c5c..51aab95 100644
--- a/clang/lib/CIR/CodeGen/CIRGenExprAggregate.cpp
+++ b/clang/lib/CIR/CodeGen/CIRGenExprAggregate.cpp
@@ -357,10 +357,97 @@ void AggExprEmitter::visitCXXParenListOrInitListExpr(
     emitArrayInit(dest.getAddress(), arrayTy, e->getType(), e, args,
                   arrayFiller);
     return;
+  } else if (e->getType()->isVariableArrayType()) {
+    cgf.cgm.errorNYI(e->getSourceRange(),
+                     "visitCXXParenListOrInitListExpr variable array type");
+    return;
+  }
+
+  if (e->getType()->isArrayType()) {
+    cgf.cgm.errorNYI(e->getSourceRange(),
+                     "visitCXXParenListOrInitListExpr array type");
+    return;
+  }
+
+  assert(e->getType()->isRecordType() && "Only support structs/unions here!");
+
+  // Do struct initialization; this code just sets each individual member
+  // to the approprate value.  This makes bitfield support automatic;
+  // the disadvantage is that the generated code is more difficult for
+  // the optimizer, especially with bitfields.
+  unsigned numInitElements = args.size();
+  RecordDecl *record = e->getType()->castAs<RecordType>()->getDecl();
+
+  // We'll need to enter cleanup scopes in case any of the element
+  // initializers throws an exception.
+  assert(!cir::MissingFeatures::requiresCleanups());
+
+  unsigned curInitIndex = 0;
+
+  // Emit initialization of base classes.
+  if (auto *cxxrd = dyn_cast<CXXRecordDecl>(record)) {
+    assert(numInitElements >= cxxrd->getNumBases() &&
+           "missing initializer for base class");
+    if (cxxrd->getNumBases() > 0) {
+      cgf.cgm.errorNYI(e->getSourceRange(),
+                       "visitCXXParenListOrInitListExpr base class init");
+      return;
+    }
+  }
+
+  LValue destLV = cgf.makeAddrLValue(dest.getAddress(), e->getType());
+
+  if (record->isUnion()) {
+    cgf.cgm.errorNYI(e->getSourceRange(),
+                     "visitCXXParenListOrInitListExpr union type");
+    return;
   }
 
-  cgf.cgm.errorNYI(
-      "visitCXXParenListOrInitListExpr Record or VariableSizeArray type");
+  // Here we iterate over the fields; this makes it simpler to both
+  // default-initialize fields and skip over unnamed fields.
+  for (const FieldDecl *field : record->fields()) {
+    // We're done once we hit the flexible array member.
+    if (field->getType()->isIncompleteArrayType())
+      break;
+
+    // Always skip anonymous bitfields.
+    if (field->isUnnamedBitField())
+      continue;
+
+    // We're done if we reach the end of the explicit initializers, we
+    // have a zeroed object, and the rest of the fields are
+    // zero-initializable.
+    if (curInitIndex == numInitElements && dest.isZeroed() &&
+        cgf.getTypes().isZeroInitializable(e->getType()))
+      break;
+    LValue lv =
+        cgf.emitLValueForFieldInitialization(destLV, field, field->getName());
+    // We never generate write-barriers for initialized fields.
+    assert(!cir::MissingFeatures::setNonGC());
+
+    if (curInitIndex < numInitElements) {
+      // Store the initializer into the field.
+      CIRGenFunction::SourceLocRAIIObject loc{
+          cgf, cgf.getLoc(record->getSourceRange())};
+      emitInitializationToLValue(args[curInitIndex++], lv);
+    } else {
+      // We're out of initializers; default-initialize to null
+      emitNullInitializationToLValue(cgf.getLoc(e->getSourceRange()), lv);
+    }
+
+    // Push a destructor if necessary.
+    // FIXME: if we have an array of structures, all explicitly
+    // initialized, we can end up pushing a linear number of cleanups.
+    if (field->getType().isDestructedType()) {
+      cgf.cgm.errorNYI(e->getSourceRange(),
+                       "visitCXXParenListOrInitListExpr destructor");
+      return;
+    }
+
+    // From classic codegen, maybe not useful for CIR:
+    // If the GEP didn't get used because of a dead zero init or something
+    // else, clean it up for -O0 builds and general tidiness.
+  }
 }
 
 // TODO(cir): This could be shared with classic codegen.
diff --git a/clang/lib/CIR/CodeGen/CIRGenExprComplex.cpp b/clang/lib/CIR/CodeGen/CIRGenExprComplex.cpp
index 7f2e2ce..02685a3 100644
--- a/clang/lib/CIR/CodeGen/CIRGenExprComplex.cpp
+++ b/clang/lib/CIR/CodeGen/CIRGenExprComplex.cpp
@@ -60,7 +60,7 @@ public:
   mlir::Value VisitDeclRefExpr(DeclRefExpr *e);
   mlir::Value VisitGenericSelectionExpr(GenericSelectionExpr *e);
   mlir::Value VisitImplicitCastExpr(ImplicitCastExpr *e);
-  mlir::Value VisitInitListExpr(const InitListExpr *e);
+  mlir::Value VisitInitListExpr(InitListExpr *e);
 
   mlir::Value VisitCompoundLiteralExpr(CompoundLiteralExpr *e) {
     return emitLoadOfLValue(e);
@@ -452,7 +452,7 @@ mlir::Value ComplexExprEmitter::VisitImplicitCastExpr(ImplicitCastExpr *e) {
   return emitCast(e->getCastKind(), e->getSubExpr(), e->getType());
 }
 
-mlir::Value ComplexExprEmitter::VisitInitListExpr(const InitListExpr *e) {
+mlir::Value ComplexExprEmitter::VisitInitListExpr(InitListExpr *e) {
   mlir::Location loc = cgf.getLoc(e->getExprLoc());
   if (e->getNumInits() == 2) {
     mlir::Value real = cgf.emitScalarExpr(e->getInit(0));
@@ -460,10 +460,8 @@ mlir::Value ComplexExprEmitter::VisitInitListExpr(const InitListExpr *e) {
     return builder.createComplexCreate(loc, real, imag);
   }
 
-  if (e->getNumInits() == 1) {
-    cgf.cgm.errorNYI("Create Complex with InitList with size 1");
-    return {};
-  }
+  if (e->getNumInits() == 1)
+    return Visit(e->getInit(0));
 
   assert(e->getNumInits() == 0 && "Unexpected number of inits");
   mlir::Type complexTy = cgf.convertType(e->getType());
diff --git a/clang/lib/CIR/Dialect/IR/CIRDialect.cpp b/clang/lib/CIR/Dialect/IR/CIRDialect.cpp
index cd77166..2213c75 100644
--- a/clang/lib/CIR/Dialect/IR/CIRDialect.cpp
+++ b/clang/lib/CIR/Dialect/IR/CIRDialect.cpp
@@ -17,6 +17,7 @@
 
 #include "mlir/Interfaces/ControlFlowInterfaces.h"
 #include "mlir/Interfaces/FunctionImplementation.h"
+#include "mlir/Support/LLVM.h"
 
 #include "clang/CIR/Dialect/IR/CIROpsDialect.cpp.inc"
 #include "clang/CIR/Dialect/IR/CIROpsEnums.cpp.inc"
@@ -338,7 +339,7 @@ static LogicalResult checkConstantTypes(mlir::Operation *op, mlir::Type opType,
   }
 
   if (mlir::isa<cir::ConstArrayAttr, cir::ConstVectorAttr,
-                cir::ConstComplexAttr>(attrType))
+                cir::ConstComplexAttr, cir::PoisonAttr>(attrType))
     return success();
 
   assert(isa<TypedAttr>(attrType) && "What else could we be looking at here?");
@@ -628,6 +629,11 @@ static Value tryFoldCastChain(cir::CastOp op) {
 }
 
 OpFoldResult cir::CastOp::fold(FoldAdaptor adaptor) {
+  if (mlir::isa_and_present<cir::PoisonAttr>(adaptor.getSrc())) {
+    // Propagate poison value
+    return cir::PoisonAttr::get(getContext(), getType());
+  }
+
   if (getSrc().getType() == getType()) {
     switch (getKind()) {
     case cir::CastKind::integral: {
@@ -1782,6 +1788,12 @@ static bool isBoolNot(cir::UnaryOp op) {
 //
 // and the argument of the first one (%0) will be used instead.
 OpFoldResult cir::UnaryOp::fold(FoldAdaptor adaptor) {
+  if (auto poison =
+          mlir::dyn_cast_if_present<cir::PoisonAttr>(adaptor.getInput())) {
+    // Propagate poison values
+    return poison;
+  }
+
   if (isBoolNot(*this))
     if (auto previous = dyn_cast_or_null<UnaryOp>(getInput().getDefiningOp()))
       if (isBoolNot(previous))
@@ -2231,6 +2243,134 @@ LogicalResult cir::ComplexImagPtrOp::verify() {
 }
 
 //===----------------------------------------------------------------------===//
+// Bit manipulation operations
+//===----------------------------------------------------------------------===//
+
+static OpFoldResult
+foldUnaryBitOp(mlir::Attribute inputAttr,
+               llvm::function_ref<llvm::APInt(const llvm::APInt &)> func,
+               bool poisonZero = false) {
+  if (mlir::isa_and_present<cir::PoisonAttr>(inputAttr)) {
+    // Propagate poison value
+    return inputAttr;
+  }
+
+  auto input = mlir::dyn_cast_if_present<IntAttr>(inputAttr);
+  if (!input)
+    return nullptr;
+
+  llvm::APInt inputValue = input.getValue();
+  if (poisonZero && inputValue.isZero())
+    return cir::PoisonAttr::get(input.getType());
+
+  llvm::APInt resultValue = func(inputValue);
+  return IntAttr::get(input.getType(), resultValue);
+}
+
+OpFoldResult BitClrsbOp::fold(FoldAdaptor adaptor) {
+  return foldUnaryBitOp(adaptor.getInput(), [](const llvm::APInt &inputValue) {
+    unsigned resultValue =
+        inputValue.getBitWidth() - inputValue.getSignificantBits();
+    return llvm::APInt(inputValue.getBitWidth(), resultValue);
+  });
+}
+
+OpFoldResult BitClzOp::fold(FoldAdaptor adaptor) {
+  return foldUnaryBitOp(
+      adaptor.getInput(),
+      [](const llvm::APInt &inputValue) {
+        unsigned resultValue = inputValue.countLeadingZeros();
+        return llvm::APInt(inputValue.getBitWidth(), resultValue);
+      },
+      getPoisonZero());
+}
+
+OpFoldResult BitCtzOp::fold(FoldAdaptor adaptor) {
+  return foldUnaryBitOp(
+      adaptor.getInput(),
+      [](const llvm::APInt &inputValue) {
+        return llvm::APInt(inputValue.getBitWidth(),
+                           inputValue.countTrailingZeros());
+      },
+      getPoisonZero());
+}
+
+OpFoldResult BitParityOp::fold(FoldAdaptor adaptor) {
+  return foldUnaryBitOp(adaptor.getInput(), [](const llvm::APInt &inputValue) {
+    return llvm::APInt(inputValue.getBitWidth(), inputValue.popcount() % 2);
+  });
+}
+
+OpFoldResult BitPopcountOp::fold(FoldAdaptor adaptor) {
+  return foldUnaryBitOp(adaptor.getInput(), [](const llvm::APInt &inputValue) {
+    return llvm::APInt(inputValue.getBitWidth(), inputValue.popcount());
+  });
+}
+
+OpFoldResult BitReverseOp::fold(FoldAdaptor adaptor) {
+  return foldUnaryBitOp(adaptor.getInput(), [](const llvm::APInt &inputValue) {
+    return inputValue.reverseBits();
+  });
+}
+
+OpFoldResult ByteSwapOp::fold(FoldAdaptor adaptor) {
+  return foldUnaryBitOp(adaptor.getInput(), [](const llvm::APInt &inputValue) {
+    return inputValue.byteSwap();
+  });
+}
+
+OpFoldResult RotateOp::fold(FoldAdaptor adaptor) {
+  if (mlir::isa_and_present<cir::PoisonAttr>(adaptor.getInput()) ||
+      mlir::isa_and_present<cir::PoisonAttr>(adaptor.getAmount())) {
+    // Propagate poison values
+    return cir::PoisonAttr::get(getType());
+  }
+
+  auto input = mlir::dyn_cast_if_present<IntAttr>(adaptor.getInput());
+  auto amount = mlir::dyn_cast_if_present<IntAttr>(adaptor.getAmount());
+  if (!input && !amount)
+    return nullptr;
+
+  // We could fold cir.rotate even if one of its two operands is not a constant:
+  //   - `cir.rotate left/right %0, 0` could be folded into just %0 even if %0
+  //     is not a constant.
+  //   - `cir.rotate left/right 0/0b111...111, %0` could be folded into 0 or
+  //     0b111...111 even if %0 is not a constant.
+
+  llvm::APInt inputValue;
+  if (input) {
+    inputValue = input.getValue();
+    if (inputValue.isZero() || inputValue.isAllOnes()) {
+      // An input value of all 0s or all 1s will not change after rotation
+      return input;
+    }
+  }
+
+  uint64_t amountValue;
+  if (amount) {
+    amountValue = amount.getValue().urem(getInput().getType().getWidth());
+    if (amountValue == 0) {
+      // A shift amount of 0 will not change the input value
+      return getInput();
+    }
+  }
+
+  if (!input || !amount)
+    return nullptr;
+
+  assert(inputValue.getBitWidth() == getInput().getType().getWidth() &&
+         "input value must have the same bit width as the input type");
+
+  llvm::APInt resultValue;
+  if (isRotateLeft())
+    resultValue = inputValue.rotl(amountValue);
+  else
+    resultValue = inputValue.rotr(amountValue);
+
+  return IntAttr::get(input.getContext(), input.getType(), resultValue);
+}
+
+//===----------------------------------------------------------------------===//
 // TableGen'd op method definitions
 //===----------------------------------------------------------------------===//
 
diff --git a/clang/lib/CIR/Dialect/Transforms/CIRCanonicalize.cpp b/clang/lib/CIR/Dialect/Transforms/CIRCanonicalize.cpp
index e505db5..2143f16 100644
--- a/clang/lib/CIR/Dialect/Transforms/CIRCanonicalize.cpp
+++ b/clang/lib/CIR/Dialect/Transforms/CIRCanonicalize.cpp
@@ -143,7 +143,8 @@ void CIRCanonicalizePass::runOnOperation() {
     if (isa<BrOp, BrCondOp, CastOp, ScopeOp, SwitchOp, SelectOp, UnaryOp,
             ComplexCreateOp, ComplexImagOp, ComplexRealOp, VecCmpOp,
             VecCreateOp, VecExtractOp, VecShuffleOp, VecShuffleDynamicOp,
-            VecTernaryOp>(op))
+            VecTernaryOp, BitClrsbOp, BitClzOp, BitCtzOp, BitParityOp,
+            BitPopcountOp, BitReverseOp, ByteSwapOp, RotateOp>(op))
       ops.push_back(op);
   });
 
diff --git a/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp b/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp
index 3cd7de0..c27b889 100644
--- a/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp
+++ b/clang/lib/CIR/Lowering/DirectToLLVM/LowerToLLVM.cpp
@@ -1027,6 +1027,12 @@ mlir::LogicalResult CIRToLLVMConstantOpLowering::matchAndRewrite(
     mlir::ConversionPatternRewriter &rewriter) const {
   mlir::Attribute attr = op.getValue();
 
+  if (mlir::isa<cir::PoisonAttr>(attr)) {
+    rewriter.replaceOpWithNewOp<mlir::LLVM::PoisonOp>(
+        op, getTypeConverter()->convertType(op.getType()));
+    return mlir::success();
+  }
+
   if (mlir::isa<mlir::IntegerType>(op.getType())) {
     // Verified cir.const operations cannot actually be of these types, but the
     // lowering pass may generate temporary cir.const operations with these
diff --git a/clang/lib/CodeGen/BackendUtil.cpp b/clang/lib/CodeGen/BackendUtil.cpp
index 1b72578..0b8b824 100644
--- a/clang/lib/CodeGen/BackendUtil.cpp
+++ b/clang/lib/CodeGen/BackendUtil.cpp
@@ -1027,12 +1027,6 @@ void EmitAssemblyHelper::RunOptimizationPipeline(
               MPM.addPass(
                   createModuleToFunctionPassAdaptor(ObjCARCExpandPass()));
           });
-      PB.registerPipelineEarlySimplificationEPCallback(
-          [](ModulePassManager &MPM, OptimizationLevel Level,
-             ThinOrFullLTOPhase) {
-            if (Level != OptimizationLevel::O0)
-              MPM.addPass(ObjCARCAPElimPass());
-          });
       PB.registerScalarOptimizerLateEPCallback(
           [](FunctionPassManager &FPM, OptimizationLevel Level) {
             if (Level != OptimizationLevel::O0)
diff --git a/clang/lib/CodeGen/CGDebugInfo.cpp b/clang/lib/CodeGen/CGDebugInfo.cpp
index a371b67..77fc3a2 100644
--- a/clang/lib/CodeGen/CGDebugInfo.cpp
+++ b/clang/lib/CodeGen/CGDebugInfo.cpp
@@ -6435,7 +6435,7 @@ CodeGenFunction::LexicalScope::~LexicalScope() {
 static std::string SanitizerHandlerToCheckLabel(SanitizerHandler Handler) {
   std::string Label;
   switch (Handler) {
-#define SANITIZER_CHECK(Enum, Name, Version)                                   \
+#define SANITIZER_CHECK(Enum, Name, Version, Msg)                              \
   case Enum:                                                                   \
     Label = "__ubsan_check_" #Name;                                            \
     break;
diff --git a/clang/lib/CodeGen/CGExpr.cpp b/clang/lib/CodeGen/CGExpr.cpp
index 85c7688..90aed79 100644
--- a/clang/lib/CodeGen/CGExpr.cpp
+++ b/clang/lib/CodeGen/CGExpr.cpp
@@ -85,6 +85,16 @@ enum VariableTypeDescriptorKind : uint16_t {
 //                        Miscellaneous Helper Methods
 //===--------------------------------------------------------------------===//
 
+static llvm::StringRef GetUBSanTrapForHandler(SanitizerHandler ID) {
+  switch (ID) {
+#define SANITIZER_CHECK(Enum, Name, Version, Msg)                              \
+  case SanitizerHandler::Enum:                                                 \
+    return Msg;
+    LIST_SANITIZER_CHECKS
+#undef SANITIZER_CHECK
+  }
+}
+
 /// CreateTempAlloca - This creates a alloca and inserts it into the entry
 /// block.
 RawAddress
@@ -3649,7 +3659,7 @@ struct SanitizerHandlerInfo {
 }
 
 const SanitizerHandlerInfo SanitizerHandlers[] = {
-#define SANITIZER_CHECK(Enum, Name, Version) {#Name, Version},
+#define SANITIZER_CHECK(Enum, Name, Version, Msg) {#Name, Version},
     LIST_SANITIZER_CHECKS
 #undef SANITIZER_CHECK
 };
@@ -3954,6 +3964,8 @@ void CodeGenFunction::EmitCfiCheckFail() {
   StartFunction(GlobalDecl(), CGM.getContext().VoidTy, F, FI, Args,
                 SourceLocation());
 
+  ApplyDebugLocation ADL = ApplyDebugLocation::CreateArtificial(*this);
+
   // This function is not affected by NoSanitizeList. This function does
   // not have a source location, but "src:*" would still apply. Revert any
   // changes to SanOpts made in StartFunction.
@@ -4051,6 +4063,15 @@ void CodeGenFunction::EmitTrapCheck(llvm::Value *Checked,
 
   llvm::BasicBlock *&TrapBB = TrapBBs[CheckHandlerID];
 
+  llvm::DILocation *TrapLocation = Builder.getCurrentDebugLocation();
+  llvm::StringRef TrapMessage = GetUBSanTrapForHandler(CheckHandlerID);
+
+  if (getDebugInfo() && !TrapMessage.empty() &&
+      CGM.getCodeGenOpts().SanitizeDebugTrapReasons && TrapLocation) {
+    TrapLocation = getDebugInfo()->CreateTrapFailureMessageFor(
+        TrapLocation, "Undefined Behavior Sanitizer", TrapMessage);
+  }
+
   NoMerge = NoMerge || !CGM.getCodeGenOpts().OptimizationLevel ||
             (CurCodeDecl && CurCodeDecl->hasAttr<OptimizeNoneAttr>());
 
@@ -4059,8 +4080,8 @@ void CodeGenFunction::EmitTrapCheck(llvm::Value *Checked,
     auto Call = TrapBB->begin();
     assert(isa<llvm::CallInst>(Call) && "Expected call in trap BB");
 
-    Call->applyMergedLocation(Call->getDebugLoc(),
-                              Builder.getCurrentDebugLocation());
+    Call->applyMergedLocation(Call->getDebugLoc(), TrapLocation);
+
     Builder.CreateCondBr(Checked, Cont, TrapBB,
                          MDHelper.createLikelyBranchWeights());
   } else {
@@ -4069,6 +4090,8 @@ void CodeGenFunction::EmitTrapCheck(llvm::Value *Checked,
                          MDHelper.createLikelyBranchWeights());
     EmitBlock(TrapBB);
 
+    ApplyDebugLocation applyTrapDI(*this, TrapLocation);
+
     llvm::CallInst *TrapCall =
         Builder.CreateCall(CGM.getIntrinsic(llvm::Intrinsic::ubsantrap),
                            llvm::ConstantInt::get(CGM.Int8Ty, CheckHandlerID));
diff --git a/clang/lib/CodeGen/SanitizerHandler.h b/clang/lib/CodeGen/SanitizerHandler.h
index bb42e39..a66e7ab 100644
--- a/clang/lib/CodeGen/SanitizerHandler.h
+++ b/clang/lib/CodeGen/SanitizerHandler.h
@@ -14,35 +14,69 @@
 #define LLVM_CLANG_LIB_CODEGEN_SANITIZER_HANDLER_H
 
 #define LIST_SANITIZER_CHECKS                                                  \
-  SANITIZER_CHECK(AddOverflow, add_overflow, 0)                                \
-  SANITIZER_CHECK(BuiltinUnreachable, builtin_unreachable, 0)                  \
-  SANITIZER_CHECK(CFICheckFail, cfi_check_fail, 0)                             \
-  SANITIZER_CHECK(DivremOverflow, divrem_overflow, 0)                          \
-  SANITIZER_CHECK(DynamicTypeCacheMiss, dynamic_type_cache_miss, 0)            \
-  SANITIZER_CHECK(FloatCastOverflow, float_cast_overflow, 0)                   \
-  SANITIZER_CHECK(FunctionTypeMismatch, function_type_mismatch, 0)             \
-  SANITIZER_CHECK(ImplicitConversion, implicit_conversion, 0)                  \
-  SANITIZER_CHECK(InvalidBuiltin, invalid_builtin, 0)                          \
-  SANITIZER_CHECK(InvalidObjCCast, invalid_objc_cast, 0)                       \
-  SANITIZER_CHECK(LoadInvalidValue, load_invalid_value, 0)                     \
-  SANITIZER_CHECK(MissingReturn, missing_return, 0)                            \
-  SANITIZER_CHECK(MulOverflow, mul_overflow, 0)                                \
-  SANITIZER_CHECK(NegateOverflow, negate_overflow, 0)                          \
-  SANITIZER_CHECK(NullabilityArg, nullability_arg, 0)                          \
-  SANITIZER_CHECK(NullabilityReturn, nullability_return, 1)                    \
-  SANITIZER_CHECK(NonnullArg, nonnull_arg, 0)                                  \
-  SANITIZER_CHECK(NonnullReturn, nonnull_return, 1)                            \
-  SANITIZER_CHECK(OutOfBounds, out_of_bounds, 0)                               \
-  SANITIZER_CHECK(PointerOverflow, pointer_overflow, 0)                        \
-  SANITIZER_CHECK(ShiftOutOfBounds, shift_out_of_bounds, 0)                    \
-  SANITIZER_CHECK(SubOverflow, sub_overflow, 0)                                \
-  SANITIZER_CHECK(TypeMismatch, type_mismatch, 1)                              \
-  SANITIZER_CHECK(AlignmentAssumption, alignment_assumption, 0)                \
-  SANITIZER_CHECK(VLABoundNotPositive, vla_bound_not_positive, 0)              \
-  SANITIZER_CHECK(BoundsSafety, bounds_safety, 0)
+  SANITIZER_CHECK(AddOverflow, add_overflow, 0, "Integer addition overflowed") \
+  SANITIZER_CHECK(BuiltinUnreachable, builtin_unreachable, 0,                  \
+                  "_builtin_unreachable(), execution reached an unreachable "  \
+                  "program point")                                             \
+  SANITIZER_CHECK(CFICheckFail, cfi_check_fail, 0,                             \
+                  "Control flow integrity check failed")                       \
+  SANITIZER_CHECK(DivremOverflow, divrem_overflow, 0,                          \
+                  "Integer divide or remainder overflowed")                    \
+  SANITIZER_CHECK(DynamicTypeCacheMiss, dynamic_type_cache_miss, 0,            \
+                  "Dynamic type cache miss, member call made on an object "    \
+                  "whose dynamic type differs from the expected type")         \
+  SANITIZER_CHECK(FloatCastOverflow, float_cast_overflow, 0,                   \
+                  "Floating-point to integer conversion overflowed")           \
+  SANITIZER_CHECK(FunctionTypeMismatch, function_type_mismatch, 0,             \
+                  "Function called with mismatched signature")                 \
+  SANITIZER_CHECK(ImplicitConversion, implicit_conversion, 0,                  \
+                  "Implicit integer conversion overflowed or lost data")       \
+  SANITIZER_CHECK(InvalidBuiltin, invalid_builtin, 0,                          \
+                  "Invalid use of builtin function")                           \
+  SANITIZER_CHECK(InvalidObjCCast, invalid_objc_cast, 0,                       \
+                  "Invalid Objective-C cast")                                  \
+  SANITIZER_CHECK(LoadInvalidValue, load_invalid_value, 0,                     \
+                  "Loaded an invalid or uninitialized value for the type")     \
+  SANITIZER_CHECK(MissingReturn, missing_return, 0,                            \
+                  "Execution reached the end of a value-returning function "   \
+                  "without returning a value")                                 \
+  SANITIZER_CHECK(MulOverflow, mul_overflow, 0,                                \
+                  "Integer multiplication overflowed")                         \
+  SANITIZER_CHECK(NegateOverflow, negate_overflow, 0,                          \
+                  "Integer negation overflowed")                               \
+  SANITIZER_CHECK(                                                             \
+      NullabilityArg, nullability_arg, 0,                                      \
+      "Passing null as an argument which is annotated with _Nonnull")          \
+  SANITIZER_CHECK(NullabilityReturn, nullability_return, 1,                    \
+                  "Returning null from a function with a return type "         \
+                  "annotated with _Nonnull")                                   \
+  SANITIZER_CHECK(NonnullArg, nonnull_arg, 0,                                  \
+                  "Passing null pointer as an argument which is declared to "  \
+                  "never be null")                                             \
+  SANITIZER_CHECK(NonnullReturn, nonnull_return, 1,                            \
+                  "Returning null pointer from a function which is declared "  \
+                  "to never return null")                                      \
+  SANITIZER_CHECK(OutOfBounds, out_of_bounds, 0, "Array index out of bounds")  \
+  SANITIZER_CHECK(PointerOverflow, pointer_overflow, 0,                        \
+                  "Pointer arithmetic overflowed bounds")                      \
+  SANITIZER_CHECK(ShiftOutOfBounds, shift_out_of_bounds, 0,                    \
+                  "Shift exponent is too large for the type")                  \
+  SANITIZER_CHECK(SubOverflow, sub_overflow, 0,                                \
+                  "Integer subtraction overflowed")                            \
+  SANITIZER_CHECK(TypeMismatch, type_mismatch, 1,                              \
+                  "Type mismatch in operation")                                \
+  SANITIZER_CHECK(AlignmentAssumption, alignment_assumption, 0,                \
+                  "Alignment assumption violated")                             \
+  SANITIZER_CHECK(                                                             \
+      VLABoundNotPositive, vla_bound_not_positive, 0,                          \
+      "Variable length array bound evaluates to non-positive value")           \
+  SANITIZER_CHECK(BoundsSafety, bounds_safety, 0,                              \
+                  "") // BoundsSafety Msg is empty because it is not considered
+                      // part of UBSan; therefore, no trap reason is emitted for
+                      // this case.
 
 enum SanitizerHandler {
-#define SANITIZER_CHECK(Enum, Name, Version) Enum,
+#define SANITIZER_CHECK(Enum, Name, Version, Msg) Enum,
   LIST_SANITIZER_CHECKS
 #undef SANITIZER_CHECK
 };
diff --git a/clang/lib/CodeGen/TargetBuiltins/WebAssembly.cpp b/clang/lib/CodeGen/TargetBuiltins/WebAssembly.cpp
index b7fd70e..33a8d8f 100644
--- a/clang/lib/CodeGen/TargetBuiltins/WebAssembly.cpp
+++ b/clang/lib/CodeGen/TargetBuiltins/WebAssembly.cpp
@@ -12,7 +12,10 @@
 
 #include "CGBuiltin.h"
 #include "clang/Basic/TargetBuiltins.h"
+#include "llvm/ADT/APInt.h"
+#include "llvm/IR/Constants.h"
 #include "llvm/IR/IntrinsicsWebAssembly.h"
+#include "llvm/Support/ErrorHandling.h"
 
 using namespace clang;
 using namespace CodeGen;
@@ -218,6 +221,64 @@ Value *CodeGenFunction::EmitWebAssemblyBuiltinExpr(unsigned BuiltinID,
     Function *Callee = CGM.getIntrinsic(Intrinsic::wasm_ref_null_func);
     return Builder.CreateCall(Callee);
   }
+  case WebAssembly::BI__builtin_wasm_test_function_pointer_signature: {
+    Value *FuncRef = EmitScalarExpr(E->getArg(0));
+
+    // Get the function type from the argument's static type
+    QualType ArgType = E->getArg(0)->getType();
+    const PointerType *PtrTy = ArgType->getAs<PointerType>();
+    assert(PtrTy && "Sema should have ensured this is a function pointer");
+
+    const FunctionType *FuncTy = PtrTy->getPointeeType()->getAs<FunctionType>();
+    assert(FuncTy && "Sema should have ensured this is a function pointer");
+
+    // In the llvm IR, we won't have access any more to the type of the function
+    // pointer so we need to insert this type information somehow. The
+    // @llvm.wasm.ref.test.func takes varargs arguments whose values are unused
+    // to indicate the type of the function to test for. See the test here:
+    // llvm/test/CodeGen/WebAssembly/ref-test-func.ll
+    //
+    // The format is: first we include the return types (since this is a C
+    // function pointer, there will be 0 or one of these) then a token type to
+    // indicate the boundary between return types and param types, then the
+    // param types.
+
+    llvm::FunctionType *LLVMFuncTy =
+        cast<llvm::FunctionType>(ConvertType(QualType(FuncTy, 0)));
+
+    unsigned NParams = LLVMFuncTy->getNumParams();
+    std::vector<Value *> Args;
+    Args.reserve(NParams + 3);
+    // The only real argument is the FuncRef
+    Args.push_back(FuncRef);
+
+    // Add the type information
+    auto addType = [this, &Args](llvm::Type *T) {
+      if (T->isVoidTy()) {
+        // Do nothing
+      } else if (T->isFloatingPointTy()) {
+        Args.push_back(ConstantFP::get(T, 0));
+      } else if (T->isIntegerTy()) {
+        Args.push_back(ConstantInt::get(T, 0));
+      } else if (T->isPointerTy()) {
+        Args.push_back(ConstantPointerNull::get(llvm::PointerType::get(
+            getLLVMContext(), T->getPointerAddressSpace())));
+      } else {
+        // TODO: Handle reference types. For now, we reject them in Sema.
+        llvm_unreachable("Unhandled type");
+      }
+    };
+
+    addType(LLVMFuncTy->getReturnType());
+    // The token type indicates the boundary between return types and param
+    // types.
+    Args.push_back(PoisonValue::get(llvm::Type::getTokenTy(getLLVMContext())));
+    for (unsigned i = 0; i < NParams; i++) {
+      addType(LLVMFuncTy->getParamType(i));
+    }
+    Function *Callee = CGM.getIntrinsic(Intrinsic::wasm_ref_test_func);
+    return Builder.CreateCall(Callee, Args);
+  }
   case WebAssembly::BI__builtin_wasm_swizzle_i8x16: {
     Value *Src = EmitScalarExpr(E->getArg(0));
     Value *Indices = EmitScalarExpr(E->getArg(1));
diff --git a/clang/lib/Driver/SanitizerArgs.cpp b/clang/lib/Driver/SanitizerArgs.cpp
index 21e4cff..98793a5 100644
--- a/clang/lib/Driver/SanitizerArgs.cpp
+++ b/clang/lib/Driver/SanitizerArgs.cpp
@@ -1382,6 +1382,12 @@ void SanitizerArgs::addArgs(const ToolChain &TC, const llvm::opt::ArgList &Args,
     CmdArgs.push_back(Args.MakeArgString("-fsanitize-annotate-debug-info=" +
                                          toString(AnnotateDebugInfo)));
 
+  if (const Arg *A =
+          Args.getLastArg(options::OPT_fsanitize_debug_trap_reasons,
+                          options::OPT_fno_sanitize_debug_trap_reasons)) {
+    CmdArgs.push_back(Args.MakeArgString(A->getAsString(Args)));
+  }
+
   addSpecialCaseListOpt(Args, CmdArgs,
                         "-fsanitize-ignorelist=", UserIgnorelistFiles);
   addSpecialCaseListOpt(Args, CmdArgs,
diff --git a/clang/lib/Driver/ToolChains/Clang.cpp b/clang/lib/Driver/ToolChains/Clang.cpp
index 7d0c142..9d882db 100644
--- a/clang/lib/Driver/ToolChains/Clang.cpp
+++ b/clang/lib/Driver/ToolChains/Clang.cpp
@@ -3881,17 +3881,17 @@ static bool RenderModulesOptions(Compilation &C, const Driver &D,
                                  const ArgList &Args, const InputInfo &Input,
                                  const InputInfo &Output, bool HaveStd20,
                                  ArgStringList &CmdArgs) {
-  bool IsCXX = types::isCXX(Input.getType());
-  bool HaveStdCXXModules = IsCXX && HaveStd20;
+  const bool IsCXX = types::isCXX(Input.getType());
+  const bool HaveStdCXXModules = IsCXX && HaveStd20;
   bool HaveModules = HaveStdCXXModules;
 
   // -fmodules enables the use of precompiled modules (off by default).
   // Users can pass -fno-cxx-modules to turn off modules support for
   // C++/Objective-C++ programs.
+  const bool AllowedInCXX = Args.hasFlag(options::OPT_fcxx_modules,
+                                         options::OPT_fno_cxx_modules, true);
   bool HaveClangModules = false;
   if (Args.hasFlag(options::OPT_fmodules, options::OPT_fno_modules, false)) {
-    bool AllowedInCXX = Args.hasFlag(options::OPT_fcxx_modules,
-                                     options::OPT_fno_cxx_modules, true);
     if (AllowedInCXX || !IsCXX) {
       CmdArgs.push_back("-fmodules");
       HaveClangModules = true;
@@ -3900,6 +3900,9 @@ static bool RenderModulesOptions(Compilation &C, const Driver &D,
 
   HaveModules |= HaveClangModules;
 
+  if (HaveModules && !AllowedInCXX)
+    CmdArgs.push_back("-fno-cxx-modules");
+
   // -fmodule-maps enables implicit reading of module map files. By default,
   // this is enabled if we are using Clang's flavor of precompiled modules.
   if (Args.hasFlag(options::OPT_fimplicit_module_maps,
diff --git a/clang/lib/Format/Format.cpp b/clang/lib/Format/Format.cpp
index 1cfa3d1..0637807 100644
--- a/clang/lib/Format/Format.cpp
+++ b/clang/lib/Format/Format.cpp
@@ -1754,7 +1754,6 @@ FormatStyle getGoogleStyle(FormatStyle::LanguageKind Language) {
   GoogleStyle.AttributeMacros.push_back("absl_nullable");
   GoogleStyle.AttributeMacros.push_back("absl_nullability_unknown");
   GoogleStyle.BreakTemplateDeclarations = FormatStyle::BTDS_Yes;
-  GoogleStyle.DerivePointerAlignment = true;
   GoogleStyle.IncludeStyle.IncludeBlocks = tooling::IncludeStyle::IBS_Regroup;
   GoogleStyle.IncludeStyle.IncludeCategories = {{"^<ext/.*\\.h>", 2, 0, false},
                                                 {"^<.*\\.h>", 1, 0, false},
@@ -1863,6 +1862,7 @@ FormatStyle getGoogleStyle(FormatStyle::LanguageKind Language) {
   } else if (Language == FormatStyle::LK_ObjC) {
     GoogleStyle.AlwaysBreakBeforeMultilineStrings = false;
     GoogleStyle.ColumnLimit = 100;
+    GoogleStyle.DerivePointerAlignment = true;
     // "Regroup" doesn't work well for ObjC yet (main header heuristic,
     // relationship between ObjC standard library headers and other heades,
     // #imports, etc.)
@@ -2639,32 +2639,44 @@ private:
 
   int countVariableAlignments(const SmallVectorImpl<AnnotatedLine *> &Lines) {
     int AlignmentDiff = 0;
+
     for (const AnnotatedLine *Line : Lines) {
       AlignmentDiff += countVariableAlignments(Line->Children);
-      for (FormatToken *Tok = Line->First; Tok && Tok->Next; Tok = Tok->Next) {
+
+      for (const auto *Tok = Line->getFirstNonComment(); Tok; Tok = Tok->Next) {
         if (Tok->isNot(TT_PointerOrReference))
           continue;
-        // Don't treat space in `void foo() &&` or `void() &&` as evidence.
-        if (const auto *Prev = Tok->getPreviousNonComment()) {
-          if (Prev->is(tok::r_paren) && Prev->MatchingParen) {
-            if (const auto *Func =
-                    Prev->MatchingParen->getPreviousNonComment()) {
-              if (Func->isOneOf(TT_FunctionDeclarationName, TT_StartOfName,
-                                TT_OverloadedOperator) ||
-                  Func->isTypeName(LangOpts)) {
-                continue;
-              }
-            }
-          }
+
+        const auto *Prev = Tok->Previous;
+        const bool PrecededByName = Prev && Prev->Tok.getIdentifierInfo();
+        const bool SpaceBefore = Tok->hasWhitespaceBefore();
+
+        // e.g. `int **`, `int*&`, etc.
+        while (Tok->Next && Tok->Next->is(TT_PointerOrReference))
+          Tok = Tok->Next;
+
+        const auto *Next = Tok->Next;
+        const bool FollowedByName = Next && Next->Tok.getIdentifierInfo();
+        const bool SpaceAfter = Next && Next->hasWhitespaceBefore();
+
+        if ((!PrecededByName && !FollowedByName) ||
+            // e.g. `int * i` or `int*i`
+            (PrecededByName && FollowedByName && SpaceBefore == SpaceAfter)) {
+          continue;
         }
-        bool SpaceBefore = Tok->hasWhitespaceBefore();
-        bool SpaceAfter = Tok->Next->hasWhitespaceBefore();
-        if (SpaceBefore && !SpaceAfter)
+
+        if ((PrecededByName && SpaceBefore) ||
+            (FollowedByName && !SpaceAfter)) {
+          // Right alignment.
           ++AlignmentDiff;
-        if (!SpaceBefore && SpaceAfter)
+        } else if ((PrecededByName && !SpaceBefore) ||
+                   (FollowedByName && SpaceAfter)) {
+          // Left alignment.
           --AlignmentDiff;
+        }
       }
     }
+
     return AlignmentDiff;
   }
 
diff --git a/clang/lib/Sema/Sema.cpp b/clang/lib/Sema/Sema.cpp
index 56608e9..d50eeff 100644
--- a/clang/lib/Sema/Sema.cpp
+++ b/clang/lib/Sema/Sema.cpp
@@ -1616,6 +1616,8 @@ void Sema::ActOnEndOfTranslationUnit() {
 
   if (!PP.isIncrementalProcessingEnabled())
     TUScope = nullptr;
+
+  checkExposure(Context.getTranslationUnitDecl());
 }
 
 
diff --git a/clang/lib/Sema/SemaModule.cpp b/clang/lib/Sema/SemaModule.cpp
index 7c982bc..98ebd70 100644
--- a/clang/lib/Sema/SemaModule.cpp
+++ b/clang/lib/Sema/SemaModule.cpp
@@ -13,6 +13,7 @@
 
 #include "clang/AST/ASTConsumer.h"
 #include "clang/AST/ASTMutationListener.h"
+#include "clang/AST/RecursiveASTVisitor.h"
 #include "clang/Lex/HeaderSearch.h"
 #include "clang/Lex/Preprocessor.h"
 #include "clang/Sema/ParsedAttr.h"
@@ -485,6 +486,7 @@ Sema::ActOnModuleDecl(SourceLocation StartLoc, SourceLocation ModuleLoc,
   // implementation unit importing its interface).  Make this module visible
   // and return the import decl to be added to the current TU.
   if (Interface) {
+    HadImportedNamedModules = true;
 
     makeTransitiveImportsVisible(getASTContext(), VisibleModules, Interface,
                                  Mod, ModuleLoc,
@@ -728,6 +730,8 @@ DeclResult Sema::ActOnModuleImport(SourceLocation StartLoc,
       getCurrentModule()->Imports.insert(Mod);
   }
 
+  HadImportedNamedModules = true;
+
   return Import;
 }
 
@@ -1102,3 +1106,471 @@ bool Sema::isCurrentModulePurview() const {
     return false;
   }
 }
+
+//===----------------------------------------------------------------------===//
+// Checking Exposure in modules                                               //
+//===----------------------------------------------------------------------===//
+
+namespace {
+class ExposureChecker {
+public:
+  ExposureChecker(Sema &S) : SemaRef(S) {}
+
+  bool checkExposure(const VarDecl *D, bool Diag);
+  bool checkExposure(const CXXRecordDecl *D, bool Diag);
+  bool checkExposure(const Stmt *S, bool Diag);
+  bool checkExposure(const FunctionDecl *D, bool Diag);
+  bool checkExposure(const NamedDecl *D, bool Diag);
+  void checkExposureInContext(const DeclContext *DC);
+  bool isExposureCandidate(const NamedDecl *D);
+
+  bool isTULocal(QualType Ty);
+  bool isTULocal(const NamedDecl *ND);
+  bool isTULocal(const Expr *E);
+
+  Sema &SemaRef;
+
+private:
+  llvm::DenseSet<const NamedDecl *> ExposureSet;
+  llvm::DenseSet<const NamedDecl *> KnownNonExposureSet;
+};
+
+bool ExposureChecker::isTULocal(QualType Ty) {
+  // [basic.link]p15:
+  // An entity is TU-local if it is
+  // - a type, type alias, namespace, namespace alias, function, variable, or
+  // template that
+  //   -- has internal linkage, or
+  return Ty->getLinkage() == Linkage::Internal;
+
+  // TODO:
+  // [basic.link]p15.2:
+  //   a type with no name that is defined outside a class-specifier, function
+  //   body, or initializer or is introduced by a defining-type-specifier that
+  //   is used to declare only TU-local entities,
+}
+
+bool ExposureChecker::isTULocal(const NamedDecl *D) {
+  if (!D)
+    return false;
+
+  // [basic.link]p15:
+  // An entity is TU-local if it is
+  // - a type, type alias, namespace, namespace alias, function, variable, or
+  // template that
+  //   -- has internal linkage, or
+  if (D->getLinkageInternal() == Linkage::Internal)
+    return true;
+
+  if (D->isInAnonymousNamespace())
+    return true;
+
+  // [basic.link]p15.1.2:
+  // does not have a name with linkage and is declared, or introduced by a
+  // lambda-expression, within the definition of a TU-local entity,
+  if (D->getLinkageInternal() == Linkage::None)
+    if (auto *ND = dyn_cast<NamedDecl>(D->getDeclContext());
+        ND && isTULocal(ND))
+      return true;
+
+  // [basic.link]p15.3, p15.4:
+  // - a specialization of a TU-local template,
+  // - a specialization of a template with any TU-local template argument, or
+  ArrayRef<TemplateArgument> TemplateArgs;
+  NamedDecl *PrimaryTemplate = nullptr;
+  if (auto *CTSD = dyn_cast<ClassTemplateSpecializationDecl>(D)) {
+    TemplateArgs = CTSD->getTemplateArgs().asArray();
+    PrimaryTemplate = CTSD->getSpecializedTemplate();
+    if (isTULocal(PrimaryTemplate))
+      return true;
+  } else if (auto *VTSD = dyn_cast<VarTemplateSpecializationDecl>(D)) {
+    TemplateArgs = VTSD->getTemplateArgs().asArray();
+    PrimaryTemplate = VTSD->getSpecializedTemplate();
+    if (isTULocal(PrimaryTemplate))
+      return true;
+  } else if (auto *FD = dyn_cast<FunctionDecl>(D)) {
+    if (auto *TAList = FD->getTemplateSpecializationArgs())
+      TemplateArgs = TAList->asArray();
+
+    PrimaryTemplate = FD->getPrimaryTemplate();
+    if (isTULocal(PrimaryTemplate))
+      return true;
+  }
+
+  if (!PrimaryTemplate)
+    // Following off, we only check for specializations.
+    return false;
+
+  if (KnownNonExposureSet.count(D))
+    return false;
+
+  for (auto &TA : TemplateArgs) {
+    switch (TA.getKind()) {
+    case TemplateArgument::Type:
+      if (isTULocal(TA.getAsType()))
+        return true;
+      break;
+    case TemplateArgument::Declaration:
+      if (isTULocal(TA.getAsDecl()))
+        return true;
+      break;
+    default:
+      break;
+    }
+  }
+
+  // [basic.link]p15.5
+  // - a specialization of a template whose (possibly instantiated) declaration
+  // is an exposure.
+  if (checkExposure(PrimaryTemplate, /*Diag=*/false))
+    return true;
+
+  // Avoid calling checkExposure again since it is expensive.
+  KnownNonExposureSet.insert(D);
+  return false;
+}
+
+bool ExposureChecker::isTULocal(const Expr *E) {
+  if (!E)
+    return false;
+
+  // [basic.link]p16:
+  //    A value or object is TU-local if either
+  //    - it is of TU-local type,
+  if (isTULocal(E->getType()))
+    return true;
+
+  E = E->IgnoreParenImpCasts();
+  // [basic.link]p16.2:
+  //   - it is, or is a pointer to, a TU-local function or the object associated
+  //   with a TU-local variable,
+  //  - it is an object of class or array type and any of its subobjects or any
+  //  of the objects or functions to which its non-static data members of
+  //  reference type refer is TU-local and is usable in constant expressions, or
+  // FIXME: But how can we know the value of pointers or arrays at compile time?
+  if (const auto *DRE = dyn_cast<DeclRefExpr>(E)) {
+    if (auto *FD = dyn_cast_or_null<FunctionDecl>(DRE->getFoundDecl()))
+      return isTULocal(FD);
+    else if (auto *VD = dyn_cast_or_null<VarDecl>(DRE->getFoundDecl()))
+      return isTULocal(VD);
+    else if (auto *RD = dyn_cast_or_null<CXXRecordDecl>(DRE->getFoundDecl()))
+      return isTULocal(RD);
+  }
+
+  // TODO:
+  // [basic.link]p16.4:
+  //  it is a reflection value that represents...
+
+  return false;
+}
+
+bool ExposureChecker::isExposureCandidate(const NamedDecl *D) {
+  if (!D)
+    return false;
+
+  // [basic.link]p17:
+  //   If a (possibly instantiated) declaration of, or a deduction guide for,
+  //   a non-TU-local entity in a module interface unit
+  //   (outside the private-module-fragment, if any) or
+  //   module partition is an exposure, the program is ill-formed.
+  Module *M = D->getOwningModule();
+  if (!M || !M->isInterfaceOrPartition())
+    return false;
+
+  if (D->isImplicit())
+    return false;
+
+  // [basic.link]p14:
+  //      A declaration is an exposure if it either names a TU-local entity
+  //      (defined below), ignoring:
+  //      ...
+  //      - friend declarations in a class definition
+  if (D->getFriendObjectKind() &&
+      isa<CXXRecordDecl>(D->getLexicalDeclContext()))
+    return false;
+
+  return true;
+}
+
+bool ExposureChecker::checkExposure(const NamedDecl *D, bool Diag) {
+  if (!isExposureCandidate(D))
+    return false;
+
+  if (auto *FD = dyn_cast<FunctionDecl>(D))
+    return checkExposure(FD, Diag);
+  if (auto *FTD = dyn_cast<FunctionTemplateDecl>(D))
+    return checkExposure(FTD->getTemplatedDecl(), Diag);
+
+  if (auto *VD = dyn_cast<VarDecl>(D))
+    return checkExposure(VD, Diag);
+  if (auto *VTD = dyn_cast<VarTemplateDecl>(D))
+    return checkExposure(VTD->getTemplatedDecl(), Diag);
+
+  if (auto *RD = dyn_cast<CXXRecordDecl>(D))
+    return checkExposure(RD, Diag);
+
+  if (auto *CTD = dyn_cast<ClassTemplateDecl>(D))
+    return checkExposure(CTD->getTemplatedDecl(), Diag);
+
+  return false;
+}
+
+bool ExposureChecker::checkExposure(const FunctionDecl *FD, bool Diag) {
+  bool IsExposure = false;
+  if (isTULocal(FD->getReturnType())) {
+    IsExposure = true;
+    if (Diag)
+      SemaRef.Diag(FD->getReturnTypeSourceRange().getBegin(),
+                   diag::warn_exposure)
+          << FD->getReturnType();
+  }
+
+  for (ParmVarDecl *Parms : FD->parameters())
+    if (isTULocal(Parms->getType())) {
+      IsExposure = true;
+      if (Diag)
+        SemaRef.Diag(Parms->getLocation(), diag::warn_exposure)
+            << Parms->getType();
+    }
+
+  bool IsImplicitInstantiation =
+      FD->getTemplateSpecializationKind() == TSK_ImplicitInstantiation;
+
+  // [basic.link]p14:
+  //      A declaration is an exposure if it either names a TU-local entity
+  //      (defined below), ignoring:
+  //      - the function-body for a non-inline function or function template
+  //      (but not the deduced return
+  //        type for a (possibly instantiated) definition of a function with a
+  //        declared return type that uses a placeholder type
+  //        ([dcl.spec.auto])),
+  Diag &=
+      (FD->isInlined() || IsImplicitInstantiation) && !FD->isDependentContext();
+
+  IsExposure |= checkExposure(FD->getBody(), Diag);
+  if (IsExposure)
+    ExposureSet.insert(FD);
+
+  return IsExposure;
+}
+
+bool ExposureChecker::checkExposure(const VarDecl *VD, bool Diag) {
+  bool IsExposure = false;
+  // [basic.link]p14:
+  //  A declaration is an exposure if it either names a TU-local entity (defined
+  //  below), ignoring:
+  //  ...
+  //  or defines a constexpr variable initialized to a TU-local value (defined
+  //  below).
+  if (VD->isConstexpr() && isTULocal(VD->getInit())) {
+    IsExposure = true;
+    if (Diag)
+      SemaRef.Diag(VD->getInit()->getExprLoc(), diag::warn_exposure)
+          << VD->getInit();
+  }
+
+  if (isTULocal(VD->getType())) {
+    IsExposure = true;
+    if (Diag)
+      SemaRef.Diag(VD->getLocation(), diag::warn_exposure) << VD->getType();
+  }
+
+  // [basic.link]p14:
+  //   ..., ignoring:
+  //   - the initializer for a variable or variable template (but not the
+  //   variable's type),
+  //
+  // Note: although the spec says to ignore the initializer for all variable,
+  // for the code we generated now for inline variables, it is dangerous if the
+  // initializer of an inline variable is TULocal.
+  Diag &= !VD->getDeclContext()->isDependentContext() && VD->isInline();
+  IsExposure |= checkExposure(VD->getInit(), Diag);
+  if (IsExposure)
+    ExposureSet.insert(VD);
+
+  return IsExposure;
+}
+
+bool ExposureChecker::checkExposure(const CXXRecordDecl *RD, bool Diag) {
+  if (!RD->hasDefinition())
+    return false;
+
+  bool IsExposure = false;
+  for (CXXMethodDecl *Method : RD->methods())
+    IsExposure |= checkExposure(Method, Diag);
+
+  for (FieldDecl *FD : RD->fields()) {
+    if (isTULocal(FD->getType())) {
+      IsExposure = true;
+      if (Diag)
+        SemaRef.Diag(FD->getLocation(), diag::warn_exposure) << FD->getType();
+    }
+  }
+
+  for (const CXXBaseSpecifier &Base : RD->bases()) {
+    if (isTULocal(Base.getType())) {
+      IsExposure = true;
+      if (Diag)
+        SemaRef.Diag(Base.getBaseTypeLoc(), diag::warn_exposure)
+            << Base.getType();
+    }
+  }
+
+  if (IsExposure)
+    ExposureSet.insert(RD);
+
+  return IsExposure;
+}
+
+template <typename CallbackTy>
+class ReferenceTULocalChecker
+    : public clang::RecursiveASTVisitor<ReferenceTULocalChecker<CallbackTy>> {
+public:
+  ReferenceTULocalChecker(ExposureChecker &C, CallbackTy &&Callback)
+      : Checker(C), Callback(std::move(Callback)) {}
+
+  bool VisitDeclRefExpr(DeclRefExpr *DRE) {
+    ValueDecl *Referenced = DRE->getDecl();
+    if (!Referenced)
+      return true;
+
+    if (!Checker.isTULocal(Referenced))
+      // We don't care if the referenced declaration is not TU-local.
+      return true;
+
+    Qualifiers Qual = DRE->getType().getQualifiers();
+    // [basic.link]p14:
+    //      A declaration is an exposure if it either names a TU-local entity
+    //      (defined below), ignoring:
+    //      ...
+    //      - any reference to a non-volatile const object ...
+    if (Qual.hasConst() && !Qual.hasVolatile())
+      return true;
+
+    // [basic.link]p14:
+    // ..., ignoring:
+    //   ...
+    //   (p14.4) - ... or reference with internal or no linkage initialized with
+    //   a constant expression that is not an odr-use
+    ASTContext &Context = Referenced->getASTContext();
+    Linkage L = Referenced->getLinkageInternal();
+    if (DRE->isNonOdrUse() && (L == Linkage::Internal || L == Linkage::None))
+      if (auto *VD = dyn_cast<VarDecl>(Referenced);
+          VD && VD->getInit() && !VD->getInit()->isValueDependent() &&
+          VD->getInit()->isConstantInitializer(Context, /*IsForRef=*/false))
+        return true;
+
+    Callback(DRE, Referenced);
+    return true;
+  }
+
+  ExposureChecker &Checker;
+  CallbackTy Callback;
+};
+
+template <typename CallbackTy>
+ReferenceTULocalChecker(ExposureChecker &, CallbackTy &&)
+    -> ReferenceTULocalChecker<CallbackTy>;
+
+bool ExposureChecker::checkExposure(const Stmt *S, bool Diag) {
+  if (!S)
+    return false;
+
+  bool HasReferencedTULocals = false;
+  ReferenceTULocalChecker Checker(
+      *this, [this, &HasReferencedTULocals, Diag](DeclRefExpr *DRE,
+                                                  ValueDecl *Referenced) {
+        if (Diag) {
+          SemaRef.Diag(DRE->getExprLoc(), diag::warn_exposure) << Referenced;
+        }
+        HasReferencedTULocals = true;
+      });
+  Checker.TraverseStmt(const_cast<Stmt *>(S));
+  return HasReferencedTULocals;
+}
+
+void ExposureChecker::checkExposureInContext(const DeclContext *DC) {
+  for (auto *TopD : DC->noload_decls()) {
+    auto *TopND = dyn_cast<NamedDecl>(TopD);
+    if (!TopND)
+      continue;
+
+    if (auto *Namespace = dyn_cast<NamespaceDecl>(TopND)) {
+      checkExposureInContext(Namespace);
+      continue;
+    }
+
+    // [basic.link]p17:
+    //   If a (possibly instantiated) declaration of, or a deduction guide for,
+    //   a non-TU-local entity in a module interface unit
+    //   (outside the private-module-fragment, if any) or
+    //   module partition is an exposure, the program is ill-formed.
+    if (!TopND->isFromASTFile() && isExposureCandidate(TopND) &&
+        !isTULocal(TopND))
+      checkExposure(TopND, /*Diag=*/true);
+  }
+}
+
+} // namespace
+
+void Sema::checkExposure(const TranslationUnitDecl *TU) {
+  if (!TU)
+    return;
+
+  ExposureChecker Checker(*this);
+
+  Module *M = TU->getOwningModule();
+  if (M && M->isInterfaceOrPartition())
+    Checker.checkExposureInContext(TU);
+
+  // [basic.link]p18:
+  //   If a declaration that appears in one translation unit names a TU-local
+  //   entity declared in another translation unit that is not a header unit,
+  //   the program is ill-formed.
+  for (auto FDAndInstantiationLocPair : PendingCheckReferenceForTULocal) {
+    FunctionDecl *FD = FDAndInstantiationLocPair.first;
+    SourceLocation PointOfInstantiation = FDAndInstantiationLocPair.second;
+
+    if (!FD->hasBody())
+      continue;
+
+    ReferenceTULocalChecker(Checker, [&, this](DeclRefExpr *DRE,
+                                               ValueDecl *Referenced) {
+      // A "defect" in current implementation. Now an implicit instantiation of
+      // a template, the instantiation is considered to be in the same module
+      // unit as the template instead of the module unit where the instantiation
+      // happens.
+      //
+      // See test/Modules/Exposre-2.cppm for example.
+      if (!Referenced->isFromASTFile())
+        return;
+
+      if (!Referenced->isInAnotherModuleUnit())
+        return;
+
+      // This is not standard conforming. But given there are too many static
+      // (inline) functions in headers in existing code, it is more user
+      // friendly to ignore them temporarily now. maybe we can have another flag
+      // for this.
+      if (Referenced->getOwningModule()->isExplicitGlobalModule() &&
+          isa<FunctionDecl>(Referenced))
+        return;
+
+      Diag(PointOfInstantiation,
+           diag::warn_reference_tu_local_entity_in_other_tu)
+          << FD << Referenced
+          << Referenced->getOwningModule()->getTopLevelModuleName();
+    }).TraverseStmt(FD->getBody());
+  }
+}
+
+void Sema::checkReferenceToTULocalFromOtherTU(
+    FunctionDecl *FD, SourceLocation PointOfInstantiation) {
+  // Checking if a declaration have any reference to TU-local entities in other
+  // TU is expensive. Try to avoid it as much as possible.
+  if (!FD || !HadImportedNamedModules)
+    return;
+
+  PendingCheckReferenceForTULocal.push_back(
+      std::make_pair(FD, PointOfInstantiation));
+}
diff --git a/clang/lib/Sema/SemaOpenACC.cpp b/clang/lib/Sema/SemaOpenACC.cpp
index 128a5db..8bfea62 100644
--- a/clang/lib/Sema/SemaOpenACC.cpp
+++ b/clang/lib/Sema/SemaOpenACC.cpp
@@ -699,11 +699,19 @@ ExprResult SemaOpenACC::ActOnVar(OpenACCDirectiveKind DK, OpenACCClauseKind CK,
   // OpenACC3.3 2.13:
   // A 'var' in a 'declare' directive must be a variable or array name.
   if ((CK == OpenACCClauseKind::UseDevice ||
-       DK == OpenACCDirectiveKind::Declare) &&
-      isa<ArraySectionExpr, ArraySubscriptExpr>(CurVarExpr)) {
-    Diag(VarExpr->getExprLoc(), diag::err_acc_not_a_var_ref_use_device_declare)
-        << (DK == OpenACCDirectiveKind::Declare);
-    return ExprError();
+       DK == OpenACCDirectiveKind::Declare)) {
+    if (isa<ArraySubscriptExpr>(CurVarExpr)) {
+      Diag(VarExpr->getExprLoc(),
+           diag::err_acc_not_a_var_ref_use_device_declare)
+          << (DK == OpenACCDirectiveKind::Declare);
+      return ExprError();
+    }
+    // As an extension, we allow 'array sections'/'sub-arrays'  here, as that is
+    // effectively defining an array, and are in common use.
+    if (isa<ArraySectionExpr>(CurVarExpr))
+      Diag(VarExpr->getExprLoc(),
+           diag::ext_acc_array_section_use_device_declare)
+          << (DK == OpenACCDirectiveKind::Declare);
   }
 
   // Sub-arrays/subscript-exprs are fine as long as the base is a
diff --git a/clang/lib/Sema/SemaOpenACCClause.cpp b/clang/lib/Sema/SemaOpenACCClause.cpp
index 3f90fe8..b54a012 100644
--- a/clang/lib/Sema/SemaOpenACCClause.cpp
+++ b/clang/lib/Sema/SemaOpenACCClause.cpp
@@ -1919,6 +1919,14 @@ ExprResult SemaOpenACC::CheckReductionVar(OpenACCDirectiveKind DirectiveKind,
           << EltTy << /*Sub array base type*/ 1;
       return ExprError();
     }
+  } else if (VarExpr->getType()->isArrayType()) {
+    // Arrays are considered an 'aggregate variable' explicitly, so are OK, no
+    // additional checking required.
+    //
+    // Glossary: Aggregate variables – a variable of any non-scalar datatype,
+    // including array or composite variables.
+    //
+    // The next branch (record decl) checks for composite variables.
   } else if (auto *RD = VarExpr->getType()->getAsRecordDecl()) {
     if (!RD->isStruct() && !RD->isClass()) {
       Diag(VarExpr->getExprLoc(), diag::err_acc_reduction_composite_type)
@@ -2246,7 +2254,13 @@ bool SemaOpenACC::CheckDeclareClause(SemaOpenACC::OpenACCParsedClause &Clause,
         continue;
       }
     } else {
-      const auto *DRE = cast<DeclRefExpr>(VarExpr);
+
+      const Expr *VarExprTemp = VarExpr;
+
+      while (const auto *ASE = dyn_cast<ArraySectionExpr>(VarExprTemp))
+        VarExprTemp = ASE->getBase()->IgnoreParenImpCasts();
+
+      const auto *DRE = cast<DeclRefExpr>(VarExprTemp);
       if (const auto *Var = dyn_cast<VarDecl>(DRE->getDecl())) {
         CurDecl = Var->getCanonicalDecl();
 
diff --git a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
index e2c3cdc..233bb65 100644
--- a/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
+++ b/clang/lib/Sema/SemaTemplateInstantiateDecl.cpp
@@ -5853,6 +5853,8 @@ void Sema::InstantiateFunctionDefinition(SourceLocation PointOfInstantiation,
     // context seems wrong. Investigate more.
     ActOnFinishFunctionBody(Function, Body.get(), /*IsInstantiation=*/true);
 
+    checkReferenceToTULocalFromOtherTU(Function, PointOfInstantiation);
+
     PerformDependentDiagnostics(PatternDecl, TemplateArgs);
 
     if (auto *Listener = getASTMutationListener())
diff --git a/clang/lib/Sema/SemaWasm.cpp b/clang/lib/Sema/SemaWasm.cpp
index 6faea24..8998492 100644
--- a/clang/lib/Sema/SemaWasm.cpp
+++ b/clang/lib/Sema/SemaWasm.cpp
@@ -227,6 +227,53 @@ bool SemaWasm::BuiltinWasmTableCopy(CallExpr *TheCall) {
   return false;
 }
 
+bool SemaWasm::BuiltinWasmTestFunctionPointerSignature(CallExpr *TheCall) {
+  if (SemaRef.checkArgCount(TheCall, 1))
+    return true;
+
+  Expr *FuncPtrArg = TheCall->getArg(0);
+  QualType ArgType = FuncPtrArg->getType();
+
+  // Check that the argument is a function pointer
+  const PointerType *PtrTy = ArgType->getAs<PointerType>();
+  if (!PtrTy) {
+    return Diag(FuncPtrArg->getBeginLoc(),
+                diag::err_typecheck_expect_function_pointer)
+           << ArgType << FuncPtrArg->getSourceRange();
+  }
+
+  const FunctionProtoType *FuncTy =
+      PtrTy->getPointeeType()->getAs<FunctionProtoType>();
+  if (!FuncTy) {
+    return Diag(FuncPtrArg->getBeginLoc(),
+                diag::err_typecheck_expect_function_pointer)
+           << ArgType << FuncPtrArg->getSourceRange();
+  }
+
+  // Check that the function pointer doesn't use reference types
+  if (FuncTy->getReturnType().isWebAssemblyReferenceType()) {
+    return Diag(
+               FuncPtrArg->getBeginLoc(),
+               diag::err_wasm_builtin_test_fp_sig_cannot_include_reference_type)
+           << 0 << FuncTy->getReturnType() << FuncPtrArg->getSourceRange();
+  }
+  auto NParams = FuncTy->getNumParams();
+  for (unsigned I = 0; I < NParams; I++) {
+    if (FuncTy->getParamType(I).isWebAssemblyReferenceType()) {
+      return Diag(
+                 FuncPtrArg->getBeginLoc(),
+                 diag::
+                     err_wasm_builtin_test_fp_sig_cannot_include_reference_type)
+             << 1 << FuncPtrArg->getSourceRange();
+    }
+  }
+
+  // Set return type to int (the result of the test)
+  TheCall->setType(getASTContext().IntTy);
+
+  return false;
+}
+
 bool SemaWasm::CheckWebAssemblyBuiltinFunctionCall(const TargetInfo &TI,
                                                    unsigned BuiltinID,
                                                    CallExpr *TheCall) {
@@ -249,6 +296,8 @@ bool SemaWasm::CheckWebAssemblyBuiltinFunctionCall(const TargetInfo &TI,
     return BuiltinWasmTableFill(TheCall);
   case WebAssembly::BI__builtin_wasm_table_copy:
     return BuiltinWasmTableCopy(TheCall);
+  case WebAssembly::BI__builtin_wasm_test_function_pointer_signature:
+    return BuiltinWasmTestFunctionPointerSignature(TheCall);
   }
 
   return false;
diff --git a/clang/lib/Serialization/ASTReader.cpp b/clang/lib/Serialization/ASTReader.cpp
index 10aedb6..f896f9f1 100644
--- a/clang/lib/Serialization/ASTReader.cpp
+++ b/clang/lib/Serialization/ASTReader.cpp
@@ -8488,6 +8488,7 @@ bool ASTReader::LoadExternalSpecializationsImpl(SpecLookupTableTy &SpecLookups,
 bool ASTReader::LoadExternalSpecializations(const Decl *D, bool OnlyPartial) {
   assert(D);
 
+  CompleteRedeclChain(D);
   bool NewSpecsFound =
       LoadExternalSpecializationsImpl(PartialSpecializationsLookups, D);
   if (OnlyPartial)
diff --git a/clang/lib/StaticAnalyzer/Checkers/DereferenceChecker.cpp b/clang/lib/StaticAnalyzer/Checkers/DereferenceChecker.cpp
index d7eea7e..152129e 100644
--- a/clang/lib/StaticAnalyzer/Checkers/DereferenceChecker.cpp
+++ b/clang/lib/StaticAnalyzer/Checkers/DereferenceChecker.cpp
@@ -25,18 +25,22 @@ using namespace clang;
 using namespace ento;
 
 namespace {
+
+class DerefBugType : public BugType {
+  StringRef ArrayMsg, FieldMsg;
+
+public:
+  DerefBugType(CheckerFrontend *FE, StringRef Desc, const char *AMsg,
+               const char *FMsg = nullptr)
+      : BugType(FE, Desc), ArrayMsg(AMsg), FieldMsg(FMsg ? FMsg : AMsg) {}
+  StringRef getArrayMsg() const { return ArrayMsg; }
+  StringRef getFieldMsg() const { return FieldMsg; }
+};
+
 class DereferenceChecker
-    : public Checker< check::Location,
-                      check::Bind,
-                      EventDispatcher<ImplicitNullDerefEvent> > {
-  enum DerefKind {
-    NullPointer,
-    UndefinedPointerValue,
-    AddressOfLabel,
-    FixedAddress,
-  };
-
-  void reportBug(DerefKind K, ProgramStateRef State, const Stmt *S,
+    : public CheckerFamily<check::Location, check::Bind,
+                           EventDispatcher<ImplicitNullDerefEvent>> {
+  void reportBug(const DerefBugType &BT, ProgramStateRef State, const Stmt *S,
                  CheckerContext &C) const;
 
   bool suppressReport(CheckerContext &C, const Expr *E) const;
@@ -52,13 +56,23 @@ public:
                              const LocationContext *LCtx,
                              bool loadedFrom = false);
 
-  bool CheckNullDereference = false;
-  bool CheckFixedDereference = false;
-
-  std::unique_ptr<BugType> BT_Null;
-  std::unique_ptr<BugType> BT_Undef;
-  std::unique_ptr<BugType> BT_Label;
-  std::unique_ptr<BugType> BT_FixedAddress;
+  CheckerFrontend NullDerefChecker, FixedDerefChecker;
+  const DerefBugType NullBug{&NullDerefChecker, "Dereference of null pointer",
+                             "a null pointer dereference",
+                             "a dereference of a null pointer"};
+  const DerefBugType UndefBug{&NullDerefChecker,
+                              "Dereference of undefined pointer value",
+                              "an undefined pointer dereference",
+                              "a dereference of an undefined pointer value"};
+  const DerefBugType LabelBug{&NullDerefChecker,
+                              "Dereference of the address of a label",
+                              "an undefined pointer dereference",
+                              "a dereference of an address of a label"};
+  const DerefBugType FixedAddressBug{&FixedDerefChecker,
+                                     "Dereference of a fixed address",
+                                     "a dereference of a fixed address"};
+
+  StringRef getDebugTag() const override { return "DereferenceChecker"; }
 };
 } // end anonymous namespace
 
@@ -158,115 +172,87 @@ static bool isDeclRefExprToReference(const Expr *E) {
   return false;
 }
 
-void DereferenceChecker::reportBug(DerefKind K, ProgramStateRef State,
-                                   const Stmt *S, CheckerContext &C) const {
-  const BugType *BT = nullptr;
-  llvm::StringRef DerefStr1;
-  llvm::StringRef DerefStr2;
-  switch (K) {
-  case DerefKind::NullPointer:
-    if (!CheckNullDereference) {
-      C.addSink();
-      return;
-    }
-    BT = BT_Null.get();
-    DerefStr1 = " results in a null pointer dereference";
-    DerefStr2 = " results in a dereference of a null pointer";
-    break;
-  case DerefKind::UndefinedPointerValue:
-    if (!CheckNullDereference) {
-      C.addSink();
+void DereferenceChecker::reportBug(const DerefBugType &BT,
+                                   ProgramStateRef State, const Stmt *S,
+                                   CheckerContext &C) const {
+  if (&BT == &FixedAddressBug) {
+    if (!FixedDerefChecker.isEnabled())
+      // Deliberately don't add a sink node if check is disabled.
+      // This situation may be valid in special cases.
       return;
-    }
-    BT = BT_Undef.get();
-    DerefStr1 = " results in an undefined pointer dereference";
-    DerefStr2 = " results in a dereference of an undefined pointer value";
-    break;
-  case DerefKind::AddressOfLabel:
-    if (!CheckNullDereference) {
+  } else {
+    if (!NullDerefChecker.isEnabled()) {
       C.addSink();
       return;
     }
-    BT = BT_Label.get();
-    DerefStr1 = " results in an undefined pointer dereference";
-    DerefStr2 = " results in a dereference of an address of a label";
-    break;
-  case DerefKind::FixedAddress:
-    // Deliberately don't add a sink node if check is disabled.
-    // This situation may be valid in special cases.
-    if (!CheckFixedDereference)
-      return;
-
-    BT = BT_FixedAddress.get();
-    DerefStr1 = " results in a dereference of a fixed address";
-    DerefStr2 = " results in a dereference of a fixed address";
-    break;
-  };
+  }
 
   // Generate an error node.
   ExplodedNode *N = C.generateErrorNode(State);
   if (!N)
     return;
 
-  SmallString<100> buf;
-  llvm::raw_svector_ostream os(buf);
+  SmallString<100> Buf;
+  llvm::raw_svector_ostream Out(Buf);
 
   SmallVector<SourceRange, 2> Ranges;
 
   switch (S->getStmtClass()) {
   case Stmt::ArraySubscriptExprClass: {
-    os << "Array access";
+    Out << "Array access";
     const ArraySubscriptExpr *AE = cast<ArraySubscriptExpr>(S);
-    AddDerefSource(os, Ranges, AE->getBase()->IgnoreParenCasts(),
-                   State.get(), N->getLocationContext());
-    os << DerefStr1;
+    AddDerefSource(Out, Ranges, AE->getBase()->IgnoreParenCasts(), State.get(),
+                   N->getLocationContext());
+    Out << " results in " << BT.getArrayMsg();
     break;
   }
   case Stmt::ArraySectionExprClass: {
-    os << "Array access";
+    Out << "Array access";
     const ArraySectionExpr *AE = cast<ArraySectionExpr>(S);
-    AddDerefSource(os, Ranges, AE->getBase()->IgnoreParenCasts(),
-                   State.get(), N->getLocationContext());
-    os << DerefStr1;
+    AddDerefSource(Out, Ranges, AE->getBase()->IgnoreParenCasts(), State.get(),
+                   N->getLocationContext());
+    Out << " results in " << BT.getArrayMsg();
     break;
   }
   case Stmt::UnaryOperatorClass: {
-    os << BT->getDescription();
+    Out << BT.getDescription();
     const UnaryOperator *U = cast<UnaryOperator>(S);
-    AddDerefSource(os, Ranges, U->getSubExpr()->IgnoreParens(),
-                   State.get(), N->getLocationContext(), true);
+    AddDerefSource(Out, Ranges, U->getSubExpr()->IgnoreParens(), State.get(),
+                   N->getLocationContext(), true);
     break;
   }
   case Stmt::MemberExprClass: {
     const MemberExpr *M = cast<MemberExpr>(S);
     if (M->isArrow() || isDeclRefExprToReference(M->getBase())) {
-      os << "Access to field '" << M->getMemberNameInfo() << "'" << DerefStr2;
-      AddDerefSource(os, Ranges, M->getBase()->IgnoreParenCasts(),
-                     State.get(), N->getLocationContext(), true);
+      Out << "Access to field '" << M->getMemberNameInfo() << "' results in "
+          << BT.getFieldMsg();
+      AddDerefSource(Out, Ranges, M->getBase()->IgnoreParenCasts(), State.get(),
+                     N->getLocationContext(), true);
     }
     break;
   }
   case Stmt::ObjCIvarRefExprClass: {
     const ObjCIvarRefExpr *IV = cast<ObjCIvarRefExpr>(S);
-    os << "Access to instance variable '" << *IV->getDecl() << "'" << DerefStr2;
-    AddDerefSource(os, Ranges, IV->getBase()->IgnoreParenCasts(),
-                   State.get(), N->getLocationContext(), true);
+    Out << "Access to instance variable '" << *IV->getDecl() << "' results in "
+        << BT.getFieldMsg();
+    AddDerefSource(Out, Ranges, IV->getBase()->IgnoreParenCasts(), State.get(),
+                   N->getLocationContext(), true);
     break;
   }
   default:
     break;
   }
 
-  auto report = std::make_unique<PathSensitiveBugReport>(
-      *BT, buf.empty() ? BT->getDescription() : buf.str(), N);
+  auto BR = std::make_unique<PathSensitiveBugReport>(
+      BT, Buf.empty() ? BT.getDescription() : Buf.str(), N);
 
-  bugreporter::trackExpressionValue(N, bugreporter::getDerefExpr(S), *report);
+  bugreporter::trackExpressionValue(N, bugreporter::getDerefExpr(S), *BR);
 
   for (SmallVectorImpl<SourceRange>::iterator
        I = Ranges.begin(), E = Ranges.end(); I!=E; ++I)
-    report->addRange(*I);
+    BR->addRange(*I);
 
-  C.emitReport(std::move(report));
+  C.emitReport(std::move(BR));
 }
 
 void DereferenceChecker::checkLocation(SVal l, bool isLoad, const Stmt* S,
@@ -275,7 +261,7 @@ void DereferenceChecker::checkLocation(SVal l, bool isLoad, const Stmt* S,
   if (l.isUndef()) {
     const Expr *DerefExpr = getDereferenceExpr(S);
     if (!suppressReport(C, DerefExpr))
-      reportBug(DerefKind::UndefinedPointerValue, C.getState(), DerefExpr, C);
+      reportBug(UndefBug, C.getState(), DerefExpr, C);
     return;
   }
 
@@ -296,7 +282,7 @@ void DereferenceChecker::checkLocation(SVal l, bool isLoad, const Stmt* S,
       // we call an "explicit" null dereference.
       const Expr *expr = getDereferenceExpr(S);
       if (!suppressReport(C, expr)) {
-        reportBug(DerefKind::NullPointer, nullState, expr, C);
+        reportBug(NullBug, nullState, expr, C);
         return;
       }
     }
@@ -314,7 +300,7 @@ void DereferenceChecker::checkLocation(SVal l, bool isLoad, const Stmt* S,
   if (location.isConstant()) {
     const Expr *DerefExpr = getDereferenceExpr(S, isLoad);
     if (!suppressReport(C, DerefExpr))
-      reportBug(DerefKind::FixedAddress, notNullState, DerefExpr, C);
+      reportBug(FixedAddressBug, notNullState, DerefExpr, C);
     return;
   }
 
@@ -330,7 +316,7 @@ void DereferenceChecker::checkBind(SVal L, SVal V, const Stmt *S,
 
   // One should never write to label addresses.
   if (auto Label = L.getAs<loc::GotoLabel>()) {
-    reportBug(DerefKind::AddressOfLabel, C.getState(), S, C);
+    reportBug(LabelBug, C.getState(), S, C);
     return;
   }
 
@@ -351,7 +337,7 @@ void DereferenceChecker::checkBind(SVal L, SVal V, const Stmt *S,
     if (!StNonNull) {
       const Expr *expr = getDereferenceExpr(S, /*IsBind=*/true);
       if (!suppressReport(C, expr)) {
-        reportBug(DerefKind::NullPointer, StNull, expr, C);
+        reportBug(NullBug, StNull, expr, C);
         return;
       }
     }
@@ -369,7 +355,7 @@ void DereferenceChecker::checkBind(SVal L, SVal V, const Stmt *S,
   if (V.isConstant()) {
     const Expr *DerefExpr = getDereferenceExpr(S, true);
     if (!suppressReport(C, DerefExpr))
-      reportBug(DerefKind::FixedAddress, State, DerefExpr, C);
+      reportBug(FixedAddressBug, State, DerefExpr, C);
     return;
   }
 
@@ -392,26 +378,8 @@ void DereferenceChecker::checkBind(SVal L, SVal V, const Stmt *S,
   C.addTransition(State, this);
 }
 
-void ento::registerDereferenceModeling(CheckerManager &Mgr) {
-  Mgr.registerChecker<DereferenceChecker>();
-}
-
-bool ento::shouldRegisterDereferenceModeling(const CheckerManager &) {
-  return true;
-}
-
 void ento::registerNullDereferenceChecker(CheckerManager &Mgr) {
-  auto *Chk = Mgr.getChecker<DereferenceChecker>();
-  Chk->CheckNullDereference = true;
-  Chk->BT_Null.reset(new BugType(Mgr.getCurrentCheckerName(),
-                                 "Dereference of null pointer",
-                                 categories::LogicError));
-  Chk->BT_Undef.reset(new BugType(Mgr.getCurrentCheckerName(),
-                                  "Dereference of undefined pointer value",
-                                  categories::LogicError));
-  Chk->BT_Label.reset(new BugType(Mgr.getCurrentCheckerName(),
-                                  "Dereference of the address of a label",
-                                  categories::LogicError));
+  Mgr.getChecker<DereferenceChecker>()->NullDerefChecker.enable(Mgr);
 }
 
 bool ento::shouldRegisterNullDereferenceChecker(const CheckerManager &) {
@@ -419,11 +387,7 @@ bool ento::shouldRegisterNullDereferenceChecker(const CheckerManager &) {
 }
 
 void ento::registerFixedAddressDereferenceChecker(CheckerManager &Mgr) {
-  auto *Chk = Mgr.getChecker<DereferenceChecker>();
-  Chk->CheckFixedDereference = true;
-  Chk->BT_FixedAddress.reset(new BugType(Mgr.getCurrentCheckerName(),
-                                         "Dereference of a fixed address",
-                                         categories::LogicError));
+  Mgr.getChecker<DereferenceChecker>()->FixedDerefChecker.enable(Mgr);
 }
 
 bool ento::shouldRegisterFixedAddressDereferenceChecker(
diff --git a/clang/test/Analysis/analyzer-enabled-checkers.c b/clang/test/Analysis/analyzer-enabled-checkers.c
index 78ee00de..a632b70 100644
--- a/clang/test/Analysis/analyzer-enabled-checkers.c
+++ b/clang/test/Analysis/analyzer-enabled-checkers.c
@@ -14,7 +14,6 @@
 // CHECK-NEXT: core.BitwiseShift
 // CHECK-NEXT: core.CallAndMessageModeling
 // CHECK-NEXT: core.CallAndMessage
-// CHECK-NEXT: core.DereferenceModeling
 // CHECK-NEXT: core.DivideZero
 // CHECK-NEXT: core.DynamicTypePropagation
 // CHECK-NEXT: core.FixedAddressDereference
diff --git a/clang/test/Analysis/std-c-library-functions-arg-enabled-checkers.c b/clang/test/Analysis/std-c-library-functions-arg-enabled-checkers.c
index 7f9c9ff..b388c31 100644
--- a/clang/test/Analysis/std-c-library-functions-arg-enabled-checkers.c
+++ b/clang/test/Analysis/std-c-library-functions-arg-enabled-checkers.c
@@ -22,7 +22,6 @@
 // CHECK-NEXT: core.BitwiseShift
 // CHECK-NEXT: core.CallAndMessageModeling
 // CHECK-NEXT: core.CallAndMessage
-// CHECK-NEXT: core.DereferenceModeling
 // CHECK-NEXT: core.DivideZero
 // CHECK-NEXT: core.DynamicTypePropagation
 // CHECK-NEXT: core.FixedAddressDereference
diff --git a/clang/test/CIR/CodeGen/complex.cpp b/clang/test/CIR/CodeGen/complex.cpp
index 0a7765f..bd7de9a 100644
--- a/clang/test/CIR/CodeGen/complex.cpp
+++ b/clang/test/CIR/CodeGen/complex.cpp
@@ -780,3 +780,22 @@ void foo29() {
 // OGCG: %[[INIT_IMAG_PTR:.*]] = getelementptr inbounds nuw { i32, i32 }, ptr %[[INIT]], i32 0, i32 1
 // OGCG: store i32 0, ptr %[[INIT_REAL_PTR]], align 4
 // OGCG: store i32 0, ptr %[[INIT_IMAG_PTR]], align 4
+
+void foo30() {
+  float _Complex a = { 1.0f };
+}
+
+// CIR: %[[A_ADDR:.*]] = cir.alloca !cir.complex<!cir.float>, !cir.ptr<!cir.complex<!cir.float>>, ["a", init]
+// CIR: %[[CONST_1F:.*]] = cir.const #cir.fp<1.000000e+00> : !cir.float
+// CIR: %[[CONST_0F:.*]] = cir.const #cir.fp<0.000000e+00> : !cir.float
+// CIR: %[[COMPLEX:.*]] = cir.complex.create %[[CONST_1F]], %[[CONST_0F]] : !cir.float -> !cir.complex<!cir.float>
+// CIR: cir.store{{.*}} %[[COMPLEX]], %[[A_ADDR]] : !cir.complex<!cir.float>, !cir.ptr<!cir.complex<!cir.float>>
+
+// LLVM: %[[A_ADDR:.*]] = alloca { float, float }, i64 1, align 4
+// LLVM: store { float, float } { float 1.000000e+00, float 0.000000e+00 }, ptr %[[A_ADDR]], align 4
+
+// OGCG:  %[[A_ADDR:.*]] = alloca { float, float }, align 4
+// OGCG: %[[A_REAL_PTR:.*]] = getelementptr inbounds nuw { float, float }, ptr %[[A_ADDR]], i32 0, i32 0
+// OGCG: %[[A_IMAG_PTR:.*]] = getelementptr inbounds nuw { float, float }, ptr %[[A_ADDR]], i32 0, i32 1
+// OGCG: store float 1.000000e+00, ptr %[[A_REAL_PTR]], align 4
+// OGCG: store float 0.000000e+00, ptr %[[A_IMAG_PTR]], align 4
diff --git a/clang/test/CIR/CodeGen/struct-init.cpp b/clang/test/CIR/CodeGen/struct-init.cpp
new file mode 100644
index 0000000..a47ef53
--- /dev/null
+++ b/clang/test/CIR/CodeGen/struct-init.cpp
@@ -0,0 +1,184 @@
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -fclangir -emit-cir %s -o %t.cir
+// RUN: FileCheck --check-prefix=CIR --input-file=%t.cir %s
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -fclangir -emit-llvm %s -o %t-cir.ll
+// RUN: FileCheck --check-prefix=LLVM --input-file=%t-cir.ll %s
+// RUN: %clang_cc1 -triple x86_64-unknown-linux-gnu -emit-llvm %s -o %t.ll
+// RUN: FileCheck --check-prefix=OGCG --input-file=%t.ll %s
+
+struct S {
+  int a, b, c;
+};
+
+void init() {
+  S s1 = {1, 2, 3};
+  S s2 = {4, 5};
+}
+
+// CIR: cir.func{{.*}} @_Z4initv()
+// CIR:   %[[S1:.*]] = cir.alloca !rec_S, !cir.ptr<!rec_S>, ["s1", init]
+// CIR:   %[[S2:.*]] = cir.alloca !rec_S, !cir.ptr<!rec_S>, ["s2", init]
+// CIR:   %[[S1_A:.*]] = cir.get_member %[[S1]][0] {name = "a"}
+// CIR:   %[[ONE:.*]] = cir.const #cir.int<1>
+// CIR:   cir.store{{.*}} %[[ONE]], %[[S1_A]]
+// CIR:   %[[S1_B:.*]] = cir.get_member %[[S1]][1] {name = "b"}
+// CIR:   %[[TWO:.*]] = cir.const #cir.int<2>
+// CIR:   cir.store{{.*}} %[[TWO]], %[[S1_B]]
+// CIR:   %[[S1_C:.*]] = cir.get_member %[[S1]][2] {name = "c"}
+// CIR:   %[[THREE:.*]] = cir.const #cir.int<3>
+// CIR:   cir.store{{.*}} %[[THREE]], %[[S1_C]]
+// CIR:   %[[S2_A:.*]] = cir.get_member %[[S2]][0] {name = "a"}
+// CIR:   %[[FOUR:.*]] = cir.const #cir.int<4>
+// CIR:   cir.store{{.*}} %[[FOUR]], %[[S2_A]]
+// CIR:   %[[S2_B:.*]] = cir.get_member %[[S2]][1] {name = "b"}
+// CIR:   %[[FIVE:.*]] = cir.const #cir.int<5>
+// CIR:   cir.store{{.*}} %[[FIVE]], %[[S2_B]]
+// CIR:   %[[S2_C:.*]] = cir.get_member %[[S2]][2] {name = "c"}
+// CIR:   %[[ZERO:.*]] = cir.const #cir.int<0>
+// CIR:   cir.store{{.*}} %[[ZERO]], %[[S2_C]]
+// CIR:   cir.return
+
+// LLVM: define{{.*}} void @_Z4initv()
+// LLVM:   %[[S1:.*]] = alloca %struct.S
+// LLVM:   %[[S2:.*]] = alloca %struct.S
+// LLVM:   %[[S1_A:.*]] = getelementptr %struct.S, ptr %[[S1]], i32 0, i32 0
+// LLVM:   store i32 1, ptr %[[S1_A]]
+// LLVM:   %[[S1_B:.*]] = getelementptr %struct.S, ptr %[[S1]], i32 0, i32 1
+// LLVM:   store i32 2, ptr %[[S1_B]]
+// LLVM:   %[[S1_C:.*]] = getelementptr %struct.S, ptr %[[S1]], i32 0, i32 2
+// LLVM:   store i32 3, ptr %[[S1_C]]
+// LLVM:   %[[S2_A:.*]] = getelementptr %struct.S, ptr %[[S2]], i32 0, i32 0
+// LLVM:   store i32 4, ptr %[[S2_A]]
+// LLVM:   %[[S2_B:.*]] = getelementptr %struct.S, ptr %[[S2]], i32 0, i32 1
+// LLVM:   store i32 5, ptr %[[S2_B]]
+// LLVM:   %[[S2_C:.*]] = getelementptr %struct.S, ptr %[[S2]], i32 0, i32 2
+// LLVM:   store i32 0, ptr %[[S2_C]]
+
+// OGCG: @__const._Z4initv.s1 = private unnamed_addr constant %struct.S { i32 1, i32 2, i32 3 }
+// OGCG: @__const._Z4initv.s2 = private unnamed_addr constant %struct.S { i32 4, i32 5, i32 0 }
+
+// OGCG: define{{.*}} void @_Z4initv()
+// OGCG:   %[[S1:.*]] = alloca %struct.S
+// OGCG:   %[[S2:.*]] = alloca %struct.S
+// OGCG:   call void @llvm.memcpy.p0.p0.i64(ptr{{.*}} %[[S1]], ptr{{.*}} @__const._Z4initv.s1, i64 12, i1 false)
+// OGCG:   call void @llvm.memcpy.p0.p0.i64(ptr{{.*}} %[[S2]], ptr{{.*}} @__const._Z4initv.s2, i64 12, i1 false)
+
+void init_var(int a, int b) {
+  S s = {a, b};
+}
+
+// CIR: cir.func{{.*}} @_Z8init_varii(%[[A_ARG:.*]]: !s32i {{.*}}, %[[B_ARG:.*]]: !s32i {{.*}})
+// CIR:   %[[A_PTR:.*]] = cir.alloca !s32i, !cir.ptr<!s32i>, ["a", init]
+// CIR:   %[[B_PTR:.*]] = cir.alloca !s32i, !cir.ptr<!s32i>, ["b", init]
+// CIR:   %[[S:.*]] = cir.alloca !rec_S, !cir.ptr<!rec_S>, ["s", init]
+// CIR:   cir.store{{.*}} %[[A_ARG]], %[[A_PTR]]
+// CIR:   cir.store{{.*}} %[[B_ARG]], %[[B_PTR]]
+// CIR:   %[[S_A:.*]] = cir.get_member %[[S]][0] {name = "a"}
+// CIR:   %[[A:.*]] = cir.load{{.*}} %[[A_PTR]]
+// CIR:   cir.store{{.*}} %[[A]], %[[S_A]]
+// CIR:   %[[S_B:.*]] = cir.get_member %[[S]][1] {name = "b"}
+// CIR:   %[[B:.*]] = cir.load{{.*}} %[[B_PTR]]
+// CIR:   cir.store{{.*}} %[[B]], %[[S_B]]
+// CIR:   cir.return
+
+// LLVM: define{{.*}} void @_Z8init_varii(i32 %[[A_ARG:.*]], i32 %[[B_ARG:.*]])
+// LLVM:   %[[A_PTR:.*]] = alloca i32
+// LLVM:   %[[B_PTR:.*]] = alloca i32
+// LLVM:   %[[S:.*]] = alloca %struct.S
+// LLVM:   store i32 %[[A_ARG]], ptr %[[A_PTR]]
+// LLVM:   store i32 %[[B_ARG]], ptr %[[B_PTR]]
+// LLVM:   %[[S_A:.*]] = getelementptr %struct.S, ptr %[[S]], i32 0, i32 0
+// LLVM:   %[[A:.*]] = load i32, ptr %[[A_PTR]] 
+// LLVM:   store i32 %[[A]], ptr %[[S_A]]
+// LLVM:   %[[S_B:.*]] = getelementptr %struct.S, ptr %[[S]], i32 0, i32 1
+// LLVM:   %[[B:.*]] = load i32, ptr %[[B_PTR]]
+// LLVM:   store i32 %[[B]], ptr %[[S_B]]
+// LLVM:   ret void
+
+// OGCG: define{{.*}} void @_Z8init_varii(i32 {{.*}} %[[A_ARG:.*]], i32 {{.*}} %[[B_ARG:.*]])
+// OGCG:   %[[A_PTR:.*]] = alloca i32
+// OGCG:   %[[B_PTR:.*]] = alloca i32
+// OGCG:   %[[S:.*]] = alloca %struct.S
+// OGCG:   store i32 %[[A_ARG]], ptr %[[A_PTR]]
+// OGCG:   store i32 %[[B_ARG]], ptr %[[B_PTR]]
+// OGCG:   %[[S_A:.*]] = getelementptr {{.*}} %struct.S, ptr %[[S]], i32 0, i32 0
+// OGCG:   %[[A:.*]] = load i32, ptr %[[A_PTR]] 
+// OGCG:   store i32 %[[A]], ptr %[[S_A]]
+// OGCG:   %[[S_B:.*]] = getelementptr {{.*}} %struct.S, ptr %[[S]], i32 0, i32 1
+// OGCG:   %[[B:.*]] = load i32, ptr %[[B_PTR]]
+// OGCG:   store i32 %[[B]], ptr %[[S_B]]
+// OGCG:   %[[S_C:.*]] = getelementptr {{.*}} %struct.S, ptr %[[S]], i32 0, i32 2
+// OGCG:   store i32 0, ptr %[[S_C]]
+// OGCG:   ret void
+
+void init_expr(int a, int b, int c) {
+  S s = {a + 1, b + 2, c + 3};
+}
+
+// CIR: cir.func{{.*}} @_Z9init_expriii(%[[A_ARG:.*]]: !s32i {{.*}}, %[[B_ARG:.*]]: !s32i {{.*}}, %[[C_ARG:.*]]: !s32i {{.*}})
+// CIR:   %[[A_PTR:.*]] = cir.alloca !s32i, !cir.ptr<!s32i>, ["a", init]
+// CIR:   %[[B_PTR:.*]] = cir.alloca !s32i, !cir.ptr<!s32i>, ["b", init]
+// CIR:   %[[C_PTR:.*]] = cir.alloca !s32i, !cir.ptr<!s32i>, ["c", init]
+// CIR:   %[[S:.*]] = cir.alloca !rec_S, !cir.ptr<!rec_S>, ["s", init]
+// CIR:   cir.store{{.*}} %[[A_ARG]], %[[A_PTR]]
+// CIR:   cir.store{{.*}} %[[B_ARG]], %[[B_PTR]]
+// CIR:   cir.store{{.*}} %[[C_ARG]], %[[C_PTR]]
+// CIR:   %[[S_A:.*]] = cir.get_member %[[S]][0] {name = "a"}
+// CIR:   %[[A:.*]] = cir.load{{.*}} %[[A_PTR]]
+// CIR:   %[[ONE:.*]] = cir.const #cir.int<1>
+// CIR:   %[[A_PLUS_ONE:.*]] = cir.binop(add, %[[A]], %[[ONE]])
+// CIR:   cir.store{{.*}} %[[A_PLUS_ONE]], %[[S_A]]
+// CIR:   %[[S_B:.*]] = cir.get_member %[[S]][1] {name = "b"}
+// CIR:   %[[B:.*]] = cir.load{{.*}} %[[B_PTR]]
+// CIR:   %[[TWO:.*]] = cir.const #cir.int<2>
+// CIR:   %[[B_PLUS_TWO:.*]] = cir.binop(add, %[[B]], %[[TWO]]) nsw : !s32i
+// CIR:   cir.store{{.*}} %[[B_PLUS_TWO]], %[[S_B]]
+// CIR:   %[[S_C:.*]] = cir.get_member %[[S]][2] {name = "c"}
+// CIR:   %[[C:.*]] = cir.load{{.*}} %[[C_PTR]]
+// CIR:   %[[THREE:.*]] = cir.const #cir.int<3>
+// CIR:   %[[C_PLUS_THREE:.*]] = cir.binop(add, %[[C]], %[[THREE]]) nsw : !s32i
+// CIR:   cir.store{{.*}} %[[C_PLUS_THREE]], %[[S_C]]
+// CIR:   cir.return
+
+// LLVM: define{{.*}} void @_Z9init_expriii(i32 %[[A_ARG:.*]], i32 %[[B_ARG:.*]], i32 %[[C_ARG:.*]])
+// LLVM:   %[[A_PTR:.*]] = alloca i32
+// LLVM:   %[[B_PTR:.*]] = alloca i32
+// LLVM:   %[[C_PTR:.*]] = alloca i32
+// LLVM:   %[[S:.*]] = alloca %struct.S
+// LLVM:   store i32 %[[A_ARG]], ptr %[[A_PTR]]
+// LLVM:   store i32 %[[B_ARG]], ptr %[[B_PTR]]
+// LLVM:   store i32 %[[C_ARG]], ptr %[[C_PTR]]
+// LLVM:   %[[S_A:.*]] = getelementptr %struct.S, ptr %[[S]], i32 0, i32 0
+// LLVM:   %[[A:.*]] = load i32, ptr %[[A_PTR]] 
+// LLVM:   %[[A_PLUS_ONE:.*]] = add nsw i32 %[[A]], 1
+// LLVM:   store i32 %[[A_PLUS_ONE]], ptr %[[S_A]]
+// LLVM:   %[[S_B:.*]] = getelementptr %struct.S, ptr %[[S]], i32 0, i32 1
+// LLVM:   %[[B:.*]] = load i32, ptr %[[B_PTR]]
+// LLVM:   %[[B_PLUS_TWO:.*]] = add nsw i32 %[[B]], 2
+// LLVM:   store i32 %[[B_PLUS_TWO]], ptr %[[S_B]]
+// LLVM:   %[[S_C:.*]] = getelementptr %struct.S, ptr %[[S]], i32 0, i32 2
+// LLVM:   %[[C:.*]] = load i32, ptr %[[C_PTR]]
+// LLVM:   %[[C_PLUS_THREE:.*]] = add nsw i32 %[[C]], 3
+// LLVM:   store i32 %[[C_PLUS_THREE]], ptr %[[S_C]]
+// LLVM:   ret void
+
+// OGCG: define{{.*}} void @_Z9init_expriii(i32 {{.*}} %[[A_ARG:.*]], i32 {{.*}} %[[B_ARG:.*]], i32 {{.*}} %[[C_ARG:.*]])
+// OGCG:   %[[A_PTR:.*]] = alloca i32
+// OGCG:   %[[B_PTR:.*]] = alloca i32
+// OGCG:   %[[C_PTR:.*]] = alloca i32
+// OGCG:   %[[S:.*]] = alloca %struct.S
+// OGCG:   store i32 %[[A_ARG]], ptr %[[A_PTR]]
+// OGCG:   store i32 %[[B_ARG]], ptr %[[B_PTR]]
+// OGCG:   store i32 %[[C_ARG]], ptr %[[C_PTR]]
+// OGCG:   %[[S_A:.*]] = getelementptr {{.*}} %struct.S, ptr %[[S]], i32 0, i32 0
+// OGCG:   %[[A:.*]] = load i32, ptr %[[A_PTR]] 
+// OGCG:   %[[A_PLUS_ONE:.*]] = add nsw i32 %[[A]], 1
+// OGCG:   store i32 %[[A_PLUS_ONE]], ptr %[[S_A]]
+// OGCG:   %[[S_B:.*]] = getelementptr {{.*}} %struct.S, ptr %[[S]], i32 0, i32 1
+// OGCG:   %[[B:.*]] = load i32, ptr %[[B_PTR]]
+// OGCG:   %[[B_PLUS_TWO:.*]] = add nsw i32 %[[B]], 2
+// OGCG:   store i32 %[[B_PLUS_TWO]], ptr %[[S_B]]
+// OGCG:   %[[S_C:.*]] = getelementptr {{.*}} %struct.S, ptr %[[S]], i32 0, i32 2
+// OGCG:   %[[C:.*]] = load i32, ptr %[[C_PTR]]
+// OGCG:   %[[C_PLUS_THREE:.*]] = add nsw i32 %[[C]], 3
+// OGCG:   store i32 %[[C_PLUS_THREE]], ptr %[[S_C]]
+// OGCG:   ret void
diff --git a/clang/test/CIR/CodeGenOpenACC/host_data.c b/clang/test/CIR/CodeGenOpenACC/host_data.c
index aeaf3d2f..fa06d2a 100644
--- a/clang/test/CIR/CodeGenOpenACC/host_data.c
+++ b/clang/test/CIR/CodeGenOpenACC/host_data.c
@@ -1,13 +1,15 @@
 // RUN: %clang_cc1 -fopenacc -Wno-openacc-self-if-potential-conflict -emit-cir -fclangir %s -o - | FileCheck %s
 
-void acc_host_data(int cond, int var1, int var2) {
-  // CHECK: cir.func{{.*}} @acc_host_data(%[[ARG_COND:.*]]: !s32i {{.*}}, %[[ARG_V1:.*]]: !s32i {{.*}}, %[[ARG_V2:.*]]: !s32i {{.*}}) {
+void acc_host_data(int cond, int var1, int var2, int *arr) {
+  // CHECK: cir.func{{.*}} @acc_host_data(%[[ARG_COND:.*]]: !s32i {{.*}}, %[[ARG_V1:.*]]: !s32i {{.*}}, %[[ARG_V2:.*]]: !s32i {{.*}}, %[[ARG_ARR:.*]]: !cir.ptr<!s32i> {{.*}}) {
   // CHECK-NEXT: %[[COND:.*]] = cir.alloca !s32i, !cir.ptr<!s32i>, ["cond", init]
   // CHECK-NEXT: %[[V1:.*]] = cir.alloca !s32i, !cir.ptr<!s32i>, ["var1", init]
   // CHECK-NEXT: %[[V2:.*]] = cir.alloca !s32i, !cir.ptr<!s32i>, ["var2", init]
+  // CHECK-NEXT: %[[ARR:.*]] = cir.alloca !cir.ptr<!s32i>, !cir.ptr<!cir.ptr<!s32i>>, ["arr", init]
   // CHECK-NEXT: cir.store %[[ARG_COND]], %[[COND]] : !s32i, !cir.ptr<!s32i>
   // CHECK-NEXT: cir.store %[[ARG_V1]], %[[V1]] : !s32i, !cir.ptr<!s32i>
   // CHECK-NEXT: cir.store %[[ARG_V2]], %[[V2]] : !s32i, !cir.ptr<!s32i>
+  // CHECK-NEXT: cir.store %[[ARG_ARR]], %[[ARR]] : !cir.ptr<!s32i>, !cir.ptr<!cir.ptr<!s32i>>
 
 #pragma acc host_data use_device(var1)
   {}
@@ -52,4 +54,18 @@ void acc_host_data(int cond, int var1, int var2) {
   // CHECK-NEXT: acc.host_data if(%[[COND_CAST]]) dataOperands(%[[USE_DEV1]], %[[USE_DEV2]] : !cir.ptr<!s32i>, !cir.ptr<!s32i>) {
   // CHECK-NEXT: acc.terminator
   // CHECK-NEXT: } attributes {ifPresent}
+
+#pragma acc host_data use_device(arr[0:var1])
+  {}
+  // CHECK-NEXT: %[[ZERO:.*]] = cir.const #cir.int<0>
+  // CHECK-NEXT: %[[ZERO_CAST:.*]] = builtin.unrealized_conversion_cast %[[ZERO]] : !s32i to si32
+  // CHECK-NEXT: %[[VAR1_LOAD:.*]] = cir.load{{.*}} %[[V1]] : !cir.ptr<!s32i>, !s32i
+  // CHECK-NEXT: %[[VAR1_CAST:.*]] = builtin.unrealized_conversion_cast %[[VAR1_LOAD]] : !s32i to si32
+  // CHECK-NEXT: %[[CONST_ZERO:.*]] = arith.constant 0
+  // CHECK-NEXT: %[[CONST_ONE:.*]] = arith.constant 1
+  // CHECK-NEXT: %[[BOUNDS:.*]] = acc.bounds lowerbound(%[[ZERO_CAST]] : si32) extent(%[[VAR1_CAST]] : si32) stride(%[[CONST_ONE]] : i64) startIdx(%[[CONST_ZERO]] : i64)
+  // CHECK-NEXT: %[[USE_DEV1:.*]] = acc.use_device varPtr(%[[ARR]] : !cir.ptr<!cir.ptr<!s32i>>) bounds(%[[BOUNDS]]) -> !cir.ptr<!cir.ptr<!s32i>> {name = "arr[0:var1]"}
+  // CHECK-NEXT: acc.host_data dataOperands(%[[USE_DEV1]] : !cir.ptr<!cir.ptr<!s32i>>)
+  // CHECK-NEXT: acc.terminator
+  // CHECK-NEXT: } loc
 }
diff --git a/clang/test/CIR/Lowering/poison.cir b/clang/test/CIR/Lowering/poison.cir
new file mode 100644
index 0000000..6f8b792
--- /dev/null
+++ b/clang/test/CIR/Lowering/poison.cir
@@ -0,0 +1,14 @@
+// RUN: cir-translate -cir-to-llvmir --disable-cc-lowering -o %t.ll %s
+// RUN: FileCheck -check-prefix=LLVM --input-file=%t.ll %s
+
+!s32i = !cir.int<s, 32>
+
+module {
+  cir.func @lower_poison() -> !s32i {
+    %0 = cir.const #cir.poison : !s32i
+    cir.return %0 : !s32i
+  }
+  // LLVM-LABEL: @lower_poison
+  // LLVM-NEXT:    ret i32 poison
+  // LLVM-NEXT:  }
+}
diff --git a/clang/test/CIR/Transforms/bit.cir b/clang/test/CIR/Transforms/bit.cir
new file mode 100644
index 0000000..fc27adb
--- /dev/null
+++ b/clang/test/CIR/Transforms/bit.cir
@@ -0,0 +1,211 @@
+// RUN: cir-opt -cir-canonicalize -cir-simplify -o %t.cir %s
+// RUN: FileCheck --input-file=%t.cir %s
+
+!s32i = !cir.int<s, 32>
+!u32i = !cir.int<u, 32>
+
+module {
+  cir.func @fold_clrsb() -> !s32i {
+    %0 = cir.const #cir.int<1> : !s32i
+    %1 = cir.clrsb %0 : !s32i
+    cir.return %1 : !s32i
+  }
+  // CHECK-LABEL: @fold_clrsb
+  // CHECK-NEXT:    %[[R:.+]] = cir.const #cir.int<30> : !s32i
+  // CHECK-NEXT:    cir.return %[[R]] : !s32i
+  // CHECK-NEXT:  }
+
+  cir.func @fold_clz() -> !u32i {
+    %0 = cir.const #cir.int<1> : !u32i
+    %1 = cir.clz %0 : !u32i
+    cir.return %1 : !u32i
+  }
+  // CHECK-LABEL: @fold_clz
+  // CHECK-NEXT:    %[[R:.+]] = cir.const #cir.int<31> : !u32i
+  // CHECK-NEXT:    cir.return %[[R]] : !u32i
+  // CHECK-NEXT:  }
+
+  cir.func @fold_clz_zero_poison() -> !u32i {
+    %0 = cir.const #cir.int<0> : !u32i
+    %1 = cir.clz %0 poison_zero : !u32i
+    cir.return %1 : !u32i
+  }
+  // CHECK-LABEL: @fold_clz_zero_poison
+  // CHECK-NEXT:    %[[R:.+]] = cir.const #cir.poison : !u32i
+  // CHECK-NEXT:    cir.return %[[R]] : !u32i
+  // CHECK-NEXT:  }
+
+  cir.func @fold_clz_zero_no_poison() -> !u32i {
+    %0 = cir.const #cir.int<0> : !u32i
+    %1 = cir.clz %0 : !u32i
+    cir.return %1 : !u32i
+  }
+  // CHECK-LABEL: @fold_clz_zero_no_poison
+  // CHECK-NEXT:    %[[R:.+]] = cir.const #cir.int<32> : !u32i
+  // CHECK-NEXT:    cir.return %[[R]] : !u32i
+  // CHECK-NEXT:  }
+
+  cir.func @fold_ctz() -> !u32i {
+    %0 = cir.const #cir.int<2> : !u32i
+    %1 = cir.ctz %0 : !u32i
+    cir.return %1 : !u32i
+  }
+  // CHECK-LABEL: @fold_ctz
+  // CHECK-NEXT:    %[[R:.+]] = cir.const #cir.int<1> : !u32i
+  // CHECK-NEXT:    cir.return %[[R]] : !u32i
+  // CHECK-NEXT:  }
+
+  cir.func @fold_ctz_zero_poison() -> !u32i {
+    %0 = cir.const #cir.int<0> : !u32i
+    %1 = cir.ctz %0 poison_zero : !u32i
+    cir.return %1 : !u32i
+  }
+  // CHECK-LABEL: @fold_ctz_zero_poison
+  // CHECK-NEXT:    %[[R:.+]] = cir.const #cir.poison : !u32i
+  // CHECK-NEXT:    cir.return %[[R]] : !u32i
+  // CHECK-NEXT:  }
+
+  cir.func @fold_ctz_zero_no_poison() -> !u32i {
+    %0 = cir.const #cir.int<0> : !u32i
+    %1 = cir.ctz %0 : !u32i
+    cir.return %1 : !u32i
+  }
+  // CHECK-LABEL: @fold_ctz_zero_no_poison
+  // CHECK-NEXT:    %[[R:.+]] = cir.const #cir.int<32> : !u32i
+  // CHECK-NEXT:    cir.return %[[R]] : !u32i
+  // CHECK-NEXT:  }
+
+  cir.func @fold_parity() -> !u32i {
+    // 0xdeadbeef is 0b1101_1110_1010_1101_1011_1110_1110_1111
+    // 0xdeadbeef contains 24 ones
+    %0 = cir.const #cir.int<0xdeadbeef> : !u32i
+    %1 = cir.parity %0 : !u32i
+    cir.return %1 : !u32i
+  }
+  // CHECK-LABEL: @fold_parity
+  // CHECK-NEXT:    %[[R:.+]] = cir.const #cir.int<0> : !u32i
+  // CHECK-NEXT:    cir.return %[[R]] : !u32i
+  // CHECK-NEXT:  }
+
+  cir.func @fold_popcount() -> !u32i {
+    // 0xdeadbeef is 0b1101_1110_1010_1101_1011_1110_1110_1111
+    // 0xdeadbeef contains 24 ones
+    %0 = cir.const #cir.int<0xdeadbeef> : !u32i
+    %1 = cir.popcount %0 : !u32i
+    cir.return %1 : !u32i
+  }
+  // CHECK-LABEL: @fold_popcount
+  // CHECK-NEXT:    %[[R:.+]] = cir.const #cir.int<24> : !u32i
+  // CHECK-NEXT:    cir.return %[[R]] : !u32i
+  // CHECK-NEXT:  }
+
+  cir.func @fold_bitreverse() -> !u32i {
+    // 0xdeadbeef is 0b1101_1110_1010_1101_1011_1110_1110_1111
+    %0 = cir.const #cir.int<0xdeadbeef> : !u32i
+    %1 = cir.bitreverse %0 : !u32i
+    cir.return %1 : !u32i
+  }
+  // CHECK-LABEL: @fold_bitreverse
+  // 4152210811 is 0b1111_0111_0111_1101_1011_0101_0111_1011
+  // CHECK-NEXT:    %[[R:.+]] = cir.const #cir.int<4152210811> : !u32i
+  // CHECK-NEXT:    cir.return %[[R]] : !u32i
+  // CHECK-NEXT:  }
+
+  cir.func @fold_byte_swap() -> !u32i {
+    %0 = cir.const #cir.int<0xdeadbeef> : !u32i
+    %1 = cir.byte_swap %0 : !u32i
+    cir.return %1 : !u32i
+  }
+  // CHECK-LABEL: @fold_byte_swap
+  // 4022250974 is 0xefbeadde
+  // CHECK-NEXT:    %[[R:.+]] = cir.const #cir.int<4022250974> : !u32i
+  // CHECK-NEXT:    cir.return %[[R]] : !u32i
+  // CHECK-NEXT:  }
+
+  cir.func @fold_input_poison() -> !s32i {
+    %0 = cir.const #cir.poison : !s32i
+    %1 = cir.clrsb %0 : !s32i
+    cir.return %1 : !s32i
+  }
+  // CHECK-LABEL: @fold_input_poison
+  // CHECK-NEXT:    %[[P:.+]] = cir.const #cir.poison : !s32i
+  // CHECK-NEXT:    cir.return %[[P]] : !s32i
+  // CHECK-NEXT:  }
+
+  cir.func @fold_rotate_input_all_zeros(%arg0 : !u32i) -> !u32i {
+    %0 = cir.const #cir.int<0> : !u32i
+    %1 = cir.rotate left %0, %arg0 : !u32i
+    cir.return %1 : !u32i
+  }
+  // CHECK-LABEL: @fold_rotate_input_all_zeros
+  // CHECK-NEXT:    %[[R:.+]] = cir.const #cir.int<0> : !u32i
+  // CHECK-NEXT:    cir.return %[[R]] : !u32i
+  // CHECK-NEXT:  }
+
+  cir.func @fold_rotate_input_all_ones(%arg0 : !u32i) -> !u32i {
+    // 4294967295 is 0b1111_1111_1111_1111_1111_1111_1111_1111
+    %0 = cir.const #cir.int<4294967295> : !u32i
+    %1 = cir.rotate left %0, %arg0 : !u32i
+    cir.return %1 : !u32i
+  }
+  // CHECK-LABEL: @fold_rotate_input_all_ones
+  // CHECK-NEXT:    %[[R:.+]] = cir.const #cir.int<4294967295> : !u32i
+  // CHECK-NEXT:    cir.return %[[R]] : !u32i
+  // CHECK-NEXT:  }
+
+  cir.func @fold_rotate_zero_amount(%arg0 : !u32i) -> !u32i {
+    %0 = cir.const #cir.int<32> : !u32i
+    %1 = cir.rotate left %arg0, %0 : !u32i
+    cir.return %1 : !u32i
+  }
+  // CHECK-LABEL: @fold_rotate_zero_amount
+  // CHECK-SAME:  (%[[R:.+]]: !u32i)
+  // CHECK-NEXT:    cir.return %[[R]] : !u32i
+  // CHECK-NEXT:  }
+
+  cir.func @fold_rotate_left() -> !u32i {
+    // 0xdeadbeef is 0b1101_1110_1010_1101_1011_1110_1110_1111
+    %0 = cir.const #cir.int<0xdeadbeef> : !u32i
+    %1 = cir.const #cir.int<8> : !u32i
+    %2 = cir.rotate left %0, %1 : !u32i
+    cir.return %2 : !u32i
+  }
+  // CHECK-LABEL: @fold_rotate_left
+  // 2914971614 is 0b1010_1101_1011_1110_1110_1111_1101_1110
+  // CHECK-NEXT:    %[[R:.+]] = cir.const #cir.int<2914971614> : !u32i
+  // CHECK-NEXT:    cir.return %[[R]] : !u32i
+  // CHECK-NEXT:  }
+
+  cir.func @fold_rotate_right() -> !u32i {
+    // 0xdeadbeef is 0b1101_1110_1010_1101_1011_1110_1110_1111
+    %0 = cir.const #cir.int<0xdeadbeef> : !u32i
+    %1 = cir.const #cir.int<8> : !u32i
+    %2 = cir.rotate right %0, %1 : !u32i
+    cir.return %2 : !u32i
+  }
+  // CHECK-LABEL: @fold_rotate_right
+  // 4260027374 is 0b1110_1111_1101_1110_1010_1101_1011_1110
+  // CHECK-NEXT:    %[[R:.+]] = cir.const #cir.int<4024348094> : !u32i
+  // CHECK-NEXT:    cir.return %[[R]] : !u32i
+  // CHECK-NEXT:  }
+
+  cir.func @fold_rotate_input_poison(%arg0 : !u32i) -> !u32i {
+    %0 = cir.const #cir.poison : !u32i
+    %1 = cir.rotate left %0, %arg0 : !u32i
+    cir.return %1 : !u32i
+  }
+  // CHECK-LABEL: @fold_rotate_input_poison
+  // CHECK-NEXT:    %[[P:.+]] = cir.const #cir.poison : !u32i
+  // CHECK-NEXT:    cir.return %[[P]] : !u32i
+  // CHECK-NEXT:  }
+
+  cir.func @fold_rotate_amount_poison(%arg0 : !u32i) -> !u32i {
+    %0 = cir.const #cir.poison : !u32i
+    %1 = cir.rotate left %arg0, %0 : !u32i
+    cir.return %1 : !u32i
+  }
+  // CHECK-LABEL: @fold_rotate_amount_poison
+  // CHECK-NEXT:    %[[P:.+]] = cir.const #cir.poison : !u32i
+  // CHECK-NEXT:    cir.return %[[P]] : !u32i
+  // CHECK-NEXT:  }
+}
diff --git a/clang/test/CIR/Transforms/canonicalize.cir b/clang/test/CIR/Transforms/canonicalize.cir
index 7ba163e..5daff11 100644
--- a/clang/test/CIR/Transforms/canonicalize.cir
+++ b/clang/test/CIR/Transforms/canonicalize.cir
@@ -39,6 +39,16 @@ module {
   // CHECK:      cir.func{{.*}} @unary_not(%arg0: !cir.bool) -> !cir.bool
   // CHECK-NEXT:   cir.return %arg0 : !cir.bool
 
+  cir.func @unary_poison() -> !s32i {
+    %0 = cir.const #cir.poison : !s32i
+    %1 = cir.unary(inc, %0) : !s32i, !s32i
+    cir.return %1 : !s32i
+  }
+  // CHECK:      @unary_poison
+  // CHECK-NEXT:   %[[P:.+]] = cir.const #cir.poison : !s32i
+  // CHECK-NEXT:   cir.return %[[P]] : !s32i
+  // CHECK-NEXT: }
+
   cir.func @cast1(%arg0: !cir.bool) -> !cir.bool {
     %0 = cir.cast(bool_to_int, %arg0 : !cir.bool), !s32i
     %1 = cir.cast(int_to_bool, %0 : !s32i), !cir.bool
@@ -70,4 +80,14 @@ module {
   // CHECK-NEXT:   %[[CAST3:.*]] = cir.cast(integral, %[[CAST2]] : !s32i), !s64i
   // CHECK-NEXT:   cir.return %[[CAST3]] : !s64i
 
+  cir.func @cast_poison() -> !s64i {
+    %0 = cir.const #cir.poison : !s32i
+    %1 = cir.cast(integral, %0 : !s32i), !s64i
+    cir.return %1 : !s64i
+  }
+  // CHECK:      @cast_poison
+  // CHECK-NEXT:   %[[P:.+]] = cir.const #cir.poison : !s64i
+  // CHECK-NEXT:   cir.return %[[P]] : !s64i
+  // CHECK-NEXT: }
+
 }
diff --git a/clang/test/CXX/basic/basic.link/p19.cppm b/clang/test/CXX/basic/basic.link/p19.cppm
new file mode 100644
index 0000000..a0772d7
--- /dev/null
+++ b/clang/test/CXX/basic/basic.link/p19.cppm
@@ -0,0 +1,71 @@
+// RUN: rm -rf %t
+// RUN: mkdir -p %t
+// RUN: split-file %s %t
+//
+// RUN: %clang_cc1 -std=c++20 -emit-reduced-module-interface -verify %t/A.cppm -o %t/A.pcm
+// RUN: %clang_cc1 -std=c++20 %t/A.cpp -fmodule-file=A=%t/A.pcm -fsyntax-only -verify
+
+//--- A.cppm
+export module A;
+static void f() {}
+inline void it() { f(); }           // expected-warning {{TU local entity 'f' is exposed}}
+static inline void its() { f(); }   // OK
+template<int> void g() { its(); }   // OK
+template void g<0>();
+
+// Developers Note: We didn't track the use in decltype(). But it should be fine
+// since the underlying type is not TU-local. So we're doing nothing bad in practice.
+decltype(f) *fp;                    // error: f (though not its type) is TU-local
+auto &fr = f;                       // OK
+// Developers Note: We didn't track the use across variables. In the current implementation,
+// we don't know the value of `fr` at compile time, so we failed to detect this.
+constexpr auto &fr2 = fr;           // error: is an exposure of f
+// Developers Note: But if it is a direct use, we are able to detect it.
+constexpr auto &fr3 = f;            // expected-warning {{TU local entity 'f' is exposed}}
+constexpr static auto fp2 = fr;     // OK
+
+struct S { void (&ref)(); } s{f};               // OK, value is TU-local
+constexpr extern struct W { S &s; } wrap{s};    // OK, value is not TU-local
+
+static auto x = []{f();};           // OK
+auto x2 = x;                        // expected-warning {{TU local entity}}
+// Developers Note: Why is this an exposure?
+int y = ([]{f();}(),0);             // error: the closure type is not TU-local
+int y2 = (x,0);                     // OK expected-warning{{left operand of comma operator has no effect}}
+
+namespace N {
+  struct A {};
+  void adl(A);
+  static void adl(int);
+}
+void adl(double);
+
+inline void h(auto x) { adl(x); }   // OK, but certain specializations are exposures
+
+// Reflection is not supported yet.
+// constexpr std::meta::info r1 = ^^g<0>;  // OK
+// namespace N2 {
+//   static constexpr std::meta::info r2 = ^^g<1>;     // OK, r2 is TU-local
+// }
+// constexpr std::meta::info r3 = ^^f;                 // error: r3 is an exposure of f
+// 
+// constexpr auto ctx = std::meta::access_context::current();
+// constexpr std::meta::info r4 =
+//   std::meta::members_of(^^N2, ctx)[0];              // error: r4 is an exposure of N2​::​r2
+
+//--- A.cpp
+module A;
+void other() {
+  g<0>();                           // OK, specialization is explicitly instantiated
+  g<1>();                           // expected-warning {{instantiation of 'g<1>' triggers reference to TU-local entity 'its' from other TU 'A'}}
+  // Developers Note: To check use of TU-local entity when overload resolution made.
+  h(N::A{});                        // error: overload set contains TU-local N​::​adl(int)
+  h(0);                             // OK, calls adl(double)
+  adl(N::A{});                      // OK; N​::​adl(int) not found, calls N​::​adl(N​::​A)
+  fr();                             // OK, calls f
+  // Developers Note: To check use of TU-local entity when we're able to detect the TUlocalness
+  // across variables.
+  constexpr auto ptr = fr;          // error: fr is not usable in constant expressions here
+
+  constexpr auto fptr = f;          // expected-error {{use of undeclared identifier 'f'}}
+}
diff --git a/clang/test/CodeGen/bounds-checking-debuginfo.c b/clang/test/CodeGen/bounds-checking-debuginfo.c
index 74c0666..bd7aedd 100644
--- a/clang/test/CodeGen/bounds-checking-debuginfo.c
+++ b/clang/test/CodeGen/bounds-checking-debuginfo.c
@@ -25,13 +25,13 @@ void d(double*);
 // CHECK-TRAP-NEXT:    [[TMP1:%.*]] = icmp ult i64 [[TMP0]], 10, !dbg [[DBG23]], !nosanitize [[META10]]
 // CHECK-TRAP-NEXT:    br i1 [[TMP1]], label %[[CONT:.*]], label %[[TRAP:.*]], !dbg [[DBG23]], !prof [[PROF27:![0-9]+]], !nosanitize [[META10]]
 // CHECK-TRAP:       [[TRAP]]:
-// CHECK-TRAP-NEXT:    call void @llvm.ubsantrap(i8 18) #[[ATTR3:[0-9]+]], !dbg [[DBG23]], !nosanitize [[META10]]
-// CHECK-TRAP-NEXT:    unreachable, !dbg [[DBG23]], !nosanitize [[META10]]
+// CHECK-TRAP-NEXT:    call void @llvm.ubsantrap(i8 18) #[[ATTR3:[0-9]+]], !dbg [[DBGTRAP:![0-9]+]], !nosanitize [[META10]]
+// CHECK-TRAP-NEXT:    unreachable, !dbg [[DBGTRAP]], !nosanitize [[META10]]
 // CHECK-TRAP:       [[CONT]]:
 // CHECK-TRAP-NEXT:    [[IDXPROM:%.*]] = sext i32 [[CALL]] to i64, !dbg [[DBG26:![0-9]+]]
 // CHECK-TRAP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [10 x double], ptr [[A]], i64 0, i64 [[IDXPROM]], !dbg [[DBG26]]
 // CHECK-TRAP-NEXT:    [[TMP2:%.*]] = load double, ptr [[ARRAYIDX]], align 8, !dbg [[DBG26]]
-// CHECK-TRAP-NEXT:    ret double [[TMP2]], !dbg [[DBG28:![0-9]+]]
+// CHECK-TRAP-NEXT:    ret double [[TMP2]], !dbg [[DBG30:![0-9]+]]
 //
 // CHECK-NOTRAP-LABEL: define dso_local double @f1(
 // CHECK-NOTRAP-SAME: i32 noundef [[B:%.*]], i32 noundef [[I:%.*]]) #[[ATTR0:[0-9]+]] !dbg [[DBG4:![0-9]+]] {
@@ -93,7 +93,9 @@ double f1(int b, int i) {
 // CHECK-TRAP: [[META25]] = !DISubroutineType(types: null)
 // CHECK-TRAP: [[DBG26]] = !DILocation(line: 66, column: 10, scope: [[DBG4]])
 // CHECK-TRAP: [[PROF27]] = !{!"branch_weights", i32 1048575, i32 1}
-// CHECK-TRAP: [[DBG28]] = !DILocation(line: 66, column: 3, scope: [[DBG4]])
+// CHECK-TRAP: [[DBGTRAP]] = !DILocation(line: 0, scope: [[TRAPMSG:![0-9]+]], inlinedAt: [[DBG23]])
+// CHECK-TRAP: [[TRAPMSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Array index out of bounds", scope: [[META5]], file: [[META5]], type: [[META25]], flags: DIFlagArtificial, spFlags: DISPFlagDefinition, unit: [[META0]])
+// CHECK-TRAP: [[DBG30]] = !DILocation(line: 66, column: 3, scope: [[DBG4]])
 //.
 // CHECK-NOTRAP: [[META0:![0-9]+]] = distinct !DICompileUnit(language: DW_LANG_C11, file: [[META1:![0-9]+]], isOptimized: false, runtimeVersion: 0, emissionKind: FullDebug, splitDebugInlining: false, nameTableKind: None)
 // CHECK-NOTRAP: [[META1]] = !DIFile(filename: "<stdin>", directory: {{.*}})
diff --git a/clang/test/CodeGen/builtins-wasm.c b/clang/test/CodeGen/builtins-wasm.c
index d8aff82..f201dfe 100644
--- a/clang/test/CodeGen/builtins-wasm.c
+++ b/clang/test/CodeGen/builtins-wasm.c
@@ -1,6 +1,6 @@
-// RUN: %clang_cc1 -triple wasm32-unknown-unknown -target-feature +reference-types -target-feature +simd128 -target-feature +relaxed-simd -target-feature +nontrapping-fptoint -target-feature +exception-handling -target-feature +bulk-memory -target-feature +atomics -target-feature +fp16 -flax-vector-conversions=none -O3 -emit-llvm -o - %s | FileCheck %s -check-prefixes WEBASSEMBLY,WEBASSEMBLY32
-// RUN: %clang_cc1 -triple wasm64-unknown-unknown -target-feature +reference-types -target-feature +simd128 -target-feature +relaxed-simd -target-feature +nontrapping-fptoint -target-feature +exception-handling -target-feature +bulk-memory -target-feature +atomics -target-feature +fp16 -flax-vector-conversions=none -O3 -emit-llvm -o - %s | FileCheck %s -check-prefixes WEBASSEMBLY,WEBASSEMBLY64
-// RUN: not %clang_cc1 -triple wasm64-unknown-unknown -target-feature +reference-types -target-feature +nontrapping-fptoint -target-feature +exception-handling -target-feature +bulk-memory -target-feature +atomics -flax-vector-conversions=none -O3 -emit-llvm -o - %s 2>&1 | FileCheck %s -check-prefixes MISSING-SIMD
+// RUN: %clang_cc1 -triple wasm32-unknown-unknown -target-feature +reference-types -target-feature +simd128 -target-feature +relaxed-simd -target-feature +nontrapping-fptoint -target-feature +exception-handling -target-feature +bulk-memory -target-feature +atomics -target-feature +gc -target-feature +fp16 -flax-vector-conversions=none -O3 -emit-llvm -o - %s | FileCheck %s -check-prefixes WEBASSEMBLY,WEBASSEMBLY32
+// RUN: %clang_cc1 -triple wasm64-unknown-unknown -target-feature +reference-types -target-feature +simd128 -target-feature +relaxed-simd -target-feature +nontrapping-fptoint -target-feature +exception-handling -target-feature +bulk-memory -target-feature +atomics -target-feature +gc  -target-feature +fp16 -flax-vector-conversions=none -O3 -emit-llvm -o - %s | FileCheck %s -check-prefixes WEBASSEMBLY,WEBASSEMBLY64
+// RUN: not %clang_cc1 -triple wasm64-unknown-unknown -target-feature +reference-types -target-feature +nontrapping-fptoint -target-feature +exception-handling -target-feature +bulk-memory -target-feature +atomics -target-feature +gc -flax-vector-conversions=none -O3 -emit-llvm -o - %s 2>&1 | FileCheck %s -check-prefixes MISSING-SIMD
 
 // SIMD convenience types
 typedef signed char i8x16 __attribute((vector_size(16)));
@@ -751,3 +751,24 @@ void *tp (void) {
   return __builtin_thread_pointer ();
   // WEBASSEMBLY: call {{.*}} @llvm.thread.pointer.p0()
 }
+
+typedef void (*Fvoid)(void);
+typedef float (*Ffloats)(float, double, int);
+typedef void (*Fpointers)(Fvoid, Ffloats, void*, int*, int***, char[5]);
+
+void use(int);
+
+void test_function_pointer_signature_void(Fvoid func) {
+  // WEBASSEMBLY:  %0 = tail call i32 (ptr, ...) @llvm.wasm.ref.test.func(ptr %func, token poison)
+  use(__builtin_wasm_test_function_pointer_signature(func));
+}
+
+void test_function_pointer_signature_floats(Ffloats func) {
+  // WEBASSEMBLY:  tail call i32 (ptr, ...) @llvm.wasm.ref.test.func(ptr %func, float 0.000000e+00, token poison, float 0.000000e+00, double 0.000000e+00, i32 0)
+  use(__builtin_wasm_test_function_pointer_signature(func));
+}
+
+void test_function_pointer_signature_pointers(Fpointers func) {
+  // WEBASSEMBLY:  %0 = tail call i32 (ptr, ...) @llvm.wasm.ref.test.func(ptr %func, token poison, ptr null, ptr null, ptr null, ptr null, ptr null, ptr null)
+  use(__builtin_wasm_test_function_pointer_signature(func));
+}
diff --git a/clang/test/CodeGen/cfi-icall-generalize-debuginfo.c b/clang/test/CodeGen/cfi-icall-generalize-debuginfo.c
index 304b605..0ffc2b9 100644
--- a/clang/test/CodeGen/cfi-icall-generalize-debuginfo.c
+++ b/clang/test/CodeGen/cfi-icall-generalize-debuginfo.c
@@ -30,11 +30,11 @@ int** f(const char *a, const char **b) {
 // UNGENERALIZED-NEXT:    [[TMP0:%.*]] = tail call i1 @llvm.type.test(ptr [[FP]], metadata !"_ZTSFPPiPKcPS2_E"), !dbg [[DBG34:![0-9]+]], !nosanitize [[META38:![0-9]+]]
 // UNGENERALIZED-NEXT:    br i1 [[TMP0]], label %[[CONT:.*]], label %[[TRAP:.*]], !dbg [[DBG34]], !prof [[PROF39:![0-9]+]], !nosanitize [[META38]]
 // UNGENERALIZED:       [[TRAP]]:
-// UNGENERALIZED-NEXT:    tail call void @llvm.ubsantrap(i8 2) #[[ATTR4:[0-9]+]], !dbg [[DBG34]], !nosanitize [[META38]]
-// UNGENERALIZED-NEXT:    unreachable, !dbg [[DBG34]], !nosanitize [[META38]]
+// UNGENERALIZED-NEXT:    tail call void @llvm.ubsantrap(i8 2) #[[ATTR4:[0-9]+]], !dbg [[DBGTRAP:![0-9]+]], !nosanitize [[META38]]
+// UNGENERALIZED-NEXT:    unreachable, !dbg [[DBGTRAP]], !nosanitize [[META38]]
 // UNGENERALIZED:       [[CONT]]:
 // UNGENERALIZED-NEXT:    [[CALL:%.*]] = tail call ptr [[FP]](ptr noundef null, ptr noundef null) #[[ATTR5:[0-9]+]], !dbg [[DBG37:![0-9]+]]
-// UNGENERALIZED-NEXT:    ret void, !dbg [[DBG40:![0-9]+]]
+// UNGENERALIZED-NEXT:    ret void, !dbg [[DBG42:![0-9]+]]
 //
 // GENERALIZED-LABEL: define dso_local void @g(
 // GENERALIZED-SAME: ptr noundef [[FP:%.*]]) local_unnamed_addr #[[ATTR1:[0-9]+]] !dbg [[DBG25:![0-9]+]] !type [[META31:![0-9]+]] !type [[META32:![0-9]+]] {
@@ -43,11 +43,11 @@ int** f(const char *a, const char **b) {
 // GENERALIZED-NEXT:    [[TMP0:%.*]] = tail call i1 @llvm.type.test(ptr [[FP]], metadata !"_ZTSFPvPKvS_E.generalized"), !dbg [[DBG34:![0-9]+]], !nosanitize [[META38:![0-9]+]]
 // GENERALIZED-NEXT:    br i1 [[TMP0]], label %[[CONT:.*]], label %[[TRAP:.*]], !dbg [[DBG34]], !prof [[PROF39:![0-9]+]], !nosanitize [[META38]]
 // GENERALIZED:       [[TRAP]]:
-// GENERALIZED-NEXT:    tail call void @llvm.ubsantrap(i8 2) #[[ATTR4:[0-9]+]], !dbg [[DBG34]], !nosanitize [[META38]]
-// GENERALIZED-NEXT:    unreachable, !dbg [[DBG34]], !nosanitize [[META38]]
+// GENERALIZED-NEXT:    tail call void @llvm.ubsantrap(i8 2) #[[ATTR4:[0-9]+]], !dbg [[DBGTRAP:![0-9]+]], !nosanitize [[META38]]
+// GENERALIZED-NEXT:    unreachable, !dbg [[DBGTRAP]], !nosanitize [[META38]]
 // GENERALIZED:       [[CONT]]:
 // GENERALIZED-NEXT:    [[CALL:%.*]] = tail call ptr [[FP]](ptr noundef null, ptr noundef null) #[[ATTR5:[0-9]+]], !dbg [[DBG37:![0-9]+]]
-// GENERALIZED-NEXT:    ret void, !dbg [[DBG40:![0-9]+]]
+// GENERALIZED-NEXT:    ret void, !dbg [[DBG42:![0-9]+]]
 //
 void g(int** (*fp)(const char *, const char **)) {
   fp(0, 0);
@@ -90,7 +90,9 @@ void g(int** (*fp)(const char *, const char **)) {
 // UNGENERALIZED: [[DBG37]] = !DILocation(line: 53, column: 3, scope: [[DBG25]])
 // UNGENERALIZED: [[META38]] = !{}
 // UNGENERALIZED: [[PROF39]] = !{!"branch_weights", i32 1048575, i32 1}
-// UNGENERALIZED: [[DBG40]] = !DILocation(line: 54, column: 1, scope: [[DBG25]])
+// UNGENERALIZED: [[DBGTRAP]] = !DILocation(line: 0, scope: [[TRAPMSG:![0-9]+]], inlinedAt: [[DBG34]])
+// UNGENERALIZED: [[TRAPMSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Control flow integrity check failed", scope: [[META11]], file: [[META11]], type: [[META36]], flags: DIFlagArtificial, spFlags: DISPFlagDefinition, unit: [[META0]])
+// UNGENERALIZED: [[DBG42]] = !DILocation(line: 54, column: 1, scope: [[DBG25]])
 //.
 // GENERALIZED: [[META0:![0-9]+]] = distinct !DICompileUnit(language: DW_LANG_C11, file: [[META1:![0-9]+]], isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, retainedTypes: [[META2:![0-9]+]], splitDebugInlining: false, nameTableKind: None)
 // GENERALIZED: [[META1]] = !DIFile(filename: "{{.*}}<stdin>", directory: {{.*}})
@@ -128,5 +130,7 @@ void g(int** (*fp)(const char *, const char **)) {
 // GENERALIZED: [[DBG37]] = !DILocation(line: 53, column: 3, scope: [[DBG25]])
 // GENERALIZED: [[META38]] = !{}
 // GENERALIZED: [[PROF39]] = !{!"branch_weights", i32 1048575, i32 1}
-// GENERALIZED: [[DBG40]] = !DILocation(line: 54, column: 1, scope: [[DBG25]])
+// GENERALIZED: [[DBGTRAP]] = !DILocation(line: 0, scope: [[TRAPMSG:![0-9]+]], inlinedAt: [[DBG34]])
+// GENERALIZED: [[TRAPMSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Control flow integrity check failed", scope: [[META11]], file: [[META11]], type: [[META36]], flags: DIFlagArtificial, spFlags: DISPFlagDefinition, unit: [[META0]])
+// GENERALIZED: [[DBG42]] = !DILocation(line: 54, column: 1, scope: [[DBG25]])
 //.
diff --git a/clang/test/CodeGen/cfi-icall-normalize2-debuginfo.c b/clang/test/CodeGen/cfi-icall-normalize2-debuginfo.c
index a2f6ee0c..258c3bf 100644
--- a/clang/test/CodeGen/cfi-icall-normalize2-debuginfo.c
+++ b/clang/test/CodeGen/cfi-icall-normalize2-debuginfo.c
@@ -15,50 +15,50 @@
 // CHECK-NEXT:    [[TMP0:%.*]] = tail call i1 @llvm.type.test(ptr [[FN]], metadata !"_ZTSFvu3i32E.normalized"), !dbg [[DBG21:![0-9]+]], !nosanitize [[META25:![0-9]+]]
 // CHECK-NEXT:    br i1 [[TMP0]], label %[[CONT:.*]], label %[[TRAP:.*]], !dbg [[DBG21]], !prof [[PROF26:![0-9]+]], !nosanitize [[META25]]
 // CHECK:       [[TRAP]]:
-// CHECK-NEXT:    tail call void @llvm.ubsantrap(i8 2) #[[ATTR3:[0-9]+]], !dbg [[DBG21]], !nosanitize [[META25]]
-// CHECK-NEXT:    unreachable, !dbg [[DBG21]], !nosanitize [[META25]]
+// CHECK-NEXT:    tail call void @llvm.ubsantrap(i8 2) #[[ATTR3:[0-9]+]], !dbg [[DBGTRAP:![0-9]+]], !nosanitize [[META25]]
+// CHECK-NEXT:    unreachable, !dbg [[DBGTRAP]], !nosanitize [[META25]]
 // CHECK:       [[CONT]]:
 // CHECK-NEXT:    tail call void [[FN]](i32 noundef [[ARG]]) #[[ATTR4:[0-9]+]], !dbg [[DBG24:![0-9]+]]
-// CHECK-NEXT:    ret void, !dbg [[DBG27:![0-9]+]]
+// CHECK-NEXT:    ret void, !dbg [[DBG29:![0-9]+]]
 //
 void foo(void (*fn)(int), int arg) {
     fn(arg);
 }
 
 // CHECK-LABEL: define dso_local void @bar(
-// CHECK-SAME: ptr noundef [[FN:%.*]], i32 noundef [[ARG1:%.*]], i32 noundef [[ARG2:%.*]]) local_unnamed_addr #[[ATTR0]] !dbg [[DBG28:![0-9]+]] !type [[META38:![0-9]+]] !type [[META39:![0-9]+]] {
+// CHECK-SAME: ptr noundef [[FN:%.*]], i32 noundef [[ARG1:%.*]], i32 noundef [[ARG2:%.*]]) local_unnamed_addr #[[ATTR0]] !dbg [[DBG30:![0-9]+]] !type [[META40:![0-9]+]] !type [[META41:![0-9]+]] {
 // CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:      #dbg_value(ptr [[FN]], [[META35:![0-9]+]], !DIExpression(), [[META40:![0-9]+]])
-// CHECK-NEXT:      #dbg_value(i32 [[ARG1]], [[META36:![0-9]+]], !DIExpression(), [[META40]])
-// CHECK-NEXT:      #dbg_value(i32 [[ARG2]], [[META37:![0-9]+]], !DIExpression(), [[META40]])
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call i1 @llvm.type.test(ptr [[FN]], metadata !"_ZTSFvu3i32S_E.normalized"), !dbg [[DBG41:![0-9]+]], !nosanitize [[META25]]
-// CHECK-NEXT:    br i1 [[TMP0]], label %[[CONT:.*]], label %[[TRAP:.*]], !dbg [[DBG41]], !prof [[PROF26]], !nosanitize [[META25]]
+// CHECK-NEXT:      #dbg_value(ptr [[FN]], [[META37:![0-9]+]], !DIExpression(), [[META42:![0-9]+]])
+// CHECK-NEXT:      #dbg_value(i32 [[ARG1]], [[META38:![0-9]+]], !DIExpression(), [[META42]])
+// CHECK-NEXT:      #dbg_value(i32 [[ARG2]], [[META39:![0-9]+]], !DIExpression(), [[META42]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call i1 @llvm.type.test(ptr [[FN]], metadata !"_ZTSFvu3i32S_E.normalized"), !dbg [[DBG43:![0-9]+]], !nosanitize [[META25]]
+// CHECK-NEXT:    br i1 [[TMP0]], label %[[CONT:.*]], label %[[TRAP:.*]], !dbg [[DBG43]], !prof [[PROF26]], !nosanitize [[META25]]
 // CHECK:       [[TRAP]]:
-// CHECK-NEXT:    tail call void @llvm.ubsantrap(i8 2) #[[ATTR3]], !dbg [[DBG41]], !nosanitize [[META25]]
-// CHECK-NEXT:    unreachable, !dbg [[DBG41]], !nosanitize [[META25]]
+// CHECK-NEXT:    tail call void @llvm.ubsantrap(i8 2) #[[ATTR3]], !dbg [[DBG45:![0-9]+]], !nosanitize [[META25]]
+// CHECK-NEXT:    unreachable, !dbg [[DBG45]], !nosanitize [[META25]]
 // CHECK:       [[CONT]]:
-// CHECK-NEXT:    tail call void [[FN]](i32 noundef [[ARG1]], i32 noundef [[ARG2]]) #[[ATTR4]], !dbg [[DBG42:![0-9]+]]
-// CHECK-NEXT:    ret void, !dbg [[DBG43:![0-9]+]]
+// CHECK-NEXT:    tail call void [[FN]](i32 noundef [[ARG1]], i32 noundef [[ARG2]]) #[[ATTR4]], !dbg [[DBG44:![0-9]+]]
+// CHECK-NEXT:    ret void, !dbg [[DBG46:![0-9]+]]
 //
 void bar(void (*fn)(int, int), int arg1, int arg2) {
     fn(arg1, arg2);
 }
 
 // CHECK-LABEL: define dso_local void @baz(
-// CHECK-SAME: ptr noundef [[FN:%.*]], i32 noundef [[ARG1:%.*]], i32 noundef [[ARG2:%.*]], i32 noundef [[ARG3:%.*]]) local_unnamed_addr #[[ATTR0]] !dbg [[DBG44:![0-9]+]] !type [[META55:![0-9]+]] !type [[META56:![0-9]+]] {
+// CHECK-SAME: ptr noundef [[FN:%.*]], i32 noundef [[ARG1:%.*]], i32 noundef [[ARG2:%.*]], i32 noundef [[ARG3:%.*]]) local_unnamed_addr #[[ATTR0]] !dbg [[DBG47:![0-9]+]] !type [[META58:![0-9]+]] !type [[META59:![0-9]+]] {
 // CHECK-NEXT:  [[ENTRY:.*:]]
-// CHECK-NEXT:      #dbg_value(ptr [[FN]], [[META51:![0-9]+]], !DIExpression(), [[META57:![0-9]+]])
-// CHECK-NEXT:      #dbg_value(i32 [[ARG1]], [[META52:![0-9]+]], !DIExpression(), [[META57]])
-// CHECK-NEXT:      #dbg_value(i32 [[ARG2]], [[META53:![0-9]+]], !DIExpression(), [[META57]])
-// CHECK-NEXT:      #dbg_value(i32 [[ARG3]], [[META54:![0-9]+]], !DIExpression(), [[META57]])
-// CHECK-NEXT:    [[TMP0:%.*]] = tail call i1 @llvm.type.test(ptr [[FN]], metadata !"_ZTSFvu3i32S_S_E.normalized"), !dbg [[DBG58:![0-9]+]], !nosanitize [[META25]]
-// CHECK-NEXT:    br i1 [[TMP0]], label %[[CONT:.*]], label %[[TRAP:.*]], !dbg [[DBG58]], !prof [[PROF26]], !nosanitize [[META25]]
+// CHECK-NEXT:      #dbg_value(ptr [[FN]], [[META54:![0-9]+]], !DIExpression(), [[META60:![0-9]+]])
+// CHECK-NEXT:      #dbg_value(i32 [[ARG1]], [[META55:![0-9]+]], !DIExpression(), [[META60]])
+// CHECK-NEXT:      #dbg_value(i32 [[ARG2]], [[META56:![0-9]+]], !DIExpression(), [[META60]])
+// CHECK-NEXT:      #dbg_value(i32 [[ARG3]], [[META57:![0-9]+]], !DIExpression(), [[META60]])
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call i1 @llvm.type.test(ptr [[FN]], metadata !"_ZTSFvu3i32S_S_E.normalized"), !dbg [[DBG61:![0-9]+]], !nosanitize [[META25]]
+// CHECK-NEXT:    br i1 [[TMP0]], label %[[CONT:.*]], label %[[TRAP:.*]], !dbg [[DBG61]], !prof [[PROF26]], !nosanitize [[META25]]
 // CHECK:       [[TRAP]]:
-// CHECK-NEXT:    tail call void @llvm.ubsantrap(i8 2) #[[ATTR3]], !dbg [[DBG58]], !nosanitize [[META25]]
-// CHECK-NEXT:    unreachable, !dbg [[DBG58]], !nosanitize [[META25]]
+// CHECK-NEXT:    tail call void @llvm.ubsantrap(i8 2) #[[ATTR3]], !dbg [[DBG63:![0-9]+]], !nosanitize [[META25]]
+// CHECK-NEXT:    unreachable, !dbg [[DBG63]], !nosanitize [[META25]]
 // CHECK:       [[CONT]]:
-// CHECK-NEXT:    tail call void [[FN]](i32 noundef [[ARG1]], i32 noundef [[ARG2]], i32 noundef [[ARG3]]) #[[ATTR4]], !dbg [[DBG59:![0-9]+]]
-// CHECK-NEXT:    ret void, !dbg [[DBG60:![0-9]+]]
+// CHECK-NEXT:    tail call void [[FN]](i32 noundef [[ARG1]], i32 noundef [[ARG2]], i32 noundef [[ARG3]]) #[[ATTR4]], !dbg [[DBG62:![0-9]+]]
+// CHECK-NEXT:    ret void, !dbg [[DBG64:![0-9]+]]
 //
 void baz(void (*fn)(int, int, int), int arg1, int arg2, int arg3) {
     fn(arg1, arg2, arg3);
@@ -87,38 +87,42 @@ void baz(void (*fn)(int, int, int), int arg1, int arg2, int arg3) {
 // CHECK: [[DBG24]] = !DILocation(line: 25, column: 5, scope: [[DBG7]])
 // CHECK: [[META25]] = !{}
 // CHECK: [[PROF26]] = !{!"branch_weights", i32 1048575, i32 1}
-// CHECK: [[DBG27]] = !DILocation(line: 26, column: 1, scope: [[DBG7]])
-// CHECK: [[DBG28]] = distinct !DISubprogram(name: "bar", scope: [[META8]], file: [[META8]], line: 43, type: [[META29:![0-9]+]], scopeLine: 43, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: [[META0]], retainedNodes: [[META34:![0-9]+]])
-// CHECK: [[META29]] = !DISubroutineType(types: [[META30:![0-9]+]])
-// CHECK: [[META30]] = !{null, [[META31:![0-9]+]], [[META14]], [[META14]]}
-// CHECK: [[META31]] = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: [[META32:![0-9]+]], size: 64)
-// CHECK: [[META32]] = !DISubroutineType(types: [[META33:![0-9]+]])
-// CHECK: [[META33]] = !{null, [[META14]], [[META14]]}
-// CHECK: [[META34]] = !{[[META35]], [[META36]], [[META37]]}
-// CHECK: [[META35]] = !DILocalVariable(name: "fn", arg: 1, scope: [[DBG28]], file: [[META8]], line: 43, type: [[META31]])
-// CHECK: [[META36]] = !DILocalVariable(name: "arg1", arg: 2, scope: [[DBG28]], file: [[META8]], line: 43, type: [[META14]])
-// CHECK: [[META37]] = !DILocalVariable(name: "arg2", arg: 3, scope: [[DBG28]], file: [[META8]], line: 43, type: [[META14]])
-// CHECK: [[META38]] = !{i64 0, !"_ZTSFvPFvu3i32S_ES_S_E.normalized"}
-// CHECK: [[META39]] = !{i64 0, !"_ZTSFvPvu3i32S0_E.normalized.generalized"}
-// CHECK: [[META40]] = !DILocation(line: 0, scope: [[DBG28]])
-// CHECK: [[DBG41]] = !DILocation(line: 0, scope: [[META22]], inlinedAt: [[DBG42]])
-// CHECK: [[DBG42]] = !DILocation(line: 44, column: 5, scope: [[DBG28]])
-// CHECK: [[DBG43]] = !DILocation(line: 45, column: 1, scope: [[DBG28]])
-// CHECK: [[DBG44]] = distinct !DISubprogram(name: "baz", scope: [[META8]], file: [[META8]], line: 63, type: [[META45:![0-9]+]], scopeLine: 63, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: [[META0]], retainedNodes: [[META50:![0-9]+]])
-// CHECK: [[META45]] = !DISubroutineType(types: [[META46:![0-9]+]])
-// CHECK: [[META46]] = !{null, [[META47:![0-9]+]], [[META14]], [[META14]], [[META14]]}
-// CHECK: [[META47]] = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: [[META48:![0-9]+]], size: 64)
+// CHECK: [[DBGTRAP]] = !DILocation(line: 0, scope: [[TRAPMSG:![0-9]+]], inlinedAt: [[DBG21]])
+// CHECK: [[TRAPMSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Control flow integrity check failed", scope: [[META8]], file: [[META8]], type: [[META23]], flags: DIFlagArtificial, spFlags: DISPFlagDefinition, unit: [[META0]])
+// CHECK: [[DBG29]] = !DILocation(line: 26, column: 1, scope: [[DBG7]])
+// CHECK: [[DBG30]] = distinct !DISubprogram(name: "bar", scope: [[META8]], file: [[META8]], line: 43, type: [[META31:![0-9]+]], scopeLine: 43, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: [[META0]], retainedNodes: [[META36:![0-9]+]])
+// CHECK: [[META31]] = !DISubroutineType(types: [[META32:![0-9]+]])
+// CHECK: [[META32]] = !{null, [[META33:![0-9]+]], [[META14]], [[META14]]}
+// CHECK: [[META33]] = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: [[META34:![0-9]+]], size: 64)
+// CHECK: [[META34]] = !DISubroutineType(types: [[META35:![0-9]+]])
+// CHECK: [[META35]] = !{null, [[META14]], [[META14]]}
+// CHECK: [[META36]] = !{[[META37]], [[META38]], [[META39]]}
+// CHECK: [[META37]] = !DILocalVariable(name: "fn", arg: 1, scope: [[DBG30]], file: [[META8]], line: 43, type: [[META33]])
+// CHECK: [[META38]] = !DILocalVariable(name: "arg1", arg: 2, scope: [[DBG30]], file: [[META8]], line: 43, type: [[META14]])
+// CHECK: [[META39]] = !DILocalVariable(name: "arg2", arg: 3, scope: [[DBG30]], file: [[META8]], line: 43, type: [[META14]])
+// CHECK: [[META40]] = !{i64 0, !"_ZTSFvPFvu3i32S_ES_S_E.normalized"}
+// CHECK: [[META41]] = !{i64 0, !"_ZTSFvPvu3i32S0_E.normalized.generalized"}
+// CHECK: [[META42]] = !DILocation(line: 0, scope: [[DBG30]])
+// CHECK: [[DBG43]] = !DILocation(line: 0, scope: [[META22]], inlinedAt: [[DBG44]])
+// CHECK: [[DBG44]] = !DILocation(line: 44, column: 5, scope: [[DBG30]])
+// CHECK: [[DBG45]] = !DILocation(line: 0, scope: [[TRAPMSG]], inlinedAt: [[DBG43]])
+// CHECK: [[DBG46]] = !DILocation(line: 45, column: 1, scope: [[DBG30]])
+// CHECK: [[DBG47]] = distinct !DISubprogram(name: "baz", scope: [[META8]], file: [[META8]], line: 63, type: [[META48:![0-9]+]], scopeLine: 63, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: [[META0]], retainedNodes: [[META53:![0-9]+]])
 // CHECK: [[META48]] = !DISubroutineType(types: [[META49:![0-9]+]])
-// CHECK: [[META49]] = !{null, [[META14]], [[META14]], [[META14]]}
-// CHECK: [[META50]] = !{[[META51]], [[META52]], [[META53]], [[META54]]}
-// CHECK: [[META51]] = !DILocalVariable(name: "fn", arg: 1, scope: [[DBG44]], file: [[META8]], line: 63, type: [[META47]])
-// CHECK: [[META52]] = !DILocalVariable(name: "arg1", arg: 2, scope: [[DBG44]], file: [[META8]], line: 63, type: [[META14]])
-// CHECK: [[META53]] = !DILocalVariable(name: "arg2", arg: 3, scope: [[DBG44]], file: [[META8]], line: 63, type: [[META14]])
-// CHECK: [[META54]] = !DILocalVariable(name: "arg3", arg: 4, scope: [[DBG44]], file: [[META8]], line: 63, type: [[META14]])
-// CHECK: [[META55]] = !{i64 0, !"_ZTSFvPFvu3i32S_S_ES_S_S_E.normalized"}
-// CHECK: [[META56]] = !{i64 0, !"_ZTSFvPvu3i32S0_S0_E.normalized.generalized"}
-// CHECK: [[META57]] = !DILocation(line: 0, scope: [[DBG44]])
-// CHECK: [[DBG58]] = !DILocation(line: 0, scope: [[META22]], inlinedAt: [[DBG59]])
-// CHECK: [[DBG59]] = !DILocation(line: 64, column: 5, scope: [[DBG44]])
-// CHECK: [[DBG60]] = !DILocation(line: 65, column: 1, scope: [[DBG44]])
+// CHECK: [[META49]] = !{null, [[META50:![0-9]+]], [[META14]], [[META14]], [[META14]]}
+// CHECK: [[META50]] = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: [[META51:![0-9]+]], size: 64)
+// CHECK: [[META51]] = !DISubroutineType(types: [[META52:![0-9]+]])
+// CHECK: [[META52]] = !{null, [[META14]], [[META14]], [[META14]]}
+// CHECK: [[META53]] = !{[[META54]], [[META55]], [[META56]], [[META57]]}
+// CHECK: [[META54]] = !DILocalVariable(name: "fn", arg: 1, scope: [[DBG47]], file: [[META8]], line: 63, type: [[META50]])
+// CHECK: [[META55]] = !DILocalVariable(name: "arg1", arg: 2, scope: [[DBG47]], file: [[META8]], line: 63, type: [[META14]])
+// CHECK: [[META56]] = !DILocalVariable(name: "arg2", arg: 3, scope: [[DBG47]], file: [[META8]], line: 63, type: [[META14]])
+// CHECK: [[META57]] = !DILocalVariable(name: "arg3", arg: 4, scope: [[DBG47]], file: [[META8]], line: 63, type: [[META14]])
+// CHECK: [[META58]] = !{i64 0, !"_ZTSFvPFvu3i32S_S_ES_S_S_E.normalized"}
+// CHECK: [[META59]] = !{i64 0, !"_ZTSFvPvu3i32S0_S0_E.normalized.generalized"}
+// CHECK: [[META60]] = !DILocation(line: 0, scope: [[DBG47]])
+// CHECK: [[DBG61]] = !DILocation(line: 0, scope: [[META22]], inlinedAt: [[DBG62]])
+// CHECK: [[DBG62]] = !DILocation(line: 64, column: 5, scope: [[DBG47]])
+// CHECK: [[DBG63]] = !DILocation(line: 0, scope: [[TRAPMSG]], inlinedAt: [[DBG61]])
+// CHECK: [[DBG64]] = !DILocation(line: 65, column: 1, scope: [[DBG47]])
 //.
diff --git a/clang/test/CodeGen/ubsan-trap-debugloc.c b/clang/test/CodeGen/ubsan-trap-debugloc.c
index 87cbfad..2f5258a 100644
--- a/clang/test/CodeGen/ubsan-trap-debugloc.c
+++ b/clang/test/CodeGen/ubsan-trap-debugloc.c
@@ -20,5 +20,8 @@ void bar(volatile int a) __attribute__((optnone)) {
 // CHECK: [[LOC]] = !DILocation(line: 0
 
 // With optimisations disabled the traps are not merged and retain accurate debug locations
-// CHECK: [[LOC2]] = !DILocation(line: 15, column: 9
-// CHECK: [[LOC3]] = !DILocation(line: 16, column: 9
+ // CHECK-DAG: [[SRC2:![0-9]+]] = !DILocation(line: 15, column: 9,
+ // CHECK-DAG: [[SRC3:![0-9]+]] = !DILocation(line: 16, column: 9,
+ // CHECK-DAG: [[LOC2]] = !DILocation(line: 0, scope: [[SCOPE2:![0-9]+]], inlinedAt: [[SRC2]])
+ // CHECK-DAG: [[LOC3]] = !DILocation(line: 0, scope: [[SCOPE3:![0-9]+]], inlinedAt: [[SRC3]])
+
diff --git a/clang/test/CodeGen/ubsan-trap-reason-add-overflow.c b/clang/test/CodeGen/ubsan-trap-reason-add-overflow.c
new file mode 100644
index 0000000..225778d
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-add-overflow.c
@@ -0,0 +1,9 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=signed-integer-overflow -fsanitize-trap=signed-integer-overflow -emit-llvm %s -o - | FileCheck %s
+
+int add_overflow(int a, int b) { return a + b; }
+
+// CHECK-LABEL: @add_overflow
+// CHECK: call void @llvm.ubsantrap(i8 0) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Integer addition overflowed"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-alignment-assumption.c b/clang/test/CodeGen/ubsan-trap-reason-alignment-assumption.c
new file mode 100644
index 0000000..3247ceb
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-alignment-assumption.c
@@ -0,0 +1,15 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=alignment -fsanitize-trap=alignment -emit-llvm %s -o - | FileCheck %s
+
+#include <stdint.h>
+int32_t *get_int(void) __attribute__((assume_aligned(16)));
+
+void retrieve_int(void) {
+  int *i = get_int();
+  *i = 7;
+}
+
+// CHECK-LABEL: @retrieve_int
+// CHECK: call void @llvm.ubsantrap(i8 23) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Alignment assumption violated"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-builtin-unreachable.c b/clang/test/CodeGen/ubsan-trap-reason-builtin-unreachable.c
new file mode 100644
index 0000000..97bd690
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-builtin-unreachable.c
@@ -0,0 +1,9 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=unreachable -fsanitize-trap=unreachable -emit-llvm %s -o - | FileCheck %s
+
+int call_builtin_unreachable(void) { __builtin_unreachable(); }
+
+// CHECK-LABEL: @call_builtin_unreachable
+// CHECK: call void @llvm.ubsantrap(i8 1) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$_builtin_unreachable(), execution reached an unreachable program point"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-cfi-check-fail.c b/clang/test/CodeGen/ubsan-trap-reason-cfi-check-fail.c
new file mode 100644
index 0000000..9304f51
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-cfi-check-fail.c
@@ -0,0 +1,25 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=cfi-icall -fsanitize-trap=cfi-icall -emit-llvm %s -o - | FileCheck %s
+
+typedef int (*fp_t)(int);
+
+int good(int x) { return x + 1; }
+
+int bad(void) { return 0; }
+
+int cfi_trigger(int a) {
+  fp_t p = good;
+  int r1 = p(a);
+
+  p = (fp_t)(void *)bad;
+  int r2 = p(a);
+
+  return r1 + r2;
+}
+
+// CHECK-LABEL: @good
+// CHECK-LABEL: @bad
+// CHECK-LABEL: @cfi_trigger
+// CHECK: call void @llvm.ubsantrap(i8 2) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Control flow integrity check failed"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-crash.cpp b/clang/test/CodeGen/ubsan-trap-reason-crash.cpp
new file mode 100644
index 0000000..6add9bf
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-crash.cpp
@@ -0,0 +1,20 @@
+// FIXME: We should emit a trap message for this case too.
+// But sometimes Clang will emit a ubsan trap into the prologue of a function,
+// at which point the debug-info locations haven't been set up yet and
+// can't hook up our artificial inline frame. [Issue #150707]
+
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=null -fsanitize-trap=null -emit-llvm %s -o - | FileCheck %s
+
+struct Foo {
+  void target() {}
+} f;
+
+void caller() {
+  f.target();
+}
+
+
+// CHECK-LABEL: @_Z6callerv
+// CHECK: call void @llvm.ubsantrap(i8 22){{.*}}!nosanitize
+// CHECK-NOT: __clang_trap_msg
diff --git a/clang/test/CodeGen/ubsan-trap-reason-div-rem-overflow.c b/clang/test/CodeGen/ubsan-trap-reason-div-rem-overflow.c
new file mode 100644
index 0000000..d0b21dd
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-div-rem-overflow.c
@@ -0,0 +1,9 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=signed-integer-overflow -fsanitize-trap=signed-integer-overflow -emit-llvm %s -o - | FileCheck %s
+
+int div_rem_overflow(int a, int b) { return a / b; }
+
+// CHECK-LABEL: @div_rem_overflow
+// CHECK: call void @llvm.ubsantrap(i8 3) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Integer divide or remainder overflowed"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-dynamic-type-cache-miss.cpp b/clang/test/CodeGen/ubsan-trap-reason-dynamic-type-cache-miss.cpp
new file mode 100644
index 0000000..f89fbdcf
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-dynamic-type-cache-miss.cpp
@@ -0,0 +1,26 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=vptr -fsanitize-trap=vptr -emit-llvm %s -o - | FileCheck %s
+
+struct A {
+  virtual void foo();
+};
+struct B {
+  virtual void bar();
+};
+
+void A::foo() {}
+void B::bar() {}
+
+int dynamic_type_cache_miss() {
+  B b;
+  A &a = reinterpret_cast<A &>(b);
+  a.foo();
+  return 0;
+}
+
+// CHECK-LABEL: @_ZN1A3fooEv
+// CHECK-LABEL: @_ZN1B3barEv
+// CHECK-LABEL: @_Z23dynamic_type_cache_missv
+// CHECK: call void @llvm.ubsantrap(i8 4) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Dynamic type cache miss, member call made on an object whose dynamic type differs from the expected type"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-flag.c b/clang/test/CodeGen/ubsan-trap-reason-flag.c
new file mode 100644
index 0000000..5cc16d1
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-flag.c
@@ -0,0 +1,22 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=signed-integer-overflow -fsanitize-trap=signed-integer-overflow -emit-llvm %s -o - \
+// RUN: | FileCheck %s --check-prefix=ANNOTATE
+
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=signed-integer-overflow -fsanitize-trap=signed-integer-overflow \
+// RUN: -fsanitize-debug-trap-reasons -emit-llvm %s -o - | FileCheck %s --check-prefix=ANNOTATE
+
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=signed-integer-overflow -fsanitize-trap=signed-integer-overflow \
+// RUN: -fno-sanitize-debug-trap-reasons -emit-llvm %s -o - | FileCheck %s --check-prefix=NO-ANNOTATE
+
+int add_overflow(int a, int b) { return a + b; }
+
+// ANNOTATE-LABEL: @add_overflow
+// ANNOTATE: call void @llvm.ubsantrap(i8 0) {{.*}}!dbg [[LOC:![0-9]+]]
+// ANNOTATE: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// ANNOTATE: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Integer addition overflowed"
+
+// NO-ANNOTATE-LABEL: @add_overflow
+// NO-ANNOTATE: call void @llvm.ubsantrap(i8 0) {{.*}}!dbg [[LOC:![0-9]+]]
+// NO-ANNOTATE-NOT: __clang_trap_msg
diff --git a/clang/test/CodeGen/ubsan-trap-reason-float-cast-overflow.c b/clang/test/CodeGen/ubsan-trap-reason-float-cast-overflow.c
new file mode 100644
index 0000000..079a191e
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-float-cast-overflow.c
@@ -0,0 +1,9 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=float-cast-overflow -fsanitize-trap=float-cast-overflow -emit-llvm %s -o - | FileCheck %s
+
+int float_cast_overflow(float x) { return (int)x; }
+
+// CHECK-LABEL: @float_cast_overflow
+// CHECK: call void @llvm.ubsantrap(i8 5) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Floating-point to integer conversion overflowed"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-function-type-mismatch.c b/clang/test/CodeGen/ubsan-trap-reason-function-type-mismatch.c
new file mode 100644
index 0000000..1727f9c
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-function-type-mismatch.c
@@ -0,0 +1,18 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=function -fsanitize-trap=function -emit-llvm %s -o - | FileCheck %s
+
+void target(void) {}
+
+int function_type_mismatch(void) {
+  int (*fp_int)(int);
+
+  fp_int = (int (*)(int))(void *)target;
+
+  return fp_int(42);
+}
+
+// CHECK-LABEL: @target
+// CHECK-LABEL: @function_type_mismatch
+// CHECK: call void @llvm.ubsantrap(i8 6) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Function called with mismatched signature"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-implicit-conversion.c b/clang/test/CodeGen/ubsan-trap-reason-implicit-conversion.c
new file mode 100644
index 0000000..43c091d
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-implicit-conversion.c
@@ -0,0 +1,11 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=implicit-unsigned-integer-truncation -fsanitize-trap=implicit-unsigned-integer-truncation -emit-llvm %s -o - | FileCheck %s
+
+unsigned long long big;
+
+unsigned implicit_conversion(void) { return big; }
+
+// CHECK-LABEL: @implicit_conversion
+// CHECK: call void @llvm.ubsantrap(i8 7) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Implicit integer conversion overflowed or lost data"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-invalid-builtin.c b/clang/test/CodeGen/ubsan-trap-reason-invalid-builtin.c
new file mode 100644
index 0000000..56cf674
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-invalid-builtin.c
@@ -0,0 +1,9 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=builtin -fsanitize-trap=builtin -emit-llvm %s -o - | FileCheck %s
+
+unsigned invalid_builtin(unsigned x) { return __builtin_clz(x); }
+
+// CHECK-LABEL: @invalid_builtin
+// CHECK: call void @llvm.ubsantrap(i8 8) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Invalid use of builtin function"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-invalid-objc-cast.m b/clang/test/CodeGen/ubsan-trap-reason-invalid-objc-cast.m
new file mode 100644
index 0000000..ed2d5ff
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-invalid-objc-cast.m
@@ -0,0 +1,32 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=objc-cast -fsanitize-trap=objc-cast -emit-llvm %s -o - | FileCheck %s
+
+@interface NSFastEnumerationState
+@end
+
+#define NSUInteger unsigned int
+
+@interface NSArray
++(NSArray*) arrayWithObjects: (id) first, ...;
+- (NSUInteger) countByEnumeratingWithState:(NSFastEnumerationState *) state 
+                                   objects:(id[]) buffer 
+                                     count:(NSUInteger) len;
+-(unsigned) count;
+@end
+@interface NSString
+-(const char*) cString;
+@end
+
+void receive_NSString(NSString*);
+
+void t0(void) {
+  NSArray *array = [NSArray arrayWithObjects: @"0", @"1", (void*)0];
+  for (NSString *i in array) {
+    receive_NSString(i);
+  }
+}
+
+// CHECK-LABEL: @t0
+// CHECK: call void @llvm.ubsantrap(i8 9) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Invalid Objective-C cast"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-load-invalid-value.c b/clang/test/CodeGen/ubsan-trap-reason-load-invalid-value.c
new file mode 100644
index 0000000..4aad832
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-load-invalid-value.c
@@ -0,0 +1,13 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=bool -fsanitize-trap=bool -emit-llvm %s -o - | FileCheck %s
+
+#include <stdbool.h>
+
+unsigned char bad_byte;
+
+bool load_invalid_value(void) { return *((bool *)&bad_byte); }
+
+// CHECK-LABEL: @load_invalid_value
+// CHECK: call void @llvm.ubsantrap(i8 10) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Loaded an invalid or uninitialized value for the type"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-missing-return.cpp b/clang/test/CodeGen/ubsan-trap-reason-missing-return.cpp
new file mode 100644
index 0000000..2818d9d
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-missing-return.cpp
@@ -0,0 +1,12 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=return -fsanitize-trap=return -emit-llvm %s -o - | FileCheck %s
+
+int missing_return(int x) {
+  if (x > 0)
+    return x;
+}
+
+// CHECK-LABEL: @_Z14missing_return
+// CHECK: call void @llvm.ubsantrap(i8 11) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Execution reached the end of a value-returning function without returning a value"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-mul-overflow.c b/clang/test/CodeGen/ubsan-trap-reason-mul-overflow.c
new file mode 100644
index 0000000..cf9a0b4
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-mul-overflow.c
@@ -0,0 +1,9 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=signed-integer-overflow -fsanitize-trap=signed-integer-overflow -emit-llvm %s -o - | FileCheck %s
+
+int mul_overflow(int a, int b) { return a * b; }
+
+// CHECK-LABEL: @mul_overflow
+// CHECK: call void @llvm.ubsantrap(i8 12) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Integer multiplication overflowed"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-negate-overflow.c b/clang/test/CodeGen/ubsan-trap-reason-negate-overflow.c
new file mode 100644
index 0000000..5534679
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-negate-overflow.c
@@ -0,0 +1,12 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=signed-integer-overflow -fsanitize-trap=signed-integer-overflow -emit-llvm %s -o - | FileCheck %s
+
+int negate_overflow() {
+  int x;
+  return -x;
+}
+
+// CHECK-LABEL: @negate_overflow
+// CHECK: call void @llvm.ubsantrap(i8 13) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Integer negation overflowed"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-nonnull-arg.c b/clang/test/CodeGen/ubsan-trap-reason-nonnull-arg.c
new file mode 100644
index 0000000..1f0f450
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-nonnull-arg.c
@@ -0,0 +1,12 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=nonnull-attribute -fsanitize-trap=nonnull-attribute -emit-llvm %s -o - | FileCheck %s
+
+__attribute__((nonnull)) void nonnull_arg(int *p) { (void)p; }
+
+void trigger_nonnull_arg() { nonnull_arg(0); }
+
+// CHECK-LABEL: @nonnull_arg
+// CHECK-LABEL: @trigger_nonnull_arg
+// CHECK: call void @llvm.ubsantrap(i8 16) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Passing null pointer as an argument which is declared to never be null"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-nonnull-return.c b/clang/test/CodeGen/ubsan-trap-reason-nonnull-return.c
new file mode 100644
index 0000000..1197b4d
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-nonnull-return.c
@@ -0,0 +1,14 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=returns-nonnull-attribute -fsanitize-trap=returns-nonnull-attribute -emit-llvm %s -o - | FileCheck %s
+
+__attribute__((returns_nonnull)) int *must_return_nonnull(int bad) {
+  if (bad)
+    return 0;
+  static int x = 1;
+  return &x;
+}
+
+// CHECK-LABEL: @must_return_nonnull
+// CHECK: call void @llvm.ubsantrap(i8 17) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Returning null pointer from a function which is declared to never return null"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-nullability-arg.c b/clang/test/CodeGen/ubsan-trap-reason-nullability-arg.c
new file mode 100644
index 0000000..2bc71de
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-nullability-arg.c
@@ -0,0 +1,14 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=nullability-arg -fsanitize-trap=nullability-arg -emit-llvm %s -o - | FileCheck %s
+
+#include <stddef.h>
+
+int nullability_arg(int *_Nonnull p) { return *p; }
+
+int trigger_nullability_arg(void) { return nullability_arg(NULL); }
+
+// CHECK-LABEL: @nullability_arg
+// CHECK-LABEL: @trigger_nullability_arg
+// CHECK: call void @llvm.ubsantrap(i8 14) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Passing null as an argument which is annotated with _Nonnull"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-nullability-return.c b/clang/test/CodeGen/ubsan-trap-reason-nullability-return.c
new file mode 100644
index 0000000..3d64c5a
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-nullability-return.c
@@ -0,0 +1,18 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=nullability-return -fsanitize-trap=nullability-return -emit-llvm %s -o - | FileCheck %s
+
+#include <stdbool.h>
+#include <stddef.h>
+
+int *_Nonnull nullability_return(bool fail) {
+  if (fail)
+    return NULL;
+
+  static int x = 0;
+  return &x;
+}
+
+// CHECK-LABEL: @nullability_return
+// CHECK: call void @llvm.ubsantrap(i8 15) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Returning null from a function with a return type annotated with _Nonnull"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-out-of-bounds.c b/clang/test/CodeGen/ubsan-trap-reason-out-of-bounds.c
new file mode 100644
index 0000000..979886d
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-out-of-bounds.c
@@ -0,0 +1,12 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=array-bounds -fsanitize-trap=array-bounds -emit-llvm %s -o - | FileCheck %s
+
+int out_of_bounds() {
+  int a[1] = {0};
+  return a[1];
+}
+
+// CHECK-LABEL: @out_of_bounds
+// CHECK: call void @llvm.ubsantrap(i8 18) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Array index out of bounds"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-pointer-overflow.c b/clang/test/CodeGen/ubsan-trap-reason-pointer-overflow.c
new file mode 100644
index 0000000..41cb487
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-pointer-overflow.c
@@ -0,0 +1,16 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=pointer-overflow -fsanitize-trap=pointer-overflow -emit-llvm %s -o - | FileCheck %s
+
+#include <stddef.h>
+#include <stdint.h>
+
+int *pointer_overflow(void) {
+  int buf[4];
+  volatile size_t n = (SIZE_MAX / sizeof(int)) - 1;
+  return buf + n;
+}
+
+// CHECK-LABEL: @pointer_overflow
+// CHECK: call void @llvm.ubsantrap(i8 19) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Pointer arithmetic overflowed bounds"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-shift-out-of-bounds.c b/clang/test/CodeGen/ubsan-trap-reason-shift-out-of-bounds.c
new file mode 100644
index 0000000..1a7465d
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-shift-out-of-bounds.c
@@ -0,0 +1,12 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=shift-base -fsanitize-trap=shift-base -emit-llvm %s -o - | FileCheck %s
+
+int shift_out_of_bounds(void) {
+  int sh = 32;
+  return 1 << sh;
+}
+
+// CHECK-LABEL: @shift_out_of_bounds
+// CHECK: call void @llvm.ubsantrap(i8 20) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Shift exponent is too large for the type"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-sub-overflow.c b/clang/test/CodeGen/ubsan-trap-reason-sub-overflow.c
new file mode 100644
index 0000000..62aa7fc
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-sub-overflow.c
@@ -0,0 +1,9 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=signed-integer-overflow -fsanitize-trap=signed-integer-overflow -emit-llvm %s -o - | FileCheck %s
+
+int sub_overflow(int a, int b) { return a - b; }
+
+// CHECK-LABEL: @sub_overflow
+// CHECK: call void @llvm.ubsantrap(i8 21) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Integer subtraction overflowed"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-type-mismatch.c b/clang/test/CodeGen/ubsan-trap-reason-type-mismatch.c
new file mode 100644
index 0000000..802ec91
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-type-mismatch.c
@@ -0,0 +1,9 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=alignment -fsanitize-trap=alignment -emit-llvm %s -o - | FileCheck %s
+
+int type_mismatch(int *p) { return *p; }
+
+// CHECK-LABEL: @type_mismatch
+// CHECK: call void @llvm.ubsantrap(i8 22) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Type mismatch in operation"
diff --git a/clang/test/CodeGen/ubsan-trap-reason-vla-bound-not-positive.c b/clang/test/CodeGen/ubsan-trap-reason-vla-bound-not-positive.c
new file mode 100644
index 0000000..ad9c408
--- /dev/null
+++ b/clang/test/CodeGen/ubsan-trap-reason-vla-bound-not-positive.c
@@ -0,0 +1,14 @@
+// RUN: %clang_cc1 -triple arm64-apple-macosx14.0.0 -O0 -debug-info-kind=standalone -dwarf-version=5 \
+// RUN: -fsanitize=vla-bound -fsanitize-trap=vla-bound -emit-llvm %s -o - | FileCheck %s
+
+int n = 0;
+
+int vla_bound_not_positive(void) {
+  int a[n];
+  return sizeof a;
+}
+
+// CHECK-LABEL: @vla_bound_not_positive
+// CHECK: call void @llvm.ubsantrap(i8 24) {{.*}}!dbg [[LOC:![0-9]+]]
+// CHECK: [[LOC]] = !DILocation(line: 0, scope: [[MSG:![0-9]+]], {{.+}})
+// CHECK: [[MSG]] = distinct !DISubprogram(name: "__clang_trap_msg$Undefined Behavior Sanitizer$Variable length array bound evaluates to non-positive value"
diff --git a/clang/test/Driver/modules.mm b/clang/test/Driver/modules.mm
index d1536c7..f0b0669 100644
--- a/clang/test/Driver/modules.mm
+++ b/clang/test/Driver/modules.mm
@@ -3,6 +3,9 @@
 // RUN: %clang -fmodules -fno-cxx-modules -### %s 2>&1 | FileCheck -check-prefix=CHECK-NO-MODULES %s
 // CHECK-NO-MODULES-NOT: -fmodules
 
+// RUN: %clang -std=c++20 -fno-cxx-modules -### %s 2>&1 | FileCheck -check-prefix=CHECK-NO-CPP-20-MODULES %s
+// CHECK-NO-CPP-20-MODULES: -fno-cxx-modules
+
 // RUN: %clang -fmodules -### %s 2>&1 | FileCheck -check-prefix=CHECK-HAS-MODULES %s
 // RUN: %clang -fmodules -fno-cxx-modules -fcxx-modules -### %s 2>&1 | FileCheck -check-prefix=CHECK-HAS-MODULES %s
 // CHECK-HAS-MODULES: -fmodules
diff --git a/clang/test/Modules/Exposure-2.cppm b/clang/test/Modules/Exposure-2.cppm
new file mode 100644
index 0000000..c09b739
--- /dev/null
+++ b/clang/test/Modules/Exposure-2.cppm
@@ -0,0 +1,27 @@
+// RUN: rm -rf %t
+// RUN: mkdir -p %t
+// RUN: split-file %s %t
+//
+// RUN: %clang_cc1 -std=c++20 -emit-reduced-module-interface %t/A.cppm -o %t/A.pcm
+// RUN: %clang_cc1 -std=c++20 %t/A.cpp -fmodule-file=A=%t/A.pcm -fsyntax-only -verify
+
+//--- A.cppm
+export module A;
+export template <class T>
+class C {};
+
+export template <class T>
+void foo() {
+    C<T> value;
+    (void) value;
+}
+
+//--- A.cpp
+// expected-no-diagnostics
+import A;
+namespace {
+class Local {};
+}
+void test() {
+    foo<Local>();
+}
diff --git a/clang/test/Modules/Exposure.cppm b/clang/test/Modules/Exposure.cppm
new file mode 100644
index 0000000..651a89e
--- /dev/null
+++ b/clang/test/Modules/Exposure.cppm
@@ -0,0 +1,25 @@
+// RUN: %clang_cc1 -std=c++20 %s -verify -fsyntax-only
+export module M;
+namespace {
+class TULocalClass {};
+}
+
+template <typename T>
+class Templ {};
+
+class C {
+    TULocalClass foo() { return TULocalClass(); } // expected-warning {{TU local entity 'TULocalClass' is exposed}}
+private:
+    TULocalClass Member; // expected-warning {{TU local entity 'TULocalClass' is exposed}}
+};
+
+static inline void its() {}
+template<int> void g() { its(); }
+
+void f0() {
+    g<1>();
+}
+
+inline void f1() {
+    g<1>(); // expected-warning {{TU local entity 'g<1>' is exposed}}
+}
diff --git a/clang/test/Modules/specializations-lazy-load-parentmap-crash.cpp b/clang/test/Modules/specializations-lazy-load-parentmap-crash.cpp
new file mode 100644
index 0000000..bd07ada
--- /dev/null
+++ b/clang/test/Modules/specializations-lazy-load-parentmap-crash.cpp
@@ -0,0 +1,99 @@
+// RUN: rm -rf %t
+// RUN: mkdir -p %t
+// RUN: split-file --leading-lines %s %t
+//
+// Prepare the BMIs.
+// RUN: %clang_cc1 -std=c++20 -triple x86_64-unknown-linux-gnu -emit-module-interface -o %t/mod_a-part1.pcm %t/mod_a-part1.cppm
+// RUN: %clang_cc1 -std=c++20 -triple x86_64-unknown-linux-gnu -emit-module-interface -o %t/mod_a-part2.pcm %t/mod_a-part2.cppm
+// RUN: %clang_cc1 -std=c++20 -triple x86_64-unknown-linux-gnu -emit-module-interface -o %t/mod_a.pcm %t/mod_a.cppm -fmodule-file=mod_a:part2=%t/mod_a-part2.pcm -fmodule-file=mod_a:part1=%t/mod_a-part1.pcm
+// RUN: %clang_cc1 -std=c++20 -triple x86_64-unknown-linux-gnu -emit-module-interface -o %t/mod_b.pcm %t/mod_b.cppm -fmodule-file=mod_a:part2=%t/mod_a-part2.pcm -fmodule-file=mod_a=%t/mod_a.pcm -fmodule-file=mod_a:part1=%t/mod_a-part1.pcm
+
+// Below are two examples to trigger the construction of the parent map (which is necessary to trigger the bug this regression test is for).
+// Using ArrayBoundV2 checker:
+// RUN: %clang_cc1 -std=c++20 -triple x86_64-unknown-linux-gnu -analyze -analyzer-checker=security,alpha.security -analyzer-output=text %t/test-array-bound-v2.cpp -fmodule-file=mod_a:part2=%t/mod_a-part2.pcm -fmodule-file=mod_a=%t/mod_a.pcm -fmodule-file=mod_a:part1=%t/mod_a-part1.pcm -fmodule-file=mod_b=%t/mod_b.pcm
+// Using a sanitized build:
+// RUN: %clang_cc1 -std=c++20 -triple x86_64-unknown-linux-gnu -fsanitize=unsigned-integer-overflow -fsanitize-undefined-ignore-overflow-pattern=all -emit-llvm -o %t/ignored %t/test-sanitized-build.cpp -fmodule-file=mod_a:part2=%t/mod_a-part2.pcm -fmodule-file=mod_a=%t/mod_a.pcm -fmodule-file=mod_a:part1=%t/mod_a-part1.pcm -fmodule-file=mod_b=%t/mod_b.pcm
+
+//--- mod_a-part1.cppm
+module;
+namespace mod_a {
+template <int> struct Important;
+}
+
+namespace mod_a {
+Important<0>& instantiate1();
+} // namespace mod_a
+export module mod_a:part1;
+
+export namespace mod_a {
+using ::mod_a::instantiate1;
+}
+
+//--- mod_a-part2.cppm
+module;
+namespace mod_a {
+template <int> struct Important;
+}
+
+namespace mod_a {
+template <int N> Important<N>& instantiate2();
+namespace part2InternalInstantiations {
+// During the construction of the parent map, we iterate over ClassTemplateDecl::specializations() for 'Important'.
+// After GH119333, the following instantiations get loaded between the call to spec_begin() and spec_end().
+// This used to invalidate the begin iterator returned by spec_begin() by the time the end iterator is returned.
+// This is a regression test for that.
+Important<1> fn1();
+Important<2> fn2();
+Important<3> fn3();
+Important<4> fn4();
+Important<5> fn5();
+Important<6> fn6();
+Important<7> fn7();
+Important<8> fn8();
+Important<9> fn9();
+Important<10> fn10();
+Important<11> fn11();
+}
+} // namespace mod_a
+export module mod_a:part2;
+
+export namespace mod_a {
+using ::mod_a::instantiate2;
+}
+
+//--- mod_a.cppm
+export module mod_a;
+export import :part1;
+export import :part2;
+
+//--- mod_b.cppm
+export module mod_b;
+import mod_a;
+
+void a() {
+  mod_a::instantiate1();
+  mod_a::instantiate2<42>();
+}
+
+//--- test-array-bound-v2.cpp
+import mod_b;
+
+extern void someFunc(char* first, char* last);
+void triggerParentMapContextCreationThroughArrayBoundV2() {
+  // This code currently causes the ArrayBoundV2 checker to create the ParentMapContext.
+  // Once it detects an access to buf[100], the checker looks through the parents to find '&' operator.
+  // (this is needed since taking the address of past-the-end pointer is allowed by the checker)
+  char buf[100];
+  someFunc(&buf[0], &buf[100]);
+}
+
+//--- test-sanitized-build.cpp
+import mod_b;
+
+extern void some();
+void triggerParentMapContextCreationThroughSanitizedBuild(unsigned i) {
+  // This code currently causes UBSan to create the ParentMapContext.
+  // UBSan currently excludes the pattern below to avoid noise, and it relies on ParentMapContext to detect it.
+  while (i--)
+    some();
+}
diff --git a/clang/test/Parser/dep_template_spec_type.cpp b/clang/test/Parser/dep_template_spec_type.cpp
new file mode 100644
index 0000000..65544bb
--- /dev/null
+++ b/clang/test/Parser/dep_template_spec_type.cpp
@@ -0,0 +1,16 @@
+// RUN: seq 100 | xargs -Ifoo %clang_cc1 -fsyntax-only -verify %s
+// expected-no-diagnostics
+// This is a regression test for a non-deterministic stack-overflow.
+
+template <typename C, typename S1, int rbits>
+typename C::A Bar(const S1& x, const C& c = C()) {
+    using T = typename C::A;
+    T result;
+
+    using PreC = typename C::template boop<T::p + rbits>;
+    using ExactC = typename C::template bap<PreC::p + 2>;
+
+    using D = typename ExactC::A;
+
+    return result;
+}
diff --git a/clang/test/Preprocessor/riscv-target-features-cv.c b/clang/test/Preprocessor/riscv-target-features-cv.c
new file mode 100644
index 0000000..a424a34
--- /dev/null
+++ b/clang/test/Preprocessor/riscv-target-features-cv.c
@@ -0,0 +1,60 @@
+// RUN: %clang --target=riscv32-unknown-linux-gnu -march=rv32i -E -dM %s \
+// RUN:   -o - | FileCheck %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu -march=rv64i -E -dM %s \
+// RUN:   -o - | FileCheck %s
+
+// CHECK-NOT: __riscv_xcvalu {{.*$}}
+// CHECK-NOT: __riscv_xcvbi {{.*$}}
+// CHECK-NOT: __riscv_xcvbitmanip {{.*$}}
+// CHECK-NOT: __riscv_xcvelw {{.*$}}
+// CHECK-NOT: __riscv_xcvmac {{.*$}}
+// CHECK-NOT: __riscv_xcvmem {{.*$}}
+// CHECK-NOT: __riscv_xcvsimd {{.*$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixcvalu -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVALU-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixcvalu -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVALU-EXT %s
+// CHECK-XCVALU-EXT: __riscv_xcvalu 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixcvbi -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVBI-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixcvbi -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVBI-EXT %s
+// CHECK-XCVBI-EXT: __riscv_xcvbi 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixcvbitmanip -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVBITMANIP-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixcvbitmanip -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVBITMANIP-EXT %s
+// CHECK-XCVBITMANIP-EXT: __riscv_xcvbitmanip 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixcvmac -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVMAC-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixcvmac -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVMAC-EXT %s
+// CHECK-XCVMAC-EXT: __riscv_xcvmac 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixcvmem -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVMEM-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixcvmem -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVMEM-EXT %s
+// CHECK-XCVMEM-EXT: __riscv_xcvmem 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixcvsimd -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVSIMD-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixcvsimd -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVSIMD-EXT %s
+// CHECK-XCVSIMD-EXT: __riscv_xcvsimd 1000000{{$}}
diff --git a/clang/test/Preprocessor/riscv-target-features-sifive.c b/clang/test/Preprocessor/riscv-target-features-sifive.c
index e4c0387..1c49b55 100644
--- a/clang/test/Preprocessor/riscv-target-features-sifive.c
+++ b/clang/test/Preprocessor/riscv-target-features-sifive.c
@@ -1,3 +1,91 @@
+// RUN: %clang --target=riscv32-unknown-linux-gnu -march=rv32i -E -dM %s \
+// RUN:   -o - | FileCheck %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu -march=rv64i -E -dM %s \
+// RUN:   -o - | FileCheck %s
+
+// CHECK-NOT: __riscv_xsfcease {{.*$}}
+// CHECK-NOT: __riscv_xsfvcp {{.*$}}
+// CHECK-NOT: __riscv_xsfvfnrclipxfqf {{.*$}}
+// CHECK-NOT: __riscv_xsfvfwmaccqqq {{.*$}}
+// CHECK-NOT: __riscv_xsfqmaccdod {{.*$}}
+// CHECK-NOT: __riscv_xsfvqmaccqoq {{.*$}}
+// CHECK-NOT: __riscv_xsifivecdiscarddlone {{.*$}}
+// CHECK-NOT: __riscv_xsifivecflushdlone {{.*$}}
+// CHECK-NOT: __riscv_xsfmm128t {{.*$}}
+// CHECK-NOT: __riscv_xsfmm16t {{.*$}}
+// CHECK-NOT: __riscv_xsfmm32a8i {{.*$}}
+// CHECK-NOT: __riscv_xsfmm32a8f {{.*$}}
+// CHECK-NOT: __riscv_xsfmm32a16f {{.*$}}
+// CHECK-NOT: __riscv_xsfmm32a32f {{.*$}}
+// CHECK-NOT: __riscv_xsfmm32a32t {{.*$}}
+// CHECK-NOT: __riscv_xsfmm64a64f {{.*$}}
+// CHECK-NOT: __riscv_xsfmm64t {{.*$}}
+// CHECK-NOT: __riscv_xsfmmbase {{.*$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixsfcease -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFCEASE-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixsfcease -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFCEASE-EXT %s
+// CHECK-XSFCEASE-EXT: __riscv_xsfcease 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixsfvcp -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVCP-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixsfvcp -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVCP-EXT %s
+// CHECK-XSFVCP-EXT: __riscv_xsfvcp 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixsfvfnrclipxfqf -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVFNRCLIPXFQF-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixsfvfnrclipxfqf -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVFNRCLIPXFQF-EXT %s
+// CHECK-XSFVFNRCLIPXFQF-EXT: __riscv_xsfvfnrclipxfqf 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixsfvfwmaccqqq -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVFWMACCQQQ-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixsfvfwmaccqqq -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVFWMACCQQQ-EXT %s
+// CHECK-XSFVFWMACCQQQ-EXT: __riscv_xsfvfwmaccqqq 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixsfvqmaccdod -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVQMACCDOD-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixsfvqmaccdod -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVQMACCDOD-EXT %s
+// CHECK-XSFVQMACCDOD-EXT: __riscv_xsfvqmaccdod 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixsfvqmaccqoq -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVQMACCQOQ-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixsfvqmaccqoq -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVQMACCQOQ-EXT %s
+// CHECK-XSFVQMACCQOQ-EXT: __riscv_xsfvqmaccqoq 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixsifivecdiscarddlone -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XSIFIVECDISCARDDLONE-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixsifivecdiscarddlone -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XSIFIVECDISCARDDLONE-EXT %s
+// CHECK-XSIFIVECDISCARDDLONE-EXT: __riscv_xsifivecdiscarddlone 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixsifivecflushdlone -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XSIFIVECFLUSHDLONE-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixsifivecflushdlone -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XSIFIVECFLUSHDLONE-EXT %s
+// CHECK-XSIFIVECFLUSHDLONE-EXT: __riscv_xsifivecflushdlone 1000000{{$}}
+
 // RUN: %clang --target=riscv32 \
 // RUN:   -march=rv32i_zve32x_xsfmm128t -E -dM %s \
 // RUN:   -o - | FileCheck --check-prefix=CHECK-XSFMM128T %s
diff --git a/clang/test/Preprocessor/riscv-target-features-thead.c b/clang/test/Preprocessor/riscv-target-features-thead.c
new file mode 100644
index 0000000..9d27d9a
--- /dev/null
+++ b/clang/test/Preprocessor/riscv-target-features-thead.c
@@ -0,0 +1,104 @@
+// RUN: %clang --target=riscv32-unknown-linux-gnu -march=rv32i -E -dM %s \
+// RUN:   -o - | FileCheck %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu -march=rv64i -E -dM %s \
+// RUN:   -o - | FileCheck %s
+
+// CHECK-NOT: __riscv_xtheadba {{.*$}}
+// CHECK-NOT: __riscv_xtheadbb {{.*$}}
+// CHECK-NOT: __riscv_xtheadbs {{.*$}}
+// CHECK-NOT: __riscv_xtheadcmo {{.*$}}
+// CHECK-NOT: __riscv_xtheadcondmov {{.*$}}
+// CHECK-NOT: __riscv_xtheadfmemidx {{.*$}}
+// CHECK-NOT: __riscv_xtheadmac {{.*$}}
+// CHECK-NOT: __riscv_xtheadmemidx {{.*$}}
+// CHECK-NOT: __riscv_xtheadmempair {{.*$}}
+// CHECK-NOT: __riscv_xtheadsync {{.*$}}
+// CHECK-NOT: __riscv_xtheadvdot {{.*$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixtheadba -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADBA-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixtheadba -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADBA-EXT %s
+// CHECK-XTHEADBA-EXT: __riscv_xtheadba 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixtheadbb -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADBB-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixtheadbb -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADBB-EXT %s
+// CHECK-XTHEADBB-EXT: __riscv_xtheadbb 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixtheadbs -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADBS-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixtheadbs -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADBS-EXT %s
+// CHECK-XTHEADBS-EXT: __riscv_xtheadbs 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixtheadcmo -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADCMO-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixtheadcmo -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADCMO-EXT %s
+// CHECK-XTHEADCMO-EXT: __riscv_xtheadcmo 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixtheadcondmov -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADCONDMOV-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixtheadcondmov -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADCONDMOV-EXT %s
+// CHECK-XTHEADCONDMOV-EXT: __riscv_xtheadcondmov 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixtheadfmemidx -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADFMEMIDX-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixtheadfmemidx -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADFMEMIDX-EXT %s
+// CHECK-XTHEADFMEMIDX-EXT: __riscv_xtheadfmemidx 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixtheadmac -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADMAC-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixtheadmac -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADMAC-EXT %s
+// CHECK-XTHEADMAC-EXT: __riscv_xtheadmac 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixtheadmemidx -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADMEMIDX-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixtheadmemidx -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADMEMIDX-EXT %s
+// CHECK-XTHEADMEMIDX-EXT: __riscv_xtheadmemidx 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixtheadmempair -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADMEMPAIR-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixtheadmempair -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADMEMPAIR-EXT %s
+// CHECK-XTHEADMEMPAIR-EXT: __riscv_xtheadmempair 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixtheadsync -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADSYNC-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixtheadsync -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADSYNC-EXT %s
+// CHECK-XTHEADSYNC-EXT: __riscv_xtheadsync 1000000{{$}}
+
+// RUN: %clang --target=riscv32-unknown-linux-gnu \
+// RUN:   -march=rv32ixtheadvdot -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADVDOT-EXT %s
+// RUN: %clang --target=riscv64-unknown-linux-gnu \
+// RUN:   -march=rv64ixtheadvdot -E -dM %s \
+// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADVDOT-EXT %s
+// CHECK-XTHEADVDOT-EXT: __riscv_xtheadvdot 1000000{{$}}
diff --git a/clang/test/Preprocessor/riscv-target-features.c b/clang/test/Preprocessor/riscv-target-features.c
index 86085c2..864d782 100644
--- a/clang/test/Preprocessor/riscv-target-features.c
+++ b/clang/test/Preprocessor/riscv-target-features.c
@@ -64,32 +64,6 @@
 // CHECK-NOT: __riscv_v_intrinsic {{.*$}}
 // CHECK-NOT: __riscv_v_min_vlen {{.*$}}
 // CHECK-NOT: __riscv_vector {{.*$}}
-// CHECK-NOT: __riscv_xcvalu {{.*$}}
-// CHECK-NOT: __riscv_xcvbi {{.*$}}
-// CHECK-NOT: __riscv_xcvbitmanip {{.*$}}
-// CHECK-NOT: __riscv_xcvelw {{.*$}}
-// CHECK-NOT: __riscv_xcvmac {{.*$}}
-// CHECK-NOT: __riscv_xcvmem {{.*$}}
-// CHECK-NOT: __riscv_xcvsimd {{.*$}}
-// CHECK-NOT: __riscv_xsfcease {{.*$}}
-// CHECK-NOT: __riscv_xsfvcp {{.*$}}
-// CHECK-NOT: __riscv_xsfvfnrclipxfqf {{.*$}}
-// CHECK-NOT: __riscv_xsfvfwmaccqqq {{.*$}}
-// CHECK-NOT: __riscv_xsfqmaccdod {{.*$}}
-// CHECK-NOT: __riscv_xsfvqmaccqoq {{.*$}}
-// CHECK-NOT: __riscv_xsifivecdiscarddlone {{.*$}}
-// CHECK-NOT: __riscv_xsifivecflushdlone {{.*$}}
-// CHECK-NOT: __riscv_xtheadba {{.*$}}
-// CHECK-NOT: __riscv_xtheadbb {{.*$}}
-// CHECK-NOT: __riscv_xtheadbs {{.*$}}
-// CHECK-NOT: __riscv_xtheadcmo {{.*$}}
-// CHECK-NOT: __riscv_xtheadcondmov {{.*$}}
-// CHECK-NOT: __riscv_xtheadfmemidx {{.*$}}
-// CHECK-NOT: __riscv_xtheadmac {{.*$}}
-// CHECK-NOT: __riscv_xtheadmemidx {{.*$}}
-// CHECK-NOT: __riscv_xtheadmempair {{.*$}}
-// CHECK-NOT: __riscv_xtheadsync {{.*$}}
-// CHECK-NOT: __riscv_xtheadvdot {{.*$}}
 // CHECK-NOT: __riscv_xventanacondops {{.*$}}
 // CHECK-NOT: __riscv_za128rs {{.*$}}
 // CHECK-NOT: __riscv_za64rs {{.*$}}
@@ -569,198 +543,6 @@
 // CHECK-V-EXT: __riscv_vector 1
 
 // RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixcvalu -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVALU-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixcvalu -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVALU-EXT %s
-// CHECK-XCVALU-EXT: __riscv_xcvalu 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixcvbi -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVBI-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixcvbi -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVBI-EXT %s
-// CHECK-XCVBI-EXT: __riscv_xcvbi 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixcvbitmanip -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVBITMANIP-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixcvbitmanip -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVBITMANIP-EXT %s
-// CHECK-XCVBITMANIP-EXT: __riscv_xcvbitmanip 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixcvmac -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVMAC-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixcvmac -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVMAC-EXT %s
-// CHECK-XCVMAC-EXT: __riscv_xcvmac 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixcvsimd -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVSIMD-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixcvsimd -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XCVSIMD-EXT %s
-// CHECK-XCVSIMD-EXT: __riscv_xcvsimd 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixsfcease -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFCEASE-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixsfcease -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFCEASE-EXT %s
-// CHECK-XSFCEASE-EXT: __riscv_xsfcease 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixsfvcp -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVCP-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixsfvcp -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVCP-EXT %s
-// CHECK-XSFVCP-EXT: __riscv_xsfvcp 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixsfvfnrclipxfqf -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVFNRCLIPXFQF-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixsfvfnrclipxfqf -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVFNRCLIPXFQF-EXT %s
-// CHECK-XSFVFNRCLIPXFQF-EXT: __riscv_xsfvfnrclipxfqf 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixsfvfwmaccqqq -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVFWMACCQQQ-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixsfvfwmaccqqq -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVFWMACCQQQ-EXT %s
-// CHECK-XSFVFWMACCQQQ-EXT: __riscv_xsfvfwmaccqqq 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixsfvqmaccdod -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVQMACCDOD-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixsfvqmaccdod -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVQMACCDOD-EXT %s
-// CHECK-XSFVQMACCDOD-EXT: __riscv_xsfvqmaccdod 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixsfvqmaccqoq -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVQMACCQOQ-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixsfvqmaccqoq -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XSFVQMACCQOQ-EXT %s
-// CHECK-XSFVQMACCQOQ-EXT: __riscv_xsfvqmaccqoq 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixsifivecdiscarddlone -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XSIFIVECDISCARDDLONE-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixsifivecdiscarddlone -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XSIFIVECDISCARDDLONE-EXT %s
-// CHECK-XSIFIVECDISCARDDLONE-EXT: __riscv_xsifivecdiscarddlone 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixsifivecflushdlone -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XSIFIVECFLUSHDLONE-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixsifivecflushdlone -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XSIFIVECFLUSHDLONE-EXT %s
-// CHECK-XSIFIVECFLUSHDLONE-EXT: __riscv_xsifivecflushdlone 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixtheadba -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADBA-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixtheadba -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADBA-EXT %s
-// CHECK-XTHEADBA-EXT: __riscv_xtheadba 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixtheadbb -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADBB-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixtheadbb -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADBB-EXT %s
-// CHECK-XTHEADBB-EXT: __riscv_xtheadbb 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixtheadbs -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADBS-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixtheadbs -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADBS-EXT %s
-// CHECK-XTHEADBS-EXT: __riscv_xtheadbs 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixtheadcmo -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADCMO-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixtheadcmo -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADCMO-EXT %s
-// CHECK-XTHEADCMO-EXT: __riscv_xtheadcmo 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixtheadcondmov -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADCONDMOV-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixtheadcondmov -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADCONDMOV-EXT %s
-// CHECK-XTHEADCONDMOV-EXT: __riscv_xtheadcondmov 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixtheadfmemidx -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADFMEMIDX-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixtheadfmemidx -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADFMEMIDX-EXT %s
-// CHECK-XTHEADFMEMIDX-EXT: __riscv_xtheadfmemidx 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixtheadmac -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADMAC-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixtheadmac -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADMAC-EXT %s
-// CHECK-XTHEADMAC-EXT: __riscv_xtheadmac 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixtheadmemidx -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADMEMIDX-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixtheadmemidx -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADMEMIDX-EXT %s
-// CHECK-XTHEADMEMIDX-EXT: __riscv_xtheadmemidx 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixtheadmempair -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADMEMPAIR-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixtheadmempair -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADMEMPAIR-EXT %s
-// CHECK-XTHEADMEMPAIR-EXT: __riscv_xtheadmempair 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixtheadsync -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADSYNC-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixtheadsync -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADSYNC-EXT %s
-// CHECK-XTHEADSYNC-EXT: __riscv_xtheadsync 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
-// RUN:   -march=rv32ixtheadvdot -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADVDOT-EXT %s
-// RUN: %clang --target=riscv64-unknown-linux-gnu \
-// RUN:   -march=rv64ixtheadvdot -E -dM %s \
-// RUN:   -o - | FileCheck --check-prefix=CHECK-XTHEADVDOT-EXT %s
-// CHECK-XTHEADVDOT-EXT: __riscv_xtheadvdot 1000000{{$}}
-
-// RUN: %clang --target=riscv32-unknown-linux-gnu \
 // RUN:   -march=rv32ixventanacondops -E -dM %s \
 // RUN:   -o - | FileCheck --check-prefix=CHECK-XVENTANACONDOPS-EXT %s
 // RUN: %clang --target=riscv64-unknown-linux-gnu \
diff --git a/clang/test/Sema/builtins-wasm.c b/clang/test/Sema/builtins-wasm.c
index 31e5291..a3486b1 100644
--- a/clang/test/Sema/builtins-wasm.c
+++ b/clang/test/Sema/builtins-wasm.c
@@ -54,3 +54,27 @@ void test_table_copy(int dst_idx, int src_idx, int nelem) {
   __builtin_wasm_table_copy(table, table, dst_idx, src_idx, table);    // expected-error {{5th argument must be an integer}}
   __builtin_wasm_table_copy(table, table, dst_idx, src_idx, nelem);
 }
+
+typedef void (*F1)(void);
+typedef int (*F2)(int);
+typedef int (*F3)(__externref_t);
+typedef __externref_t (*F4)(int);
+
+void test_function_pointer_signature() {
+  // Test argument count validation
+  (void)__builtin_wasm_test_function_pointer_signature(); // expected-error {{too few arguments to function call, expected 1, have 0}}
+  (void)__builtin_wasm_test_function_pointer_signature((F1)0, (F2)0); // expected-error {{too many arguments to function call, expected 1, have 2}}
+
+  // // Test argument type validation - should require function pointer
+  (void)__builtin_wasm_test_function_pointer_signature((void*)0); // expected-error {{used type 'void *' where function pointer is required}}
+  (void)__builtin_wasm_test_function_pointer_signature((int)0);   // expected-error {{used type 'int' where function pointer is required}}
+  (void)__builtin_wasm_test_function_pointer_signature((F3)0);   // expected-error {{not supported for function pointers with a reference type parameter}}
+  (void)__builtin_wasm_test_function_pointer_signature((F4)0);   // expected-error {{not supported for function pointers with a reference type return value}}
+
+  // // Test valid usage
+  int res = __builtin_wasm_test_function_pointer_signature((F1)0);
+  res = __builtin_wasm_test_function_pointer_signature((F2)0);
+
+  // Test return type
+  _Static_assert(EXPR_HAS_TYPE(__builtin_wasm_test_function_pointer_signature((F1)0), int), "");
+}
diff --git a/clang/test/SemaOpenACC/compute-construct-reduction-clause.c b/clang/test/SemaOpenACC/compute-construct-reduction-clause.c
index 6a77f07..995b6d3 100644
--- a/clang/test/SemaOpenACC/compute-construct-reduction-clause.c
+++ b/clang/test/SemaOpenACC/compute-construct-reduction-clause.c
@@ -68,7 +68,6 @@ void uses(unsigned Parm) {
 #pragma acc parallel reduction(&: ChC)
   while (1);
 
-  // expected-error@+1{{OpenACC 'reduction' variable must be of scalar type, sub-array, or a composite of scalar types; type is 'int[5]'}}
 #pragma acc parallel reduction(&: Array)
   while (1);
 
@@ -76,7 +75,7 @@ void uses(unsigned Parm) {
   while (1);
 
   struct CompositeHasComposite ChCArray[5];
-  // expected-error@+1{{OpenACC 'reduction' variable must be of scalar type, sub-array, or a composite of scalar types; sub-array base type is 'struct CompositeHasComposite'}}
+  // expected-error@+1{{OpenACC 'reduction' variable must be of scalar type, aggregate, sub-array, or a composite of scalar types; sub-array base type is 'struct CompositeHasComposite'}}
 #pragma acc parallel reduction(&: CoS, Array[I], ChCArray[0:I])
   while (1);
 
diff --git a/clang/test/SemaOpenACC/compute-construct-reduction-clause.cpp b/clang/test/SemaOpenACC/compute-construct-reduction-clause.cpp
index 3e972b0..b40268c 100644
--- a/clang/test/SemaOpenACC/compute-construct-reduction-clause.cpp
+++ b/clang/test/SemaOpenACC/compute-construct-reduction-clause.cpp
@@ -70,7 +70,6 @@ void uses(unsigned Parm) {
   // expected-note@#COS_FIELD{{invalid field is here}}
 #pragma acc parallel reduction(&: ChC)
   while (1);
-  // expected-error@+1{{OpenACC 'reduction' variable must be of scalar type, sub-array, or a composite of scalar types; type is 'int[5]'}}
 #pragma acc parallel reduction(&: Array)
   while (1);
 
@@ -140,10 +139,8 @@ void TemplUses(T Parm, U CoS, V ChC) {
   // expected-note@#COS_FIELD{{invalid field is here}}
 #pragma acc parallel reduction(&: ChC)
   while (1);
-  // expected-error@+1{{OpenACC 'reduction' variable must be of scalar type, sub-array, or a composite of scalar types; type is 'int[5]'}}
 #pragma acc parallel reduction(&: Array)
   while (1);
-  // expected-error@+1{{OpenACC 'reduction' variable must be of scalar type, sub-array, or a composite of scalar types; type is 'int[5]'}}
 #pragma acc parallel reduction(&: NonDepArray)
   while (1);
 
diff --git a/clang/test/SemaOpenACC/data-construct-use_device-clause.c b/clang/test/SemaOpenACC/data-construct-use_device-clause.c
index 9239757..65eaf4e 100644
--- a/clang/test/SemaOpenACC/data-construct-use_device-clause.c
+++ b/clang/test/SemaOpenACC/data-construct-use_device-clause.c
@@ -1,4 +1,4 @@
-// RUN: %clang_cc1 %s -fopenacc -verify
+// RUN: %clang_cc1 %s -fopenacc -verify -Wopenacc-extension
 
 typedef struct IsComplete {
   struct S { int A; } CompositeMember;
@@ -23,7 +23,7 @@ void uses(int IntParam, short *PointerParam, float ArrayParam[5], Complete Compo
 #pragma acc host_data use_device(LocalComposite.ScalarMember, LocalComposite.ScalarMember)
   ;
 
-  // expected-error@+1{{OpenACC variable in 'use_device' clause is not a valid variable name or array name}}
+  // expected-warning@+1{{sub-array as a variable in 'use_device' clause is not a valid variable name or array name}}
 #pragma acc host_data use_device(LocalArray[2:1])
 
   // expected-error@+1{{OpenACC variable in 'use_device' clause is not a valid variable name or array name}}
@@ -35,12 +35,12 @@ void uses(int IntParam, short *PointerParam, float ArrayParam[5], Complete Compo
   ;
 
   // expected-error@+2{{OpenACC sub-array length is unspecified and cannot be inferred because the subscripted value is not an array}}
-  // expected-error@+1{{OpenACC variable in 'use_device' clause is not a valid variable name or array name}}
+  // expected-warning@+1{{sub-array as a variable in 'use_device' clause is not a valid variable name or array name}}
 #pragma acc host_data use_device(PointerParam[2:])
   ;
 
   // expected-error@+2{{OpenACC sub-array specified range [2:5] would be out of the range of the subscripted array size of 5}}
-  // expected-error@+1{{OpenACC variable in 'use_device' clause is not a valid variable name or array name}}
+  // expected-warning@+1{{sub-array as a variable in 'use_device' clause is not a valid variable name or array name}}
 #pragma acc host_data use_device(ArrayParam[2:5])
   ;
 
diff --git a/clang/test/SemaOpenACC/data-construct.cpp b/clang/test/SemaOpenACC/data-construct.cpp
index 394ebb0..da7b80a 100644
--- a/clang/test/SemaOpenACC/data-construct.cpp
+++ b/clang/test/SemaOpenACC/data-construct.cpp
@@ -1,4 +1,4 @@
-// RUN: %clang_cc1 %s -fopenacc -verify -Wno-empty-body -Wno-unused-value
+// RUN: %clang_cc1 %s -fopenacc -verify -Wno-empty-body -Wno-unused-value -Wopenacc-extension
 
 void HasStmt() {
   {
@@ -185,7 +185,7 @@ void HostDataRules() {
 #pragma acc host_data use_device(Array)
   ;
 
-  // expected-error@+1{{OpenACC variable in 'use_device' clause is not a valid variable name or array name}}
+  // expected-warning@+1{{sub-array as a variable in 'use_device' clause is not a valid variable name or array name}}
 #pragma acc host_data use_device(Array[1:1])
   ;
 
diff --git a/clang/test/SemaOpenACC/declare-construct.cpp b/clang/test/SemaOpenACC/declare-construct.cpp
index 6f21aed..6828ecd 100644
--- a/clang/test/SemaOpenACC/declare-construct.cpp
+++ b/clang/test/SemaOpenACC/declare-construct.cpp
@@ -1,7 +1,8 @@
-// RUN: %clang_cc1 %s -fopenacc -verify
+// RUN: %clang_cc1 %s -fopenacc -verify -Wopenacc-extension
 
 int *Global;
 int GlobalArray[5];
+int GlobalArray2[5];
 // expected-error@+1{{no valid clauses specified in OpenACC 'declare' directive}}
 #pragma acc declare
 namespace NS {
@@ -265,8 +266,8 @@ void use() {
 
 // expected-error@+1{{OpenACC variable on 'declare' construct is not a valid variable name or array name}}
 #pragma acc declare create(GlobalArray[0])
-// expected-error@+1{{OpenACC variable on 'declare' construct is not a valid variable name or array name}}
-#pragma acc declare create(GlobalArray[0: 1])
+// expected-warning@+1{{sub-array as a variable on 'declare' construct is not a valid variable name or array name}}
+#pragma acc declare create(GlobalArray[0: 1]) // #GLOBALARRAYREF
 
 struct S { int I; };
 // expected-error@+1{{OpenACC variable on 'declare' construct is not a valid variable name or array name}}
@@ -288,8 +289,12 @@ void ExternVar() {
 #pragma acc declare copy(I) copyin(I2), copyout(I3), create(I4), present(I5), deviceptr(I6), device_resident(I7), link(I8)
 }
 
+// expected-error@+2{{variable referenced in 'link' clause of OpenACC 'declare' directive was already referenced}}
+// expected-note@#GLOBALARRAYREF{{previous reference is here}}
+#pragma acc declare link(GlobalArray)
+
 // Link can only have global, namespace, or extern vars.
-#pragma acc declare link(Global, GlobalArray)
+#pragma acc declare link(Global, GlobalArray2)
 
 struct Struct2 {
   static const int StaticMem = 5;
diff --git a/clang/test/SemaOpenACC/loop-construct-reduction-clause.cpp b/clang/test/SemaOpenACC/loop-construct-reduction-clause.cpp
index bbcfffb..00bcd74 100644
--- a/clang/test/SemaOpenACC/loop-construct-reduction-clause.cpp
+++ b/clang/test/SemaOpenACC/loop-construct-reduction-clause.cpp
@@ -36,7 +36,6 @@ void uses() {
 
 #pragma acc serial
   {
-  // expected-error@+1{{OpenACC 'reduction' variable must be of scalar type, sub-array, or a composite of scalar types; type is 'int[5]'}}
 #pragma acc loop reduction(+:Array)
     for(int i = 0; i < 5; ++i){}
   }
@@ -172,7 +171,6 @@ void templ_uses() {
 
 #pragma acc serial
   {
-  // expected-error@+1{{OpenACC 'reduction' variable must be of scalar type, sub-array, or a composite of scalar types; type is 'int[5]'}}
 #pragma acc loop reduction(+:Array)
     for(int i = 0; i < 5; ++i){}
   }
diff --git a/clang/unittests/Format/FormatTest.cpp b/clang/unittests/Format/FormatTest.cpp
index c20d099..9c5aa11 100644
--- a/clang/unittests/Format/FormatTest.cpp
+++ b/clang/unittests/Format/FormatTest.cpp
@@ -8593,10 +8593,10 @@ TEST_F(FormatTest, BreaksFunctionDeclarations) {
                "operator<<(const SomeLooooooooooooooooooooooooogType &other);");
   verifyGoogleFormat(
       "SomeLoooooooooooooooooooooooooooooogType operator>>(\n"
-      "    const SomeLooooooooogType &a, const SomeLooooooooogType &b);");
+      "    const SomeLooooooooogType& a, const SomeLooooooooogType& b);");
   verifyGoogleFormat(
       "SomeLoooooooooooooooooooooooooooooogType operator<<(\n"
-      "    const SomeLooooooooogType &a, const SomeLooooooooogType &b);");
+      "    const SomeLooooooooogType& a, const SomeLooooooooogType& b);");
 
   verifyFormat("void aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa(\n"
                "    int aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa = 1);");
@@ -8605,7 +8605,7 @@ TEST_F(FormatTest, BreaksFunctionDeclarations) {
   verifyGoogleFormat(
       "typename aaaaaaaaaa<aaaaaa>::aaaaaaaaaaa\n"
       "aaaaaaaaaa<aaaaaa>::aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa(\n"
-      "    bool *aaaaaaaaaaaaaaaaaa, bool *aa) {}");
+      "    bool* aaaaaaaaaaaaaaaaaa, bool* aa) {}");
   verifyGoogleFormat("template <typename T>\n"
                      "aaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaaa\n"
                      "aaaaaaaaaaaaaaaaaaaaaaa<T>::aaaaaaaaaaaaa(\n"
@@ -12114,7 +12114,12 @@ TEST_F(FormatTest, UnderstandsFunctionRefQualification) {
            "void b() const &;\n";
   verifyFormat(Prefix + "int *x;", Prefix + "int* x;", DerivePointerAlignment);
 
-  verifyGoogleFormat("MACRO(int*, std::function<void() &&>);");
+  constexpr StringRef Code("MACRO(int*, std::function<void() &&>);");
+  verifyFormat(Code, DerivePointerAlignment);
+
+  auto Style = getGoogleStyle();
+  Style.DerivePointerAlignment = true;
+  verifyFormat(Code, Style);
 }
 
 TEST_F(FormatTest, PointerAlignmentFallback) {
@@ -12915,27 +12920,31 @@ TEST_F(FormatTest, UnderstandsEllipsis) {
 }
 
 TEST_F(FormatTest, AdaptivelyFormatsPointersAndReferences) {
+  auto Style = getGoogleStyle();
+  EXPECT_FALSE(Style.DerivePointerAlignment);
+  Style.DerivePointerAlignment = true;
+
   verifyFormat("int *a;\n"
                "int *a;\n"
                "int *a;",
                "int *a;\n"
                "int* a;\n"
                "int *a;",
-               getGoogleStyle());
+               Style);
   verifyFormat("int* a;\n"
                "int* a;\n"
                "int* a;",
                "int* a;\n"
                "int* a;\n"
                "int *a;",
-               getGoogleStyle());
+               Style);
   verifyFormat("int *a;\n"
                "int *a;\n"
                "int *a;",
                "int *a;\n"
                "int * a;\n"
                "int *  a;",
-               getGoogleStyle());
+               Style);
   verifyFormat("auto x = [] {\n"
                "  int *a;\n"
                "  int *a;\n"
@@ -12944,7 +12953,7 @@ TEST_F(FormatTest, AdaptivelyFormatsPointersAndReferences) {
                "auto x=[]{int *a;\n"
                "int * a;\n"
                "int *  a;};",
-               getGoogleStyle());
+               Style);
 }
 
 TEST_F(FormatTest, UnderstandsRvalueReferences) {
@@ -13080,7 +13089,7 @@ TEST_F(FormatTest, FormatsCasts) {
   verifyFormat("virtual void foo(char &) const;");
   verifyFormat("virtual void foo(int *a, char *) const;");
   verifyFormat("int a = sizeof(int *) + b;");
-  verifyGoogleFormat("int a = alignof(int *) + b;");
+  verifyGoogleFormat("int a = alignof(int*) + b;");
   verifyFormat("bool b = f(g<int>) && c;");
   verifyFormat("typedef void (*f)(int i) func;");
   verifyFormat("void operator++(int) noexcept;");
@@ -25449,7 +25458,7 @@ TEST_F(FormatTest, AtomicQualifier) {
   verifyFormat("struct foo {\n"
                "  int a1;\n"
                "  _Atomic(a) a2;\n"
-               "  _Atomic(_Atomic(int) *const) a3;\n"
+               "  _Atomic(_Atomic(int)* const) a3;\n"
                "};",
                Google);
   verifyFormat("_Atomic(uint64_t) a;");
diff --git a/compiler-rt/lib/sanitizer_common/sanitizer_platform_limits_posix.cpp b/compiler-rt/lib/sanitizer_common/sanitizer_platform_limits_posix.cpp
index ef4b8a8..ea8cc30 100644
--- a/compiler-rt/lib/sanitizer_common/sanitizer_platform_limits_posix.cpp
+++ b/compiler-rt/lib/sanitizer_common/sanitizer_platform_limits_posix.cpp
@@ -779,15 +779,11 @@ unsigned struct_ElfW_Phdr_sz = sizeof(Elf_Phdr);
   unsigned IOCTL_SOUND_PCM_WRITE_FILTER = SOUND_PCM_WRITE_FILTER;
 #endif // SOUND_VERSION
   unsigned IOCTL_TCFLSH = TCFLSH;
-  unsigned IOCTL_TCGETA = TCGETA;
 #    if SANITIZER_TERMIOS_IOCTL_CONSTANTS
   unsigned IOCTL_TCGETS = TCGETS;
 #    endif
   unsigned IOCTL_TCSBRK = TCSBRK;
   unsigned IOCTL_TCSBRKP = TCSBRKP;
-  unsigned IOCTL_TCSETA = TCSETA;
-  unsigned IOCTL_TCSETAF = TCSETAF;
-  unsigned IOCTL_TCSETAW = TCSETAW;
 #    if SANITIZER_TERMIOS_IOCTL_CONSTANTS
   unsigned IOCTL_TCSETS = TCSETS;
   unsigned IOCTL_TCSETSF = TCSETSF;
diff --git a/compiler-rt/lib/sanitizer_common/sanitizer_platform_limits_posix.h b/compiler-rt/lib/sanitizer_common/sanitizer_platform_limits_posix.h
index 9e11dd9..2496652 100644
--- a/compiler-rt/lib/sanitizer_common/sanitizer_platform_limits_posix.h
+++ b/compiler-rt/lib/sanitizer_common/sanitizer_platform_limits_posix.h
@@ -1338,12 +1338,8 @@ extern unsigned IOCTL_SNDCTL_COPR_SENDMSG;
 extern unsigned IOCTL_SNDCTL_COPR_WCODE;
 extern unsigned IOCTL_SNDCTL_COPR_WDATA;
 extern unsigned IOCTL_TCFLSH;
-extern unsigned IOCTL_TCGETA;
 extern unsigned IOCTL_TCSBRK;
 extern unsigned IOCTL_TCSBRKP;
-extern unsigned IOCTL_TCSETA;
-extern unsigned IOCTL_TCSETAF;
-extern unsigned IOCTL_TCSETAW;
 #    if SANITIZER_TERMIOS_IOCTL_CONSTANTS
 extern unsigned IOCTL_TCGETS;
 extern unsigned IOCTL_TCSETS;
diff --git a/compiler-rt/test/rtsan/pthread_cond_wait.cpp b/compiler-rt/test/rtsan/pthread_cond_wait.cpp
new file mode 100644
index 0000000..29afbf4b
--- /dev/null
+++ b/compiler-rt/test/rtsan/pthread_cond_wait.cpp
@@ -0,0 +1,48 @@
+// RUN: %clangxx -fsanitize=realtime %s -o %t
+// RUN: %run %t 2>&1 | FileCheck %s
+
+// UNSUPPORTED: ios
+
+// Intent: Ensures that pthread_cond_signal does not segfault under rtsan
+// See issue #146120
+
+#include <condition_variable>
+#include <future>
+#include <mutex>
+#include <thread>
+
+#include <iostream>
+
+int main() {
+  std::cout << "Entry to main!" << std::endl;
+
+
+  // TODO: This is disabled because it does cause a test failure
+  /*
+  std::mutex mut;
+  std::condition_variable cv;
+  bool go{false};
+
+  const auto fut = std::async(std::launch::async, [&] {
+    std::this_thread::sleep_for(std::chrono::milliseconds(100));
+    {
+      std::unique_lock<std::mutex> lock(mut);
+      go = true;
+    }
+    cv.notify_one();
+  });
+
+  std::unique_lock<std::mutex> lock(mut);
+  // normal wait is fine
+  // cv.wait(lock, [&] { return go; });
+  // but timed wait could segfault
+
+  // NOTE: When a fix for the pthread_cond issue #146120 is fixed, uncomment this line
+  //cv.wait_for(lock, std::chrono::milliseconds(200), [&] { return go; });
+  */
+
+  std::cout << "Exit from main!" << std::endl;
+}
+
+// CHECK: Entry to main!
+// CHECK-NEXT: Exit from main!
diff --git a/flang-rt/cmake/modules/AddFlangRT.cmake b/flang-rt/cmake/modules/AddFlangRT.cmake
index e51590f..5854160 100644
--- a/flang-rt/cmake/modules/AddFlangRT.cmake
+++ b/flang-rt/cmake/modules/AddFlangRT.cmake
@@ -286,27 +286,6 @@ function (add_flangrt_library name)
       target_compile_options(${tgtname} PUBLIC -U_LIBCPP_ENABLE_ASSERTIONS)
     endif ()
 
-    # Flang/Clang (including clang-cl) -compiled programs targeting the MSVC ABI
-    # should only depend on msvcrt/ucrt. LLVM still emits libgcc/compiler-rt
-    # functions in some cases like 128-bit integer math (__udivti3, __modti3,
-    # __fixsfti, __floattidf, ...) that msvc does not support. We are injecting a
-    # dependency to Compiler-RT's builtin library where these are implemented.
-    if (MSVC AND CMAKE_CXX_COMPILER_ID MATCHES "Clang")
-      if (FLANG_RT_BUILTINS_LIBRARY)
-        target_compile_options(${tgtname} PRIVATE "$<$<COMPILE_LANGUAGE:CXX,C>:-Xclang>" "$<$<COMPILE_LANGUAGE:CXX,C>:--dependent-lib=${FLANG_RT_BUILTINS_LIBRARY}>")
-      endif ()
-    endif ()
-    if (MSVC AND CMAKE_Fortran_COMPILER_ID STREQUAL "LLVMFlang")
-      if (FLANG_RT_BUILTINS_LIBRARY)
-        target_compile_options(${tgtname} PRIVATE "$<$<COMPILE_LANGUAGE:Fortran>:-Xflang>" "$<$<COMPILE_LANGUAGE:Fortran>:--dependent-lib=${FLANG_RT_BUILTINS_LIBRARY}>")
-      else ()
-        message(WARNING "Did not find libclang_rt.builtins.lib.
-          LLVM may emit builtins that are not implemented in msvcrt/ucrt and
-          instead falls back to builtins from Compiler-RT. Linking with ${tgtname}
-          may result in a linker error.")
-      endif ()
-    endif ()
-
     # Non-GTest unittests depend on LLVMSupport
     if (ARG_LINK_TO_LLVM)
       if (LLVM_LINK_LLVM_DYLIB)
diff --git a/flang-rt/include/flang-rt/runtime/io-stmt.h b/flang-rt/include/flang-rt/runtime/io-stmt.h
index bfd5f69..95b2ee7 100644
--- a/flang-rt/include/flang-rt/runtime/io-stmt.h
+++ b/flang-rt/include/flang-rt/runtime/io-stmt.h
@@ -84,6 +84,9 @@ public:
   // This design avoids virtual member functions and function pointers,
   // which may not have good support in some runtime environments.
 
+  RT_API_ATTRS const NonTbpDefinedIoTable *nonTbpDefinedIoTable() const;
+  RT_API_ATTRS void set_nonTbpDefinedIoTable(const NonTbpDefinedIoTable *);
+
   // CompleteOperation() is the last opportunity to raise an I/O error.
   // It is called by EndIoStatement(), but it can be invoked earlier to
   // catch errors for (e.g.) GetIoMsg() and GetNewUnit().  If called
@@ -363,6 +366,13 @@ public:
   using IoErrorHandler::IoErrorHandler;
 
   RT_API_ATTRS bool completedOperation() const { return completedOperation_; }
+  RT_API_ATTRS const NonTbpDefinedIoTable *nonTbpDefinedIoTable() const {
+    return nonTbpDefinedIoTable_;
+  }
+  RT_API_ATTRS void set_nonTbpDefinedIoTable(
+      const NonTbpDefinedIoTable *table) {
+    nonTbpDefinedIoTable_ = table;
+  }
 
   RT_API_ATTRS void CompleteOperation() { completedOperation_ = true; }
   RT_API_ATTRS int EndIoStatement() { return GetIoStat(); }
@@ -397,6 +407,11 @@ public:
 
 protected:
   bool completedOperation_{false};
+
+private:
+  // Original NonTbpDefinedIoTable argument to Input/OutputDerivedType,
+  // saved here so that it can also be used in child I/O statements.
+  const NonTbpDefinedIoTable *nonTbpDefinedIoTable_{nullptr};
 };
 
 // Common state for list-directed & NAMELIST I/O, both internal & external
@@ -630,8 +645,10 @@ class ChildIoStatementState : public IoStatementBase,
 public:
   RT_API_ATTRS ChildIoStatementState(
       ChildIo &, const char *sourceFile = nullptr, int sourceLine = 0);
+  RT_API_ATTRS const NonTbpDefinedIoTable *nonTbpDefinedIoTable() const;
+  RT_API_ATTRS void set_nonTbpDefinedIoTable(const NonTbpDefinedIoTable *);
   RT_API_ATTRS ChildIo &child() { return child_; }
-  RT_API_ATTRS MutableModes &mutableModes();
+  RT_API_ATTRS MutableModes &mutableModes() { return mutableModes_; }
   RT_API_ATTRS ConnectionState &GetConnectionState();
   RT_API_ATTRS ExternalFileUnit *GetExternalFileUnit() const;
   RT_API_ATTRS int EndIoStatement();
@@ -644,6 +661,7 @@ public:
 
 private:
   ChildIo &child_;
+  MutableModes mutableModes_;
 };
 
 template <Direction DIR, typename CHAR>
@@ -654,7 +672,6 @@ public:
   RT_API_ATTRS ChildFormattedIoStatementState(ChildIo &, const CharType *format,
       std::size_t formatLength, const Descriptor *formatDescriptor = nullptr,
       const char *sourceFile = nullptr, int sourceLine = 0);
-  RT_API_ATTRS MutableModes &mutableModes() { return mutableModes_; }
   RT_API_ATTRS void CompleteOperation();
   RT_API_ATTRS int EndIoStatement();
   RT_API_ATTRS bool AdvanceRecord(int = 1);
@@ -664,7 +681,6 @@ public:
   }
 
 private:
-  MutableModes mutableModes_;
   FormatControl<ChildFormattedIoStatementState> format_;
 };
 
@@ -840,7 +856,7 @@ private:
 };
 
 class InquireIOLengthState : public NoUnitIoStatementState,
-                             public OutputStatementState {
+                             public IoDirectionState<Direction::Output> {
 public:
   RT_API_ATTRS InquireIOLengthState(
       const char *sourceFile = nullptr, int sourceLine = 0);
diff --git a/flang-rt/lib/runtime/descriptor-io.cpp b/flang-rt/lib/runtime/descriptor-io.cpp
index 3868c8d..e22fc79 100644
--- a/flang-rt/lib/runtime/descriptor-io.cpp
+++ b/flang-rt/lib/runtime/descriptor-io.cpp
@@ -47,15 +47,19 @@ static RT_API_ATTRS Fortran::common::optional<bool> DefinedFormattedIo(
     const typeInfo::DerivedType &derived,
     const typeInfo::SpecialBinding &special,
     const SubscriptValue subscripts[]) {
-  Fortran::common::optional<DataEdit> peek{
-      io.GetNextDataEdit(0 /*to peek at it*/)};
+  // Look at the next data edit descriptor.  If this is list-directed I/O, the
+  // "maxRepeat=0" argument will prevent the input from advancing over an
+  // initial '(' that shouldn't be consumed now as the start of a real part.
+  Fortran::common::optional<DataEdit> peek{io.GetNextDataEdit(/*maxRepeat=*/0)};
   if (peek &&
       (peek->descriptor == DataEdit::DefinedDerivedType ||
-          peek->descriptor == DataEdit::ListDirected)) {
+          peek->descriptor == DataEdit::ListDirected ||
+          peek->descriptor == DataEdit::ListDirectedRealPart)) {
     // Defined formatting
     IoErrorHandler &handler{io.GetIoErrorHandler()};
-    DataEdit edit{*io.GetNextDataEdit(1)}; // now consume it; no repeats
-    RUNTIME_CHECK(handler, edit.descriptor == peek->descriptor);
+    DataEdit edit{peek->descriptor == DataEdit::ListDirectedRealPart
+            ? *peek
+            : *io.GetNextDataEdit(1)};
     char ioType[2 + edit.maxIoTypeChars];
     auto ioTypeLen{std::size_t{2} /*"DT"*/ + edit.ioTypeChars};
     if (edit.descriptor == DataEdit::DefinedDerivedType) {
@@ -836,13 +840,23 @@ template RT_API_ATTRS int DescriptorIoTicket<Direction::Input>::Continue(
 
 template <Direction DIR>
 RT_API_ATTRS bool DescriptorIO(IoStatementState &io,
-    const Descriptor &descriptor, const NonTbpDefinedIoTable *table) {
+    const Descriptor &descriptor, const NonTbpDefinedIoTable *originalTable) {
   bool anyIoTookPlace{false};
+  const NonTbpDefinedIoTable *defaultTable{io.nonTbpDefinedIoTable()};
+  const NonTbpDefinedIoTable *table{originalTable};
+  if (!table) {
+    table = defaultTable;
+  } else if (table != defaultTable) {
+    io.set_nonTbpDefinedIoTable(table); // for nested I/O
+  }
   WorkQueue workQueue{io.GetIoErrorHandler()};
   if (workQueue.BeginDescriptorIo<DIR>(io, descriptor, table, anyIoTookPlace) ==
       StatContinue) {
     workQueue.Run();
   }
+  if (defaultTable != table) {
+    io.set_nonTbpDefinedIoTable(defaultTable);
+  }
   return anyIoTookPlace;
 }
 
diff --git a/flang-rt/lib/runtime/edit-input.cpp b/flang-rt/lib/runtime/edit-input.cpp
index 1355767..0c2a4cc 100644
--- a/flang-rt/lib/runtime/edit-input.cpp
+++ b/flang-rt/lib/runtime/edit-input.cpp
@@ -37,9 +37,7 @@ static RT_API_ATTRS bool CheckCompleteListDirectedField(
   if (edit.IsListDirected()) {
     std::size_t byteCount;
     if (auto ch{io.GetCurrentChar(byteCount)}) {
-      if (IsCharValueSeparator(edit, *ch)) {
-        return true;
-      } else {
+      if (!IsCharValueSeparator(edit, *ch)) {
         const auto &connection{io.GetConnectionState()};
         io.GetIoErrorHandler().SignalError(IostatBadListDirectedInputSeparator,
             "invalid character (0x%x) after list-directed input value, "
@@ -49,12 +47,9 @@ static RT_API_ATTRS bool CheckCompleteListDirectedField(
             static_cast<int>(connection.currentRecordNumber));
         return false;
       }
-    } else {
-      return true; // end of record: ok
     }
-  } else {
-    return true;
   }
+  return true;
 }
 
 template <int LOG2_BASE>
@@ -349,8 +344,8 @@ static RT_API_ATTRS ScannedRealInput ScanRealInput(
   }
   bool bzMode{(edit.modes.editingFlags & blankZero) != 0};
   int exponent{0};
-  if (!next || (!bzMode && *next == ' ') ||
-      (!(edit.modes.editingFlags & decimalComma) && *next == ',')) {
+  const char32_t comma{GetSeparatorChar(edit)};
+  if (!next || (!bzMode && *next == ' ') || *next == comma) {
     if (!edit.IsListDirected() && !io.GetConnectionState().IsAtEOF()) {
       // An empty/blank field means zero when not list-directed.
       // A fixed-width field containing only a sign is also zero;
@@ -375,21 +370,37 @@ static RT_API_ATTRS ScannedRealInput ScanRealInput(
         Put(*next);
       }
     }
-    if (next && *next == '(') { // NaN(...)
-      Put('(');
-      int depth{1};
-      while (true) {
-        next = io.NextInField(remaining, edit);
-        if (depth == 0) {
-          break;
-        } else if (!next) {
-          return {}; // error
-        } else if (*next == '(') {
-          ++depth;
-        } else if (*next == ')') {
-          --depth;
+    if (first == 'N' && (!next || *next == '(') &&
+        remaining.value_or(1) > 0) { // NaN(...)?
+      std::size_t byteCount{0};
+      if (!next) { // NextInField won't return '(' for list-directed
+        next = io.GetCurrentChar(byteCount);
+      }
+      if (next && *next == '(') {
+        int depth{1};
+        while (true) {
+          if (*next >= 'a' && *next <= 'z') {
+            *next = *next - 'a' + 'A';
+          }
+          Put(*next);
+          io.HandleRelativePosition(byteCount);
+          io.GotChar(byteCount);
+          if (remaining) {
+            *remaining -= byteCount;
+          }
+          if (depth == 0) {
+            break; // done
+          }
+          next = io.GetCurrentChar(byteCount);
+          if (!next || remaining.value_or(1) < 1) {
+            return {}; // error
+          } else if (*next == '(') {
+            ++depth;
+          } else if (*next == ')') {
+            --depth;
+          }
         }
-        Put(*next);
+        next = io.NextInField(remaining, edit);
       }
     }
   } else if (first == radixPointChar || (first >= '0' && first <= '9') ||
@@ -521,9 +532,11 @@ static RT_API_ATTRS ScannedRealInput ScanRealInput(
       io.SkipSpaces(remaining);
       next = io.NextInField(remaining, edit);
     }
-    if (!next) { // NextInField fails on separators like ')'
+    if (!next || *next == ')') { // NextInField fails on separators like ')'
       std::size_t byteCount{0};
-      next = io.GetCurrentChar(byteCount);
+      if (!next) {
+        next = io.GetCurrentChar(byteCount);
+      }
       if (next && *next == ')') {
         io.HandleRelativePosition(byteCount);
       }
@@ -532,7 +545,7 @@ static RT_API_ATTRS ScannedRealInput ScanRealInput(
     while (next && (*next == ' ' || *next == '\t')) {
       next = io.NextInField(remaining, edit);
     }
-    if (next && (*next != ',' || (edit.modes.editingFlags & decimalComma))) {
+    if (next && *next != comma) {
       return {}; // error: unused nonblank character in fixed-width field
     }
   }
diff --git a/flang-rt/lib/runtime/io-stmt.cpp b/flang-rt/lib/runtime/io-stmt.cpp
index 8056c8d..5667d67 100644
--- a/flang-rt/lib/runtime/io-stmt.cpp
+++ b/flang-rt/lib/runtime/io-stmt.cpp
@@ -526,6 +526,17 @@ Fortran::common::optional<DataEdit> IoStatementState::GetNextDataEdit(int n) {
       [&](auto &x) { return x.get().GetNextDataEdit(*this, n); }, u_);
 }
 
+const NonTbpDefinedIoTable *IoStatementState::nonTbpDefinedIoTable() const {
+  return common::visit(
+      [&](auto &x) { return x.get().nonTbpDefinedIoTable(); }, u_);
+}
+
+void IoStatementState::set_nonTbpDefinedIoTable(
+    const NonTbpDefinedIoTable *table) {
+  common::visit(
+      [&](auto &x) { return x.get().set_nonTbpDefinedIoTable(table); }, u_);
+}
+
 bool IoStatementState::Emit(
     const char *data, std::size_t bytes, std::size_t elementBytes) {
   return common::visit(
@@ -633,10 +644,10 @@ IoStatementState::FastAsciiField IoStatementState::GetUpcomingFastAsciiField() {
   if (!connection.isUTF8 && connection.internalIoCharKind <= 1) {
     const char *p{nullptr};
     if (std::size_t bytes{GetNextInputBytes(p)}) {
-      return FastAsciiField(connection, p, bytes);
+      return FastAsciiField{connection, p, bytes};
     }
   }
-  return FastAsciiField(connection);
+  return FastAsciiField{connection};
 }
 
 Fortran::common::optional<char32_t> IoStatementState::NextInField(
@@ -920,9 +931,12 @@ ListDirectedStatementState<Direction::Input>::GetNextDataEdit(
       fastField.connection().positionInRecord = start;
     }
   }
-  if (!imaginaryPart_ && ch && *ch == '(') {
-    realPart_ = true;
-    fastField.connection().HandleRelativePosition(byteCount);
+  if (!imaginaryPart_ && edit.descriptor == DataEdit::ListDirected && ch &&
+      *ch == '(') {
+    if (maxRepeat > 0) { // not being peeked at fram DefinedFormattedIo()
+      realPart_ = true;
+      fastField.connection().HandleRelativePosition(byteCount);
+    }
     edit.descriptor = DataEdit::ListDirectedRealPart;
   }
   return edit;
@@ -952,12 +966,24 @@ bool ExternalUnformattedIoStatementState<DIR>::Receive(
 template <Direction DIR>
 ChildIoStatementState<DIR>::ChildIoStatementState(
     ChildIo &child, const char *sourceFile, int sourceLine)
-    : IoStatementBase{sourceFile, sourceLine}, child_{child} {}
+    : IoStatementBase{sourceFile, sourceLine}, child_{child},
+      mutableModes_{child.parent().mutableModes()} {}
+
+template <Direction DIR>
+const NonTbpDefinedIoTable *
+ChildIoStatementState<DIR>::nonTbpDefinedIoTable() const {
+#if !defined(RT_DEVICE_AVOID_RECURSION)
+  return child_.parent().nonTbpDefinedIoTable();
+#else
+  ReportUnsupportedChildIo();
+#endif
+}
 
 template <Direction DIR>
-MutableModes &ChildIoStatementState<DIR>::mutableModes() {
+void ChildIoStatementState<DIR>::set_nonTbpDefinedIoTable(
+    const NonTbpDefinedIoTable *table) {
 #if !defined(RT_DEVICE_AVOID_RECURSION)
-  return child_.parent().mutableModes();
+  child_.parent().set_nonTbpDefinedIoTable(table);
 #else
   ReportUnsupportedChildIo();
 #endif
@@ -1030,9 +1056,7 @@ ChildFormattedIoStatementState<DIR, CHAR>::ChildFormattedIoStatementState(
     ChildIo &child, const CHAR *format, std::size_t formatLength,
     const Descriptor *formatDescriptor, const char *sourceFile, int sourceLine)
     : ChildIoStatementState<DIR>{child, sourceFile, sourceLine},
-      mutableModes_{child.parent().mutableModes()}, format_{*this, format,
-                                                        formatLength,
-                                                        formatDescriptor} {}
+      format_{*this, format, formatLength, formatDescriptor} {}
 
 template <Direction DIR, typename CHAR>
 void ChildFormattedIoStatementState<DIR, CHAR>::CompleteOperation() {
diff --git a/flang-rt/test/NonGtestUnit/lit.cfg.py b/flang-rt/test/NonGtestUnit/lit.cfg.py
index 4bee709..8da36ad 100644
--- a/flang-rt/test/NonGtestUnit/lit.cfg.py
+++ b/flang-rt/test/NonGtestUnit/lit.cfg.py
@@ -8,8 +8,7 @@ import lit.formats
 config.name = "flang-rt-OldUnit"
 
 # suffixes: A list of file extensions to treat as test files.
-# On Windows, ".exe" also matches the GTests and will execited redundantly.
-config.suffixes = [".test", ".exe"]
+config.suffixes = [".test", ".test.exe"]
 
 # test_source_root: The root path where unit test binaries are located.
 config.test_source_root = os.path.join(config.flangrt_binary_dir, "unittests")
diff --git a/flang-rt/unittests/CMakeLists.txt b/flang-rt/unittests/CMakeLists.txt
index 5282196..831bc8a 100644
--- a/flang-rt/unittests/CMakeLists.txt
+++ b/flang-rt/unittests/CMakeLists.txt
@@ -60,6 +60,27 @@ function(add_flangrt_unittest_offload_properties target)
   endif()
 endfunction()
 
+# flang-rt on Windows requires compiler-rt for some symbols. For binaries built
+# with flang this dependency is added by the flang driver, but since the unit
+# tests are built with clang we need to add the dependency manually.
+function(add_flangrt_dependent_libs target)
+  if (MSVC AND CMAKE_CXX_COMPILER_ID MATCHES "Clang")
+    if (FLANG_RT_BUILTINS_LIBRARY)
+      target_compile_options(${target} PRIVATE "$<$<COMPILE_LANGUAGE:CXX,C>:-Xclang>" "$<$<COMPILE_LANGUAGE:CXX,C>:--dependent-lib=${FLANG_RT_BUILTINS_LIBRARY}>")
+    endif ()
+  endif ()
+  if (MSVC AND CMAKE_Fortran_COMPILER_ID STREQUAL "LLVMFlang")
+    if (FLANG_RT_BUILTINS_LIBRARY)
+      target_compile_options(${target} PRIVATE "$<$<COMPILE_LANGUAGE:Fortran>:-Xflang>" "$<$<COMPILE_LANGUAGE:Fortran>:--dependent-lib=${FLANG_RT_BUILTINS_LIBRARY}>")
+    else ()
+      message(WARNING "Did not find libclang_rt.builtins.lib.
+        LLVM may emit builtins that are not implemented in msvcrt/ucrt and
+        instead falls back to builtins from Compiler-RT. Linking with ${tgtname}
+        may result in a linker error.")
+    endif ()
+  endif ()
+endfunction()
+
 
 function(add_flangrt_unittest test_dirname)
   cmake_parse_arguments(ARG
@@ -72,6 +93,7 @@ function(add_flangrt_unittest test_dirname)
 
   target_link_libraries(${test_dirname} PRIVATE ${ARG_LINK_LIBS})
   add_flangrt_unittest_offload_properties(${test_dirname})
+  add_flangrt_dependent_libs(${test_dirname})
 
   # Required because LLVMSupport is compiled with this option.
   # FIXME: According to CMake documentation, this is the default. Why is it
@@ -99,6 +121,7 @@ function(add_flangrt_nongtest_unittest test_name)
   set_target_properties(${test_name}${suffix} PROPERTIES FOLDER "Flang-RT/Tests/Unit")
 
   target_link_libraries(${test_name}${suffix} PRIVATE NonGTestTesting ${ARG_LINK_LIBS})
+  add_flangrt_dependent_libs(${test_name}${suffix})
 
   if(NOT ARG_SLOW_TEST)
     add_dependencies(FlangRTUnitTests ${test_name}${suffix})
diff --git a/flang-rt/unittests/Runtime/NumericalFormatTest.cpp b/flang-rt/unittests/Runtime/NumericalFormatTest.cpp
index 73245dc..4e5fdcc 100644
--- a/flang-rt/unittests/Runtime/NumericalFormatTest.cpp
+++ b/flang-rt/unittests/Runtime/NumericalFormatTest.cpp
@@ -921,6 +921,7 @@ TEST(IOApiTests, EditDoubleInputValues) {
       {"(BZ,F18.0)", "           .      ", 0x0, 0},
       {"(BZ,F18.0)", "           . e +1 ", 0x0, 0},
       {"(DC,F18.0)", "              12,5", 0x4029000000000000, 0},
+      {"(DC,F18.0)", "             12,5;", 0x4029000000000000, 0},
       {"(EX22.0)", "0X0P0                 ", 0x0, 0}, // +0.
       {"(EX22.0)", "-0X0P0                ", 0x8000000000000000, 0}, // -0.
       {"(EX22.0)", "0X.8P1                ", 0x3ff0000000000000, 0}, // 1.0
@@ -964,6 +965,7 @@ TEST(IOApiTests, EditDoubleInputValues) {
       {"(RU,E9.1)", " 1.0E-325", 0x1, 0},
       {"(E9.1)", "-1.0E-325", 0x8000000000000000, 0},
       {"(RD,E9.1)", "-1.0E-325", 0x8000000000000001, 0},
+      {"(F7.0))", "+NaN(q)", 0x7ff8000000000000, 0},
   };
   for (auto const &[format, data, want, iostat] : testCases) {
     auto cookie{IONAME(BeginInternalFormattedInput)(
diff --git a/flang/docs/F202X.md b/flang/docs/F202X.md
index 67ea7fd..7504561 100644
--- a/flang/docs/F202X.md
+++ b/flang/docs/F202X.md
@@ -268,7 +268,6 @@ Addressing some issues and omissions in intrinsic modules:
  * LOGICAL8/16/32/64 and REAL16
  * IEEE module facilities upgraded to match latest IEEE FP standard
  * C_F_STRPOINTER, F_C_STRING for NUL-terminated strings
- * C_F_POINTER(LOWER=)
 
 #### Intrinsic Procedure Extensions
 
diff --git a/flang/docs/ReleaseNotes.md b/flang/docs/ReleaseNotes.md
index f0c9562..99dc41c 100644
--- a/flang/docs/ReleaseNotes.md
+++ b/flang/docs/ReleaseNotes.md
@@ -24,6 +24,8 @@ page](https://llvm.org/releases/).
 
 ## Major New Features
 
+* Support for LOWER= argument for C_F_POINTER.
+
 ## Bug Fixes
 
 ## Non-comprehensive list of changes in this release
diff --git a/flang/include/flang/Common/target-rounding.h b/flang/include/flang/Common/target-rounding.h
index f503b22..9041d74 100644
--- a/flang/include/flang/Common/target-rounding.h
+++ b/flang/include/flang/Common/target-rounding.h
@@ -21,7 +21,7 @@ struct Rounding {
   // (viz., fail to set the Underflow flag when an inexact product of a
   // multiplication is rounded up to a normal number from a subnormal
   // in some rounding modes)
-#if __x86_64__ || __riscv || __loongarch__
+#if __x86_64__ || _M_X64 || __riscv || __loongarch__
   bool x86CompatibleBehavior{true};
 #else
   bool x86CompatibleBehavior{false};
diff --git a/flang/include/flang/Testing/fp-testing.h b/flang/include/flang/Testing/fp-testing.h
index c65766b..5d7de3c 100644
--- a/flang/include/flang/Testing/fp-testing.h
+++ b/flang/include/flang/Testing/fp-testing.h
@@ -27,7 +27,7 @@ public:
 
 private:
   fenv_t originalFenv_;
-#if __x86_64__
+#if __x86_64__ || _M_X64
   unsigned int originalMxcsr;
 #endif
 };
diff --git a/flang/lib/Evaluate/intrinsics.cpp b/flang/lib/Evaluate/intrinsics.cpp
index 768e4ba..c37a7f90 100644
--- a/flang/lib/Evaluate/intrinsics.cpp
+++ b/flang/lib/Evaluate/intrinsics.cpp
@@ -3077,10 +3077,11 @@ IntrinsicProcTable::Implementation::HandleC_F_Pointer(
     ActualArguments &arguments, FoldingContext &context) const {
   characteristics::Procedure::Attrs attrs;
   attrs.set(characteristics::Procedure::Attr::Subroutine);
-  static const char *const keywords[]{"cptr", "fptr", "shape", nullptr};
+  static const char *const keywords[]{
+      "cptr", "fptr", "shape", "lower", nullptr};
   characteristics::DummyArguments dummies;
-  if (CheckAndRearrangeArguments(arguments, context.messages(), keywords, 1)) {
-    CHECK(arguments.size() == 3);
+  if (CheckAndRearrangeArguments(arguments, context.messages(), keywords, 2)) {
+    CHECK(arguments.size() == 4);
     if (const auto *expr{arguments[0].value().UnwrapExpr()}) {
       // General semantic checks will catch an actual argument that's not
       // scalar.
@@ -3173,11 +3174,30 @@ IntrinsicProcTable::Implementation::HandleC_F_Pointer(
           }
         }
       }
+      if (arguments[3] && fptrRank == 0) {
+        context.messages().Say(arguments[3]->sourceLocation(),
+            "LOWER= argument to C_F_POINTER() may not appear when FPTR= is scalar"_err_en_US);
+      } else if (arguments[3]) {
+        if (const auto *argExpr{arguments[3].value().UnwrapExpr()}) {
+          if (argExpr->Rank() > 1) {
+            context.messages().Say(arguments[3]->sourceLocation(),
+                "LOWER= argument to C_F_POINTER() must be a rank-one array."_err_en_US);
+          } else if (argExpr->Rank() == 1) {
+            if (auto constShape{GetConstantShape(context, *argExpr)}) {
+              if (constShape->At(ConstantSubscripts{1}).ToInt64() != fptrRank) {
+                context.messages().Say(arguments[3]->sourceLocation(),
+                    "LOWER= argument to C_F_POINTER() must have size equal to the rank of FPTR="_err_en_US);
+              }
+            }
+          }
+        }
+      }
     }
   }
   if (dummies.size() == 2) {
+    // Handle SHAPE
     DynamicType shapeType{TypeCategory::Integer, defaults_.sizeIntegerKind()};
-    if (arguments[2]) {
+    if (arguments.size() >= 3 && arguments[2]) {
       if (auto type{arguments[2]->GetType()}) {
         if (type->category() == TypeCategory::Integer) {
           shapeType = *type;
@@ -3189,6 +3209,22 @@ IntrinsicProcTable::Implementation::HandleC_F_Pointer(
     shape.intent = common::Intent::In;
     shape.attrs.set(characteristics::DummyDataObject::Attr::Optional);
     dummies.emplace_back("shape"s, std::move(shape));
+
+    // Handle LOWER
+    DynamicType lowerType{TypeCategory::Integer, defaults_.sizeIntegerKind()};
+    if (arguments.size() >= 4 && arguments[3]) {
+      if (auto type{arguments[3]->GetType()}) {
+        if (type->category() == TypeCategory::Integer) {
+          lowerType = *type;
+        }
+      }
+    }
+    characteristics::DummyDataObject lower{
+        characteristics::TypeAndShape{lowerType, 1}};
+    lower.intent = common::Intent::In;
+    lower.attrs.set(characteristics::DummyDataObject::Attr::Optional);
+    dummies.emplace_back("lower"s, std::move(lower));
+
     return SpecificCall{
         SpecificIntrinsic{"__builtin_c_f_pointer"s,
             characteristics::Procedure{std::move(dummies), attrs}},
diff --git a/flang/lib/Frontend/CMakeLists.txt b/flang/lib/Frontend/CMakeLists.txt
index 96ba27a..fa0d5ec 100644
--- a/flang/lib/Frontend/CMakeLists.txt
+++ b/flang/lib/Frontend/CMakeLists.txt
@@ -45,6 +45,7 @@ add_flang_library(flangFrontend
   LINK_COMPONENTS
   Passes
   Analysis
+  BitReader
   Extensions
   IRPrinter
   IRReader
diff --git a/flang/lib/Optimizer/Builder/IntrinsicCall.cpp b/flang/lib/Optimizer/Builder/IntrinsicCall.cpp
index 4753d0a..e62ed48 100644
--- a/flang/lib/Optimizer/Builder/IntrinsicCall.cpp
+++ b/flang/lib/Optimizer/Builder/IntrinsicCall.cpp
@@ -371,7 +371,8 @@ static constexpr IntrinsicHandler handlers[]{
      &I::genCFPointer,
      {{{"cptr", asValue},
        {"fptr", asInquired},
-       {"shape", asAddr, handleDynamicOptional}}},
+       {"shape", asAddr, handleDynamicOptional},
+       {"lower", asAddr, handleDynamicOptional}}},
      /*isElemental=*/false},
     {"c_f_procpointer",
      &I::genCFProcPointer,
@@ -3438,7 +3439,7 @@ IntrinsicLibrary::genCDevLoc(mlir::Type resultType,
 
 // C_F_POINTER
 void IntrinsicLibrary::genCFPointer(llvm::ArrayRef<fir::ExtendedValue> args) {
-  assert(args.size() == 3);
+  assert(args.size() == 4);
   // Handle CPTR argument
   // Get the value of the C address or the result of a reference to C_LOC.
   mlir::Value cPtr = fir::getBase(args[0]);
@@ -3453,9 +3454,12 @@ void IntrinsicLibrary::genCFPointer(llvm::ArrayRef<fir::ExtendedValue> args) {
     mlir::Value addr =
         builder.createConvert(loc, fPtr->getMemTy(), cPtrAddrVal);
     mlir::SmallVector<mlir::Value> extents;
+    mlir::SmallVector<mlir::Value> lbounds;
     if (box.hasRank()) {
       assert(isStaticallyPresent(args[2]) &&
              "FPTR argument must be an array if SHAPE argument exists");
+
+      // Handle and unpack SHAPE argument
       mlir::Value shape = fir::getBase(args[2]);
       int arrayRank = box.rank();
       mlir::Type shapeElementType =
@@ -3468,17 +3472,31 @@ void IntrinsicLibrary::genCFPointer(llvm::ArrayRef<fir::ExtendedValue> args) {
         mlir::Value load = fir::LoadOp::create(builder, loc, var);
         extents.push_back(builder.createConvert(loc, idxType, load));
       }
+
+      // Handle and unpack LOWER argument if present
+      if (isStaticallyPresent(args[3])) {
+        mlir::Value lower = fir::getBase(args[3]);
+        mlir::Type lowerElementType =
+            fir::unwrapSequenceType(fir::unwrapPassByRefType(lower.getType()));
+        for (int i = 0; i < arrayRank; ++i) {
+          mlir::Value index = builder.createIntegerConstant(loc, idxType, i);
+          mlir::Value var = builder.create<fir::CoordinateOp>(
+              loc, builder.getRefType(lowerElementType), lower, index);
+          mlir::Value load = builder.create<fir::LoadOp>(loc, var);
+          lbounds.push_back(builder.createConvert(loc, idxType, load));
+        }
+      }
     }
     if (box.isCharacter()) {
       mlir::Value len = box.nonDeferredLenParams()[0];
       if (box.hasRank())
-        return fir::CharArrayBoxValue{addr, len, extents};
+        return fir::CharArrayBoxValue{addr, len, extents, lbounds};
       return fir::CharBoxValue{addr, len};
     }
     if (box.isDerivedWithLenParameters())
       TODO(loc, "get length parameters of derived type");
     if (box.hasRank())
-      return fir::ArrayBoxValue{addr, extents};
+      return fir::ArrayBoxValue{addr, extents, lbounds};
     return addr;
   };
 
diff --git a/flang/lib/Optimizer/Dialect/Support/KindMapping.cpp b/flang/lib/Optimizer/Dialect/Support/KindMapping.cpp
index 30c6030..6cf8adb 100644
--- a/flang/lib/Optimizer/Dialect/Support/KindMapping.cpp
+++ b/flang/lib/Optimizer/Dialect/Support/KindMapping.cpp
@@ -12,6 +12,7 @@
 
 #include "flang/Optimizer/Dialect/Support/KindMapping.h"
 #include "mlir/Dialect/LLVMIR/LLVMDialect.h"
+#include "llvm/IR/LLVMContext.h"
 #include "llvm/Support/CommandLine.h"
 
 /// Allow the user to set the FIR intrinsic type kind value to LLVM type
diff --git a/flang/lib/Optimizer/OpenMP/MapInfoFinalization.cpp b/flang/lib/Optimizer/OpenMP/MapInfoFinalization.cpp
index 03ca3d8..57be863 100644
--- a/flang/lib/Optimizer/OpenMP/MapInfoFinalization.cpp
+++ b/flang/lib/Optimizer/OpenMP/MapInfoFinalization.cpp
@@ -137,32 +137,41 @@ class MapInfoFinalizationPass
         !fir::factory::isOptionalArgument(descriptor.getDefiningOp()))
       return descriptor;
 
-    mlir::Value &slot = localBoxAllocas[descriptor.getDefiningOp()];
-    if (slot) {
-      return slot;
+    mlir::Value &alloca = localBoxAllocas[descriptor.getDefiningOp()];
+    mlir::Location loc = boxMap->getLoc();
+
+    if (!alloca) {
+      // The fir::BoxOffsetOp only works with !fir.ref<!fir.box<...>> types, as
+      // allowing it to access non-reference box operations can cause some
+      // problematic SSA IR. However, in the case of assumed shape's the type
+      // is not a !fir.ref, in these cases to retrieve the appropriate
+      // !fir.ref<!fir.box<...>> to access the data we need to map we must
+      // perform an alloca and then store to it and retrieve the data from the
+      // new alloca.
+      mlir::OpBuilder::InsertPoint insPt = builder.saveInsertionPoint();
+      mlir::Block *allocaBlock = builder.getAllocaBlock();
+      assert(allocaBlock && "No alloca block found for this top level op");
+      builder.setInsertionPointToStart(allocaBlock);
+
+      mlir::Type allocaType = descriptor.getType();
+      if (fir::isBoxAddress(allocaType))
+        allocaType = fir::unwrapRefType(allocaType);
+      alloca = fir::AllocaOp::create(builder, loc, allocaType);
+      builder.restoreInsertionPoint(insPt);
     }
 
-    // The fir::BoxOffsetOp only works with !fir.ref<!fir.box<...>> types, as
-    // allowing it to access non-reference box operations can cause some
-    // problematic SSA IR. However, in the case of assumed shape's the type
-    // is not a !fir.ref, in these cases to retrieve the appropriate
-    // !fir.ref<!fir.box<...>> to access the data we need to map we must
-    // perform an alloca and then store to it and retrieve the data from the new
-    // alloca.
-    mlir::OpBuilder::InsertPoint insPt = builder.saveInsertionPoint();
-    mlir::Block *allocaBlock = builder.getAllocaBlock();
-    mlir::Location loc = boxMap->getLoc();
-    assert(allocaBlock && "No alloca block found for this top level op");
-    builder.setInsertionPointToStart(allocaBlock);
-
-    mlir::Type allocaType = descriptor.getType();
-    if (fir::isBoxAddress(allocaType))
-      allocaType = fir::unwrapRefType(allocaType);
-    auto alloca = fir::AllocaOp::create(builder, loc, allocaType);
-    builder.restoreInsertionPoint(insPt);
     // We should only emit a store if the passed in data is present, it is
     // possible a user passes in no argument to an optional parameter, in which
     // case we cannot store or we'll segfault on the emitted memcpy.
+    // TODO: We currently emit a present -> load/store every time we use a
+    // mapped value that requires a local allocation, this isn't the most
+    // efficient, although, it is more correct in a lot of situations. One
+    // such situation is emitting a this series of instructions in separate
+    // segments of a branch (e.g. two target regions in separate else/if branch
+    // mapping the same function argument), however, it would be nice to be able
+    // to optimize these situations e.g. raising the load/store out of the
+    // branch if possible. But perhaps this is best left to lower level
+    // optimisation passes.
     auto isPresent =
         fir::IsPresentOp::create(builder, loc, builder.getI1Type(), descriptor);
     builder.genIfOp(loc, {}, isPresent, false)
@@ -171,7 +180,7 @@ class MapInfoFinalizationPass
           fir::StoreOp::create(builder, loc, descriptor, alloca);
         })
         .end();
-    return slot = alloca;
+    return alloca;
   }
 
   /// Function that generates a FIR operation accessing the descriptor's
diff --git a/flang/lib/Parser/openmp-parsers.cpp b/flang/lib/Parser/openmp-parsers.cpp
index 25a692d..1c62614 100644
--- a/flang/lib/Parser/openmp-parsers.cpp
+++ b/flang/lib/Parser/openmp-parsers.cpp
@@ -1222,11 +1222,14 @@ struct StrictlyStructuredBlockParser {
   using resultType = Block;
 
   std::optional<resultType> Parse(ParseState &state) const {
-    if (auto epc{Parser<ExecutionPartConstruct>{}.Parse(state)}) {
-      if (IsFortranBlockConstruct(*epc)) {
-        Block block;
-        block.emplace_back(std::move(*epc));
-        return std::move(block);
+    // Detect BLOCK construct without parsing the entire thing.
+    if (lookAhead(skipStuffBeforeStatement >> "BLOCK"_tok).Parse(state)) {
+      if (auto epc{Parser<ExecutionPartConstruct>{}.Parse(state)}) {
+        if (IsFortranBlockConstruct(*epc)) {
+          Block block;
+          block.emplace_back(std::move(*epc));
+          return std::move(block);
+        }
       }
     }
     return std::nullopt;
@@ -1237,6 +1240,10 @@ struct LooselyStructuredBlockParser {
   using resultType = Block;
 
   std::optional<resultType> Parse(ParseState &state) const {
+    // Detect BLOCK construct without parsing the entire thing.
+    if (lookAhead(skipStuffBeforeStatement >> "BLOCK"_tok).Parse(state)) {
+      return std::nullopt;
+    }
     Block body;
     if (auto epc{attempt(Parser<ExecutionPartConstruct>{}).Parse(state)}) {
       if (!IsFortranBlockConstruct(*epc)) {
diff --git a/flang/lib/Semantics/resolve-names.cpp b/flang/lib/Semantics/resolve-names.cpp
index b326860..d08c669 100644
--- a/flang/lib/Semantics/resolve-names.cpp
+++ b/flang/lib/Semantics/resolve-names.cpp
@@ -2351,7 +2351,8 @@ bool AttrsVisitor::IsConflictingAttr(Attr attrName) {
       HaveAttrConflict(attrName, Attr::PASS, Attr::NOPASS) || // C781
       HaveAttrConflict(attrName, Attr::PURE, Attr::IMPURE) ||
       HaveAttrConflict(attrName, Attr::PUBLIC, Attr::PRIVATE) ||
-      HaveAttrConflict(attrName, Attr::RECURSIVE, Attr::NON_RECURSIVE);
+      HaveAttrConflict(attrName, Attr::RECURSIVE, Attr::NON_RECURSIVE) ||
+      HaveAttrConflict(attrName, Attr::INTRINSIC, Attr::EXTERNAL);
 }
 bool AttrsVisitor::CheckAndSet(Attr attrName) {
   if (IsConflictingAttr(attrName) || IsDuplicateAttr(attrName)) {
diff --git a/flang/lib/Testing/fp-testing.cpp b/flang/lib/Testing/fp-testing.cpp
index 5e1728e..56335f1 100644
--- a/flang/lib/Testing/fp-testing.cpp
+++ b/flang/lib/Testing/fp-testing.cpp
@@ -11,7 +11,7 @@
 #include <cstdio>
 #include <cstdlib>
 #include <cstring>
-#if __x86_64__
+#if __x86_64__ || _M_X64
 #include <xmmintrin.h>
 #endif
 
@@ -19,7 +19,7 @@ using Fortran::common::RealFlag;
 using Fortran::common::RoundingMode;
 
 ScopedHostFloatingPointEnvironment::ScopedHostFloatingPointEnvironment(
-#if __x86_64__
+#if __x86_64__ || _M_X64
     bool treatSubnormalOperandsAsZero, bool flushSubnormalResultsToZero
 #else
     bool, bool
@@ -38,7 +38,7 @@ ScopedHostFloatingPointEnvironment::ScopedHostFloatingPointEnvironment(
     std::abort();
   }
 
-#if __x86_64__
+#if __x86_64__ || _M_X64
   originalMxcsr = _mm_getcsr();
   unsigned int currentMxcsr{originalMxcsr};
   if (treatSubnormalOperandsAsZero) {
@@ -72,7 +72,7 @@ ScopedHostFloatingPointEnvironment::~ScopedHostFloatingPointEnvironment() {
         stderr, "fesetenv() failed: %s\n", llvm::sys::StrError(errno).c_str());
     std::abort();
   }
-#if __x86_64__
+#if __x86_64__ || _M_X64
   _mm_setcsr(originalMxcsr);
 #endif
 }
diff --git a/flang/test/CMakeLists.txt b/flang/test/CMakeLists.txt
index 8520bec..40f4a2e 100644
--- a/flang/test/CMakeLists.txt
+++ b/flang/test/CMakeLists.txt
@@ -65,6 +65,7 @@ set(FLANG_TEST_DEPENDS
   bbc
   FortranDecimal
 )
+
 if (NOT FLANG_STANDALONE_BUILD)
   list(APPEND FLANG_TEST_DEPENDS
     llvm-config
@@ -79,7 +80,7 @@ if (NOT FLANG_STANDALONE_BUILD)
   )
 endif ()
 
-if (LLVM_BUILD_EXAMPLES AND LLVM_ENABLE_PLUGINS AND NOT WIN32 AND NOT FLANG_STANDALONE_BUILD)
+if (LLVM_BUILD_EXAMPLES AND LLVM_ENABLE_PLUGINS AND NOT (WIN32 OR CYGWIN) AND NOT FLANG_STANDALONE_BUILD)
   list(APPEND FLANG_TEST_DEPENDS Bye)
 endif()
 
diff --git a/flang/test/Lower/Intrinsics/c_f_pointer.f90 b/flang/test/Lower/Intrinsics/c_f_pointer.f90
index 67817e3..c1f1d79 100644
--- a/flang/test/Lower/Intrinsics/c_f_pointer.f90
+++ b/flang/test/Lower/Intrinsics/c_f_pointer.f90
@@ -140,3 +140,78 @@ subroutine dynamic_shape_size_2(cptr, fptr, shape, n)
 ! CHECK:         %[[VAL_16:.*]] = fir.shape %[[VAL_11]], %[[VAL_15]] : (index, index) -> !fir.shape<2>
   call c_f_pointer(cptr, fptr, shape)
 end subroutine
+
+! CHECK-LABEL: func.func @_QPdynamic_shape_lower(
+subroutine dynamic_shape_lower(cptr, fpr, shape, lower)
+  use iso_c_binding
+  type(c_ptr)  :: cptr
+  real, pointer :: fptr(:, :)
+  integer :: n
+  integer :: shape(:)
+  integer :: lower(:)
+! CHECK: %[[C_0:.*]] = arith.constant 0 : index
+! CHECK: %[[VAL_2:.*]] = fir.shape %[[C_0]], %[[C_0]] : (index, index) -> !fir.shape<2>
+! CHECK: %[[VAL_3:.*]] = fir.embox %[[VAL_1:.*]](%[[VAL_2]]) : (!fir.ptr<!fir.array<?x?xf32>>, !fir.shape<2>) -> !fir.box<!fir.ptr<!fir.array<?x?xf32>>>
+! CHECK: fir.store %[[VAL_3]] to %[[VAL_0:.*]] : !fir.ref<!fir.box<!fir.ptr<!fir.array<?x?xf32>>>>
+! CHECK: %[[VAL_4:.*]] = fir.alloca i32 {bindc_name = "n", uniq_name = "_QFdynamic_shape_lowerEn"}
+! CHECK: %[[VAL_5:.*]] = fir.coordinate_of %[[ARG_0:.*]], __address : (!fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>) -> !fir.ref<i64>
+! CHECK: %[[VAL_6:.*]] = fir.load %[[VAL_5]] : !fir.ref<i64>
+! CHECK: %[[VAL_7:.*]] = fir.convert %[[VAL_6]] : (i64) -> !fir.ptr<!fir.array<?x?xf32>>
+! CHECK: %[[C_0:.*]]_0 = arith.constant 0 : index
+! CHECK: %[[VAL_8:.*]] = fir.coordinate_of %[[ARG_2:.*]], %[[C_0]]_0 : (!fir.box<!fir.array<?xi32>>, index) -> !fir.ref<i32>
+! CHECK: %[[VAL_9:.*]] = fir.load %[[VAL_8]] : !fir.ref<i32>
+! CHECK: %[[VAL_10:.*]] = fir.convert %[[VAL_9]] : (i32) -> index
+! CHECK: %[[C_1:.*]] = arith.constant 1 : index
+! CHECK: %[[VAL_11:.*]] = fir.coordinate_of %[[ARG_2:.*]], %[[C_1]] : (!fir.box<!fir.array<?xi32>>, index) -> !fir.ref<i32>
+! CHECK: %[[VAL_12:.*]] = fir.load %[[VAL_11]] : !fir.ref<i32>
+! CHECK: %[[VAL_13:.*]] = fir.convert %[[VAL_12]] : (i32) -> index
+! CHECK: %[[C_0:.*]]_1 = arith.constant 0 : index
+! CHECK: %[[VAL_14:.*]] = fir.coordinate_of %[[ARG_3:.*]], %[[C_0]]_1 : (!fir.box<!fir.array<?xi32>>, index) -> !fir.ref<i32>
+! CHECK: %[[VAL_15:.*]] = fir.load %[[VAL_14]] : !fir.ref<i32>
+! CHECK: %[[VAL_16:.*]] = fir.convert %[[VAL_15]] : (i32) -> index
+! CHECK: %[[C_1:.*]]_2 = arith.constant 1 : index
+! CHECK: %[[VAL_17:.*]] = fir.coordinate_of %[[ARG_3:.*]], %[[C_1]]_2 : (!fir.box<!fir.array<?xi32>>, index) -> !fir.ref<i32>
+! CHECK: %[[VAL_18:.*]] = fir.load %[[VAL_17]] : !fir.ref<i32>
+! CHECK: %[[VAL_19:.*]] = fir.convert %[[VAL_18]] : (i32) -> index
+! CHECK: %[[VAL_20:.*]] = fir.shape_shift %[[VAL_16]], %[[VAL_10]], %[[VAL_19]], %[[VAL_13]] : (index, index, index, index) -> !fir.shapeshift<2>
+! CHECK: %[[VAL_21:.*]] = fir.embox %[[VAL_7]](%[[VAL_20]]) : (!fir.ptr<!fir.array<?x?xf32>>, !fir.shapeshift<2>) -> !fir.box<!fir.ptr<!fir.array<?x?xf32>>>
+! CHECK: fir.store %[[VAL_21:.*]] to %[[VAL_0]] : !fir.ref<!fir.box<!fir.ptr<!fir.array<?x?xf32>>>>
+  call c_f_pointer(cptr, fptr, shape, lower)
+end subroutine dynamic_shape_lower
+
+! CHECK-LABEL: func.func @_QPdynamic_shape_lower_2(
+subroutine dynamic_shape_lower_2(cptr, fpr, shape, lower, n)
+  use iso_c_binding
+  type(c_ptr)  :: cptr
+  real, pointer :: fptr(:, :)
+  integer :: n
+  integer :: shape(n)
+  integer :: lower(n)
+!CHECK: %[[C_0:.*]] = arith.constant 0 : index
+!CHECK: %[[VAL_2:.*]] = fir.shape %[[C_0]], %[[C_0]] : (index, index) -> !fir.shape<2>
+!CHECK: %[[VAL_3:.*]] = fir.embox %[[ARG1:.*]](%[[VAL_2]]) : (!fir.ptr<!fir.array<?x?xf32>>, !fir.shape<2>) -> !fir.box<!fir.ptr<!fir.array<?x?xf32>>>
+!CHECK: fir.store %[[VAL_3]] to %[[VAL_0:.*]] : !fir.ref<!fir.box<!fir.ptr<!fir.array<?x?xf32>>>>
+!CHECK: %[[VAL_4:.*]] = fir.coordinate_of %[[ARG_0:.*]], __address : (!fir.ref<!fir.type<_QM__fortran_builtinsT__builtin_c_ptr{__address:i64}>>) -> !fir.ref<i64>
+!CHECK: %[[VAL_5:.*]] = fir.load %[[VAL_4]] : !fir.ref<i64>
+!CHECK: %[[VAL_6:.*]] = fir.convert %[[VAL_5]] : (i64) -> !fir.ptr<!fir.array<?x?xf32>>
+!CHECK: %[[C_0:.*]]_0 = arith.constant 0 : index
+!CHECK: %[[VAL_7:.*]] = fir.coordinate_of %[[ARG_2:.*]], %[[C_0]]_0 : (!fir.ref<!fir.array<?xi32>>, index) -> !fir.ref<i32>
+!CHECK: %[[VAL_8:.*]] = fir.load %[[VAL_7]] : !fir.ref<i32>
+!CHECK: %[[VAL_9:.*]] = fir.convert %[[VAL_8]] : (i32) -> index
+!CHECK: %[[C_1:.*]] = arith.constant 1 : index
+!CHECK: %[[VAL_10:.*]] = fir.coordinate_of %[[ARG_2]], %[[C_1]] : (!fir.ref<!fir.array<?xi32>>, index) -> !fir.ref<i32>
+!CHECK: %[[VAL_11:.*]] = fir.load %[[VAL_10]] : !fir.ref<i32>
+!CHECK: %[[VAL_12:.*]] = fir.convert %[[VAL_11]] : (i32) -> index
+!CHECK: %[[C_0:.*]]_1 = arith.constant 0 : index
+!CHECK: %[[VAL_13:.*]] = fir.coordinate_of %[[ARG_3:.*]], %[[C_0]]_1 : (!fir.ref<!fir.array<?xi32>>, index) -> !fir.ref<i32>
+!CHECK: %[[VAL_14:.*]] = fir.load %[[VAL_13]] : !fir.ref<i32>
+!CHECK: %[[VAL_15:.*]] = fir.convert %[[VAL_14]] : (i32) -> index
+!CHECK: %[[C_1:.*]]_2 = arith.constant 1 : index
+!CHECK: %[[VAL_16:.*]] = fir.coordinate_of %[[ARG_3]], %[[C_1]]_2 : (!fir.ref<!fir.array<?xi32>>, index) -> !fir.ref<i32>
+!CHECK: %[[VAL_17:.*]] = fir.load %[[VAL_16]] : !fir.ref<i32>
+!CHECK: %[[VAL_18:.*]] = fir.convert %[[VAL_17]] : (i32) -> index
+!CHECK: %[[VAL_19:.*]] = fir.shape_shift %[[VAL_15]], %[[VAL_9]], %[[VAL_18]], %[[VAL_12]] : (index, index, index, index) -> !fir.shapeshift<2>
+!CHECK: %[[VAL_20:.*]] = fir.embox %[[VAL_6]](%[[VAL_19]]) : (!fir.ptr<!fir.array<?x?xf32>>, !fir.shapeshift<2>)
+!CHECK: fir.store %[[VAL_20]] to %[[VAL_0]] : !fir.ref<!fir.box<!fir.ptr<!fir.array<?x?xf32>>>>
+  call c_f_pointer(cptr, fptr, shape, lower)
+end subroutine dynamic_shape_lower_2
diff --git a/flang/test/Lower/OpenMP/optional-argument-map-3.f90 b/flang/test/Lower/OpenMP/optional-argument-map-3.f90
new file mode 100644
index 0000000..7e2a24e
--- /dev/null
+++ b/flang/test/Lower/OpenMP/optional-argument-map-3.f90
@@ -0,0 +1,46 @@
+!RUN: %flang_fc1 -emit-hlfir -fopenmp %s -o - | FileCheck %s
+
+module mod
+contains
+   subroutine foo(dt, switch)
+      implicit none
+      real(4), dimension(:), intent(inout) :: dt
+      logical, intent(in) :: switch
+      integer :: dim, idx
+
+      if (switch) then
+!$omp target teams distribute parallel do
+         do idx = 1, 100
+            dt(idx) = 20
+         end do
+      else
+!$omp target teams distribute parallel do
+         do idx = 1, 100
+            dt(idx) = 30
+         end do
+      end if
+   end subroutine foo
+end module
+
+! CHECK-LABEL:   func.func @{{.*}}(
+! CHECK-SAME:      %[[ARG0:.*]]: !fir.box<!fir.array<?xf32>> {fir.bindc_name = "dt"},
+! CHECK:           %[[VAL_0:.*]] = fir.alloca !fir.box<!fir.array<?xf32>>
+! CHECK:           %[[VAL_1:.*]]:2 = hlfir.declare %[[ARG0]] dummy_scope {{.*}}
+! CHECK:           fir.if %{{.*}} {
+! CHECK:             %[[VAL_2:.*]] = fir.is_present %[[VAL_1]]#1 : (!fir.box<!fir.array<?xf32>>) -> i1
+! CHECK:             fir.if %[[VAL_2]] {
+! CHECK:               fir.store %[[VAL_1]]#1 to %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<?xf32>>>
+! CHECK:             }
+! CHECK:             %[[VAL_3:.*]] = fir.box_offset %[[VAL_0]] base_addr : (!fir.ref<!fir.box<!fir.array<?xf32>>>) -> !fir.llvm_ptr<!fir.ref<!fir.array<?xf32>>>
+! CHECK:             %[[VAL_4:.*]] = omp.map.info var_ptr(%[[VAL_0]] : !fir.ref<!fir.box<!fir.array<?xf32>>>, f32) map_clauses(implicit, tofrom) capture(ByRef) var_ptr_ptr(%[[VAL_3]] : !fir.llvm_ptr<!fir.ref<!fir.array<?xf32>>>) bounds(%{{.*}}) -> !fir.llvm_ptr<!fir.ref<!fir.array<?xf32>>> {name = ""}
+! CHECK:             %[[VAL_5:.*]] = omp.map.info var_ptr(%[[VAL_0]] : !fir.ref<!fir.box<!fir.array<?xf32>>>, !fir.box<!fir.array<?xf32>>) map_clauses(implicit, to) capture(ByRef) members(%[[VAL_4]] : [0] : !fir.llvm_ptr<!fir.ref<!fir.array<?xf32>>>) -> !fir.ref<!fir.array<?xf32>> {name = "dt"}
+! CHECK:             omp.target host_eval({{.*}}) map_entries({{.*}}%[[VAL_5]] -> {{.*}}, %[[VAL_4]] -> {{.*}} : {{.*}}) {
+! CHECK:           } else {
+! CHECK:             %[[VAL_6:.*]] = fir.is_present %[[VAL_1]]#1 : (!fir.box<!fir.array<?xf32>>) -> i1
+! CHECK:             fir.if %[[VAL_6]] {
+! CHECK:               fir.store %[[VAL_1]]#1 to %[[VAL_0]] : !fir.ref<!fir.box<!fir.array<?xf32>>>
+! CHECK:             }
+! CHECK:             %[[VAL_7:.*]] = fir.box_offset %[[VAL_0]] base_addr : (!fir.ref<!fir.box<!fir.array<?xf32>>>) -> !fir.llvm_ptr<!fir.ref<!fir.array<?xf32>>>
+! CHECK:             %[[VAL_8:.*]] = omp.map.info var_ptr(%[[VAL_0]] : !fir.ref<!fir.box<!fir.array<?xf32>>>, f32) map_clauses(implicit, tofrom) capture(ByRef) var_ptr_ptr(%[[VAL_7]] : !fir.llvm_ptr<!fir.ref<!fir.array<?xf32>>>) bounds(%{{.*}}) -> !fir.llvm_ptr<!fir.ref<!fir.array<?xf32>>> {name = ""}
+! CHECK:             %[[VAL_9:.*]] = omp.map.info var_ptr(%[[VAL_0]] : !fir.ref<!fir.box<!fir.array<?xf32>>>, !fir.box<!fir.array<?xf32>>) map_clauses(implicit, to) capture(ByRef) members(%[[VAL_8]] : [0] : !fir.llvm_ptr<!fir.ref<!fir.array<?xf32>>>) -> !fir.ref<!fir.array<?xf32>> {name = "dt"}
+! CHECK:             omp.target host_eval({{.*}}) map_entries({{.*}}, %[[VAL_9]] ->{{.*}}, %[[VAL_8]] -> {{.*}} : {{.*}}) {
diff --git a/flang/test/NonGtestUnit/lit.cfg.py b/flang/test/NonGtestUnit/lit.cfg.py
index 39ae19f..407b393 100644
--- a/flang/test/NonGtestUnit/lit.cfg.py
+++ b/flang/test/NonGtestUnit/lit.cfg.py
@@ -4,7 +4,7 @@ import lit.Test
 
 config.name = "flang-OldUnit"
 
-config.suffixes = [".test"]
+config.suffixes = [".test", ".test.exe"]
 
 config.test_source_root = os.path.join(config.flang_obj_root, "unittests")
 config.test_exec_root = config.test_source_root
diff --git a/flang/test/Semantics/c_f_pointer.f90 b/flang/test/Semantics/c_f_pointer.f90
index e2d0053..8a22175 100644
--- a/flang/test/Semantics/c_f_pointer.f90
+++ b/flang/test/Semantics/c_f_pointer.f90
@@ -54,4 +54,12 @@ program test
   call c_f_pointer(scalarC, c2ptr)
   !WARNING: FPTR= argument to C_F_POINTER() should not have the non-interoperable intrinsic type or kind CHARACTER(KIND=4,LEN=1_8) [-Winteroperability]
   call c_f_pointer(scalarC, unicodePtr)
+
+  !ERROR: SHAPE= argument to C_F_POINTER() may not appear when FPTR= is scalar
+  !ERROR: LOWER= argument to C_F_POINTER() may not appear when FPTR= is scalar
+  call c_f_pointer(scalarC, scalarIntF, [1_8], [0_8])
+  !ERROR: LOWER= argument to C_F_POINTER() must be a rank-one array.
+  call c_f_pointer(scalarC, arrayIntF, shape=[1_8], lower=rankTwoArray)
+  !ERROR: SHAPE= argument to C_F_POINTER() must appear when FPTR= is an array
+  call c_f_pointer(scalarC, arrayIntF, lower=[0])
 end program
diff --git a/flang/test/Semantics/resolve126.f90 b/flang/test/Semantics/resolve126.f90
new file mode 100644
index 0000000..6313f8c
--- /dev/null
+++ b/flang/test/Semantics/resolve126.f90
@@ -0,0 +1,7 @@
+! RUN: %python %S/test_errors.py %s %flang_fc1
+!ERROR: Attributes 'INTRINSIC' and 'EXTERNAL' conflict with each other
+real, external, intrinsic :: exp
+!ERROR: Symbol 'sin' cannot have both EXTERNAL and INTRINSIC attributes
+external sin
+intrinsic sin
+end
diff --git a/libc/config/linux/x86_64/entrypoints.txt b/libc/config/linux/x86_64/entrypoints.txt
index e8f59c9..5e8278e 100644
--- a/libc/config/linux/x86_64/entrypoints.txt
+++ b/libc/config/linux/x86_64/entrypoints.txt
@@ -1264,6 +1264,7 @@ if(LLVM_LIBC_FULL_BUILD)
     # wchar.h entrypoints
     libc.src.wchar.mblen
     libc.src.wchar.mbrlen
+    libc.src.wchar.mbsinit
     libc.src.wchar.mbrtowc
     libc.src.wchar.mbtowc
     libc.src.wchar.wcrtomb
diff --git a/libc/fuzzing/math/CMakeLists.txt b/libc/fuzzing/math/CMakeLists.txt
index be63fe4..19416fc 100644
--- a/libc/fuzzing/math/CMakeLists.txt
+++ b/libc/fuzzing/math/CMakeLists.txt
@@ -205,3 +205,30 @@ add_libc_fuzzer(
   DEPENDS
     libc.src.math.cbrt
 )
+
+add_libc_fuzzer(
+  fsqrt_fuzz
+  NEED_MPFR
+  SRCS
+    fsqrt_fuzz.cpp
+  DEPENDS
+    libc.src.math.fsqrt
+)
+
+add_libc_fuzzer(
+  f16sqrt_fuzz
+  NEED_MPFR
+  SRCS
+    f16sqrt_fuzz.cpp
+  DEPENDS
+    libc.src.math.f16sqrt
+)
+
+add_libc_fuzzer(
+  hypot_fuzz
+  NEED_MPFR
+  SRCS
+    hypot_fuzz.cpp
+  DEPENDS
+    libc.src.math.hypot
+)
diff --git a/libc/fuzzing/math/f16sqrt_fuzz.cpp b/libc/fuzzing/math/f16sqrt_fuzz.cpp
new file mode 100644
index 0000000..9a097a5
--- /dev/null
+++ b/libc/fuzzing/math/f16sqrt_fuzz.cpp
@@ -0,0 +1,56 @@
+//===-- f16sqrt_fuzz.cpp --------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+///
+/// Fuzzing test for llvm-libc f16sqrt implementation.
+///
+//===----------------------------------------------------------------------===//
+
+#include "src/math/f16sqrt.h"
+#include "utils/MPFRWrapper/mpfr_inc.h"
+#include <cstdint>
+#include <cstring>
+#include <iostream>
+#include <math.h>
+
+extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) {
+  mpfr_t input;
+  mpfr_t out;
+  mpfr_init2(input, 53);
+  mpfr_init2(out, 128);
+  for (size_t i = 0; i < size / sizeof(double); ++i) {
+    double x;
+    std::memcpy(&x, data, sizeof(double));
+    data += sizeof(double);
+
+    // remove NaN, inf, and values outside the accepted range
+    if (isnan(x) || isinf(x) || x < 0)
+      continue;
+    // signed zeros already tested in unit tests
+    if (signbit(x) && x == 0.0)
+      continue;
+
+    mpfr_set_d(input, x, MPFR_RNDN);
+    mpfr_sqrt(out, input, MPFR_RNDN);
+    float16 to_compare = mpfr_get_d(out, MPFR_RNDN);
+
+    float16 result = LIBC_NAMESPACE::f16sqrt(x);
+
+    if (result != to_compare) {
+      std::cout << std::hexfloat << "Failing input: " << x << std::endl;
+      std::cout << std::hexfloat
+                << "Failing output: " << static_cast<float>(result)
+                << std::endl;
+      std::cout << std::hexfloat
+                << "Expected: " << static_cast<float>(to_compare) << std::endl;
+      __builtin_trap();
+    }
+  }
+  mpfr_clear(input);
+  mpfr_clear(out);
+  return 0;
+}
diff --git a/libc/fuzzing/math/fsqrt_fuzz.cpp b/libc/fuzzing/math/fsqrt_fuzz.cpp
new file mode 100644
index 0000000..06bb231
--- /dev/null
+++ b/libc/fuzzing/math/fsqrt_fuzz.cpp
@@ -0,0 +1,53 @@
+//===-- fsqrt_fuzz.cpp ----------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+///
+/// Fuzzing test for llvm-libc fsqrt implementation.
+///
+//===----------------------------------------------------------------------===//
+
+#include "src/math/fsqrt.h"
+#include "utils/MPFRWrapper/mpfr_inc.h"
+#include <cstdint>
+#include <cstring>
+#include <iostream>
+#include <math.h>
+
+extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) {
+  mpfr_t input;
+  mpfr_t out;
+  mpfr_init2(input, 53);
+  mpfr_init2(out, 128);
+  for (size_t i = 0; i < size / sizeof(double); ++i) {
+    double x;
+    std::memcpy(&x, data, sizeof(double));
+    data += sizeof(double);
+
+    // remove NaN, inf, and values outside the accepted range
+    if (isnan(x) || isinf(x) || x < 0)
+      continue;
+    // signed zeros already tested in unit tests
+    if (signbit(x) && x == 0.0)
+      continue;
+
+    mpfr_set_d(input, x, MPFR_RNDN);
+    mpfr_sqrt(out, input, MPFR_RNDN);
+    float to_compare = mpfr_get_flt(out, MPFR_RNDN);
+
+    float result = LIBC_NAMESPACE::fsqrt(x);
+
+    if (result != to_compare) {
+      std::cout << std::hexfloat << "Failing input: " << x << std::endl;
+      std::cout << std::hexfloat << "Failing output: " << result << std::endl;
+      std::cout << std::hexfloat << "Expected: " << to_compare << std::endl;
+      __builtin_trap();
+    }
+  }
+  mpfr_clear(input);
+  mpfr_clear(out);
+  return 0;
+}
diff --git a/libc/fuzzing/math/hypot_fuzz.cpp b/libc/fuzzing/math/hypot_fuzz.cpp
new file mode 100644
index 0000000..6129e41
--- /dev/null
+++ b/libc/fuzzing/math/hypot_fuzz.cpp
@@ -0,0 +1,64 @@
+//===-- hypot_fuzz.cpp ----------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+///
+/// Fuzzing test for llvm-libc hypot implementation.
+///
+//===----------------------------------------------------------------------===//
+
+#include "src/math/hypot.h"
+#include "utils/MPFRWrapper/mpfr_inc.h"
+#include <cstdint>
+#include <cstring>
+#include <iostream>
+#include <math.h>
+
+extern "C" int LLVMFuzzerTestOneInput(const uint8_t *data, size_t size) {
+  mpfr_t in_x;
+  mpfr_t in_y;
+  mpfr_t out;
+  mpfr_init2(in_x, 53);
+  mpfr_init2(in_y, 53);
+  mpfr_init2(out, 128);
+
+  for (size_t i = 0; i < size / (2 * sizeof(double)); ++i) {
+    double x;
+    double y;
+
+    std::memcpy(&x, data, sizeof(double));
+    data += sizeof(double);
+    std::memcpy(&y, data, sizeof(double));
+    data += sizeof(double);
+
+    // remove NaN, inf, and signed zeros
+    if (isnan(x) || isinf(x) || (signbit(x) && x == 0.0))
+      return 0;
+    if (isnan(y) || isinf(y) || (signbit(y) && y == 0.0))
+      return 0;
+
+    mpfr_set_d(in_x, x, MPFR_RNDN);
+    mpfr_set_d(in_y, y, MPFR_RNDN);
+
+    int output = mpfr_hypot(out, in_x, in_y, MPFR_RNDN);
+    mpfr_subnormalize(out, output, MPFR_RNDN);
+    double to_compare = mpfr_get_d(out, MPFR_RNDN);
+
+    double result = LIBC_NAMESPACE::hypot(x, y);
+
+    if (result != to_compare) {
+      std::cout << std::hexfloat << "Failing x: " << x << std::endl;
+      std::cout << std::hexfloat << "Failing y: " << y << std::endl;
+      std::cout << std::hexfloat << "Failing output: " << result << std::endl;
+      std::cout << std::hexfloat << "Expected: " << to_compare << std::endl;
+      __builtin_trap();
+    }
+  }
+  mpfr_clear(in_x);
+  mpfr_clear(in_y);
+  mpfr_clear(out);
+  return 0;
+}
diff --git a/libc/include/wchar.yaml b/libc/include/wchar.yaml
index 0285f19..6e1f595 100644
--- a/libc/include/wchar.yaml
+++ b/libc/include/wchar.yaml
@@ -53,6 +53,12 @@ functions:
       - type: wchar_t *__restrict
       - type: const char *__restrict
       - type: size_t
+  - name: mbsinit
+    standards:
+      - stdc
+    return_type: int
+    arguments:
+      - type: mbstate_t *
   - name: mblen
     standards:
       - stdc
diff --git a/libc/shared/math.h b/libc/shared/math.h
index e3c674c..e01c8db 100644
--- a/libc/shared/math.h
+++ b/libc/shared/math.h
@@ -16,6 +16,10 @@
 #include "math/acosf16.h"
 #include "math/acoshf.h"
 #include "math/acoshf16.h"
+#include "math/acospif16.h"
+#include "math/asin.h"
+#include "math/asinf.h"
+#include "math/asinf16.h"
 #include "math/erff.h"
 #include "math/exp.h"
 #include "math/exp10.h"
diff --git a/libc/shared/math/acospif16.h b/libc/shared/math/acospif16.h
new file mode 100644
index 0000000..38225f2
--- /dev/null
+++ b/libc/shared/math/acospif16.h
@@ -0,0 +1,29 @@
+//===-- Shared acospif16 function -------------------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SHARED_MATH_ACOSPIF16_H
+#define LLVM_LIBC_SHARED_MATH_ACOSPIF16_H
+
+#include "include/llvm-libc-macros/float16-macros.h"
+
+#ifdef LIBC_TYPES_HAS_FLOAT16
+
+#include "shared/libc_common.h"
+#include "src/__support/math/acospif16.h"
+
+namespace LIBC_NAMESPACE_DECL {
+namespace shared {
+
+using math::acospif16;
+
+} // namespace shared
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LIBC_TYPES_HAS_FLOAT16
+
+#endif // LLVM_LIBC_SHARED_MATH_ACOSPIF16_H
diff --git a/libc/shared/math/asin.h b/libc/shared/math/asin.h
new file mode 100644
index 0000000..0b2c8ea6
--- /dev/null
+++ b/libc/shared/math/asin.h
@@ -0,0 +1,23 @@
+//===-- Shared asin function ------------------------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SHARED_MATH_ASIN_H
+#define LLVM_LIBC_SHARED_MATH_ASIN_H
+
+#include "shared/libc_common.h"
+#include "src/__support/math/asin.h"
+
+namespace LIBC_NAMESPACE_DECL {
+namespace shared {
+
+using math::asin;
+
+} // namespace shared
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LLVM_LIBC_SHARED_MATH_ASIN_H
diff --git a/libc/shared/math/asinf.h b/libc/shared/math/asinf.h
new file mode 100644
index 0000000..ac051bd
--- /dev/null
+++ b/libc/shared/math/asinf.h
@@ -0,0 +1,23 @@
+//===-- Shared asinf function -----------------------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SHARED_MATH_ASINF_H
+#define LLVM_LIBC_SHARED_MATH_ASINF_H
+
+#include "shared/libc_common.h"
+#include "src/__support/math/asinf.h"
+
+namespace LIBC_NAMESPACE_DECL {
+namespace shared {
+
+using math::asinf;
+
+} // namespace shared
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LLVM_LIBC_SHARED_MATH_ASINF_H
diff --git a/libc/shared/math/asinf16.h b/libc/shared/math/asinf16.h
new file mode 100644
index 0000000..d545e26
--- /dev/null
+++ b/libc/shared/math/asinf16.h
@@ -0,0 +1,28 @@
+//===-- Shared asinf16 function ---------------------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SHARED_MATH_ASINF16_H
+#define LLVM_LIBC_SHARED_MATH_ASINF16_H
+
+#include "shared/libc_common.h"
+
+#ifdef LIBC_TYPES_HAS_FLOAT16
+
+#include "src/__support/math/asinf16.h"
+
+namespace LIBC_NAMESPACE_DECL {
+namespace shared {
+
+using math::asinf16;
+
+} // namespace shared
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LIBC_TYPES_HAS_FLOAT16
+
+#endif // LLVM_LIBC_SHARED_MATH_ASINF_H
diff --git a/libc/shared/math/exp10f16.h b/libc/shared/math/exp10f16.h
index af00787..d6ba067 100644
--- a/libc/shared/math/exp10f16.h
+++ b/libc/shared/math/exp10f16.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef LLVM_LIBC_SHARED_MATH_EXP10F_H
-#define LLVM_LIBC_SHARED_MATH_EXP10F_H
+#ifndef LLVM_LIBC_SHARED_MATH_EXP10F16_H
+#define LLVM_LIBC_SHARED_MATH_EXP10F16_H
 
 #include "include/llvm-libc-macros/float16-macros.h"
 #include "shared/libc_common.h"
@@ -26,4 +26,4 @@ using math::exp10f16;
 
 #endif // LIBC_TYPES_HAS_FLOAT16
 
-#endif // LLVM_LIBC_SHARED_MATH_EXP10F_H
+#endif // LLVM_LIBC_SHARED_MATH_EXP10F16_H
diff --git a/libc/src/__support/FPUtil/CMakeLists.txt b/libc/src/__support/FPUtil/CMakeLists.txt
index 94f8b95..6e447fc 100644
--- a/libc/src/__support/FPUtil/CMakeLists.txt
+++ b/libc/src/__support/FPUtil/CMakeLists.txt
@@ -7,6 +7,7 @@ add_header_library(
     libc.hdr.fenv_macros
     libc.hdr.math_macros
     libc.hdr.stdint_proxy
+    libc.src.__support.CPP.type_traits
     libc.src.__support.macros.attributes
     libc.src.errno.errno
 )
diff --git a/libc/src/__support/FPUtil/FEnvImpl.h b/libc/src/__support/FPUtil/FEnvImpl.h
index 7691088..7bd5643 100644
--- a/libc/src/__support/FPUtil/FEnvImpl.h
+++ b/libc/src/__support/FPUtil/FEnvImpl.h
@@ -12,6 +12,7 @@
 #include "hdr/fenv_macros.h"
 #include "hdr/math_macros.h"
 #include "hdr/types/fenv_t.h"
+#include "src/__support/CPP/type_traits.h"
 #include "src/__support/libc_errno.h"
 #include "src/__support/macros/attributes.h" // LIBC_INLINE
 #include "src/__support/macros/config.h"
@@ -72,40 +73,58 @@ LIBC_INLINE int set_env(const fenv_t *) { return 0; }
 namespace LIBC_NAMESPACE_DECL {
 namespace fputil {
 
-LIBC_INLINE static int clear_except_if_required([[maybe_unused]] int excepts) {
+LIBC_INLINE static constexpr int
+clear_except_if_required([[maybe_unused]] int excepts) {
+  if (cpp::is_constant_evaluated()) {
+    return 0;
+  } else {
 #ifndef LIBC_MATH_HAS_NO_EXCEPT
-  if (math_errhandling & MATH_ERREXCEPT)
-    return clear_except(excepts);
+    if (math_errhandling & MATH_ERREXCEPT)
+      return clear_except(excepts);
 #endif // LIBC_MATH_HAS_NO_EXCEPT
-  return 0;
+    return 0;
+  }
 }
 
-LIBC_INLINE static int set_except_if_required([[maybe_unused]] int excepts) {
+LIBC_INLINE static constexpr int
+set_except_if_required([[maybe_unused]] int excepts) {
+  if (cpp::is_constant_evaluated()) {
+    return 0;
+  } else {
 #ifndef LIBC_MATH_HAS_NO_EXCEPT
-  if (math_errhandling & MATH_ERREXCEPT)
-    return set_except(excepts);
+    if (math_errhandling & MATH_ERREXCEPT)
+      return set_except(excepts);
 #endif // LIBC_MATH_HAS_NO_EXCEPT
-  return 0;
+    return 0;
+  }
 }
 
-LIBC_INLINE static int raise_except_if_required([[maybe_unused]] int excepts) {
+LIBC_INLINE static constexpr int
+raise_except_if_required([[maybe_unused]] int excepts) {
+  if (cpp::is_constant_evaluated()) {
+    return 0;
+  } else {
 #ifndef LIBC_MATH_HAS_NO_EXCEPT
-  if (math_errhandling & MATH_ERREXCEPT)
+    if (math_errhandling & MATH_ERREXCEPT)
 #ifdef LIBC_TARGET_ARCH_IS_X86_64
-    return raise_except</*SKIP_X87_FPU*/ true>(excepts);
+      return raise_except</*SKIP_X87_FPU*/ true>(excepts);
 #else  // !LIBC_TARGET_ARCH_IS_X86
-    return raise_except(excepts);
+      return raise_except(excepts);
 #endif // LIBC_TARGET_ARCH_IS_X86
 
 #endif // LIBC_MATH_HAS_NO_EXCEPT
-  return 0;
+    return 0;
+  }
 }
 
-LIBC_INLINE static void set_errno_if_required([[maybe_unused]] int err) {
+LIBC_INLINE static constexpr void
+set_errno_if_required([[maybe_unused]] int err) {
+  if (!cpp::is_constant_evaluated()) {
 #ifndef LIBC_MATH_HAS_NO_ERRNO
-  if (math_errhandling & MATH_ERRNO)
-    libc_errno = err;
+    if (math_errhandling & MATH_ERRNO)
+      libc_errno = err;
 #endif // LIBC_MATH_HAS_NO_ERRNO
+  }
 }
 
 } // namespace fputil
diff --git a/libc/src/__support/math/CMakeLists.txt b/libc/src/__support/math/CMakeLists.txt
index 926bbd5..1050938 100644
--- a/libc/src/__support/math/CMakeLists.txt
+++ b/libc/src/__support/math/CMakeLists.txt
@@ -4,7 +4,6 @@ add_header_library(
     acos.h
   DEPENDS
     .asin_utils
-    libc.src.__support.math.asin_utils
     libc.src.__support.FPUtil.double_double
     libc.src.__support.FPUtil.dyadic_float
     libc.src.__support.FPUtil.fenv_impl
@@ -96,6 +95,21 @@ add_header_library(
 )
 
 add_header_library(
+  acospif16
+  HDRS
+    acospif16.h
+  DEPENDS
+    libc.src.__support.FPUtil.cast
+    libc.src.__support.FPUtil.fenv_impl
+    libc.src.__support.FPUtil.fp_bits
+    libc.src.__support.FPUtil.multiply_add
+    libc.src.__support.FPUtil.polyeval
+    libc.src.__support.FPUtil.sqrt
+    libc.src.__support.macros.optimization
+    libc.src.__support.macros.properties.types
+)
+
+add_header_library(
   asin_utils
   HDRS
     asin_utils.h
@@ -110,6 +124,53 @@ add_header_library(
 )
 
 add_header_library(
+  asin
+  HDRS
+    asin.h
+  DEPENDS
+    .asin_utils
+    libc.src.__support.FPUtil.double_double
+    libc.src.__support.FPUtil.dyadic_float
+    libc.src.__support.FPUtil.fenv_impl
+    libc.src.__support.FPUtil.fp_bits
+    libc.src.__support.FPUtil.multiply_add
+    libc.src.__support.FPUtil.polyeval
+    libc.src.__support.FPUtil.sqrt
+    libc.src.__support.macros.optimization
+    libc.src.__support.macros.properties.cpu_features
+)
+
+add_header_library(
+  asinf
+  HDRS
+    asinf.h
+  DEPENDS
+    .inv_trigf_utils
+    libc.src.__support.FPUtil.fenv_impl
+    libc.src.__support.FPUtil.fp_bits
+    libc.src.__support.FPUtil.except_value_utils
+    libc.src.__support.FPUtil.multiply_add
+    libc.src.__support.FPUtil.sqrt
+    libc.src.__support.macros.config
+    libc.src.__support.macros.optimization
+    libc.src.__support.macros.properties.cpu_features
+)
+
+add_header_library(
+  asinf16
+  HDRS
+    asinf16.h
+  DEPENDS
+    libc.src.__support.FPUtil.fenv_impl
+    libc.src.__support.FPUtil.fp_bits
+    libc.src.__support.FPUtil.polyeval
+    libc.src.__support.FPUtil.cast
+    libc.src.__support.FPUtil.multiply_add
+    libc.src.__support.FPUtil.sqrt
+    libc.src.__support.macros.optimization
+)
+
+add_header_library(
   erff
   HDRS
     erff.h
@@ -358,11 +419,10 @@ add_header_library(
   DEPENDS
     .exp10f16_utils
     libc.src.__support.FPUtil.fp_bits
-    src.__support.FPUtil.FEnvImpl
-    src.__support.FPUtil.FPBits
-    src.__support.FPUtil.cast
-    src.__support.FPUtil.rounding_mode
-    src.__support.FPUtil.except_value_utils
-    src.__support.macros.optimization
-    src.__support.macros.properties.cpu_features
+    libc.src.__support.FPUtil.fenv_impl
+    libc.src.__support.FPUtil.cast
+    libc.src.__support.FPUtil.rounding_mode
+    libc.src.__support.FPUtil.except_value_utils
+    libc.src.__support.macros.optimization
+    libc.src.__support.macros.properties.cpu_features
 )
diff --git a/libc/src/__support/math/acos.h b/libc/src/__support/math/acos.h
index a52ead7..0e1e413 100644
--- a/libc/src/__support/math/acos.h
+++ b/libc/src/__support/math/acos.h
@@ -24,7 +24,7 @@ namespace LIBC_NAMESPACE_DECL {
 
 namespace math {
 
-static constexpr double acos(double x) {
+LIBC_INLINE static constexpr double acos(double x) {
   using DoubleDouble = fputil::DoubleDouble;
   using namespace asin_internal;
   using FPBits = fputil::FPBits<double>;
diff --git a/libc/src/__support/math/acosf.h b/libc/src/__support/math/acosf.h
index 153087e..7a0c0e5 100644
--- a/libc/src/__support/math/acosf.h
+++ b/libc/src/__support/math/acosf.h
@@ -45,7 +45,7 @@ static constexpr fputil::ExceptValues<float, N_EXCEPTS> ACOSF_EXCEPTS = {{
 
 } // namespace acosf_internal
 
-static constexpr float acosf(float x) {
+LIBC_INLINE static constexpr float acosf(float x) {
   using namespace acosf_internal;
   using namespace inv_trigf_utils_internal;
   using FPBits = typename fputil::FPBits<float>;
diff --git a/libc/src/__support/math/acosf16.h b/libc/src/__support/math/acosf16.h
index 58d3761..3f0e002 100644
--- a/libc/src/__support/math/acosf16.h
+++ b/libc/src/__support/math/acosf16.h
@@ -26,7 +26,7 @@ namespace LIBC_NAMESPACE_DECL {
 
 namespace math {
 
-static constexpr float16 acosf16(float16 x) {
+LIBC_INLINE static constexpr float16 acosf16(float16 x) {
 
   // Generated by Sollya using the following command:
   // > round(pi/2, SG, RN);
diff --git a/libc/src/__support/math/acoshf.h b/libc/src/__support/math/acoshf.h
index f18f169..4e00311 100644
--- a/libc/src/__support/math/acoshf.h
+++ b/libc/src/__support/math/acoshf.h
@@ -21,7 +21,7 @@ namespace LIBC_NAMESPACE_DECL {
 
 namespace math {
 
-static constexpr float acoshf(float x) {
+LIBC_INLINE static constexpr float acoshf(float x) {
   using namespace acoshf_internal;
   using FPBits_t = typename fputil::FPBits<float>;
   FPBits_t xbits(x);
diff --git a/libc/src/__support/math/acoshf16.h b/libc/src/__support/math/acoshf16.h
index 15e7f6a..e5be2a8 100644
--- a/libc/src/__support/math/acoshf16.h
+++ b/libc/src/__support/math/acoshf16.h
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef LLVM_LIBC_SRC___SUPPORT_MATH_ACOSHF_H
-#define LLVM_LIBC_SRC___SUPPORT_MATH_ACOSHF_H
+#ifndef LLVM_LIBC_SRC___SUPPORT_MATH_ACOSHF16_H
+#define LLVM_LIBC_SRC___SUPPORT_MATH_ACOSHF16_H
 
 #include "include/llvm-libc-macros/float16-macros.h"
 
@@ -28,7 +28,7 @@ namespace LIBC_NAMESPACE_DECL {
 
 namespace math {
 
-static constexpr float16 acoshf16(float16 x) {
+LIBC_INLINE static constexpr float16 acoshf16(float16 x) {
 
   using namespace acoshf_internal;
   constexpr size_t N_EXCEPTS = 2;
@@ -120,4 +120,4 @@ static constexpr float16 acoshf16(float16 x) {
 
 #endif // LIBC_TYPES_HAS_FLOAT16
 
-#endif // LLVM_LIBC_SRC___SUPPORT_MATH_ACOSHF_H
+#endif // LLVM_LIBC_SRC___SUPPORT_MATH_ACOSHF16_H
diff --git a/libc/src/__support/math/acospif16.h b/libc/src/__support/math/acospif16.h
new file mode 100644
index 0000000..cf29c76
--- /dev/null
+++ b/libc/src/__support/math/acospif16.h
@@ -0,0 +1,147 @@
+//===-- Implementation header for acospif16 ---------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC___SUPPORT_MATH_ACOSPIF16_H
+#define LLVM_LIBC_SRC___SUPPORT_MATH_ACOSPIF16_H
+
+#include "include/llvm-libc-macros/float16-macros.h"
+
+#ifdef LIBC_TYPES_HAS_FLOAT16
+
+#include "src/__support/FPUtil/FEnvImpl.h"
+#include "src/__support/FPUtil/FPBits.h"
+#include "src/__support/FPUtil/PolyEval.h"
+#include "src/__support/FPUtil/cast.h"
+#include "src/__support/FPUtil/multiply_add.h"
+#include "src/__support/FPUtil/sqrt.h"
+#include "src/__support/macros/optimization.h"
+
+namespace LIBC_NAMESPACE_DECL {
+
+namespace math {
+
+LIBC_INLINE static constexpr float16 acospif16(float16 x) {
+  using FPBits = fputil::FPBits<float16>;
+  FPBits xbits(x);
+
+  uint16_t x_u = xbits.uintval();
+  uint16_t x_abs = x_u & 0x7fff;
+  uint16_t x_sign = x_u >> 15;
+
+  // |x| > 0x1p0, |x| > 1, or x is NaN.
+  if (LIBC_UNLIKELY(x_abs > 0x3c00)) {
+    // acospif16(NaN) = NaN
+    if (xbits.is_nan()) {
+      if (xbits.is_signaling_nan()) {
+        fputil::raise_except_if_required(FE_INVALID);
+        return FPBits::quiet_nan().get_val();
+      }
+
+      return x;
+    }
+
+    // 1 < |x| <= +inf
+    fputil::raise_except_if_required(FE_INVALID);
+    fputil::set_errno_if_required(EDOM);
+
+    return FPBits::quiet_nan().get_val();
+  }
+
+  // |x| == 0x1p0, x is 1 or -1
+  // if x is (-)1, return 1
+  // if x is (+)1, return 0
+  if (LIBC_UNLIKELY(x_abs == 0x3c00))
+    return fputil::cast<float16>(x_sign ? 1.0f : 0.0f);
+
+  float xf = x;
+  float xsq = xf * xf;
+
+  // Degree-6 minimax polynomial coefficients of asin(x) generated by Sollya
+  // with: > P = fpminimax(asin(x)/(pi * x), [|0, 2, 4, 6, 8|], [|SG...|], [0,
+  // 0.5]);
+  constexpr float POLY_COEFFS[5] = {0x1.45f308p-2f, 0x1.b2900cp-5f,
+                                    0x1.897e36p-6f, 0x1.9efafcp-7f,
+                                    0x1.06d884p-6f};
+  // |x| <= 0x1p-1, |x| <= 0.5
+  if (x_abs <= 0x3800) {
+    // if x is 0, return 0.5
+    if (LIBC_UNLIKELY(x_abs == 0))
+      return fputil::cast<float16>(0.5f);
+
+    // Note that: acos(x) = pi/2 + asin(-x) = pi/2 - asin(x), then
+    //            acospi(x) = 0.5 - asin(x)/pi
+    float interm =
+        fputil::polyeval(xsq, POLY_COEFFS[0], POLY_COEFFS[1], POLY_COEFFS[2],
+                         POLY_COEFFS[3], POLY_COEFFS[4]);
+
+    return fputil::cast<float16>(fputil::multiply_add(-xf, interm, 0.5f));
+  }
+
+  // When |x| > 0.5, assume that 0.5 < |x| <= 1
+  //
+  // Step-by-step range-reduction proof:
+  // 1:  Let y = asin(x), such that, x = sin(y)
+  // 2:  From complimentary angle identity:
+  //       x = sin(y) = cos(pi/2 - y)
+  // 3:  Let z = pi/2 - y, such that x = cos(z)
+  // 4:  From double angle formula; cos(2A) = 1 - 2 * sin^2(A):
+  //       z = 2A, z/2 = A
+  //       cos(z) = 1 - 2 * sin^2(z/2)
+  // 5:  Make sin(z/2) subject of the formula:
+  //       sin(z/2) = sqrt((1 - cos(z))/2)
+  // 6:  Recall [3]; x = cos(z). Therefore:
+  //       sin(z/2) = sqrt((1 - x)/2)
+  // 7:  Let u = (1 - x)/2
+  // 8:  Therefore:
+  //       asin(sqrt(u)) = z/2
+  //       2 * asin(sqrt(u)) = z
+  // 9:  Recall [3]; z = pi/2 - y. Therefore:
+  //       y = pi/2 - z
+  //       y = pi/2 - 2 * asin(sqrt(u))
+  // 10: Recall [1], y = asin(x). Therefore:
+  //       asin(x) = pi/2 - 2 * asin(sqrt(u))
+  // 11: Recall that: acos(x) = pi/2 + asin(-x) = pi/2 - asin(x)
+  //     Therefore:
+  //       acos(x) = pi/2 - (pi/2 - 2 * asin(sqrt(u)))
+  //       acos(x) = 2 * asin(sqrt(u))
+  //       acospi(x) = 2 * (asin(sqrt(u)) / pi)
+  //
+  // THE RANGE REDUCTION, HOW?
+  // 12: Recall [7], u = (1 - x)/2
+  // 13: Since 0.5 < x <= 1, therefore:
+  //       0 <= u <= 0.25 and 0 <= sqrt(u) <= 0.5
+  //
+  // Hence, we can reuse the same [0, 0.5] domain polynomial approximation for
+  // Step [11] as `sqrt(u)` is in range.
+  // When -1 < x <= -0.5, the identity:
+  //       acos(x) = pi - acos(-x)
+  //       acospi(x) = 1 - acos(-x)/pi
+  // allows us to compute for the negative x value (lhs)
+  // with a positive x value instead (rhs).
+
+  float xf_abs = (xf < 0 ? -xf : xf);
+  float u = fputil::multiply_add(-0.5f, xf_abs, 0.5f);
+  float sqrt_u = fputil::sqrt<float>(u);
+
+  float asin_sqrt_u =
+      sqrt_u * fputil::polyeval(u, POLY_COEFFS[0], POLY_COEFFS[1],
+                                POLY_COEFFS[2], POLY_COEFFS[3], POLY_COEFFS[4]);
+
+  // Same as acos(x), but devided the expression with pi
+  return fputil::cast<float16>(
+      x_sign ? fputil::multiply_add(-2.0f, asin_sqrt_u, 1.0f)
+             : 2.0f * asin_sqrt_u);
+}
+
+} // namespace math
+
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LIBC_TYPES_HAS_FLOAT16
+
+#endif // LLVM_LIBC_SRC___SUPPORT_MATH_ACOSPIF16_H
diff --git a/libc/src/__support/math/asin.h b/libc/src/__support/math/asin.h
new file mode 100644
index 0000000..5e06d04
--- /dev/null
+++ b/libc/src/__support/math/asin.h
@@ -0,0 +1,297 @@
+//===-- Implementation header for asin --------------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC___SUPPORT_MATH_ASIN_H
+#define LLVM_LIBC_SRC___SUPPORT_MATH_ASIN_H
+
+#include "asin_utils.h"
+#include "src/__support/FPUtil/FEnvImpl.h"
+#include "src/__support/FPUtil/FPBits.h"
+#include "src/__support/FPUtil/double_double.h"
+#include "src/__support/FPUtil/dyadic_float.h"
+#include "src/__support/FPUtil/multiply_add.h"
+#include "src/__support/FPUtil/sqrt.h"
+#include "src/__support/macros/config.h"
+#include "src/__support/macros/optimization.h"            // LIBC_UNLIKELY
+#include "src/__support/macros/properties/cpu_features.h" // LIBC_TARGET_CPU_HAS_FMA
+#include "src/__support/math/asin_utils.h"
+
+namespace LIBC_NAMESPACE_DECL {
+
+namespace math {
+
+LIBC_INLINE static constexpr double asin(double x) {
+  using namespace asin_internal;
+  using FPBits = fputil::FPBits<double>;
+
+  FPBits xbits(x);
+  int x_exp = xbits.get_biased_exponent();
+
+  // |x| < 0.5.
+  if (x_exp < FPBits::EXP_BIAS - 1) {
+    // |x| < 2^-26.
+    if (LIBC_UNLIKELY(x_exp < FPBits::EXP_BIAS - 26)) {
+      // When |x| < 2^-26, the relative error of the approximation asin(x) ~ x
+      // is:
+      //   |asin(x) - x| / |asin(x)| < |x^3| / (6|x|)
+      //                             = x^2 / 6
+      //                             < 2^-54
+      //                             < epsilon(1)/2.
+      // So the correctly rounded values of asin(x) are:
+      //   = x + sign(x)*eps(x) if rounding mode = FE_TOWARDZERO,
+      //                        or (rounding mode = FE_UPWARD and x is
+      //                        negative),
+      //   = x otherwise.
+      // To simplify the rounding decision and make it more efficient, we use
+      //   fma(x, 2^-54, x) instead.
+      // Note: to use the formula x + 2^-54*x to decide the correct rounding, we
+      // do need fma(x, 2^-54, x) to prevent underflow caused by 2^-54*x when
+      // |x| < 2^-1022. For targets without FMA instructions, when x is close to
+      // denormal range, we normalize x,
+#if defined(LIBC_MATH_HAS_SKIP_ACCURATE_PASS)
+      return x;
+#elif defined(LIBC_TARGET_CPU_HAS_FMA_DOUBLE)
+      return fputil::multiply_add(x, 0x1.0p-54, x);
+#else
+      if (xbits.abs().uintval() == 0)
+        return x;
+      // Get sign(x) * min_normal.
+      FPBits eps_bits = FPBits::min_normal();
+      eps_bits.set_sign(xbits.sign());
+      double eps = eps_bits.get_val();
+      double normalize_const = (x_exp == 0) ? eps : 0.0;
+      double scaled_normal =
+          fputil::multiply_add(x + normalize_const, 0x1.0p54, eps);
+      return fputil::multiply_add(scaled_normal, 0x1.0p-54, -normalize_const);
+#endif // LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+    }
+
+#ifdef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+    return x * asin_eval(x * x);
+#else
+    using Float128 = fputil::DyadicFloat<128>;
+    using DoubleDouble = fputil::DoubleDouble;
+
+    unsigned idx = 0;
+    DoubleDouble x_sq = fputil::exact_mult(x, x);
+    double err = xbits.abs().get_val() * 0x1.0p-51;
+    // Polynomial approximation:
+    //   p ~ asin(x)/x
+
+    DoubleDouble p = asin_eval(x_sq, idx, err);
+    // asin(x) ~ x * (ASIN_COEFFS[idx][0] + p)
+    DoubleDouble r0 = fputil::exact_mult(x, p.hi);
+    double r_lo = fputil::multiply_add(x, p.lo, r0.lo);
+
+    // Ziv's accuracy test.
+
+    double r_upper = r0.hi + (r_lo + err);
+    double r_lower = r0.hi + (r_lo - err);
+
+    if (LIBC_LIKELY(r_upper == r_lower))
+      return r_upper;
+
+    // Ziv's accuracy test failed, perform 128-bit calculation.
+
+    // Recalculate mod 1/64.
+    idx = static_cast<unsigned>(fputil::nearest_integer(x_sq.hi * 0x1.0p6));
+
+    // Get x^2 - idx/64 exactly.  When FMA is available, double-double
+    // multiplication will be correct for all rounding modes.  Otherwise we use
+    // Float128 directly.
+    Float128 x_f128(x);
+
+#ifdef LIBC_TARGET_CPU_HAS_FMA_DOUBLE
+    // u = x^2 - idx/64
+    Float128 u_hi(
+        fputil::multiply_add(static_cast<double>(idx), -0x1.0p-6, x_sq.hi));
+    Float128 u = fputil::quick_add(u_hi, Float128(x_sq.lo));
+#else
+    Float128 x_sq_f128 = fputil::quick_mul(x_f128, x_f128);
+    Float128 u = fputil::quick_add(
+        x_sq_f128, Float128(static_cast<double>(idx) * (-0x1.0p-6)));
+#endif // LIBC_TARGET_CPU_HAS_FMA_DOUBLE
+
+    Float128 p_f128 = asin_eval(u, idx);
+    Float128 r = fputil::quick_mul(x_f128, p_f128);
+
+    return static_cast<double>(r);
+#endif // LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+  }
+  // |x| >= 0.5
+
+  double x_abs = xbits.abs().get_val();
+
+  // Maintaining the sign:
+  constexpr double SIGN[2] = {1.0, -1.0};
+  double x_sign = SIGN[xbits.is_neg()];
+
+  // |x| >= 1
+  if (LIBC_UNLIKELY(x_exp >= FPBits::EXP_BIAS)) {
+    // x = +-1, asin(x) = +- pi/2
+    if (x_abs == 1.0) {
+      // return +- pi/2
+      return fputil::multiply_add(x_sign, PI_OVER_TWO.hi,
+                                  x_sign * PI_OVER_TWO.lo);
+    }
+    // |x| > 1, return NaN.
+    if (xbits.is_quiet_nan())
+      return x;
+
+    // Set domain error for non-NaN input.
+    if (!xbits.is_nan())
+      fputil::set_errno_if_required(EDOM);
+
+    fputil::raise_except_if_required(FE_INVALID);
+    return FPBits::quiet_nan().get_val();
+  }
+
+  // When |x| >= 0.5, we perform range reduction as follow:
+  //
+  // Assume further that 0.5 <= x < 1, and let:
+  //   y = asin(x)
+  // We will use the double angle formula:
+  //   cos(2y) = 1 - 2 sin^2(y)
+  // and the complement angle identity:
+  //   x = sin(y) = cos(pi/2 - y)
+  //              = 1 - 2 sin^2 (pi/4 - y/2)
+  // So:
+  //   sin(pi/4 - y/2) = sqrt( (1 - x)/2 )
+  // And hence:
+  //   pi/4 - y/2 = asin( sqrt( (1 - x)/2 ) )
+  // Equivalently:
+  //   asin(x) = y = pi/2 - 2 * asin( sqrt( (1 - x)/2 ) )
+  // Let u = (1 - x)/2, then:
+  //   asin(x) = pi/2 - 2 * asin( sqrt(u) )
+  // Moreover, since 0.5 <= x < 1:
+  //   0 < u <= 1/4, and 0 < sqrt(u) <= 0.5,
+  // And hence we can reuse the same polynomial approximation of asin(x) when
+  // |x| <= 0.5:
+  //   asin(x) ~ pi/2 - 2 * sqrt(u) * P(u),
+
+  // u = (1 - |x|)/2
+  double u = fputil::multiply_add(x_abs, -0.5, 0.5);
+  // v_hi + v_lo ~ sqrt(u).
+  // Let:
+  //   h = u - v_hi^2 = (sqrt(u) - v_hi) * (sqrt(u) + v_hi)
+  // Then:
+  //   sqrt(u) = v_hi + h / (sqrt(u) + v_hi)
+  //           ~ v_hi + h / (2 * v_hi)
+  // So we can use:
+  //   v_lo = h / (2 * v_hi).
+  // Then,
+  //   asin(x) ~ pi/2 - 2*(v_hi + v_lo) * P(u)
+  double v_hi = fputil::sqrt<double>(u);
+
+#ifdef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+  double p = asin_eval(u);
+  double r = x_sign * fputil::multiply_add(-2.0 * v_hi, p, PI_OVER_TWO.hi);
+  return r;
+#else
+
+#ifdef LIBC_TARGET_CPU_HAS_FMA_DOUBLE
+  double h = fputil::multiply_add(v_hi, -v_hi, u);
+#else
+  DoubleDouble v_hi_sq = fputil::exact_mult(v_hi, v_hi);
+  double h = (u - v_hi_sq.hi) - v_hi_sq.lo;
+#endif // LIBC_TARGET_CPU_HAS_FMA_DOUBLE
+
+  // Scale v_lo and v_hi by 2 from the formula:
+  //   vh = v_hi * 2
+  //   vl = 2*v_lo = h / v_hi.
+  double vh = v_hi * 2.0;
+  double vl = h / v_hi;
+
+  // Polynomial approximation:
+  //   p ~ asin(sqrt(u))/sqrt(u)
+  unsigned idx = 0;
+  double err = vh * 0x1.0p-51;
+
+  DoubleDouble p = asin_eval(DoubleDouble{0.0, u}, idx, err);
+
+  // Perform computations in double-double arithmetic:
+  //   asin(x) = pi/2 - (v_hi + v_lo) * (ASIN_COEFFS[idx][0] + p)
+  DoubleDouble r0 = fputil::quick_mult(DoubleDouble{vl, vh}, p);
+  DoubleDouble r = fputil::exact_add(PI_OVER_TWO.hi, -r0.hi);
+
+  double r_lo = PI_OVER_TWO.lo - r0.lo + r.lo;
+
+  // Ziv's accuracy test.
+
+#ifdef LIBC_TARGET_CPU_HAS_FMA_DOUBLE
+  double r_upper = fputil::multiply_add(
+      r.hi, x_sign, fputil::multiply_add(r_lo, x_sign, err));
+  double r_lower = fputil::multiply_add(
+      r.hi, x_sign, fputil::multiply_add(r_lo, x_sign, -err));
+#else
+  r_lo *= x_sign;
+  r.hi *= x_sign;
+  double r_upper = r.hi + (r_lo + err);
+  double r_lower = r.hi + (r_lo - err);
+#endif // LIBC_TARGET_CPU_HAS_FMA_DOUBLE
+
+  if (LIBC_LIKELY(r_upper == r_lower))
+    return r_upper;
+
+  // Ziv's accuracy test failed, we redo the computations in Float128.
+  // Recalculate mod 1/64.
+  idx = static_cast<unsigned>(fputil::nearest_integer(u * 0x1.0p6));
+
+  // After the first step of Newton-Raphson approximating v = sqrt(u), we have
+  // that:
+  //   sqrt(u) = v_hi + h / (sqrt(u) + v_hi)
+  //      v_lo = h / (2 * v_hi)
+  // With error:
+  //   sqrt(u) - (v_hi + v_lo) = h * ( 1/(sqrt(u) + v_hi) - 1/(2*v_hi) )
+  //                           = -h^2 / (2*v * (sqrt(u) + v)^2).
+  // Since:
+  //   (sqrt(u) + v_hi)^2 ~ (2sqrt(u))^2 = 4u,
+  // we can add another correction term to (v_hi + v_lo) that is:
+  //   v_ll = -h^2 / (2*v_hi * 4u)
+  //        = -v_lo * (h / 4u)
+  //        = -vl * (h / 8u),
+  // making the errors:
+  //   sqrt(u) - (v_hi + v_lo + v_ll) = O(h^3)
+  // well beyond 128-bit precision needed.
+
+  // Get the rounding error of vl = 2 * v_lo ~ h / vh
+  // Get full product of vh * vl
+#ifdef LIBC_TARGET_CPU_HAS_FMA_DOUBLE
+  double vl_lo = fputil::multiply_add(-v_hi, vl, h) / v_hi;
+#else
+  DoubleDouble vh_vl = fputil::exact_mult(v_hi, vl);
+  double vl_lo = ((h - vh_vl.hi) - vh_vl.lo) / v_hi;
+#endif // LIBC_TARGET_CPU_HAS_FMA_DOUBLE
+  // vll = 2*v_ll = -vl * (h / (4u)).
+  double t = h * (-0.25) / u;
+  double vll = fputil::multiply_add(vl, t, vl_lo);
+  // m_v = -(v_hi + v_lo + v_ll).
+  Float128 m_v = fputil::quick_add(
+      Float128(vh), fputil::quick_add(Float128(vl), Float128(vll)));
+  m_v.sign = Sign::NEG;
+
+  // Perform computations in Float128:
+  //   asin(x) = pi/2 - (v_hi + v_lo + vll) * P(u).
+  Float128 y_f128(fputil::multiply_add(static_cast<double>(idx), -0x1.0p-6, u));
+
+  Float128 p_f128 = asin_eval(y_f128, idx);
+  Float128 r0_f128 = fputil::quick_mul(m_v, p_f128);
+  Float128 r_f128 = fputil::quick_add(PI_OVER_TWO_F128, r0_f128);
+
+  if (xbits.is_neg())
+    r_f128.sign = Sign::NEG;
+
+  return static_cast<double>(r_f128);
+#endif // LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+}
+
+} // namespace math
+
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LLVM_LIBC_SRC___SUPPORT_MATH_ASIN_H
diff --git a/libc/src/__support/math/asinf.h b/libc/src/__support/math/asinf.h
new file mode 100644
index 0000000..bfa0dc3
--- /dev/null
+++ b/libc/src/__support/math/asinf.h
@@ -0,0 +1,175 @@
+//===-- Implementation header for asinf -------------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC___SUPPORT_MATH_ASINF_H
+#define LLVM_LIBC_SRC___SUPPORT_MATH_ASINF_H
+
+#include "inv_trigf_utils.h"
+#include "src/__support/FPUtil/FEnvImpl.h"
+#include "src/__support/FPUtil/FPBits.h"
+#include "src/__support/FPUtil/except_value_utils.h"
+#include "src/__support/FPUtil/multiply_add.h"
+#include "src/__support/FPUtil/sqrt.h"
+#include "src/__support/macros/config.h"
+#include "src/__support/macros/optimization.h"            // LIBC_UNLIKELY
+#include "src/__support/macros/properties/cpu_features.h" // LIBC_TARGET_CPU_HAS_FMA
+
+namespace LIBC_NAMESPACE_DECL {
+
+namespace math {
+
+LIBC_INLINE static constexpr float asinf(float x) {
+  using namespace inv_trigf_utils_internal;
+
+#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+  constexpr size_t N_EXCEPTS = 2;
+
+  // Exceptional values when |x| <= 0.5
+  constexpr fputil::ExceptValues<float, N_EXCEPTS> ASINF_EXCEPTS_LO = {{
+      // (inputs, RZ output, RU offset, RD offset, RN offset)
+      // x = 0x1.137f0cp-5, asinf(x) = 0x1.138c58p-5 (RZ)
+      {0x3d09bf86, 0x3d09c62c, 1, 0, 1},
+      // x = 0x1.cbf43cp-4, asinf(x) = 0x1.cced1cp-4 (RZ)
+      {0x3de5fa1e, 0x3de6768e, 1, 0, 0},
+  }};
+
+  // Exceptional values when 0.5 < |x| <= 1
+  constexpr fputil::ExceptValues<float, N_EXCEPTS> ASINF_EXCEPTS_HI = {{
+      // (inputs, RZ output, RU offset, RD offset, RN offset)
+      // x = 0x1.107434p-1, asinf(x) = 0x1.1f4b64p-1 (RZ)
+      {0x3f083a1a, 0x3f0fa5b2, 1, 0, 0},
+      // x = 0x1.ee836cp-1, asinf(x) = 0x1.4f0654p0 (RZ)
+      {0x3f7741b6, 0x3fa7832a, 1, 0, 0},
+  }};
+#endif // !LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+
+  using namespace inv_trigf_utils_internal;
+  using FPBits = typename fputil::FPBits<float>;
+
+  FPBits xbits(x);
+  uint32_t x_uint = xbits.uintval();
+  uint32_t x_abs = xbits.uintval() & 0x7fff'ffffU;
+  constexpr double SIGN[2] = {1.0, -1.0};
+  uint32_t x_sign = x_uint >> 31;
+
+  // |x| <= 0.5-ish
+  if (x_abs < 0x3f04'471dU) {
+    // |x| < 0x1.d12edp-12
+    if (LIBC_UNLIKELY(x_abs < 0x39e8'9768U)) {
+      // When |x| < 2^-12, the relative error of the approximation asin(x) ~ x
+      // is:
+      //   |asin(x) - x| / |asin(x)| < |x^3| / (6|x|)
+      //                             = x^2 / 6
+      //                             < 2^-25
+      //                             < epsilon(1)/2.
+      // So the correctly rounded values of asin(x) are:
+      //   = x + sign(x)*eps(x) if rounding mode = FE_TOWARDZERO,
+      //                        or (rounding mode = FE_UPWARD and x is
+      //                        negative),
+      //   = x otherwise.
+      // To simplify the rounding decision and make it more efficient, we use
+      //   fma(x, 2^-25, x) instead.
+      // An exhaustive test shows that this formula work correctly for all
+      // rounding modes up to |x| < 0x1.d12edp-12.
+      // Note: to use the formula x + 2^-25*x to decide the correct rounding, we
+      // do need fma(x, 2^-25, x) to prevent underflow caused by 2^-25*x when
+      // |x| < 2^-125. For targets without FMA instructions, we simply use
+      // double for intermediate results as it is more efficient than using an
+      // emulated version of FMA.
+#if defined(LIBC_TARGET_CPU_HAS_FMA_FLOAT)
+      return fputil::multiply_add(x, 0x1.0p-25f, x);
+#else
+      double xd = static_cast<double>(x);
+      return static_cast<float>(fputil::multiply_add(xd, 0x1.0p-25, xd));
+#endif // LIBC_TARGET_CPU_HAS_FMA_FLOAT
+    }
+
+#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+    // Check for exceptional values
+    if (auto r = ASINF_EXCEPTS_LO.lookup_odd(x_abs, x_sign);
+        LIBC_UNLIKELY(r.has_value()))
+      return r.value();
+#endif // !LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+
+    // For |x| <= 0.5, we approximate asinf(x) by:
+    //   asin(x) = x * P(x^2)
+    // Where P(X^2) = Q(X) is a degree-20 minimax even polynomial approximating
+    // asin(x)/x on [0, 0.5] generated by Sollya with:
+    // > Q = fpminimax(asin(x)/x, [|0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20|],
+    //                 [|1, D...|], [0, 0.5]);
+    // An exhaustive test shows that this approximation works well up to a
+    // little more than 0.5.
+    double xd = static_cast<double>(x);
+    double xsq = xd * xd;
+    double x3 = xd * xsq;
+    double r = asin_eval(xsq);
+    return static_cast<float>(fputil::multiply_add(x3, r, xd));
+  }
+
+  // |x| > 1, return NaNs.
+  if (LIBC_UNLIKELY(x_abs > 0x3f80'0000U)) {
+    if (xbits.is_signaling_nan()) {
+      fputil::raise_except_if_required(FE_INVALID);
+      return FPBits::quiet_nan().get_val();
+    }
+
+    if (x_abs <= 0x7f80'0000U) {
+      fputil::set_errno_if_required(EDOM);
+      fputil::raise_except_if_required(FE_INVALID);
+    }
+
+    return FPBits::quiet_nan().get_val();
+  }
+
+#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+  // Check for exceptional values
+  if (auto r = ASINF_EXCEPTS_HI.lookup_odd(x_abs, x_sign);
+      LIBC_UNLIKELY(r.has_value()))
+    return r.value();
+#endif // !LIBC_MATH_HAS_SKIP_ACCURATE_PASS
+
+  // When |x| > 0.5, we perform range reduction as follow:
+  //
+  // Assume further that 0.5 < x <= 1, and let:
+  //   y = asin(x)
+  // We will use the double angle formula:
+  //   cos(2y) = 1 - 2 sin^2(y)
+  // and the complement angle identity:
+  //   x = sin(y) = cos(pi/2 - y)
+  //              = 1 - 2 sin^2 (pi/4 - y/2)
+  // So:
+  //   sin(pi/4 - y/2) = sqrt( (1 - x)/2 )
+  // And hence:
+  //   pi/4 - y/2 = asin( sqrt( (1 - x)/2 ) )
+  // Equivalently:
+  //   asin(x) = y = pi/2 - 2 * asin( sqrt( (1 - x)/2 ) )
+  // Let u = (1 - x)/2, then:
+  //   asin(x) = pi/2 - 2 * asin( sqrt(u) )
+  // Moreover, since 0.5 < x <= 1:
+  //   0 <= u < 1/4, and 0 <= sqrt(u) < 0.5,
+  // And hence we can reuse the same polynomial approximation of asin(x) when
+  // |x| <= 0.5:
+  //   asin(x) ~ pi/2 - 2 * sqrt(u) * P(u),
+
+  xbits.set_sign(Sign::POS);
+  double sign = SIGN[x_sign];
+  double xd = static_cast<double>(xbits.get_val());
+  double u = fputil::multiply_add(-0.5, xd, 0.5);
+  double c1 = sign * (-2 * fputil::sqrt<double>(u));
+  double c2 = fputil::multiply_add(sign, M_MATH_PI_2, c1);
+  double c3 = c1 * u;
+
+  double r = asin_eval(u);
+  return static_cast<float>(fputil::multiply_add(c3, r, c2));
+}
+
+} // namespace math
+
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LLVM_LIBC_SRC___SUPPORT_MATH_ASINF_H
diff --git a/libc/src/__support/math/asinf16.h b/libc/src/__support/math/asinf16.h
new file mode 100644
index 0000000..3d032a4
--- /dev/null
+++ b/libc/src/__support/math/asinf16.h
@@ -0,0 +1,146 @@
+//===-- Implementation header for asinf16 -----------------------*- C++ -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC___SUPPORT_MATH_ASINF16_H
+#define LLVM_LIBC_SRC___SUPPORT_MATH_ASINF16_H
+
+#include "include/llvm-libc-macros/float16-macros.h"
+
+#ifdef LIBC_TYPES_HAS_FLOAT16
+
+#include "src/__support/FPUtil/FEnvImpl.h"
+#include "src/__support/FPUtil/FPBits.h"
+#include "src/__support/FPUtil/PolyEval.h"
+#include "src/__support/FPUtil/cast.h"
+#include "src/__support/FPUtil/multiply_add.h"
+#include "src/__support/FPUtil/sqrt.h"
+#include "src/__support/macros/optimization.h"
+
+namespace LIBC_NAMESPACE_DECL {
+
+namespace math {
+
+LIBC_INLINE static constexpr float16 asinf16(float16 x) {
+
+  // Generated by Sollya using the following command:
+  // > round(pi/2, D, RN);
+  constexpr float PI_2 = 0x1.921fb54442d18p0f;
+
+  using FPBits = fputil::FPBits<float16>;
+  FPBits xbits(x);
+
+  uint16_t x_u = xbits.uintval();
+  uint16_t x_abs = x_u & 0x7fff;
+  float xf = x;
+
+  // |x| > 0x1p0, |x| > 1, or x is NaN.
+  if (LIBC_UNLIKELY(x_abs > 0x3c00)) {
+    // asinf16(NaN) = NaN
+    if (xbits.is_nan()) {
+      if (xbits.is_signaling_nan()) {
+        fputil::raise_except_if_required(FE_INVALID);
+        return FPBits::quiet_nan().get_val();
+      }
+
+      return x;
+    }
+
+    // 1 < |x| <= +/-inf
+    fputil::raise_except_if_required(FE_INVALID);
+    fputil::set_errno_if_required(EDOM);
+
+    return FPBits::quiet_nan().get_val();
+  }
+
+  float xsq = xf * xf;
+
+  // |x| <= 0x1p-1, |x| <= 0.5
+  if (x_abs <= 0x3800) {
+    // asinf16(+/-0) = +/-0
+    if (LIBC_UNLIKELY(x_abs == 0))
+      return x;
+
+    // Exhaustive tests show that,
+    // for |x| <= 0x1.878p-9, when:
+    // x > 0, and rounding upward, or
+    // x < 0, and rounding downward, then,
+    // asin(x) = x * 2^-11 + x
+    // else, in other rounding modes,
+    // asin(x) = x
+    if (LIBC_UNLIKELY(x_abs <= 0x1a1e)) {
+      int rounding = fputil::quick_get_round();
+
+      if ((xbits.is_pos() && rounding == FE_UPWARD) ||
+          (xbits.is_neg() && rounding == FE_DOWNWARD))
+        return fputil::cast<float16>(fputil::multiply_add(xf, 0x1.0p-11f, xf));
+      return x;
+    }
+
+    // Degree-6 minimax odd polynomial of asin(x) generated by Sollya with:
+    // > P = fpminimax(asin(x)/x, [|0, 2, 4, 6, 8|], [|SG...|], [0, 0.5]);
+    float result =
+        fputil::polyeval(xsq, 0x1.000002p0f, 0x1.554c2ap-3f, 0x1.3541ccp-4f,
+                         0x1.43b2d6p-5f, 0x1.a0d73ep-5f);
+    return fputil::cast<float16>(xf * result);
+  }
+
+  // When |x| > 0.5, assume that 0.5 < |x| <= 1,
+  //
+  // Step-by-step range-reduction proof:
+  // 1:  Let y = asin(x), such that, x = sin(y)
+  // 2:  From complimentary angle identity:
+  //       x = sin(y) = cos(pi/2 - y)
+  // 3:  Let z = pi/2 - y, such that x = cos(z)
+  // 4:  From double angle formula; cos(2A) = 1 - sin^2(A):
+  //       z = 2A, z/2 = A
+  //       cos(z) = 1 - 2 * sin^2(z/2)
+  // 5:  Make sin(z/2) subject of the formula:
+  //       sin(z/2) = sqrt((1 - cos(z))/2)
+  // 6:  Recall [3]; x = cos(z). Therefore:
+  //       sin(z/2) = sqrt((1 - x)/2)
+  // 7:  Let u = (1 - x)/2
+  // 8:  Therefore:
+  //       asin(sqrt(u)) = z/2
+  //       2 * asin(sqrt(u)) = z
+  // 9:  Recall [3], z = pi/2 - y. Therefore:
+  //       y = pi/2 - z
+  //       y = pi/2 - 2 * asin(sqrt(u))
+  // 10: Recall [1], y = asin(x). Therefore:
+  //       asin(x) = pi/2 - 2 * asin(sqrt(u))
+  //
+  // WHY?
+  // 11: Recall [7], u = (1 - x)/2
+  // 12: Since 0.5 < x <= 1, therefore:
+  //       0 <= u <= 0.25 and 0 <= sqrt(u) <= 0.5
+  //
+  // Hence, we can reuse the same [0, 0.5] domain polynomial approximation for
+  // Step [10] as `sqrt(u)` is in range.
+
+  // 0x1p-1 < |x| <= 0x1p0, 0.5 < |x| <= 1.0
+  float xf_abs = (xf < 0 ? -xf : xf);
+  float sign = (xbits.uintval() >> 15 == 1 ? -1.0 : 1.0);
+  float u = fputil::multiply_add(-0.5f, xf_abs, 0.5f);
+  float u_sqrt = fputil::sqrt<float>(u);
+
+  // Degree-6 minimax odd polynomial of asin(x) generated by Sollya with:
+  // > P = fpminimax(asin(x)/x, [|0, 2, 4, 6, 8|], [|SG...|], [0, 0.5]);
+  float asin_sqrt_u =
+      u_sqrt * fputil::polyeval(u, 0x1.000002p0f, 0x1.554c2ap-3f,
+                                0x1.3541ccp-4f, 0x1.43b2d6p-5f, 0x1.a0d73ep-5f);
+
+  return fputil::cast<float16>(sign *
+                               fputil::multiply_add(-2.0f, asin_sqrt_u, PI_2));
+}
+
+} // namespace math
+
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LIBC_TYPES_HAS_FLOAT16
+
+#endif // LLVM_LIBC_SRC___SUPPORT_MATH_ASINF16_H
diff --git a/libc/src/__support/math/erff.h b/libc/src/__support/math/erff.h
index e54ec77..b81be30 100644
--- a/libc/src/__support/math/erff.h
+++ b/libc/src/__support/math/erff.h
@@ -19,7 +19,7 @@ namespace LIBC_NAMESPACE_DECL {
 
 namespace math {
 
-static constexpr float erff(float x) {
+LIBC_INLINE static constexpr float erff(float x) {
 
   // Polynomials approximating erf(x)/x on ( k/8, (k + 1)/8 ) generated by
   // Sollya with: > P = fpminimax(erf(x)/x, [|0, 2, 4, 6, 8, 10, 12, 14|],
diff --git a/libc/src/__support/math/exp.h b/libc/src/__support/math/exp.h
index ff59ff7..83638e8 100644
--- a/libc/src/__support/math/exp.h
+++ b/libc/src/__support/math/exp.h
@@ -67,7 +67,7 @@ namespace {
 // Return expm1(dx) / x ~ 1 + dx / 2 + dx^2 / 6 + dx^3 / 24.
 // For |dx| < 2^-13 + 2^-30:
 //   | output - expm1(dx) / dx | < 2^-51.
-static double poly_approx_d(double dx) {
+LIBC_INLINE static double poly_approx_d(double dx) {
   // dx^2
   double dx2 = dx * dx;
   // c0 = 1 + dx / 2
@@ -85,7 +85,7 @@ static double poly_approx_d(double dx) {
 // Return exp(dx) ~ 1 + dx + dx^2 / 2 + ... + dx^6 / 720
 // For |dx| < 2^-13 + 2^-30:
 //   | output - exp(dx) | < 2^-101
-static DoubleDouble poly_approx_dd(const DoubleDouble &dx) {
+LIBC_INLINE static DoubleDouble poly_approx_dd(const DoubleDouble &dx) {
   // Taylor polynomial.
   constexpr DoubleDouble COEFFS[] = {
       {0, 0x1p0},                                      // 1
@@ -106,7 +106,7 @@ static DoubleDouble poly_approx_dd(const DoubleDouble &dx) {
 // Return exp(dx) ~ 1 + dx + dx^2 / 2 + ... + dx^7 / 5040
 // For |dx| < 2^-13 + 2^-30:
 //   | output - exp(dx) | < 2^-126.
-static Float128 poly_approx_f128(const Float128 &dx) {
+LIBC_INLINE static Float128 poly_approx_f128(const Float128 &dx) {
   constexpr Float128 COEFFS_128[]{
       {Sign::POS, -127, 0x80000000'00000000'00000000'00000000_u128}, // 1.0
       {Sign::POS, -127, 0x80000000'00000000'00000000'00000000_u128}, // 1.0
@@ -127,7 +127,7 @@ static Float128 poly_approx_f128(const Float128 &dx) {
 // Compute exp(x) using 128-bit precision.
 // TODO(lntue): investigate triple-double precision implementation for this
 // step.
-static Float128 exp_f128(double x, double kd, int idx1, int idx2) {
+LIBC_INLINE static Float128 exp_f128(double x, double kd, int idx1, int idx2) {
   // Recalculate dx:
 
   double t1 = fputil::multiply_add(kd, MLOG_2_EXP2_M12_HI, x); // exact
@@ -160,8 +160,8 @@ static Float128 exp_f128(double x, double kd, int idx1, int idx2) {
 }
 
 // Compute exp(x) with double-double precision.
-static DoubleDouble exp_double_double(double x, double kd,
-                                      const DoubleDouble &exp_mid) {
+LIBC_INLINE static DoubleDouble exp_double_double(double x, double kd,
+                                                  const DoubleDouble &exp_mid) {
   // Recalculate dx:
   //   dx = x - k * 2^-12 * log(2)
   double t1 = fputil::multiply_add(kd, MLOG_2_EXP2_M12_HI, x); // exact
@@ -184,7 +184,7 @@ static DoubleDouble exp_double_double(double x, double kd,
 
 // Check for exceptional cases when
 // |x| <= 2^-53 or x < log(2^-1075) or x >= 0x1.6232bdd7abcd3p+9
-static double set_exceptional(double x) {
+LIBC_INLINE static double set_exceptional(double x) {
   using FPBits = typename fputil::FPBits<double>;
   FPBits xbits(x);
 
@@ -234,7 +234,7 @@ static double set_exceptional(double x) {
 
 namespace math {
 
-static double exp(double x) {
+LIBC_INLINE static double exp(double x) {
   using FPBits = typename fputil::FPBits<double>;
   FPBits xbits(x);
 
diff --git a/libc/src/__support/math/exp10.h b/libc/src/__support/math/exp10.h
index fa60e40c..12a09d7 100644
--- a/libc/src/__support/math/exp10.h
+++ b/libc/src/__support/math/exp10.h
@@ -83,7 +83,8 @@ LIBC_INLINE static double exp10_poly_approx_d(double dx) {
 // > P = fpminimax((10^x - 1)/x, 5, [|DD...|], [-2^-14, 2^-14]);
 // Error bounds:
 //   | output - 10^(dx) | < 2^-101
-static constexpr DoubleDouble exp10_poly_approx_dd(const DoubleDouble &dx) {
+LIBC_INLINE static constexpr DoubleDouble
+exp10_poly_approx_dd(const DoubleDouble &dx) {
   // Taylor polynomial.
   constexpr DoubleDouble COEFFS[] = {
       {0, 0x1p0},
@@ -105,7 +106,8 @@ static constexpr DoubleDouble exp10_poly_approx_dd(const DoubleDouble &dx) {
 // Return exp(dx) ~ 1 + a0 * dx + a1 * dx^2 + ... + a6 * dx^7
 // For |dx| < 2^-14:
 //   | output - 10^dx | < 1.5 * 2^-124.
-static constexpr Float128 exp10_poly_approx_f128(const Float128 &dx) {
+LIBC_INLINE static constexpr Float128
+exp10_poly_approx_f128(const Float128 &dx) {
   constexpr Float128 COEFFS_128[]{
       {Sign::POS, -127, 0x80000000'00000000'00000000'00000000_u128}, // 1.0
       {Sign::POS, -126, 0x935d8ddd'aaa8ac16'ea56d62b'82d30a2d_u128},
@@ -126,7 +128,8 @@ static constexpr Float128 exp10_poly_approx_f128(const Float128 &dx) {
 // Compute 10^(x) using 128-bit precision.
 // TODO(lntue): investigate triple-double precision implementation for this
 // step.
-static Float128 exp10_f128(double x, double kd, int idx1, int idx2) {
+LIBC_INLINE static Float128 exp10_f128(double x, double kd, int idx1,
+                                       int idx2) {
   double t1 = fputil::multiply_add(kd, MLOG10_2_EXP2_M12_HI, x); // exact
   double t2 = kd * MLOG10_2_EXP2_M12_MID_32;                     // exact
   double t3 = kd * MLOG10_2_EXP2_M12_LO; // Error < 2^-144
@@ -157,8 +160,8 @@ static Float128 exp10_f128(double x, double kd, int idx1, int idx2) {
 }
 
 // Compute 10^x with double-double precision.
-static DoubleDouble exp10_double_double(double x, double kd,
-                                        const DoubleDouble &exp_mid) {
+LIBC_INLINE static DoubleDouble
+exp10_double_double(double x, double kd, const DoubleDouble &exp_mid) {
   // Recalculate dx:
   //   dx = x - k * 2^-12 * log10(2)
   double t1 = fputil::multiply_add(kd, MLOG10_2_EXP2_M12_HI, x); // exact
@@ -180,7 +183,7 @@ static DoubleDouble exp10_double_double(double x, double kd,
 #endif // LIBC_MATH_HAS_SKIP_ACCURATE_PASS
 
 // When output is denormal.
-static double exp10_denorm(double x) {
+LIBC_INLINE static double exp10_denorm(double x) {
   // Range reduction.
   double tmp = fputil::multiply_add(x, LOG2_10, 0x1.8000'0000'4p21);
   int k = static_cast<int>(cpp::bit_cast<uint64_t>(tmp) >> 19);
@@ -234,7 +237,7 @@ static double exp10_denorm(double x) {
 //  * x >= log10(2^1024)
 //  * x <= log10(2^-1022)
 //  * x is inf or nan
-static constexpr double exp10_set_exceptional(double x) {
+LIBC_INLINE static constexpr double exp10_set_exceptional(double x) {
   using FPBits = typename fputil::FPBits<double>;
   FPBits xbits(x);
 
@@ -285,7 +288,7 @@ static constexpr double exp10_set_exceptional(double x) {
 
 namespace math {
 
-static constexpr double exp10(double x) {
+LIBC_INLINE static constexpr double exp10(double x) {
   using FPBits = typename fputil::FPBits<double>;
   FPBits xbits(x);
 
diff --git a/libc/src/__support/math/exp10f.h b/libc/src/__support/math/exp10f.h
index 807b4f0..76ae197 100644
--- a/libc/src/__support/math/exp10f.h
+++ b/libc/src/__support/math/exp10f.h
@@ -20,7 +20,7 @@
 namespace LIBC_NAMESPACE_DECL {
 namespace math {
 
-static constexpr float exp10f(float x) {
+LIBC_INLINE static constexpr float exp10f(float x) {
   using FPBits = typename fputil::FPBits<float>;
   FPBits xbits(x);
 
diff --git a/libc/src/__support/math/exp10f16.h b/libc/src/__support/math/exp10f16.h
index 0d8b125..3eca867 100644
--- a/libc/src/__support/math/exp10f16.h
+++ b/libc/src/__support/math/exp10f16.h
@@ -57,7 +57,7 @@ static constexpr fputil::ExceptValues<float16, N_EXP10F16_EXCEPTS>
     }};
 #endif // !LIBC_MATH_HAS_SKIP_ACCURATE_PASS
 
-static constexpr float16 exp10f16(float16 x) {
+LIBC_INLINE static constexpr float16 exp10f16(float16 x) {
   using FPBits = fputil::FPBits<float16>;
   FPBits x_bits(x);
 
diff --git a/libc/src/__support/math/exp10f16_utils.h b/libc/src/__support/math/exp10f16_utils.h
index bffb81b..5952a41 100644
--- a/libc/src/__support/math/exp10f16_utils.h
+++ b/libc/src/__support/math/exp10f16_utils.h
@@ -19,8 +19,7 @@
 
 namespace LIBC_NAMESPACE_DECL {
 
-LIBC_INLINE static constexpr ExpRangeReduction
-exp10_range_reduction(float16 x) {
+LIBC_INLINE static ExpRangeReduction exp10_range_reduction(float16 x) {
   // For -8 < x < 5, to compute 10^x, we perform the following range reduction:
   // find hi, mid, lo, such that:
   //   x = (hi + mid) * log2(10) + lo, in which
diff --git a/libc/src/__support/math/exp10f_utils.h b/libc/src/__support/math/exp10f_utils.h
index c30def9..010a2f1 100644
--- a/libc/src/__support/math/exp10f_utils.h
+++ b/libc/src/__support/math/exp10f_utils.h
@@ -89,7 +89,7 @@ struct Exp10Base : public ExpBase {
                                        0x1.0470591dff149p1, 0x1.2bd7c0a9fbc4dp0,
                                        0x1.1429e74a98f43p-1};
 
-  static double powb_lo(double dx) {
+  LIBC_INLINE static double powb_lo(double dx) {
     using fputil::multiply_add;
     double dx2 = dx * dx;
     // c0 = 1 + COEFFS[0] * dx
diff --git a/libc/src/__support/math/exp_utils.h b/libc/src/__support/math/exp_utils.h
index fc9ab10..ef408ed 100644
--- a/libc/src/__support/math/exp_utils.h
+++ b/libc/src/__support/math/exp_utils.h
@@ -22,8 +22,8 @@ namespace LIBC_NAMESPACE_DECL {
 // So if we scale x up by 2^1022, we can use
 //   double(1.0 + 2^1022 * x) - 1.0 to test how x is rounded in denormal range.
 template <bool SKIP_ZIV_TEST = false>
-static constexpr cpp::optional<double> ziv_test_denorm(int hi, double mid,
-                                                       double lo, double err) {
+LIBC_INLINE static constexpr cpp::optional<double>
+ziv_test_denorm(int hi, double mid, double lo, double err) {
   using FPBits = typename fputil::FPBits<double>;
 
   // Scaling factor = 1/(min normal number) = 2^1022
diff --git a/libc/src/__support/math/expf.h b/libc/src/__support/math/expf.h
index 88c1514..f7e11be 100644
--- a/libc/src/__support/math/expf.h
+++ b/libc/src/__support/math/expf.h
@@ -24,7 +24,7 @@ namespace LIBC_NAMESPACE_DECL {
 
 namespace math {
 
-static constexpr float expf(float x) {
+LIBC_INLINE static constexpr float expf(float x) {
   using FPBits = typename fputil::FPBits<float>;
   FPBits xbits(x);
 
diff --git a/libc/src/__support/math/expf16.h b/libc/src/__support/math/expf16.h
index ded28c7..14302a7 100644
--- a/libc/src/__support/math/expf16.h
+++ b/libc/src/__support/math/expf16.h
@@ -31,7 +31,7 @@ namespace LIBC_NAMESPACE_DECL {
 
 namespace math {
 
-static constexpr float16 expf16(float16 x) {
+LIBC_INLINE static constexpr float16 expf16(float16 x) {
 #ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
   constexpr fputil::ExceptValues<float16, 2> EXPF16_EXCEPTS_LO = {{
       // (input, RZ output, RU offset, RD offset, RN offset)
diff --git a/libc/src/__support/math/expf16_utils.h b/libc/src/__support/math/expf16_utils.h
index 8a2fc94..4204dab7 100644
--- a/libc/src/__support/math/expf16_utils.h
+++ b/libc/src/__support/math/expf16_utils.h
@@ -47,7 +47,8 @@ struct ExpRangeReduction {
   float exp_lo;
 };
 
-[[maybe_unused]] static ExpRangeReduction exp_range_reduction(float16 x) {
+[[maybe_unused]] LIBC_INLINE static ExpRangeReduction
+exp_range_reduction(float16 x) {
   // For -18 < x < 12, to compute exp(x), we perform the following range
   // reduction: find hi, mid, lo, such that:
   //   x = hi + mid + lo, in which
diff --git a/libc/src/__support/math/frexpf.h b/libc/src/__support/math/frexpf.h
index 4d2f494..7834a12 100644
--- a/libc/src/__support/math/frexpf.h
+++ b/libc/src/__support/math/frexpf.h
@@ -17,7 +17,7 @@ namespace LIBC_NAMESPACE_DECL {
 
 namespace math {
 
-static constexpr float frexpf(float x, int *exp) {
+LIBC_INLINE static constexpr float frexpf(float x, int *exp) {
   return fputil::frexp(x, *exp);
 }
 
diff --git a/libc/src/__support/math/frexpf128.h b/libc/src/__support/math/frexpf128.h
index 2fd5bc4..5218b26 100644
--- a/libc/src/__support/math/frexpf128.h
+++ b/libc/src/__support/math/frexpf128.h
@@ -21,7 +21,7 @@ namespace LIBC_NAMESPACE_DECL {
 
 namespace math {
 
-static constexpr float128 frexpf128(float128 x, int *exp) {
+LIBC_INLINE static constexpr float128 frexpf128(float128 x, int *exp) {
   return fputil::frexp(x, *exp);
 }
 
diff --git a/libc/src/__support/math/frexpf16.h b/libc/src/__support/math/frexpf16.h
index 8deeba0..530b61a 100644
--- a/libc/src/__support/math/frexpf16.h
+++ b/libc/src/__support/math/frexpf16.h
@@ -21,7 +21,7 @@ namespace LIBC_NAMESPACE_DECL {
 
 namespace math {
 
-static constexpr float16 frexpf16(float16 x, int *exp) {
+LIBC_INLINE static constexpr float16 frexpf16(float16 x, int *exp) {
   return fputil::frexp(x, *exp);
 }
 
diff --git a/libc/src/__support/math/ldexpf.h b/libc/src/__support/math/ldexpf.h
index 3a5ec1d..9ef5d96 100644
--- a/libc/src/__support/math/ldexpf.h
+++ b/libc/src/__support/math/ldexpf.h
@@ -17,7 +17,7 @@ namespace LIBC_NAMESPACE_DECL {
 
 namespace math {
 
-static constexpr float ldexpf(float x, int exp) {
+LIBC_INLINE static constexpr float ldexpf(float x, int exp) {
   return fputil::ldexp(x, exp);
 }
 
diff --git a/libc/src/__support/math/ldexpf128.h b/libc/src/__support/math/ldexpf128.h
index 3625830..4fba20c 100644
--- a/libc/src/__support/math/ldexpf128.h
+++ b/libc/src/__support/math/ldexpf128.h
@@ -21,7 +21,7 @@ namespace LIBC_NAMESPACE_DECL {
 
 namespace math {
 
-static constexpr float128 ldexpf128(float128 x, int exp) {
+LIBC_INLINE static constexpr float128 ldexpf128(float128 x, int exp) {
   return fputil::ldexp(x, exp);
 }
 
diff --git a/libc/src/__support/math/ldexpf16.h b/libc/src/__support/math/ldexpf16.h
index fbead87..d978d22 100644
--- a/libc/src/__support/math/ldexpf16.h
+++ b/libc/src/__support/math/ldexpf16.h
@@ -21,7 +21,7 @@ namespace LIBC_NAMESPACE_DECL {
 
 namespace math {
 
-static constexpr float16 ldexpf16(float16 x, int exp) {
+LIBC_INLINE static constexpr float16 ldexpf16(float16 x, int exp) {
   return fputil::ldexp(x, exp);
 }
 
diff --git a/libc/src/math/generic/CMakeLists.txt b/libc/src/math/generic/CMakeLists.txt
index a001d99..d4d268c 100644
--- a/libc/src/math/generic/CMakeLists.txt
+++ b/libc/src/math/generic/CMakeLists.txt
@@ -3958,13 +3958,7 @@ add_entrypoint_object(
   HDRS
     ../asinf.h
   DEPENDS
-    libc.src.__support.FPUtil.except_value_utils
-    libc.src.__support.FPUtil.fp_bits
-    libc.src.__support.FPUtil.multiply_add
-    libc.src.__support.FPUtil.polyeval
-    libc.src.__support.FPUtil.sqrt
-    libc.src.__support.macros.optimization
-    libc.src.__support.math.inv_trigf_utils
+    libc.src.__support.math.asinf
 )
 
 add_entrypoint_object(
@@ -3974,16 +3968,7 @@ add_entrypoint_object(
   HDRS
     ../asinf16.h
   DEPENDS
-    libc.hdr.errno_macros
-    libc.hdr.fenv_macros
-    libc.src.__support.FPUtil.cast
-    libc.src.__support.FPUtil.fenv_impl
-    libc.src.__support.FPUtil.fp_bits
-    libc.src.__support.FPUtil.multiply_add
-    libc.src.__support.FPUtil.polyeval
-    libc.src.__support.FPUtil.sqrt
-    libc.src.__support.macros.optimization
-    libc.src.__support.macros.properties.types
+    libc.src.__support.math.asinf16
 )
 
 add_entrypoint_object(
@@ -3993,16 +3978,7 @@ add_entrypoint_object(
   HDRS
     ../asin.h
   DEPENDS
-    libc.src.__support.math.asin_utils
-    libc.src.__support.FPUtil.double_double
-    libc.src.__support.FPUtil.dyadic_float
-    libc.src.__support.FPUtil.fenv_impl
-    libc.src.__support.FPUtil.fp_bits
-    libc.src.__support.FPUtil.multiply_add
-    libc.src.__support.FPUtil.polyeval
-    libc.src.__support.FPUtil.sqrt
-    libc.src.__support.macros.optimization
-    libc.src.__support.macros.properties.cpu_features
+    libc.src.__support.math.asin
 )
 
 add_entrypoint_object(
@@ -4043,16 +4019,8 @@ add_entrypoint_object(
   HDRS
     ../acospif16.h
   DEPENDS
-    libc.hdr.errno_macros
-    libc.hdr.fenv_macros
-    libc.src.__support.FPUtil.cast
-    libc.src.__support.FPUtil.fenv_impl
-    libc.src.__support.FPUtil.fp_bits
-    libc.src.__support.FPUtil.multiply_add
-    libc.src.__support.FPUtil.polyeval
-    libc.src.__support.FPUtil.sqrt
-    libc.src.__support.macros.optimization
-    libc.src.__support.macros.properties.types
+    libc.src.__support.math.acospif16
+    libc.src.errno.errno
 )
 
 add_header_library(
diff --git a/libc/src/math/generic/acospif16.cpp b/libc/src/math/generic/acospif16.cpp
index bfdf169..09cbd99 100644
--- a/libc/src/math/generic/acospif16.cpp
+++ b/libc/src/math/generic/acospif16.cpp
@@ -7,128 +7,12 @@
 //===----------------------------------------------------------------------===//
 
 #include "src/math/acospif16.h"
-#include "hdr/errno_macros.h"
-#include "hdr/fenv_macros.h"
-#include "src/__support/FPUtil/FEnvImpl.h"
-#include "src/__support/FPUtil/FPBits.h"
-#include "src/__support/FPUtil/PolyEval.h"
-#include "src/__support/FPUtil/cast.h"
-#include "src/__support/FPUtil/multiply_add.h"
-#include "src/__support/FPUtil/sqrt.h"
-#include "src/__support/macros/optimization.h"
+#include "src/__support/math/acospif16.h"
 
 namespace LIBC_NAMESPACE_DECL {
 
 LLVM_LIBC_FUNCTION(float16, acospif16, (float16 x)) {
-  using FPBits = fputil::FPBits<float16>;
-  FPBits xbits(x);
-
-  uint16_t x_u = xbits.uintval();
-  uint16_t x_abs = x_u & 0x7fff;
-  uint16_t x_sign = x_u >> 15;
-
-  // |x| > 0x1p0, |x| > 1, or x is NaN.
-  if (LIBC_UNLIKELY(x_abs > 0x3c00)) {
-    // acospif16(NaN) = NaN
-    if (xbits.is_nan()) {
-      if (xbits.is_signaling_nan()) {
-        fputil::raise_except_if_required(FE_INVALID);
-        return FPBits::quiet_nan().get_val();
-      }
-
-      return x;
-    }
-
-    // 1 < |x| <= +inf
-    fputil::raise_except_if_required(FE_INVALID);
-    fputil::set_errno_if_required(EDOM);
-
-    return FPBits::quiet_nan().get_val();
-  }
-
-  // |x| == 0x1p0, x is 1 or -1
-  // if x is (-)1, return 1
-  // if x is (+)1, return 0
-  if (LIBC_UNLIKELY(x_abs == 0x3c00))
-    return fputil::cast<float16>(x_sign ? 1.0f : 0.0f);
-
-  float xf = x;
-  float xsq = xf * xf;
-
-  // Degree-6 minimax polynomial coefficients of asin(x) generated by Sollya
-  // with: > P = fpminimax(asin(x)/(pi * x), [|0, 2, 4, 6, 8|], [|SG...|], [0,
-  // 0.5]);
-  constexpr float POLY_COEFFS[5] = {0x1.45f308p-2f, 0x1.b2900cp-5f,
-                                    0x1.897e36p-6f, 0x1.9efafcp-7f,
-                                    0x1.06d884p-6f};
-  // |x| <= 0x1p-1, |x| <= 0.5
-  if (x_abs <= 0x3800) {
-    // if x is 0, return 0.5
-    if (LIBC_UNLIKELY(x_abs == 0))
-      return fputil::cast<float16>(0.5f);
-
-    // Note that: acos(x) = pi/2 + asin(-x) = pi/2 - asin(x), then
-    //            acospi(x) = 0.5 - asin(x)/pi
-    float interm =
-        fputil::polyeval(xsq, POLY_COEFFS[0], POLY_COEFFS[1], POLY_COEFFS[2],
-                         POLY_COEFFS[3], POLY_COEFFS[4]);
-
-    return fputil::cast<float16>(fputil::multiply_add(-xf, interm, 0.5f));
-  }
-
-  // When |x| > 0.5, assume that 0.5 < |x| <= 1
-  //
-  // Step-by-step range-reduction proof:
-  // 1:  Let y = asin(x), such that, x = sin(y)
-  // 2:  From complimentary angle identity:
-  //       x = sin(y) = cos(pi/2 - y)
-  // 3:  Let z = pi/2 - y, such that x = cos(z)
-  // 4:  From double angle formula; cos(2A) = 1 - 2 * sin^2(A):
-  //       z = 2A, z/2 = A
-  //       cos(z) = 1 - 2 * sin^2(z/2)
-  // 5:  Make sin(z/2) subject of the formula:
-  //       sin(z/2) = sqrt((1 - cos(z))/2)
-  // 6:  Recall [3]; x = cos(z). Therefore:
-  //       sin(z/2) = sqrt((1 - x)/2)
-  // 7:  Let u = (1 - x)/2
-  // 8:  Therefore:
-  //       asin(sqrt(u)) = z/2
-  //       2 * asin(sqrt(u)) = z
-  // 9:  Recall [3]; z = pi/2 - y. Therefore:
-  //       y = pi/2 - z
-  //       y = pi/2 - 2 * asin(sqrt(u))
-  // 10: Recall [1], y = asin(x). Therefore:
-  //       asin(x) = pi/2 - 2 * asin(sqrt(u))
-  // 11: Recall that: acos(x) = pi/2 + asin(-x) = pi/2 - asin(x)
-  //     Therefore:
-  //       acos(x) = pi/2 - (pi/2 - 2 * asin(sqrt(u)))
-  //       acos(x) = 2 * asin(sqrt(u))
-  //       acospi(x) = 2 * (asin(sqrt(u)) / pi)
-  //
-  // THE RANGE REDUCTION, HOW?
-  // 12: Recall [7], u = (1 - x)/2
-  // 13: Since 0.5 < x <= 1, therefore:
-  //       0 <= u <= 0.25 and 0 <= sqrt(u) <= 0.5
-  //
-  // Hence, we can reuse the same [0, 0.5] domain polynomial approximation for
-  // Step [11] as `sqrt(u)` is in range.
-  // When -1 < x <= -0.5, the identity:
-  //       acos(x) = pi - acos(-x)
-  //       acospi(x) = 1 - acos(-x)/pi
-  // allows us to compute for the negative x value (lhs)
-  // with a positive x value instead (rhs).
-
-  float xf_abs = (xf < 0 ? -xf : xf);
-  float u = fputil::multiply_add(-0.5f, xf_abs, 0.5f);
-  float sqrt_u = fputil::sqrt<float>(u);
-
-  float asin_sqrt_u =
-      sqrt_u * fputil::polyeval(u, POLY_COEFFS[0], POLY_COEFFS[1],
-                                POLY_COEFFS[2], POLY_COEFFS[3], POLY_COEFFS[4]);
-
-  // Same as acos(x), but devided the expression with pi
-  return fputil::cast<float16>(
-      x_sign ? fputil::multiply_add(-2.0f, asin_sqrt_u, 1.0f)
-             : 2.0f * asin_sqrt_u);
+  return math::acospif16(x);
 }
+
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/asin.cpp b/libc/src/math/generic/asin.cpp
index d286fce..b5ba9ea 100644
--- a/libc/src/math/generic/asin.cpp
+++ b/libc/src/math/generic/asin.cpp
@@ -7,23 +7,10 @@
 //===----------------------------------------------------------------------===//
 
 #include "src/math/asin.h"
-#include "src/__support/FPUtil/FEnvImpl.h"
-#include "src/__support/FPUtil/FPBits.h"
-#include "src/__support/FPUtil/PolyEval.h"
-#include "src/__support/FPUtil/double_double.h"
-#include "src/__support/FPUtil/dyadic_float.h"
-#include "src/__support/FPUtil/multiply_add.h"
-#include "src/__support/FPUtil/sqrt.h"
-#include "src/__support/macros/config.h"
-#include "src/__support/macros/optimization.h"            // LIBC_UNLIKELY
-#include "src/__support/macros/properties/cpu_features.h" // LIBC_TARGET_CPU_HAS_FMA
-#include "src/__support/math/asin_utils.h"
+#include "src/__support/math/asin.h"
 
 namespace LIBC_NAMESPACE_DECL {
 
-using DoubleDouble = fputil::DoubleDouble;
-using Float128 = fputil::DyadicFloat<128>;
-
 LLVM_LIBC_FUNCTION(double, asin, (double x)) {
   using namespace asin_internal;
   using FPBits = fputil::FPBits<double>;
diff --git a/libc/src/math/generic/asinf.cpp b/libc/src/math/generic/asinf.cpp
index 77d6de9..9c6766f 100644
--- a/libc/src/math/generic/asinf.cpp
+++ b/libc/src/math/generic/asinf.cpp
@@ -7,161 +7,10 @@
 //===----------------------------------------------------------------------===//
 
 #include "src/math/asinf.h"
-#include "src/__support/FPUtil/FEnvImpl.h"
-#include "src/__support/FPUtil/FPBits.h"
-#include "src/__support/FPUtil/PolyEval.h"
-#include "src/__support/FPUtil/except_value_utils.h"
-#include "src/__support/FPUtil/multiply_add.h"
-#include "src/__support/FPUtil/sqrt.h"
-#include "src/__support/macros/config.h"
-#include "src/__support/macros/optimization.h"            // LIBC_UNLIKELY
-#include "src/__support/macros/properties/cpu_features.h" // LIBC_TARGET_CPU_HAS_FMA
-
-#include "src/__support/math/inv_trigf_utils.h"
+#include "src/__support/math/asinf.h"
 
 namespace LIBC_NAMESPACE_DECL {
 
-#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
-static constexpr size_t N_EXCEPTS = 2;
-
-// Exceptional values when |x| <= 0.5
-static constexpr fputil::ExceptValues<float, N_EXCEPTS> ASINF_EXCEPTS_LO = {{
-    // (inputs, RZ output, RU offset, RD offset, RN offset)
-    // x = 0x1.137f0cp-5, asinf(x) = 0x1.138c58p-5 (RZ)
-    {0x3d09bf86, 0x3d09c62c, 1, 0, 1},
-    // x = 0x1.cbf43cp-4, asinf(x) = 0x1.cced1cp-4 (RZ)
-    {0x3de5fa1e, 0x3de6768e, 1, 0, 0},
-}};
-
-// Exceptional values when 0.5 < |x| <= 1
-static constexpr fputil::ExceptValues<float, N_EXCEPTS> ASINF_EXCEPTS_HI = {{
-    // (inputs, RZ output, RU offset, RD offset, RN offset)
-    // x = 0x1.107434p-1, asinf(x) = 0x1.1f4b64p-1 (RZ)
-    {0x3f083a1a, 0x3f0fa5b2, 1, 0, 0},
-    // x = 0x1.ee836cp-1, asinf(x) = 0x1.4f0654p0 (RZ)
-    {0x3f7741b6, 0x3fa7832a, 1, 0, 0},
-}};
-#endif // !LIBC_MATH_HAS_SKIP_ACCURATE_PASS
-
-LLVM_LIBC_FUNCTION(float, asinf, (float x)) {
-  using namespace inv_trigf_utils_internal;
-  using FPBits = typename fputil::FPBits<float>;
-
-  FPBits xbits(x);
-  uint32_t x_uint = xbits.uintval();
-  uint32_t x_abs = xbits.uintval() & 0x7fff'ffffU;
-  constexpr double SIGN[2] = {1.0, -1.0};
-  uint32_t x_sign = x_uint >> 31;
-
-  // |x| <= 0.5-ish
-  if (x_abs < 0x3f04'471dU) {
-    // |x| < 0x1.d12edp-12
-    if (LIBC_UNLIKELY(x_abs < 0x39e8'9768U)) {
-      // When |x| < 2^-12, the relative error of the approximation asin(x) ~ x
-      // is:
-      //   |asin(x) - x| / |asin(x)| < |x^3| / (6|x|)
-      //                             = x^2 / 6
-      //                             < 2^-25
-      //                             < epsilon(1)/2.
-      // So the correctly rounded values of asin(x) are:
-      //   = x + sign(x)*eps(x) if rounding mode = FE_TOWARDZERO,
-      //                        or (rounding mode = FE_UPWARD and x is
-      //                        negative),
-      //   = x otherwise.
-      // To simplify the rounding decision and make it more efficient, we use
-      //   fma(x, 2^-25, x) instead.
-      // An exhaustive test shows that this formula work correctly for all
-      // rounding modes up to |x| < 0x1.d12edp-12.
-      // Note: to use the formula x + 2^-25*x to decide the correct rounding, we
-      // do need fma(x, 2^-25, x) to prevent underflow caused by 2^-25*x when
-      // |x| < 2^-125. For targets without FMA instructions, we simply use
-      // double for intermediate results as it is more efficient than using an
-      // emulated version of FMA.
-#if defined(LIBC_TARGET_CPU_HAS_FMA_FLOAT)
-      return fputil::multiply_add(x, 0x1.0p-25f, x);
-#else
-      double xd = static_cast<double>(x);
-      return static_cast<float>(fputil::multiply_add(xd, 0x1.0p-25, xd));
-#endif // LIBC_TARGET_CPU_HAS_FMA_FLOAT
-    }
-
-#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
-    // Check for exceptional values
-    if (auto r = ASINF_EXCEPTS_LO.lookup_odd(x_abs, x_sign);
-        LIBC_UNLIKELY(r.has_value()))
-      return r.value();
-#endif // !LIBC_MATH_HAS_SKIP_ACCURATE_PASS
-
-    // For |x| <= 0.5, we approximate asinf(x) by:
-    //   asin(x) = x * P(x^2)
-    // Where P(X^2) = Q(X) is a degree-20 minimax even polynomial approximating
-    // asin(x)/x on [0, 0.5] generated by Sollya with:
-    // > Q = fpminimax(asin(x)/x, [|0, 2, 4, 6, 8, 10, 12, 14, 16, 18, 20|],
-    //                 [|1, D...|], [0, 0.5]);
-    // An exhaustive test shows that this approximation works well up to a
-    // little more than 0.5.
-    double xd = static_cast<double>(x);
-    double xsq = xd * xd;
-    double x3 = xd * xsq;
-    double r = asin_eval(xsq);
-    return static_cast<float>(fputil::multiply_add(x3, r, xd));
-  }
-
-  // |x| > 1, return NaNs.
-  if (LIBC_UNLIKELY(x_abs > 0x3f80'0000U)) {
-    if (xbits.is_signaling_nan()) {
-      fputil::raise_except_if_required(FE_INVALID);
-      return FPBits::quiet_nan().get_val();
-    }
-
-    if (x_abs <= 0x7f80'0000U) {
-      fputil::set_errno_if_required(EDOM);
-      fputil::raise_except_if_required(FE_INVALID);
-    }
-
-    return FPBits::quiet_nan().get_val();
-  }
-
-#ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
-  // Check for exceptional values
-  if (auto r = ASINF_EXCEPTS_HI.lookup_odd(x_abs, x_sign);
-      LIBC_UNLIKELY(r.has_value()))
-    return r.value();
-#endif // !LIBC_MATH_HAS_SKIP_ACCURATE_PASS
-
-  // When |x| > 0.5, we perform range reduction as follow:
-  //
-  // Assume further that 0.5 < x <= 1, and let:
-  //   y = asin(x)
-  // We will use the double angle formula:
-  //   cos(2y) = 1 - 2 sin^2(y)
-  // and the complement angle identity:
-  //   x = sin(y) = cos(pi/2 - y)
-  //              = 1 - 2 sin^2 (pi/4 - y/2)
-  // So:
-  //   sin(pi/4 - y/2) = sqrt( (1 - x)/2 )
-  // And hence:
-  //   pi/4 - y/2 = asin( sqrt( (1 - x)/2 ) )
-  // Equivalently:
-  //   asin(x) = y = pi/2 - 2 * asin( sqrt( (1 - x)/2 ) )
-  // Let u = (1 - x)/2, then:
-  //   asin(x) = pi/2 - 2 * asin( sqrt(u) )
-  // Moreover, since 0.5 < x <= 1:
-  //   0 <= u < 1/4, and 0 <= sqrt(u) < 0.5,
-  // And hence we can reuse the same polynomial approximation of asin(x) when
-  // |x| <= 0.5:
-  //   asin(x) ~ pi/2 - 2 * sqrt(u) * P(u),
-
-  xbits.set_sign(Sign::POS);
-  double sign = SIGN[x_sign];
-  double xd = static_cast<double>(xbits.get_val());
-  double u = fputil::multiply_add(-0.5, xd, 0.5);
-  double c1 = sign * (-2 * fputil::sqrt<double>(u));
-  double c2 = fputil::multiply_add(sign, M_MATH_PI_2, c1);
-  double c3 = c1 * u;
-
-  double r = asin_eval(u);
-  return static_cast<float>(fputil::multiply_add(c3, r, c2));
-}
+LLVM_LIBC_FUNCTION(float, asinf, (float x)) { return math::asinf(x); }
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/asinf16.cpp b/libc/src/math/generic/asinf16.cpp
index 518c384..af8dbfe 100644
--- a/libc/src/math/generic/asinf16.cpp
+++ b/libc/src/math/generic/asinf16.cpp
@@ -7,127 +7,10 @@
 //===----------------------------------------------------------------------===//
 
 #include "src/math/asinf16.h"
-#include "hdr/errno_macros.h"
-#include "hdr/fenv_macros.h"
-#include "src/__support/FPUtil/FEnvImpl.h"
-#include "src/__support/FPUtil/FPBits.h"
-#include "src/__support/FPUtil/PolyEval.h"
-#include "src/__support/FPUtil/cast.h"
-#include "src/__support/FPUtil/multiply_add.h"
-#include "src/__support/FPUtil/sqrt.h"
-#include "src/__support/macros/optimization.h"
+#include "src/__support/math/asinf16.h"
 
 namespace LIBC_NAMESPACE_DECL {
 
-// Generated by Sollya using the following command:
-// > round(pi/2, D, RN);
-static constexpr float PI_2 = 0x1.921fb54442d18p0f;
-
-LLVM_LIBC_FUNCTION(float16, asinf16, (float16 x)) {
-  using FPBits = fputil::FPBits<float16>;
-  FPBits xbits(x);
-
-  uint16_t x_u = xbits.uintval();
-  uint16_t x_abs = x_u & 0x7fff;
-  float xf = x;
-
-  // |x| > 0x1p0, |x| > 1, or x is NaN.
-  if (LIBC_UNLIKELY(x_abs > 0x3c00)) {
-    // asinf16(NaN) = NaN
-    if (xbits.is_nan()) {
-      if (xbits.is_signaling_nan()) {
-        fputil::raise_except_if_required(FE_INVALID);
-        return FPBits::quiet_nan().get_val();
-      }
-
-      return x;
-    }
-
-    // 1 < |x| <= +/-inf
-    fputil::raise_except_if_required(FE_INVALID);
-    fputil::set_errno_if_required(EDOM);
-
-    return FPBits::quiet_nan().get_val();
-  }
-
-  float xsq = xf * xf;
-
-  // |x| <= 0x1p-1, |x| <= 0.5
-  if (x_abs <= 0x3800) {
-    // asinf16(+/-0) = +/-0
-    if (LIBC_UNLIKELY(x_abs == 0))
-      return x;
-
-    // Exhaustive tests show that,
-    // for |x| <= 0x1.878p-9, when:
-    // x > 0, and rounding upward, or
-    // x < 0, and rounding downward, then,
-    // asin(x) = x * 2^-11 + x
-    // else, in other rounding modes,
-    // asin(x) = x
-    if (LIBC_UNLIKELY(x_abs <= 0x1a1e)) {
-      int rounding = fputil::quick_get_round();
-
-      if ((xbits.is_pos() && rounding == FE_UPWARD) ||
-          (xbits.is_neg() && rounding == FE_DOWNWARD))
-        return fputil::cast<float16>(fputil::multiply_add(xf, 0x1.0p-11f, xf));
-      return x;
-    }
-
-    // Degree-6 minimax odd polynomial of asin(x) generated by Sollya with:
-    // > P = fpminimax(asin(x)/x, [|0, 2, 4, 6, 8|], [|SG...|], [0, 0.5]);
-    float result =
-        fputil::polyeval(xsq, 0x1.000002p0f, 0x1.554c2ap-3f, 0x1.3541ccp-4f,
-                         0x1.43b2d6p-5f, 0x1.a0d73ep-5f);
-    return fputil::cast<float16>(xf * result);
-  }
-
-  // When |x| > 0.5, assume that 0.5 < |x| <= 1,
-  //
-  // Step-by-step range-reduction proof:
-  // 1:  Let y = asin(x), such that, x = sin(y)
-  // 2:  From complimentary angle identity:
-  //       x = sin(y) = cos(pi/2 - y)
-  // 3:  Let z = pi/2 - y, such that x = cos(z)
-  // 4:  From double angle formula; cos(2A) = 1 - sin^2(A):
-  //       z = 2A, z/2 = A
-  //       cos(z) = 1 - 2 * sin^2(z/2)
-  // 5:  Make sin(z/2) subject of the formula:
-  //       sin(z/2) = sqrt((1 - cos(z))/2)
-  // 6:  Recall [3]; x = cos(z). Therefore:
-  //       sin(z/2) = sqrt((1 - x)/2)
-  // 7:  Let u = (1 - x)/2
-  // 8:  Therefore:
-  //       asin(sqrt(u)) = z/2
-  //       2 * asin(sqrt(u)) = z
-  // 9:  Recall [3], z = pi/2 - y. Therefore:
-  //       y = pi/2 - z
-  //       y = pi/2 - 2 * asin(sqrt(u))
-  // 10: Recall [1], y = asin(x). Therefore:
-  //       asin(x) = pi/2 - 2 * asin(sqrt(u))
-  //
-  // WHY?
-  // 11: Recall [7], u = (1 - x)/2
-  // 12: Since 0.5 < x <= 1, therefore:
-  //       0 <= u <= 0.25 and 0 <= sqrt(u) <= 0.5
-  //
-  // Hence, we can reuse the same [0, 0.5] domain polynomial approximation for
-  // Step [10] as `sqrt(u)` is in range.
-
-  // 0x1p-1 < |x| <= 0x1p0, 0.5 < |x| <= 1.0
-  float xf_abs = (xf < 0 ? -xf : xf);
-  float sign = (xbits.uintval() >> 15 == 1 ? -1.0 : 1.0);
-  float u = fputil::multiply_add(-0.5f, xf_abs, 0.5f);
-  float u_sqrt = fputil::sqrt<float>(u);
-
-  // Degree-6 minimax odd polynomial of asin(x) generated by Sollya with:
-  // > P = fpminimax(asin(x)/x, [|0, 2, 4, 6, 8|], [|SG...|], [0, 0.5]);
-  float asin_sqrt_u =
-      u_sqrt * fputil::polyeval(u, 0x1.000002p0f, 0x1.554c2ap-3f,
-                                0x1.3541ccp-4f, 0x1.43b2d6p-5f, 0x1.a0d73ep-5f);
-
-  return fputil::cast<float16>(sign *
-                               fputil::multiply_add(-2.0f, asin_sqrt_u, PI_2));
-}
+LLVM_LIBC_FUNCTION(float16, asinf16, (float16 x)) { return math::asinf16(x); }
 
 } // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/math/generic/range_reduction_double_common.h b/libc/src/math/generic/range_reduction_double_common.h
index f3dcdb9..a93ee25 100644
--- a/libc/src/math/generic/range_reduction_double_common.h
+++ b/libc/src/math/generic/range_reduction_double_common.h
@@ -278,7 +278,7 @@ private:
 };
 
 #ifndef LIBC_MATH_HAS_SKIP_ACCURATE_PASS
-static Float128 range_reduction_small_f128(double x) {
+LIBC_INLINE static Float128 range_reduction_small_f128(double x) {
   constexpr Float128 PI_OVER_128_F128 = {
       Sign::POS, -133, 0xc90f'daa2'2168'c234'c4c6'628b'80dc'1cd1_u128};
   constexpr double ONE_TWENTY_EIGHT_OVER_PI_D = 0x1.45f306dc9c883p5;
diff --git a/libc/src/stdio/baremetal/CMakeLists.txt b/libc/src/stdio/baremetal/CMakeLists.txt
index e879230..548938f 100644
--- a/libc/src/stdio/baremetal/CMakeLists.txt
+++ b/libc/src/stdio/baremetal/CMakeLists.txt
@@ -72,6 +72,7 @@ add_entrypoint_object(
     ../scanf.h
   DEPENDS
     .scanf_internal
+    libc.include.inttypes
     libc.src.stdio.scanf_core.scanf_main
     libc.src.__support.arg_list
     libc.src.__support.OSUtil.osutil
diff --git a/libc/src/stdio/scanf_core/CMakeLists.txt b/libc/src/stdio/scanf_core/CMakeLists.txt
index dee125c..561180c 100644
--- a/libc/src/stdio/scanf_core/CMakeLists.txt
+++ b/libc/src/stdio/scanf_core/CMakeLists.txt
@@ -35,6 +35,7 @@ add_header_library(
     core_structs.h
   DEPENDS
     .scanf_config
+    libc.include.inttypes
     libc.src.__support.CPP.string_view
     libc.src.__support.CPP.bitset
     libc.src.__support.FPUtil.fp_bits
@@ -97,6 +98,7 @@ add_header_library(
   DEPENDS
     .reader
     .core_structs
+    libc.include.inttypes
     libc.src.__support.common
     libc.src.__support.ctype_utils
     libc.src.__support.CPP.bitset
diff --git a/libc/src/wchar/CMakeLists.txt b/libc/src/wchar/CMakeLists.txt
index 43f44a9..49f4a1b 100644
--- a/libc/src/wchar/CMakeLists.txt
+++ b/libc/src/wchar/CMakeLists.txt
@@ -137,6 +137,21 @@ add_entrypoint_object(
 )
 
 add_entrypoint_object(
+  mbsinit
+  SRCS
+    mbsinit.cpp
+  HDRS
+    mbsinit.h
+  DEPENDS
+    libc.hdr.types.wchar_t
+    libc.hdr.types.mbstate_t
+    libc.src.__support.common
+    libc.src.__support.macros.config
+    libc.src.__support.wchar.character_converter
+    libc.src.__support.wchar.mbstate
+)
+
+add_entrypoint_object(
   mbrtowc
   SRCS
     mbrtowc.cpp
diff --git a/libc/src/wchar/mbsinit.cpp b/libc/src/wchar/mbsinit.cpp
new file mode 100644
index 0000000..23ba542
--- /dev/null
+++ b/libc/src/wchar/mbsinit.cpp
@@ -0,0 +1,26 @@
+//===-- Implementation of mbsinit -----------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "src/wchar/mbsinit.h"
+
+#include "hdr/types/mbstate_t.h"
+#include "src/__support/common.h"
+#include "src/__support/macros/config.h"
+#include "src/__support/wchar/character_converter.h"
+#include "src/__support/wchar/mbstate.h"
+
+namespace LIBC_NAMESPACE_DECL {
+
+LLVM_LIBC_FUNCTION(int, mbsinit, (mbstate_t * ps)) {
+  if (ps == nullptr)
+    return true;
+  internal::CharacterConverter cr(reinterpret_cast<internal::mbstate *>(ps));
+  return cr.isValidState() && cr.isEmpty();
+}
+
+} // namespace LIBC_NAMESPACE_DECL
diff --git a/libc/src/wchar/mbsinit.h b/libc/src/wchar/mbsinit.h
new file mode 100644
index 0000000..fa6be0f
--- /dev/null
+++ b/libc/src/wchar/mbsinit.h
@@ -0,0 +1,22 @@
+//===-- Implementation header for mbsinit ---------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef LLVM_LIBC_SRC_WCHAR_MBSINIT_H
+#define LLVM_LIBC_SRC_WCHAR_MBSINIT_H
+
+#include "hdr/types/mbstate_t.h"
+#include "hdr/types/size_t.h"
+#include "src/__support/macros/config.h"
+
+namespace LIBC_NAMESPACE_DECL {
+
+int mbsinit(mbstate_t *ps);
+
+} // namespace LIBC_NAMESPACE_DECL
+
+#endif // LLVM_LIBC_SRC_WCHAR_MBSINIT_H
diff --git a/libc/test/shared/CMakeLists.txt b/libc/test/shared/CMakeLists.txt
index 89b607d..e93fbb9 100644
--- a/libc/test/shared/CMakeLists.txt
+++ b/libc/test/shared/CMakeLists.txt
@@ -12,6 +12,10 @@ add_fp_unittest(
     libc.src.__support.math.acosf16
     libc.src.__support.math.acoshf
     libc.src.__support.math.acoshf16
+    libc.src.__support.math.acospif16
+    libc.src.__support.math.asin
+    libc.src.__support.math.asinf
+    libc.src.__support.math.asinf16
     libc.src.__support.math.erff
     libc.src.__support.math.exp
     libc.src.__support.math.exp10
diff --git a/libc/test/shared/shared_math_test.cpp b/libc/test/shared/shared_math_test.cpp
index 8d3cebd..51d38ec 100644
--- a/libc/test/shared/shared_math_test.cpp
+++ b/libc/test/shared/shared_math_test.cpp
@@ -8,64 +8,72 @@
 
 #include "shared/math.h"
 #include "test/UnitTest/FPMatcher.h"
+#include "test/UnitTest/Test.h"
 
 #ifdef LIBC_TYPES_HAS_FLOAT16
 
 TEST(LlvmLibcSharedMathTest, AllFloat16) {
   int exponent;
 
-  EXPECT_FP_EQ(0x0p+0f, LIBC_NAMESPACE::shared::acoshf16(1.0f));
+  EXPECT_FP_EQ(0x0p+0f16, LIBC_NAMESPACE::shared::acoshf16(1.0f16));
+  EXPECT_FP_EQ(0x0p+0f16, LIBC_NAMESPACE::shared::acospif16(1.0f16));
+  EXPECT_FP_EQ(0x0p+0f16, LIBC_NAMESPACE::shared::asinf16(0.0f16));
 
   EXPECT_FP_EQ(0x1p+0f16, LIBC_NAMESPACE::shared::exp10f16(0.0f16));
 
   EXPECT_FP_EQ(0x1p+0f16, LIBC_NAMESPACE::shared::expf16(0.0f16));
 
-  ASSERT_FP_EQ(float16(8 << 5), LIBC_NAMESPACE::shared::ldexpf16(float(8), 5));
+  ASSERT_FP_EQ(float16(8 << 5), LIBC_NAMESPACE::shared::ldexpf16(8.0f16, 5));
   ASSERT_FP_EQ(float16(-1 * (8 << 5)),
-               LIBC_NAMESPACE::shared::ldexpf16(float(-8), 5));
+               LIBC_NAMESPACE::shared::ldexpf16(-8.0f16, 5));
 
-  EXPECT_FP_EQ_ALL_ROUNDING(0.75f16,
-                            LIBC_NAMESPACE::shared::frexpf16(24.0f, &exponent));
+  EXPECT_FP_EQ_ALL_ROUNDING(
+      0.75f16, LIBC_NAMESPACE::shared::frexpf16(24.0f16, &exponent));
   EXPECT_EQ(exponent, 5);
 
   EXPECT_FP_EQ(0x1.921fb6p+0f16, LIBC_NAMESPACE::shared::acosf16(0.0f16));
 }
 
-#endif
+#endif // LIBC_TYPES_HAS_FLOAT16
 
 TEST(LlvmLibcSharedMathTest, AllFloat) {
   int exponent;
 
   EXPECT_FP_EQ(0x1.921fb6p+0, LIBC_NAMESPACE::shared::acosf(0.0f));
+  EXPECT_FP_EQ(0x0p+0f, LIBC_NAMESPACE::shared::acoshf(1.0f));
+  EXPECT_FP_EQ(0x0p+0f, LIBC_NAMESPACE::shared::asinf(0.0f));
+  EXPECT_FP_EQ(0x0p+0f, LIBC_NAMESPACE::shared::erff(0.0f));
   EXPECT_FP_EQ(0x1p+0f, LIBC_NAMESPACE::shared::exp10f(0.0f));
   EXPECT_FP_EQ(0x1p+0f, LIBC_NAMESPACE::shared::expf(0.0f));
-  EXPECT_FP_EQ(0x0p+0f, LIBC_NAMESPACE::shared::erff(0.0f));
-  EXPECT_FP_EQ(0x0p+0f, LIBC_NAMESPACE::shared::acoshf(1.0f));
 
   EXPECT_FP_EQ_ALL_ROUNDING(0.75f,
                             LIBC_NAMESPACE::shared::frexpf(24.0f, &exponent));
   EXPECT_EQ(exponent, 5);
 
-  ASSERT_FP_EQ(float(8 << 5), LIBC_NAMESPACE::shared::ldexpf(float(8), 5));
-  ASSERT_FP_EQ(float(-1 * (8 << 5)),
-               LIBC_NAMESPACE::shared::ldexpf(float(-8), 5));
+  ASSERT_FP_EQ(float(8 << 5), LIBC_NAMESPACE::shared::ldexpf(8.0f, 5));
+  ASSERT_FP_EQ(float(-1 * (8 << 5)), LIBC_NAMESPACE::shared::ldexpf(-8.0f, 5));
 }
 
 TEST(LlvmLibcSharedMathTest, AllDouble) {
   EXPECT_FP_EQ(0x1.921fb54442d18p+0, LIBC_NAMESPACE::shared::acos(0.0));
+  EXPECT_FP_EQ(0x0p+0, LIBC_NAMESPACE::shared::asin(0.0));
   EXPECT_FP_EQ(0x1p+0, LIBC_NAMESPACE::shared::exp(0.0));
   EXPECT_FP_EQ(0x1p+0, LIBC_NAMESPACE::shared::exp10(0.0));
 }
 
+#ifdef LIBC_TYPES_HAS_FLOAT128
+
 TEST(LlvmLibcSharedMathTest, AllFloat128) {
   int exponent;
 
-  EXPECT_FP_EQ_ALL_ROUNDING(
-      float128(0.75), LIBC_NAMESPACE::shared::frexpf128(24.0f, &exponent));
+  EXPECT_FP_EQ_ALL_ROUNDING(float128(0.75), LIBC_NAMESPACE::shared::frexpf128(
+                                                float128(24), &exponent));
   EXPECT_EQ(exponent, 5);
 
   ASSERT_FP_EQ(float128(8 << 5),
-               LIBC_NAMESPACE::shared::ldexpf128(float(8), 5));
+               LIBC_NAMESPACE::shared::ldexpf128(float128(8), 5));
   ASSERT_FP_EQ(float128(-1 * (8 << 5)),
-               LIBC_NAMESPACE::shared::ldexpf128(float(-8), 5));
+               LIBC_NAMESPACE::shared::ldexpf128(float128(-8), 5));
 }
+
+#endif // LIBC_TYPES_HAS_FLOAT128
diff --git a/libc/test/src/wchar/CMakeLists.txt b/libc/test/src/wchar/CMakeLists.txt
index f420ecc..fad89dc 100644
--- a/libc/test/src/wchar/CMakeLists.txt
+++ b/libc/test/src/wchar/CMakeLists.txt
@@ -99,7 +99,22 @@ add_libc_test(
     libc.src.string.memset
     libc.src.wchar.mbrlen
     libc.hdr.types.mbstate_t
+    libc.hdr.types.wchar_t
     libc.test.UnitTest.ErrnoCheckingTest
+)   
+
+add_libc_test(
+  mbsinit_test
+  SUITE
+    libc_wchar_unittests
+  SRCS
+    mbsinit_test.cpp
+  DEPENDS
+    libc.src.string.memset
+    libc.src.wchar.mbsinit
+    libc.src.wchar.mbrtowc
+    libc.hdr.types.mbstate_t
+    libc.hdr.types.wchar_t
 )
 
 add_libc_test(
diff --git a/libc/test/src/wchar/mbsinit_test.cpp b/libc/test/src/wchar/mbsinit_test.cpp
new file mode 100644
index 0000000..ecb48aa
--- /dev/null
+++ b/libc/test/src/wchar/mbsinit_test.cpp
@@ -0,0 +1,33 @@
+//===-- Unittests for mbsinit ---------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#include "hdr/types/wchar_t.h"
+#include "src/string/memset.h"
+#include "src/wchar/mbrtowc.h"
+#include "src/wchar/mbsinit.h"
+#include "test/UnitTest/Test.h"
+
+TEST(LlvmLibcMBSInitTest, EmptyState) {
+  mbstate_t ps;
+  LIBC_NAMESPACE::memset(&ps, 0, sizeof(mbstate_t));
+  ASSERT_NE(LIBC_NAMESPACE::mbsinit(&ps), 0);
+  ASSERT_NE(LIBC_NAMESPACE::mbsinit(nullptr), 0);
+}
+
+TEST(LlvmLibcMBSInitTest, ConversionTest) {
+  const char *src = "\xf0\x9f\xa4\xa3"; // 4 byte emoji
+  wchar_t dest[2];
+  mbstate_t ps;
+  LIBC_NAMESPACE::memset(&ps, 0, sizeof(mbstate_t));
+
+  ASSERT_NE(LIBC_NAMESPACE::mbsinit(&ps), 0);
+  LIBC_NAMESPACE::mbrtowc(dest, src, 2, &ps); // partial conversion
+  ASSERT_EQ(LIBC_NAMESPACE::mbsinit(&ps), 0);
+  LIBC_NAMESPACE::mbrtowc(dest, src + 2, 2, &ps); // complete conversion
+  ASSERT_NE(LIBC_NAMESPACE::mbsinit(&ps), 0);     // state should be reset now
+}
diff --git a/libcxx/include/__algorithm/copy.h b/libcxx/include/__algorithm/copy.h
index ea98031..6387728 100644
--- a/libcxx/include/__algorithm/copy.h
+++ b/libcxx/include/__algorithm/copy.h
@@ -182,7 +182,7 @@ struct __copy_impl {
     }
   };
 
-  template <class _InIter, class _OutIter, __enable_if_t<__is_segmented_iterator<_InIter>::value, int> = 0>
+  template <class _InIter, class _OutIter, __enable_if_t<__is_segmented_iterator_v<_InIter>, int> = 0>
   _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 pair<_InIter, _OutIter>
   operator()(_InIter __first, _InIter __last, _OutIter __result) const {
     std::__for_each_segment(__first, __last, _CopySegment<_InIter, _OutIter>(__result));
@@ -192,7 +192,7 @@ struct __copy_impl {
   template <class _InIter,
             class _OutIter,
             __enable_if_t<__has_random_access_iterator_category<_InIter>::value &&
-                              !__is_segmented_iterator<_InIter>::value && __is_segmented_iterator<_OutIter>::value,
+                              !__is_segmented_iterator_v<_InIter> && __is_segmented_iterator_v<_OutIter>,
                           int> = 0>
   _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 pair<_InIter, _OutIter>
   operator()(_InIter __first, _InIter __last, _OutIter __result) const {
diff --git a/libcxx/include/__algorithm/copy_backward.h b/libcxx/include/__algorithm/copy_backward.h
index 9f89064..807c64b 100644
--- a/libcxx/include/__algorithm/copy_backward.h
+++ b/libcxx/include/__algorithm/copy_backward.h
@@ -170,7 +170,7 @@ struct __copy_backward_impl {
     return std::make_pair(std::move(__original_last_iter), std::move(__result));
   }
 
-  template <class _InIter, class _OutIter, __enable_if_t<__is_segmented_iterator<_InIter>::value, int> = 0>
+  template <class _InIter, class _OutIter, __enable_if_t<__is_segmented_iterator_v<_InIter>, int> = 0>
   _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 pair<_InIter, _OutIter>
   operator()(_InIter __first, _InIter __last, _OutIter __result) const {
     using _Traits = __segmented_iterator_traits<_InIter>;
@@ -200,7 +200,7 @@ struct __copy_backward_impl {
   template <class _InIter,
             class _OutIter,
             __enable_if_t<__has_random_access_iterator_category<_InIter>::value &&
-                              !__is_segmented_iterator<_InIter>::value && __is_segmented_iterator<_OutIter>::value,
+                              !__is_segmented_iterator_v<_InIter> && __is_segmented_iterator_v<_OutIter>,
                           int> = 0>
   _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 pair<_InIter, _OutIter>
   operator()(_InIter __first, _InIter __last, _OutIter __result) const {
diff --git a/libcxx/include/__algorithm/find.h b/libcxx/include/__algorithm/find.h
index a7d9374..8c8cb58 100644
--- a/libcxx/include/__algorithm/find.h
+++ b/libcxx/include/__algorithm/find.h
@@ -149,7 +149,7 @@ struct __find_segment;
 template <class _SegmentedIterator,
           class _Tp,
           class _Proj,
-          __enable_if_t<__is_segmented_iterator<_SegmentedIterator>::value, int> = 0>
+          __enable_if_t<__is_segmented_iterator_v<_SegmentedIterator>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 _SegmentedIterator
 __find(_SegmentedIterator __first, _SegmentedIterator __last, const _Tp& __value, _Proj& __proj) {
   return std::__find_segment_if(std::move(__first), std::move(__last), __find_segment<_Tp>(__value), __proj);
diff --git a/libcxx/include/__algorithm/for_each.h b/libcxx/include/__algorithm/for_each.h
index 4167eec..e31fcae 100644
--- a/libcxx/include/__algorithm/for_each.h
+++ b/libcxx/include/__algorithm/for_each.h
@@ -39,7 +39,7 @@ __for_each(_InputIterator __first, _Sent __last, _Func& __f, _Proj& __proj) {
 template <class _SegmentedIterator,
           class _Func,
           class _Proj,
-          __enable_if_t<__is_segmented_iterator<_SegmentedIterator>::value, int> = 0>
+          __enable_if_t<__is_segmented_iterator_v<_SegmentedIterator>, int> = 0>
 _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 _SegmentedIterator
 __for_each(_SegmentedIterator __first, _SegmentedIterator __last, _Func& __func, _Proj& __proj) {
   using __local_iterator_t = typename __segmented_iterator_traits<_SegmentedIterator>::__local_iterator;
diff --git a/libcxx/include/__algorithm/for_each_n.h b/libcxx/include/__algorithm/for_each_n.h
index 9a6c6bb..04650e1 100644
--- a/libcxx/include/__algorithm/for_each_n.h
+++ b/libcxx/include/__algorithm/for_each_n.h
@@ -37,8 +37,8 @@ template <class _InputIterator,
           class _Func,
           class _Proj,
           __enable_if_t<!__has_random_access_iterator_category<_InputIterator>::value &&
-                            _Or< _Not<__is_segmented_iterator<_InputIterator> >,
-                                 _Not<__has_random_access_local_iterator<_InputIterator> > >::value,
+                            _Or<integral_constant<bool, !__is_segmented_iterator_v<_InputIterator> >,
+                                _Not<__has_random_access_local_iterator<_InputIterator> > >::value,
                         int> = 0>
 _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX20 _InputIterator
 __for_each_n(_InputIterator __first, _Size __orig_n, _Func& __f, _Proj& __proj) {
@@ -71,7 +71,7 @@ template <class _SegmentedIterator,
           class _Func,
           class _Proj,
           __enable_if_t<!__has_random_access_iterator_category<_SegmentedIterator>::value &&
-                            __is_segmented_iterator<_SegmentedIterator>::value &&
+                            __is_segmented_iterator_v<_SegmentedIterator> &&
                             __has_random_access_iterator_category<
                                 typename __segmented_iterator_traits<_SegmentedIterator>::__local_iterator>::value,
                         int> = 0>
diff --git a/libcxx/include/__algorithm/for_each_n_segment.h b/libcxx/include/__algorithm/for_each_n_segment.h
index 1b522fb..a433df5 100644
--- a/libcxx/include/__algorithm/for_each_n_segment.h
+++ b/libcxx/include/__algorithm/for_each_n_segment.h
@@ -27,7 +27,7 @@ _LIBCPP_BEGIN_NAMESPACE_STD
 template <class _SegmentedIterator, class _Size, class _Functor>
 _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 _SegmentedIterator
 __for_each_n_segment(_SegmentedIterator __first, _Size __orig_n, _Functor __func) {
-  static_assert(__is_segmented_iterator<_SegmentedIterator>::value &&
+  static_assert(__is_segmented_iterator_v<_SegmentedIterator> &&
                     __has_random_access_iterator_category<
                         typename __segmented_iterator_traits<_SegmentedIterator>::__local_iterator>::value,
                 "__for_each_n_segment only works with segmented iterators with random-access local iterators");
diff --git a/libcxx/include/__algorithm/move.h b/libcxx/include/__algorithm/move.h
index a3320e9..73b780d 100644
--- a/libcxx/include/__algorithm/move.h
+++ b/libcxx/include/__algorithm/move.h
@@ -65,7 +65,7 @@ struct __move_impl {
     }
   };
 
-  template <class _InIter, class _OutIter, __enable_if_t<__is_segmented_iterator<_InIter>::value, int> = 0>
+  template <class _InIter, class _OutIter, __enable_if_t<__is_segmented_iterator_v<_InIter>, int> = 0>
   _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 pair<_InIter, _OutIter>
   operator()(_InIter __first, _InIter __last, _OutIter __result) const {
     std::__for_each_segment(__first, __last, _MoveSegment<_InIter, _OutIter>(__result));
@@ -75,7 +75,7 @@ struct __move_impl {
   template <class _InIter,
             class _OutIter,
             __enable_if_t<__has_random_access_iterator_category<_InIter>::value &&
-                              !__is_segmented_iterator<_InIter>::value && __is_segmented_iterator<_OutIter>::value,
+                              !__is_segmented_iterator_v<_InIter> && __is_segmented_iterator_v<_OutIter>,
                           int> = 0>
   _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 pair<_InIter, _OutIter>
   operator()(_InIter __first, _InIter __last, _OutIter __result) const {
diff --git a/libcxx/include/__algorithm/move_backward.h b/libcxx/include/__algorithm/move_backward.h
index 14482fe..e3e61c7b 100644
--- a/libcxx/include/__algorithm/move_backward.h
+++ b/libcxx/include/__algorithm/move_backward.h
@@ -51,7 +51,7 @@ struct __move_backward_impl {
     return std::make_pair(std::move(__original_last_iter), std::move(__result));
   }
 
-  template <class _InIter, class _OutIter, __enable_if_t<__is_segmented_iterator<_InIter>::value, int> = 0>
+  template <class _InIter, class _OutIter, __enable_if_t<__is_segmented_iterator_v<_InIter>, int> = 0>
   _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 pair<_InIter, _OutIter>
   operator()(_InIter __first, _InIter __last, _OutIter __result) const {
     using _Traits = __segmented_iterator_traits<_InIter>;
@@ -81,7 +81,7 @@ struct __move_backward_impl {
   template <class _InIter,
             class _OutIter,
             __enable_if_t<__has_random_access_iterator_category<_InIter>::value &&
-                              !__is_segmented_iterator<_InIter>::value && __is_segmented_iterator<_OutIter>::value,
+                              !__is_segmented_iterator_v<_InIter> && __is_segmented_iterator_v<_OutIter>,
                           int> = 0>
   _LIBCPP_HIDE_FROM_ABI _LIBCPP_CONSTEXPR_SINCE_CXX14 pair<_InIter, _OutIter>
   operator()(_InIter __first, _InIter __last, _OutIter __result) const {
diff --git a/libcxx/include/__iterator/segmented_iterator.h b/libcxx/include/__iterator/segmented_iterator.h
index af27a7b..5df9737 100644
--- a/libcxx/include/__iterator/segmented_iterator.h
+++ b/libcxx/include/__iterator/segmented_iterator.h
@@ -67,13 +67,13 @@ struct __segmented_iterator_traits;
 */
 
 template <class _Tp, size_t = 0>
-struct __has_specialization : false_type {};
+inline const bool __has_specialization_v = false;
 
 template <class _Tp>
-struct __has_specialization<_Tp, sizeof(_Tp) * 0> : true_type {};
+inline const bool __has_specialization_v<_Tp, sizeof(_Tp) * 0> = true;
 
 template <class _Iterator>
-using __is_segmented_iterator _LIBCPP_NODEBUG = __has_specialization<__segmented_iterator_traits<_Iterator> >;
+inline const bool __is_segmented_iterator_v = __has_specialization_v<__segmented_iterator_traits<_Iterator> >;
 
 template <class _SegmentedIterator>
 struct __has_random_access_local_iterator
diff --git a/libcxx/include/deque b/libcxx/include/deque
index e33e7d3..395a107 100644
--- a/libcxx/include/deque
+++ b/libcxx/include/deque
@@ -461,9 +461,8 @@ private:
       __deque_iterator<_ValueType, _Pointer, _Reference, _MapPointer, _DiffType, _BlockSize>;
 
 public:
-  using __is_segmented_iterator _LIBCPP_NODEBUG = true_type;
-  using __segment_iterator _LIBCPP_NODEBUG      = _MapPointer;
-  using __local_iterator _LIBCPP_NODEBUG        = _Pointer;
+  using __segment_iterator _LIBCPP_NODEBUG = _MapPointer;
+  using __local_iterator _LIBCPP_NODEBUG   = _Pointer;
 
   static _LIBCPP_HIDE_FROM_ABI __segment_iterator __segment(_Iterator __iter) { return __iter.__m_iter_; }
   static _LIBCPP_HIDE_FROM_ABI __local_iterator __local(_Iterator __iter) { return __iter.__ptr_; }
diff --git a/libcxx/test/libcxx/containers/sequences/deque/segmented_iterator.compile.pass.cpp b/libcxx/test/libcxx/containers/sequences/deque/segmented_iterator.compile.pass.cpp
index 7ae67d8..b050da4 100644
--- a/libcxx/test/libcxx/containers/sequences/deque/segmented_iterator.compile.pass.cpp
+++ b/libcxx/test/libcxx/containers/sequences/deque/segmented_iterator.compile.pass.cpp
@@ -6,7 +6,9 @@
 //
 //===----------------------------------------------------------------------===//
 
+// XFAIL: FROZEN-CXX03-HEADERS-FIXME
+
 #include <deque>
 
 using DequeIterator = typename std::deque<int>::iterator;
-static_assert(std::__is_segmented_iterator<DequeIterator>::value, "");
+static_assert(std::__is_segmented_iterator_v<DequeIterator>, "");
diff --git a/libcxx/test/libcxx/ranges/range.adaptors/range.join/segmented_iterator.compile.pass.cpp b/libcxx/test/libcxx/ranges/range.adaptors/range.join/segmented_iterator.compile.pass.cpp
index 6cd17c2..4e4074f 100644
--- a/libcxx/test/libcxx/ranges/range.adaptors/range.join/segmented_iterator.compile.pass.cpp
+++ b/libcxx/test/libcxx/ranges/range.adaptors/range.join/segmented_iterator.compile.pass.cpp
@@ -14,4 +14,4 @@
 
 using JoinView = decltype(std::views::join(std::declval<std::vector<std::vector<int>>&>()));
 using JoinIter = std::ranges::iterator_t<JoinView>;
-static_assert(std::__is_segmented_iterator<JoinIter>::value);
+static_assert(std::__is_segmented_iterator_v<JoinIter>);
diff --git a/libcxx/test/std/library/description/conventions/customization.point.object/cpo.compile.pass.cpp b/libcxx/test/std/library/description/conventions/customization.point.object/cpo.compile.pass.cpp
index 4e24dbe..678483b 100644
--- a/libcxx/test/std/library/description/conventions/customization.point.object/cpo.compile.pass.cpp
+++ b/libcxx/test/std/library/description/conventions/customization.point.object/cpo.compile.pass.cpp
@@ -15,6 +15,7 @@
 #include <concepts>
 #include <iterator>
 #include <ranges>
+#include <sstream>
 #include <type_traits>
 #include <utility>
 
@@ -45,7 +46,10 @@ constexpr bool test(CPO& o, Args&&...) {
 
 int a[10];
 int arrays[10][10];
-//std::pair<int, int> pairs[10];
+std::pair<int, int> pairs[10];
+#ifndef TEST_HAS_NO_LOCALIZATION
+std::istringstream stream;
+#endif
 
 // [concept.swappable]
 static_assert(test(std::ranges::swap, a, a));
@@ -81,25 +85,50 @@ static_assert(test(std::ranges::ssize, a));
 // views::empty<T> is not a CPO
 static_assert(test(std::views::iota, 1));
 static_assert(test(std::views::iota, 1, 10));
-//static_assert(test(std::views::istream<int>, 1);
+#ifndef TEST_HAS_NO_LOCALIZATION
+static_assert(test(std::views::istream<int>, stream));
+#endif
 static_assert(test(std::views::single, 4));
 
+#if TEST_STD_VER >= 23
+static_assert(test(std::views::repeat, 1));
+#endif
+
 // [range.adaptors]
 static_assert(test(std::views::all, a));
 static_assert(test(std::views::common, a));
 static_assert(test(std::views::counted, a, 10));
 static_assert(test(std::views::drop, a, 10));
-//static_assert(test(std::views::drop_while, a, [](int x){ return x < 10; }));
-//static_assert(test(std::views::elements<0>, pairs));
+static_assert(test(std::views::drop_while, a, [](int x) { return x < 10; }));
+static_assert(test(std::views::elements<0>, pairs));
 static_assert(test(std::views::filter, a, [](int x) { return x < 10; }));
 static_assert(test(std::views::join, arrays));
-//static_assert(test(std::views::split, a, 4));
+static_assert(test(std::views::keys, pairs));
 static_assert(test(std::views::lazy_split, a, 4));
 static_assert(test(std::views::reverse, a));
+static_assert(test(std::views::split, a, 4));
 static_assert(test(std::views::take, a, 10));
-//static_assert(test(std::views::take_while, a, [](int x){ return x < 10; }));
+static_assert(test(std::views::take_while, a, [](int x) { return x < 10; }));
 static_assert(test(std::views::transform, a, [](int x) { return x + 1; }));
+static_assert(test(std::views::values, pairs));
 
 #if TEST_STD_VER >= 23
+// static_assert(test(std::views::adjacent_transform<2>, [](int x, int y) { return x + y; }, a));
+// static_assert(test(std::views::adjacent<2>, a));
+// static_assert(test(std::views::as_const, a));
+static_assert(test(std::views::as_rvalue, a));
+// static_assert(test(std::views::cartesian_product, a, a, a));
+static_assert(test(std::views::chunk_by, a, [](int x, int y) { return x < y; }));
+// static_assert(test(std::views::chunk, a, 1));
+// static_assert(test(std::views::enumerate, a));
+static_assert(test(std::views::join_with, 1));
+// static_assert(test(std::views::stride, a, 1));
 static_assert(test(std::views::zip_transform, [](int x, int y) { return x + y; }, a, a));
+static_assert(test(std::views::zip, a, a));
+#endif
+
+#if TEST_STD_VER >= 26
+// static_assert(test(std::views::cache_latest, a));
+// static_assert(test(std::views::concat, a, a));
+// static_assert(test(std::views::to_input, a));
 #endif
diff --git a/lld/COFF/InputFiles.cpp b/lld/COFF/InputFiles.cpp
index 2a6b63c..c08099b 100644
--- a/lld/COFF/InputFiles.cpp
+++ b/lld/COFF/InputFiles.cpp
@@ -403,6 +403,11 @@ SectionChunk *ObjFile::readSection(uint32_t sectionNumber,
     return nullptr;
   }
 
+  // Those sections are generated by -fembed-bitcode and do not need to be kept
+  // in executable files.
+  if (name == ".llvmbc" || name == ".llvmcmd")
+    return nullptr;
+
   // Object files may have DWARF debug info or MS CodeView debug info
   // (or both).
   //
diff --git a/lld/ELF/Target.h b/lld/ELF/Target.h
index 93f1592..f4a6d83 100644
--- a/lld/ELF/Target.h
+++ b/lld/ELF/Target.h
@@ -338,21 +338,23 @@ inline uint64_t overwriteULEB128(uint8_t *bufLoc, uint64_t val) {
 #pragma clang diagnostic ignored "-Wgnu-zero-variadic-macro-arguments"
 #endif
 #define invokeELFT(f, ...)                                                     \
-  switch (ctx.arg.ekind) {                                                     \
-  case lld::elf::ELF32LEKind:                                                  \
-    f<llvm::object::ELF32LE>(__VA_ARGS__);                                     \
-    break;                                                                     \
-  case lld::elf::ELF32BEKind:                                                  \
-    f<llvm::object::ELF32BE>(__VA_ARGS__);                                     \
-    break;                                                                     \
-  case lld::elf::ELF64LEKind:                                                  \
-    f<llvm::object::ELF64LE>(__VA_ARGS__);                                     \
-    break;                                                                     \
-  case lld::elf::ELF64BEKind:                                                  \
-    f<llvm::object::ELF64BE>(__VA_ARGS__);                                     \
-    break;                                                                     \
-  default:                                                                     \
-    llvm_unreachable("unknown ctx.arg.ekind");                                 \
-  }
+  do {                                                                         \
+    switch (ctx.arg.ekind) {                                                   \
+    case lld::elf::ELF32LEKind:                                                \
+      f<llvm::object::ELF32LE>(__VA_ARGS__);                                   \
+      break;                                                                   \
+    case lld::elf::ELF32BEKind:                                                \
+      f<llvm::object::ELF32BE>(__VA_ARGS__);                                   \
+      break;                                                                   \
+    case lld::elf::ELF64LEKind:                                                \
+      f<llvm::object::ELF64LE>(__VA_ARGS__);                                   \
+      break;                                                                   \
+    case lld::elf::ELF64BEKind:                                                \
+      f<llvm::object::ELF64BE>(__VA_ARGS__);                                   \
+      break;                                                                   \
+    default:                                                                   \
+      llvm_unreachable("unknown ctx.arg.ekind");                               \
+    }                                                                          \
+  } while (0)
 
 #endif
diff --git a/lld/test/COFF/embed-bitcode.test b/lld/test/COFF/embed-bitcode.test
new file mode 100644
index 0000000..10f88c5
--- /dev/null
+++ b/lld/test/COFF/embed-bitcode.test
@@ -0,0 +1,30 @@
+# RUN: yaml2obj %s -o %t.obj
+# RUN: lld-link /entry:main /subsystem:console /out:%t.exe %t.obj
+# RUN: llvm-readobj -S %t.exe | FileCheck %s
+
+# CHECK-NOT: Name: .llvmbc
+# CHECK-NOT: Name: .llvmcmd
+
+--- !COFF
+header:
+  Machine:         IMAGE_FILE_MACHINE_AMD64
+
+sections:
+  - Name:            .text
+    Characteristics: [ IMAGE_SCN_CNT_CODE, IMAGE_SCN_MEM_EXECUTE, IMAGE_SCN_MEM_READ ]
+    SectionData:     "C3"
+  - Name:           .llvmbc
+    Characteristics: [ IMAGE_SCN_MEM_DISCARDABLE ]
+    SectionData:     "4243C0DE"
+  - Name:           .llvmcmd
+    Characteristics: [ IMAGE_SCN_MEM_DISCARDABLE ]
+    SectionData:     "2D63633100"
+
+symbols:
+  - Name:            main
+    Value:           0
+    SectionNumber:   1
+    SimpleType:      IMAGE_SYM_TYPE_NULL
+    ComplexType:     IMAGE_SYM_DTYPE_FUNCTION
+    StorageClass:    IMAGE_SYM_CLASS_EXTERNAL
+...
diff --git a/lld/test/COFF/nodefaultlib.test b/lld/test/COFF/nodefaultlib.test
index ceeb1f3..fbf6b43 100644
--- a/lld/test/COFF/nodefaultlib.test
+++ b/lld/test/COFF/nodefaultlib.test
@@ -1,5 +1,6 @@
-# RUN: cp %p/Inputs/hello64.obj %T
-# RUN: cp %p/Inputs/std64.lib %T
+# RUN: mkdir -p %t.dir
+# RUN: cp %p/Inputs/hello64.obj %t.dir
+# RUN: cp %p/Inputs/std64.lib %t.dir
 
 # RUN: not lld-link /out:%t.exe /entry:main /subsystem:console \
 # RUN:   hello64.obj /defaultlib:std64.lib >& %t.log
@@ -9,12 +10,12 @@
 # RUN:   hello64 /defaultlib:std64.lib >& %t.log
 # RUN: FileCheck -DMSG=%errc_ENOENT -check-prefix=CHECK2 %s < %t.log
 
-# RUN: lld-link /libpath:%T /out:%t.exe /entry:main \
+# RUN: lld-link /libpath:%t.dir /out:%t.exe /entry:main \
 # RUN:   /subsystem:console hello64.obj /defaultlib:std64.lib \
 # RUN:   /nodefaultlib:std64.lib >& %t.log || true
 # RUN: FileCheck -check-prefix=CHECK3 %s < %t.log
 
-# RUN: lld-link /libpath:%T /out:%t.exe /entry:main \
+# RUN: lld-link /libpath:%t.dir /out:%t.exe /entry:main \
 # RUN:   /subsystem:console hello64.obj /defaultlib:std64 \
 # RUN:   /nodefaultlib:std64.lib >& %t.log || true
 # RUN: FileCheck -check-prefix=CHECK3 %s < %t.log
@@ -24,10 +25,10 @@ CHECK2: error: could not open 'hello64': [[MSG]]
 CHECK3: error: undefined symbol: MessageBoxA
 CHECK3-NEXT: >>> referenced by {{.*}}hello64.obj:(main)
 
-# RUN: lld-link /libpath:%T /out:%t.exe /entry:main \
+# RUN: lld-link /libpath:%t.dir /out:%t.exe /entry:main \
 # RUN:   /subsystem:console hello64.obj /defaultlib:std64.lib
 
-# RUN: env LIB=%T lld-link /out:%t.exe /entry:main \
+# RUN: env LIB=%t.dir lld-link /out:%t.exe /entry:main \
 # RUN:   /subsystem:console hello64.obj /defaultlib:std64.lib
 
 MSVC stamps uppercase references in OBJ directives, thus ensure that passing lowercase 'libcmt' and 'oldnames' to /nodefaultlib works.
@@ -37,11 +38,11 @@ MSVC stamps uppercase references in OBJ directives, thus ensure that passing low
 UPPERCASE-NOT: OLDNAMES
 UPPERCASE-NOT: LIBCMT
 
-# RUN: yaml2obj -o %T/defaultlib.obj %p/Inputs/defaultlib.yaml
+# RUN: yaml2obj -o %t.dir/defaultlib.obj %p/Inputs/defaultlib.yaml
 # RUN: mkdir -p %t.dir/sysroot/VC/Tools/MSVC/1.1.1.1/lib/x64
 # RUN: cp %p/Inputs/ret42.lib %t.dir/sysroot/VC/Tools/MSVC/1.1.1.1/lib/x64/default.lib
 
-# RUN: lld-link /winsysroot:%t.dir/sysroot /out:%t.exe /entry:main /subsystem:console %T/defaultlib.obj
-# RUN: not lld-link /winsysroot:%t.dir/sysroot /out:%t.exe /entry:main /subsystem:console /nodefaultlib:default.lib %T/defaultlib.obj 2>&1 | FileCheck -check-prefix=CHECK4 %s
+# RUN: lld-link /winsysroot:%t.dir/sysroot /out:%t.exe /entry:main /subsystem:console %t.dir/defaultlib.obj
+# RUN: not lld-link /winsysroot:%t.dir/sysroot /out:%t.exe /entry:main /subsystem:console /nodefaultlib:default.lib %t.dir/defaultlib.obj 2>&1 | FileCheck -check-prefix=CHECK4 %s
 
 CHECK4: error: <root>: undefined symbol: main
diff --git a/lld/test/COFF/pdb-options.test b/lld/test/COFF/pdb-options.test
index 70f6cbf..840040e 100644
--- a/lld/test/COFF/pdb-options.test
+++ b/lld/test/COFF/pdb-options.test
@@ -19,8 +19,10 @@
 ; If /DEBUG is specified but not /pdb, it uses a default name in the current
 ; directory.  This is a bit hacky since but we need to be IN our test specific
 ; temporary directory when we run this command or we can't test this
-# RUN: cd %T
-# RUN: lld-link /DEBUG /entry:main /nodefaultlib %t1.obj %t2.obj
-# RUN: ls %t1.pdb
-# RUN: rm %t*
-# RUN: cd %T/..
+# RUN: mkdir -p %t.dir
+# RUN: cp %t1.obj %t.dir/1.obj
+# RUN: cp %t2.obj %t.dir/2.obj
+# RUN: cd %t.dir
+# RUN: lld-link /DEBUG /entry:main /nodefaultlib %t.dir/1.obj %t.dir/2.obj
+# RUN: ls %t.dir/1.pdb
+# RUN: rm -r %t*
diff --git a/lld/test/COFF/pdb-type-server-invalid-signature.yaml b/lld/test/COFF/pdb-type-server-invalid-signature.yaml
index 8f1528f..aedcefa 100644
--- a/lld/test/COFF/pdb-type-server-invalid-signature.yaml
+++ b/lld/test/COFF/pdb-type-server-invalid-signature.yaml
@@ -19,9 +19,10 @@
 # VALID-SIGNATURE-NOT: The signature does not match; the file(s) might be out of date
 
 # Test an invalid path reference to a PDB type server; as a fallback LLD should try to load the PDB in the same path as the OBJ
-# RUN: yaml2obj %S/Inputs/pdb-type-server-invalid-path.yaml -o %t3.obj
-# RUN: cp %S/Inputs/pdb-diff-cl.pdb %T
-# RUN: lld-link %t3.obj -out:%t3.exe -debug -pdb:%t3.pdb -nodefaultlib -entry:main 2>&1 | FileCheck -DMSG=%errc_ENOENT %s -check-prefix=INVALID-PATH -allow-empty
+# RUN: mkdir -p %t.dir
+# RUN: yaml2obj %S/Inputs/pdb-type-server-invalid-path.yaml -o %t.dir/3.obj
+# RUN: cp %S/Inputs/pdb-diff-cl.pdb %t.dir/pdb-diff-cl.pdb
+# RUN: lld-link %t.dir/3.obj -out:%t3.exe -debug -pdb:%t3.pdb -nodefaultlib -entry:main 2>&1 | FileCheck -DMSG=%errc_ENOENT %s -check-prefix=INVALID-PATH -allow-empty
 
 # INVALID-PATH-NOT: warning: Cannot use debug info for '{{.*}}3.obj' [LNK4099]
 # INVALID-PATH-NOT: failed to load reference 'c:\some_invalid_path_AABB98765\pdb-diff-cl.pdb': [[MSG]]
diff --git a/lld/test/COFF/wrap-lto-2.ll b/lld/test/COFF/wrap-lto-2.ll
index c50feab..4c74ff2 100644
--- a/lld/test/COFF/wrap-lto-2.ll
+++ b/lld/test/COFF/wrap-lto-2.ll
@@ -23,28 +23,30 @@
 ;; the wrapped symbol, when LTO or ThinLTO is involved. It checks for various
 ;; combinations of bitcode and regular objects.
 
+; RUN: mkdir -p %t.dir
+
 ;; LTO + LTO
-; RUN: lld-link -out:%t.bc-bc.exe %t.main.bc -libpath:%T %t.bc.lib -entry:entry -subsystem:console -wrap:bar -debug:symtab -lldsavetemps
+; RUN: lld-link -out:%t.bc-bc.exe %t.main.bc -libpath:%t.dir %t.bc.lib -entry:entry -subsystem:console -wrap:bar -debug:symtab -lldsavetemps
 ; RUN: llvm-objdump -d %t.bc-bc.exe | FileCheck %s --check-prefixes=CHECK,JMP
 
 ;; LTO + Object
-; RUN: lld-link -out:%t.bc-obj.exe %t.main.bc -libpath:%T %t.obj.lib -entry:entry -subsystem:console -wrap:bar -debug:symtab -lldsavetemps
+; RUN: lld-link -out:%t.bc-obj.exe %t.main.bc -libpath:%t.dir %t.obj.lib -entry:entry -subsystem:console -wrap:bar -debug:symtab -lldsavetemps
 ; RUN: llvm-objdump -d %t.bc-obj.exe | FileCheck %s --check-prefixes=CHECK,JMP
 
 ;; Object + LTO
-; RUN: lld-link -out:%t.obj-bc.exe %t.main.obj -libpath:%T %t.bc.lib -entry:entry -subsystem:console -wrap:bar -debug:symtab -lldsavetemps
+; RUN: lld-link -out:%t.obj-bc.exe %t.main.obj -libpath:%t.dir %t.bc.lib -entry:entry -subsystem:console -wrap:bar -debug:symtab -lldsavetemps
 ; RUN: llvm-objdump -d %t.obj-bc.exe | FileCheck %s --check-prefixes=CHECK,CALL
 
 ;; ThinLTO + ThinLTO
-; RUN: lld-link -out:%t.thin-thin.exe %t.main.thin -libpath:%T %t.thin.lib -entry:entry -subsystem:console -wrap:bar -debug:symtab -lldsavetemps
+; RUN: lld-link -out:%t.thin-thin.exe %t.main.thin -libpath:%t.dir %t.thin.lib -entry:entry -subsystem:console -wrap:bar -debug:symtab -lldsavetemps
 ; RUN: llvm-objdump -d %t.thin-thin.exe | FileCheck %s --check-prefixes=CHECK,JMP
 
 ;; ThinLTO + Object
-; RUN: lld-link -out:%t.thin-obj.exe %t.main.thin -libpath:%T %t.obj.lib -entry:entry -subsystem:console -wrap:bar -debug:symtab -lldsavetemps
+; RUN: lld-link -out:%t.thin-obj.exe %t.main.thin -libpath:%t.dir %t.obj.lib -entry:entry -subsystem:console -wrap:bar -debug:symtab -lldsavetemps
 ; RUN: llvm-objdump -d %t.thin-obj.exe | FileCheck %s --check-prefixes=CHECK,JMP
 
 ;; Object + ThinLTO
-; RUN: lld-link -out:%t.obj-thin.exe %t.main.obj -libpath:%T %t.thin.lib -entry:entry -subsystem:console -wrap:bar -debug:symtab -lldsavetemps
+; RUN: lld-link -out:%t.obj-thin.exe %t.main.obj -libpath:%t.dir %t.thin.lib -entry:entry -subsystem:console -wrap:bar -debug:symtab -lldsavetemps
 ; RUN: llvm-objdump -d %t.obj-thin.exe | FileCheck %s --check-prefixes=CHECK,CALL
 
 ;; Make sure that calls in entry() are not eliminated and that bar is
diff --git a/lld/test/COFF/wrap-with-archive.s b/lld/test/COFF/wrap-with-archive.s
index 96b244a..d8a4fdb 100644
--- a/lld/test/COFF/wrap-with-archive.s
+++ b/lld/test/COFF/wrap-with-archive.s
@@ -6,7 +6,8 @@
 // RUN: rm -f %t.lib
 // RUN: llvm-ar rcs %t.lib %t.wrap.obj %t.other.obj
 
-// RUN: lld-link -out:%t.exe %t.main.obj -libpath:%T %t.lib -entry:entry -subsystem:console -wrap:foo
+// RUN: mkdir -p %t.dir
+// RUN: lld-link -out:%t.exe %t.main.obj -libpath:%t.dir %t.lib -entry:entry -subsystem:console -wrap:foo
 
 // Note: No real definition of foo exists here, but that works fine as long
 // as there's no actual references to __real_foo.
diff --git a/lld/test/wasm/lto/save-temps.ll b/lld/test/wasm/lto/save-temps.ll
index 773978e..e5e96d3 100644
--- a/lld/test/wasm/lto/save-temps.ll
+++ b/lld/test/wasm/lto/save-temps.ll
@@ -1,4 +1,5 @@
-; RUN: cd %T
+; RUN: mkdir -p %t.dir
+; RUN: cd %t.dir
 ; RUN: rm -f a.out a.out.lto.bc a.out.lto.o
 ; RUN: llvm-as %s -o %t.o
 ; RUN: llvm-as %p/Inputs/save-temps.ll -o %t2.o
diff --git a/lldb/cmake/modules/LLDBConfig.cmake b/lldb/cmake/modules/LLDBConfig.cmake
index 1bc494a..fc84e58 100644
--- a/lldb/cmake/modules/LLDBConfig.cmake
+++ b/lldb/cmake/modules/LLDBConfig.cmake
@@ -327,6 +327,8 @@ endif()
 # lldb-rpc sources in the first phase of host build so that they can
 # get built using the just-built Clang toolchain in the second phase.
 if (NOT DEFINED LLDB_CAN_USE_LLDB_RPC_SERVER)
+  set(LLDB_CAN_USE_LLDB_RPC_SERVER OFF)
+else()
   if ((CMAKE_CROSSCOMPILING OR LLVM_HOST_TRIPLE MATCHES "${LLVM_DEFAULT_TARGET_TRIPLE}") AND
       CMAKE_SYSTEM_NAME MATCHES "AIX|Android|Darwin|FreeBSD|Linux|NetBSD|OpenBSD|Windows")
     set(LLDB_CAN_USE_LLDB_RPC_SERVER ON)
@@ -335,11 +337,16 @@ if (NOT DEFINED LLDB_CAN_USE_LLDB_RPC_SERVER)
   endif()
 endif()
 
-if (CMAKE_CROSSCOMPILING)
-  set(LLDB_BUILD_LLDBRPC OFF CACHE BOOL "")
-  get_host_tool_path(lldb-rpc-gen LLDB_RPC_GEN_EXE lldb_rpc_gen_exe lldb_rpc_gen_target)
+
+if (NOT DEFINED LLDB_BUILD_LLDBRPC)
+  set(LLDB_BUILD_LLDBRPC OFF)
 else()
-  set(LLDB_BUILD_LLDBRPC ON CACHE BOOL "")
+  if (CMAKE_CROSSCOMPILING)
+    set(LLDB_BUILD_LLDBRPC OFF CACHE BOOL "")
+    get_host_tool_path(lldb-rpc-gen LLDB_RPC_GEN_EXE lldb_rpc_gen_exe lldb_rpc_gen_target)
+  else()
+    set(LLDB_BUILD_LLDBRPC ON CACHE BOOL "")
+  endif()
 endif()
 
 include(LLDBGenerateConfig)
diff --git a/lldb/docs/index.rst b/lldb/docs/index.rst
index d6c68b1..a981c0a 100644
--- a/lldb/docs/index.rst
+++ b/lldb/docs/index.rst
@@ -86,7 +86,7 @@ current state, follow the links to their respective issues:
 
 * `RISC-V <https://github.com/llvm/llvm-project/issues/55383>`_
 * `LoongArch <https://github.com/llvm/llvm-project/issues/112693>`_
-* `WebAssembly <https://github.com/llvm/llvm-project/issues/150449`_
+* `WebAssembly <https://github.com/llvm/llvm-project/issues/150449>`_
 
 Get Involved
 ------------
diff --git a/lldb/include/lldb/Target/Process.h b/lldb/include/lldb/Target/Process.h
index 637b077..7e66e31 100644
--- a/lldb/include/lldb/Target/Process.h
+++ b/lldb/include/lldb/Target/Process.h
@@ -2619,7 +2619,7 @@ void PruneThreadPlans();
 
   void ResetExtendedCrashInfoDict() {
     // StructuredData::Dictionary is add only, so we have to make a new one:
-    m_crash_info_dict_sp.reset(new StructuredData::Dictionary());
+    m_crash_info_dict_sp = std::make_shared<StructuredData::Dictionary>();
   }
 
   size_t AddImageToken(lldb::addr_t image_ptr);
diff --git a/lldb/include/lldb/Target/StackFrameList.h b/lldb/include/lldb/Target/StackFrameList.h
index 8a66296..e5a6e942 100644
--- a/lldb/include/lldb/Target/StackFrameList.h
+++ b/lldb/include/lldb/Target/StackFrameList.h
@@ -174,6 +174,11 @@ protected:
   /// change the frame if this is the first time GetSelectedFrame is called.
   std::optional<uint32_t> m_selected_frame_idx;
 
+  /// Protect access to m_selected_frame_idx. Always acquire after m_list_mutex
+  /// to avoid lock inversion. A recursive mutex because GetSelectedFrameIndex
+  /// may indirectly call SetSelectedFrame.
+  std::recursive_mutex m_selected_frame_mutex;
+
   /// The number of concrete frames fetched while filling the frame list. This
   /// is only used when synthetic frames are enabled.
   uint32_t m_concrete_frames_fetched;
diff --git a/lldb/source/API/SBType.cpp b/lldb/source/API/SBType.cpp
index 00f2871..f58902d 100644
--- a/lldb/source/API/SBType.cpp
+++ b/lldb/source/API/SBType.cpp
@@ -184,7 +184,7 @@ SBType SBType::GetPointerType() {
   if (!IsValid())
     return SBType();
 
-  return SBType(TypeImplSP(new TypeImpl(m_opaque_sp->GetPointerType())));
+  return SBType(std::make_shared<TypeImpl>(m_opaque_sp->GetPointerType()));
 }
 
 SBType SBType::GetPointeeType() {
@@ -192,7 +192,7 @@ SBType SBType::GetPointeeType() {
 
   if (!IsValid())
     return SBType();
-  return SBType(TypeImplSP(new TypeImpl(m_opaque_sp->GetPointeeType())));
+  return SBType(std::make_shared<TypeImpl>(m_opaque_sp->GetPointeeType()));
 }
 
 SBType SBType::GetReferenceType() {
@@ -200,7 +200,7 @@ SBType SBType::GetReferenceType() {
 
   if (!IsValid())
     return SBType();
-  return SBType(TypeImplSP(new TypeImpl(m_opaque_sp->GetReferenceType())));
+  return SBType(std::make_shared<TypeImpl>(m_opaque_sp->GetReferenceType()));
 }
 
 SBType SBType::GetTypedefedType() {
@@ -208,7 +208,7 @@ SBType SBType::GetTypedefedType() {
 
   if (!IsValid())
     return SBType();
-  return SBType(TypeImplSP(new TypeImpl(m_opaque_sp->GetTypedefedType())));
+  return SBType(std::make_shared<TypeImpl>(m_opaque_sp->GetTypedefedType()));
 }
 
 SBType SBType::GetDereferencedType() {
@@ -216,7 +216,7 @@ SBType SBType::GetDereferencedType() {
 
   if (!IsValid())
     return SBType();
-  return SBType(TypeImplSP(new TypeImpl(m_opaque_sp->GetDereferencedType())));
+  return SBType(std::make_shared<TypeImpl>(m_opaque_sp->GetDereferencedType()));
 }
 
 SBType SBType::GetArrayElementType() {
@@ -224,8 +224,8 @@ SBType SBType::GetArrayElementType() {
 
   if (!IsValid())
     return SBType();
-  return SBType(TypeImplSP(new TypeImpl(
-      m_opaque_sp->GetCompilerType(true).GetArrayElementType(nullptr))));
+  return SBType(std::make_shared<TypeImpl>(
+      m_opaque_sp->GetCompilerType(true).GetArrayElementType(nullptr)));
 }
 
 SBType SBType::GetArrayType(uint64_t size) {
@@ -233,8 +233,8 @@ SBType SBType::GetArrayType(uint64_t size) {
 
   if (!IsValid())
     return SBType();
-  return SBType(TypeImplSP(
-      new TypeImpl(m_opaque_sp->GetCompilerType(true).GetArrayType(size))));
+  return SBType(std::make_shared<TypeImpl>(
+      m_opaque_sp->GetCompilerType(true).GetArrayType(size)));
 }
 
 SBType SBType::GetVectorElementType() {
@@ -245,7 +245,7 @@ SBType SBType::GetVectorElementType() {
     CompilerType vector_element_type;
     if (m_opaque_sp->GetCompilerType(true).IsVectorType(&vector_element_type,
                                                         nullptr))
-      type_sb.SetSP(TypeImplSP(new TypeImpl(vector_element_type)));
+      type_sb.SetSP(std::make_shared<TypeImpl>(vector_element_type));
   }
   return type_sb;
 }
@@ -421,14 +421,14 @@ lldb::SBType SBType::GetUnqualifiedType() {
 
   if (!IsValid())
     return SBType();
-  return SBType(TypeImplSP(new TypeImpl(m_opaque_sp->GetUnqualifiedType())));
+  return SBType(std::make_shared<TypeImpl>(m_opaque_sp->GetUnqualifiedType()));
 }
 
 lldb::SBType SBType::GetCanonicalType() {
   LLDB_INSTRUMENT_VA(this);
 
   if (IsValid())
-    return SBType(TypeImplSP(new TypeImpl(m_opaque_sp->GetCanonicalType())));
+    return SBType(std::make_shared<TypeImpl>(m_opaque_sp->GetCanonicalType()));
   return SBType();
 }
 
@@ -508,7 +508,7 @@ SBTypeMember SBType::GetDirectBaseClassAtIndex(uint32_t idx) {
             idx, &bit_offset);
     if (base_class_type.IsValid())
       sb_type_member.reset(new TypeMemberImpl(
-          TypeImplSP(new TypeImpl(base_class_type)), bit_offset));
+          std::make_shared<TypeImpl>(base_class_type), bit_offset));
   }
   return sb_type_member;
 }
@@ -524,7 +524,7 @@ SBTypeMember SBType::GetVirtualBaseClassAtIndex(uint32_t idx) {
             idx, &bit_offset);
     if (base_class_type.IsValid())
       sb_type_member.reset(new TypeMemberImpl(
-          TypeImplSP(new TypeImpl(base_class_type)), bit_offset));
+          std::make_shared<TypeImpl>(base_class_type), bit_offset));
   }
   return sb_type_member;
 }
@@ -546,16 +546,15 @@ SBTypeEnumMemberList SBType::GetEnumMembers() {
   if (IsValid()) {
     CompilerType this_type(m_opaque_sp->GetCompilerType(true));
     if (this_type.IsValid()) {
-      this_type.ForEachEnumerator([&sb_enum_member_list](
-                                      const CompilerType &integer_type,
-                                      ConstString name,
-                                      const llvm::APSInt &value) -> bool {
-        SBTypeEnumMember enum_member(
-            lldb::TypeEnumMemberImplSP(new TypeEnumMemberImpl(
-                lldb::TypeImplSP(new TypeImpl(integer_type)), name, value)));
-        sb_enum_member_list.Append(enum_member);
-        return true; // Keep iterating
-      });
+      this_type.ForEachEnumerator(
+          [&sb_enum_member_list](const CompilerType &integer_type,
+                                 ConstString name,
+                                 const llvm::APSInt &value) -> bool {
+            SBTypeEnumMember enum_member(std::make_shared<TypeEnumMemberImpl>(
+                std::make_shared<TypeImpl>(integer_type), name, value));
+            sb_enum_member_list.Append(enum_member);
+            return true; // Keep iterating
+          });
     }
   }
   return sb_enum_member_list;
@@ -578,9 +577,9 @@ SBTypeMember SBType::GetFieldAtIndex(uint32_t idx) {
         ConstString name;
         if (!name_sstr.empty())
           name.SetCString(name_sstr.c_str());
-        sb_type_member.reset(
-            new TypeMemberImpl(TypeImplSP(new TypeImpl(field_type)), bit_offset,
-                               name, bitfield_bit_size, is_bitfield));
+        sb_type_member.reset(new TypeMemberImpl(
+            std::make_shared<TypeImpl>(field_type), bit_offset, name,
+            bitfield_bit_size, is_bitfield));
       }
     }
   }
@@ -978,7 +977,7 @@ SBType SBTypeMemberFunction::GetType() {
 
   SBType sb_type;
   if (m_opaque_sp) {
-    sb_type.SetSP(lldb::TypeImplSP(new TypeImpl(m_opaque_sp->GetType())));
+    sb_type.SetSP(std::make_shared<TypeImpl>(m_opaque_sp->GetType()));
   }
   return sb_type;
 }
@@ -988,7 +987,7 @@ lldb::SBType SBTypeMemberFunction::GetReturnType() {
 
   SBType sb_type;
   if (m_opaque_sp) {
-    sb_type.SetSP(lldb::TypeImplSP(new TypeImpl(m_opaque_sp->GetReturnType())));
+    sb_type.SetSP(std::make_shared<TypeImpl>(m_opaque_sp->GetReturnType()));
   }
   return sb_type;
 }
@@ -1007,7 +1006,7 @@ lldb::SBType SBTypeMemberFunction::GetArgumentTypeAtIndex(uint32_t i) {
   SBType sb_type;
   if (m_opaque_sp) {
     sb_type.SetSP(
-        lldb::TypeImplSP(new TypeImpl(m_opaque_sp->GetArgumentAtIndex(i))));
+        std::make_shared<TypeImpl>(m_opaque_sp->GetArgumentAtIndex(i)));
   }
   return sb_type;
 }
diff --git a/lldb/source/API/SBTypeFilter.cpp b/lldb/source/API/SBTypeFilter.cpp
index f1b5bc9..e10d269 100644
--- a/lldb/source/API/SBTypeFilter.cpp
+++ b/lldb/source/API/SBTypeFilter.cpp
@@ -19,7 +19,7 @@ using namespace lldb_private;
 SBTypeFilter::SBTypeFilter() { LLDB_INSTRUMENT_VA(this); }
 
 SBTypeFilter::SBTypeFilter(uint32_t options)
-    : m_opaque_sp(TypeFilterImplSP(new TypeFilterImpl(options))) {
+    : m_opaque_sp(std::make_shared<TypeFilterImpl>(options)) {
   LLDB_INSTRUMENT_VA(this, options);
 }
 
diff --git a/lldb/source/API/SBTypeNameSpecifier.cpp b/lldb/source/API/SBTypeNameSpecifier.cpp
index 308b1cd..dd81720 100644
--- a/lldb/source/API/SBTypeNameSpecifier.cpp
+++ b/lldb/source/API/SBTypeNameSpecifier.cpp
@@ -38,8 +38,8 @@ SBTypeNameSpecifier::SBTypeNameSpecifier(SBType type) {
   LLDB_INSTRUMENT_VA(this, type);
 
   if (type.IsValid())
-    m_opaque_sp = TypeNameSpecifierImplSP(
-        new TypeNameSpecifierImpl(type.m_opaque_sp->GetCompilerType(true)));
+    m_opaque_sp = std::make_shared<TypeNameSpecifierImpl>(
+        type.m_opaque_sp->GetCompilerType(true));
 }
 
 SBTypeNameSpecifier::SBTypeNameSpecifier(const lldb::SBTypeNameSpecifier &rhs)
diff --git a/lldb/source/API/SBTypeSynthetic.cpp b/lldb/source/API/SBTypeSynthetic.cpp
index 19a4c53..5ebc884 100644
--- a/lldb/source/API/SBTypeSynthetic.cpp
+++ b/lldb/source/API/SBTypeSynthetic.cpp
@@ -24,8 +24,8 @@ SBTypeSynthetic SBTypeSynthetic::CreateWithClassName(const char *data,
 
   if (!data || data[0] == 0)
     return SBTypeSynthetic();
-  return SBTypeSynthetic(ScriptedSyntheticChildrenSP(
-      new ScriptedSyntheticChildren(options, data, "")));
+  return SBTypeSynthetic(
+      std::make_shared<ScriptedSyntheticChildren>(options, data, ""));
 }
 
 SBTypeSynthetic SBTypeSynthetic::CreateWithScriptCode(const char *data,
@@ -34,8 +34,8 @@ SBTypeSynthetic SBTypeSynthetic::CreateWithScriptCode(const char *data,
 
   if (!data || data[0] == 0)
     return SBTypeSynthetic();
-  return SBTypeSynthetic(ScriptedSyntheticChildrenSP(
-      new ScriptedSyntheticChildren(options, "", data)));
+  return SBTypeSynthetic(
+      std::make_shared<ScriptedSyntheticChildren>(options, "", data));
 }
 
 SBTypeSynthetic::SBTypeSynthetic(const lldb::SBTypeSynthetic &rhs)
diff --git a/lldb/source/API/SBValue.cpp b/lldb/source/API/SBValue.cpp
index d878eb4..e300ece 100644
--- a/lldb/source/API/SBValue.cpp
+++ b/lldb/source/API/SBValue.cpp
@@ -1120,11 +1120,11 @@ void SBValue::SetSP(const lldb::ValueObjectSP &sp) {
       lldb::DynamicValueType use_dynamic = target_sp->GetPreferDynamicValue();
       bool use_synthetic =
           target_sp->TargetProperties::GetEnableSyntheticValue();
-      m_opaque_sp = ValueImplSP(new ValueImpl(sp, use_dynamic, use_synthetic));
+      m_opaque_sp = std::make_shared<ValueImpl>(sp, use_dynamic, use_synthetic);
     } else
-      m_opaque_sp = ValueImplSP(new ValueImpl(sp, eNoDynamicValues, true));
+      m_opaque_sp = std::make_shared<ValueImpl>(sp, eNoDynamicValues, true);
   } else
-    m_opaque_sp = ValueImplSP(new ValueImpl(sp, eNoDynamicValues, false));
+    m_opaque_sp = std::make_shared<ValueImpl>(sp, eNoDynamicValues, false);
 }
 
 void SBValue::SetSP(const lldb::ValueObjectSP &sp,
@@ -1155,14 +1155,14 @@ void SBValue::SetSP(const lldb::ValueObjectSP &sp, bool use_synthetic) {
 
 void SBValue::SetSP(const lldb::ValueObjectSP &sp,
                     lldb::DynamicValueType use_dynamic, bool use_synthetic) {
-  m_opaque_sp = ValueImplSP(new ValueImpl(sp, use_dynamic, use_synthetic));
+  m_opaque_sp = std::make_shared<ValueImpl>(sp, use_dynamic, use_synthetic);
 }
 
 void SBValue::SetSP(const lldb::ValueObjectSP &sp,
                     lldb::DynamicValueType use_dynamic, bool use_synthetic,
                     const char *name) {
   m_opaque_sp =
-      ValueImplSP(new ValueImpl(sp, use_dynamic, use_synthetic, name));
+      std::make_shared<ValueImpl>(sp, use_dynamic, use_synthetic, name);
 }
 
 bool SBValue::GetExpressionPath(SBStream &description) {
diff --git a/lldb/source/Breakpoint/Breakpoint.cpp b/lldb/source/Breakpoint/Breakpoint.cpp
index d757bc4..1544bf8 100644
--- a/lldb/source/Breakpoint/Breakpoint.cpp
+++ b/lldb/source/Breakpoint/Breakpoint.cpp
@@ -83,8 +83,7 @@ StructuredData::ObjectSP Breakpoint::SerializeToStructuredData() {
   if (!m_name_list.empty()) {
     StructuredData::ArraySP names_array_sp(new StructuredData::Array());
     for (auto name : m_name_list) {
-      names_array_sp->AddItem(
-          StructuredData::StringSP(new StructuredData::String(name)));
+      names_array_sp->AddItem(std::make_shared<StructuredData::String>(name));
     }
     breakpoint_contents_sp->AddItem(Breakpoint::GetKey(OptionNames::Names),
                                     names_array_sp);
diff --git a/lldb/source/Breakpoint/BreakpointResolverName.cpp b/lldb/source/Breakpoint/BreakpointResolverName.cpp
index edde1c9..21024a4 100644
--- a/lldb/source/Breakpoint/BreakpointResolverName.cpp
+++ b/lldb/source/Breakpoint/BreakpointResolverName.cpp
@@ -195,10 +195,10 @@ StructuredData::ObjectSP BreakpointResolverName::SerializeToStructuredData() {
     StructuredData::ArraySP names_sp(new StructuredData::Array());
     StructuredData::ArraySP name_masks_sp(new StructuredData::Array());
     for (auto lookup : m_lookups) {
-      names_sp->AddItem(StructuredData::StringSP(
-          new StructuredData::String(lookup.GetName().GetStringRef())));
-      name_masks_sp->AddItem(StructuredData::UnsignedIntegerSP(
-          new StructuredData::UnsignedInteger(lookup.GetNameTypeMask())));
+      names_sp->AddItem(std::make_shared<StructuredData::String>(
+          lookup.GetName().GetStringRef()));
+      name_masks_sp->AddItem(std::make_shared<StructuredData::UnsignedInteger>(
+          lookup.GetNameTypeMask()));
     }
     options_dict_sp->AddItem(GetKey(OptionNames::SymbolNameArray), names_sp);
     options_dict_sp->AddItem(GetKey(OptionNames::NameMaskArray), name_masks_sp);
diff --git a/lldb/source/Commands/CommandObjectCommands.cpp b/lldb/source/Commands/CommandObjectCommands.cpp
index 10dc273..3049eb8 100644
--- a/lldb/source/Commands/CommandObjectCommands.cpp
+++ b/lldb/source/Commands/CommandObjectCommands.cpp
@@ -25,6 +25,7 @@
 #include "lldb/Utility/Args.h"
 #include "lldb/Utility/StringList.h"
 #include "llvm/ADT/StringRef.h"
+#include <memory>
 #include <optional>
 
 using namespace lldb;
@@ -467,7 +468,7 @@ protected:
     // Verify & handle any options/arguments passed to the alias command
 
     OptionArgVectorSP option_arg_vector_sp =
-        OptionArgVectorSP(new OptionArgVector);
+        std::make_shared<OptionArgVector>();
 
     const bool include_aliases = true;
     // Look up the command using command's name first.  This is to resolve
@@ -543,7 +544,7 @@ protected:
     CommandObject *cmd_obj = command_obj_sp.get();
     CommandObject *sub_cmd_obj = nullptr;
     OptionArgVectorSP option_arg_vector_sp =
-        OptionArgVectorSP(new OptionArgVector);
+        std::make_shared<OptionArgVector>();
 
     while (cmd_obj->IsMultiwordObject() && !args.empty()) {
       auto sub_command = args[0].ref();
@@ -2504,9 +2505,9 @@ protected:
 
     CommandObjectSP new_cmd_sp;
     if (m_options.m_class_name.empty()) {
-      new_cmd_sp.reset(new CommandObjectPythonFunction(
+      new_cmd_sp = std::make_shared<CommandObjectPythonFunction>(
           m_interpreter, m_cmd_name, m_options.m_funct_name,
-          m_options.m_short_help, m_synchronicity, m_completion_type));
+          m_options.m_short_help, m_synchronicity, m_completion_type);
     } else {
       ScriptInterpreter *interpreter = GetDebugger().GetScriptInterpreter();
       if (!interpreter) {
@@ -2528,9 +2529,9 @@ protected:
         if (!result.Succeeded())
           return;
       } else
-        new_cmd_sp.reset(new CommandObjectScriptingObjectRaw(
+        new_cmd_sp = std::make_shared<CommandObjectScriptingObjectRaw>(
             m_interpreter, m_cmd_name, cmd_obj_sp, m_synchronicity,
-            m_completion_type));
+            m_completion_type);
     }
     
     // Assume we're going to succeed...
diff --git a/lldb/source/Commands/CommandObjectFrame.cpp b/lldb/source/Commands/CommandObjectFrame.cpp
index 7e42ef2..5692699 100644
--- a/lldb/source/Commands/CommandObjectFrame.cpp
+++ b/lldb/source/Commands/CommandObjectFrame.cpp
@@ -901,10 +901,9 @@ void CommandObjectFrameRecognizerAdd::DoExecute(Args &command,
       StackFrameRecognizerSP(new ScriptedStackFrameRecognizer(
           interpreter, m_options.m_class_name.c_str()));
   if (m_options.m_regex) {
-    auto module =
-        RegularExpressionSP(new RegularExpression(m_options.m_module));
+    auto module = std::make_shared<RegularExpression>(m_options.m_module);
     auto func =
-        RegularExpressionSP(new RegularExpression(m_options.m_symbols.front()));
+        std::make_shared<RegularExpression>(m_options.m_symbols.front());
     GetTarget().GetFrameRecognizerManager().AddRecognizer(
         recognizer_sp, module, func, Mangled::NamePreference::ePreferDemangled,
         m_options.m_first_instruction_only);
diff --git a/lldb/source/Core/IOHandlerCursesGUI.cpp b/lldb/source/Core/IOHandlerCursesGUI.cpp
index 7f0e0fc..3976630 100644
--- a/lldb/source/Core/IOHandlerCursesGUI.cpp
+++ b/lldb/source/Core/IOHandlerCursesGUI.cpp
@@ -22,6 +22,7 @@
 #if defined(__APPLE__)
 #include <deque>
 #endif
+#include <memory>
 #include <string>
 
 #include "lldb/Core/Debugger.h"
@@ -6536,7 +6537,7 @@ public:
       if (process && process->IsAlive() &&
           StateIsStoppedState(process->GetState(), true)) {
         if (submenus.size() == 7)
-          menu.AddSubmenu(MenuSP(new Menu(Menu::Type::Separator)));
+          menu.AddSubmenu(std::make_shared<Menu>(Menu::Type::Separator));
         else if (submenus.size() > 8)
           submenus.erase(submenus.begin() + 8, submenus.end());
 
@@ -6558,9 +6559,9 @@ public:
             if (queue_name && queue_name[0])
               thread_menu_title.Printf(" %s", queue_name);
           }
-          menu.AddSubmenu(
-              MenuSP(new Menu(thread_menu_title.GetString().str().c_str(),
-                              nullptr, menu_char, thread_sp->GetID())));
+          menu.AddSubmenu(std::make_shared<Menu>(
+              thread_menu_title.GetString().str().c_str(), nullptr, menu_char,
+              thread_sp->GetID()));
         }
       } else if (submenus.size() > 7) {
         // Remove the separator and any other thread submenu items that were
@@ -7573,70 +7574,67 @@ void IOHandlerCursesGUI::Activate() {
     MenuSP exit_menuitem_sp(
         new Menu("Exit", nullptr, 'x', ApplicationDelegate::eMenuID_LLDBExit));
     exit_menuitem_sp->SetCannedResult(MenuActionResult::Quit);
-    lldb_menu_sp->AddSubmenu(MenuSP(new Menu(
-        "About LLDB", nullptr, 'a', ApplicationDelegate::eMenuID_LLDBAbout)));
-    lldb_menu_sp->AddSubmenu(MenuSP(new Menu(Menu::Type::Separator)));
+    lldb_menu_sp->AddSubmenu(std::make_shared<Menu>(
+        "About LLDB", nullptr, 'a', ApplicationDelegate::eMenuID_LLDBAbout));
+    lldb_menu_sp->AddSubmenu(std::make_shared<Menu>(Menu::Type::Separator));
     lldb_menu_sp->AddSubmenu(exit_menuitem_sp);
 
     MenuSP target_menu_sp(new Menu("Target", "F2", KEY_F(2),
                                    ApplicationDelegate::eMenuID_Target));
-    target_menu_sp->AddSubmenu(MenuSP(new Menu(
-        "Create", nullptr, 'c', ApplicationDelegate::eMenuID_TargetCreate)));
-    target_menu_sp->AddSubmenu(MenuSP(new Menu(
-        "Delete", nullptr, 'd', ApplicationDelegate::eMenuID_TargetDelete)));
+    target_menu_sp->AddSubmenu(std::make_shared<Menu>(
+        "Create", nullptr, 'c', ApplicationDelegate::eMenuID_TargetCreate));
+    target_menu_sp->AddSubmenu(std::make_shared<Menu>(
+        "Delete", nullptr, 'd', ApplicationDelegate::eMenuID_TargetDelete));
 
     MenuSP process_menu_sp(new Menu("Process", "F3", KEY_F(3),
                                     ApplicationDelegate::eMenuID_Process));
-    process_menu_sp->AddSubmenu(MenuSP(new Menu(
-        "Attach", nullptr, 'a', ApplicationDelegate::eMenuID_ProcessAttach)));
-    process_menu_sp->AddSubmenu(
-        MenuSP(new Menu("Detach and resume", nullptr, 'd',
-                        ApplicationDelegate::eMenuID_ProcessDetachResume)));
-    process_menu_sp->AddSubmenu(
-        MenuSP(new Menu("Detach suspended", nullptr, 's',
-                        ApplicationDelegate::eMenuID_ProcessDetachSuspended)));
-    process_menu_sp->AddSubmenu(MenuSP(new Menu(
-        "Launch", nullptr, 'l', ApplicationDelegate::eMenuID_ProcessLaunch)));
-    process_menu_sp->AddSubmenu(MenuSP(new Menu(Menu::Type::Separator)));
+    process_menu_sp->AddSubmenu(std::make_shared<Menu>(
+        "Attach", nullptr, 'a', ApplicationDelegate::eMenuID_ProcessAttach));
+    process_menu_sp->AddSubmenu(std::make_shared<Menu>(
+        "Detach and resume", nullptr, 'd',
+        ApplicationDelegate::eMenuID_ProcessDetachResume));
+    process_menu_sp->AddSubmenu(std::make_shared<Menu>(
+        "Detach suspended", nullptr, 's',
+        ApplicationDelegate::eMenuID_ProcessDetachSuspended));
+    process_menu_sp->AddSubmenu(std::make_shared<Menu>(
+        "Launch", nullptr, 'l', ApplicationDelegate::eMenuID_ProcessLaunch));
+    process_menu_sp->AddSubmenu(std::make_shared<Menu>(Menu::Type::Separator));
     process_menu_sp->AddSubmenu(
-        MenuSP(new Menu("Continue", nullptr, 'c',
-                        ApplicationDelegate::eMenuID_ProcessContinue)));
-    process_menu_sp->AddSubmenu(MenuSP(new Menu(
-        "Halt", nullptr, 'h', ApplicationDelegate::eMenuID_ProcessHalt)));
-    process_menu_sp->AddSubmenu(MenuSP(new Menu(
-        "Kill", nullptr, 'k', ApplicationDelegate::eMenuID_ProcessKill)));
+        std::make_shared<Menu>("Continue", nullptr, 'c',
+                               ApplicationDelegate::eMenuID_ProcessContinue));
+    process_menu_sp->AddSubmenu(std::make_shared<Menu>(
+        "Halt", nullptr, 'h', ApplicationDelegate::eMenuID_ProcessHalt));
+    process_menu_sp->AddSubmenu(std::make_shared<Menu>(
+        "Kill", nullptr, 'k', ApplicationDelegate::eMenuID_ProcessKill));
 
     MenuSP thread_menu_sp(new Menu("Thread", "F4", KEY_F(4),
                                    ApplicationDelegate::eMenuID_Thread));
-    thread_menu_sp->AddSubmenu(MenuSP(new Menu(
-        "Step In", nullptr, 'i', ApplicationDelegate::eMenuID_ThreadStepIn)));
+    thread_menu_sp->AddSubmenu(std::make_shared<Menu>(
+        "Step In", nullptr, 'i', ApplicationDelegate::eMenuID_ThreadStepIn));
     thread_menu_sp->AddSubmenu(
-        MenuSP(new Menu("Step Over", nullptr, 'v',
-                        ApplicationDelegate::eMenuID_ThreadStepOver)));
-    thread_menu_sp->AddSubmenu(MenuSP(new Menu(
-        "Step Out", nullptr, 'o', ApplicationDelegate::eMenuID_ThreadStepOut)));
+        std::make_shared<Menu>("Step Over", nullptr, 'v',
+                               ApplicationDelegate::eMenuID_ThreadStepOver));
+    thread_menu_sp->AddSubmenu(std::make_shared<Menu>(
+        "Step Out", nullptr, 'o', ApplicationDelegate::eMenuID_ThreadStepOut));
 
     MenuSP view_menu_sp(
         new Menu("View", "F5", KEY_F(5), ApplicationDelegate::eMenuID_View));
+    view_menu_sp->AddSubmenu(std::make_shared<Menu>(
+        "Backtrace", nullptr, 't', ApplicationDelegate::eMenuID_ViewBacktrace));
+    view_menu_sp->AddSubmenu(std::make_shared<Menu>(
+        "Registers", nullptr, 'r', ApplicationDelegate::eMenuID_ViewRegisters));
+    view_menu_sp->AddSubmenu(std::make_shared<Menu>(
+        "Source", nullptr, 's', ApplicationDelegate::eMenuID_ViewSource));
+    view_menu_sp->AddSubmenu(std::make_shared<Menu>(
+        "Variables", nullptr, 'v', ApplicationDelegate::eMenuID_ViewVariables));
     view_menu_sp->AddSubmenu(
-        MenuSP(new Menu("Backtrace", nullptr, 't',
-                        ApplicationDelegate::eMenuID_ViewBacktrace)));
-    view_menu_sp->AddSubmenu(
-        MenuSP(new Menu("Registers", nullptr, 'r',
-                        ApplicationDelegate::eMenuID_ViewRegisters)));
-    view_menu_sp->AddSubmenu(MenuSP(new Menu(
-        "Source", nullptr, 's', ApplicationDelegate::eMenuID_ViewSource)));
-    view_menu_sp->AddSubmenu(
-        MenuSP(new Menu("Variables", nullptr, 'v',
-                        ApplicationDelegate::eMenuID_ViewVariables)));
-    view_menu_sp->AddSubmenu(
-        MenuSP(new Menu("Breakpoints", nullptr, 'b',
-                        ApplicationDelegate::eMenuID_ViewBreakpoints)));
+        std::make_shared<Menu>("Breakpoints", nullptr, 'b',
+                               ApplicationDelegate::eMenuID_ViewBreakpoints));
 
     MenuSP help_menu_sp(
         new Menu("Help", "F6", KEY_F(6), ApplicationDelegate::eMenuID_Help));
-    help_menu_sp->AddSubmenu(MenuSP(new Menu(
-        "GUI Help", nullptr, 'g', ApplicationDelegate::eMenuID_HelpGUIHelp)));
+    help_menu_sp->AddSubmenu(std::make_shared<Menu>(
+        "GUI Help", nullptr, 'g', ApplicationDelegate::eMenuID_HelpGUIHelp));
 
     m_app_up->Initialize();
     WindowSP &main_window_sp = m_app_up->GetMainWindow();
diff --git a/lldb/source/DataFormatters/FormatManager.cpp b/lldb/source/DataFormatters/FormatManager.cpp
index 122f230..7862fb8 100644
--- a/lldb/source/DataFormatters/FormatManager.cpp
+++ b/lldb/source/DataFormatters/FormatManager.cpp
@@ -422,9 +422,8 @@ FormatManager::GetCategory(ConstString category_name, bool can_create) {
   if (!can_create)
     return lldb::TypeCategoryImplSP();
 
-  m_categories_map.Add(
-      category_name,
-      lldb::TypeCategoryImplSP(new TypeCategoryImpl(this, category_name)));
+  m_categories_map.Add(category_name,
+                       std::make_shared<TypeCategoryImpl>(this, category_name));
   return GetCategory(category_name);
 }
 
diff --git a/lldb/source/DataFormatters/TypeCategoryMap.cpp b/lldb/source/DataFormatters/TypeCategoryMap.cpp
index 719264b..9412dd3 100644
--- a/lldb/source/DataFormatters/TypeCategoryMap.cpp
+++ b/lldb/source/DataFormatters/TypeCategoryMap.cpp
@@ -19,7 +19,7 @@ TypeCategoryMap::TypeCategoryMap(IFormatChangeListener *lst)
     : m_map_mutex(), listener(lst), m_map(), m_active_categories() {
   ConstString default_cs("default");
   lldb::TypeCategoryImplSP default_sp =
-      lldb::TypeCategoryImplSP(new TypeCategoryImpl(listener, default_cs));
+      std::make_shared<TypeCategoryImpl>(listener, default_cs);
   Add(default_cs, default_sp);
   Enable(default_cs, First);
 }
diff --git a/lldb/source/DataFormatters/ValueObjectPrinter.cpp b/lldb/source/DataFormatters/ValueObjectPrinter.cpp
index 05fcc4d..c2f8bb3 100644
--- a/lldb/source/DataFormatters/ValueObjectPrinter.cpp
+++ b/lldb/source/DataFormatters/ValueObjectPrinter.cpp
@@ -16,6 +16,7 @@
 #include "lldb/ValueObject/ValueObject.h"
 #include "llvm/Support/MathExtras.h"
 #include <cstdint>
+#include <memory>
 
 using namespace lldb;
 using namespace lldb_private;
@@ -62,10 +63,9 @@ void ValueObjectPrinter::Init(
   m_summary.assign("");
   m_error.assign("");
   m_val_summary_ok = false;
-  m_printed_instance_pointers =
-      printed_instance_pointers
-          ? printed_instance_pointers
-          : InstancePointersSetSP(new InstancePointersSet());
+  m_printed_instance_pointers = printed_instance_pointers
+                                    ? printed_instance_pointers
+                                    : std::make_shared<InstancePointersSet>();
   SetupMostSpecializedValue();
 }
 
diff --git a/lldb/source/Host/common/FileSystem.cpp b/lldb/source/Host/common/FileSystem.cpp
index 5153a0a9..00919fe 100644
--- a/lldb/source/Host/common/FileSystem.cpp
+++ b/lldb/source/Host/common/FileSystem.cpp
@@ -35,6 +35,7 @@
 
 #include <algorithm>
 #include <fstream>
+#include <memory>
 #include <optional>
 #include <vector>
 
@@ -288,8 +289,7 @@ FileSystem::CreateWritableDataBuffer(const llvm::Twine &path, uint64_t size,
                                                             is_volatile);
   if (!buffer)
     return {};
-  return std::shared_ptr<WritableDataBufferLLVM>(
-      new WritableDataBufferLLVM(std::move(buffer)));
+  return std::make_shared<WritableDataBufferLLVM>(std::move(buffer));
 }
 
 std::shared_ptr<DataBuffer>
@@ -300,7 +300,7 @@ FileSystem::CreateDataBuffer(const llvm::Twine &path, uint64_t size,
       GetMemoryBuffer<llvm::MemoryBuffer>(path, size, offset, is_volatile);
   if (!buffer)
     return {};
-  return std::shared_ptr<DataBufferLLVM>(new DataBufferLLVM(std::move(buffer)));
+  return std::make_shared<DataBufferLLVM>(std::move(buffer));
 }
 
 std::shared_ptr<WritableDataBuffer>
diff --git a/lldb/source/Host/windows/MainLoopWindows.cpp b/lldb/source/Host/windows/MainLoopWindows.cpp
index c1a0182..c0b1079 100644
--- a/lldb/source/Host/windows/MainLoopWindows.cpp
+++ b/lldb/source/Host/windows/MainLoopWindows.cpp
@@ -223,7 +223,7 @@ MainLoopWindows::RegisterReadObject(const IOObjectSP &object_sp,
 
   if (m_read_fds.find(waitable_handle) != m_read_fds.end()) {
     error = Status::FromErrorStringWithFormat(
-        "File descriptor %d already monitored.", waitable_handle);
+        "File descriptor %p already monitored.", waitable_handle);
     return nullptr;
   }
 
@@ -235,7 +235,7 @@ MainLoopWindows::RegisterReadObject(const IOObjectSP &object_sp,
   } else {
     DWORD file_type = GetFileType(waitable_handle);
     if (file_type != FILE_TYPE_PIPE) {
-      error = Status::FromErrorStringWithFormat("Unsupported file type %d",
+      error = Status::FromErrorStringWithFormat("Unsupported file type %ld",
                                                 file_type);
       return nullptr;
     }
diff --git a/lldb/source/Interpreter/CommandInterpreter.cpp b/lldb/source/Interpreter/CommandInterpreter.cpp
index da545f1..a0080cf 100644
--- a/lldb/source/Interpreter/CommandInterpreter.cpp
+++ b/lldb/source/Interpreter/CommandInterpreter.cpp
@@ -135,8 +135,7 @@ CommandInterpreter::CommandInterpreter(Debugger &debugger,
                                        bool synchronous_execution)
     : Broadcaster(debugger.GetBroadcasterManager(),
                   CommandInterpreter::GetStaticBroadcasterClass().str()),
-      Properties(
-          OptionValuePropertiesSP(new OptionValueProperties("interpreter"))),
+      Properties(std::make_shared<OptionValueProperties>("interpreter")),
       IOHandlerDelegate(IOHandlerDelegate::Completion::LLDBCommand),
       m_debugger(debugger), m_synchronous_execution(true),
       m_skip_lldbinit_files(false), m_skip_app_init_files(false),
diff --git a/lldb/source/Plugins/ExpressionParser/Clang/ClangASTImporter.h b/lldb/source/Plugins/ExpressionParser/Clang/ClangASTImporter.h
index 6231f0f..47b137a 100644
--- a/lldb/source/Plugins/ExpressionParser/Clang/ClangASTImporter.h
+++ b/lldb/source/Plugins/ExpressionParser/Clang/ClangASTImporter.h
@@ -225,7 +225,7 @@ public:
     ContextMetadataMap::iterator context_md_iter = m_metadata_map.find(dst_ctx);
 
     if (context_md_iter == m_metadata_map.end()) {
-      context_md = ASTContextMetadataSP(new ASTContextMetadata(dst_ctx));
+      context_md = std::make_shared<ASTContextMetadata>(dst_ctx);
       m_metadata_map[dst_ctx] = context_md;
     } else {
       context_md = context_md_iter->second;
@@ -438,7 +438,7 @@ public:
 
     if (context_md_iter == m_metadata_map.end()) {
       ASTContextMetadataSP context_md =
-          ASTContextMetadataSP(new ASTContextMetadata(dst_ctx));
+          std::make_shared<ASTContextMetadata>(dst_ctx);
       m_metadata_map[dst_ctx] = context_md;
       return context_md;
     }
@@ -462,7 +462,7 @@ public:
 
     if (delegate_iter == delegates.end()) {
       ImporterDelegateSP delegate =
-          ImporterDelegateSP(new ASTImporterDelegate(*this, dst_ctx, src_ctx));
+          std::make_shared<ASTImporterDelegate>(*this, dst_ctx, src_ctx);
       delegates[src_ctx] = delegate;
       return delegate;
     }
diff --git a/lldb/source/Plugins/Language/CPlusPlus/CPlusPlusLanguage.cpp b/lldb/source/Plugins/Language/CPlusPlus/CPlusPlusLanguage.cpp
index e69f2f6..46753c5 100644
--- a/lldb/source/Plugins/Language/CPlusPlus/CPlusPlusLanguage.cpp
+++ b/lldb/source/Plugins/Language/CPlusPlus/CPlusPlusLanguage.cpp
@@ -1363,6 +1363,28 @@ static void RegisterStdStringSummaryProvider(
       summary_sp);
 }
 
+static void RegisterStdStringViewSummaryProvider(
+    const lldb::TypeCategoryImplSP &category_sp, llvm::StringRef string_ty,
+    llvm::StringRef char_ty, lldb::TypeSummaryImplSP summary_sp) {
+  // std::string_view
+  category_sp->AddTypeSummary(
+      std::make_shared<lldb_private::TypeNameSpecifierImpl>(
+          string_ty, eFormatterMatchExact),
+      summary_sp);
+
+  // std::basic_string_view<char, std::char_traits<char>>
+  // NativePDB has spaces at different positions compared to PDB and DWARF, so
+  // use a regex and make them optional.
+  category_sp->AddTypeSummary(
+      std::make_shared<lldb_private::TypeNameSpecifierImpl>(
+          llvm::formatv(
+              "^std::basic_string_view<{0}, ?std::char_traits<{0}> ?>$",
+              char_ty)
+              .str(),
+          eFormatterMatchRegex),
+      summary_sp);
+}
+
 static void LoadLibStdcppFormatters(lldb::TypeCategoryImplSP cpp_category_sp) {
   if (!cpp_category_sp)
     return;
@@ -1863,6 +1885,36 @@ static void LoadMsvcStlFormatters(lldb::TypeCategoryImplSP cpp_category_sp) {
           MsvcStlStringSummaryProvider<StringElementType::UTF32>,
           "MSVC STL std::u32string summary provider"));
 
+  RegisterStdStringViewSummaryProvider(
+      cpp_category_sp, "std::string_view", "char",
+      std::make_shared<CXXFunctionSummaryFormat>(
+          stl_summary_flags,
+          MsvcStlStringViewSummaryProvider<StringElementType::ASCII>,
+          "MSVC STL std::string_view summary provider"));
+  RegisterStdStringViewSummaryProvider(
+      cpp_category_sp, "std::u8string_view", "char8_t",
+      std::make_shared<CXXFunctionSummaryFormat>(
+          stl_summary_flags,
+          MsvcStlStringViewSummaryProvider<StringElementType::UTF8>,
+          "MSVC STL std::u8string_view summary provider"));
+  RegisterStdStringViewSummaryProvider(
+      cpp_category_sp, "std::u16string_view", "char16_t",
+      std::make_shared<CXXFunctionSummaryFormat>(
+          stl_summary_flags,
+          MsvcStlStringViewSummaryProvider<StringElementType::UTF16>,
+          "MSVC STL std::u16string_view summary provider"));
+  RegisterStdStringViewSummaryProvider(
+      cpp_category_sp, "std::u32string_view", "char32_t",
+      std::make_shared<CXXFunctionSummaryFormat>(
+          stl_summary_flags,
+          MsvcStlStringViewSummaryProvider<StringElementType::UTF32>,
+          "MSVC STL std::u32string_view summary provider"));
+  RegisterStdStringViewSummaryProvider(
+      cpp_category_sp, "std::wstring_view", "wchar_t",
+      std::make_shared<CXXFunctionSummaryFormat>(
+          stl_summary_flags, MsvcStlWStringViewSummaryProvider,
+          "MSVC STL std::wstring_view summary provider"));
+
   stl_summary_flags.SetDontShowChildren(false);
 
   AddCXXSynthetic(cpp_category_sp, MsvcStlAtomicSyntheticFrontEndCreator,
diff --git a/lldb/source/Plugins/Language/CPlusPlus/MsvcStl.cpp b/lldb/source/Plugins/Language/CPlusPlus/MsvcStl.cpp
index 5d0f20b..8fb305b 100644
--- a/lldb/source/Plugins/Language/CPlusPlus/MsvcStl.cpp
+++ b/lldb/source/Plugins/Language/CPlusPlus/MsvcStl.cpp
@@ -112,6 +112,33 @@ static bool formatStringImpl(ValueObject &valobj, Stream &stream,
   return true;
 }
 
+template <StringPrinter::StringElementType element_type>
+static bool formatStringViewImpl(ValueObject &valobj, Stream &stream,
+                                 const TypeSummaryOptions &summary_options,
+                                 std::string prefix_token) {
+  auto data_sp = valobj.GetChildMemberWithName("_Mydata");
+  auto size_sp = valobj.GetChildMemberWithName("_Mysize");
+  if (!data_sp || !size_sp)
+    return false;
+
+  bool success = false;
+  uint64_t size = size_sp->GetValueAsUnsigned(0, &success);
+  if (!success) {
+    stream << "Summary Unavailable";
+    return true;
+  }
+
+  StreamString scratch_stream;
+  success = StringBufferSummaryProvider<element_type>(
+      scratch_stream, summary_options, data_sp, size, prefix_token);
+
+  if (success)
+    stream << scratch_stream.GetData();
+  else
+    stream << "Summary Unavailable";
+  return true;
+}
+
 bool lldb_private::formatters::IsMsvcStlStringType(ValueObject &valobj) {
   std::vector<uint32_t> indexes;
   return valobj.GetCompilerType().GetIndexOfChildMemberWithName("_Mypair", true,
@@ -153,3 +180,39 @@ bool lldb_private::formatters::MsvcStlStringSummaryProvider<
   return MsvcStlStringSummaryProviderImpl<StringElementType::UTF32>(
       valobj, stream, summary_options, "U");
 }
+
+bool lldb_private::formatters::MsvcStlWStringViewSummaryProvider(
+    ValueObject &valobj, Stream &stream,
+    const TypeSummaryOptions &summary_options) {
+  return formatStringViewImpl<StringElementType::UTF16>(valobj, stream,
+                                                        summary_options, "L");
+}
+
+template <>
+bool lldb_private::formatters::MsvcStlStringViewSummaryProvider<
+    StringElementType::ASCII>(ValueObject &valobj, Stream &stream,
+                              const TypeSummaryOptions &summary_options) {
+  return formatStringViewImpl<StringElementType::ASCII>(valobj, stream,
+                                                        summary_options, "");
+}
+template <>
+bool lldb_private::formatters::MsvcStlStringViewSummaryProvider<
+    StringElementType::UTF8>(ValueObject &valobj, Stream &stream,
+                             const TypeSummaryOptions &summary_options) {
+  return formatStringViewImpl<StringElementType::UTF8>(valobj, stream,
+                                                       summary_options, "u8");
+}
+template <>
+bool lldb_private::formatters::MsvcStlStringViewSummaryProvider<
+    StringElementType::UTF16>(ValueObject &valobj, Stream &stream,
+                              const TypeSummaryOptions &summary_options) {
+  return formatStringViewImpl<StringElementType::UTF16>(valobj, stream,
+                                                        summary_options, "u");
+}
+template <>
+bool lldb_private::formatters::MsvcStlStringViewSummaryProvider<
+    StringElementType::UTF32>(ValueObject &valobj, Stream &stream,
+                              const TypeSummaryOptions &summary_options) {
+  return formatStringViewImpl<StringElementType::UTF32>(valobj, stream,
+                                                        summary_options, "U");
+}
diff --git a/lldb/source/Plugins/Language/CPlusPlus/MsvcStl.h b/lldb/source/Plugins/Language/CPlusPlus/MsvcStl.h
index 490794c..8a49181 100644
--- a/lldb/source/Plugins/Language/CPlusPlus/MsvcStl.h
+++ b/lldb/source/Plugins/Language/CPlusPlus/MsvcStl.h
@@ -29,6 +29,15 @@ bool MsvcStlWStringSummaryProvider(
     ValueObject &valobj, Stream &stream,
     const TypeSummaryOptions &options); // VC 2015+ std::wstring
 
+template <StringPrinter::StringElementType element_type>
+bool MsvcStlStringViewSummaryProvider(
+    ValueObject &valobj, Stream &stream,
+    const TypeSummaryOptions &summary_options); // std::{u8,u16,u32}?string_view
+
+bool MsvcStlWStringViewSummaryProvider(
+    ValueObject &valobj, Stream &stream,
+    const TypeSummaryOptions &options); // std::wstring_view
+
 // MSVC STL std::shared_ptr<> and std::weak_ptr<>
 bool IsMsvcStlSmartPointer(ValueObject &valobj);
 bool MsvcStlSmartPointerSummaryProvider(ValueObject &valobj, Stream &stream,
diff --git a/lldb/source/Plugins/LanguageRuntime/ObjC/AppleObjCRuntime/AppleObjCRuntimeV2.cpp b/lldb/source/Plugins/LanguageRuntime/ObjC/AppleObjCRuntime/AppleObjCRuntimeV2.cpp
index 4fcdebe..cca721e 100644
--- a/lldb/source/Plugins/LanguageRuntime/ObjC/AppleObjCRuntime/AppleObjCRuntimeV2.cpp
+++ b/lldb/source/Plugins/LanguageRuntime/ObjC/AppleObjCRuntime/AppleObjCRuntimeV2.cpp
@@ -3462,7 +3462,7 @@ public:
         *exception, eValueTypeVariableArgument);
     exception = exception->GetDynamicValue(eDynamicDontRunTarget);
 
-    m_arguments = ValueObjectListSP(new ValueObjectList());
+    m_arguments = std::make_shared<ValueObjectList>();
     m_arguments->Append(exception);
 
     m_stop_desc = "hit Objective-C exception";
diff --git a/lldb/source/Plugins/Platform/MacOSX/PlatformDarwinKernel.cpp b/lldb/source/Plugins/Platform/MacOSX/PlatformDarwinKernel.cpp
index 54869001..07c5a52 100644
--- a/lldb/source/Plugins/Platform/MacOSX/PlatformDarwinKernel.cpp
+++ b/lldb/source/Plugins/Platform/MacOSX/PlatformDarwinKernel.cpp
@@ -805,7 +805,7 @@ Status PlatformDarwinKernel::GetSharedModuleKernel(
     if (FileSystem::Instance().Exists(possible_kernel)) {
       ModuleSpec kern_spec(possible_kernel);
       kern_spec.GetUUID() = module_spec.GetUUID();
-      module_sp.reset(new Module(kern_spec));
+      module_sp = std::make_shared<Module>(kern_spec);
       if (module_sp && module_sp->GetObjectFile() &&
           module_sp->MatchesModuleSpec(kern_spec)) {
         // The dSYM is next to the binary (that's the only
@@ -835,7 +835,7 @@ Status PlatformDarwinKernel::GetSharedModuleKernel(
       kern_spec.GetUUID() = module_spec.GetUUID();
       kern_spec.GetSymbolFileSpec() = possible_kernel_dsym;
 
-      module_sp.reset(new Module(kern_spec));
+      module_sp = std::make_shared<Module>(kern_spec);
       if (module_sp && module_sp->GetObjectFile() &&
           module_sp->MatchesModuleSpec(kern_spec)) {
         if (did_create_ptr)
diff --git a/lldb/source/Plugins/ScriptInterpreter/Python/ScriptInterpreterPython.cpp b/lldb/source/Plugins/ScriptInterpreter/Python/ScriptInterpreterPython.cpp
index 0c864dc..457182a 100644
--- a/lldb/source/Plugins/ScriptInterpreter/Python/ScriptInterpreterPython.cpp
+++ b/lldb/source/Plugins/ScriptInterpreter/Python/ScriptInterpreterPython.cpp
@@ -1496,7 +1496,7 @@ lldb::ValueObjectListSP ScriptInterpreterPythonImpl::GetRecognizedArguments(
   }
   if (py_return.get()) {
     PythonList result_list(PyRefType::Borrowed, py_return.get());
-    ValueObjectListSP result = ValueObjectListSP(new ValueObjectList());
+    ValueObjectListSP result = std::make_shared<ValueObjectList>();
     for (size_t i = 0; i < result_list.GetSize(); i++) {
       PyObject *item = result_list.GetItemAtIndex(i).get();
       lldb::SBValue *sb_value_ptr =
@@ -3047,7 +3047,7 @@ bool ScriptInterpreterPythonImpl::SetOptionValueForCommandObject(
 
   lldb::ExecutionContextRefSP exe_ctx_ref_sp;
   if (exe_ctx)
-    exe_ctx_ref_sp.reset(new ExecutionContextRef(exe_ctx));
+    exe_ctx_ref_sp = std::make_shared<ExecutionContextRef>(exe_ctx);
   PythonObject ctx_ref_obj = SWIGBridge::ToSWIGWrapper(exe_ctx_ref_sp);
 
   bool py_return = unwrapOrSetPythonException(As<bool>(
diff --git a/lldb/source/Plugins/StructuredData/DarwinLog/StructuredDataDarwinLog.cpp b/lldb/source/Plugins/StructuredData/DarwinLog/StructuredDataDarwinLog.cpp
index 82f18c5..867f6a6 100644
--- a/lldb/source/Plugins/StructuredData/DarwinLog/StructuredDataDarwinLog.cpp
+++ b/lldb/source/Plugins/StructuredData/DarwinLog/StructuredDataDarwinLog.cpp
@@ -574,8 +574,7 @@ public:
       return config_sp;
 
     // Handle source stream flags.
-    auto source_flags_sp =
-        StructuredData::DictionarySP(new StructuredData::Dictionary());
+    auto source_flags_sp = std::make_shared<StructuredData::Dictionary>();
     config_sp->AddItem("source-flags", source_flags_sp);
 
     source_flags_sp->AddBooleanItem("any-process", m_include_any_process);
@@ -591,8 +590,7 @@ public:
 
     // Handle filter rules
     if (!m_filter_rules.empty()) {
-      auto json_filter_rules_sp =
-          StructuredData::ArraySP(new StructuredData::Array);
+      auto json_filter_rules_sp = std::make_shared<StructuredData::Array>();
       config_sp->AddItem("filter-rules", json_filter_rules_sp);
       for (auto &rule_sp : m_filter_rules) {
         if (!rule_sp)
diff --git a/lldb/source/Plugins/SymbolLocator/DebugSymbols/SymbolLocatorDebugSymbols.cpp b/lldb/source/Plugins/SymbolLocator/DebugSymbols/SymbolLocatorDebugSymbols.cpp
index e5ba029..1e51dda 100644
--- a/lldb/source/Plugins/SymbolLocator/DebugSymbols/SymbolLocatorDebugSymbols.cpp
+++ b/lldb/source/Plugins/SymbolLocator/DebugSymbols/SymbolLocatorDebugSymbols.cpp
@@ -46,6 +46,7 @@
 #include <cstring>
 #include <dirent.h>
 #include <dlfcn.h>
+#include <memory>
 #include <optional>
 #include <pwd.h>
 
@@ -188,7 +189,7 @@ std::optional<ModuleSpec> SymbolLocatorDebugSymbols::LocateExecutableObjectFile(
             exe_spec.GetFileSpec() = module_spec.GetFileSpec();
             exe_spec.GetUUID() = module_spec.GetUUID();
             ModuleSP module_sp;
-            module_sp.reset(new Module(exe_spec));
+            module_sp = std::make_shared<Module>(exe_spec);
             if (module_sp && module_sp->GetObjectFile() &&
                 module_sp->MatchesModuleSpec(exe_spec)) {
               success = true;
@@ -630,7 +631,7 @@ static int LocateMacOSXFilesUsingDebugSymbols(const ModuleSpec &module_spec,
             exe_spec.GetFileSpec() = module_spec.GetFileSpec();
             exe_spec.GetUUID() = module_spec.GetUUID();
             ModuleSP module_sp;
-            module_sp.reset(new Module(exe_spec));
+            module_sp = std::make_shared<Module>(exe_spec);
             if (module_sp && module_sp->GetObjectFile() &&
                 module_sp->MatchesModuleSpec(exe_spec)) {
               success = true;
diff --git a/lldb/source/Plugins/SystemRuntime/MacOSX/AbortWithPayloadFrameRecognizer.cpp b/lldb/source/Plugins/SystemRuntime/MacOSX/AbortWithPayloadFrameRecognizer.cpp
index 7cf8750..aaa4ccd 100644
--- a/lldb/source/Plugins/SystemRuntime/MacOSX/AbortWithPayloadFrameRecognizer.cpp
+++ b/lldb/source/Plugins/SystemRuntime/MacOSX/AbortWithPayloadFrameRecognizer.cpp
@@ -128,7 +128,7 @@ AbortWithPayloadFrameRecognizer::RecognizeFrame(lldb::StackFrameSP frame_sp) {
 
   Value *cur_value;
   StackFrame *frame = frame_sp.get();
-  ValueObjectListSP arguments_sp = ValueObjectListSP(new ValueObjectList());
+  ValueObjectListSP arguments_sp = std::make_shared<ValueObjectList>();
 
   auto add_to_arguments = [&](llvm::StringRef name, Value *value,
                               bool dynamic) {
diff --git a/lldb/source/Symbol/Type.cpp b/lldb/source/Symbol/Type.cpp
index 0a886e5..b7adae4 100644
--- a/lldb/source/Symbol/Type.cpp
+++ b/lldb/source/Symbol/Type.cpp
@@ -9,6 +9,7 @@
 #include <algorithm>
 #include <cstdio>
 #include <iterator>
+#include <memory>
 #include <optional>
 
 #include "lldb/Core/Module.h"
@@ -246,7 +247,7 @@ public:
   TypeAppendVisitor(TypeListImpl &type_list) : m_type_list(type_list) {}
 
   bool operator()(const lldb::TypeSP &type) {
-    m_type_list.Append(TypeImplSP(new TypeImpl(type)));
+    m_type_list.Append(std::make_shared<TypeImpl>(type));
     return true;
   }
 
diff --git a/lldb/source/Target/InstrumentationRuntime.cpp b/lldb/source/Target/InstrumentationRuntime.cpp
index 9da06e8..d5a2e2e 100644
--- a/lldb/source/Target/InstrumentationRuntime.cpp
+++ b/lldb/source/Target/InstrumentationRuntime.cpp
@@ -73,5 +73,5 @@ void InstrumentationRuntime::ModulesDidLoad(
 lldb::ThreadCollectionSP
 InstrumentationRuntime::GetBacktracesFromExtendedStopInfo(
     StructuredData::ObjectSP info) {
-  return ThreadCollectionSP(new ThreadCollection());
+  return std::make_shared<ThreadCollection>();
 }
diff --git a/lldb/source/Target/StackFrameList.cpp b/lldb/source/Target/StackFrameList.cpp
index 16cd254..aedfc52 100644
--- a/lldb/source/Target/StackFrameList.cpp
+++ b/lldb/source/Target/StackFrameList.cpp
@@ -783,6 +783,8 @@ void StackFrameList::SelectMostRelevantFrame() {
 
 uint32_t
 StackFrameList::GetSelectedFrameIndex(SelectMostRelevant select_most_relevant) {
+  std::lock_guard<std::recursive_mutex> guard(m_selected_frame_mutex);
+
   if (!m_selected_frame_idx && select_most_relevant)
     SelectMostRelevantFrame();
   if (!m_selected_frame_idx) {
@@ -798,6 +800,8 @@ StackFrameList::GetSelectedFrameIndex(SelectMostRelevant select_most_relevant) {
 
 uint32_t StackFrameList::SetSelectedFrame(lldb_private::StackFrame *frame) {
   std::shared_lock<std::shared_mutex> guard(m_list_mutex);
+  std::lock_guard<std::recursive_mutex> selected_frame_guard(
+      m_selected_frame_mutex);
 
   const_iterator pos;
   const_iterator begin = m_frames.begin();
@@ -851,6 +855,8 @@ void StackFrameList::Clear() {
   std::unique_lock<std::shared_mutex> guard(m_list_mutex);
   m_frames.clear();
   m_concrete_frames_fetched = 0;
+  std::lock_guard<std::recursive_mutex> selected_frame_guard(
+      m_selected_frame_mutex);
   m_selected_frame_idx.reset();
 }
 
diff --git a/lldb/source/Target/StackFrameRecognizer.cpp b/lldb/source/Target/StackFrameRecognizer.cpp
index d23c1fa..9d5116c 100644
--- a/lldb/source/Target/StackFrameRecognizer.cpp
+++ b/lldb/source/Target/StackFrameRecognizer.cpp
@@ -41,7 +41,7 @@ ScriptedStackFrameRecognizer::RecognizeFrame(lldb::StackFrameSP frame) {
 
   ValueObjectListSP args =
       m_interpreter->GetRecognizedArguments(m_python_object_sp, frame);
-  auto args_synthesized = ValueObjectListSP(new ValueObjectList());
+  auto args_synthesized = std::make_shared<ValueObjectList>();
   if (args) {
     for (const auto &o : args->GetObjects())
       args_synthesized->Append(ValueObjectRecognizerSynthesizedValue::Create(
diff --git a/lldb/source/Target/ThreadPlanStepRange.cpp b/lldb/source/Target/ThreadPlanStepRange.cpp
index 78e1270..dca96cc 100644
--- a/lldb/source/Target/ThreadPlanStepRange.cpp
+++ b/lldb/source/Target/ThreadPlanStepRange.cpp
@@ -428,8 +428,8 @@ bool ThreadPlanStepRange::SetNextBranchBreakpoint() {
             top_most_line_entry.line = call_site.GetLine();
             top_most_line_entry.column = call_site.GetColumn();
             FileSpec call_site_file_spec = call_site.GetFile();
-            top_most_line_entry.original_file_sp.reset(
-                new SupportFile(call_site_file_spec));
+            top_most_line_entry.original_file_sp =
+                std::make_shared<SupportFile>(call_site_file_spec);
             top_most_line_entry.range = range;
             top_most_line_entry.file_sp.reset();
             top_most_line_entry.ApplyFileMappings(
diff --git a/lldb/test/API/functionalities/data-formatter/data-formatter-stl/generic/string_view/TestDataFormatterStdStringView.py b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/generic/string_view/TestDataFormatterStdStringView.py
index 8c053f0..1811418 100644
--- a/lldb/test/API/functionalities/data-formatter/data-formatter-stl/generic/string_view/TestDataFormatterStdStringView.py
+++ b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/generic/string_view/TestDataFormatterStdStringView.py
@@ -78,9 +78,6 @@ class StdStringViewDataFormatterTestCase(TestBase):
             "u32_string", type="std::u32string_view", summary='U"🍄🍅🍆🍌"'
         )
         self.expect_var_path("u32_empty", type="std::u32string_view", summary='U""')
-        self.expect_var_path(
-            "oops", type="std::string_view", summary='"Hellooo World\\n"'
-        )
 
         # GetSummary returns None so can't be checked by expect_var_path, so we
         # use the str representation instead
@@ -163,3 +160,8 @@ class StdStringViewDataFormatterTestCase(TestBase):
     def test_libcxx(self):
         self.build(dictionary={"USE_LIBCPP": 1})
         self.do_test()
+
+    @add_test_categories(["msvcstl"])
+    def test_msvcstl(self):
+        self.build()
+        self.do_test()
diff --git a/lldb/test/API/functionalities/data-formatter/data-formatter-stl/generic/string_view/main.cpp b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/generic/string_view/main.cpp
index 1e164786b..e8781fb 100644
--- a/lldb/test/API/functionalities/data-formatter/data-formatter-stl/generic/string_view/main.cpp
+++ b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/generic/string_view/main.cpp
@@ -95,7 +95,6 @@ int main() {
   std::string_view *null_str = nullptr;
 
   std::string hello = "Hellooo ";
-  std::string_view oops = hello + "World\n";
 
   q_source[0] = 'H'; // Set break point at this line.
 
diff --git a/lldb/test/API/functionalities/data-formatter/data-formatter-stl/generic/u8string_view/Makefile b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/generic/u8string_view/Makefile
new file mode 100644
index 0000000..4f79c0a
--- /dev/null
+++ b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/generic/u8string_view/Makefile
@@ -0,0 +1,4 @@
+CXX_SOURCES := main.cpp
+CXXFLAGS_EXTRAS := -std=c++20
+
+include Makefile.rules
diff --git a/lldb/test/API/functionalities/data-formatter/data-formatter-stl/generic/u8string_view/TestDataFormatterStdU8StringView.py b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/generic/u8string_view/TestDataFormatterStdU8StringView.py
new file mode 100644
index 0000000..1e35a0f
--- /dev/null
+++ b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/generic/u8string_view/TestDataFormatterStdU8StringView.py
@@ -0,0 +1,44 @@
+# coding=utf8
+"""
+Test std::u8string_view summary.
+"""
+
+
+import lldb
+from lldbsuite.test.decorators import *
+from lldbsuite.test.lldbtest import *
+from lldbsuite.test import lldbutil
+
+
+class StdU8StringViewDataFormatterTestCase(TestBase):
+    def do_test(self):
+        lldbutil.run_to_source_breakpoint(
+            self, "Set break point at this line.", lldb.SBFileSpec("main.cpp")
+        )
+
+        self.expect(
+            "frame variable",
+            substrs=[
+                '(std::u8string_view) u8_string_small = u8"🍄"',
+                '(std::u8string_view) u8_string = u8"❤️👍📄📁😃🧑‍🌾"',
+                '(std::u8string_view) u8_empty = u8""',
+                '(std::u8string_view) u8_text = u8"ABCd"',
+            ],
+        )
+
+    @expectedFailureAll(bugnumber="No libc++ formatters for std::u8string_view yet.")
+    @add_test_categories(["libc++"])
+    def test_libcxx(self):
+        self.build(dictionary={"USE_LIBCPP": 1})
+        self.do_test()
+
+    @expectedFailureAll(bugnumber="No libstdc++ formatters for std::u8string_view yet.")
+    @add_test_categories(["libstdcxx"])
+    def test_libstdcxx(self):
+        self.build(dictionary={"USE_LIBSTDCPP": 1})
+        self.do_test()
+
+    @add_test_categories(["msvcstl"])
+    def test_msvc(self):
+        self.build()
+        self.do_test()
diff --git a/lldb/test/API/functionalities/data-formatter/data-formatter-stl/generic/u8string_view/main.cpp b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/generic/u8string_view/main.cpp
new file mode 100644
index 0000000..458e783
--- /dev/null
+++ b/lldb/test/API/functionalities/data-formatter/data-formatter-stl/generic/u8string_view/main.cpp
@@ -0,0 +1,12 @@
+#include <cstdio>
+#include <string_view>
+
+int main() {
+  std::u8string_view u8_string_small(u8"🍄");
+  std::u8string_view u8_string(u8"❤️👍📄📁😃🧑‍🌾");
+  std::u8string_view u8_empty(u8"");
+  std::u8string_view u8_text(u8"ABC");
+  u8_text = u8"ABCd";
+
+  std::puts("// Set break point at this line.");
+}
diff --git a/lldb/tools/debugserver/source/RNBRemote.cpp b/lldb/tools/debugserver/source/RNBRemote.cpp
index 8be384c..102b2ab 100644
--- a/lldb/tools/debugserver/source/RNBRemote.cpp
+++ b/lldb/tools/debugserver/source/RNBRemote.cpp
@@ -21,6 +21,7 @@
 #include <mach/exception_types.h>
 #include <mach/mach_vm.h>
 #include <mach/task_info.h>
+#include <memory>
 #include <pwd.h>
 #include <string>
 #include <sys/stat.h>
@@ -5410,9 +5411,8 @@ RNBRemote::GetJSONThreadsInfo(bool threads_with_valid_stop_info_only) {
             JSONGenerator::ArraySP medata_array_sp(new JSONGenerator::Array());
             for (nub_size_t i = 0;
                  i < tid_stop_info.details.exception.data_count; ++i) {
-              medata_array_sp->AddItem(
-                  JSONGenerator::IntegerSP(new JSONGenerator::Integer(
-                      tid_stop_info.details.exception.data[i])));
+              medata_array_sp->AddItem(std::make_shared<JSONGenerator::Integer>(
+                  tid_stop_info.details.exception.data[i]));
             }
             thread_dict_sp->AddItem("medata", medata_array_sp);
           }
diff --git a/llvm/CMakeLists.txt b/llvm/CMakeLists.txt
index 903461b..e874d76 100644
--- a/llvm/CMakeLists.txt
+++ b/llvm/CMakeLists.txt
@@ -1190,6 +1190,8 @@ if (LLVM_HAVE_TFLITE)
   find_package(tensorflow-lite REQUIRED)
 endif()
 
+set(LLVM_ENABLE_PROFCHECK OFF CACHE BOOL "Enable profile checking in test tools")
+
 # For up-to-date instructions for installing the Tensorflow dependency, refer to
 # the bot setup script: https://github.com/google/ml-compiler-opt/blob/main/buildbot/buildbot_init.sh
 # Specifically, assuming python3 is installed:
diff --git a/llvm/cmake/modules/LLVMProcessSources.cmake b/llvm/cmake/modules/LLVMProcessSources.cmake
index 0670d60..a7f9517 100644
--- a/llvm/cmake/modules/LLVMProcessSources.cmake
+++ b/llvm/cmake/modules/LLVMProcessSources.cmake
@@ -58,6 +58,15 @@ function(llvm_process_sources OUT_VAR)
   set(sources ${ARG_UNPARSED_ARGUMENTS})
   llvm_check_source_file_list(${sources})
 
+  foreach(fn ${sources})
+    get_filename_component(suf ${fn} EXT)
+    if("${suf}" STREQUAL ".cpp" OR "${suf}" STREQUAL ".c")
+      get_filename_component(short_name ${fn} NAME)
+      set_source_files_properties(${fn} PROPERTIES COMPILE_DEFINITIONS "__SHORT_FILE__=\"${short_name}\"")
+    endif()
+  endforeach()
+
+
   # This adds .td and .h files to the Visual Studio solution:
   add_td_sources(sources)
   find_all_header_files(hdrs "${ARG_ADDITIONAL_HEADER_DIRS}")
diff --git a/llvm/docs/GettingStarted.rst b/llvm/docs/GettingStarted.rst
index 3036dae..e4dbb64b 100644
--- a/llvm/docs/GettingStarted.rst
+++ b/llvm/docs/GettingStarted.rst
@@ -240,8 +240,10 @@ Linux              x86\ :sup:`1`         GCC, Clang
 Linux              amd64                 GCC, Clang
 Linux              ARM                   GCC, Clang
 Linux              AArch64               GCC, Clang
+Linux              LoongArch             GCC, Clang
 Linux              Mips                  GCC, Clang
 Linux              PowerPC               GCC, Clang
+Linux              RISC-V                GCC, Clang
 Linux              SystemZ               GCC, Clang
 Solaris            V9 (Ultrasparc)       GCC
 DragonFlyBSD       amd64                 GCC, Clang
diff --git a/llvm/docs/ProgrammersManual.rst b/llvm/docs/ProgrammersManual.rst
index 68490c8..9ddeebd 100644
--- a/llvm/docs/ProgrammersManual.rst
+++ b/llvm/docs/ProgrammersManual.rst
@@ -932,7 +932,7 @@ In some contexts, certain types of errors are known to be benign. For example,
 when walking an archive, some clients may be happy to skip over badly formatted
 object files rather than terminating the walk immediately. Skipping badly
 formatted objects could be achieved using an elaborate handler method, but the
-Error.h header provides two utilities that make this idiom much cleaner: the
+``Error.h`` header provides two utilities that make this idiom much cleaner: the
 type inspection method, ``isA``, and the ``consumeError`` function:
 
 .. code-block:: c++
@@ -1073,7 +1073,7 @@ relatively natural use of C++ iterator/loop idioms.
 .. _function_apis:
 
 More information on Error and its related utilities can be found in the
-Error.h header file.
+``Error.h`` header file.
 
 Passing functions and other callable objects
 --------------------------------------------
@@ -1224,7 +1224,7 @@ Then you can run your pass like this:
 
 Of course, in practice, you should only set ``DEBUG_TYPE`` at the top of a file,
 to specify the debug type for the entire module. Be careful that you only do
-this after including Debug.h and not around any #include of headers. Also, you
+this after including ``Debug.h`` and not around any #include of headers. Also, you
 should use names more meaningful than "foo" and "bar", because there is no
 system in place to ensure that names do not conflict. If two different modules
 use the same string, they will all be turned on when the name is specified.
@@ -1579,18 +1579,18 @@ llvm/ADT/SmallVector.h
 ``SmallVector<Type, N>`` is a simple class that looks and smells just like
 ``vector<Type>``: it supports efficient iteration, lays out elements in memory
 order (so you can do pointer arithmetic between elements), supports efficient
-push_back/pop_back operations, supports efficient random access to its elements,
+``push_back``/``pop_back`` operations, supports efficient random access to its elements,
 etc.
 
-The main advantage of SmallVector is that it allocates space for some number of
-elements (N) **in the object itself**.  Because of this, if the SmallVector is
+The main advantage of ``SmallVector`` is that it allocates space for some number of
+elements (N) **in the object itself**.  Because of this, if the ``SmallVector`` is
 dynamically smaller than N, no malloc is performed.  This can be a big win in
 cases where the malloc/free call is far more expensive than the code that
 fiddles around with the elements.
 
 This is good for vectors that are "usually small" (e.g. the number of
 predecessors/successors of a block is usually less than 8).  On the other hand,
-this makes the size of the SmallVector itself large, so you don't want to
+this makes the size of the ``SmallVector`` itself large, so you don't want to
 allocate lots of them (doing so will waste a lot of space).  As such,
 SmallVectors are most useful when on the stack.
 
@@ -1600,21 +1600,21 @@ omitting the ``N``). This will choose a default number of
 inlined elements reasonable for allocation on the stack (for example, trying
 to keep ``sizeof(SmallVector<T>)`` around 64 bytes).
 
-SmallVector also provides a nice portable and efficient replacement for
+``SmallVector`` also provides a nice portable and efficient replacement for
 ``alloca``.
 
-SmallVector has grown a few other minor advantages over std::vector, causing
+``SmallVector`` has grown a few other minor advantages over ``std::vector``, causing
 ``SmallVector<Type, 0>`` to be preferred over ``std::vector<Type>``.
 
-#. std::vector is exception-safe, and some implementations have pessimizations
-   that copy elements when SmallVector would move them.
+#. ``std::vector`` is exception-safe, and some implementations have pessimizations
+   that copy elements when ``SmallVector`` would move them.
 
-#. SmallVector understands ``std::is_trivially_copyable<Type>`` and uses realloc aggressively.
+#. ``SmallVector`` understands ``std::is_trivially_copyable<Type>`` and uses realloc aggressively.
 
-#. Many LLVM APIs take a SmallVectorImpl as an out parameter (see the note
+#. Many LLVM APIs take a ``SmallVectorImpl`` as an out parameter (see the note
    below).
 
-#. SmallVector with N equal to 0 is smaller than std::vector on 64-bit
+#. ``SmallVector`` with N equal to 0 is smaller than ``std::vector`` on 64-bit
    platforms, since it uses ``unsigned`` (instead of ``void*``) for its size
    and capacity.
 
@@ -1698,11 +1698,11 @@ non-ordered manner.
 ^^^^^^^^
 
 ``std::vector<T>`` is well loved and respected.  However, ``SmallVector<T, 0>``
-is often a better option due to the advantages listed above.  std::vector is
+is often a better option due to the advantages listed above.  ``std::vector`` is
 still useful when you need to store more than ``UINT32_MAX`` elements or when
 interfacing with code that expects vectors :).
 
-One worthwhile note about std::vector: avoid code like this:
+One worthwhile note about ``std::vector``: avoid code like this:
 
 .. code-block:: c++
 
@@ -1749,10 +1749,10 @@ extremely high constant factor, particularly for small data types.
 ``std::list`` also only supports bidirectional iteration, not random access
 iteration.
 
-In exchange for this high cost, std::list supports efficient access to both ends
+In exchange for this high cost, ``std::list`` supports efficient access to both ends
 of the list (like ``std::deque``, but unlike ``std::vector`` or
 ``SmallVector``).  In addition, the iterator invalidation characteristics of
-std::list are stronger than that of a vector class: inserting or removing an
+``std::list`` are stronger than that of a vector class: inserting or removing an
 element into the list does not invalidate iterator or pointers to other elements
 in the list.
 
@@ -1895,7 +1895,7 @@ Note that it is generally preferred to *not* pass strings around as ``const
 char*``'s.  These have a number of problems, including the fact that they
 cannot represent embedded nul ("\0") characters, and do not have a length
 available efficiently.  The general replacement for '``const char*``' is
-StringRef.
+``StringRef``.
 
 For more information on choosing string containers for APIs, please see
 :ref:`Passing Strings <string_apis>`.
@@ -1905,41 +1905,41 @@ For more information on choosing string containers for APIs, please see
 llvm/ADT/StringRef.h
 ^^^^^^^^^^^^^^^^^^^^
 
-The StringRef class is a simple value class that contains a pointer to a
+The ``StringRef`` class is a simple value class that contains a pointer to a
 character and a length, and is quite related to the :ref:`ArrayRef
 <dss_arrayref>` class (but specialized for arrays of characters).  Because
-StringRef carries a length with it, it safely handles strings with embedded nul
+``StringRef`` carries a length with it, it safely handles strings with embedded nul
 characters in it, getting the length does not require a strlen call, and it even
 has very convenient APIs for slicing and dicing the character range that it
 represents.
 
-StringRef is ideal for passing simple strings around that are known to be live,
-either because they are C string literals, std::string, a C array, or a
-SmallVector.  Each of these cases has an efficient implicit conversion to
-StringRef, which doesn't result in a dynamic strlen being executed.
+``StringRef`` is ideal for passing simple strings around that are known to be live,
+either because they are C string literals, ``std::string``, a C array, or a
+``SmallVector``.  Each of these cases has an efficient implicit conversion to
+``StringRef``, which doesn't result in a dynamic ``strlen`` being executed.
 
-StringRef has a few major limitations which make more powerful string containers
+``StringRef`` has a few major limitations which make more powerful string containers
 useful:
 
-#. You cannot directly convert a StringRef to a 'const char*' because there is
-   no way to add a trailing nul (unlike the .c_str() method on various stronger
+#. You cannot directly convert a ``StringRef`` to a 'const char*' because there is
+   no way to add a trailing nul (unlike the ``.c_str()`` method on various stronger
    classes).
 
-#. StringRef doesn't own or keep alive the underlying string bytes.
+#. ``StringRef`` doesn't own or keep alive the underlying string bytes.
    As such it can easily lead to dangling pointers, and is not suitable for
-   embedding in datastructures in most cases (instead, use an std::string or
+   embedding in datastructures in most cases (instead, use an ``std::string`` or
    something like that).
 
-#. For the same reason, StringRef cannot be used as the return value of a
-   method if the method "computes" the result string.  Instead, use std::string.
+#. For the same reason, ``StringRef`` cannot be used as the return value of a
+   method if the method "computes" the result string.  Instead, use ``std::string``.
 
-#. StringRef's do not allow you to mutate the pointed-to string bytes and it
+#. ``StringRef``'s do not allow you to mutate the pointed-to string bytes and it
    doesn't allow you to insert or remove bytes from the range.  For editing
    operations like this, it interoperates with the :ref:`Twine <dss_twine>`
    class.
 
 Because of its strengths and limitations, it is very common for a function to
-take a StringRef and for a method on an object to return a StringRef that points
+take a ``StringRef`` and for a method on an object to return a ``StringRef`` that points
 into some string that it owns.
 
 .. _dss_twine:
@@ -1979,25 +1979,25 @@ behavior and will probably crash:
   const Twine &Tmp = X + "." + Twine(i);
   foo(Tmp);
 
-... because the temporaries are destroyed before the call.  That said, Twine's
-are much more efficient than intermediate std::string temporaries, and they work
-really well with StringRef.  Just be aware of their limitations.
+... because the temporaries are destroyed before the call.  That said, ``Twine``'s
+are much more efficient than intermediate ``std::string`` temporaries, and they work
+really well with ``StringRef``.  Just be aware of their limitations.
 
 .. _dss_smallstring:
 
 llvm/ADT/SmallString.h
 ^^^^^^^^^^^^^^^^^^^^^^
 
-SmallString is a subclass of :ref:`SmallVector <dss_smallvector>` that adds some
-convenience APIs like += that takes StringRef's.  SmallString avoids allocating
+``SmallString`` is a subclass of :ref:`SmallVector <dss_smallvector>` that adds some
+convenience APIs like += that takes ``StringRef``'s.  ``SmallString`` avoids allocating
 memory in the case when the preallocated space is enough to hold its data, and
 it calls back to general heap allocation when required.  Since it owns its data,
 it is very safe to use and supports full mutation of the string.
 
-Like SmallVector's, the big downside to SmallString is their sizeof.  While they
+Like ``SmallVector``'s, the big downside to ``SmallString`` is their sizeof.  While they
 are optimized for small strings, they themselves are not particularly small.
 This means that they work great for temporary scratch buffers on the stack, but
-should not generally be put into the heap: it is very rare to see a SmallString
+should not generally be put into the heap: it is very rare to see a ``SmallString``
 as the member of a frequently-allocated heap data structure or returned
 by-value.
 
@@ -2006,18 +2006,18 @@ by-value.
 std::string
 ^^^^^^^^^^^
 
-The standard C++ std::string class is a very general class that (like
-SmallString) owns its underlying data.  sizeof(std::string) is very reasonable
+The standard C++ ``std::string`` class is a very general class that (like
+``SmallString``) owns its underlying data.  sizeof(std::string) is very reasonable
 so it can be embedded into heap data structures and returned by-value.  On the
-other hand, std::string is highly inefficient for inline editing (e.g.
+other hand, ``std::string`` is highly inefficient for inline editing (e.g.
 concatenating a bunch of stuff together) and because it is provided by the
 standard library, its performance characteristics depend a lot of the host
 standard library (e.g. libc++ and MSVC provide a highly optimized string class,
 GCC contains a really slow implementation).
 
-The major disadvantage of std::string is that almost every operation that makes
+The major disadvantage of ``std::string`` is that almost every operation that makes
 them larger can allocate memory, which is slow.  As such, it is better to use
-SmallVector or Twine as a scratch buffer, but then use std::string to persist
+``SmallVector`` or ``Twine`` as a scratch buffer, but then use ``std::string`` to persist
 the result.
 
 .. _ds_set:
@@ -2035,8 +2035,8 @@ A sorted 'vector'
 ^^^^^^^^^^^^^^^^^
 
 If you intend to insert a lot of elements, then do a lot of queries, a great
-approach is to use an std::vector (or other sequential container) with
-std::sort+std::unique to remove duplicates.  This approach works really well if
+approach is to use an ``std::vector`` (or other sequential container) with
+``std::sort``+``std::unique`` to remove duplicates.  This approach works really well if
 your usage pattern has these two distinct phases (insert then query), and can be
 coupled with a good choice of :ref:`sequential container <ds_sequential>`.
 
@@ -2102,11 +2102,11 @@ copy-construction, which :ref:`SmallSet <dss_smallset>` and :ref:`SmallPtrSet
 llvm/ADT/DenseSet.h
 ^^^^^^^^^^^^^^^^^^^
 
-DenseSet is a simple quadratically probed hash table.  It excels at supporting
+``DenseSet`` is a simple quadratically probed hash table.  It excels at supporting
 small values: it uses a single allocation to hold all of the pairs that are
-currently inserted in the set.  DenseSet is a great way to unique small values
+currently inserted in the set.  ``DenseSet`` is a great way to unique small values
 that are not simple pointers (use :ref:`SmallPtrSet <dss_smallptrset>` for
-pointers).  Note that DenseSet has the same requirements for the value type that
+pointers).  Note that ``DenseSet`` has the same requirements for the value type that
 :ref:`DenseMap <dss_densemap>` has.
 
 .. _dss_sparseset:
@@ -2128,12 +2128,12 @@ data structures.
 llvm/ADT/SparseMultiSet.h
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 
-SparseMultiSet adds multiset behavior to SparseSet, while retaining SparseSet's
-desirable attributes. Like SparseSet, it typically uses a lot of memory, but
+``SparseMultiSet`` adds multiset behavior to ``SparseSet``, while retaining ``SparseSet``'s
+desirable attributes. Like ``SparseSet``, it typically uses a lot of memory, but
 provides operations that are almost as fast as a vector.  Typical keys are
 physical registers, virtual registers, or numbered basic blocks.
 
-SparseMultiSet is useful for algorithms that need very fast
+``SparseMultiSet`` is useful for algorithms that need very fast
 clear/find/insert/erase of the entire collection, and iteration over sets of
 elements sharing a key. It is often a more efficient choice than using composite
 data structures (e.g. vector-of-vectors, map-of-vectors). It is not intended for
@@ -2144,10 +2144,10 @@ building composite data structures.
 llvm/ADT/FoldingSet.h
 ^^^^^^^^^^^^^^^^^^^^^
 
-FoldingSet is an aggregate class that is really good at uniquing
+``FoldingSet`` is an aggregate class that is really good at uniquing
 expensive-to-create or polymorphic objects.  It is a combination of a chained
 hash table with intrusive links (uniqued objects are required to inherit from
-FoldingSetNode) that uses :ref:`SmallVector <dss_smallvector>` as part of its ID
+``FoldingSetNode``) that uses :ref:`SmallVector <dss_smallvector>` as part of its ID
 process.
 
 Consider a case where you want to implement a "getOrCreateFoo" method for a
@@ -2157,14 +2157,14 @@ operands), but we don't want to 'new' a node, then try inserting it into a set
 only to find out it already exists, at which point we would have to delete it
 and return the node that already exists.
 
-To support this style of client, FoldingSet perform a query with a
-FoldingSetNodeID (which wraps SmallVector) that can be used to describe the
+To support this style of client, ``FoldingSet`` perform a query with a
+``FoldingSetNodeID`` (which wraps ``SmallVector``) that can be used to describe the
 element that we want to query for.  The query either returns the element
 matching the ID or it returns an opaque ID that indicates where insertion should
 take place.  Construction of the ID usually does not require heap traffic.
 
-Because FoldingSet uses intrusive links, it can support polymorphic objects in
-the set (for example, you can have SDNode instances mixed with LoadSDNodes).
+Because ``FoldingSet`` uses intrusive links, it can support polymorphic objects in
+the set (for example, you can have ``SDNode`` instances mixed with ``LoadSDNodes``).
 Because the elements are individually allocated, pointers to the elements are
 stable: inserting or removing elements does not invalidate any pointers to other
 elements.
@@ -2175,7 +2175,7 @@ elements.
 ^^^^^
 
 ``std::set`` is a reasonable all-around set class, which is decent at many
-things but great at nothing.  std::set allocates memory for each element
+things but great at nothing.  ``std::set`` allocates memory for each element
 inserted (thus it is very malloc intensive) and typically stores three pointers
 per element in the set (thus adding a large amount of per-element space
 overhead).  It offers guaranteed log(n) performance, which is not particularly
@@ -2183,12 +2183,12 @@ fast from a complexity standpoint (particularly if the elements of the set are
 expensive to compare, like strings), and has extremely high constant factors for
 lookup, insertion and removal.
 
-The advantages of std::set are that its iterators are stable (deleting or
+The advantages of ``std::set`` are that its iterators are stable (deleting or
 inserting an element from the set does not affect iterators or pointers to other
 elements) and that iteration over the set is guaranteed to be in sorted order.
 If the elements in the set are large, then the relative overhead of the pointers
 and malloc traffic is not a big deal, but if the elements of the set are small,
-std::set is almost never a good choice.
+``std::set`` is almost never a good choice.
 
 .. _dss_setvector:
 
@@ -2242,11 +2242,11 @@ produces a lot of malloc traffic.  It should be avoided.
 llvm/ADT/ImmutableSet.h
 ^^^^^^^^^^^^^^^^^^^^^^^
 
-ImmutableSet is an immutable (functional) set implementation based on an AVL
+``ImmutableSet`` is an immutable (functional) set implementation based on an AVL
 tree.  Adding or removing elements is done through a Factory object and results
-in the creation of a new ImmutableSet object.  If an ImmutableSet already exists
+in the creation of a new ``ImmutableSet`` object.  If an ``ImmutableSet`` already exists
 with the given contents, then the existing one is returned; equality is compared
-with a FoldingSetNodeID.  The time and space complexity of add or remove
+with a ``FoldingSetNodeID``.  The time and space complexity of add or remove
 operations is logarithmic in the size of the original set.
 
 There is no method for returning an element of the set, you can only check for
@@ -2257,11 +2257,11 @@ membership.
 Other Set-Like Container Options
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 
-The STL provides several other options, such as std::multiset and
-std::unordered_set.  We never use containers like unordered_set because
+The STL provides several other options, such as ``std::multiset`` and
+``std::unordered_set``.  We never use containers like ``unordered_set`` because
 they are generally very expensive (each insertion requires a malloc).
 
-std::multiset is useful if you're not interested in elimination of duplicates,
+``std::multiset`` is useful if you're not interested in elimination of duplicates,
 but has all the drawbacks of :ref:`std::set <dss_set>`.  A sorted vector
 (where you don't delete duplicate entries) or some other approach is almost
 always better.
@@ -2282,7 +2282,7 @@ A sorted 'vector'
 If your usage pattern follows a strict insert-then-query approach, you can
 trivially use the same approach as :ref:`sorted vectors for set-like containers
 <dss_sortedvectorset>`.  The only difference is that your query function (which
-uses std::lower_bound to get efficient log(n) lookup) should only compare the
+uses ``std::lower_bound`` to get efficient log(n) lookup) should only compare the
 key, not both the key and value.  This yields the same advantages as sorted
 vectors for sets.
 
@@ -2293,11 +2293,11 @@ llvm/ADT/StringMap.h
 
 Strings are commonly used as keys in maps, and they are difficult to support
 efficiently: they are variable length, inefficient to hash and compare when
-long, expensive to copy, etc.  StringMap is a specialized container designed to
+long, expensive to copy, etc.  ``StringMap`` is a specialized container designed to
 cope with these issues.  It supports mapping an arbitrary range of bytes to an
 arbitrary other object.
 
-The StringMap implementation uses a quadratically-probed hash table, where the
+The ``StringMap`` implementation uses a quadratically-probed hash table, where the
 buckets store a pointer to the heap allocated entries (and some other stuff).
 The entries in the map must be heap allocated because the strings are variable
 length.  The string data (key) and the element object (value) are stored in the
@@ -2305,26 +2305,26 @@ same allocation with the string data immediately after the element object.
 This container guarantees the "``(char*)(&Value+1)``" points to the key string
 for a value.
 
-The StringMap is very fast for several reasons: quadratic probing is very cache
+The ``StringMap`` is very fast for several reasons: quadratic probing is very cache
 efficient for lookups, the hash value of strings in buckets is not recomputed
-when looking up an element, StringMap rarely has to touch the memory for
+when looking up an element, ``StringMap`` rarely has to touch the memory for
 unrelated objects when looking up a value (even when hash collisions happen),
 hash table growth does not recompute the hash values for strings already in the
 table, and each pair in the map is store in a single allocation (the string data
 is stored in the same allocation as the Value of a pair).
 
-StringMap also provides query methods that take byte ranges, so it only ever
+``StringMap`` also provides query methods that take byte ranges, so it only ever
 copies a string if a value is inserted into the table.
 
-StringMap iteration order, however, is not guaranteed to be deterministic, so
-any uses which require that should instead use a std::map.
+``StringMap`` iteration order, however, is not guaranteed to be deterministic, so
+any uses which require that should instead use a ``std::map``.
 
 .. _dss_indexmap:
 
 llvm/ADT/IndexedMap.h
 ^^^^^^^^^^^^^^^^^^^^^
 
-IndexedMap is a specialized container for mapping small dense integers (or
+``IndexedMap`` is a specialized container for mapping small dense integers (or
 values that can be mapped to small dense integers) to some other type.  It is
 internally implemented as a vector with a mapping function that maps the keys
 to the dense integer range.
@@ -2338,27 +2338,27 @@ virtual register ID).
 llvm/ADT/DenseMap.h
 ^^^^^^^^^^^^^^^^^^^
 
-DenseMap is a simple quadratically probed hash table.  It excels at supporting
+``DenseMap`` is a simple quadratically probed hash table.  It excels at supporting
 small keys and values: it uses a single allocation to hold all of the pairs
-that are currently inserted in the map.  DenseMap is a great way to map
+that are currently inserted in the map.  ``DenseMap`` is a great way to map
 pointers to pointers, or map other small types to each other.
 
-There are several aspects of DenseMap that you should be aware of, however.
-The iterators in a DenseMap are invalidated whenever an insertion occurs,
-unlike map.  Also, because DenseMap allocates space for a large number of
+There are several aspects of ``DenseMap`` that you should be aware of, however.
+The iterators in a ``DenseMap`` are invalidated whenever an insertion occurs,
+unlike ``map``.  Also, because ``DenseMap`` allocates space for a large number of
 key/value pairs (it starts with 64 by default), it will waste a lot of space if
 your keys or values are large.  Finally, you must implement a partial
-specialization of DenseMapInfo for the key that you want, if it isn't already
-supported.  This is required to tell DenseMap about two special marker values
+specialization of ``DenseMapInfo`` for the key that you want, if it isn't already
+supported.  This is required to tell ``DenseMap`` about two special marker values
 (which can never be inserted into the map) that it needs internally.
 
-DenseMap's find_as() method supports lookup operations using an alternate key
+``DenseMap``'s ``find_as()`` method supports lookup operations using an alternate key
 type.  This is useful in cases where the normal key type is expensive to
-construct, but cheap to compare against.  The DenseMapInfo is responsible for
+construct, but cheap to compare against.  The ``DenseMapInfo`` is responsible for
 defining the appropriate comparison and hashing methods for each alternate key
 type used.
 
-DenseMap.h also contains a SmallDenseMap variant, that similar to
+``DenseMap.h`` also contains a ``SmallDenseMap`` variant, that similar to
 :ref:`SmallVector <dss_smallvector>` performs no heap allocation until the
 number of elements in the template parameter N are exceeded.
 
@@ -2404,12 +2404,12 @@ further additions.
 <map>
 ^^^^^
 
-std::map has similar characteristics to :ref:`std::set <dss_set>`: it uses a
+``std::map`` has similar characteristics to :ref:`std::set <dss_set>`: it uses a
 single allocation per pair inserted into the map, it offers log(n) lookup with
 an extremely large constant factor, imposes a space penalty of 3 pointers per
 pair in the map, etc.
 
-std::map is most useful when your keys or values are very large, if you need to
+``std::map`` is most useful when your keys or values are very large, if you need to
 iterate over the collection in sorted order, or if you need stable iterators
 into the map (i.e. they don't get invalidated if an insertion or deletion of
 another element takes place).
@@ -2419,7 +2419,7 @@ another element takes place).
 llvm/ADT/MapVector.h
 ^^^^^^^^^^^^^^^^^^^^
 
-``MapVector<KeyT,ValueT>`` provides a subset of the DenseMap interface.  The
+``MapVector<KeyT,ValueT>`` provides a subset of the ``DenseMap`` interface.  The
 main difference is that the iteration order is guaranteed to be the insertion
 order, making it an easy (but somewhat expensive) solution for non-deterministic
 iteration over maps of pointers.
@@ -2463,12 +2463,12 @@ operations is logarithmic in the size of the original map.
 Other Map-Like Container Options
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 
-The STL provides several other options, such as std::multimap and
-std::unordered_map.  We never use containers like unordered_map because
+The STL provides several other options, such as ``std::multimap`` and
+``std::unordered_map``.  We never use containers like ``unordered_map`` because
 they are generally very expensive (each insertion requires a malloc).
 
-std::multimap is useful if you want to map a key to multiple values, but has all
-the drawbacks of std::map.  A sorted vector or some other approach is almost
+``std::multimap`` is useful if you want to map a key to multiple values, but has all
+the drawbacks of ``std::map``.  A sorted vector or some other approach is almost
 always better.
 
 .. _ds_bit:
diff --git a/llvm/docs/ReleaseNotes.md b/llvm/docs/ReleaseNotes.md
index 48d2ef1..021f321 100644
--- a/llvm/docs/ReleaseNotes.md
+++ b/llvm/docs/ReleaseNotes.md
@@ -68,6 +68,12 @@ Changes to TableGen
 Changes to Interprocedural Optimizations
 ----------------------------------------
 
+Changes to Vectorizers
+----------------------------------------
+
+* Added initial support for copyable elements in SLP, which models copyable
+  elements as add <element>, 0, i.e. uses identity constants for missing lanes.
+
 Changes to the AArch64 Backend
 ------------------------------
 
diff --git a/llvm/docs/YamlIO.rst b/llvm/docs/YamlIO.rst
index 420adb8..c5079d8 100644
--- a/llvm/docs/YamlIO.rst
+++ b/llvm/docs/YamlIO.rst
@@ -8,10 +8,10 @@ YAML I/O
 Introduction to YAML
 ====================
 
-YAML is a human readable data serialization language.  The full YAML language
+YAML is a human-readable data serialization language.  The full YAML language
 spec can be read at `yaml.org
 <http://www.yaml.org/spec/1.2/spec.html#Introduction>`_.  The simplest form of
-yaml is just "scalars", "mappings", and "sequences".  A scalar is any number
+YAML is just "scalars", "mappings", and "sequences".  A scalar is any number
 or string.  The pound/hash symbol (#) begins a comment line.   A mapping is
 a set of key-value pairs where the key ends with a colon.  For example:
 
@@ -49,10 +49,10 @@ of mappings in which one of the mapping values is itself a sequence:
         - PowerPC
         - x86
 
-Sometime sequences are known to be short and the one entry per line is too
-verbose, so YAML offers an alternate syntax for sequences called a "Flow
+Sometimes sequences are known to be short and the one entry per line is too
+verbose, so YAML offers an alternative syntax for sequences called a "Flow
 Sequence" in which you put comma separated sequence elements into square
-brackets.  The above example could then be simplified to :
+brackets.  The above example could then be simplified to:
 
 
 .. code-block:: yaml
@@ -78,21 +78,21 @@ YAML I/O assumes you have some "native" data structures which you want to be
 able to dump as YAML and recreate from YAML.  The first step is to try
 writing example YAML for your data structures. You may find after looking at
 possible YAML representations that a direct mapping of your data structures
-to YAML is not very readable.  Often the fields are not in the order that
+to YAML is not very readable.  Often, the fields are not in an order that
 a human would find readable.  Or the same information is replicated in multiple
 locations, making it hard for a human to write such YAML correctly.
 
 In relational database theory there is a design step called normalization in
 which you reorganize fields and tables.  The same considerations need to
 go into the design of your YAML encoding.  But, you may not want to change
-your existing native data structures.  Therefore, when writing out YAML
+your existing native data structures.  Therefore, when writing out YAML,
 there may be a normalization step, and when reading YAML there would be a
 corresponding denormalization step.
 
-YAML I/O uses a non-invasive, traits based design.  YAML I/O defines some
+YAML I/O uses a non-invasive, traits-based design.  YAML I/O defines some
 abstract base templates.  You specialize those templates on your data types.
-For instance, if you have an enumerated type FooBar you could specialize
-ScalarEnumerationTraits on that type and define the ``enumeration()`` method:
+For instance, if you have an enumerated type ``FooBar`` you could specialize
+``ScalarEnumerationTraits`` on that type and define the ``enumeration()`` method:
 
 .. code-block:: c++
 
@@ -107,12 +107,12 @@ ScalarEnumerationTraits on that type and define the ``enumeration()`` method:
     };
 
 
-As with all YAML I/O template specializations, the ScalarEnumerationTraits is used for
+As with all YAML I/O template specializations, the ``ScalarEnumerationTraits`` is used for
 both reading and writing YAML. That is, the mapping between in-memory enum
 values and the YAML string representation is only in one place.
 This assures that the code for writing and parsing of YAML stays in sync.
 
-To specify a YAML mappings, you define a specialization on
+To specify YAML mappings, you define a specialization on
 ``llvm::yaml::MappingTraits``.
 If your native data structure happens to be a struct that is already normalized,
 then the specialization is simple.  For example:
@@ -131,7 +131,7 @@ then the specialization is simple.  For example:
     };
 
 
-A YAML sequence is automatically inferred if you data type has ``begin()``/``end()``
+A YAML sequence is automatically inferred if your data type has ``begin()``/``end()``
 iterators and a ``push_back()`` method.  Therefore any of the STL containers
 (such as ``std::vector<>``) will automatically translate to YAML sequences.
 
@@ -197,7 +197,7 @@ Error Handling
 When parsing a YAML document, if the input does not match your schema (as
 expressed in your ``XxxTraits<>`` specializations).  YAML I/O
 will print out an error message and your Input object's ``error()`` method will
-return true. For instance the following document:
+return true. For instance, the following document:
 
 .. code-block:: yaml
 
@@ -244,8 +244,8 @@ The following types have built-in support in YAML I/O:
 * uint16_t
 * uint8_t
 
-That is, you can use those types in fields of MappingTraits or as element type
-in sequence.  When reading, YAML I/O will validate that the string found
+That is, you can use those types in fields of ``MappingTraits`` or as the element type
+in a sequence.  When reading, YAML I/O will validate that the string found
 is convertible to that type and error out if not.
 
 
@@ -255,7 +255,7 @@ Given that YAML I/O is trait based, the selection of how to convert your data
 to YAML is based on the type of your data.  But in C++ type matching, typedefs
 do not generate unique type names.  That means if you have two typedefs of
 unsigned int, to YAML I/O both types look exactly like unsigned int.  To
-facilitate make unique type names, YAML I/O provides a macro which is used
+facilitate making unique type names, YAML I/O provides a macro which is used
 like a typedef on built-in types, but expands to create a class with conversion
 operators to and from the base type.  For example:
 
@@ -271,7 +271,7 @@ is that you can now specify traits on them to get different YAML conversions.
 
 Hex types
 ---------
-An example use of a unique type is that YAML I/O provides fixed sized unsigned
+An example use of a unique type is that YAML I/O provides fixed-sized unsigned
 integers that are written with YAML I/O as hexadecimal instead of the decimal
 format used by the built-in integer types:
 
@@ -280,7 +280,7 @@ format used by the built-in integer types:
 * Hex16
 * Hex8
 
-You can use ``llvm::yaml::Hex32`` instead of ``uint32_t`` and the only different will
+You can use ``llvm::yaml::Hex32`` instead of ``uint32_t`` and the only difference will
 be that when YAML I/O writes out that type it will be formatted in hexadecimal.
 
 
@@ -288,7 +288,7 @@ ScalarEnumerationTraits
 -----------------------
 YAML I/O supports translating between in-memory enumerations and a set of string
 values in YAML documents. This is done by specializing ``ScalarEnumerationTraits<>``
-on your enumeration type and define an ``enumeration()`` method.
+on your enumeration type and defining an ``enumeration()`` method.
 For instance, suppose you had an enumeration of CPUs and a struct with it as
 a field:
 
@@ -306,7 +306,7 @@ a field:
     };
 
 To support reading and writing of this enumeration, you can define a
-ScalarEnumerationTraits specialization on CPUs, which can then be used
+``ScalarEnumerationTraits`` specialization on CPUs, which can then be used
 as a field type:
 
 .. code-block:: c++
@@ -356,7 +356,7 @@ had the following bit flags defined:
 
     LLVM_YAML_STRONG_TYPEDEF(uint32_t, MyFlags)
 
-To support reading and writing of MyFlags, you specialize ScalarBitSetTraits<>
+To support reading and writing of MyFlags, you specialize ``ScalarBitSetTraits<>``
 on MyFlags and provide the bit values and their names.
 
 .. code-block:: c++
@@ -399,7 +399,7 @@ the above schema, a same valid YAML document is:
     name:    Tom
     flags:   [ pointy, flat ]
 
-Sometimes a "flags" field might contains an enumeration part
+Sometimes a "flags" field might contain an enumeration part
 defined by a bit-mask.
 
 .. code-block:: c++
@@ -415,7 +415,7 @@ defined by a bit-mask.
       flagsCPU2 = 16
     };
 
-To support reading and writing such fields, you need to use the maskedBitSet()
+To support reading and writing such fields, you need to use the ``maskedBitSet()``
 method and provide the bit values, their names and the enumeration mask.
 
 .. code-block:: c++
@@ -438,8 +438,8 @@ to the flow sequence.
 
 Custom Scalar
 -------------
-Sometimes for readability a scalar needs to be formatted in a custom way. For
-instance your internal data structure may use an integer for time (seconds since
+Sometimes, for readability, a scalar needs to be formatted in a custom way. For
+instance, your internal data structure may use an integer for time (seconds since
 some epoch), but in YAML it would be much nicer to express that integer in
 some time format (e.g. 4-May-2012 10:30pm).  YAML I/O has a way to support
 custom formatting and parsing of scalar types by specializing ``ScalarTraits<>`` on
@@ -487,13 +487,13 @@ your data type. The library doesn't provide any built-in support for block
 scalar I/O for types like ``std::string`` and ``llvm::StringRef`` as they are already
 supported by YAML I/O and use the ordinary scalar notation by default.
 
-BlockScalarTraits specializations are very similar to the
-ScalarTraits specialization - YAML I/O will provide the native type and your
+``BlockScalarTraits`` specializations are very similar to the
+``ScalarTraits`` specialization - YAML I/O will provide the native type and your
 specialization must create a temporary ``llvm::StringRef`` when writing, and
 it will also provide an ``llvm::StringRef`` that has the value of that block scalar
 and your specialization must convert that to your native data type when reading.
 An example of a custom type with an appropriate specialization of
-BlockScalarTraits is shown below:
+``BlockScalarTraits`` is shown below:
 
 .. code-block:: c++
 
@@ -621,7 +621,7 @@ Polar which specifies a position in polar coordinates (distance, angle):
       float angle;
     };
 
-but you've decided the normalized YAML for should be in x,y coordinates. That
+but you've decided the normalized YAML form should be in x,y coordinates. That
 is, you want the yaml to look like:
 
 .. code-block:: yaml
@@ -762,19 +762,19 @@ tag on a map. Using this functionality it is even possible to support different
 mappings, as long as they are convertible.
 
 To check a tag, inside your ``mapping()`` method you can use ``io.mapTag()`` to specify
-what the tag should be.  This will also add that tag when writing yaml.
+what the tag should be.  This will also add that tag when writing YAML.
 
 Validation
 ----------
 
 Sometimes in a YAML map, each key/value pair is valid, but the combination is
 not.  This is similar to something having no syntax errors, but still having
-semantic errors.  To support semantic level checking, YAML I/O allows
+semantic errors.  To support semantic-level checking, YAML I/O allows
 an optional ``validate()`` method in a MappingTraits template specialization.
 
 When parsing YAML, the ``validate()`` method is call *after* all key/values in
 the map have been processed. Any error message returned by the ``validate()``
-method during input will be printed just a like a syntax error would be printed.
+method during input will be printed just like a syntax error would be printed.
 When writing YAML, the ``validate()`` method is called *before* the YAML
 key/values  are written.  Any error during output will trigger an ``assert()``
 because it is a programming error to have invalid struct values.
@@ -827,7 +827,7 @@ add "static const bool flow = true;". For instance:
       static const bool flow = true;
     }
 
-Flow mappings are subject to line wrapping according to the Output object
+Flow mappings are subject to line wrapping according to the ``Output`` object
 configuration.
 
 Sequence
@@ -850,7 +850,7 @@ The ``size()`` method returns how many elements are currently in your sequence.
 The ``element()`` method returns a reference to the i'th element in the sequence.
 When parsing YAML, the ``element()`` method may be called with an index one bigger
 than the current size.  Your ``element()`` method should allocate space for one
-more element (using default constructor if element is a C++ object) and returns
+more element (using default constructor if element is a C++ object) and return
 a reference to that new allocated space.
 
 
@@ -919,8 +919,8 @@ trait for you document list type.  The trait has the same methods as
 
 User Context Data
 =================
-When an ``llvm::yaml::Input`` or ``llvm::yaml::Output`` object is created their
-constructors take an optional "context" parameter.  This is a pointer to
+When an ``llvm::yaml::Input`` or ``llvm::yaml::Output`` object is created, its
+constructor takes an optional "context" parameter.  This is a pointer to
 whatever state information you might need.
 
 For instance, in a previous example we showed how the conversion type for a
@@ -930,10 +930,10 @@ of an outer mapping?  That is where the "context" parameter comes in. You
 can set values in the context in the outer map's ``mapping()`` method and
 retrieve those values in the inner map's ``mapping()`` method.
 
-The context value is just a void*.  All your traits which use the context
+The context value is just a ``void*``.  All your traits which use the context
 and operate on your native data types, need to agree what the context value
 actually is.  It could be a pointer to an object or struct which your various
-traits use to shared context sensitive information.
+traits use to share context sensitive information.
 
 
 Output
@@ -941,7 +941,7 @@ Output
 
 The ``llvm::yaml::Output`` class is used to generate a YAML document from your
 in-memory data structures, using traits defined on your data types.
-To instantiate an Output object you need an ``llvm::raw_ostream``, an optional
+To instantiate an ``Output`` object you need an ``llvm::raw_ostream``, an optional
 context pointer and an optional wrapping column:
 
 .. code-block:: c++
@@ -950,10 +950,10 @@ context pointer and an optional wrapping column:
       public:
         Output(llvm::raw_ostream &, void *context = NULL, int WrapColumn = 70);
 
-Once you have an Output object, you can use the C++ stream operator on it
+Once you have an ``Output`` object, you can use the C++ stream operator on it
 to write your native data as YAML. One thing to recall is that a YAML file
 can contain multiple "documents".  If the top level data structure you are
-streaming as YAML is a mapping, scalar, or sequence, then Output assumes you
+streaming as YAML is a mapping, scalar, or sequence, then ``Output`` assumes you
 are generating one document and wraps the mapping output
 with  "``---``" and trailing "``...``".
 
@@ -1009,7 +1009,7 @@ Input
 =====
 
 The ``llvm::yaml::Input`` class is used to parse YAML document(s) into your native
-data structures. To instantiate an Input
+data structures. To instantiate an ``Input``
 object you need a ``StringRef`` to the entire YAML file, and optionally a context
 pointer:
 
@@ -1019,12 +1019,12 @@ pointer:
       public:
         Input(StringRef inputContent, void *context=NULL);
 
-Once you have an Input object, you can use the C++ stream operator to read
+Once you have an ``Input`` object, you can use the C++ stream operator to read
 the document(s).  If you expect there might be multiple YAML documents in
-one file, you'll need to specialize DocumentListTraits on a list of your
+one file, you'll need to specialize ``DocumentListTraits`` on a list of your
 document type and stream in that document list type.  Otherwise you can
 just stream in the document type.  Also, you can check if there was
-any syntax errors in the YAML be calling the ``error()`` method on the Input
+any syntax errors in the YAML by calling the ``error()`` method on the ``Input``
 object.  For example:
 
 .. code-block:: c++
diff --git a/llvm/include/llvm/Analysis/DXILResource.h b/llvm/include/llvm/Analysis/DXILResource.h
index 9e2dc1a..956dcbc 100644
--- a/llvm/include/llvm/Analysis/DXILResource.h
+++ b/llvm/include/llvm/Analysis/DXILResource.h
@@ -359,6 +359,8 @@ public:
              std::tie(RHS.RecordID, RHS.Space, RHS.LowerBound, RHS.Size);
     }
     bool overlapsWith(const ResourceBinding &RHS) const {
+      if (Size == UINT32_MAX)
+        return LowerBound < RHS.LowerBound;
       return Space == RHS.Space && LowerBound + Size - 1 >= RHS.LowerBound;
     }
   };
diff --git a/llvm/include/llvm/Analysis/ScalarEvolutionPatternMatch.h b/llvm/include/llvm/Analysis/ScalarEvolutionPatternMatch.h
index 09e3945..bff7707 100644
--- a/llvm/include/llvm/Analysis/ScalarEvolutionPatternMatch.h
+++ b/llvm/include/llvm/Analysis/ScalarEvolutionPatternMatch.h
@@ -160,6 +160,12 @@ m_scev_ZExt(const Op0_t &Op0) {
   return m_scev_Unary<SCEVZeroExtendExpr>(Op0);
 }
 
+template <typename Op0_t>
+inline SCEVUnaryExpr_match<SCEVPtrToIntExpr, Op0_t>
+m_scev_PtrToInt(const Op0_t &Op0) {
+  return SCEVUnaryExpr_match<SCEVPtrToIntExpr, Op0_t>(Op0);
+}
+
 /// Match a binary SCEV.
 template <typename SCEVTy, typename Op0_t, typename Op1_t>
 struct SCEVBinaryExpr_match {
diff --git a/llvm/include/llvm/Config/llvm-config.h.cmake b/llvm/include/llvm/Config/llvm-config.h.cmake
index a683229..39136bc 100644
--- a/llvm/include/llvm/Config/llvm-config.h.cmake
+++ b/llvm/include/llvm/Config/llvm-config.h.cmake
@@ -101,6 +101,9 @@
 /* Define if LLVM is using tflite */
 #cmakedefine LLVM_HAVE_TFLITE
 
+/* Define if we want to check profile consistency in lit tests */
+#cmakedefine LLVM_ENABLE_PROFCHECK
+
 /* Define to 1 if you have the <sysexits.h> header file. */
 #cmakedefine HAVE_SYSEXITS_H ${HAVE_SYSEXITS_H}
 
diff --git a/llvm/include/llvm/IR/DiagnosticInfo.h b/llvm/include/llvm/IR/DiagnosticInfo.h
index 862be04..5f7225e 100644
--- a/llvm/include/llvm/IR/DiagnosticInfo.h
+++ b/llvm/include/llvm/IR/DiagnosticInfo.h
@@ -68,6 +68,7 @@ enum DiagnosticKind {
   DK_StackSize,
   DK_Linker,
   DK_Lowering,
+  DK_LegalizationFailure,
   DK_DebugMetadataVersion,
   DK_DebugMetadataInvalid,
   DK_Instrumentation,
@@ -383,6 +384,30 @@ private:
   DiagnosticLocation Loc;
 };
 
+class LLVM_ABI DiagnosticInfoLegalizationFailure
+    : public DiagnosticInfoWithLocationBase {
+private:
+  /// Message to be reported.
+  const Twine &MsgStr;
+
+public:
+  DiagnosticInfoLegalizationFailure(const Twine &MsgStr LLVM_LIFETIME_BOUND,
+                                    const Function &Fn,
+                                    const DiagnosticLocation &Loc,
+                                    DiagnosticSeverity Severity = DS_Error)
+      : DiagnosticInfoWithLocationBase(DK_LegalizationFailure, Severity, Fn,
+                                       Loc),
+        MsgStr(MsgStr) {}
+
+  const Twine &getMsgStr() const { return MsgStr; }
+
+  void print(DiagnosticPrinter &DP) const override;
+
+  static bool classof(const DiagnosticInfo *DI) {
+    return DI->getKind() == DK_LegalizationFailure;
+  }
+};
+
 class LLVM_ABI DiagnosticInfoGenericWithLoc
     : public DiagnosticInfoWithLocationBase {
 private:
diff --git a/llvm/include/llvm/MC/MCAsmInfo.h b/llvm/include/llvm/MC/MCAsmInfo.h
index 71da048..6c12cd3 100644
--- a/llvm/include/llvm/MC/MCAsmInfo.h
+++ b/llvm/include/llvm/MC/MCAsmInfo.h
@@ -35,6 +35,7 @@ class MCStreamer;
 class MCSubtargetInfo;
 class MCSymbol;
 class MCValue;
+class Triple;
 class raw_ostream;
 
 namespace WinEH {
@@ -485,6 +486,9 @@ public:
   /// syntactically correct.
   virtual bool isValidUnquotedName(StringRef Name) const;
 
+  virtual void printSwitchToSection(const MCSection &, uint32_t Subsection,
+                                    const Triple &, raw_ostream &) const {}
+
   /// Return true if the .section directive should be omitted when
   /// emitting \p SectionName.  For example:
   ///
@@ -494,6 +498,10 @@ public:
   /// returns true  => .text
   virtual bool shouldOmitSectionDirective(StringRef SectionName) const;
 
+  // Return true if a .align directive should use "optimized nops" to fill
+  // instead of 0s.
+  virtual bool useCodeAlign(const MCSection &Sec) const { return false; }
+
   bool usesSunStyleELFSectionSwitchSyntax() const {
     return SunStyleELFSectionSwitchSyntax;
   }
diff --git a/llvm/include/llvm/MC/MCAsmInfoCOFF.h b/llvm/include/llvm/MC/MCAsmInfoCOFF.h
index 1dfb475..dc7832c 100644
--- a/llvm/include/llvm/MC/MCAsmInfoCOFF.h
+++ b/llvm/include/llvm/MC/MCAsmInfoCOFF.h
@@ -15,6 +15,9 @@ namespace llvm {
 
 class MCAsmInfoCOFF : public MCAsmInfo {
   virtual void anchor();
+  void printSwitchToSection(const MCSection &, uint32_t, const Triple &,
+                            raw_ostream &) const final;
+  bool useCodeAlign(const MCSection &Sec) const final;
 
 protected:
   explicit MCAsmInfoCOFF();
diff --git a/llvm/include/llvm/MC/MCAsmInfoDarwin.h b/llvm/include/llvm/MC/MCAsmInfoDarwin.h
index 4ca62b3..12bc3e9 100644
--- a/llvm/include/llvm/MC/MCAsmInfoDarwin.h
+++ b/llvm/include/llvm/MC/MCAsmInfoDarwin.h
@@ -21,6 +21,9 @@ namespace llvm {
 class MCAsmInfoDarwin : public MCAsmInfo {
 public:
   explicit MCAsmInfoDarwin();
+  void printSwitchToSection(const MCSection &, uint32_t, const Triple &,
+                            raw_ostream &) const final;
+  bool useCodeAlign(const MCSection &Sec) const final;
 
   /// True if the section is atomized using the symbols in it.
   /// This is false if the section is atomized based on its contents (MachO' __TEXT,__cstring for
diff --git a/llvm/include/llvm/MC/MCAsmInfoELF.h b/llvm/include/llvm/MC/MCAsmInfoELF.h
index 408d4df..c05e4ad 100644
--- a/llvm/include/llvm/MC/MCAsmInfoELF.h
+++ b/llvm/include/llvm/MC/MCAsmInfoELF.h
@@ -16,6 +16,9 @@ namespace llvm {
 class MCAsmInfoELF : public MCAsmInfo {
   virtual void anchor();
   MCSection *getNonexecutableStackSection(MCContext &Ctx) const final;
+  void printSwitchToSection(const MCSection &, uint32_t, const Triple &,
+                            raw_ostream &) const final;
+  bool useCodeAlign(const MCSection &Sec) const final;
 
 protected:
   MCAsmInfoELF();
diff --git a/llvm/include/llvm/MC/MCAsmInfoGOFF.h b/llvm/include/llvm/MC/MCAsmInfoGOFF.h
index 1f3b263..e62d2ae 100644
--- a/llvm/include/llvm/MC/MCAsmInfoGOFF.h
+++ b/llvm/include/llvm/MC/MCAsmInfoGOFF.h
@@ -19,7 +19,8 @@
 
 namespace llvm {
 class MCAsmInfoGOFF : public MCAsmInfo {
-  virtual void anchor();
+  void printSwitchToSection(const MCSection &, uint32_t, const Triple &,
+                            raw_ostream &) const final;
 
 protected:
   MCAsmInfoGOFF();
diff --git a/llvm/include/llvm/MC/MCAsmInfoWasm.h b/llvm/include/llvm/MC/MCAsmInfoWasm.h
index 3afc610..d98de6c 100644
--- a/llvm/include/llvm/MC/MCAsmInfoWasm.h
+++ b/llvm/include/llvm/MC/MCAsmInfoWasm.h
@@ -13,7 +13,8 @@
 
 namespace llvm {
 class MCAsmInfoWasm : public MCAsmInfo {
-  virtual void anchor();
+  void printSwitchToSection(const MCSection &, uint32_t, const Triple &,
+                            raw_ostream &) const final;
 
 protected:
   MCAsmInfoWasm();
diff --git a/llvm/include/llvm/MC/MCAsmInfoXCOFF.h b/llvm/include/llvm/MC/MCAsmInfoXCOFF.h
index 5483899..fd1ae82 100644
--- a/llvm/include/llvm/MC/MCAsmInfoXCOFF.h
+++ b/llvm/include/llvm/MC/MCAsmInfoXCOFF.h
@@ -14,10 +14,11 @@
 namespace llvm {
 
 class MCAsmInfoXCOFF : public MCAsmInfo {
-  virtual void anchor();
-
 protected:
   MCAsmInfoXCOFF();
+  void printSwitchToSection(const MCSection &, uint32_t, const Triple &,
+                            raw_ostream &) const final;
+  bool useCodeAlign(const MCSection &Sec) const final;
 
 public:
   // Return true only when C is an acceptable character inside a
diff --git a/llvm/include/llvm/MC/MCMachObjectWriter.h b/llvm/include/llvm/MC/MCMachObjectWriter.h
index 51e4df5..170e2e7 100644
--- a/llvm/include/llvm/MC/MCMachObjectWriter.h
+++ b/llvm/include/llvm/MC/MCMachObjectWriter.h
@@ -16,7 +16,7 @@
 #include "llvm/MC/MCExpr.h"
 #include "llvm/MC/MCLinkerOptimizationHint.h"
 #include "llvm/MC/MCObjectWriter.h"
-#include "llvm/MC/MCSection.h"
+#include "llvm/MC/MCSectionMachO.h"
 #include "llvm/MC/StringTableBuilder.h"
 #include "llvm/Support/Compiler.h"
 #include "llvm/Support/EndianStream.h"
@@ -276,7 +276,7 @@ public:
                                uint64_t SectionDataSize, uint32_t MaxProt,
                                uint32_t InitProt);
 
-  void writeSection(const MCAssembler &Asm, const MCSection &Sec,
+  void writeSection(const MCAssembler &Asm, const MCSectionMachO &Sec,
                     uint64_t VMAddr, uint64_t FileOffset, unsigned Flags,
                     uint64_t RelocationsStart, unsigned NumRelocations);
 
diff --git a/llvm/include/llvm/MC/MCObjectStreamer.h b/llvm/include/llvm/MC/MCObjectStreamer.h
index aea93e9..bbbee15 100644
--- a/llvm/include/llvm/MC/MCObjectStreamer.h
+++ b/llvm/include/llvm/MC/MCObjectStreamer.h
@@ -102,7 +102,6 @@ public:
   void emitSLEB128Value(const MCExpr *Value) override;
   void emitWeakReference(MCSymbol *Alias, const MCSymbol *Target) override;
   void changeSection(MCSection *Section, uint32_t Subsection = 0) override;
-  void switchSectionNoPrint(MCSection *Section) override;
   void emitInstruction(const MCInst &Inst, const MCSubtargetInfo &STI) override;
 
   /// Emit an instruction to a special fragment, because this instruction
diff --git a/llvm/include/llvm/MC/MCSection.h b/llvm/include/llvm/MC/MCSection.h
index 125f849..7989310 100644
--- a/llvm/include/llvm/MC/MCSection.h
+++ b/llvm/include/llvm/MC/MCSection.h
@@ -540,17 +540,6 @@ public:
   friend class MCFragment;
   static constexpr unsigned NonUniqueID = ~0U;
 
-  enum SectionVariant {
-    SV_COFF = 0,
-    SV_ELF,
-    SV_GOFF,
-    SV_MachO,
-    SV_Wasm,
-    SV_XCOFF,
-    SV_SPIRV,
-    SV_DXContainer,
-  };
-
   struct iterator {
     MCFragment *F = nullptr;
     iterator() = default;
@@ -570,6 +559,8 @@ private:
   // At parse time, this holds the fragment list of the current subsection. At
   // layout time, this holds the concatenated fragment lists of all subsections.
   FragList *CurFragList;
+  // In many object file formats, this denotes the section symbol. In Mach-O,
+  // this denotes an optional temporary label at the section start.
   MCSymbol *Begin;
   MCSymbol *End = nullptr;
   /// The alignment requirement of this section.
@@ -604,12 +595,8 @@ private:
 protected:
   // TODO Make Name private when possible.
   StringRef Name;
-  SectionVariant Variant;
 
-  MCSection(SectionVariant V, StringRef Name, bool IsText, bool IsBss,
-            MCSymbol *Begin);
-  // Protected non-virtual dtor prevents destroy through a base class pointer.
-  ~MCSection() {}
+  MCSection(StringRef Name, bool IsText, bool IsBss, MCSymbol *Begin);
 
 public:
   MCSection(const MCSection &) = delete;
@@ -618,8 +605,6 @@ public:
   StringRef getName() const { return Name; }
   bool isText() const { return IsText; }
 
-  SectionVariant getVariant() const { return Variant; }
-
   MCSymbol *getBeginSymbol() { return Begin; }
   const MCSymbol *getBeginSymbol() const {
     return const_cast<MCSection *>(this)->getBeginSymbol();
@@ -661,14 +646,6 @@ public:
   void dump(DenseMap<const MCFragment *, SmallVector<const MCSymbol *, 0>>
                 *FragToSyms = nullptr) const;
 
-  virtual void printSwitchToSection(const MCAsmInfo &MAI, const Triple &T,
-                                    raw_ostream &OS,
-                                    uint32_t Subsection) const = 0;
-
-  /// Return true if a .align directive should use "optimized nops" to fill
-  /// instead of 0s.
-  virtual bool useCodeAlign() const = 0;
-
   /// Check whether this section is "virtual", that is has no actual object
   /// file contents.
   bool isBssSection() const { return IsBss; }
diff --git a/llvm/include/llvm/MC/MCSectionCOFF.h b/llvm/include/llvm/MC/MCSectionCOFF.h
index f979413a..71efc41 100644
--- a/llvm/include/llvm/MC/MCSectionCOFF.h
+++ b/llvm/include/llvm/MC/MCSectionCOFF.h
@@ -51,11 +51,12 @@ class MCSectionCOFF final : public MCSection {
 
 private:
   friend class MCContext;
+  friend class MCAsmInfoCOFF;
   // The storage of Name is owned by MCContext's COFFUniquingMap.
   MCSectionCOFF(StringRef Name, unsigned Characteristics,
                 MCSymbol *COMDATSymbol, int Selection, unsigned UniqueID,
                 MCSymbol *Begin)
-      : MCSection(SV_COFF, Name, Characteristics & COFF::IMAGE_SCN_CNT_CODE,
+      : MCSection(Name, Characteristics & COFF::IMAGE_SCN_CNT_CODE,
                   Characteristics & COFF::IMAGE_SCN_CNT_UNINITIALIZED_DATA,
                   Begin),
         Characteristics(Characteristics), COMDATSymbol(COMDATSymbol),
@@ -67,7 +68,7 @@ private:
 public:
   /// Decides whether a '.section' directive should be printed before the
   /// section name
-  bool shouldOmitSectionDirective(StringRef Name, const MCAsmInfo &MAI) const;
+  bool shouldOmitSectionDirective(StringRef Name) const;
 
   unsigned getCharacteristics() const { return Characteristics; }
   MCSymbol *getCOMDATSymbol() const { return COMDATSymbol; }
@@ -78,11 +79,6 @@ public:
   bool isUnique() const { return UniqueID != NonUniqueID; }
   unsigned getUniqueID() const { return UniqueID; }
 
-  void printSwitchToSection(const MCAsmInfo &MAI, const Triple &T,
-                            raw_ostream &OS,
-                            uint32_t Subsection) const override;
-  bool useCodeAlign() const override;
-
   unsigned getOrAssignWinCFISectionID(unsigned *NextID) const {
     if (WinCFISectionID == ~0U)
       WinCFISectionID = (*NextID)++;
@@ -92,8 +88,6 @@ public:
   static bool isImplicitlyDiscardable(StringRef Name) {
     return Name.starts_with(".debug");
   }
-
-  static bool classof(const MCSection *S) { return S->getVariant() == SV_COFF; }
 };
 
 } // end namespace llvm
diff --git a/llvm/include/llvm/MC/MCSectionDXContainer.h b/llvm/include/llvm/MC/MCSectionDXContainer.h
index 723b477..7d8e0c5 100644
--- a/llvm/include/llvm/MC/MCSectionDXContainer.h
+++ b/llvm/include/llvm/MC/MCSectionDXContainer.h
@@ -24,13 +24,7 @@ class MCSectionDXContainer final : public MCSection {
   friend class MCContext;
 
   MCSectionDXContainer(StringRef Name, SectionKind K, MCSymbol *Begin)
-      : MCSection(SV_DXContainer, Name, K.isText(), /*IsVirtual=*/false,
-                  Begin) {}
-
-public:
-  void printSwitchToSection(const MCAsmInfo &, const Triple &, raw_ostream &,
-                            uint32_t) const override;
-  bool useCodeAlign() const override { return false; }
+      : MCSection(Name, K.isText(), /*IsVirtual=*/false, Begin) {}
 };
 
 } // end namespace llvm
diff --git a/llvm/include/llvm/MC/MCSectionELF.h b/llvm/include/llvm/MC/MCSectionELF.h
index 64a4daf..f089dd9 100644
--- a/llvm/include/llvm/MC/MCSectionELF.h
+++ b/llvm/include/llvm/MC/MCSectionELF.h
@@ -52,14 +52,15 @@ class MCSectionELF final : public MCSection {
 
 private:
   friend class MCContext;
+  friend class MCAsmInfoELF;
 
   // The storage of Name is owned by MCContext's ELFUniquingMap.
   MCSectionELF(StringRef Name, unsigned type, unsigned flags,
                unsigned entrySize, const MCSymbolELF *group, bool IsComdat,
                unsigned UniqueID, MCSymbol *Begin,
                const MCSymbolELF *LinkedToSym)
-      : MCSection(SV_ELF, Name, flags & ELF::SHF_EXECINSTR,
-                  type == ELF::SHT_NOBITS, Begin),
+      : MCSection(Name, flags & ELF::SHF_EXECINSTR, type == ELF::SHT_NOBITS,
+                  Begin),
         Type(type), Flags(flags), UniqueID(UniqueID), EntrySize(entrySize),
         Group(group, IsComdat), LinkedToSym(LinkedToSym) {
     assert((!(Flags & ELF::SHF_GROUP) || Group.getPointer()) &&
@@ -69,10 +70,6 @@ private:
   }
 
 public:
-  /// Decides whether a '.section' directive should be printed before the
-  /// section name
-  bool shouldOmitSectionDirective(StringRef Name, const MCAsmInfo &MAI) const;
-
   unsigned getType() const { return Type; }
   unsigned getFlags() const { return Flags; }
   unsigned getEntrySize() const { return EntrySize; }
@@ -80,11 +77,6 @@ public:
   const MCSymbolELF *getGroup() const { return Group.getPointer(); }
   bool isComdat() const { return Group.getInt(); }
 
-  void printSwitchToSection(const MCAsmInfo &MAI, const Triple &T,
-                            raw_ostream &OS,
-                            uint32_t Subsection) const override;
-  bool useCodeAlign() const override;
-
   bool isUnique() const { return UniqueID != NonUniqueID; }
   unsigned getUniqueID() const { return UniqueID; }
 
@@ -100,10 +92,6 @@ public:
   std::pair<uint64_t, uint64_t> getOffsets() const {
     return std::make_pair(StartOffset, EndOffset);
   }
-
-  static bool classof(const MCSection *S) {
-    return S->getVariant() == SV_ELF;
-  }
 };
 
 } // end namespace llvm
diff --git a/llvm/include/llvm/MC/MCSectionGOFF.h b/llvm/include/llvm/MC/MCSectionGOFF.h
index b166397..2136148 100644
--- a/llvm/include/llvm/MC/MCSectionGOFF.h
+++ b/llvm/include/llvm/MC/MCSectionGOFF.h
@@ -52,36 +52,28 @@ class LLVM_ABI MCSectionGOFF final : public MCSection {
   mutable unsigned Emitted : 1;
 
   friend class MCContext;
+  friend class MCAsmInfoGOFF;
   friend class MCSymbolGOFF;
 
   MCSectionGOFF(StringRef Name, SectionKind K, bool IsVirtual,
                 GOFF::SDAttr SDAttributes, MCSectionGOFF *Parent)
-      : MCSection(SV_GOFF, Name, K.isText(), IsVirtual, nullptr),
-        Parent(Parent), SDAttributes(SDAttributes),
-        SymbolType(GOFF::ESD_ST_SectionDefinition), IsBSS(K.isBSS()),
-        RequiresNonZeroLength(0), Emitted(0) {}
+      : MCSection(Name, K.isText(), IsVirtual, nullptr), Parent(Parent),
+        SDAttributes(SDAttributes), SymbolType(GOFF::ESD_ST_SectionDefinition),
+        IsBSS(K.isBSS()), RequiresNonZeroLength(0), Emitted(0) {}
 
   MCSectionGOFF(StringRef Name, SectionKind K, bool IsVirtual,
                 GOFF::EDAttr EDAttributes, MCSectionGOFF *Parent)
-      : MCSection(SV_GOFF, Name, K.isText(), IsVirtual, nullptr),
-        Parent(Parent), EDAttributes(EDAttributes),
-        SymbolType(GOFF::ESD_ST_ElementDefinition), IsBSS(K.isBSS()),
-        RequiresNonZeroLength(0), Emitted(0) {}
+      : MCSection(Name, K.isText(), IsVirtual, nullptr), Parent(Parent),
+        EDAttributes(EDAttributes), SymbolType(GOFF::ESD_ST_ElementDefinition),
+        IsBSS(K.isBSS()), RequiresNonZeroLength(0), Emitted(0) {}
 
   MCSectionGOFF(StringRef Name, SectionKind K, bool IsVirtual,
                 GOFF::PRAttr PRAttributes, MCSectionGOFF *Parent)
-      : MCSection(SV_GOFF, Name, K.isText(), IsVirtual, nullptr),
-        Parent(Parent), PRAttributes(PRAttributes),
-        SymbolType(GOFF::ESD_ST_PartReference), IsBSS(K.isBSS()),
-        RequiresNonZeroLength(0), Emitted(0) {}
+      : MCSection(Name, K.isText(), IsVirtual, nullptr), Parent(Parent),
+        PRAttributes(PRAttributes), SymbolType(GOFF::ESD_ST_PartReference),
+        IsBSS(K.isBSS()), RequiresNonZeroLength(0), Emitted(0) {}
 
 public:
-  void printSwitchToSection(const MCAsmInfo &MAI, const Triple &T,
-                            raw_ostream &OS,
-                            uint32_t Subsection) const override;
-
-  bool useCodeAlign() const override { return false; }
-
   // Return the parent section.
   MCSectionGOFF *getParent() const { return Parent; }
 
@@ -123,8 +115,6 @@ public:
   bool requiresNonZeroLength() const { return RequiresNonZeroLength; }
 
   void setName(StringRef SectionName) { Name = SectionName; }
-
-  static bool classof(const MCSection *S) { return S->getVariant() == SV_GOFF; }
 };
 } // end namespace llvm
 
diff --git a/llvm/include/llvm/MC/MCSectionMachO.h b/llvm/include/llvm/MC/MCSectionMachO.h
index 4312175..a65d7e0 100644
--- a/llvm/include/llvm/MC/MCSectionMachO.h
+++ b/llvm/include/llvm/MC/MCSectionMachO.h
@@ -23,6 +23,8 @@ namespace llvm {
 /// This represents a section on a Mach-O system (used by Mac OS X).  On a Mac
 /// system, these are also described in /usr/include/mach-o/loader.h.
 class LLVM_ABI MCSectionMachO final : public MCSection {
+  friend class MCContext;
+  friend class MCAsmInfoDarwin;
   char SegmentName[16];  // Not necessarily null terminated!
 
   /// This is the SECTION_TYPE and SECTION_ATTRIBUTES field of a section, drawn
@@ -42,7 +44,6 @@ class LLVM_ABI MCSectionMachO final : public MCSection {
 
   MCSectionMachO(StringRef Segment, StringRef Section, unsigned TAA,
                  unsigned reserved2, SectionKind K, MCSymbol *Begin);
-  friend class MCContext;
 public:
 
   StringRef getSegmentName() const {
@@ -76,21 +77,12 @@ public:
                                      bool &TAAParsed,     // Out.
                                      unsigned &StubSize); // Out.
 
-  void printSwitchToSection(const MCAsmInfo &MAI, const Triple &T,
-                            raw_ostream &OS,
-                            uint32_t Subsection) const override;
-  bool useCodeAlign() const override;
-
   void allocAtoms();
   const MCSymbol *getAtom(size_t I) const;
   void setAtom(size_t I, const MCSymbol *Sym);
 
   unsigned getLayoutOrder() const { return LayoutOrder; }
   void setLayoutOrder(unsigned Value) { LayoutOrder = Value; }
-
-  static bool classof(const MCSection *S) {
-    return S->getVariant() == SV_MachO;
-  }
 };
 
 } // end namespace llvm
diff --git a/llvm/include/llvm/MC/MCSectionSPIRV.h b/llvm/include/llvm/MC/MCSectionSPIRV.h
index 091114a..6850965 100644
--- a/llvm/include/llvm/MC/MCSectionSPIRV.h
+++ b/llvm/include/llvm/MC/MCSectionSPIRV.h
@@ -18,22 +18,13 @@
 
 namespace llvm {
 
-class MCSymbol;
-
 class MCSectionSPIRV final : public MCSection {
   friend class MCContext;
 
   MCSectionSPIRV()
-      : MCSection(SV_SPIRV, "", /*IsText=*/true, /*IsVirtual=*/false,
+      : MCSection("", /*IsText=*/true, /*IsVirtual=*/false,
                   /*Begin=*/nullptr) {}
   // TODO: Add StringRef Name to MCSectionSPIRV.
-
-public:
-  ~MCSectionSPIRV() = default;
-  void printSwitchToSection(const MCAsmInfo &MAI, const Triple &T,
-                            raw_ostream &OS,
-                            uint32_t Subsection) const override {}
-  bool useCodeAlign() const override { return false; }
 };
 
 } // end namespace llvm
diff --git a/llvm/include/llvm/MC/MCSectionWasm.h b/llvm/include/llvm/MC/MCSectionWasm.h
index 4523818..5ec01ed 100644
--- a/llvm/include/llvm/MC/MCSectionWasm.h
+++ b/llvm/include/llvm/MC/MCSectionWasm.h
@@ -49,26 +49,18 @@ class MCSectionWasm final : public MCSection {
 
   // The storage of Name is owned by MCContext's WasmUniquingMap.
   friend class MCContext;
+  friend class MCAsmInfoWasm;
   MCSectionWasm(StringRef Name, SectionKind K, unsigned SegmentFlags,
                 const MCSymbolWasm *Group, unsigned UniqueID, MCSymbol *Begin)
-      : MCSection(SV_Wasm, Name, K.isText(), /*IsVirtual=*/false, Begin),
+      : MCSection(Name, K.isText(), /*IsVirtual=*/false, Begin),
         UniqueID(UniqueID), Group(Group),
         IsWasmData(K.isReadOnly() || K.isWriteable()),
         IsMetadata(K.isMetadata()), SegmentFlags(SegmentFlags) {}
 
 public:
-  /// Decides whether a '.section' directive should be printed before the
-  /// section name
-  bool shouldOmitSectionDirective(StringRef Name, const MCAsmInfo &MAI) const;
-
   const MCSymbolWasm *getGroup() const { return Group; }
   unsigned getSegmentFlags() const { return SegmentFlags; }
 
-  void printSwitchToSection(const MCAsmInfo &MAI, const Triple &T,
-                            raw_ostream &OS,
-                            uint32_t Subsection) const override;
-  bool useCodeAlign() const override;
-
   bool isWasmData() const { return IsWasmData; }
   bool isMetadata() const { return IsMetadata; }
 
@@ -89,7 +81,6 @@ public:
     assert(isWasmData());
     IsPassive = V;
   }
-  static bool classof(const MCSection *S) { return S->getVariant() == SV_Wasm; }
 };
 
 } // end namespace llvm
diff --git a/llvm/include/llvm/MC/MCSectionXCOFF.h b/llvm/include/llvm/MC/MCSectionXCOFF.h
index 499df6b5..0571f95 100644
--- a/llvm/include/llvm/MC/MCSectionXCOFF.h
+++ b/llvm/include/llvm/MC/MCSectionXCOFF.h
@@ -31,6 +31,7 @@ namespace llvm {
 //    implemented yet.
 class MCSectionXCOFF final : public MCSection {
   friend class MCContext;
+  friend class MCAsmInfoXCOFF;
 
   std::optional<XCOFF::CsectProperties> CsectProp;
   MCSymbolXCOFF *const QualName;
@@ -46,7 +47,7 @@ class MCSectionXCOFF final : public MCSection {
                  XCOFF::SymbolType ST, SectionKind K, MCSymbolXCOFF *QualName,
                  MCSymbol *Begin, StringRef SymbolTableName,
                  bool MultiSymbolsAllowed)
-      : MCSection(SV_XCOFF, Name, K.isText(),
+      : MCSection(Name, K.isText(),
                   /*IsVirtual=*/ST == XCOFF::XTY_CM && SMC != XCOFF::XMC_TD,
                   Begin),
         CsectProp(XCOFF::CsectProperties(SMC, ST)), QualName(QualName),
@@ -77,7 +78,7 @@ class MCSectionXCOFF final : public MCSection {
                  XCOFF::DwarfSectionSubtypeFlags DwarfSubtypeFlags,
                  MCSymbol *Begin, StringRef SymbolTableName,
                  bool MultiSymbolsAllowed)
-      : MCSection(SV_XCOFF, Name, K.isText(), /*IsVirtual=*/false, Begin),
+      : MCSection(Name, K.isText(), /*IsVirtual=*/false, Begin),
         QualName(QualName), SymbolTableName(SymbolTableName),
         DwarfSubtypeFlags(DwarfSubtypeFlags),
         MultiSymbolsAllowed(MultiSymbolsAllowed), Kind(K) {
@@ -95,10 +96,6 @@ class MCSectionXCOFF final : public MCSection {
 public:
   ~MCSectionXCOFF();
 
-  static bool classof(const MCSection *S) {
-    return S->getVariant() == SV_XCOFF;
-  }
-
   XCOFF::StorageMappingClass getMappingClass() const {
     assert(isCsect() && "Only csect section has mapping class property!");
     return CsectProp->MappingClass;
@@ -115,10 +112,6 @@ public:
   }
   MCSymbolXCOFF *getQualNameSymbol() const { return QualName; }
 
-  void printSwitchToSection(const MCAsmInfo &MAI, const Triple &T,
-                            raw_ostream &OS,
-                            uint32_t Subsection) const override;
-  bool useCodeAlign() const override;
   StringRef getSymbolTableName() const { return SymbolTableName; }
   void setSymbolTableName(StringRef STN) { SymbolTableName = STN; }
   bool isMultiSymbolsAllowed() const { return MultiSymbolsAllowed; }
diff --git a/llvm/include/llvm/MC/MCStreamer.h b/llvm/include/llvm/MC/MCStreamer.h
index 8ccce80..79c715e 100644
--- a/llvm/include/llvm/MC/MCStreamer.h
+++ b/llvm/include/llvm/MC/MCStreamer.h
@@ -459,7 +459,7 @@ public:
   bool switchSection(MCSection *Section, const MCExpr *);
 
   /// Similar to switchSection, but does not print the section directive.
-  virtual void switchSectionNoPrint(MCSection *Section);
+  void switchSectionNoPrint(MCSection *Section);
 
   /// Create the default sections and set the initial one.
   virtual void initSections(bool NoExecStack, const MCSubtargetInfo &STI);
diff --git a/llvm/include/llvm/MC/MCSymbolELF.h b/llvm/include/llvm/MC/MCSymbolELF.h
index eba9964..7c271e7 100644
--- a/llvm/include/llvm/MC/MCSymbolELF.h
+++ b/llvm/include/llvm/MC/MCSymbolELF.h
@@ -13,6 +13,7 @@
 
 namespace llvm {
 class MCSymbolELF : public MCSymbol {
+  friend class MCAsmInfoELF;
   /// An expression describing how to calculate the size of a symbol. If a
   /// symbol has no size this field will be NULL.
   const MCExpr *SymbolSize = nullptr;
diff --git a/llvm/include/llvm/Support/DebugLog.h b/llvm/include/llvm/Support/DebugLog.h
index 9556bf2..19d3098 100644
--- a/llvm/include/llvm/Support/DebugLog.h
+++ b/llvm/include/llvm/Support/DebugLog.h
@@ -26,35 +26,107 @@ namespace llvm {
 //              << "] " << "Bitset contains: " << Bitset << "\n");
 #define LDBG() DEBUGLOG_WITH_STREAM_AND_TYPE(llvm::dbgs(), DEBUG_TYPE)
 
-#define DEBUGLOG_WITH_STREAM_AND_TYPE(STREAM, TYPE)                            \
+#define DEBUGLOG_WITH_STREAM_TYPE_AND_FILE(STREAM, TYPE, FILE)                 \
   for (bool _c = (::llvm::DebugFlag && ::llvm::isCurrentDebugType(TYPE)); _c;  \
        _c = false)                                                             \
-  ::llvm::impl::LogWithNewline(TYPE, __FILE__, __LINE__, (STREAM))
+  ::llvm::impl::raw_ldbg_ostream{                                              \
+      ::llvm::impl::computePrefix(TYPE, FILE, __LINE__), (STREAM)}             \
+      .asLvalue()
+// When __SHORT_FILE__ is not defined, the File is the full path,
+// otherwise __SHORT_FILE__ is defined in CMake to provide the file name
+// without the path prefix.
+#if defined(__SHORT_FILE__)
+#define DEBUGLOG_WITH_STREAM_AND_TYPE(STREAM, TYPE)                            \
+  DEBUGLOG_WITH_STREAM_TYPE_AND_FILE(STREAM, TYPE, __SHORT_FILE__)
+#else
+#define DEBUGLOG_WITH_STREAM_AND_TYPE(STREAM, TYPE)                            \
+  DEBUGLOG_WITH_STREAM_TYPE_AND_FILE(                                          \
+      STREAM, TYPE, ::llvm::impl::LogWithNewline::getShortFileName(__FILE__))
+#endif
 
 namespace impl {
-class LogWithNewline {
+
+/// A raw_ostream that tracks `\n` and print the prefix.
+class LLVM_ABI raw_ldbg_ostream final : public raw_ostream {
+  std::string Prefix;
+  raw_ostream &Os;
+  bool HasPendingNewline = true;
+
+  /// Split the line on newlines and insert the prefix before each newline.
+  /// Forward everything to the underlying stream.
+  void write_impl(const char *Ptr, size_t Size) final {
+    auto Str = StringRef(Ptr, Size);
+    // Handle the initial prefix.
+    if (!Str.empty())
+      writeWithPrefix(StringRef());
+
+    auto Eol = Str.find('\n');
+    while (Eol != StringRef::npos) {
+      StringRef Line = Str.take_front(Eol + 1);
+      if (!Line.empty())
+        writeWithPrefix(Line);
+      HasPendingNewline = true;
+      Str = Str.drop_front(Eol + 1);
+      Eol = Str.find('\n');
+    }
+    if (!Str.empty())
+      writeWithPrefix(Str);
+  }
+  void emitPrefix() { Os.write(Prefix.c_str(), Prefix.size()); }
+  void writeWithPrefix(StringRef Str) {
+    if (HasPendingNewline) {
+      emitPrefix();
+      HasPendingNewline = false;
+    }
+    Os.write(Str.data(), Str.size());
+  }
+
 public:
-  LogWithNewline(const char *debug_type, const char *file, int line,
-                 raw_ostream &os)
-      : os(os) {
-    if (debug_type)
-      os << debug_type << " ";
-    os << "[" << file << ":" << line << "] ";
+  explicit raw_ldbg_ostream(std::string Prefix, raw_ostream &Os)
+      : Prefix(std::move(Prefix)), Os(Os) {
+    SetUnbuffered();
+  }
+  ~raw_ldbg_ostream() final {
+    flushEol();
+    Os << '\n';
   }
-  ~LogWithNewline() { os << '\n'; }
-  template <typename T> raw_ostream &operator<<(const T &t) && {
-    return os << t;
+  void flushEol() {
+    if (HasPendingNewline) {
+      emitPrefix();
+      HasPendingNewline = false;
+    }
   }
 
-  // Prevent copying, as this class manages newline responsibility and is
-  // intended for use as a temporary.
-  LogWithNewline(const LogWithNewline &) = delete;
-  LogWithNewline &operator=(const LogWithNewline &) = delete;
-  LogWithNewline &operator=(LogWithNewline &&) = delete;
+  /// Forward the current_pos method to the underlying stream.
+  uint64_t current_pos() const final { return Os.tell(); }
 
-private:
-  raw_ostream &os;
+  /// Some of the `<<` operators expect an lvalue, so we trick the type system.
+  raw_ldbg_ostream &asLvalue() { return *this; }
 };
+
+/// Remove the path prefix from the file name.
+static LLVM_ATTRIBUTE_UNUSED constexpr const char *
+getShortFileName(const char *path) {
+  const char *filename = path;
+  for (const char *p = path; *p != '\0'; ++p) {
+    if (*p == '/' || *p == '\\')
+      filename = p + 1;
+  }
+  return filename;
+}
+
+/// Compute the prefix for the debug log in the form of:
+/// "[DebugType] File:Line "
+/// Where the File is the file name without the path prefix.
+static LLVM_ATTRIBUTE_UNUSED std::string
+computePrefix(const char *DebugType, const char *File, int Line) {
+  std::string Prefix;
+  raw_string_ostream OsPrefix(Prefix);
+  if (DebugType)
+    OsPrefix << "[" << DebugType << "] ";
+  OsPrefix << File << ":" << Line << " ";
+  return OsPrefix.str();
+}
 } // end namespace impl
 #else
 // As others in Debug, When compiling without assertions, the -debug-* options
diff --git a/llvm/include/llvm/Support/ThreadPool.h b/llvm/include/llvm/Support/ThreadPool.h
index 9272760..c26681c 100644
--- a/llvm/include/llvm/Support/ThreadPool.h
+++ b/llvm/include/llvm/Support/ThreadPool.h
@@ -149,10 +149,6 @@ public:
   /// number of threads!
   unsigned getMaxConcurrency() const override { return MaxThreadCount; }
 
-  // TODO: Remove, misleading legacy name warning!
-  LLVM_DEPRECATED("Use getMaxConcurrency instead", "getMaxConcurrency")
-  unsigned getThreadCount() const { return MaxThreadCount; }
-
   /// Returns true if the current thread is a worker thread of this thread pool.
   bool isWorkerThread() const;
 
@@ -233,10 +229,6 @@ public:
   /// Returns always 1: there is no concurrency.
   unsigned getMaxConcurrency() const override { return 1; }
 
-  // TODO: Remove, misleading legacy name warning!
-  LLVM_DEPRECATED("Use getMaxConcurrency instead", "getMaxConcurrency")
-  unsigned getThreadCount() const { return 1; }
-
   /// Returns true if the current thread is a worker thread of this thread pool.
   bool isWorkerThread() const;
 
diff --git a/llvm/include/llvm/TextAPI/SymbolSet.h b/llvm/include/llvm/TextAPI/SymbolSet.h
index cd30663..a04cb35 100644
--- a/llvm/include/llvm/TextAPI/SymbolSet.h
+++ b/llvm/include/llvm/TextAPI/SymbolSet.h
@@ -92,6 +92,7 @@ private:
 
 public:
   SymbolSet() = default;
+  ~SymbolSet();
   LLVM_ABI Symbol *addGlobal(EncodeKind Kind, StringRef Name, SymbolFlags Flags,
                              const Target &Targ);
   size_t size() const { return Symbols.size(); }
diff --git a/llvm/include/llvm/Transforms/ObjCARC.h b/llvm/include/llvm/Transforms/ObjCARC.h
index c927513..c4b4c4f 100644
--- a/llvm/include/llvm/Transforms/ObjCARC.h
+++ b/llvm/include/llvm/Transforms/ObjCARC.h
@@ -35,10 +35,6 @@ struct ObjCARCContractPass : public PassInfoMixin<ObjCARCContractPass> {
   LLVM_ABI PreservedAnalyses run(Function &F, FunctionAnalysisManager &AM);
 };
 
-struct ObjCARCAPElimPass : public PassInfoMixin<ObjCARCAPElimPass> {
-  LLVM_ABI PreservedAnalyses run(Module &M, ModuleAnalysisManager &AM);
-};
-
 struct ObjCARCExpandPass : public PassInfoMixin<ObjCARCExpandPass> {
   LLVM_ABI PreservedAnalyses run(Function &M, FunctionAnalysisManager &AM);
 };
diff --git a/llvm/lib/Analysis/ConstantFolding.cpp b/llvm/lib/Analysis/ConstantFolding.cpp
index ec78386..759c553 100644
--- a/llvm/lib/Analysis/ConstantFolding.cpp
+++ b/llvm/lib/Analysis/ConstantFolding.cpp
@@ -929,12 +929,11 @@ Constant *SymbolicallyEvaluateGEP(const GEPOperator *GEP,
     if (!AllConstantInt)
       break;
 
-    // TODO: Try to intersect two inrange attributes?
-    if (!InRange) {
-      InRange = GEP->getInRange();
-      if (InRange)
-        // Adjust inrange by offset until now.
-        InRange = InRange->sextOrTrunc(BitWidth).subtract(Offset);
+    // Adjust inrange offset and intersect inrange attributes
+    if (auto GEPRange = GEP->getInRange()) {
+      auto AdjustedGEPRange = GEPRange->sextOrTrunc(BitWidth).subtract(Offset);
+      InRange =
+          InRange ? InRange->intersectWith(AdjustedGEPRange) : AdjustedGEPRange;
     }
 
     Ptr = cast<Constant>(GEP->getOperand(0));
diff --git a/llvm/lib/Analysis/DXILResource.cpp b/llvm/lib/Analysis/DXILResource.cpp
index 2da6468..1959ab6 100644
--- a/llvm/lib/Analysis/DXILResource.cpp
+++ b/llvm/lib/Analysis/DXILResource.cpp
@@ -1079,15 +1079,16 @@ void DXILResourceBindingInfo::populate(Module &M, DXILResourceTypeMap &DRTM) {
       // add new space
       S = &BS->Spaces.emplace_back(B.Space);
 
-    // the space is full - set flag to report overlapping binding later
-    if (S->FreeRanges.empty()) {
+    // The space is full - there are no free slots left, or the rest of the
+    // slots are taken by an unbounded array. Set flag to report overlapping
+    // binding later.
+    if (S->FreeRanges.empty() || S->FreeRanges.back().UpperBound < UINT32_MAX) {
       OverlappingBinding = true;
       continue;
     }
 
     // adjust the last free range lower bound, split it in two, or remove it
     BindingRange &LastFreeRange = S->FreeRanges.back();
-    assert(LastFreeRange.UpperBound == UINT32_MAX);
     if (LastFreeRange.LowerBound == B.LowerBound) {
       if (B.UpperBound < UINT32_MAX)
         LastFreeRange.LowerBound = B.UpperBound + 1;
diff --git a/llvm/lib/Analysis/DependenceAnalysis.cpp b/llvm/lib/Analysis/DependenceAnalysis.cpp
index dd9a44b..f1473b2 100644
--- a/llvm/lib/Analysis/DependenceAnalysis.cpp
+++ b/llvm/lib/Analysis/DependenceAnalysis.cpp
@@ -3383,6 +3383,10 @@ bool DependenceInfo::tryDelinearize(Instruction *Src, Instruction *Dst,
                                     SrcSubscripts, DstSubscripts))
     return false;
 
+  assert(isLoopInvariant(SrcBase, SrcLoop) &&
+         isLoopInvariant(DstBase, DstLoop) &&
+         "Expected SrcBase and DstBase to be loop invariant");
+
   int Size = SrcSubscripts.size();
   LLVM_DEBUG({
     dbgs() << "\nSrcSubscripts: ";
@@ -3666,6 +3670,19 @@ DependenceInfo::depends(Instruction *Src, Instruction *Dst,
                                         SCEVUnionPredicate(Assume, *SE));
   }
 
+  // Even if the base pointers are the same, they may not be loop-invariant. It
+  // could lead to incorrect results, as we're analyzing loop-carried
+  // dependencies. Src and Dst can be in different loops, so we need to check
+  // the base pointer is invariant in both loops.
+  Loop *SrcLoop = LI->getLoopFor(Src->getParent());
+  Loop *DstLoop = LI->getLoopFor(Dst->getParent());
+  if (!isLoopInvariant(SrcBase, SrcLoop) ||
+      !isLoopInvariant(DstBase, DstLoop)) {
+    LLVM_DEBUG(dbgs() << "The base pointer is not loop invariant.\n");
+    return std::make_unique<Dependence>(Src, Dst,
+                                        SCEVUnionPredicate(Assume, *SE));
+  }
+
   uint64_t EltSize = SrcLoc.Size.toRaw();
   const SCEV *SrcEv = SE->getMinusSCEV(SrcSCEV, SrcBase);
   const SCEV *DstEv = SE->getMinusSCEV(DstSCEV, DstBase);
diff --git a/llvm/lib/Analysis/InstructionSimplify.cpp b/llvm/lib/Analysis/InstructionSimplify.cpp
index 82530e7..5907e21 100644
--- a/llvm/lib/Analysis/InstructionSimplify.cpp
+++ b/llvm/lib/Analysis/InstructionSimplify.cpp
@@ -5366,7 +5366,7 @@ static Value *simplifyCastInst(unsigned CastOpc, Value *Op, Type *Ty,
     Type *MidTy = CI->getType();
     Type *DstTy = Ty;
     if (Src->getType() == Ty) {
-      auto FirstOp = static_cast<Instruction::CastOps>(CI->getOpcode());
+      auto FirstOp = CI->getOpcode();
       auto SecondOp = static_cast<Instruction::CastOps>(CastOpc);
       Type *SrcIntPtrTy =
           SrcTy->isPtrOrPtrVectorTy() ? Q.DL.getIntPtrType(SrcTy) : nullptr;
diff --git a/llvm/lib/Analysis/MemoryDependenceAnalysis.cpp b/llvm/lib/Analysis/MemoryDependenceAnalysis.cpp
index 3aa9909..2b0f212 100644
--- a/llvm/lib/Analysis/MemoryDependenceAnalysis.cpp
+++ b/llvm/lib/Analysis/MemoryDependenceAnalysis.cpp
@@ -983,33 +983,37 @@ MemDepResult MemoryDependenceResults::getNonLocalInfoForBlock(
 static void
 SortNonLocalDepInfoCache(MemoryDependenceResults::NonLocalDepInfo &Cache,
                          unsigned NumSortedEntries) {
-  switch (Cache.size() - NumSortedEntries) {
-  case 0:
-    // done, no new entries.
-    break;
-  case 2: {
-    // Two new entries, insert the last one into place.
-    NonLocalDepEntry Val = Cache.back();
-    Cache.pop_back();
-    MemoryDependenceResults::NonLocalDepInfo::iterator Entry =
-        std::upper_bound(Cache.begin(), Cache.end() - 1, Val);
-    Cache.insert(Entry, Val);
-    [[fallthrough]];
+
+  // If only one entry, don't sort.
+  if (Cache.size() < 2)
+    return;
+
+  unsigned s = Cache.size() - NumSortedEntries;
+
+  // If the cache is already sorted, don't sort it again.
+  if (s == 0)
+    return;
+
+  // If no entry is sorted, sort the whole cache.
+  if (NumSortedEntries == 0) {
+    llvm::sort(Cache);
+    return;
   }
-  case 1:
-    // One new entry, Just insert the new value at the appropriate position.
-    if (Cache.size() != 1) {
+
+  // If the number of unsorted entires is small and the cache size is big, using
+  // insertion sort is faster. Here use Log2_32 to quickly choose the sort
+  // method.
+  if (s < Log2_32(Cache.size())) {
+    while (s > 0) {
       NonLocalDepEntry Val = Cache.back();
       Cache.pop_back();
       MemoryDependenceResults::NonLocalDepInfo::iterator Entry =
-          llvm::upper_bound(Cache, Val);
+          std::upper_bound(Cache.begin(), Cache.end() - s + 1, Val);
       Cache.insert(Entry, Val);
+      s--;
     }
-    break;
-  default:
-    // Added many values, do a full scale sort.
+  } else {
     llvm::sort(Cache);
-    break;
   }
 }
 
diff --git a/llvm/lib/Analysis/TargetLibraryInfo.cpp b/llvm/lib/Analysis/TargetLibraryInfo.cpp
index e475be2..6e92766 100644
--- a/llvm/lib/Analysis/TargetLibraryInfo.cpp
+++ b/llvm/lib/Analysis/TargetLibraryInfo.cpp
@@ -875,6 +875,34 @@ static void initializeLibCalls(TargetLibraryInfoImpl &TLI, const Triple &T,
     TLI.setUnavailable(LibFunc_toascii);
   }
 
+  if (T.isOSFreeBSD()) {
+    TLI.setAvailable(LibFunc_dunder_strtok_r);
+    TLI.setAvailable(LibFunc_memalign);
+    TLI.setAvailable(LibFunc_fputc_unlocked);
+    TLI.setAvailable(LibFunc_fputs_unlocked);
+    TLI.setAvailable(LibFunc_fread_unlocked);
+    TLI.setAvailable(LibFunc_fwrite_unlocked);
+    TLI.setAvailable(LibFunc_getc_unlocked);
+    TLI.setAvailable(LibFunc_getchar_unlocked);
+    TLI.setAvailable(LibFunc_putc_unlocked);
+    TLI.setAvailable(LibFunc_putchar_unlocked);
+
+    TLI.setUnavailable(LibFunc___kmpc_alloc_shared);
+    TLI.setUnavailable(LibFunc___kmpc_free_shared);
+    TLI.setUnavailable(LibFunc_dunder_strndup);
+    TLI.setUnavailable(LibFunc_memccpy_chk);
+    TLI.setUnavailable(LibFunc_strlen_chk);
+    TLI.setUnavailable(LibFunc_fmaximum_num);
+    TLI.setUnavailable(LibFunc_fmaximum_numf);
+    TLI.setUnavailable(LibFunc_fmaximum_numl);
+    TLI.setUnavailable(LibFunc_fminimum_num);
+    TLI.setUnavailable(LibFunc_fminimum_numf);
+    TLI.setUnavailable(LibFunc_fminimum_numl);
+    TLI.setUnavailable(LibFunc_roundeven);
+    TLI.setUnavailable(LibFunc_roundevenf);
+    TLI.setUnavailable(LibFunc_roundevenl);
+  }
+
   // As currently implemented in clang, NVPTX code has no standard library to
   // speak of.  Headers provide a standard-ish library implementation, but many
   // of the signatures are wrong -- for example, many libm functions are not
diff --git a/llvm/lib/CodeGen/AsmPrinter/AIXException.cpp b/llvm/lib/CodeGen/AsmPrinter/AIXException.cpp
index 5d7c97a..6356d71 100644
--- a/llvm/lib/CodeGen/AsmPrinter/AIXException.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/AIXException.cpp
@@ -37,8 +37,8 @@ void AIXException::emitExceptionInfoTable(const MCSymbol *LSDA,
   //   unsigned long personality;  /* Pointer to the personality routine */
   //   }
 
-  auto *EHInfo =
-      cast<MCSectionXCOFF>(Asm->getObjFileLowering().getCompactUnwindSection());
+  auto *EHInfo = static_cast<MCSectionXCOFF *>(
+      Asm->getObjFileLowering().getCompactUnwindSection());
   if (Asm->TM.getFunctionSections()) {
     // If option -ffunction-sections is on, append the function name to the
     // name of EH Info Table csect so that each function has its own EH Info
diff --git a/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp b/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
index f1d3e96..6166271 100644
--- a/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/AsmPrinter.cpp
@@ -4221,10 +4221,11 @@ MCSymbol *AsmPrinter::GetCPISymbol(unsigned CPID) const {
       SectionKind Kind = CPE.getSectionKind(&DL);
       const Constant *C = CPE.Val.ConstVal;
       Align Alignment = CPE.Alignment;
-      if (const MCSectionCOFF *S = dyn_cast<MCSectionCOFF>(
-              getObjFileLowering().getSectionForConstant(DL, Kind, C,
-                                                         Alignment))) {
-        if (MCSymbol *Sym = S->getCOMDATSymbol()) {
+      auto *S =
+          getObjFileLowering().getSectionForConstant(DL, Kind, C, Alignment);
+      if (S && TM.getTargetTriple().isOSBinFormatCOFF()) {
+        if (MCSymbol *Sym =
+                static_cast<const MCSectionCOFF *>(S)->getCOMDATSymbol()) {
           if (Sym->isUndefined())
             OutStreamer->emitSymbolAttribute(Sym, MCSA_Global);
           return Sym;
diff --git a/llvm/lib/CodeGen/AsmPrinter/CodeViewDebug.cpp b/llvm/lib/CodeGen/AsmPrinter/CodeViewDebug.cpp
index 8abeb56..c5d6e40 100644
--- a/llvm/lib/CodeGen/AsmPrinter/CodeViewDebug.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/CodeViewDebug.cpp
@@ -1051,10 +1051,10 @@ void CodeViewDebug::switchToDebugSectionForSymbol(const MCSymbol *GVSym) {
   // comdat key. A section may be comdat because of -ffunction-sections or
   // because it is comdat in the IR.
   MCSectionCOFF *GVSec =
-      GVSym ? dyn_cast<MCSectionCOFF>(&GVSym->getSection()) : nullptr;
+      GVSym ? static_cast<MCSectionCOFF *>(&GVSym->getSection()) : nullptr;
   const MCSymbol *KeySym = GVSec ? GVSec->getCOMDATSymbol() : nullptr;
 
-  MCSectionCOFF *DebugSec = cast<MCSectionCOFF>(
+  auto *DebugSec = static_cast<MCSectionCOFF *>(
       CompilerInfoAsm->getObjFileLowering().getCOFFDebugSymbolsSection());
   DebugSec = OS.getContext().getAssociativeCOFFSection(DebugSec, KeySym);
 
diff --git a/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp b/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
index 11b8576..7188833 100644
--- a/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
+++ b/llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp
@@ -972,10 +972,9 @@ void DwarfDebug::constructCallSiteEntryDIEs(const DISubprogram &SP,
       // the call graph which could lead to some target function. For tail
       // calls, no return PC information is needed, unless tuning for GDB in
       // DWARF4 mode in which case we fake a return PC for compatibility.
-      const MCSymbol *PCAddr =
-          (!IsTail || CU.useGNUAnalogForDwarf5Feature())
-              ? const_cast<MCSymbol *>(getLabelAfterInsn(TopLevelCallMI))
-              : nullptr;
+      const MCSymbol *PCAddr = (!IsTail || CU.useGNUAnalogForDwarf5Feature())
+                                   ? getLabelAfterInsn(TopLevelCallMI)
+                                   : nullptr;
 
       // For tail calls, it's necessary to record the address of the branch
       // instruction so that the debugger can show where the tail call occurred.
diff --git a/llvm/lib/CodeGen/CodeGenPrepare.cpp b/llvm/lib/CodeGen/CodeGenPrepare.cpp
index c21058c..416c56d 100644
--- a/llvm/lib/CodeGen/CodeGenPrepare.cpp
+++ b/llvm/lib/CodeGen/CodeGenPrepare.cpp
@@ -2095,6 +2095,10 @@ static bool isRemOfLoopIncrementWithLoopInvariant(
   if (!L->isLoopInvariant(RemAmt))
     return false;
 
+  // Only works if the AddOffset is a loop invaraint
+  if (AddOffset && !L->isLoopInvariant(AddOffset))
+    return false;
+
   // Is the PHI a loop increment?
   auto LoopIncrInfo = getIVIncrement(PN, LI);
   if (!LoopIncrInfo)
diff --git a/llvm/lib/CodeGen/InterleavedAccessPass.cpp b/llvm/lib/CodeGen/InterleavedAccessPass.cpp
index c2839d4..5e50898 100644
--- a/llvm/lib/CodeGen/InterleavedAccessPass.cpp
+++ b/llvm/lib/CodeGen/InterleavedAccessPass.cpp
@@ -634,6 +634,9 @@ bool InterleavedAccessImpl::lowerDeinterleaveIntrinsic(
                       << " and factor = " << Factor << "\n");
   } else {
     assert(II);
+    if (II->getIntrinsicID() != Intrinsic::masked_load &&
+        II->getIntrinsicID() != Intrinsic::vp_load)
+      return false;
 
     // Check mask operand. Handle both all-true/false and interleaved mask.
     Mask = getMask(getMaskOperand(II), Factor, getDeinterleavedVectorType(DI));
@@ -673,6 +676,9 @@ bool InterleavedAccessImpl::lowerInterleaveIntrinsic(
 
   Value *Mask = nullptr;
   if (II) {
+    if (II->getIntrinsicID() != Intrinsic::masked_store &&
+        II->getIntrinsicID() != Intrinsic::vp_store)
+      return false;
     // Check mask operand. Handle both all-true/false and interleaved mask.
     Mask = getMask(getMaskOperand(II), Factor,
                    cast<VectorType>(InterleaveValues[0]->getType()));
diff --git a/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp b/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp
index e5704c0..583a85a 100644
--- a/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp
+++ b/llvm/lib/CodeGen/SelectionDAG/LegalizeIntegerTypes.cpp
@@ -22,6 +22,7 @@
 #include "llvm/CodeGen/StackMaps.h"
 #include "llvm/CodeGen/TargetLowering.h"
 #include "llvm/IR/DerivedTypes.h"
+#include "llvm/IR/DiagnosticInfo.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/KnownBits.h"
 #include "llvm/Support/raw_ostream.h"
@@ -357,6 +358,9 @@ void DAGTypeLegalizer::PromoteIntegerResult(SDNode *N, unsigned ResNo) {
   case ISD::PATCHPOINT:
     Res = PromoteIntRes_PATCHPOINT(N);
     break;
+  case ISD::READ_REGISTER:
+    Res = PromoteIntRes_READ_REGISTER(N);
+    break;
   }
 
   // If the result is null then the sub-method took care of registering it.
@@ -2076,6 +2080,9 @@ bool DAGTypeLegalizer::PromoteIntegerOperand(SDNode *N, unsigned OpNo) {
   case ISD::PATCHPOINT:
     Res = PromoteIntOp_PATCHPOINT(N, OpNo);
     break;
+  case ISD::WRITE_REGISTER:
+    Res = PromoteIntOp_WRITE_REGISTER(N, OpNo);
+    break;
   case ISD::EXPERIMENTAL_VP_STRIDED_LOAD:
   case ISD::EXPERIMENTAL_VP_STRIDED_STORE:
     Res = PromoteIntOp_VP_STRIDED(N, OpNo);
@@ -2853,6 +2860,15 @@ SDValue DAGTypeLegalizer::PromoteIntOp_PATCHPOINT(SDNode *N, unsigned OpNo) {
   return SDValue(DAG.UpdateNodeOperands(N, NewOps), 0);
 }
 
+SDValue DAGTypeLegalizer::PromoteIntOp_WRITE_REGISTER(SDNode *N,
+                                                      unsigned OpNo) {
+  const Function &Fn = DAG.getMachineFunction().getFunction();
+  Fn.getContext().diagnose(DiagnosticInfoLegalizationFailure(
+      "cannot use llvm.write_register with illegal type", Fn,
+      N->getDebugLoc()));
+  return N->getOperand(0);
+}
+
 SDValue DAGTypeLegalizer::PromoteIntOp_VP_STRIDED(SDNode *N, unsigned OpNo) {
   assert((N->getOpcode() == ISD::EXPERIMENTAL_VP_STRIDED_LOAD && OpNo == 3) ||
          (N->getOpcode() == ISD::EXPERIMENTAL_VP_STRIDED_STORE && OpNo == 4));
@@ -3127,6 +3143,10 @@ void DAGTypeLegalizer::ExpandIntegerResult(SDNode *N, unsigned ResNo) {
   case ISD::VSCALE:
     ExpandIntRes_VSCALE(N, Lo, Hi);
     break;
+
+  case ISD::READ_REGISTER:
+    ExpandIntRes_READ_REGISTER(N, Lo, Hi);
+    break;
   }
 
   // If Lo/Hi is null, the sub-method took care of registering results etc.
@@ -5471,6 +5491,18 @@ void DAGTypeLegalizer::ExpandIntRes_VSCALE(SDNode *N, SDValue &Lo,
   SplitInteger(Res, Lo, Hi);
 }
 
+void DAGTypeLegalizer::ExpandIntRes_READ_REGISTER(SDNode *N, SDValue &Lo,
+                                                  SDValue &Hi) {
+  const Function &Fn = DAG.getMachineFunction().getFunction();
+  Fn.getContext().diagnose(DiagnosticInfoLegalizationFailure(
+      "cannot use llvm.read_register with illegal type", Fn, N->getDebugLoc()));
+  ReplaceValueWith(SDValue(N, 1), N->getOperand(0));
+  EVT LoVT, HiVT;
+  std::tie(LoVT, HiVT) = DAG.GetSplitDestVTs(N->getValueType(0));
+  Lo = DAG.getPOISON(LoVT);
+  Hi = DAG.getPOISON(HiVT);
+}
+
 //===----------------------------------------------------------------------===//
 //  Integer Operand Expansion
 //===----------------------------------------------------------------------===//
@@ -5537,6 +5569,9 @@ bool DAGTypeLegalizer::ExpandIntegerOperand(SDNode *N, unsigned OpNo) {
   case ISD::EXPERIMENTAL_VP_STRIDED_STORE:
     Res = ExpandIntOp_VP_STRIDED(N, OpNo);
     break;
+  case ISD::WRITE_REGISTER:
+    Res = ExpandIntOp_WRITE_REGISTER(N, OpNo);
+    break;
   }
 
   // If the result is null, the sub-method took care of registering results etc.
@@ -5935,6 +5970,15 @@ SDValue DAGTypeLegalizer::ExpandIntOp_VP_STRIDED(SDNode *N, unsigned OpNo) {
   return SDValue(DAG.UpdateNodeOperands(N, NewOps), 0);
 }
 
+SDValue DAGTypeLegalizer::ExpandIntOp_WRITE_REGISTER(SDNode *N, unsigned OpNo) {
+  const Function &Fn = DAG.getMachineFunction().getFunction();
+  Fn.getContext().diagnose(DiagnosticInfoLegalizationFailure(
+      "cannot use llvm.write_register with illegal type", Fn,
+      N->getDebugLoc()));
+
+  return N->getOperand(0);
+}
+
 SDValue DAGTypeLegalizer::PromoteIntRes_VECTOR_SPLICE(SDNode *N) {
   SDLoc dl(N);
 
@@ -6332,6 +6376,16 @@ SDValue DAGTypeLegalizer::PromoteIntRes_PATCHPOINT(SDNode *N) {
   return Res.getValue(0);
 }
 
+SDValue DAGTypeLegalizer::PromoteIntRes_READ_REGISTER(SDNode *N) {
+  const Function &Fn = DAG.getMachineFunction().getFunction();
+  Fn.getContext().diagnose(DiagnosticInfoLegalizationFailure(
+      "cannot use llvm.read_register with illegal type", Fn, N->getDebugLoc()));
+
+  EVT NVT = TLI.getTypeToTransformTo(*DAG.getContext(), N->getValueType(0));
+  ReplaceValueWith(SDValue(N, 1), N->getOperand(0));
+  return DAG.getPOISON(NVT);
+}
+
 SDValue DAGTypeLegalizer::PromoteIntOp_EXTRACT_VECTOR_ELT(SDNode *N) {
   SDLoc dl(N);
   SDValue V0 = GetPromotedInteger(N->getOperand(0));
diff --git a/llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h b/llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h
index 9b53724..2e13b18 100644
--- a/llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h
+++ b/llvm/lib/CodeGen/SelectionDAG/LegalizeTypes.h
@@ -378,6 +378,7 @@ private:
   SDValue PromoteIntRes_VPFunnelShift(SDNode *N);
   SDValue PromoteIntRes_IS_FPCLASS(SDNode *N);
   SDValue PromoteIntRes_PATCHPOINT(SDNode *N);
+  SDValue PromoteIntRes_READ_REGISTER(SDNode *N);
   SDValue PromoteIntRes_VECTOR_FIND_LAST_ACTIVE(SDNode *N);
   SDValue PromoteIntRes_GET_ACTIVE_LANE_MASK(SDNode *N);
   SDValue PromoteIntRes_PARTIAL_REDUCE_MLA(SDNode *N);
@@ -428,6 +429,7 @@ private:
   SDValue PromoteIntOp_SET_ROUNDING(SDNode *N);
   SDValue PromoteIntOp_STACKMAP(SDNode *N, unsigned OpNo);
   SDValue PromoteIntOp_PATCHPOINT(SDNode *N, unsigned OpNo);
+  SDValue PromoteIntOp_WRITE_REGISTER(SDNode *N, unsigned OpNo);
   SDValue PromoteIntOp_VP_STRIDED(SDNode *N, unsigned OpNo);
   SDValue PromoteIntOp_VP_SPLICE(SDNode *N, unsigned OpNo);
   SDValue PromoteIntOp_VECTOR_HISTOGRAM(SDNode *N, unsigned OpNo);
@@ -511,6 +513,7 @@ private:
   void ExpandIntRes_FunnelShift       (SDNode *N, SDValue &Lo, SDValue &Hi);
 
   void ExpandIntRes_VSCALE            (SDNode *N, SDValue &Lo, SDValue &Hi);
+  void ExpandIntRes_READ_REGISTER(SDNode *N, SDValue &Lo, SDValue &Hi);
 
   void ExpandShiftByConstant(SDNode *N, const APInt &Amt,
                              SDValue &Lo, SDValue &Hi);
@@ -534,6 +537,7 @@ private:
   SDValue ExpandIntOp_STACKMAP(SDNode *N, unsigned OpNo);
   SDValue ExpandIntOp_PATCHPOINT(SDNode *N, unsigned OpNo);
   SDValue ExpandIntOp_VP_STRIDED(SDNode *N, unsigned OpNo);
+  SDValue ExpandIntOp_WRITE_REGISTER(SDNode *N, unsigned OpNo);
 
   void IntegerExpandSetCCOperands(SDValue &NewLHS, SDValue &NewRHS,
                                   ISD::CondCode &CCCode, const SDLoc &dl);
diff --git a/llvm/lib/CodeGen/TargetLoweringBase.cpp b/llvm/lib/CodeGen/TargetLoweringBase.cpp
index d4a3455..68b8a00 100644
--- a/llvm/lib/CodeGen/TargetLoweringBase.cpp
+++ b/llvm/lib/CodeGen/TargetLoweringBase.cpp
@@ -806,7 +806,17 @@ void TargetLoweringBase::initActions() {
                         ISD::SDIVFIX,        ISD::SDIVFIXSAT,
                         ISD::UDIVFIX,        ISD::UDIVFIXSAT,
                         ISD::FP_TO_SINT_SAT, ISD::FP_TO_UINT_SAT,
-                        ISD::IS_FPCLASS},
+                        ISD::IS_FPCLASS,     ISD::FCBRT,
+                        ISD::FLOG,           ISD::FLOG2,
+                        ISD::FLOG10,         ISD::FEXP,
+                        ISD::FEXP2,          ISD::FEXP10,
+                        ISD::FFLOOR,         ISD::FNEARBYINT,
+                        ISD::FCEIL,          ISD::FRINT,
+                        ISD::FTRUNC,         ISD::FROUNDEVEN,
+                        ISD::FTAN,           ISD::FACOS,
+                        ISD::FASIN,          ISD::FATAN,
+                        ISD::FCOSH,          ISD::FSINH,
+                        ISD::FTANH,          ISD::FATAN2},
                        VT, Expand);
 
     // Overflow operations default to expand
@@ -852,13 +862,12 @@ void TargetLoweringBase::initActions() {
 
     // These operations default to expand for vector types.
     if (VT.isVector())
-      setOperationAction(
-          {ISD::FCOPYSIGN, ISD::SIGN_EXTEND_INREG, ISD::ANY_EXTEND_VECTOR_INREG,
-           ISD::SIGN_EXTEND_VECTOR_INREG, ISD::ZERO_EXTEND_VECTOR_INREG,
-           ISD::SPLAT_VECTOR, ISD::LRINT, ISD::LLRINT, ISD::LROUND,
-           ISD::LLROUND, ISD::FTAN, ISD::FACOS, ISD::FASIN, ISD::FATAN,
-           ISD::FCOSH, ISD::FSINH, ISD::FTANH, ISD::FATAN2},
-          VT, Expand);
+      setOperationAction({ISD::FCOPYSIGN, ISD::SIGN_EXTEND_INREG,
+                          ISD::ANY_EXTEND_VECTOR_INREG,
+                          ISD::SIGN_EXTEND_VECTOR_INREG,
+                          ISD::ZERO_EXTEND_VECTOR_INREG, ISD::SPLAT_VECTOR,
+                          ISD::LRINT, ISD::LLRINT, ISD::LROUND, ISD::LLROUND},
+                         VT, Expand);
 
       // Constrained floating-point operations default to expand.
 #define DAG_INSTRUCTION(NAME, NARG, ROUND_MODE, INTRINSIC, DAGN)               \
@@ -914,15 +923,6 @@ void TargetLoweringBase::initActions() {
                      {MVT::bf16, MVT::f16, MVT::f32, MVT::f64, MVT::f80, MVT::f128},
                      Expand);
 
-  // These library functions default to expand.
-  setOperationAction({ISD::FCBRT,      ISD::FLOG,  ISD::FLOG2,  ISD::FLOG10,
-                      ISD::FEXP,       ISD::FEXP2, ISD::FEXP10, ISD::FFLOOR,
-                      ISD::FNEARBYINT, ISD::FCEIL, ISD::FRINT,  ISD::FTRUNC,
-                      ISD::FROUNDEVEN, ISD::FTAN,  ISD::FACOS,  ISD::FASIN,
-                      ISD::FATAN,      ISD::FCOSH, ISD::FSINH,  ISD::FTANH,
-                      ISD::FATAN2},
-                     {MVT::f32, MVT::f64, MVT::f128}, Expand);
-
   // Insert custom handling default for llvm.canonicalize.*.
   setOperationAction(ISD::FCANONICALIZE,
                      {MVT::f16, MVT::f32, MVT::f64, MVT::f128}, Expand);
diff --git a/llvm/lib/CodeGen/TargetLoweringObjectFileImpl.cpp b/llvm/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
index a40ceaa..725e951 100644
--- a/llvm/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
+++ b/llvm/lib/CodeGen/TargetLoweringObjectFileImpl.cpp
@@ -995,7 +995,7 @@ MCSection *TargetLoweringObjectFileELF::getSectionForLSDA(
   if (!LSDASection || (!F.hasComdat() && !TM.getFunctionSections()))
     return LSDASection;
 
-  const auto *LSDA = cast<MCSectionELF>(LSDASection);
+  const auto *LSDA = static_cast<const MCSectionELF *>(LSDASection);
   unsigned Flags = LSDA->getFlags();
   const MCSymbolELF *LinkedToSym = nullptr;
   StringRef Group;
@@ -1734,7 +1734,8 @@ MCSection *TargetLoweringObjectFileCOFF::getExplicitSectionGlobal(
       Name == getInstrProfSectionName(IPSK_covdata, Triple::COFF,
                                       /*AddSegmentInfo=*/false) ||
       Name == getInstrProfSectionName(IPSK_covname, Triple::COFF,
-                                      /*AddSegmentInfo=*/false))
+                                      /*AddSegmentInfo=*/false) ||
+      Name == ".llvmbc" || Name == ".llvmcmd")
     Kind = SectionKind::getMetadata();
   int Selection = 0;
   unsigned Characteristics = getCOFFSectionFlags(Kind, TM);
@@ -2054,14 +2055,14 @@ MCSection *TargetLoweringObjectFileCOFF::getStaticCtorSection(
     unsigned Priority, const MCSymbol *KeySym) const {
   return getCOFFStaticStructorSection(
       getContext(), getContext().getTargetTriple(), true, Priority, KeySym,
-      cast<MCSectionCOFF>(StaticCtorSection));
+      static_cast<MCSectionCOFF *>(StaticCtorSection));
 }
 
 MCSection *TargetLoweringObjectFileCOFF::getStaticDtorSection(
     unsigned Priority, const MCSymbol *KeySym) const {
   return getCOFFStaticStructorSection(
       getContext(), getContext().getTargetTriple(), false, Priority, KeySym,
-      cast<MCSectionCOFF>(StaticDtorSection));
+      static_cast<MCSectionCOFF *>(StaticDtorSection));
 }
 
 const MCExpr *TargetLoweringObjectFileCOFF::lowerRelativeReference(
@@ -2388,23 +2389,25 @@ TargetLoweringObjectFileXCOFF::getTargetSymbol(const GlobalValue *GV,
   // here.
   if (const GlobalObject *GO = dyn_cast<GlobalObject>(GV)) {
     if (GO->isDeclarationForLinker())
-      return cast<MCSectionXCOFF>(getSectionForExternalReference(GO, TM))
+      return static_cast<const MCSectionXCOFF *>(
+                 getSectionForExternalReference(GO, TM))
           ->getQualNameSymbol();
 
     if (const GlobalVariable *GVar = dyn_cast<GlobalVariable>(GV))
       if (GVar->hasAttribute("toc-data"))
-        return cast<MCSectionXCOFF>(
+        return static_cast<const MCSectionXCOFF *>(
                    SectionForGlobal(GVar, SectionKind::getData(), TM))
             ->getQualNameSymbol();
 
     SectionKind GOKind = getKindForGlobal(GO, TM);
     if (GOKind.isText())
-      return cast<MCSectionXCOFF>(
+      return static_cast<const MCSectionXCOFF *>(
                  getSectionForFunctionDescriptor(cast<Function>(GO), TM))
           ->getQualNameSymbol();
     if ((TM.getDataSections() && !GO->hasSection()) || GO->hasCommonLinkage() ||
         GOKind.isBSSLocal() || GOKind.isThreadBSSLocal())
-      return cast<MCSectionXCOFF>(SectionForGlobal(GO, GOKind, TM))
+      return static_cast<const MCSectionXCOFF *>(
+                 SectionForGlobal(GO, GOKind, TM))
           ->getQualNameSymbol();
   }
 
@@ -2740,7 +2743,7 @@ MCSection *TargetLoweringObjectFileXCOFF::getSectionForTOCEntry(
 
 MCSection *TargetLoweringObjectFileXCOFF::getSectionForLSDA(
     const Function &F, const MCSymbol &FnSym, const TargetMachine &TM) const {
-  auto *LSDA = cast<MCSectionXCOFF>(LSDASection);
+  auto *LSDA = static_cast<MCSectionXCOFF *>(LSDASection);
   if (TM.getFunctionSections()) {
     // If option -ffunction-sections is on, append the function name to the
     // name of the LSDA csect so that each function has its own LSDA csect.
diff --git a/llvm/lib/ExecutionEngine/RuntimeDyld/RuntimeDyld.cpp b/llvm/lib/ExecutionEngine/RuntimeDyld/RuntimeDyld.cpp
index b3798f1..a8559e7 100644
--- a/llvm/lib/ExecutionEngine/RuntimeDyld/RuntimeDyld.cpp
+++ b/llvm/lib/ExecutionEngine/RuntimeDyld/RuntimeDyld.cpp
@@ -183,7 +183,7 @@ RuntimeDyldImpl::loadObjectImpl(const object::ObjectFile &Obj) {
   std::lock_guard<sys::Mutex> locked(lock);
 
   // Save information about our target
-  Arch = (Triple::ArchType)Obj.getArch();
+  Arch = Obj.getArch();
   IsTargetLittleEndian = Obj.isLittleEndian();
   setMipsABI(Obj);
 
@@ -1361,18 +1361,17 @@ std::unique_ptr<RuntimeDyld::LoadedObjectInfo>
 RuntimeDyld::loadObject(const ObjectFile &Obj) {
   if (!Dyld) {
     if (Obj.isELF())
-      Dyld =
-          createRuntimeDyldELF(static_cast<Triple::ArchType>(Obj.getArch()),
-                               MemMgr, Resolver, ProcessAllSections,
-                               std::move(NotifyStubEmitted));
+      Dyld = createRuntimeDyldELF(Obj.getArch(), MemMgr, Resolver,
+                                  ProcessAllSections,
+                                  std::move(NotifyStubEmitted));
     else if (Obj.isMachO())
-      Dyld = createRuntimeDyldMachO(
-               static_cast<Triple::ArchType>(Obj.getArch()), MemMgr, Resolver,
-               ProcessAllSections, std::move(NotifyStubEmitted));
+      Dyld = createRuntimeDyldMachO(Obj.getArch(), MemMgr, Resolver,
+                                    ProcessAllSections,
+                                    std::move(NotifyStubEmitted));
     else if (Obj.isCOFF())
-      Dyld = createRuntimeDyldCOFF(
-               static_cast<Triple::ArchType>(Obj.getArch()), MemMgr, Resolver,
-               ProcessAllSections, std::move(NotifyStubEmitted));
+      Dyld = createRuntimeDyldCOFF(Obj.getArch(), MemMgr, Resolver,
+                                   ProcessAllSections,
+                                   std::move(NotifyStubEmitted));
     else
       report_fatal_error("Incompatible object format!");
   }
diff --git a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
index 840ca83..7928772 100644
--- a/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
+++ b/llvm/lib/Frontend/OpenMP/OMPIRBuilder.cpp
@@ -2617,7 +2617,7 @@ void OpenMPIRBuilder::emitReductionListCopy(
 Expected<Function *> OpenMPIRBuilder::emitInterWarpCopyFunction(
     const LocationDescription &Loc, ArrayRef<ReductionInfo> ReductionInfos,
     AttributeList FuncAttrs) {
-  IRBuilder<>::InsertPointGuard IPG(Builder);
+  InsertPointTy SavedIP = Builder.saveIP();
   LLVMContext &Ctx = M.getContext();
   FunctionType *FuncTy = FunctionType::get(
       Builder.getVoidTy(), {Builder.getPtrTy(), Builder.getInt32Ty()},
@@ -2630,7 +2630,6 @@ Expected<Function *> OpenMPIRBuilder::emitInterWarpCopyFunction(
   WcFunc->addParamAttr(1, Attribute::NoUndef);
   BasicBlock *EntryBB = BasicBlock::Create(M.getContext(), "entry", WcFunc);
   Builder.SetInsertPoint(EntryBB);
-  Builder.SetCurrentDebugLocation(llvm::DebugLoc());
 
   // ReduceList: thread local Reduce list.
   // At the stage of the computation when this function is called, partially
@@ -2845,6 +2844,7 @@ Expected<Function *> OpenMPIRBuilder::emitInterWarpCopyFunction(
   }
 
   Builder.CreateRetVoid();
+  Builder.restoreIP(SavedIP);
 
   return WcFunc;
 }
@@ -2853,7 +2853,6 @@ Function *OpenMPIRBuilder::emitShuffleAndReduceFunction(
     ArrayRef<ReductionInfo> ReductionInfos, Function *ReduceFn,
     AttributeList FuncAttrs) {
   LLVMContext &Ctx = M.getContext();
-  IRBuilder<>::InsertPointGuard IPG(Builder);
   FunctionType *FuncTy =
       FunctionType::get(Builder.getVoidTy(),
                         {Builder.getPtrTy(), Builder.getInt16Ty(),
@@ -2872,7 +2871,6 @@ Function *OpenMPIRBuilder::emitShuffleAndReduceFunction(
   SarFunc->addParamAttr(3, Attribute::SExt);
   BasicBlock *EntryBB = BasicBlock::Create(M.getContext(), "entry", SarFunc);
   Builder.SetInsertPoint(EntryBB);
-  Builder.SetCurrentDebugLocation(llvm::DebugLoc());
 
   // Thread local Reduce list used to host the values of data to be reduced.
   Argument *ReduceListArg = SarFunc->getArg(0);
@@ -3019,7 +3017,7 @@ Function *OpenMPIRBuilder::emitShuffleAndReduceFunction(
 Function *OpenMPIRBuilder::emitListToGlobalCopyFunction(
     ArrayRef<ReductionInfo> ReductionInfos, Type *ReductionsBufferTy,
     AttributeList FuncAttrs) {
-  IRBuilder<>::InsertPointGuard IPG(Builder);
+  OpenMPIRBuilder::InsertPointTy OldIP = Builder.saveIP();
   LLVMContext &Ctx = M.getContext();
   FunctionType *FuncTy = FunctionType::get(
       Builder.getVoidTy(),
@@ -3035,7 +3033,6 @@ Function *OpenMPIRBuilder::emitListToGlobalCopyFunction(
 
   BasicBlock *EntryBlock = BasicBlock::Create(Ctx, "entry", LtGCFunc);
   Builder.SetInsertPoint(EntryBlock);
-  Builder.SetCurrentDebugLocation(llvm::DebugLoc());
 
   // Buffer: global reduction buffer.
   Argument *BufferArg = LtGCFunc->getArg(0);
@@ -3123,13 +3120,14 @@ Function *OpenMPIRBuilder::emitListToGlobalCopyFunction(
   }
 
   Builder.CreateRetVoid();
+  Builder.restoreIP(OldIP);
   return LtGCFunc;
 }
 
 Function *OpenMPIRBuilder::emitListToGlobalReduceFunction(
     ArrayRef<ReductionInfo> ReductionInfos, Function *ReduceFn,
     Type *ReductionsBufferTy, AttributeList FuncAttrs) {
-  IRBuilder<>::InsertPointGuard IPG(Builder);
+  OpenMPIRBuilder::InsertPointTy OldIP = Builder.saveIP();
   LLVMContext &Ctx = M.getContext();
   FunctionType *FuncTy = FunctionType::get(
       Builder.getVoidTy(),
@@ -3145,7 +3143,6 @@ Function *OpenMPIRBuilder::emitListToGlobalReduceFunction(
 
   BasicBlock *EntryBlock = BasicBlock::Create(Ctx, "entry", LtGRFunc);
   Builder.SetInsertPoint(EntryBlock);
-  Builder.SetCurrentDebugLocation(llvm::DebugLoc());
 
   // Buffer: global reduction buffer.
   Argument *BufferArg = LtGRFunc->getArg(0);
@@ -3206,13 +3203,14 @@ Function *OpenMPIRBuilder::emitListToGlobalReduceFunction(
   Builder.CreateCall(ReduceFn, {LocalReduceListAddrCast, ReduceList})
       ->addFnAttr(Attribute::NoUnwind);
   Builder.CreateRetVoid();
+  Builder.restoreIP(OldIP);
   return LtGRFunc;
 }
 
 Function *OpenMPIRBuilder::emitGlobalToListCopyFunction(
     ArrayRef<ReductionInfo> ReductionInfos, Type *ReductionsBufferTy,
     AttributeList FuncAttrs) {
-  IRBuilder<>::InsertPointGuard IPG(Builder);
+  OpenMPIRBuilder::InsertPointTy OldIP = Builder.saveIP();
   LLVMContext &Ctx = M.getContext();
   FunctionType *FuncTy = FunctionType::get(
       Builder.getVoidTy(),
@@ -3228,7 +3226,6 @@ Function *OpenMPIRBuilder::emitGlobalToListCopyFunction(
 
   BasicBlock *EntryBlock = BasicBlock::Create(Ctx, "entry", LtGCFunc);
   Builder.SetInsertPoint(EntryBlock);
-  Builder.SetCurrentDebugLocation(llvm::DebugLoc());
 
   // Buffer: global reduction buffer.
   Argument *BufferArg = LtGCFunc->getArg(0);
@@ -3314,13 +3311,14 @@ Function *OpenMPIRBuilder::emitGlobalToListCopyFunction(
   }
 
   Builder.CreateRetVoid();
+  Builder.restoreIP(OldIP);
   return LtGCFunc;
 }
 
 Function *OpenMPIRBuilder::emitGlobalToListReduceFunction(
     ArrayRef<ReductionInfo> ReductionInfos, Function *ReduceFn,
     Type *ReductionsBufferTy, AttributeList FuncAttrs) {
-  IRBuilder<>::InsertPointGuard IPG(Builder);
+  OpenMPIRBuilder::InsertPointTy OldIP = Builder.saveIP();
   LLVMContext &Ctx = M.getContext();
   auto *FuncTy = FunctionType::get(
       Builder.getVoidTy(),
@@ -3336,7 +3334,6 @@ Function *OpenMPIRBuilder::emitGlobalToListReduceFunction(
 
   BasicBlock *EntryBlock = BasicBlock::Create(Ctx, "entry", LtGRFunc);
   Builder.SetInsertPoint(EntryBlock);
-  Builder.SetCurrentDebugLocation(llvm::DebugLoc());
 
   // Buffer: global reduction buffer.
   Argument *BufferArg = LtGRFunc->getArg(0);
@@ -3397,6 +3394,7 @@ Function *OpenMPIRBuilder::emitGlobalToListReduceFunction(
   Builder.CreateCall(ReduceFn, {ReduceList, ReductionList})
       ->addFnAttr(Attribute::NoUnwind);
   Builder.CreateRetVoid();
+  Builder.restoreIP(OldIP);
   return LtGRFunc;
 }
 
@@ -3409,7 +3407,6 @@ std::string OpenMPIRBuilder::getReductionFuncName(StringRef Name) const {
 Expected<Function *> OpenMPIRBuilder::createReductionFunction(
     StringRef ReducerName, ArrayRef<ReductionInfo> ReductionInfos,
     ReductionGenCBKind ReductionGenCBKind, AttributeList FuncAttrs) {
-  IRBuilder<>::InsertPointGuard IPG(Builder);
   auto *FuncTy = FunctionType::get(Builder.getVoidTy(),
                                    {Builder.getPtrTy(), Builder.getPtrTy()},
                                    /* IsVarArg */ false);
@@ -3422,7 +3419,6 @@ Expected<Function *> OpenMPIRBuilder::createReductionFunction(
   BasicBlock *EntryBB =
       BasicBlock::Create(M.getContext(), "entry", ReductionFunc);
   Builder.SetInsertPoint(EntryBB);
-  Builder.SetCurrentDebugLocation(llvm::DebugLoc());
 
   // Need to alloca memory here and deal with the pointers before getting
   // LHS/RHS pointers out
@@ -3750,12 +3746,10 @@ static Error populateReductionFunction(
     Function *ReductionFunc,
     ArrayRef<OpenMPIRBuilder::ReductionInfo> ReductionInfos,
     IRBuilder<> &Builder, ArrayRef<bool> IsByRef, bool IsGPU) {
-  IRBuilder<>::InsertPointGuard IPG(Builder);
   Module *Module = ReductionFunc->getParent();
   BasicBlock *ReductionFuncBlock =
       BasicBlock::Create(Module->getContext(), "", ReductionFunc);
   Builder.SetInsertPoint(ReductionFuncBlock);
-  Builder.SetCurrentDebugLocation(llvm::DebugLoc());
   Value *LHSArrayPtr = nullptr;
   Value *RHSArrayPtr = nullptr;
   if (IsGPU) {
diff --git a/llvm/lib/IR/DiagnosticInfo.cpp b/llvm/lib/IR/DiagnosticInfo.cpp
index b94dcac..4f37624 100644
--- a/llvm/lib/IR/DiagnosticInfo.cpp
+++ b/llvm/lib/IR/DiagnosticInfo.cpp
@@ -81,6 +81,10 @@ void DiagnosticInfoInlineAsm::print(DiagnosticPrinter &DP) const {
     DP << " at line " << getLocCookie();
 }
 
+void DiagnosticInfoLegalizationFailure::print(DiagnosticPrinter &DP) const {
+  DP << getLocationStr() << ": " << getMsgStr();
+}
+
 DiagnosticInfoRegAllocFailure::DiagnosticInfoRegAllocFailure(
     const Twine &MsgStr, const Function &Fn, const DiagnosticLocation &DL,
     DiagnosticSeverity Severity)
diff --git a/llvm/lib/MC/CMakeLists.txt b/llvm/lib/MC/CMakeLists.txt
index d662c42..18a85b3 100644
--- a/llvm/lib/MC/CMakeLists.txt
+++ b/llvm/lib/MC/CMakeLists.txt
@@ -43,13 +43,7 @@ add_llvm_component_library(LLVMMC
   MCRegisterInfo.cpp
   MCSchedule.cpp
   MCSection.cpp
-  MCSectionCOFF.cpp
-  MCSectionDXContainer.cpp
-  MCSectionELF.cpp
-  MCSectionGOFF.cpp
   MCSectionMachO.cpp
-  MCSectionWasm.cpp
-  MCSectionXCOFF.cpp
   MCStreamer.cpp
   MCSPIRVStreamer.cpp
   MCSubtargetInfo.cpp
diff --git a/llvm/lib/MC/ELFObjectWriter.cpp b/llvm/lib/MC/ELFObjectWriter.cpp
index 4510d4db..ae8dffc 100644
--- a/llvm/lib/MC/ELFObjectWriter.cpp
+++ b/llvm/lib/MC/ELFObjectWriter.cpp
@@ -1087,7 +1087,8 @@ uint64_t ELFWriter::writeObject() {
       // Remember the offset into the file for this section.
       const uint64_t SecStart = align(RelSection->getAlign());
 
-      writeRelocations(cast<MCSectionELF>(*RelSection->getLinkedToSection()));
+      writeRelocations(
+          static_cast<const MCSectionELF &>(*RelSection->getLinkedToSection()));
 
       uint64_t SecEnd = W.OS.tell();
       RelSection->setOffsets(SecStart, SecEnd);
@@ -1260,7 +1261,7 @@ bool ELFObjectWriter::useSectionSymbol(const MCValue &Val,
   // that it pointed to another string and subtracting 42 at runtime will
   // produce the wrong value.
   if (Sym->isInSection()) {
-    auto &Sec = cast<MCSectionELF>(Sym->getSection());
+    auto &Sec = static_cast<const MCSectionELF &>(Sym->getSection());
     unsigned Flags = Sec.getFlags();
     if (Flags & ELF::SHF_MERGE) {
       if (C != 0)
@@ -1312,13 +1313,14 @@ bool ELFObjectWriter::checkRelocation(SMLoc Loc, const MCSectionELF *From,
 void ELFObjectWriter::recordRelocation(const MCFragment &F,
                                        const MCFixup &Fixup, MCValue Target,
                                        uint64_t &FixedValue) {
-  const MCSectionELF &Section = cast<MCSectionELF>(*F.getParent());
+  auto &Section = static_cast<const MCSectionELF &>(*F.getParent());
   MCContext &Ctx = getContext();
 
   const auto *SymA = cast_or_null<MCSymbolELF>(Target.getAddSym());
-  const MCSectionELF *SecA = (SymA && SymA->isInSection())
-                                 ? cast<MCSectionELF>(&SymA->getSection())
-                                 : nullptr;
+  const MCSectionELF *SecA =
+      (SymA && SymA->isInSection())
+          ? static_cast<const MCSectionELF *>(&SymA->getSection())
+          : nullptr;
   if (DwoOS && !checkRelocation(Fixup.getLoc(), &Section, SecA))
     return;
 
diff --git a/llvm/lib/MC/GOFFObjectWriter.cpp b/llvm/lib/MC/GOFFObjectWriter.cpp
index 1871f5f..88188f3 100644
--- a/llvm/lib/MC/GOFFObjectWriter.cpp
+++ b/llvm/lib/MC/GOFFObjectWriter.cpp
@@ -336,7 +336,7 @@ void GOFFWriter::defineSymbols() {
   unsigned Ordinal = 0;
   // Process all sections.
   for (MCSection &S : Asm) {
-    auto &Section = cast<MCSectionGOFF>(S);
+    auto &Section = static_cast<MCSectionGOFF &>(S);
     Section.setOrdinal(++Ordinal);
     defineSectionSymbols(Section);
   }
diff --git a/llvm/lib/MC/MCAsmInfoCOFF.cpp b/llvm/lib/MC/MCAsmInfoCOFF.cpp
index 0b8781c..54717df 100644
--- a/llvm/lib/MC/MCAsmInfoCOFF.cpp
+++ b/llvm/lib/MC/MCAsmInfoCOFF.cpp
@@ -12,7 +12,13 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/MC/MCAsmInfoCOFF.h"
+#include "llvm/BinaryFormat/COFF.h"
 #include "llvm/MC/MCDirectives.h"
+#include "llvm/MC/MCSection.h"
+#include "llvm/MC/MCSectionCOFF.h"
+#include "llvm/MC/MCSymbol.h"
+#include "llvm/Support/raw_ostream.h"
+#include <cassert>
 
 using namespace llvm;
 
@@ -49,6 +55,10 @@ MCAsmInfoCOFF::MCAsmInfoCOFF() {
   HasCOFFComdatConstants = true;
 }
 
+bool MCAsmInfoCOFF::useCodeAlign(const MCSection &Sec) const {
+  return Sec.isText();
+}
+
 void MCAsmInfoMicrosoft::anchor() {}
 
 MCAsmInfoMicrosoft::MCAsmInfoMicrosoft() = default;
@@ -64,3 +74,101 @@ MCAsmInfoGNUCOFF::MCAsmInfoGNUCOFF() {
   // We don't create constants in comdat sections for MinGW.
   HasCOFFComdatConstants = false;
 }
+
+bool MCSectionCOFF::shouldOmitSectionDirective(StringRef Name) const {
+  if (COMDATSymbol || isUnique())
+    return false;
+
+  // FIXME: Does .section .bss/.data/.text work everywhere??
+  if (Name == ".text" || Name == ".data" || Name == ".bss")
+    return true;
+
+  return false;
+}
+
+void MCSectionCOFF::setSelection(int Selection) const {
+  assert(Selection != 0 && "invalid COMDAT selection type");
+  this->Selection = Selection;
+  Characteristics |= COFF::IMAGE_SCN_LNK_COMDAT;
+}
+
+void MCAsmInfoCOFF::printSwitchToSection(const MCSection &Section, uint32_t,
+                                         const Triple &T,
+                                         raw_ostream &OS) const {
+  auto &Sec = static_cast<const MCSectionCOFF &>(Section);
+  // standard sections don't require the '.section'
+  if (Sec.shouldOmitSectionDirective(Sec.getName())) {
+    OS << '\t' << Sec.getName() << '\n';
+    return;
+  }
+
+  OS << "\t.section\t" << Sec.getName() << ",\"";
+  if (Sec.getCharacteristics() & COFF::IMAGE_SCN_CNT_INITIALIZED_DATA)
+    OS << 'd';
+  if (Sec.getCharacteristics() & COFF::IMAGE_SCN_CNT_UNINITIALIZED_DATA)
+    OS << 'b';
+  if (Sec.getCharacteristics() & COFF::IMAGE_SCN_MEM_EXECUTE)
+    OS << 'x';
+  if (Sec.getCharacteristics() & COFF::IMAGE_SCN_MEM_WRITE)
+    OS << 'w';
+  else if (Sec.getCharacteristics() & COFF::IMAGE_SCN_MEM_READ)
+    OS << 'r';
+  else
+    OS << 'y';
+  if (Sec.getCharacteristics() & COFF::IMAGE_SCN_LNK_REMOVE)
+    OS << 'n';
+  if (Sec.getCharacteristics() & COFF::IMAGE_SCN_MEM_SHARED)
+    OS << 's';
+  if ((Sec.getCharacteristics() & COFF::IMAGE_SCN_MEM_DISCARDABLE) &&
+      !Sec.isImplicitlyDiscardable(Sec.getName()))
+    OS << 'D';
+  if (Sec.getCharacteristics() & COFF::IMAGE_SCN_LNK_INFO)
+    OS << 'i';
+  OS << '"';
+
+  // unique should be tail of .section directive.
+  if (Sec.isUnique() && !Sec.COMDATSymbol)
+    OS << ",unique," << Sec.UniqueID;
+
+  if (Sec.getCharacteristics() & COFF::IMAGE_SCN_LNK_COMDAT) {
+    if (Sec.COMDATSymbol)
+      OS << ",";
+    else
+      OS << "\n\t.linkonce\t";
+    switch (Sec.Selection) {
+    case COFF::IMAGE_COMDAT_SELECT_NODUPLICATES:
+      OS << "one_only";
+      break;
+    case COFF::IMAGE_COMDAT_SELECT_ANY:
+      OS << "discard";
+      break;
+    case COFF::IMAGE_COMDAT_SELECT_SAME_SIZE:
+      OS << "same_size";
+      break;
+    case COFF::IMAGE_COMDAT_SELECT_EXACT_MATCH:
+      OS << "same_contents";
+      break;
+    case COFF::IMAGE_COMDAT_SELECT_ASSOCIATIVE:
+      OS << "associative";
+      break;
+    case COFF::IMAGE_COMDAT_SELECT_LARGEST:
+      OS << "largest";
+      break;
+    case COFF::IMAGE_COMDAT_SELECT_NEWEST:
+      OS << "newest";
+      break;
+    default:
+      assert(false && "unsupported COFF selection type");
+      break;
+    }
+    if (Sec.COMDATSymbol) {
+      OS << ",";
+      Sec.COMDATSymbol->print(OS, this);
+    }
+  }
+
+  if (Sec.isUnique() && Sec.COMDATSymbol)
+    OS << ",unique," << Sec.UniqueID;
+
+  OS << '\n';
+}
diff --git a/llvm/lib/MC/MCAsmInfoDarwin.cpp b/llvm/lib/MC/MCAsmInfoDarwin.cpp
index 9cba775..e156fa0 100644
--- a/llvm/lib/MC/MCAsmInfoDarwin.cpp
+++ b/llvm/lib/MC/MCAsmInfoDarwin.cpp
@@ -85,3 +85,8 @@ MCAsmInfoDarwin::MCAsmInfoDarwin() {
   DwarfUsesRelocationsAcrossSections = false;
   SetDirectiveSuppressesReloc = true;
 }
+
+bool MCAsmInfoDarwin::useCodeAlign(const MCSection &Sec) const {
+  return static_cast<const MCSectionMachO &>(Sec).hasAttribute(
+      MachO::S_ATTR_PURE_INSTRUCTIONS);
+}
diff --git a/llvm/lib/MC/MCAsmInfoELF.cpp b/llvm/lib/MC/MCAsmInfoELF.cpp
index 7eb89ef..cdae9d7 100644
--- a/llvm/lib/MC/MCAsmInfoELF.cpp
+++ b/llvm/lib/MC/MCAsmInfoELF.cpp
@@ -12,9 +12,16 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/MC/MCAsmInfoELF.h"
+#include "llvm/ADT/Twine.h"
 #include "llvm/BinaryFormat/ELF.h"
+#include "llvm/MC/MCAsmInfo.h"
 #include "llvm/MC/MCContext.h"
+#include "llvm/MC/MCExpr.h"
 #include "llvm/MC/MCSectionELF.h"
+#include "llvm/Support/ErrorHandling.h"
+#include "llvm/Support/raw_ostream.h"
+#include "llvm/TargetParser/Triple.h"
+#include <cassert>
 
 using namespace llvm;
 
@@ -28,9 +35,198 @@ MCSection *MCAsmInfoELF::getNonexecutableStackSection(MCContext &Ctx) const {
   return Ctx.getELFSection(".note.GNU-stack", ELF::SHT_PROGBITS, 0);
 }
 
+bool MCAsmInfoELF::useCodeAlign(const MCSection &Sec) const {
+  return static_cast<const MCSectionELF &>(Sec).getFlags() & ELF::SHF_EXECINSTR;
+}
+
 MCAsmInfoELF::MCAsmInfoELF() {
   HasIdentDirective = true;
   WeakRefDirective = "\t.weak\t";
   PrivateGlobalPrefix = ".L";
   PrivateLabelPrefix = ".L";
 }
+
+static void printName(raw_ostream &OS, StringRef Name) {
+  if (Name.find_first_not_of("0123456789_."
+                             "abcdefghijklmnopqrstuvwxyz"
+                             "ABCDEFGHIJKLMNOPQRSTUVWXYZ") == Name.npos) {
+    OS << Name;
+    return;
+  }
+  OS << '"';
+  for (const char *B = Name.begin(), *E = Name.end(); B < E; ++B) {
+    if (*B == '"') // Unquoted "
+      OS << "\\\"";
+    else if (*B != '\\') // Neither " or backslash
+      OS << *B;
+    else if (B + 1 == E) // Trailing backslash
+      OS << "\\\\";
+    else {
+      OS << B[0] << B[1]; // Quoted character
+      ++B;
+    }
+  }
+  OS << '"';
+}
+
+void MCAsmInfoELF::printSwitchToSection(const MCSection &Section,
+                                        uint32_t Subsection, const Triple &T,
+                                        raw_ostream &OS) const {
+  auto &Sec = static_cast<const MCSectionELF &>(Section);
+  if (!Sec.isUnique() && shouldOmitSectionDirective(Sec.getName())) {
+    OS << '\t' << Sec.getName();
+    if (Subsection)
+      OS << '\t' << Subsection;
+    OS << '\n';
+    return;
+  }
+
+  OS << "\t.section\t";
+  printName(OS, Sec.getName());
+
+  // Handle the weird solaris syntax if desired.
+  if (usesSunStyleELFSectionSwitchSyntax() && !(Sec.Flags & ELF::SHF_MERGE)) {
+    if (Sec.Flags & ELF::SHF_ALLOC)
+      OS << ",#alloc";
+    if (Sec.Flags & ELF::SHF_EXECINSTR)
+      OS << ",#execinstr";
+    if (Sec.Flags & ELF::SHF_WRITE)
+      OS << ",#write";
+    if (Sec.Flags & ELF::SHF_EXCLUDE)
+      OS << ",#exclude";
+    if (Sec.Flags & ELF::SHF_TLS)
+      OS << ",#tls";
+    OS << '\n';
+    return;
+  }
+
+  OS << ",\"";
+  if (Sec.Flags & ELF::SHF_ALLOC)
+    OS << 'a';
+  if (Sec.Flags & ELF::SHF_EXCLUDE)
+    OS << 'e';
+  if (Sec.Flags & ELF::SHF_EXECINSTR)
+    OS << 'x';
+  if (Sec.Flags & ELF::SHF_WRITE)
+    OS << 'w';
+  if (Sec.Flags & ELF::SHF_MERGE)
+    OS << 'M';
+  if (Sec.Flags & ELF::SHF_STRINGS)
+    OS << 'S';
+  if (Sec.Flags & ELF::SHF_TLS)
+    OS << 'T';
+  if (Sec.Flags & ELF::SHF_LINK_ORDER)
+    OS << 'o';
+  if (Sec.Flags & ELF::SHF_GROUP)
+    OS << 'G';
+  if (Sec.Flags & ELF::SHF_GNU_RETAIN)
+    OS << 'R';
+
+  // If there are os-specific flags, print them.
+  if (T.isOSSolaris())
+    if (Sec.Flags & ELF::SHF_SUNW_NODISCARD)
+      OS << 'R';
+
+  // If there are tarSec.get-specific flags, print them.
+  Triple::ArchType Arch = T.getArch();
+  if (Arch == Triple::xcore) {
+    if (Sec.Flags & ELF::XCORE_SHF_CP_SECTION)
+      OS << 'c';
+    if (Sec.Flags & ELF::XCORE_SHF_DP_SECTION)
+      OS << 'd';
+  } else if (T.isARM() || T.isThumb()) {
+    if (Sec.Flags & ELF::SHF_ARM_PURECODE)
+      OS << 'y';
+  } else if (T.isAArch64()) {
+    if (Sec.Flags & ELF::SHF_AARCH64_PURECODE)
+      OS << 'y';
+  } else if (Arch == Triple::hexagon) {
+    if (Sec.Flags & ELF::SHF_HEX_GPREL)
+      OS << 's';
+  } else if (Arch == Triple::x86_64) {
+    if (Sec.Flags & ELF::SHF_X86_64_LARGE)
+      OS << 'l';
+  }
+
+  OS << '"';
+
+  OS << ',';
+
+  // If comment string is '@', e.g. as on ARM - use '%' instead
+  if (getCommentString()[0] == '@')
+    OS << '%';
+  else
+    OS << '@';
+
+  if (Sec.Type == ELF::SHT_INIT_ARRAY)
+    OS << "init_array";
+  else if (Sec.Type == ELF::SHT_FINI_ARRAY)
+    OS << "fini_array";
+  else if (Sec.Type == ELF::SHT_PREINIT_ARRAY)
+    OS << "preinit_array";
+  else if (Sec.Type == ELF::SHT_NOBITS)
+    OS << "nobits";
+  else if (Sec.Type == ELF::SHT_NOTE)
+    OS << "note";
+  else if (Sec.Type == ELF::SHT_PROGBITS)
+    OS << "progbits";
+  else if (Sec.Type == ELF::SHT_X86_64_UNWIND)
+    OS << "unwind";
+  else if (Sec.Type == ELF::SHT_MIPS_DWARF)
+    // Print hex value of the flag while we do not have
+    // any standard symbolic representation of the flag.
+    OS << "0x7000001e";
+  else if (Sec.Type == ELF::SHT_LLVM_ODRTAB)
+    OS << "llvm_odrtab";
+  else if (Sec.Type == ELF::SHT_LLVM_LINKER_OPTIONS)
+    OS << "llvm_linker_options";
+  else if (Sec.Type == ELF::SHT_LLVM_CALL_GRAPH_PROFILE)
+    OS << "llvm_call_graph_profile";
+  else if (Sec.Type == ELF::SHT_LLVM_DEPENDENT_LIBRARIES)
+    OS << "llvm_dependent_libraries";
+  else if (Sec.Type == ELF::SHT_LLVM_SYMPART)
+    OS << "llvm_sympart";
+  else if (Sec.Type == ELF::SHT_LLVM_BB_ADDR_MAP)
+    OS << "llvm_bb_addr_map";
+  else if (Sec.Type == ELF::SHT_LLVM_OFFLOADING)
+    OS << "llvm_offloading";
+  else if (Sec.Type == ELF::SHT_LLVM_LTO)
+    OS << "llvm_lto";
+  else if (Sec.Type == ELF::SHT_LLVM_JT_SIZES)
+    OS << "llvm_jt_sizes";
+  else if (Sec.Type == ELF::SHT_LLVM_CFI_JUMP_TABLE)
+    OS << "llvm_cfi_jump_table";
+  else
+    OS << "0x" << Twine::utohexstr(Sec.Type);
+
+  if (Sec.EntrySize) {
+    assert((Sec.Flags & ELF::SHF_MERGE) ||
+           Sec.Type == ELF::SHT_LLVM_CFI_JUMP_TABLE);
+    OS << "," << Sec.EntrySize;
+  }
+
+  if (Sec.Flags & ELF::SHF_LINK_ORDER) {
+    OS << ",";
+    if (Sec.LinkedToSym)
+      printName(OS, Sec.LinkedToSym->getName());
+    else
+      OS << '0';
+  }
+
+  if (Sec.Flags & ELF::SHF_GROUP) {
+    OS << ",";
+    printName(OS, Sec.Group.getPointer()->getName());
+    if (Sec.isComdat())
+      OS << ",comdat";
+  }
+
+  if (Sec.isUnique())
+    OS << ",unique," << Sec.UniqueID;
+
+  OS << '\n';
+
+  if (Subsection) {
+    OS << "\t.subsection\t" << Subsection;
+    OS << '\n';
+  }
+}
diff --git a/llvm/lib/MC/MCAsmInfoGOFF.cpp b/llvm/lib/MC/MCAsmInfoGOFF.cpp
index 3c81a46..0a5d1927 100644
--- a/llvm/lib/MC/MCAsmInfoGOFF.cpp
+++ b/llvm/lib/MC/MCAsmInfoGOFF.cpp
@@ -13,11 +13,12 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/MC/MCAsmInfoGOFF.h"
+#include "llvm/BinaryFormat/GOFF.h"
+#include "llvm/MC/MCSectionGOFF.h"
+#include "llvm/Support/raw_ostream.h"
 
 using namespace llvm;
 
-void MCAsmInfoGOFF::anchor() {}
-
 MCAsmInfoGOFF::MCAsmInfoGOFF() {
   Data64bitsDirective = "\t.quad\t";
   HasDotTypeDotSizeDirective = false;
@@ -25,3 +26,136 @@ MCAsmInfoGOFF::MCAsmInfoGOFF() {
   PrivateLabelPrefix = "L#";
   ZeroDirective = "\t.space\t";
 }
+
+static void emitCATTR(raw_ostream &OS, StringRef Name, GOFF::ESDRmode Rmode,
+                      GOFF::ESDAlignment Alignment,
+                      GOFF::ESDLoadingBehavior LoadBehavior,
+                      GOFF::ESDExecutable Executable, bool IsReadOnly,
+                      uint32_t SortKey, uint8_t FillByteValue,
+                      StringRef PartName) {
+  OS << Name << " CATTR ";
+  OS << "ALIGN(" << static_cast<unsigned>(Alignment) << "),"
+     << "FILL(" << static_cast<unsigned>(FillByteValue) << ")";
+  switch (LoadBehavior) {
+  case GOFF::ESD_LB_Deferred:
+    OS << ",DEFLOAD";
+    break;
+  case GOFF::ESD_LB_NoLoad:
+    OS << ",NOLOAD";
+    break;
+  default:
+    break;
+  }
+  switch (Executable) {
+  case GOFF::ESD_EXE_CODE:
+    OS << ",EXECUTABLE";
+    break;
+  case GOFF::ESD_EXE_DATA:
+    OS << ",NOTEXECUTABLE";
+    break;
+  default:
+    break;
+  }
+  if (IsReadOnly)
+    OS << ",READONLY";
+  if (Rmode != GOFF::ESD_RMODE_None) {
+    OS << ',';
+    OS << "RMODE(";
+    switch (Rmode) {
+    case GOFF::ESD_RMODE_24:
+      OS << "24";
+      break;
+    case GOFF::ESD_RMODE_31:
+      OS << "31";
+      break;
+    case GOFF::ESD_RMODE_64:
+      OS << "64";
+      break;
+    case GOFF::ESD_RMODE_None:
+      break;
+    }
+    OS << ')';
+  }
+  if (SortKey)
+    OS << ",PRIORITY(" << SortKey << ")";
+  if (!PartName.empty())
+    OS << ",PART(" << PartName << ")";
+  OS << '\n';
+}
+
+static void emitXATTR(raw_ostream &OS, StringRef Name,
+                      GOFF::ESDLinkageType Linkage,
+                      GOFF::ESDExecutable Executable,
+                      GOFF::ESDBindingScope BindingScope) {
+  OS << Name << " XATTR ";
+  OS << "LINKAGE(" << (Linkage == GOFF::ESD_LT_OS ? "OS" : "XPLINK") << "),";
+  if (Executable != GOFF::ESD_EXE_Unspecified)
+    OS << "REFERENCE(" << (Executable == GOFF::ESD_EXE_CODE ? "CODE" : "DATA")
+       << "),";
+  if (BindingScope != GOFF::ESD_BSC_Unspecified) {
+    OS << "SCOPE(";
+    switch (BindingScope) {
+    case GOFF::ESD_BSC_Section:
+      OS << "SECTION";
+      break;
+    case GOFF::ESD_BSC_Module:
+      OS << "MODULE";
+      break;
+    case GOFF::ESD_BSC_Library:
+      OS << "LIBRARY";
+      break;
+    case GOFF::ESD_BSC_ImportExport:
+      OS << "EXPORT";
+      break;
+    default:
+      break;
+    }
+    OS << ')';
+  }
+  OS << '\n';
+}
+
+void MCAsmInfoGOFF::printSwitchToSection(const MCSection &Section,
+                                         uint32_t Subsection, const Triple &T,
+                                         raw_ostream &OS) const {
+  auto &Sec =
+      const_cast<MCSectionGOFF &>(static_cast<const MCSectionGOFF &>(Section));
+  switch (Sec.SymbolType) {
+  case GOFF::ESD_ST_SectionDefinition: {
+    OS << Sec.getName() << " CSECT\n";
+    Sec.Emitted = true;
+    break;
+  }
+  case GOFF::ESD_ST_ElementDefinition: {
+    printSwitchToSection(*Sec.getParent(), Subsection, T, OS);
+    if (!Sec.Emitted) {
+      emitCATTR(OS, Sec.getName(), Sec.EDAttributes.Rmode,
+                Sec.EDAttributes.Alignment, Sec.EDAttributes.LoadBehavior,
+                GOFF::ESD_EXE_Unspecified, Sec.EDAttributes.IsReadOnly, 0,
+                Sec.EDAttributes.FillByteValue, StringRef());
+      Sec.Emitted = true;
+    } else
+      OS << Sec.getName() << " CATTR\n";
+    break;
+  }
+  case GOFF::ESD_ST_PartReference: {
+    MCSectionGOFF *ED = Sec.getParent();
+    printSwitchToSection(*ED->getParent(), Subsection, T, OS);
+    if (!Sec.Emitted) {
+      emitCATTR(OS, ED->getName(), ED->getEDAttributes().Rmode,
+                ED->EDAttributes.Alignment, ED->EDAttributes.LoadBehavior,
+                Sec.PRAttributes.Executable, ED->EDAttributes.IsReadOnly,
+                Sec.PRAttributes.SortKey, ED->EDAttributes.FillByteValue,
+                Sec.getName());
+      emitXATTR(OS, Sec.getName(), Sec.PRAttributes.Linkage,
+                Sec.PRAttributes.Executable, Sec.PRAttributes.BindingScope);
+      ED->Emitted = true;
+      Sec.Emitted = true;
+    } else
+      OS << ED->getName() << " CATTR PART(" << Sec.getName() << ")\n";
+    break;
+  }
+  default:
+    llvm_unreachable("Wrong section type");
+  }
+}
diff --git a/llvm/lib/MC/MCAsmInfoWasm.cpp b/llvm/lib/MC/MCAsmInfoWasm.cpp
index ce6ec7e..5e44f48 100644
--- a/llvm/lib/MC/MCAsmInfoWasm.cpp
+++ b/llvm/lib/MC/MCAsmInfoWasm.cpp
@@ -12,9 +12,11 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/MC/MCAsmInfoWasm.h"
-using namespace llvm;
+#include "llvm/MC/MCSectionWasm.h"
+#include "llvm/MC/MCSymbolWasm.h"
+#include "llvm/Support/raw_ostream.h"
 
-void MCAsmInfoWasm::anchor() {}
+using namespace llvm;
 
 MCAsmInfoWasm::MCAsmInfoWasm() {
   HasIdentDirective = true;
@@ -23,3 +25,80 @@ MCAsmInfoWasm::MCAsmInfoWasm() {
   PrivateGlobalPrefix = ".L";
   PrivateLabelPrefix = ".L";
 }
+
+static void printName(raw_ostream &OS, StringRef Name) {
+  if (Name.find_first_not_of("0123456789_."
+                             "abcdefghijklmnopqrstuvwxyz"
+                             "ABCDEFGHIJKLMNOPQRSTUVWXYZ") == Name.npos) {
+    OS << Name;
+    return;
+  }
+  OS << '"';
+  for (const char *B = Name.begin(), *E = Name.end(); B < E; ++B) {
+    if (*B == '"') // Unquoted "
+      OS << "\\\"";
+    else if (*B != '\\') // Neither " or backslash
+      OS << *B;
+    else if (B + 1 == E) // Trailing backslash
+      OS << "\\\\";
+    else {
+      OS << B[0] << B[1]; // Quoted character
+      ++B;
+    }
+  }
+  OS << '"';
+}
+
+void MCAsmInfoWasm::printSwitchToSection(const MCSection &Section,
+                                         uint32_t Subsection, const Triple &T,
+                                         raw_ostream &OS) const {
+  auto &Sec = static_cast<const MCSectionWasm &>(Section);
+  if (shouldOmitSectionDirective(Sec.getName())) {
+    OS << '\t' << Sec.getName();
+    if (Subsection)
+      OS << '\t' << Subsection;
+    OS << '\n';
+    return;
+  }
+
+  OS << "\t.section\t";
+  printName(OS, Sec.getName());
+  OS << ",\"";
+
+  if (Sec.IsPassive)
+    OS << 'p';
+  if (Sec.Group)
+    OS << 'G';
+  if (Sec.SegmentFlags & wasm::WASM_SEG_FLAG_STRINGS)
+    OS << 'S';
+  if (Sec.SegmentFlags & wasm::WASM_SEG_FLAG_TLS)
+    OS << 'T';
+  if (Sec.SegmentFlags & wasm::WASM_SEG_FLAG_RETAIN)
+    OS << 'R';
+
+  OS << '"';
+
+  OS << ',';
+
+  // If comment string is '@', e.g. as on ARM - use '%' instead
+  if (getCommentString()[0] == '@')
+    OS << '%';
+  else
+    OS << '@';
+
+  // TODO: Print section type.
+
+  if (Sec.Group) {
+    OS << ",";
+    printName(OS, Sec.Group->getName());
+    OS << ",comdat";
+  }
+
+  if (Sec.isUnique())
+    OS << ",unique," << Sec.UniqueID;
+
+  OS << '\n';
+
+  if (Subsection)
+    OS << "\t.subsection\t" << Subsection << '\n';
+}
diff --git a/llvm/lib/MC/MCAsmInfoXCOFF.cpp b/llvm/lib/MC/MCAsmInfoXCOFF.cpp
index 6ef11ba..0403b44 100644
--- a/llvm/lib/MC/MCAsmInfoXCOFF.cpp
+++ b/llvm/lib/MC/MCAsmInfoXCOFF.cpp
@@ -8,7 +8,11 @@
 
 #include "llvm/MC/MCAsmInfoXCOFF.h"
 #include "llvm/ADT/StringExtras.h"
+#include "llvm/MC/MCAsmInfo.h"
+#include "llvm/MC/MCSectionXCOFF.h"
 #include "llvm/Support/CommandLine.h"
+#include "llvm/Support/Format.h"
+#include "llvm/Support/raw_ostream.h"
 
 using namespace llvm;
 
@@ -16,8 +20,6 @@ namespace llvm {
 extern cl::opt<cl::boolOrDefault> UseLEB128Directives;
 }
 
-void MCAsmInfoXCOFF::anchor() {}
-
 MCAsmInfoXCOFF::MCAsmInfoXCOFF() {
   IsAIX = true;
   IsLittleEndian = false;
@@ -56,3 +58,121 @@ bool MCAsmInfoXCOFF::isAcceptableChar(char C) const {
   // any combination of these.
   return isAlnum(C) || C == '_' || C == '.';
 }
+
+bool MCAsmInfoXCOFF::useCodeAlign(const MCSection &Sec) const {
+  return static_cast<const MCSectionXCOFF &>(Sec).getKind().isText();
+}
+
+MCSectionXCOFF::~MCSectionXCOFF() = default;
+
+void MCSectionXCOFF::printCsectDirective(raw_ostream &OS) const {
+  OS << "\t.csect " << QualName->getName() << "," << Log2(getAlign()) << '\n';
+}
+
+void MCAsmInfoXCOFF::printSwitchToSection(const MCSection &Section, uint32_t,
+                                          const Triple &T,
+                                          raw_ostream &OS) const {
+  auto &Sec = static_cast<const MCSectionXCOFF &>(Section);
+  if (Sec.getKind().isText()) {
+    if (Sec.getMappingClass() != XCOFF::XMC_PR)
+      report_fatal_error("Unhandled storage-mapping class for .text csect");
+
+    Sec.printCsectDirective(OS);
+    return;
+  }
+
+  if (Sec.getKind().isReadOnly()) {
+    if (Sec.getMappingClass() != XCOFF::XMC_RO &&
+        Sec.getMappingClass() != XCOFF::XMC_TD)
+      report_fatal_error("Unhandled storage-mapping class for .rodata csect.");
+    Sec.printCsectDirective(OS);
+    return;
+  }
+
+  if (Sec.getKind().isReadOnlyWithRel()) {
+    if (Sec.getMappingClass() != XCOFF::XMC_RW &&
+        Sec.getMappingClass() != XCOFF::XMC_RO &&
+        Sec.getMappingClass() != XCOFF::XMC_TD)
+      report_fatal_error(
+          "Unexepected storage-mapping class for ReadOnlyWithRel kind");
+    Sec.printCsectDirective(OS);
+    return;
+  }
+
+  // Initialized TLS data.
+  if (Sec.getKind().isThreadData()) {
+    // We only expect XMC_TL here for initialized TLS data.
+    if (Sec.getMappingClass() != XCOFF::XMC_TL)
+      report_fatal_error("Unhandled storage-mapping class for .tdata csect.");
+    Sec.printCsectDirective(OS);
+    return;
+  }
+
+  if (Sec.getKind().isData()) {
+    switch (Sec.getMappingClass()) {
+    case XCOFF::XMC_RW:
+    case XCOFF::XMC_DS:
+    case XCOFF::XMC_TD:
+      Sec.printCsectDirective(OS);
+      break;
+    case XCOFF::XMC_TC:
+    case XCOFF::XMC_TE:
+      break;
+    case XCOFF::XMC_TC0:
+      OS << "\t.toc\n";
+      break;
+    default:
+      report_fatal_error("Unhandled storage-mapping class for .data csect.");
+    }
+    return;
+  }
+
+  if (Sec.isCsect() && Sec.getMappingClass() == XCOFF::XMC_TD) {
+    // Common csect type (uninitialized storage) does not have to print
+    // csect directive for section switching unless it is local.
+    if (Sec.getKind().isCommon() && !Sec.getKind().isBSSLocal())
+      return;
+
+    assert(Sec.getKind().isBSS() && "Unexpected section kind for toc-data");
+    Sec.printCsectDirective(OS);
+    return;
+  }
+  // Common csect type (uninitialized storage) does not have to print csect
+  // directive for section switching.
+  if (Sec.isCsect() && Sec.getCSectType() == XCOFF::XTY_CM) {
+    assert((Sec.getMappingClass() == XCOFF::XMC_RW ||
+            Sec.getMappingClass() == XCOFF::XMC_BS ||
+            Sec.getMappingClass() == XCOFF::XMC_UL) &&
+           "Generated a storage-mapping class for a common/bss/tbss csect we "
+           "don't "
+           "understand how to switch to.");
+    // Common symbols and local zero-initialized symbols for TLS and Non-TLS are
+    // eligible for .bss/.tbss csect, getKind().isThreadBSS() is used to
+    // cover TLS common and zero-initialized local symbols since linkage type
+    // (in the GlobalVariable) is not accessible in this class.
+    assert((Sec.getKind().isBSSLocal() || Sec.getKind().isCommon() ||
+            Sec.getKind().isThreadBSS()) &&
+           "wrong symbol type for .bss/.tbss csect");
+    // Don't have to print a directive for switching to section for commons
+    // and zero-initialized TLS data. The '.comm' and '.lcomm' directives of the
+    // variable will create the needed csect.
+    return;
+  }
+
+  // Zero-initialized TLS data with weak or external linkage are not eligible to
+  // be put into common csect.
+  if (Sec.getKind().isThreadBSS()) {
+    Sec.printCsectDirective(OS);
+    return;
+  }
+
+  // XCOFF debug sections.
+  if (Sec.getKind().isMetadata() && Sec.isDwarfSect()) {
+    OS << "\n\t.dwsect " << format("0x%" PRIx32, *Sec.getDwarfSubtypeFlags())
+       << '\n';
+    OS << Sec.getName() << ':' << '\n';
+    return;
+  }
+
+  report_fatal_error("Printing for this SectionKind is unimplemented.");
+}
diff --git a/llvm/lib/MC/MCAsmStreamer.cpp b/llvm/lib/MC/MCAsmStreamer.cpp
index 7119ef4..da51da4 100644
--- a/llvm/lib/MC/MCAsmStreamer.cpp
+++ b/llvm/lib/MC/MCAsmStreamer.cpp
@@ -532,8 +532,8 @@ void MCAsmStreamer::switchSection(MCSection *Section, uint32_t Subsection) {
     if (MCTargetStreamer *TS = getTargetStreamer()) {
       TS->changeSection(Cur.first, Section, Subsection, OS);
     } else {
-      Section->printSwitchToSection(*MAI, getContext().getTargetTriple(), OS,
-                                    Subsection);
+      MAI->printSwitchToSection(*Section, Subsection,
+                                getContext().getTargetTriple(), OS);
     }
   }
   MCStreamer::switchSection(Section, Subsection);
@@ -543,7 +543,7 @@ bool MCAsmStreamer::popSection() {
   if (!MCStreamer::popSection())
     return false;
   auto [Sec, Subsec] = getCurrentSection();
-  Sec->printSwitchToSection(*MAI, getContext().getTargetTriple(), OS, Subsec);
+  MAI->printSwitchToSection(*Sec, Subsec, getContext().getTargetTriple(), OS);
   return true;
 }
 
@@ -1105,7 +1105,7 @@ void MCAsmStreamer::emitZerofill(MCSection *Section, MCSymbol *Symbol,
   // Note: a .zerofill directive does not switch sections.
   OS << ".zerofill ";
 
-  assert(Section->getVariant() == MCSection::SV_MachO &&
+  assert(getContext().getObjectFileType() == MCContext::IsMachO &&
          ".zerofill is a Mach-O specific directive");
   // This is a mach-o specific directive.
 
@@ -1130,7 +1130,7 @@ void MCAsmStreamer::emitTBSSSymbol(MCSection *Section, MCSymbol *Symbol,
 
   // Instead of using the Section we'll just use the shortcut.
 
-  assert(Section->getVariant() == MCSection::SV_MachO &&
+  assert(getContext().getObjectFileType() == MCContext::IsMachO &&
          ".zerofill is a Mach-O specific directive");
   // This is a mach-o specific directive and section.
 
diff --git a/llvm/lib/MC/MCContext.cpp b/llvm/lib/MC/MCContext.cpp
index c0448ae..39bf628 100644
--- a/llvm/lib/MC/MCContext.cpp
+++ b/llvm/lib/MC/MCContext.cpp
@@ -767,8 +767,8 @@ MCSectionCOFF *MCContext::getCOFFSection(StringRef Section,
     if (Selection != COFF::IMAGE_COMDAT_SELECT_ASSOCIATIVE &&
         COMDATSymbol->isDefined() &&
         (!COMDATSymbol->isInSection() ||
-         cast<MCSectionCOFF>(COMDATSymbol->getSection()).getCOMDATSymbol() !=
-             COMDATSymbol))
+         static_cast<const MCSectionCOFF &>(COMDATSymbol->getSection())
+                 .getCOMDATSymbol() != COMDATSymbol))
       reportError(SMLoc(), "invalid symbol redefinition");
   }
 
diff --git a/llvm/lib/MC/MCExpr.cpp b/llvm/lib/MC/MCExpr.cpp
index dbb2fd1..c24c82d 100644
--- a/llvm/lib/MC/MCExpr.cpp
+++ b/llvm/lib/MC/MCExpr.cpp
@@ -346,17 +346,16 @@ static void attemptToFoldSymbolOffsetDifference(const MCAssembler *Asm,
       Displacement *= -1;
     }
 
-    // Track whether B is before a relaxable instruction and whether A is after
-    // a relaxable instruction. If SA and SB are separated by a linker-relaxable
-    // instruction, the difference cannot be resolved as it may be changed by
-    // the linker.
+    // Track whether B is before a relaxable instruction/alignment and whether A
+    // is after a relaxable instruction/alignment. If SA and SB are separated by
+    // a linker-relaxable instruction/alignment, the difference cannot be
+    // resolved as it may be changed by the linker.
     bool BBeforeRelax = false, AAfterRelax = false;
     for (auto F = FB; F; F = F->getNext()) {
-      auto DF = F->getKind() == MCFragment::FT_Data ? F : nullptr;
-      if (DF && DF->isLinkerRelaxable()) {
-        if (&*F != FB || SBOffset != DF->getContents().size())
+      if (F && F->isLinkerRelaxable()) {
+        if (&*F != FB || SBOffset != F->getSize())
           BBeforeRelax = true;
-        if (&*F != FA || SAOffset == DF->getContents().size())
+        if (&*F != FA || SAOffset == F->getSize())
           AAfterRelax = true;
         if (BBeforeRelax && AAfterRelax)
           return;
@@ -370,17 +369,15 @@ static void attemptToFoldSymbolOffsetDifference(const MCAssembler *Asm,
       }
 
       int64_t Num;
-      if (DF) {
-        Displacement += DF->getContents().size();
-      } else if (F->getKind() == MCFragment::FT_Relaxable &&
+      if (F->getKind() == MCFragment::FT_Data) {
+        Displacement += F->getFixedSize();
+      } else if ((F->getKind() == MCFragment::FT_Relaxable ||
+                  F->getKind() == MCFragment::FT_Align) &&
                  Asm->hasFinalLayout()) {
         // Before finishLayout, a relaxable fragment's size is indeterminate.
         // After layout, during relocation generation, it can be treated as a
         // data fragment.
         Displacement += F->getSize();
-      } else if (F->getKind() == MCFragment::FT_Align && Layout &&
-                 F->isLinkerRelaxable()) {
-        Displacement += Asm->computeFragmentSize(*F);
       } else if (auto *FF = dyn_cast<MCFillFragment>(F);
                  FF && FF->getNumValues().evaluateAsAbsolute(Num)) {
         Displacement += Num * FF->getValueSize();
diff --git a/llvm/lib/MC/MCFragment.cpp b/llvm/lib/MC/MCFragment.cpp
index 3c395e5..6cbdf74 100644
--- a/llvm/lib/MC/MCFragment.cpp
+++ b/llvm/lib/MC/MCFragment.cpp
@@ -35,7 +35,7 @@ MCFragment::MCFragment(FragmentType Kind, bool HasInstructions)
 }
 
 const MCSymbol *MCFragment::getAtom() const {
-  return cast<MCSectionMachO>(Parent)->getAtom(LayoutOrder);
+  return static_cast<const MCSectionMachO *>(Parent)->getAtom(LayoutOrder);
 }
 
 #if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
diff --git a/llvm/lib/MC/MCMachOStreamer.cpp b/llvm/lib/MC/MCMachOStreamer.cpp
index 4934815..1074669 100644
--- a/llvm/lib/MC/MCMachOStreamer.cpp
+++ b/llvm/lib/MC/MCMachOStreamer.cpp
@@ -443,13 +443,13 @@ void MCMachOStreamer::finishImpl() {
   // Set the fragment atom associations by tracking the last seen atom defining
   // symbol.
   for (MCSection &Sec : getAssembler()) {
-    cast<MCSectionMachO>(Sec).allocAtoms();
+    static_cast<MCSectionMachO &>(Sec).allocAtoms();
     const MCSymbol *CurrentAtom = nullptr;
     size_t I = 0;
     for (MCFragment &Frag : Sec) {
       if (const MCSymbol *Symbol = DefiningSymbolMap.lookup(&Frag))
         CurrentAtom = Symbol;
-      cast<MCSectionMachO>(Sec).setAtom(I++, CurrentAtom);
+      static_cast<MCSectionMachO &>(Sec).setAtom(I++, CurrentAtom);
     }
   }
 
diff --git a/llvm/lib/MC/MCObjectStreamer.cpp b/llvm/lib/MC/MCObjectStreamer.cpp
index 37b6629..9c7b05b 100644
--- a/llvm/lib/MC/MCObjectStreamer.cpp
+++ b/llvm/lib/MC/MCObjectStreamer.cpp
@@ -280,11 +280,6 @@ void MCObjectStreamer::changeSection(MCSection *Section, uint32_t Subsection) {
   }
 }
 
-void MCObjectStreamer::switchSectionNoPrint(MCSection *Section) {
-  MCStreamer::switchSectionNoPrint(Section);
-  changeSection(Section, 0);
-}
-
 void MCObjectStreamer::emitAssignment(MCSymbol *Symbol, const MCExpr *Value) {
   getAssembler().registerSymbol(*Symbol);
   MCStreamer::emitAssignment(Symbol, Value);
@@ -345,31 +340,33 @@ void MCObjectStreamer::emitInstToData(const MCInst &Inst,
   MCFragment *F = getCurrentFragment();
 
   // Append the instruction to the data fragment.
-  size_t FixupStartIndex = F->getFixups().size();
   size_t CodeOffset = F->getContents().size();
   SmallVector<MCFixup, 1> Fixups;
   getAssembler().getEmitter().encodeInstruction(
       Inst, F->getContentsForAppending(), Fixups, STI);
   F->doneAppending();
-  if (!Fixups.empty())
-    F->appendFixups(Fixups);
   F->setHasInstructions(STI);
 
+  if (Fixups.empty())
+    return;
   bool MarkedLinkerRelaxable = false;
-  for (auto &Fixup : MutableArrayRef(F->getFixups()).slice(FixupStartIndex)) {
+  for (auto &Fixup : Fixups) {
     Fixup.setOffset(Fixup.getOffset() + CodeOffset);
-    if (!Fixup.isLinkerRelaxable())
+    if (!Fixup.isLinkerRelaxable() || MarkedLinkerRelaxable)
       continue;
-    F->setLinkerRelaxable();
+    MarkedLinkerRelaxable = true;
+    // Set the fragment's order within the subsection for use by
+    // MCAssembler::relaxAlign.
+    auto *Sec = F->getParent();
+    if (!Sec->isLinkerRelaxable())
+      Sec->setLinkerRelaxable();
     // Do not add data after a linker-relaxable instruction. The difference
     // between a new label and a label at or before the linker-relaxable
     // instruction cannot be resolved at assemble-time.
-    if (!MarkedLinkerRelaxable) {
-      MarkedLinkerRelaxable = true;
-      getCurrentSectionOnly()->setLinkerRelaxable();
-      newFragment();
-    }
+    F->setLinkerRelaxable();
+    newFragment();
   }
+  F->appendFixups(Fixups);
 }
 
 void MCObjectStreamer::emitInstToFragment(const MCInst &Inst,
diff --git a/llvm/lib/MC/MCParser/AsmParser.cpp b/llvm/lib/MC/MCParser/AsmParser.cpp
index eda5e8c..9f64a98 100644
--- a/llvm/lib/MC/MCParser/AsmParser.cpp
+++ b/llvm/lib/MC/MCParser/AsmParser.cpp
@@ -3413,7 +3413,7 @@ bool AsmParser::parseDirectiveAlign(bool IsPow2, uint8_t ValueSize) {
 
   // Check whether we should use optimal code alignment for this .align
   // directive.
-  if (Section->useCodeAlign() && !HasFillExpr) {
+  if (MAI.useCodeAlign(*Section) && !HasFillExpr) {
     getStreamer().emitCodeAlignment(
         Align(Alignment), &getTargetParser().getSTI(), MaxBytesToFill);
   } else {
diff --git a/llvm/lib/MC/MCParser/ELFAsmParser.cpp b/llvm/lib/MC/MCParser/ELFAsmParser.cpp
index c7c3df3..2e251cc 100644
--- a/llvm/lib/MC/MCParser/ELFAsmParser.cpp
+++ b/llvm/lib/MC/MCParser/ELFAsmParser.cpp
@@ -644,8 +644,8 @@ EndStmt:
   }
 
   if (UseLastGroup) {
-    if (const MCSectionELF *Section =
-            cast_or_null<MCSectionELF>(getStreamer().getCurrentSectionOnly()))
+    if (auto *Section = static_cast<const MCSectionELF *>(
+            getStreamer().getCurrentSectionOnly()))
       if (const MCSymbol *Group = Section->getGroup()) {
         GroupName = Group->getName();
         IsComdat = Section->isComdat();
diff --git a/llvm/lib/MC/MCParser/MasmParser.cpp b/llvm/lib/MC/MCParser/MasmParser.cpp
index f4684e6..780289e 100644
--- a/llvm/lib/MC/MCParser/MasmParser.cpp
+++ b/llvm/lib/MC/MCParser/MasmParser.cpp
@@ -4228,8 +4228,7 @@ bool MasmParser::emitAlignTo(int64_t Alignment) {
     // Check whether we should use optimal code alignment for this align
     // directive.
     const MCSection *Section = getStreamer().getCurrentSectionOnly();
-    assert(Section && "must have section to emit alignment");
-    if (Section->useCodeAlign()) {
+    if (MAI.useCodeAlign(*Section)) {
       getStreamer().emitCodeAlignment(Align(Alignment),
                                       &getTargetParser().getSTI(),
                                       /*MaxBytesToEmit=*/0);
diff --git a/llvm/lib/MC/MCParser/WasmAsmParser.cpp b/llvm/lib/MC/MCParser/WasmAsmParser.cpp
index 1f824b8..d97f4f5 100644
--- a/llvm/lib/MC/MCParser/WasmAsmParser.cpp
+++ b/llvm/lib/MC/MCParser/WasmAsmParser.cpp
@@ -252,7 +252,7 @@ public:
     if (TypeName == "function") {
       WasmSym->setType(wasm::WASM_SYMBOL_TYPE_FUNCTION);
       auto *Current =
-          cast<MCSectionWasm>(getStreamer().getCurrentSectionOnly());
+          static_cast<MCSectionWasm *>(getStreamer().getCurrentSectionOnly());
       if (Current->getGroup())
         WasmSym->setComdat(true);
     } else if (TypeName == "global")
diff --git a/llvm/lib/MC/MCSection.cpp b/llvm/lib/MC/MCSection.cpp
index e738a22..4f28267 100644
--- a/llvm/lib/MC/MCSection.cpp
+++ b/llvm/lib/MC/MCSection.cpp
@@ -18,10 +18,9 @@
 
 using namespace llvm;
 
-MCSection::MCSection(SectionVariant V, StringRef Name, bool IsText, bool IsBss,
-                     MCSymbol *Begin)
+MCSection::MCSection(StringRef Name, bool IsText, bool IsBss, MCSymbol *Begin)
     : Begin(Begin), HasInstructions(false), IsRegistered(false), IsText(IsText),
-      IsBss(IsBss), LinkerRelaxable(false), Name(Name), Variant(V) {
+      IsBss(IsBss), LinkerRelaxable(false), Name(Name) {
   DummyFragment.setParent(this);
 }
 
diff --git a/llvm/lib/MC/MCSectionCOFF.cpp b/llvm/lib/MC/MCSectionCOFF.cpp
deleted file mode 100644
index 5bf1473..0000000
--- a/llvm/lib/MC/MCSectionCOFF.cpp
+++ /dev/null
@@ -1,117 +0,0 @@
-//===- lib/MC/MCSectionCOFF.cpp - COFF Code Section Representation --------===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-#include "llvm/MC/MCSectionCOFF.h"
-#include "llvm/BinaryFormat/COFF.h"
-#include "llvm/MC/MCSymbol.h"
-#include "llvm/Support/raw_ostream.h"
-#include <cassert>
-
-using namespace llvm;
-
-// shouldOmitSectionDirective - Decides whether a '.section' directive
-// should be printed before the section name
-bool MCSectionCOFF::shouldOmitSectionDirective(StringRef Name,
-                                               const MCAsmInfo &MAI) const {
-  if (COMDATSymbol || isUnique())
-    return false;
-
-  // FIXME: Does .section .bss/.data/.text work everywhere??
-  if (Name == ".text" || Name == ".data" || Name == ".bss")
-    return true;
-
-  return false;
-}
-
-void MCSectionCOFF::setSelection(int Selection) const {
-  assert(Selection != 0 && "invalid COMDAT selection type");
-  this->Selection = Selection;
-  Characteristics |= COFF::IMAGE_SCN_LNK_COMDAT;
-}
-
-void MCSectionCOFF::printSwitchToSection(const MCAsmInfo &MAI, const Triple &T,
-                                         raw_ostream &OS,
-                                         uint32_t Subsection) const {
-  // standard sections don't require the '.section'
-  if (shouldOmitSectionDirective(getName(), MAI)) {
-    OS << '\t' << getName() << '\n';
-    return;
-  }
-
-  OS << "\t.section\t" << getName() << ",\"";
-  if (getCharacteristics() & COFF::IMAGE_SCN_CNT_INITIALIZED_DATA)
-    OS << 'd';
-  if (getCharacteristics() & COFF::IMAGE_SCN_CNT_UNINITIALIZED_DATA)
-    OS << 'b';
-  if (getCharacteristics() & COFF::IMAGE_SCN_MEM_EXECUTE)
-    OS << 'x';
-  if (getCharacteristics() & COFF::IMAGE_SCN_MEM_WRITE)
-    OS << 'w';
-  else if (getCharacteristics() & COFF::IMAGE_SCN_MEM_READ)
-    OS << 'r';
-  else
-    OS << 'y';
-  if (getCharacteristics() & COFF::IMAGE_SCN_LNK_REMOVE)
-    OS << 'n';
-  if (getCharacteristics() & COFF::IMAGE_SCN_MEM_SHARED)
-    OS << 's';
-  if ((getCharacteristics() & COFF::IMAGE_SCN_MEM_DISCARDABLE) &&
-      !isImplicitlyDiscardable(getName()))
-    OS << 'D';
-  if (getCharacteristics() & COFF::IMAGE_SCN_LNK_INFO)
-    OS << 'i';
-  OS << '"';
-
-  // unique should be tail of .section directive.
-  if (isUnique() && !COMDATSymbol)
-    OS << ",unique," << UniqueID;
-
-  if (getCharacteristics() & COFF::IMAGE_SCN_LNK_COMDAT) {
-    if (COMDATSymbol)
-      OS << ",";
-    else
-      OS << "\n\t.linkonce\t";
-    switch (Selection) {
-      case COFF::IMAGE_COMDAT_SELECT_NODUPLICATES:
-        OS << "one_only";
-        break;
-      case COFF::IMAGE_COMDAT_SELECT_ANY:
-        OS << "discard";
-        break;
-      case COFF::IMAGE_COMDAT_SELECT_SAME_SIZE:
-        OS << "same_size";
-        break;
-      case COFF::IMAGE_COMDAT_SELECT_EXACT_MATCH:
-        OS << "same_contents";
-        break;
-      case COFF::IMAGE_COMDAT_SELECT_ASSOCIATIVE:
-        OS << "associative";
-        break;
-      case COFF::IMAGE_COMDAT_SELECT_LARGEST:
-        OS << "largest";
-        break;
-      case COFF::IMAGE_COMDAT_SELECT_NEWEST:
-        OS << "newest";
-        break;
-      default:
-        assert(false && "unsupported COFF selection type");
-        break;
-    }
-    if (COMDATSymbol) {
-      OS << ",";
-      COMDATSymbol->print(OS, &MAI);
-    }
-  }
-
-  if (isUnique() && COMDATSymbol)
-    OS << ",unique," << UniqueID;
-
-  OS << '\n';
-}
-
-bool MCSectionCOFF::useCodeAlign() const { return isText(); }
diff --git a/llvm/lib/MC/MCSectionDXContainer.cpp b/llvm/lib/MC/MCSectionDXContainer.cpp
deleted file mode 100644
index 7eee59d..0000000
--- a/llvm/lib/MC/MCSectionDXContainer.cpp
+++ /dev/null
@@ -1,15 +0,0 @@
-//===- lib/MC/MCSectionDXContainer.cpp - DXContainer Section --------------===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-#include "llvm/MC/MCSectionDXContainer.h"
-
-using namespace llvm;
-
-void MCSectionDXContainer::printSwitchToSection(const MCAsmInfo &,
-                                                const Triple &, raw_ostream &,
-                                                uint32_t) const {}
diff --git a/llvm/lib/MC/MCSectionELF.cpp b/llvm/lib/MC/MCSectionELF.cpp
deleted file mode 100644
index ef33f9c..0000000
--- a/llvm/lib/MC/MCSectionELF.cpp
+++ /dev/null
@@ -1,217 +0,0 @@
-//===- lib/MC/MCSectionELF.cpp - ELF Code Section Representation ----------===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-#include "llvm/MC/MCSectionELF.h"
-#include "llvm/ADT/Twine.h"
-#include "llvm/BinaryFormat/ELF.h"
-#include "llvm/MC/MCAsmInfo.h"
-#include "llvm/MC/MCExpr.h"
-#include "llvm/Support/ErrorHandling.h"
-#include "llvm/Support/raw_ostream.h"
-#include "llvm/TargetParser/Triple.h"
-#include <cassert>
-
-using namespace llvm;
-
-// Decides whether a '.section' directive
-// should be printed before the section name.
-bool MCSectionELF::shouldOmitSectionDirective(StringRef Name,
-                                              const MCAsmInfo &MAI) const {
-  if (isUnique())
-    return false;
-
-  return MAI.shouldOmitSectionDirective(Name);
-}
-
-static void printName(raw_ostream &OS, StringRef Name) {
-  if (Name.find_first_not_of("0123456789_."
-                             "abcdefghijklmnopqrstuvwxyz"
-                             "ABCDEFGHIJKLMNOPQRSTUVWXYZ") == Name.npos) {
-    OS << Name;
-    return;
-  }
-  OS << '"';
-  for (const char *B = Name.begin(), *E = Name.end(); B < E; ++B) {
-    if (*B == '"') // Unquoted "
-      OS << "\\\"";
-    else if (*B != '\\') // Neither " or backslash
-      OS << *B;
-    else if (B + 1 == E) // Trailing backslash
-      OS << "\\\\";
-    else {
-      OS << B[0] << B[1]; // Quoted character
-      ++B;
-    }
-  }
-  OS << '"';
-}
-
-void MCSectionELF::printSwitchToSection(const MCAsmInfo &MAI, const Triple &T,
-                                        raw_ostream &OS,
-                                        uint32_t Subsection) const {
-  if (shouldOmitSectionDirective(getName(), MAI)) {
-    OS << '\t' << getName();
-    if (Subsection)
-      OS << '\t' << Subsection;
-    OS << '\n';
-    return;
-  }
-
-  OS << "\t.section\t";
-  printName(OS, getName());
-
-  // Handle the weird solaris syntax if desired.
-  if (MAI.usesSunStyleELFSectionSwitchSyntax() &&
-      !(Flags & ELF::SHF_MERGE)) {
-    if (Flags & ELF::SHF_ALLOC)
-      OS << ",#alloc";
-    if (Flags & ELF::SHF_EXECINSTR)
-      OS << ",#execinstr";
-    if (Flags & ELF::SHF_WRITE)
-      OS << ",#write";
-    if (Flags & ELF::SHF_EXCLUDE)
-      OS << ",#exclude";
-    if (Flags & ELF::SHF_TLS)
-      OS << ",#tls";
-    OS << '\n';
-    return;
-  }
-
-  OS << ",\"";
-  if (Flags & ELF::SHF_ALLOC)
-    OS << 'a';
-  if (Flags & ELF::SHF_EXCLUDE)
-    OS << 'e';
-  if (Flags & ELF::SHF_EXECINSTR)
-    OS << 'x';
-  if (Flags & ELF::SHF_WRITE)
-    OS << 'w';
-  if (Flags & ELF::SHF_MERGE)
-    OS << 'M';
-  if (Flags & ELF::SHF_STRINGS)
-    OS << 'S';
-  if (Flags & ELF::SHF_TLS)
-    OS << 'T';
-  if (Flags & ELF::SHF_LINK_ORDER)
-    OS << 'o';
-  if (Flags & ELF::SHF_GROUP)
-    OS << 'G';
-  if (Flags & ELF::SHF_GNU_RETAIN)
-    OS << 'R';
-
-  // If there are os-specific flags, print them.
-  if (T.isOSSolaris())
-    if (Flags & ELF::SHF_SUNW_NODISCARD)
-      OS << 'R';
-
-  // If there are target-specific flags, print them.
-  Triple::ArchType Arch = T.getArch();
-  if (Arch == Triple::xcore) {
-    if (Flags & ELF::XCORE_SHF_CP_SECTION)
-      OS << 'c';
-    if (Flags & ELF::XCORE_SHF_DP_SECTION)
-      OS << 'd';
-  } else if (T.isARM() || T.isThumb()) {
-    if (Flags & ELF::SHF_ARM_PURECODE)
-      OS << 'y';
-  } else if (T.isAArch64()) {
-    if (Flags & ELF::SHF_AARCH64_PURECODE)
-      OS << 'y';
-  } else if (Arch == Triple::hexagon) {
-    if (Flags & ELF::SHF_HEX_GPREL)
-      OS << 's';
-  } else if (Arch == Triple::x86_64) {
-    if (Flags & ELF::SHF_X86_64_LARGE)
-      OS << 'l';
-  }
-
-  OS << '"';
-
-  OS << ',';
-
-  // If comment string is '@', e.g. as on ARM - use '%' instead
-  if (MAI.getCommentString()[0] == '@')
-    OS << '%';
-  else
-    OS << '@';
-
-  if (Type == ELF::SHT_INIT_ARRAY)
-    OS << "init_array";
-  else if (Type == ELF::SHT_FINI_ARRAY)
-    OS << "fini_array";
-  else if (Type == ELF::SHT_PREINIT_ARRAY)
-    OS << "preinit_array";
-  else if (Type == ELF::SHT_NOBITS)
-    OS << "nobits";
-  else if (Type == ELF::SHT_NOTE)
-    OS << "note";
-  else if (Type == ELF::SHT_PROGBITS)
-    OS << "progbits";
-  else if (Type == ELF::SHT_X86_64_UNWIND)
-    OS << "unwind";
-  else if (Type == ELF::SHT_MIPS_DWARF)
-    // Print hex value of the flag while we do not have
-    // any standard symbolic representation of the flag.
-    OS << "0x7000001e";
-  else if (Type == ELF::SHT_LLVM_ODRTAB)
-    OS << "llvm_odrtab";
-  else if (Type == ELF::SHT_LLVM_LINKER_OPTIONS)
-    OS << "llvm_linker_options";
-  else if (Type == ELF::SHT_LLVM_CALL_GRAPH_PROFILE)
-    OS << "llvm_call_graph_profile";
-  else if (Type == ELF::SHT_LLVM_DEPENDENT_LIBRARIES)
-    OS << "llvm_dependent_libraries";
-  else if (Type == ELF::SHT_LLVM_SYMPART)
-    OS << "llvm_sympart";
-  else if (Type == ELF::SHT_LLVM_BB_ADDR_MAP)
-    OS << "llvm_bb_addr_map";
-  else if (Type == ELF::SHT_LLVM_OFFLOADING)
-    OS << "llvm_offloading";
-  else if (Type == ELF::SHT_LLVM_LTO)
-    OS << "llvm_lto";
-  else if (Type == ELF::SHT_LLVM_JT_SIZES)
-    OS << "llvm_jt_sizes";
-  else if (Type == ELF::SHT_LLVM_CFI_JUMP_TABLE)
-    OS << "llvm_cfi_jump_table";
-  else
-    OS << "0x" << Twine::utohexstr(Type);
-
-  if (EntrySize) {
-    assert((Flags & ELF::SHF_MERGE) || Type == ELF::SHT_LLVM_CFI_JUMP_TABLE);
-    OS << "," << EntrySize;
-  }
-
-  if (Flags & ELF::SHF_LINK_ORDER) {
-    OS << ",";
-    if (LinkedToSym)
-      printName(OS, LinkedToSym->getName());
-    else
-      OS << '0';
-  }
-
-  if (Flags & ELF::SHF_GROUP) {
-    OS << ",";
-    printName(OS, Group.getPointer()->getName());
-    if (isComdat())
-      OS << ",comdat";
-  }
-
-  if (isUnique())
-    OS << ",unique," << UniqueID;
-
-  OS << '\n';
-
-  if (Subsection) {
-    OS << "\t.subsection\t" << Subsection;
-    OS << '\n';
-  }
-}
-
-bool MCSectionELF::useCodeAlign() const {
-  return getFlags() & ELF::SHF_EXECINSTR;
-}
diff --git a/llvm/lib/MC/MCSectionGOFF.cpp b/llvm/lib/MC/MCSectionGOFF.cpp
deleted file mode 100644
index 8163e5b..0000000
--- a/llvm/lib/MC/MCSectionGOFF.cpp
+++ /dev/null
@@ -1,143 +0,0 @@
-//===- MCSectionGOFF.cpp - GOFF Code Section Representation ---------------===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-#include "llvm/MC/MCSectionGOFF.h"
-#include "llvm/BinaryFormat/GOFF.h"
-#include "llvm/Support/raw_ostream.h"
-
-using namespace llvm;
-
-static void emitCATTR(raw_ostream &OS, StringRef Name, GOFF::ESDRmode Rmode,
-                      GOFF::ESDAlignment Alignment,
-                      GOFF::ESDLoadingBehavior LoadBehavior,
-                      GOFF::ESDExecutable Executable, bool IsReadOnly,
-                      uint32_t SortKey, uint8_t FillByteValue,
-                      StringRef PartName) {
-  OS << Name << " CATTR ";
-  OS << "ALIGN(" << static_cast<unsigned>(Alignment) << "),"
-     << "FILL(" << static_cast<unsigned>(FillByteValue) << ")";
-  switch (LoadBehavior) {
-  case GOFF::ESD_LB_Deferred:
-    OS << ",DEFLOAD";
-    break;
-  case GOFF::ESD_LB_NoLoad:
-    OS << ",NOLOAD";
-    break;
-  default:
-    break;
-  }
-  switch (Executable) {
-  case GOFF::ESD_EXE_CODE:
-    OS << ",EXECUTABLE";
-    break;
-  case GOFF::ESD_EXE_DATA:
-    OS << ",NOTEXECUTABLE";
-    break;
-  default:
-    break;
-  }
-  if (IsReadOnly)
-    OS << ",READONLY";
-  if (Rmode != GOFF::ESD_RMODE_None) {
-    OS << ',';
-    OS << "RMODE(";
-    switch (Rmode) {
-    case GOFF::ESD_RMODE_24:
-      OS << "24";
-      break;
-    case GOFF::ESD_RMODE_31:
-      OS << "31";
-      break;
-    case GOFF::ESD_RMODE_64:
-      OS << "64";
-      break;
-    case GOFF::ESD_RMODE_None:
-      break;
-    }
-    OS << ')';
-  }
-  if (SortKey)
-    OS << ",PRIORITY(" << SortKey << ")";
-  if (!PartName.empty())
-    OS << ",PART(" << PartName << ")";
-  OS << '\n';
-}
-
-static void emitXATTR(raw_ostream &OS, StringRef Name,
-                      GOFF::ESDLinkageType Linkage,
-                      GOFF::ESDExecutable Executable,
-                      GOFF::ESDBindingScope BindingScope) {
-  OS << Name << " XATTR ";
-  OS << "LINKAGE(" << (Linkage == GOFF::ESD_LT_OS ? "OS" : "XPLINK") << "),";
-  if (Executable != GOFF::ESD_EXE_Unspecified)
-    OS << "REFERENCE(" << (Executable == GOFF::ESD_EXE_CODE ? "CODE" : "DATA")
-       << "),";
-  if (BindingScope != GOFF::ESD_BSC_Unspecified) {
-    OS << "SCOPE(";
-    switch (BindingScope) {
-    case GOFF::ESD_BSC_Section:
-      OS << "SECTION";
-      break;
-    case GOFF::ESD_BSC_Module:
-      OS << "MODULE";
-      break;
-    case GOFF::ESD_BSC_Library:
-      OS << "LIBRARY";
-      break;
-    case GOFF::ESD_BSC_ImportExport:
-      OS << "EXPORT";
-      break;
-    default:
-      break;
-    }
-    OS << ')';
-  }
-  OS << '\n';
-}
-
-void MCSectionGOFF::printSwitchToSection(const MCAsmInfo &MAI, const Triple &T,
-                                         raw_ostream &OS,
-                                         uint32_t Subsection) const {
-  switch (SymbolType) {
-  case GOFF::ESD_ST_SectionDefinition: {
-    OS << Name << " CSECT\n";
-    Emitted = true;
-    break;
-  }
-  case GOFF::ESD_ST_ElementDefinition: {
-    getParent()->printSwitchToSection(MAI, T, OS, Subsection);
-    if (!Emitted) {
-      emitCATTR(OS, Name, EDAttributes.Rmode, EDAttributes.Alignment,
-                EDAttributes.LoadBehavior, GOFF::ESD_EXE_Unspecified,
-                EDAttributes.IsReadOnly, 0, EDAttributes.FillByteValue,
-                StringRef());
-      Emitted = true;
-    } else
-      OS << Name << " CATTR\n";
-    break;
-  }
-  case GOFF::ESD_ST_PartReference: {
-    MCSectionGOFF *ED = getParent();
-    ED->getParent()->printSwitchToSection(MAI, T, OS, Subsection);
-    if (!Emitted) {
-      emitCATTR(OS, ED->getName(), ED->getEDAttributes().Rmode,
-                ED->EDAttributes.Alignment, ED->EDAttributes.LoadBehavior,
-                PRAttributes.Executable, ED->EDAttributes.IsReadOnly,
-                PRAttributes.SortKey, ED->EDAttributes.FillByteValue, Name);
-      emitXATTR(OS, Name, PRAttributes.Linkage, PRAttributes.Executable,
-                PRAttributes.BindingScope);
-      ED->Emitted = true;
-      Emitted = true;
-    } else
-      OS << ED->getName() << " CATTR PART(" << Name << ")\n";
-    break;
-  }
-  default:
-    llvm_unreachable("Wrong section type");
-  }
-}
\ No newline at end of file
diff --git a/llvm/lib/MC/MCSectionMachO.cpp b/llvm/lib/MC/MCSectionMachO.cpp
index 67453ce..67c8235 100644
--- a/llvm/lib/MC/MCSectionMachO.cpp
+++ b/llvm/lib/MC/MCSectionMachO.cpp
@@ -7,6 +7,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "llvm/MC/MCSectionMachO.h"
+#include "llvm/MC/MCAsmInfoDarwin.h"
 #include "llvm/MC/SectionKind.h"
 #include "llvm/Support/raw_ostream.h"
 
@@ -92,7 +93,7 @@ ENTRY("" /*FIXME*/,          S_ATTR_LOC_RELOC)
 MCSectionMachO::MCSectionMachO(StringRef Segment, StringRef Section,
                                unsigned TAA, unsigned reserved2, SectionKind K,
                                MCSymbol *Begin)
-    : MCSection(SV_MachO, Section, K.isText(),
+    : MCSection(Section, K.isText(),
                 MachO::isVirtualSection(TAA & MachO::SECTION_TYPE), Begin),
       TypeAndAttributes(TAA), Reserved2(reserved2) {
   assert(Segment.size() <= 16 && Section.size() <= 16 &&
@@ -105,19 +106,20 @@ MCSectionMachO::MCSectionMachO(StringRef Segment, StringRef Section,
   }
 }
 
-void MCSectionMachO::printSwitchToSection(const MCAsmInfo &MAI, const Triple &T,
-                                          raw_ostream &OS,
-                                          uint32_t Subsection) const {
-  OS << "\t.section\t" << getSegmentName() << ',' << getName();
+void MCAsmInfoDarwin::printSwitchToSection(const MCSection &Section, uint32_t,
+                                           const Triple &T,
+                                           raw_ostream &OS) const {
+  auto &Sec = static_cast<const MCSectionMachO &>(Section);
+  OS << "\t.section\t" << Sec.getSegmentName() << ',' << Sec.getName();
 
   // Get the section type and attributes.
-  unsigned TAA = getTypeAndAttributes();
+  unsigned TAA = Sec.getTypeAndAttributes();
   if (TAA == 0) {
     OS << '\n';
     return;
   }
 
-  MachO::SectionType SectionType = getType();
+  MachO::SectionType SectionType = Sec.getType();
   assert(SectionType <= MachO::LAST_KNOWN_SECTION_TYPE &&
          "Invalid SectionType specified!");
 
@@ -135,8 +137,8 @@ void MCSectionMachO::printSwitchToSection(const MCAsmInfo &MAI, const Triple &T,
   if (SectionAttrs == 0) {
     // If we have a S_SYMBOL_STUBS size specified, print it along with 'none' as
     // the attribute specifier.
-    if (Reserved2 != 0)
-      OS << ",none," << Reserved2;
+    if (Sec.Reserved2 != 0)
+      OS << ",none," << Sec.Reserved2;
     OS << '\n';
     return;
   }
@@ -164,15 +166,11 @@ void MCSectionMachO::printSwitchToSection(const MCAsmInfo &MAI, const Triple &T,
   assert(SectionAttrs == 0 && "Unknown section attributes!");
 
   // If we have a S_SYMBOL_STUBS size specified, print it.
-  if (Reserved2 != 0)
-    OS << ',' << Reserved2;
+  if (Sec.Reserved2 != 0)
+    OS << ',' << Sec.Reserved2;
   OS << '\n';
 }
 
-bool MCSectionMachO::useCodeAlign() const {
-  return hasAttribute(MachO::S_ATTR_PURE_INSTRUCTIONS);
-}
-
 /// ParseSectionSpecifier - Parse the section specifier indicated by "Spec".
 /// This is a string that can appear after a .section directive in a mach-o
 /// flavored .s file.  If successful, this fills in the specified Out
diff --git a/llvm/lib/MC/MCSectionWasm.cpp b/llvm/lib/MC/MCSectionWasm.cpp
deleted file mode 100644
index e25af1c..0000000
--- a/llvm/lib/MC/MCSectionWasm.cpp
+++ /dev/null
@@ -1,101 +0,0 @@
-//===- lib/MC/MCSectionWasm.cpp - Wasm Code Section Representation --------===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-#include "llvm/MC/MCSectionWasm.h"
-#include "llvm/MC/MCAsmInfo.h"
-#include "llvm/MC/MCExpr.h"
-#include "llvm/MC/MCSymbolWasm.h"
-#include "llvm/Support/raw_ostream.h"
-
-using namespace llvm;
-
-// Decides whether a '.section' directive
-// should be printed before the section name.
-bool MCSectionWasm::shouldOmitSectionDirective(StringRef Name,
-                                               const MCAsmInfo &MAI) const {
-  return MAI.shouldOmitSectionDirective(Name);
-}
-
-static void printName(raw_ostream &OS, StringRef Name) {
-  if (Name.find_first_not_of("0123456789_."
-                             "abcdefghijklmnopqrstuvwxyz"
-                             "ABCDEFGHIJKLMNOPQRSTUVWXYZ") == Name.npos) {
-    OS << Name;
-    return;
-  }
-  OS << '"';
-  for (const char *B = Name.begin(), *E = Name.end(); B < E; ++B) {
-    if (*B == '"') // Unquoted "
-      OS << "\\\"";
-    else if (*B != '\\') // Neither " or backslash
-      OS << *B;
-    else if (B + 1 == E) // Trailing backslash
-      OS << "\\\\";
-    else {
-      OS << B[0] << B[1]; // Quoted character
-      ++B;
-    }
-  }
-  OS << '"';
-}
-
-void MCSectionWasm::printSwitchToSection(const MCAsmInfo &MAI, const Triple &T,
-                                         raw_ostream &OS,
-                                         uint32_t Subsection) const {
-
-  if (shouldOmitSectionDirective(getName(), MAI)) {
-    OS << '\t' << getName();
-    if (Subsection)
-      OS << '\t' << Subsection;
-    OS << '\n';
-    return;
-  }
-
-  OS << "\t.section\t";
-  printName(OS, getName());
-  OS << ",\"";
-
-  if (IsPassive)
-    OS << 'p';
-  if (Group)
-    OS << 'G';
-  if (SegmentFlags & wasm::WASM_SEG_FLAG_STRINGS)
-    OS << 'S';
-  if (SegmentFlags & wasm::WASM_SEG_FLAG_TLS)
-    OS << 'T';
-  if (SegmentFlags & wasm::WASM_SEG_FLAG_RETAIN)
-    OS << 'R';
-
-  OS << '"';
-
-  OS << ',';
-
-  // If comment string is '@', e.g. as on ARM - use '%' instead
-  if (MAI.getCommentString()[0] == '@')
-    OS << '%';
-  else
-    OS << '@';
-
-  // TODO: Print section type.
-
-  if (Group) {
-    OS << ",";
-    printName(OS, Group->getName());
-    OS << ",comdat";
-  }
-
-  if (isUnique())
-    OS << ",unique," << UniqueID;
-
-  OS << '\n';
-
-  if (Subsection)
-    OS << "\t.subsection\t" << Subsection << '\n';
-}
-
-bool MCSectionWasm::useCodeAlign() const { return false; }
diff --git a/llvm/lib/MC/MCSectionXCOFF.cpp b/llvm/lib/MC/MCSectionXCOFF.cpp
deleted file mode 100644
index 41043b2..0000000
--- a/llvm/lib/MC/MCSectionXCOFF.cpp
+++ /dev/null
@@ -1,134 +0,0 @@
-//===- lib/MC/MCSectionXCOFF.cpp - XCOFF Code Section Representation ------===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-
-#include "llvm/MC/MCSectionXCOFF.h"
-#include "llvm/MC/MCAsmInfo.h"
-#include "llvm/Support/Format.h"
-#include "llvm/Support/raw_ostream.h"
-namespace llvm {
-class MCExpr;
-class Triple;
-} // namespace llvm
-
-using namespace llvm;
-
-MCSectionXCOFF::~MCSectionXCOFF() = default;
-
-void MCSectionXCOFF::printCsectDirective(raw_ostream &OS) const {
-  OS << "\t.csect " << QualName->getName() << "," << Log2(getAlign()) << '\n';
-}
-
-void MCSectionXCOFF::printSwitchToSection(const MCAsmInfo &MAI, const Triple &T,
-                                          raw_ostream &OS,
-                                          uint32_t Subsection) const {
-  if (getKind().isText()) {
-    if (getMappingClass() != XCOFF::XMC_PR)
-      report_fatal_error("Unhandled storage-mapping class for .text csect");
-
-    printCsectDirective(OS);
-    return;
-  }
-
-  if (getKind().isReadOnly()) {
-    if (getMappingClass() != XCOFF::XMC_RO &&
-        getMappingClass() != XCOFF::XMC_TD)
-      report_fatal_error("Unhandled storage-mapping class for .rodata csect.");
-    printCsectDirective(OS);
-    return;
-  }
-
-  if (getKind().isReadOnlyWithRel()) {
-    if (getMappingClass() != XCOFF::XMC_RW &&
-        getMappingClass() != XCOFF::XMC_RO &&
-        getMappingClass() != XCOFF::XMC_TD)
-      report_fatal_error(
-          "Unexepected storage-mapping class for ReadOnlyWithRel kind");
-    printCsectDirective(OS);
-    return;
-  }
-
-  // Initialized TLS data.
-  if (getKind().isThreadData()) {
-    // We only expect XMC_TL here for initialized TLS data.
-    if (getMappingClass() != XCOFF::XMC_TL)
-      report_fatal_error("Unhandled storage-mapping class for .tdata csect.");
-    printCsectDirective(OS);
-    return;
-  }
-
-  if (getKind().isData()) {
-    switch (getMappingClass()) {
-    case XCOFF::XMC_RW:
-    case XCOFF::XMC_DS:
-    case XCOFF::XMC_TD:
-      printCsectDirective(OS);
-      break;
-    case XCOFF::XMC_TC:
-    case XCOFF::XMC_TE:
-      break;
-    case XCOFF::XMC_TC0:
-      OS << "\t.toc\n";
-      break;
-    default:
-      report_fatal_error(
-          "Unhandled storage-mapping class for .data csect.");
-    }
-    return;
-  }
-
-  if (isCsect() && getMappingClass() == XCOFF::XMC_TD) {
-    // Common csect type (uninitialized storage) does not have to print csect
-    // directive for section switching unless it is local.
-    if (getKind().isCommon() && !getKind().isBSSLocal())
-      return;
-
-    assert(getKind().isBSS() && "Unexpected section kind for toc-data");
-    printCsectDirective(OS);
-    return;
-  }
-  // Common csect type (uninitialized storage) does not have to print csect
-  // directive for section switching.
-  if (isCsect() && getCSectType() == XCOFF::XTY_CM) {
-    assert((getMappingClass() == XCOFF::XMC_RW ||
-            getMappingClass() == XCOFF::XMC_BS ||
-            getMappingClass() == XCOFF::XMC_UL) &&
-           "Generated a storage-mapping class for a common/bss/tbss csect we "
-           "don't "
-           "understand how to switch to.");
-    // Common symbols and local zero-initialized symbols for TLS and Non-TLS are
-    // eligible for .bss/.tbss csect, getKind().isThreadBSS() is used to cover
-    // TLS common and zero-initialized local symbols since linkage type (in the
-    // GlobalVariable) is not accessible in this class.
-    assert((getKind().isBSSLocal() || getKind().isCommon() ||
-            getKind().isThreadBSS()) &&
-           "wrong symbol type for .bss/.tbss csect");
-    // Don't have to print a directive for switching to section for commons and
-    // zero-initialized TLS data. The '.comm' and '.lcomm' directives of the
-    // variable will create the needed csect.
-    return;
-  }
-
-  // Zero-initialized TLS data with weak or external linkage are not eligible to
-  // be put into common csect.
-  if (getKind().isThreadBSS()) {
-    printCsectDirective(OS);
-    return;
-  }
-
-  // XCOFF debug sections.
-  if (getKind().isMetadata() && isDwarfSect()) {
-    OS << "\n\t.dwsect " << format("0x%" PRIx32, *getDwarfSubtypeFlags())
-       << '\n';
-    OS << getName() << ':' << '\n';
-    return;
-  }
-
-  report_fatal_error("Printing for this SectionKind is unimplemented.");
-}
-
-bool MCSectionXCOFF::useCodeAlign() const { return getKind().isText(); }
diff --git a/llvm/lib/MC/MCStreamer.cpp b/llvm/lib/MC/MCStreamer.cpp
index 899a7df..bc73981 100644
--- a/llvm/lib/MC/MCStreamer.cpp
+++ b/llvm/lib/MC/MCStreamer.cpp
@@ -56,12 +56,11 @@ void MCTargetStreamer::finish() {}
 
 void MCTargetStreamer::emitConstantPools() {}
 
-void MCTargetStreamer::changeSection(const MCSection *CurSection,
-                                     MCSection *Section, uint32_t Subsection,
-                                     raw_ostream &OS) {
-  Section->printSwitchToSection(*Streamer.getContext().getAsmInfo(),
-                                Streamer.getContext().getTargetTriple(), OS,
-                                Subsection);
+void MCTargetStreamer::changeSection(const MCSection *, MCSection *Sec,
+                                     uint32_t Subsection, raw_ostream &OS) {
+  auto &MAI = *Streamer.getContext().getAsmInfo();
+  MAI.printSwitchToSection(*Sec, Subsection,
+                           Streamer.getContext().getTargetTriple(), OS);
 }
 
 void MCTargetStreamer::emitDwarfFileDirective(StringRef Directive) {
@@ -838,8 +837,8 @@ static MCSection *getWinCFISection(MCContext &Context, unsigned *NextWinCFIID,
   if (TextSec == Context.getObjectFileInfo()->getTextSection())
     return MainCFISec;
 
-  const auto *TextSecCOFF = cast<MCSectionCOFF>(TextSec);
-  auto *MainCFISecCOFF = cast<MCSectionCOFF>(MainCFISec);
+  const auto *TextSecCOFF = static_cast<const MCSectionCOFF *>(TextSec);
+  auto *MainCFISecCOFF = static_cast<MCSectionCOFF *>(MainCFISec);
   unsigned UniqueID = TextSecCOFF->getOrAssignWinCFISectionID(NextWinCFIID);
 
   // If this section is COMDAT, this unwind section should be COMDAT associative
@@ -1314,11 +1313,20 @@ void MCStreamer::emitZerofill(MCSection *, MCSymbol *, uint64_t, Align, SMLoc) {
 }
 void MCStreamer::emitTBSSSymbol(MCSection *Section, MCSymbol *Symbol,
                                 uint64_t Size, Align ByteAlignment) {}
+
 void MCStreamer::changeSection(MCSection *Sec, uint32_t) {
   CurFrag = &Sec->getDummyFragment();
-  if (auto *Sym = Sec->getBeginSymbol())
-    Sym->setFragment(&Sec->getDummyFragment());
+  auto *Sym = Sec->getBeginSymbol();
+  if (!Sym || !Sym->isUndefined())
+    return;
+  // In Mach-O, DWARF sections use Begin as a temporary label, requiring a label
+  // definition, unlike section symbols in other file formats.
+  if (getContext().getObjectFileType() == MCContext::IsMachO)
+    emitLabel(Sym);
+  else
+    Sym->setFragment(CurFrag);
 }
+
 void MCStreamer::emitWeakReference(MCSymbol *Alias, const MCSymbol *Symbol) {}
 void MCStreamer::emitBytes(StringRef Data) {}
 void MCStreamer::emitBinaryData(StringRef Data) { emitBytes(Data); }
diff --git a/llvm/lib/MC/MCWasmStreamer.cpp b/llvm/lib/MC/MCWasmStreamer.cpp
index 5891420c..e3ef111 100644
--- a/llvm/lib/MC/MCWasmStreamer.cpp
+++ b/llvm/lib/MC/MCWasmStreamer.cpp
@@ -58,7 +58,7 @@ void MCWasmStreamer::emitLabelAtPos(MCSymbol *S, SMLoc Loc, MCFragment &F,
 
 void MCWasmStreamer::changeSection(MCSection *Section, uint32_t Subsection) {
   MCAssembler &Asm = getAssembler();
-  auto *SectionWasm = cast<MCSectionWasm>(Section);
+  auto *SectionWasm = static_cast<const MCSectionWasm *>(Section);
   const MCSymbol *Grp = SectionWasm->getGroup();
   if (Grp)
     Asm.registerSymbol(*Grp);
diff --git a/llvm/lib/MC/MCWinCOFFStreamer.cpp b/llvm/lib/MC/MCWinCOFFStreamer.cpp
index 9369bea..1ffe25c 100644
--- a/llvm/lib/MC/MCWinCOFFStreamer.cpp
+++ b/llvm/lib/MC/MCWinCOFFStreamer.cpp
@@ -157,7 +157,8 @@ void MCWinCOFFStreamer::changeSection(MCSection *Section, uint32_t Subsection) {
   // Ensure that the first and the second symbols relative to the section are
   // the section symbol and the COMDAT symbol.
   getAssembler().registerSymbol(*Section->getBeginSymbol());
-  if (auto *Sym = cast<MCSectionCOFF>(Section)->getCOMDATSymbol())
+  if (auto *Sym =
+          static_cast<const MCSectionCOFF *>(Section)->getCOMDATSymbol())
     getAssembler().registerSymbol(*Sym);
 }
 
diff --git a/llvm/lib/MC/MCXCOFFStreamer.cpp b/llvm/lib/MC/MCXCOFFStreamer.cpp
index 78e4b95..898ac5d 100644
--- a/llvm/lib/MC/MCXCOFFStreamer.cpp
+++ b/llvm/lib/MC/MCXCOFFStreamer.cpp
@@ -38,7 +38,7 @@ XCOFFObjectWriter &MCXCOFFStreamer::getWriter() {
 
 void MCXCOFFStreamer::changeSection(MCSection *Section, uint32_t Subsection) {
   MCObjectStreamer::changeSection(Section, Subsection);
-  auto *Sec = cast<MCSectionXCOFF>(Section);
+  auto *Sec = static_cast<const MCSectionXCOFF *>(Section);
   // We might miss calculating the symbols difference as absolute value before
   // adding fixups when symbol_A without the fragment set is the csect itself
   // and symbol_B is in it.
diff --git a/llvm/lib/MC/MachObjectWriter.cpp b/llvm/lib/MC/MachObjectWriter.cpp
index 48d2fc6..7b5c3c0 100644
--- a/llvm/lib/MC/MachObjectWriter.cpp
+++ b/llvm/lib/MC/MachObjectWriter.cpp
@@ -126,7 +126,8 @@ uint64_t MachObjectWriter::getSymbolAddress(const MCSymbol &S) const {
 uint64_t MachObjectWriter::getPaddingSize(const MCAssembler &Asm,
                                           const MCSection *Sec) const {
   uint64_t EndAddr = getSectionAddress(Sec) + Asm.getSectionAddressSize(*Sec);
-  unsigned Next = cast<MCSectionMachO>(Sec)->getLayoutOrder() + 1;
+  unsigned Next =
+      static_cast<const MCSectionMachO *>(Sec)->getLayoutOrder() + 1;
   if (Next >= SectionOrder.size())
     return 0;
 
@@ -259,15 +260,12 @@ void MachObjectWriter::writeSegmentLoadCommand(
 }
 
 void MachObjectWriter::writeSection(const MCAssembler &Asm,
-                                    const MCSection &Sec, uint64_t VMAddr,
+                                    const MCSectionMachO &Sec, uint64_t VMAddr,
                                     uint64_t FileOffset, unsigned Flags,
                                     uint64_t RelocationsStart,
                                     unsigned NumRelocations) {
-  uint64_t SectionSize = Asm.getSectionAddressSize(Sec);
-  const MCSectionMachO &Section = cast<MCSectionMachO>(Sec);
-
   // The offset is unused for virtual sections.
-  if (Section.isBssSection()) {
+  if (Sec.isBssSection()) {
     assert(Asm.getSectionFileSize(Sec) == 0 && "Invalid file size!");
     FileOffset = 0;
   }
@@ -275,11 +273,11 @@ void MachObjectWriter::writeSection(const MCAssembler &Asm,
   // struct section (68 bytes) or
   // struct section_64 (80 bytes)
 
+  uint64_t SectionSize = Asm.getSectionAddressSize(Sec);
   uint64_t Start = W.OS.tell();
   (void) Start;
-
-  writeWithPadding(Section.getName(), 16);
-  writeWithPadding(Section.getSegmentName(), 16);
+  writeWithPadding(Sec.getName(), 16);
+  writeWithPadding(Sec.getSegmentName(), 16);
   if (is64Bit()) {
     W.write<uint64_t>(VMAddr);      // address
     W.write<uint64_t>(SectionSize); // size
@@ -290,14 +288,14 @@ void MachObjectWriter::writeSection(const MCAssembler &Asm,
   assert(isUInt<32>(FileOffset) && "Cannot encode offset of section");
   W.write<uint32_t>(FileOffset);
 
-  W.write<uint32_t>(Log2(Section.getAlign()));
+  W.write<uint32_t>(Log2(Sec.getAlign()));
   assert((!NumRelocations || isUInt<32>(RelocationsStart)) &&
          "Cannot encode offset of relocations");
   W.write<uint32_t>(NumRelocations ? RelocationsStart : 0);
   W.write<uint32_t>(NumRelocations);
   W.write<uint32_t>(Flags);
   W.write<uint32_t>(IndirectSymBase.lookup(&Sec)); // reserved1
-  W.write<uint32_t>(Section.getStubSize()); // reserved2
+  W.write<uint32_t>(Sec.getStubSize());            // reserved2
   if (is64Bit())
     W.write<uint32_t>(0); // reserved3
 
@@ -531,7 +529,7 @@ void MachObjectWriter::bindIndirectSymbols(MCAssembler &Asm) {
   // Report errors for use of .indirect_symbol not in a symbol pointer section
   // or stub section.
   for (IndirectSymbolData &ISD : IndirectSymbols) {
-    const MCSectionMachO &Section = cast<MCSectionMachO>(*ISD.Section);
+    const MCSectionMachO &Section = static_cast<MCSectionMachO &>(*ISD.Section);
 
     if (Section.getType() != MachO::S_NON_LAZY_SYMBOL_POINTERS &&
         Section.getType() != MachO::S_LAZY_SYMBOL_POINTERS &&
@@ -545,7 +543,7 @@ void MachObjectWriter::bindIndirectSymbols(MCAssembler &Asm) {
 
   // Bind non-lazy symbol pointers first.
   for (auto [IndirectIndex, ISD] : enumerate(IndirectSymbols)) {
-    const auto &Section = cast<MCSectionMachO>(*ISD.Section);
+    const auto &Section = static_cast<MCSectionMachO &>(*ISD.Section);
 
     if (Section.getType() != MachO::S_NON_LAZY_SYMBOL_POINTERS &&
         Section.getType() !=  MachO::S_THREAD_LOCAL_VARIABLE_POINTERS)
@@ -559,7 +557,7 @@ void MachObjectWriter::bindIndirectSymbols(MCAssembler &Asm) {
 
   // Then lazy symbol pointers and symbol stubs.
   for (auto [IndirectIndex, ISD] : enumerate(IndirectSymbols)) {
-    const auto &Section = cast<MCSectionMachO>(*ISD.Section);
+    const auto &Section = static_cast<MCSectionMachO &>(*ISD.Section);
 
     if (Section.getType() != MachO::S_LAZY_SYMBOL_POINTERS &&
         Section.getType() != MachO::S_SYMBOL_STUBS)
@@ -684,13 +682,13 @@ void MachObjectWriter::computeSectionAddresses(const MCAssembler &Asm) {
   for (MCSection &Sec : Asm) {
     if (!Sec.isBssSection()) {
       SectionOrder.push_back(&Sec);
-      cast<MCSectionMachO>(Sec).setLayoutOrder(i++);
+      static_cast<MCSectionMachO &>(Sec).setLayoutOrder(i++);
     }
   }
   for (MCSection &Sec : Asm) {
     if (Sec.isBssSection()) {
       SectionOrder.push_back(&Sec);
-      cast<MCSectionMachO>(Sec).setLayoutOrder(i++);
+      static_cast<MCSectionMachO &>(Sec).setLayoutOrder(i++);
     }
   }
 
@@ -907,7 +905,7 @@ uint64_t MachObjectWriter::writeObject() {
   // ... and then the section headers.
   uint64_t RelocTableEnd = SectionDataStart + SectionDataFileSize;
   for (const MCSection &Section : Asm) {
-    const auto &Sec = cast<MCSectionMachO>(Section);
+    const auto &Sec = static_cast<const MCSectionMachO &>(Section);
     std::vector<RelAndSymbol> &Relocs = Relocations[&Sec];
     unsigned NumRelocs = Relocs.size();
     uint64_t SectionStart = SectionDataStart + getSectionAddress(&Sec);
diff --git a/llvm/lib/MC/WasmObjectWriter.cpp b/llvm/lib/MC/WasmObjectWriter.cpp
index 3b99af4..bfd6334 100644
--- a/llvm/lib/MC/WasmObjectWriter.cpp
+++ b/llvm/lib/MC/WasmObjectWriter.cpp
@@ -480,7 +480,7 @@ void WasmObjectWriter::recordRelocation(const MCFragment &F,
   // The WebAssembly backend should never generate FKF_IsPCRel fixups
   assert(!Fixup.isPCRel());
 
-  const auto &FixupSection = cast<MCSectionWasm>(*F.getParent());
+  const auto &FixupSection = static_cast<MCSectionWasm &>(*F.getParent());
   uint64_t C = Target.getConstant();
   uint64_t FixupOffset = Asm->getFragmentOffset(F) + Fixup.getOffset();
   MCContext &Ctx = getContext();
diff --git a/llvm/lib/MC/WinCOFFObjectWriter.cpp b/llvm/lib/MC/WinCOFFObjectWriter.cpp
index 6ad4334..856850d 100644
--- a/llvm/lib/MC/WinCOFFObjectWriter.cpp
+++ b/llvm/lib/MC/WinCOFFObjectWriter.cpp
@@ -373,7 +373,7 @@ void WinCOFFWriter::defineSymbol(const MCSymbol &MCSym) {
   COFFSection *Sec = nullptr;
   MCSectionCOFF *MCSec = nullptr;
   if (Base && Base->getFragment()) {
-    MCSec = cast<MCSectionCOFF>(Base->getFragment()->getParent());
+    MCSec = static_cast<MCSectionCOFF *>(Base->getFragment()->getParent());
     Sec = SectionMap[MCSec];
   }
 
@@ -1057,7 +1057,8 @@ uint64_t WinCOFFWriter::writeObject() {
       continue;
     }
 
-    const auto *AssocMCSec = cast<MCSectionCOFF>(&AssocMCSym->getSection());
+    const auto *AssocMCSec =
+        static_cast<const MCSectionCOFF *>(&AssocMCSym->getSection());
     assert(SectionMap.count(AssocMCSec));
     COFFSection *AssocSec = SectionMap[AssocMCSec];
 
diff --git a/llvm/lib/MC/XCOFFObjectWriter.cpp b/llvm/lib/MC/XCOFFObjectWriter.cpp
index 2f6785f..65f543b 100644
--- a/llvm/lib/MC/XCOFFObjectWriter.cpp
+++ b/llvm/lib/MC/XCOFFObjectWriter.cpp
@@ -550,13 +550,13 @@ CsectGroup &XCOFFWriter::getCsectGroup(const MCSectionXCOFF *MCSec) {
 
 static MCSectionXCOFF *getContainingCsect(const MCSymbolXCOFF *XSym) {
   if (XSym->isDefined())
-    return cast<MCSectionXCOFF>(XSym->getFragment()->getParent());
+    return static_cast<MCSectionXCOFF *>(XSym->getFragment()->getParent());
   return XSym->getRepresentedCsect();
 }
 
 void XCOFFWriter::executePostLayoutBinding() {
   for (const auto &S : *Asm) {
-    const auto *MCSec = cast<const MCSectionXCOFF>(&S);
+    auto *MCSec = static_cast<const MCSectionXCOFF *>(&S);
     assert(!SectionMap.contains(MCSec) && "Cannot add a section twice.");
 
     // If the name does not fit in the storage provided in the symbol table
@@ -747,7 +747,7 @@ void XCOFFWriter::recordRelocation(const MCFragment &F, const MCFixup &Fixup,
       FixedValue = TOCEntryOffset;
     }
   } else if (Type == XCOFF::RelocationType::R_RBR) {
-    MCSectionXCOFF *ParentSec = cast<MCSectionXCOFF>(F.getParent());
+    auto *ParentSec = static_cast<MCSectionXCOFF *>(F.getParent());
     assert((SymASec->getMappingClass() == XCOFF::XMC_PR &&
             ParentSec->getMappingClass() == XCOFF::XMC_PR) &&
            "Only XMC_PR csect may have the R_RBR relocation.");
@@ -768,7 +768,7 @@ void XCOFFWriter::recordRelocation(const MCFragment &F, const MCFixup &Fixup,
   }
 
   XCOFFRelocation Reloc = {Index, FixupOffsetInCsect, SignAndSize, Type};
-  MCSectionXCOFF *RelocationSec = cast<MCSectionXCOFF>(F.getParent());
+  auto *RelocationSec = static_cast<MCSectionXCOFF *>(F.getParent());
   assert(SectionMap.contains(RelocationSec) &&
          "Expected containing csect to exist in map.");
   SectionMap[RelocationSec]->Relocations.push_back(Reloc);
diff --git a/llvm/lib/Object/IRSymtab.cpp b/llvm/lib/Object/IRSymtab.cpp
index 2579fa3..0f19495 100644
--- a/llvm/lib/Object/IRSymtab.cpp
+++ b/llvm/lib/Object/IRSymtab.cpp
@@ -8,11 +8,11 @@
 
 #include "llvm/Object/IRSymtab.h"
 #include "llvm/ADT/ArrayRef.h"
-#include "llvm/ADT/DenseMap.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/ADT/SmallString.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/StringRef.h"
+#include "llvm/ADT/StringSet.h"
 #include "llvm/Bitcode/BitcodeReader.h"
 #include "llvm/Config/llvm-config.h"
 #include "llvm/IR/Comdat.h"
@@ -213,9 +213,10 @@ Expected<int> Builder::getComdatIndex(const Comdat *C, const Module *M) {
   return P.first->second;
 }
 
-static DenseSet<StringRef> buildPreservedSymbolsSet(const Triple &TT) {
-  DenseSet<StringRef> PreservedSymbolSet(std::begin(PreservedSymbols),
-                                         std::end(PreservedSymbols));
+static StringSet<> buildPreservedSymbolsSet(const Triple &TT) {
+  StringSet<> PreservedSymbolSet;
+  PreservedSymbolSet.insert(std::begin(PreservedSymbols),
+                            std::end(PreservedSymbols));
   // FIXME: Do we need to pass in ABI fields from TargetOptions?
   RTLIB::RuntimeLibcallsInfo Libcalls(TT);
   for (RTLIB::LibcallImpl Impl : Libcalls.getLibcallImpls()) {
@@ -280,7 +281,7 @@ Error Builder::addSymbol(const ModuleSymbolTable &Msymtab,
 
   setStr(Sym.IRName, GV->getName());
 
-  static const DenseSet<StringRef> PreservedSymbolsSet =
+  static const StringSet<> PreservedSymbolsSet =
       buildPreservedSymbolsSet(GV->getParent()->getTargetTriple());
   bool IsPreservedSymbol = PreservedSymbolsSet.contains(GV->getName());
 
diff --git a/llvm/lib/Passes/PassRegistry.def b/llvm/lib/Passes/PassRegistry.def
index bb7ccdb..fd89583 100644
--- a/llvm/lib/Passes/PassRegistry.def
+++ b/llvm/lib/Passes/PassRegistry.def
@@ -119,7 +119,6 @@ MODULE_PASS("module-inline", ModuleInlinerPass())
 MODULE_PASS("name-anon-globals", NameAnonGlobalPass())
 MODULE_PASS("no-op-module", NoOpModulePass())
 MODULE_PASS("nsan", NumericalStabilitySanitizerPass())
-MODULE_PASS("objc-arc-apelim", ObjCARCAPElimPass())
 MODULE_PASS("openmp-opt", OpenMPOptPass())
 MODULE_PASS("openmp-opt-postlink",
             OpenMPOptPass(ThinOrFullLTOPhase::FullLTOPostLink))
diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
index 802e4a9..8685d7a0 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
+++ b/llvm/lib/Target/AArch64/AArch64InstrInfo.cpp
@@ -20,7 +20,6 @@
 #include "Utils/AArch64BaseInfo.h"
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/STLExtras.h"
-#include "llvm/ADT/SmallSet.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/CodeGen/CFIInstBuilder.h"
 #include "llvm/CodeGen/LivePhysRegs.h"
@@ -36,7 +35,6 @@
 #include "llvm/CodeGen/MachineRegisterInfo.h"
 #include "llvm/CodeGen/RegisterScavenging.h"
 #include "llvm/CodeGen/StackMaps.h"
-#include "llvm/CodeGen/TargetOpcodes.h"
 #include "llvm/CodeGen/TargetRegisterInfo.h"
 #include "llvm/CodeGen/TargetSubtargetInfo.h"
 #include "llvm/IR/DebugInfoMetadata.h"
@@ -7354,9 +7352,6 @@ bool AArch64InstrInfo::isThroughputPattern(unsigned Pattern) const {
   case AArch64MachineCombinerPattern::MULSUBv2i32_indexed_OP2:
   case AArch64MachineCombinerPattern::MULSUBv4i32_indexed_OP1:
   case AArch64MachineCombinerPattern::MULSUBv4i32_indexed_OP2:
-  case AArch64MachineCombinerPattern::GATHER_LANE_i32:
-  case AArch64MachineCombinerPattern::GATHER_LANE_i16:
-  case AArch64MachineCombinerPattern::GATHER_LANE_i8:
     return true;
   } // end switch (Pattern)
   return false;
@@ -7397,252 +7392,11 @@ static bool getMiscPatterns(MachineInstr &Root,
   return false;
 }
 
-static bool getGatherPattern(MachineInstr &Root,
-                             SmallVectorImpl<unsigned> &Patterns,
-                             unsigned LoadLaneOpCode, unsigned NumLanes) {
-  const MachineFunction *MF = Root.getMF();
-
-  // Early exit if optimizing for size.
-  if (MF->getFunction().hasMinSize())
-    return false;
-
-  const MachineRegisterInfo &MRI = MF->getRegInfo();
-  const TargetRegisterInfo *TRI = MF->getSubtarget().getRegisterInfo();
-
-  // The root of the pattern must load into the last lane of the vector.
-  if (Root.getOperand(2).getImm() != NumLanes - 1)
-    return false;
-
-  // Check that we have load into all lanes except lane 0.
-  // For each load we also want to check that:
-  // 1. It has a single non-debug use (since we will be replacing the virtual
-  // register)
-  // 2. That the addressing mode only uses a single offset register.
-  auto *CurrInstr = MRI.getUniqueVRegDef(Root.getOperand(1).getReg());
-  auto Range = llvm::seq<unsigned>(1, NumLanes - 1);
-  SmallSet<unsigned, 4> RemainingLanes(Range.begin(), Range.end());
-  while (!RemainingLanes.empty() && CurrInstr &&
-         CurrInstr->getOpcode() == LoadLaneOpCode &&
-         MRI.hasOneNonDBGUse(CurrInstr->getOperand(0).getReg()) &&
-         CurrInstr->getNumOperands() == 4) {
-    RemainingLanes.erase(CurrInstr->getOperand(2).getImm());
-    CurrInstr = MRI.getUniqueVRegDef(CurrInstr->getOperand(1).getReg());
-  }
-
-  if (!RemainingLanes.empty())
-    return false;
-
-  // Match the SUBREG_TO_REG sequence.
-  if (CurrInstr->getOpcode() != TargetOpcode::SUBREG_TO_REG)
-    return false;
-
-  // Verify that the subreg to reg loads an integer into the first lane.
-  auto Lane0LoadReg = CurrInstr->getOperand(2).getReg();
-  unsigned SingleLaneSizeInBits = 128 / NumLanes;
-  if (TRI->getRegSizeInBits(Lane0LoadReg, MRI) != SingleLaneSizeInBits)
-    return false;
-
-  // Verify that it also has a single non debug use.
-  if (!MRI.hasOneNonDBGUse(Lane0LoadReg))
-    return false;
-
-  switch (NumLanes) {
-  case 4:
-    Patterns.push_back(AArch64MachineCombinerPattern::GATHER_LANE_i32);
-    break;
-  case 8:
-    Patterns.push_back(AArch64MachineCombinerPattern::GATHER_LANE_i16);
-    break;
-  case 16:
-    Patterns.push_back(AArch64MachineCombinerPattern::GATHER_LANE_i8);
-    break;
-  default:
-    llvm_unreachable("Got bad number of lanes for gather pattern.");
-  }
-
-  return true;
-}
-
-/// Search for patterns where we use LD1 instructions to load into
-/// separate lanes of an 128 bit Neon register. We can increase Memory Level
-/// Parallelism by loading into 2 Neon registers instead.
-static bool getLoadPatterns(MachineInstr &Root,
-                            SmallVectorImpl<unsigned> &Patterns) {
-
-  // The pattern searches for loads into single lanes.
-  switch (Root.getOpcode()) {
-  case AArch64::LD1i32:
-    return getGatherPattern(Root, Patterns, Root.getOpcode(), 4);
-  case AArch64::LD1i16:
-    return getGatherPattern(Root, Patterns, Root.getOpcode(), 8);
-  case AArch64::LD1i8:
-    return getGatherPattern(Root, Patterns, Root.getOpcode(), 16);
-  default:
-    return false;
-  }
-}
-
-static void
-generateGatherPattern(MachineInstr &Root,
-                      SmallVectorImpl<MachineInstr *> &InsInstrs,
-                      SmallVectorImpl<MachineInstr *> &DelInstrs,
-                      DenseMap<Register, unsigned> &InstrIdxForVirtReg,
-                      unsigned Pattern, unsigned NumLanes) {
-
-  MachineFunction &MF = *Root.getParent()->getParent();
-  MachineRegisterInfo &MRI = MF.getRegInfo();
-  const TargetInstrInfo *TII = MF.getSubtarget().getInstrInfo();
-
-  // Gather the initial load instructions to build the pattern
-  SmallVector<MachineInstr *, 16> LoadToLaneInstrs;
-  MachineInstr *CurrInstr = &Root;
-  for (unsigned i = 0; i < NumLanes - 1; ++i) {
-    LoadToLaneInstrs.push_back(CurrInstr);
-    CurrInstr = MRI.getUniqueVRegDef(CurrInstr->getOperand(1).getReg());
-  }
-
-  // Sort the load instructions according to the lane.
-  llvm::sort(LoadToLaneInstrs,
-             [](const MachineInstr *A, const MachineInstr *B) {
-               return A->getOperand(2).getImm() > B->getOperand(2).getImm();
-             });
-
-  MachineInstr *SubregToReg = CurrInstr;
-  LoadToLaneInstrs.push_back(
-      MRI.getUniqueVRegDef(SubregToReg->getOperand(2).getReg()));
-  auto LoadToLaneInstrsAscending = llvm::reverse(LoadToLaneInstrs);
-
-  const TargetRegisterClass *FPR128RegClass =
-      MRI.getRegClass(Root.getOperand(0).getReg());
-
-  auto LoadLaneToRegister = [&](MachineInstr *OriginalInstr,
-                                Register SrcRegister, unsigned Lane,
-                                Register OffsetRegister) {
-    auto NewRegister = MRI.createVirtualRegister(FPR128RegClass);
-    MachineInstrBuilder LoadIndexIntoRegister =
-        BuildMI(MF, MIMetadata(*OriginalInstr), TII->get(Root.getOpcode()),
-                NewRegister)
-            .addReg(SrcRegister)
-            .addImm(Lane)
-            .addReg(OffsetRegister, getKillRegState(true));
-    InstrIdxForVirtReg.insert(std::make_pair(NewRegister, InsInstrs.size()));
-    InsInstrs.push_back(LoadIndexIntoRegister);
-    return NewRegister;
-  };
-
-  // Helper to create load instruction based on opcode
-  auto CreateLoadInstruction = [&](unsigned NumLanes, Register DestReg,
-                                   Register OffsetReg) -> MachineInstrBuilder {
-    unsigned Opcode;
-    switch (NumLanes) {
-    case 4:
-      Opcode = AArch64::LDRSui;
-      break;
-    case 8:
-      Opcode = AArch64::LDRHui;
-      break;
-    case 16:
-      Opcode = AArch64::LDRBui;
-      break;
-    default:
-      llvm_unreachable(
-          "Got unsupported number of lanes in machine-combiner gather pattern");
-    }
-    // Immediate offset load
-    return BuildMI(MF, MIMetadata(Root), TII->get(Opcode), DestReg)
-        .addReg(OffsetReg)
-        .addImm(0); // immediate offset
-  };
-
-  // Load the remaining lanes into register 0.
-  auto LanesToLoadToReg0 =
-      llvm::make_range(LoadToLaneInstrsAscending.begin() + 1,
-                       LoadToLaneInstrsAscending.begin() + NumLanes / 2);
-  auto PrevReg = SubregToReg->getOperand(0).getReg();
-  for (auto [Index, LoadInstr] : llvm::enumerate(LanesToLoadToReg0)) {
-    PrevReg = LoadLaneToRegister(LoadInstr, PrevReg, Index + 1,
-                                 LoadInstr->getOperand(3).getReg());
-    DelInstrs.push_back(LoadInstr);
-  }
-  auto LastLoadReg0 = PrevReg;
-
-  // First load into register 1. Perform a LDRSui to zero out the upper lanes in
-  // a single instruction.
-  auto Lane0Load = *LoadToLaneInstrsAscending.begin();
-  auto OriginalSplitLoad =
-      *std::next(LoadToLaneInstrsAscending.begin(), NumLanes / 2);
-  auto DestRegForMiddleIndex = MRI.createVirtualRegister(
-      MRI.getRegClass(Lane0Load->getOperand(0).getReg()));
-
-  MachineInstrBuilder MiddleIndexLoadInstr =
-      CreateLoadInstruction(NumLanes, DestRegForMiddleIndex,
-                            OriginalSplitLoad->getOperand(3).getReg());
-
-  InstrIdxForVirtReg.insert(
-      std::make_pair(DestRegForMiddleIndex, InsInstrs.size()));
-  InsInstrs.push_back(MiddleIndexLoadInstr);
-  DelInstrs.push_back(OriginalSplitLoad);
-
-  // Subreg To Reg instruction for register 1.
-  auto DestRegForSubregToReg = MRI.createVirtualRegister(FPR128RegClass);
-  unsigned SubregType;
-  switch (NumLanes) {
-  case 4:
-    SubregType = AArch64::ssub;
-    break;
-  case 8:
-    SubregType = AArch64::hsub;
-    break;
-  case 16:
-    SubregType = AArch64::bsub;
-    break;
-  default:
-    llvm_unreachable(
-        "Got invalid NumLanes for machine-combiner gather pattern");
-  }
-
-  auto SubRegToRegInstr =
-      BuildMI(MF, MIMetadata(Root), TII->get(SubregToReg->getOpcode()),
-              DestRegForSubregToReg)
-          .addImm(0)
-          .addReg(DestRegForMiddleIndex, getKillRegState(true))
-          .addImm(SubregType);
-  InstrIdxForVirtReg.insert(
-      std::make_pair(DestRegForSubregToReg, InsInstrs.size()));
-  InsInstrs.push_back(SubRegToRegInstr);
-
-  // Load remaining lanes into register 1.
-  auto LanesToLoadToReg1 =
-      llvm::make_range(LoadToLaneInstrsAscending.begin() + NumLanes / 2 + 1,
-                       LoadToLaneInstrsAscending.end());
-  PrevReg = SubRegToRegInstr->getOperand(0).getReg();
-  for (auto [Index, LoadInstr] : llvm::enumerate(LanesToLoadToReg1)) {
-    PrevReg = LoadLaneToRegister(LoadInstr, PrevReg, Index + 1,
-                                 LoadInstr->getOperand(3).getReg());
-    if (Index == NumLanes / 2 - 2) {
-      break;
-    }
-    DelInstrs.push_back(LoadInstr);
-  }
-  auto LastLoadReg1 = PrevReg;
-
-  // Create the final zip instruction to combine the results.
-  MachineInstrBuilder ZipInstr =
-      BuildMI(MF, MIMetadata(Root), TII->get(AArch64::ZIP1v2i64),
-              Root.getOperand(0).getReg())
-          .addReg(LastLoadReg0)
-          .addReg(LastLoadReg1);
-  InsInstrs.push_back(ZipInstr);
-}
-
 CombinerObjective
 AArch64InstrInfo::getCombinerObjective(unsigned Pattern) const {
   switch (Pattern) {
   case AArch64MachineCombinerPattern::SUBADD_OP1:
   case AArch64MachineCombinerPattern::SUBADD_OP2:
-  case AArch64MachineCombinerPattern::GATHER_LANE_i32:
-  case AArch64MachineCombinerPattern::GATHER_LANE_i16:
-  case AArch64MachineCombinerPattern::GATHER_LANE_i8:
     return CombinerObjective::MustReduceDepth;
   default:
     return TargetInstrInfo::getCombinerObjective(Pattern);
@@ -7672,10 +7426,6 @@ bool AArch64InstrInfo::getMachineCombinerPatterns(
   if (getMiscPatterns(Root, Patterns))
     return true;
 
-  // Load patterns
-  if (getLoadPatterns(Root, Patterns))
-    return true;
-
   return TargetInstrInfo::getMachineCombinerPatterns(Root, Patterns,
                                                      DoRegPressureReduce);
 }
@@ -8931,21 +8681,6 @@ void AArch64InstrInfo::genAlternativeCodeSequence(
     MUL = genFNegatedMAD(MF, MRI, TII, Root, InsInstrs);
     break;
   }
-  case AArch64MachineCombinerPattern::GATHER_LANE_i32: {
-    generateGatherPattern(Root, InsInstrs, DelInstrs, InstrIdxForVirtReg,
-                          Pattern, 4);
-    break;
-  }
-  case AArch64MachineCombinerPattern::GATHER_LANE_i16: {
-    generateGatherPattern(Root, InsInstrs, DelInstrs, InstrIdxForVirtReg,
-                          Pattern, 8);
-    break;
-  }
-  case AArch64MachineCombinerPattern::GATHER_LANE_i8: {
-    generateGatherPattern(Root, InsInstrs, DelInstrs, InstrIdxForVirtReg,
-                          Pattern, 16);
-    break;
-  }
 
   } // end switch (Pattern)
   // Record MUL and ADD/SUB for deletion
diff --git a/llvm/lib/Target/AArch64/AArch64InstrInfo.h b/llvm/lib/Target/AArch64/AArch64InstrInfo.h
index 02734866..7c255da 100644
--- a/llvm/lib/Target/AArch64/AArch64InstrInfo.h
+++ b/llvm/lib/Target/AArch64/AArch64InstrInfo.h
@@ -172,10 +172,6 @@ enum AArch64MachineCombinerPattern : unsigned {
   FMULv8i16_indexed_OP2,
 
   FNMADD,
-
-  GATHER_LANE_i32,
-  GATHER_LANE_i16,
-  GATHER_LANE_i8
 };
 class AArch64InstrInfo final : public AArch64GenInstrInfo {
   const AArch64RegisterInfo RI;
diff --git a/llvm/lib/Target/AArch64/AArch64TargetObjectFile.cpp b/llvm/lib/Target/AArch64/AArch64TargetObjectFile.cpp
index c218831..85de2d5 100644
--- a/llvm/lib/Target/AArch64/AArch64TargetObjectFile.cpp
+++ b/llvm/lib/Target/AArch64/AArch64TargetObjectFile.cpp
@@ -36,7 +36,7 @@ void AArch64_ELFTargetObjectFile::Initialize(MCContext &Ctx,
   // SHF_AARCH64_PURECODE flag set if the "+execute-only" target feature is
   // present.
   if (TM.getMCSubtargetInfo()->hasFeature(AArch64::FeatureExecuteOnly)) {
-    auto *Text = cast<MCSectionELF>(TextSection);
+    auto *Text = static_cast<MCSectionELF *>(TextSection);
     Text->setFlags(Text->getFlags() | ELF::SHF_AARCH64_PURECODE);
   }
 }
diff --git a/llvm/lib/Target/AArch64/MCTargetDesc/AArch64ELFStreamer.cpp b/llvm/lib/Target/AArch64/MCTargetDesc/AArch64ELFStreamer.cpp
index 08f547a..6257e99 100644
--- a/llvm/lib/Target/AArch64/MCTargetDesc/AArch64ELFStreamer.cpp
+++ b/llvm/lib/Target/AArch64/MCTargetDesc/AArch64ELFStreamer.cpp
@@ -523,7 +523,8 @@ void AArch64TargetELFStreamer::finish() {
   // mark it execute-only if it is empty and there is at least one
   // execute-only section in the object.
   if (any_of(Asm, [](const MCSection &Sec) {
-        return cast<MCSectionELF>(Sec).getFlags() & ELF::SHF_AARCH64_PURECODE;
+        return static_cast<const MCSectionELF &>(Sec).getFlags() &
+               ELF::SHF_AARCH64_PURECODE;
       })) {
     auto *Text =
         static_cast<MCSectionELF *>(Ctx.getObjectFileInfo()->getTextSection());
diff --git a/llvm/lib/Target/AArch64/MCTargetDesc/AArch64MachObjectWriter.cpp b/llvm/lib/Target/AArch64/MCTargetDesc/AArch64MachObjectWriter.cpp
index 1ac340a..a22a17a 100644
--- a/llvm/lib/Target/AArch64/MCTargetDesc/AArch64MachObjectWriter.cpp
+++ b/llvm/lib/Target/AArch64/MCTargetDesc/AArch64MachObjectWriter.cpp
@@ -132,7 +132,8 @@ static bool canUseLocalRelocation(const MCSectionMachO &Section,
   // But only if they don't point to a few forbidden sections.
   if (!Symbol.isInSection())
     return true;
-  const MCSectionMachO &RefSec = cast<MCSectionMachO>(Symbol.getSection());
+  const MCSectionMachO &RefSec =
+      static_cast<MCSectionMachO &>(Symbol.getSection());
   if (RefSec.getType() == MachO::S_CSTRING_LITERALS)
     return false;
 
diff --git a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
index e3ca09e..f25ce87 100644
--- a/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/AMDGPUISelLowering.cpp
@@ -391,8 +391,9 @@ AMDGPUTargetLowering::AMDGPUTargetLowering(const TargetMachine &TM,
   // Library functions.  These default to Expand, but we have instructions
   // for them.
   setOperationAction({ISD::FCEIL, ISD::FPOW, ISD::FABS, ISD::FFLOOR,
-                      ISD::FROUNDEVEN, ISD::FTRUNC, ISD::FMINNUM, ISD::FMAXNUM},
-                     MVT::f32, Legal);
+                      ISD::FROUNDEVEN, ISD::FTRUNC},
+                     {MVT::f16, MVT::f32}, Legal);
+  setOperationAction({ISD::FMINNUM, ISD::FMAXNUM}, MVT::f32, Legal);
 
   setOperationAction(ISD::FLOG2, MVT::f32, Custom);
   setOperationAction(ISD::FROUND, {MVT::f32, MVT::f64}, Custom);
@@ -412,9 +413,10 @@ AMDGPUTargetLowering::AMDGPUTargetLowering(const TargetMachine &TM,
 
   setOperationAction(ISD::FREM, {MVT::f16, MVT::f32, MVT::f64}, Custom);
 
-  if (Subtarget->has16BitInsts())
+  if (Subtarget->has16BitInsts()) {
     setOperationAction(ISD::IS_FPCLASS, {MVT::f16, MVT::f32, MVT::f64}, Legal);
-  else {
+    setOperationAction({ISD::FLOG2, ISD::FEXP2}, MVT::f16, Legal);
+  } else {
     setOperationAction(ISD::IS_FPCLASS, {MVT::f32, MVT::f64}, Legal);
     setOperationAction({ISD::FLOG2, ISD::FEXP2}, MVT::f16, Custom);
   }
diff --git a/llvm/lib/Target/AMDGPU/GCNRegPressure.cpp b/llvm/lib/Target/AMDGPU/GCNRegPressure.cpp
index 7d6723a..334afd3 100644
--- a/llvm/lib/Target/AMDGPU/GCNRegPressure.cpp
+++ b/llvm/lib/Target/AMDGPU/GCNRegPressure.cpp
@@ -38,7 +38,11 @@ bool llvm::isEqual(const GCNRPTracker::LiveRegSet &S1,
 
 unsigned GCNRegPressure::getRegKind(const TargetRegisterClass *RC,
                                     const SIRegisterInfo *STI) {
-  return STI->isSGPRClass(RC) ? SGPR : (STI->isAGPRClass(RC) ? AGPR : VGPR);
+  return STI->isSGPRClass(RC)
+             ? SGPR
+             : (STI->isAGPRClass(RC)
+                    ? AGPR
+                    : (STI->isVectorSuperClass(RC) ? AVGPR : VGPR));
 }
 
 void GCNRegPressure::inc(unsigned Reg,
diff --git a/llvm/lib/Target/AMDGPU/GCNRegPressure.h b/llvm/lib/Target/AMDGPU/GCNRegPressure.h
index 3749b6d..ea33a22 100644
--- a/llvm/lib/Target/AMDGPU/GCNRegPressure.h
+++ b/llvm/lib/Target/AMDGPU/GCNRegPressure.h
@@ -29,43 +29,57 @@ class raw_ostream;
 class SlotIndex;
 
 struct GCNRegPressure {
-  enum RegKind { SGPR, VGPR, AGPR, TOTAL_KINDS };
+  enum RegKind { SGPR, VGPR, AGPR, AVGPR, TOTAL_KINDS };
 
   GCNRegPressure() {
     clear();
   }
 
-  bool empty() const { return !Value[SGPR] && !Value[VGPR] && !Value[AGPR]; }
+  bool empty() const {
+    return !Value[SGPR] && !Value[VGPR] && !Value[AGPR] && !Value[AVGPR];
+  }
 
   void clear() { std::fill(&Value[0], &Value[ValueArraySize], 0); }
 
   /// \returns the SGPR32 pressure
   unsigned getSGPRNum() const { return Value[SGPR]; }
-  /// \returns the aggregated ArchVGPR32, AccVGPR32 pressure dependent upon \p
-  /// UnifiedVGPRFile
+  /// \returns the aggregated ArchVGPR32, AccVGPR32, and Pseudo AVGPR pressure
+  /// dependent upon \p UnifiedVGPRFile
   unsigned getVGPRNum(bool UnifiedVGPRFile) const {
     if (UnifiedVGPRFile) {
-      return Value[AGPR] ? getUnifiedVGPRNum(Value[VGPR], Value[AGPR])
-                         : Value[VGPR];
+      return Value[AGPR]
+                 ? getUnifiedVGPRNum(Value[VGPR], Value[AGPR], Value[AVGPR])
+                 : Value[VGPR] + Value[AVGPR];
     }
-    return std::max(Value[VGPR], Value[AGPR]);
+    // AVGPR assignment priority is based on the width of the register. Account
+    // AVGPR pressure as VGPR.
+    return std::max(Value[VGPR] + Value[AVGPR], Value[AGPR]);
   }
 
   /// Returns the aggregated VGPR pressure, assuming \p NumArchVGPRs ArchVGPRs
-  /// and \p NumAGPRs AGPRS, for a target with a unified VGPR file.
+  /// \p NumAGPRs AGPRS, and \p NumAVGPRs AVGPRs for a target with a unified
+  /// VGPR file.
   inline static unsigned getUnifiedVGPRNum(unsigned NumArchVGPRs,
-                                           unsigned NumAGPRs) {
-    return alignTo(NumArchVGPRs, AMDGPU::IsaInfo::getArchVGPRAllocGranule()) +
+                                           unsigned NumAGPRs,
+                                           unsigned NumAVGPRs) {
+
+    // Assume AVGPRs will be assigned as VGPRs.
+    return alignTo(NumArchVGPRs + NumAVGPRs,
+                   AMDGPU::IsaInfo::getArchVGPRAllocGranule()) +
            NumAGPRs;
   }
 
-  /// \returns the ArchVGPR32 pressure
-  unsigned getArchVGPRNum() const { return Value[VGPR]; }
+  /// \returns the ArchVGPR32 pressure, plus the AVGPRS which we assume will be
+  /// allocated as VGPR
+  unsigned getArchVGPRNum() const { return Value[VGPR] + Value[AVGPR]; }
   /// \returns the AccVGPR32 pressure
   unsigned getAGPRNum() const { return Value[AGPR]; }
+  /// \returns the AVGPR32 pressure
+  unsigned getAVGPRNum() const { return Value[AVGPR]; }
 
   unsigned getVGPRTuplesWeight() const {
-    return std::max(Value[TOTAL_KINDS + VGPR], Value[TOTAL_KINDS + AGPR]);
+    return std::max(Value[TOTAL_KINDS + VGPR] + Value[TOTAL_KINDS + AVGPR],
+                    Value[TOTAL_KINDS + AGPR]);
   }
   unsigned getSGPRTuplesWeight() const { return Value[TOTAL_KINDS + SGPR]; }
 
diff --git a/llvm/lib/Target/AMDGPU/GCNSchedStrategy.cpp b/llvm/lib/Target/AMDGPU/GCNSchedStrategy.cpp
index 3c3247a..ce1ce68 100644
--- a/llvm/lib/Target/AMDGPU/GCNSchedStrategy.cpp
+++ b/llvm/lib/Target/AMDGPU/GCNSchedStrategy.cpp
@@ -1914,8 +1914,8 @@ void PreRARematStage::rematerialize() {
     unsigned DefRegion = MIRegion.at(DefMI);
 
     // Rematerialize DefMI to its use block.
-    TII->reMaterialize(*InsertPos->getParent(), InsertPos, Reg, 0, *DefMI,
-                       *DAG.TRI);
+    TII->reMaterialize(*InsertPos->getParent(), InsertPos, Reg,
+                       AMDGPU::NoSubRegister, *DefMI, *DAG.TRI);
     Remat.RematMI = &*std::prev(InsertPos);
     DAG.LIS->InsertMachineInstrInMaps(*Remat.RematMI);
 
@@ -2066,7 +2066,8 @@ void PreRARematStage::finalizeGCNSchedStage() {
     // Re-rematerialize MI at the end of its original region. Note that it may
     // not be rematerialized exactly in the same position as originally within
     // the region, but it should not matter much.
-    TII->reMaterialize(*MBB, InsertPos, Reg, 0, RematMI, *DAG.TRI);
+    TII->reMaterialize(*MBB, InsertPos, Reg, AMDGPU::NoSubRegister, RematMI,
+                       *DAG.TRI);
     MachineInstr *NewMI = &*std::prev(InsertPos);
     DAG.LIS->InsertMachineInstrInMaps(*NewMI);
 
diff --git a/llvm/lib/Target/AMDGPU/GCNSubtarget.cpp b/llvm/lib/Target/AMDGPU/GCNSubtarget.cpp
index 9a2bab1..0a0a107 100644
--- a/llvm/lib/Target/AMDGPU/GCNSubtarget.cpp
+++ b/llvm/lib/Target/AMDGPU/GCNSubtarget.cpp
@@ -537,6 +537,63 @@ unsigned GCNSubtarget::getMaxNumVGPRs(const MachineFunction &MF) const {
   return getMaxNumVGPRs(MF.getFunction());
 }
 
+std::pair<unsigned, unsigned>
+GCNSubtarget::getMaxNumVectorRegs(const Function &F) const {
+  const unsigned MaxVectorRegs = getMaxNumVGPRs(F);
+
+  unsigned MaxNumVGPRs = MaxVectorRegs;
+  unsigned MaxNumAGPRs = 0;
+
+  // On GFX90A, the number of VGPRs and AGPRs need not be equal. Theoretically,
+  // a wave may have up to 512 total vector registers combining together both
+  // VGPRs and AGPRs. Hence, in an entry function without calls and without
+  // AGPRs used within it, it is possible to use the whole vector register
+  // budget for VGPRs.
+  //
+  // TODO: it shall be possible to estimate maximum AGPR/VGPR pressure and split
+  //       register file accordingly.
+  if (hasGFX90AInsts()) {
+    unsigned MinNumAGPRs = 0;
+    const unsigned TotalNumAGPRs = AMDGPU::AGPR_32RegClass.getNumRegs();
+    const unsigned TotalNumVGPRs = AMDGPU::VGPR_32RegClass.getNumRegs();
+
+    const std::pair<unsigned, unsigned> DefaultNumAGPR = {~0u, ~0u};
+
+    // TODO: The lower bound should probably force the number of required
+    // registers up, overriding amdgpu-waves-per-eu.
+    std::tie(MinNumAGPRs, MaxNumAGPRs) =
+        AMDGPU::getIntegerPairAttribute(F, "amdgpu-agpr-alloc", DefaultNumAGPR,
+                                        /*OnlyFirstRequired=*/true);
+
+    if (MinNumAGPRs == DefaultNumAGPR.first) {
+      // Default to splitting half the registers if AGPRs are required.
+      MinNumAGPRs = MaxNumAGPRs = MaxVectorRegs / 2;
+    } else {
+      // Align to accum_offset's allocation granularity.
+      MinNumAGPRs = alignTo(MinNumAGPRs, 4);
+
+      MinNumAGPRs = std::min(MinNumAGPRs, TotalNumAGPRs);
+    }
+
+    // Clamp values to be inbounds of our limits, and ensure min <= max.
+
+    MaxNumAGPRs = std::min(std::max(MinNumAGPRs, MaxNumAGPRs), MaxVectorRegs);
+    MinNumAGPRs = std::min(std::min(MinNumAGPRs, TotalNumAGPRs), MaxNumAGPRs);
+
+    MaxNumVGPRs = std::min(MaxVectorRegs - MinNumAGPRs, TotalNumVGPRs);
+    MaxNumAGPRs = std::min(MaxVectorRegs - MaxNumVGPRs, MaxNumAGPRs);
+
+    assert(MaxNumVGPRs + MaxNumAGPRs <= MaxVectorRegs &&
+           MaxNumAGPRs <= TotalNumAGPRs && MaxNumVGPRs <= TotalNumVGPRs &&
+           "invalid register counts");
+  } else if (hasMAIInsts()) {
+    // On gfx908 the number of AGPRs always equals the number of VGPRs.
+    MaxNumAGPRs = MaxNumVGPRs = MaxVectorRegs;
+  }
+
+  return std::pair(MaxNumVGPRs, MaxNumAGPRs);
+}
+
 void GCNSubtarget::adjustSchedDependency(
     SUnit *Def, int DefOpIdx, SUnit *Use, int UseOpIdx, SDep &Dep,
     const TargetSchedModel *SchedModel) const {
diff --git a/llvm/lib/Target/AMDGPU/GCNSubtarget.h b/llvm/lib/Target/AMDGPU/GCNSubtarget.h
index 88a269f..785ede3 100644
--- a/llvm/lib/Target/AMDGPU/GCNSubtarget.h
+++ b/llvm/lib/Target/AMDGPU/GCNSubtarget.h
@@ -1667,6 +1667,10 @@ public:
     return getMaxNumVGPRs(F);
   }
 
+  /// Return a pair of maximum numbers of VGPRs and AGPRs that meet the number
+  /// of waves per execution unit required for the function \p MF.
+  std::pair<unsigned, unsigned> getMaxNumVectorRegs(const Function &F) const;
+
   /// \returns Maximum number of VGPRs that meets number of waves per execution
   /// unit requirement for function \p MF, or number of VGPRs explicitly
   /// requested using "amdgpu-num-vgpr" attribute attached to function \p MF.
diff --git a/llvm/lib/Target/AMDGPU/SIFoldOperands.cpp b/llvm/lib/Target/AMDGPU/SIFoldOperands.cpp
index e5d1eaa..b77da4d 100644
--- a/llvm/lib/Target/AMDGPU/SIFoldOperands.cpp
+++ b/llvm/lib/Target/AMDGPU/SIFoldOperands.cpp
@@ -1062,9 +1062,13 @@ bool SIFoldOperandsImpl::tryFoldRegSeqSplat(
     switch (OpTy) {
     case AMDGPU::OPERAND_REG_INLINE_AC_INT32:
     case AMDGPU::OPERAND_REG_INLINE_AC_FP32:
+    case AMDGPU::OPERAND_REG_INLINE_C_INT32:
+    case AMDGPU::OPERAND_REG_INLINE_C_FP32:
       OpRC = TRI->getSubRegisterClass(OpRC, AMDGPU::sub0);
       break;
     case AMDGPU::OPERAND_REG_INLINE_AC_FP64:
+    case AMDGPU::OPERAND_REG_INLINE_C_FP64:
+    case AMDGPU::OPERAND_REG_INLINE_C_INT64:
       OpRC = TRI->getSubRegisterClass(OpRC, AMDGPU::sub0_sub1);
       break;
     default:
diff --git a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
index 0eee7ad..8d51ec6 100644
--- a/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
+++ b/llvm/lib/Target/AMDGPU/SIISelLowering.cpp
@@ -14179,6 +14179,8 @@ SDValue SITargetLowering::performMinMaxCombine(SDNode *N,
         Op0.getOpcode() == AMDGPUISD::FMAX_LEGACY)) &&
       (VT == MVT::f32 || VT == MVT::f64 ||
        (VT == MVT::f16 && Subtarget->has16BitInsts()) ||
+       (VT == MVT::bf16 && Subtarget->hasBF16PackedInsts()) ||
+       (VT == MVT::v2bf16 && Subtarget->hasBF16PackedInsts()) ||
        (VT == MVT::v2f16 && Subtarget->hasVOP3PInsts())) &&
       Op0.hasOneUse()) {
     if (SDValue Res = performFPMed3ImmCombine(DAG, SDLoc(N), Op0, Op1))
diff --git a/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp b/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
index dd3f2fe..520c321 100644
--- a/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInsertWaitcnts.cpp
@@ -552,7 +552,7 @@ public:
         (!Inst.mayLoad() || SIInstrInfo::isAtomicNoRet(Inst))) {
       // FLAT and SCRATCH instructions may access scratch. Other VMEM
       // instructions do not.
-      if (SIInstrInfo::isFLAT(Inst) && mayAccessScratchThroughFlat(Inst))
+      if (TII->mayAccessScratchThroughFlat(Inst))
         return SCRATCH_WRITE_ACCESS;
       return VMEM_WRITE_ACCESS;
     }
@@ -565,7 +565,6 @@ public:
 
   bool mayAccessVMEMThroughFlat(const MachineInstr &MI) const;
   bool mayAccessLDSThroughFlat(const MachineInstr &MI) const;
-  bool mayAccessScratchThroughFlat(const MachineInstr &MI) const;
   bool isVmemAccess(const MachineInstr &MI) const;
   bool generateWaitcntInstBefore(MachineInstr &MI,
                                  WaitcntBrackets &ScoreBrackets,
@@ -2160,32 +2159,6 @@ bool SIInsertWaitcnts::mayAccessLDSThroughFlat(const MachineInstr &MI) const {
   return false;
 }
 
-// This is a flat memory operation. Check to see if it has memory tokens for
-// either scratch or FLAT.
-bool SIInsertWaitcnts::mayAccessScratchThroughFlat(
-    const MachineInstr &MI) const {
-  assert(TII->isFLAT(MI));
-
-  // SCRATCH instructions always access scratch.
-  if (TII->isFLATScratch(MI))
-    return true;
-
-  // GLOBAL instructions never access scratch.
-  if (TII->isFLATGlobal(MI))
-    return false;
-
-  // If there are no memory operands then conservatively assume the flat
-  // operation may access scratch.
-  if (MI.memoperands_empty())
-    return true;
-
-  // See if any memory operand specifies an address space that involves scratch.
-  return any_of(MI.memoperands(), [](const MachineMemOperand *Memop) {
-    unsigned AS = Memop->getAddrSpace();
-    return AS == AMDGPUAS::PRIVATE_ADDRESS || AS == AMDGPUAS::FLAT_ADDRESS;
-  });
-}
-
 bool SIInsertWaitcnts::isVmemAccess(const MachineInstr &MI) const {
   return (TII->isFLAT(MI) && mayAccessVMEMThroughFlat(MI)) ||
          (TII->isVMEM(MI) && !AMDGPU::getMUBUFIsBufferInv(MI.getOpcode()));
diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
index 40e6871..2aa6b4e 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.cpp
@@ -2508,7 +2508,20 @@ bool SIInstrInfo::expandPostRAPseudo(MachineInstr &MI) const {
           .addReg(DstHi);
     }
     break;
+
+  case AMDGPU::V_MAX_BF16_PSEUDO_e64:
+    assert(ST.hasBF16PackedInsts());
+    MI.setDesc(get(AMDGPU::V_PK_MAX_NUM_BF16));
+    MI.addOperand(MachineOperand::CreateImm(0)); // op_sel
+    MI.addOperand(MachineOperand::CreateImm(0)); // neg_lo
+    MI.addOperand(MachineOperand::CreateImm(0)); // neg_hi
+    auto Op0 = getNamedOperand(MI, AMDGPU::OpName::src0_modifiers);
+    Op0->setImm(Op0->getImm() | SISrcMods::OP_SEL_1);
+    auto Op1 = getNamedOperand(MI, AMDGPU::OpName::src1_modifiers);
+    Op1->setImm(Op1->getImm() | SISrcMods::OP_SEL_1);
+    break;
   }
+
   return true;
 }
 
@@ -2733,49 +2746,47 @@ static MachineInstr *swapImmOperands(MachineInstr &MI,
 }
 
 bool SIInstrInfo::isLegalToSwap(const MachineInstr &MI, unsigned OpIdx0,
-                                const MachineOperand *MO0, unsigned OpIdx1,
-                                const MachineOperand *MO1) const {
+                                unsigned OpIdx1) const {
   const MCInstrDesc &InstDesc = MI.getDesc();
   const MCOperandInfo &OpInfo0 = InstDesc.operands()[OpIdx0];
   const MCOperandInfo &OpInfo1 = InstDesc.operands()[OpIdx1];
-  const TargetRegisterClass *DefinedRC1 =
-      OpInfo1.RegClass != -1 ? RI.getRegClass(OpInfo1.RegClass) : nullptr;
-  const TargetRegisterClass *DefinedRC0 =
-      OpInfo1.RegClass != -1 ? RI.getRegClass(OpInfo0.RegClass) : nullptr;
 
   unsigned Opc = MI.getOpcode();
   int Src0Idx = AMDGPU::getNamedOperandIdx(Opc, AMDGPU::OpName::src0);
 
+  const MachineOperand &MO0 = MI.getOperand(OpIdx0);
+  const MachineOperand &MO1 = MI.getOperand(OpIdx1);
+
   // Swap doesn't breach constant bus or literal limits
   // It may move literal to position other than src0, this is not allowed
   // pre-gfx10 However, most test cases need literals in Src0 for VOP
   // FIXME: After gfx9, literal can be in place other than Src0
   if (isVALU(MI)) {
-    if ((int)OpIdx0 == Src0Idx && !MO0->isReg() &&
-        !isInlineConstant(*MO0, OpInfo1))
+    if ((int)OpIdx0 == Src0Idx && !MO0.isReg() &&
+        !isInlineConstant(MO0, OpInfo1))
       return false;
-    if ((int)OpIdx1 == Src0Idx && !MO1->isReg() &&
-        !isInlineConstant(*MO1, OpInfo0))
+    if ((int)OpIdx1 == Src0Idx && !MO1.isReg() &&
+        !isInlineConstant(MO1, OpInfo0))
       return false;
   }
 
-  if ((int)OpIdx1 != Src0Idx && MO0->isReg()) {
-    if (!DefinedRC1)
+  if ((int)OpIdx1 != Src0Idx && MO0.isReg()) {
+    if (OpInfo1.RegClass == -1)
       return OpInfo1.OperandType == MCOI::OPERAND_UNKNOWN;
-    return isLegalRegOperand(MI, OpIdx1, *MO0) &&
-           (!MO1->isReg() || isLegalRegOperand(MI, OpIdx0, *MO1));
+    return isLegalRegOperand(MI, OpIdx1, MO0) &&
+           (!MO1.isReg() || isLegalRegOperand(MI, OpIdx0, MO1));
   }
-  if ((int)OpIdx0 != Src0Idx && MO1->isReg()) {
-    if (!DefinedRC0)
+  if ((int)OpIdx0 != Src0Idx && MO1.isReg()) {
+    if (OpInfo0.RegClass == -1)
       return OpInfo0.OperandType == MCOI::OPERAND_UNKNOWN;
-    return (!MO0->isReg() || isLegalRegOperand(MI, OpIdx1, *MO0)) &&
-           isLegalRegOperand(MI, OpIdx0, *MO1);
+    return (!MO0.isReg() || isLegalRegOperand(MI, OpIdx1, MO0)) &&
+           isLegalRegOperand(MI, OpIdx0, MO1);
   }
 
   // No need to check 64-bit literals since swapping does not bring new
   // 64-bit literals into current instruction to fold to 32-bit
 
-  return isImmOperandLegal(MI, OpIdx1, *MO0);
+  return isImmOperandLegal(MI, OpIdx1, MO0);
 }
 
 MachineInstr *SIInstrInfo::commuteInstructionImpl(MachineInstr &MI, bool NewMI,
@@ -2797,12 +2808,12 @@ MachineInstr *SIInstrInfo::commuteInstructionImpl(MachineInstr &MI, bool NewMI,
            static_cast<int>(Src1Idx) &&
          "inconsistency with findCommutedOpIndices");
 
-  MachineOperand &Src0 = MI.getOperand(Src0Idx);
-  MachineOperand &Src1 = MI.getOperand(Src1Idx);
-  if (!isLegalToSwap(MI, Src0Idx, &Src0, Src1Idx, &Src1)) {
+  if (!isLegalToSwap(MI, Src0Idx, Src1Idx))
     return nullptr;
-  }
+
   MachineInstr *CommutedMI = nullptr;
+  MachineOperand &Src0 = MI.getOperand(Src0Idx);
+  MachineOperand &Src1 = MI.getOperand(Src1Idx);
   if (Src0.isReg() && Src1.isReg()) {
     // Be sure to copy the source modifiers to the right place.
     CommutedMI =
@@ -4238,6 +4249,32 @@ bool SIInstrInfo::isAlwaysGDS(uint16_t Opcode) const {
          Opcode == AMDGPU::DS_SUB_GS_REG_RTN || isGWS(Opcode);
 }
 
+bool SIInstrInfo::mayAccessScratchThroughFlat(const MachineInstr &MI) const {
+  if (!isFLAT(MI) || isFLATGlobal(MI))
+    return false;
+
+  // If scratch is not initialized, we can never access it.
+  if (MI.getMF()->getFunction().hasFnAttribute("amdgpu-no-flat-scratch-init"))
+    return false;
+
+  // SCRATCH instructions always access scratch.
+  if (isFLATScratch(MI))
+    return true;
+
+  // If there are no memory operands then conservatively assume the flat
+  // operation may access scratch.
+  if (MI.memoperands_empty())
+    return true;
+
+  // TODO (?): Does this need to be taught how to read noalias.addrspace ?
+
+  // See if any memory operand specifies an address space that involves scratch.
+  return any_of(MI.memoperands(), [](const MachineMemOperand *Memop) {
+    unsigned AS = Memop->getAddrSpace();
+    return AS == AMDGPUAS::PRIVATE_ADDRESS || AS == AMDGPUAS::FLAT_ADDRESS;
+  });
+}
+
 bool SIInstrInfo::modifiesModeRegister(const MachineInstr &MI) {
   // Skip the full operand and register alias search modifiesRegister
   // does. There's only a handful of instructions that touch this, it's only an
diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.h b/llvm/lib/Target/AMDGPU/SIInstrInfo.h
index 800ea9a..e042b59 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.h
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.h
@@ -197,8 +197,7 @@ protected:
                            AMDGPU::OpName Src0OpName, MachineOperand &Src1,
                            AMDGPU::OpName Src1OpName) const;
   bool isLegalToSwap(const MachineInstr &MI, unsigned fromIdx,
-                     const MachineOperand *fromMO, unsigned toIdx,
-                     const MachineOperand *toMO) const;
+                     unsigned toIdx) const;
   MachineInstr *commuteInstructionImpl(MachineInstr &MI, bool NewMI,
                                        unsigned OpIdx0,
                                        unsigned OpIdx1) const override;
@@ -679,6 +678,12 @@ public:
     return get(Opcode).TSFlags & SIInstrFlags::FLAT;
   }
 
+  /// \returns true for SCRATCH_ instructions, or FLAT_ instructions with
+  /// SCRATCH_ memory operands.
+  /// Conservatively correct; will return true if \p MI cannot be proven
+  /// to not hit scratch.
+  bool mayAccessScratchThroughFlat(const MachineInstr &MI) const;
+
   static bool isBlockLoadStore(uint16_t Opcode) {
     switch (Opcode) {
     case AMDGPU::SI_BLOCK_SPILL_V1024_SAVE:
diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.td b/llvm/lib/Target/AMDGPU/SIInstrInfo.td
index b0be3f86..83b0490 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.td
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.td
@@ -2865,6 +2865,7 @@ def VOP_F16_F16_I16 : VOPProfile <[f16, f16, i16, untyped]>;
 def VOP_F16_F16_I32 : VOPProfile <[f16, f16, i32, untyped]>;
 def VOP_I16_I16_I16 : VOPProfile <[i16, i16, i16, untyped]>;
 def VOP_I16_I16_I16_ARITH : VOPProfile <[i16, i16, i16, untyped], /*EnableClamp=*/1>;
+def VOP_BF16_BF16_BF16 : VOPProfile <[bf16, bf16, bf16, untyped]>;
 
 def VOP_I16_I16_I16_I16 : VOPProfile <[i16, i16, i16, i16, untyped]>;
 def VOP_F16_F16_F16_F16 : VOPProfile <[f16, f16, f16, f16, untyped]>;
diff --git a/llvm/lib/Target/AMDGPU/SIInstructions.td b/llvm/lib/Target/AMDGPU/SIInstructions.td
index d05be8f..54fa192 100644
--- a/llvm/lib/Target/AMDGPU/SIInstructions.td
+++ b/llvm/lib/Target/AMDGPU/SIInstructions.td
@@ -1894,6 +1894,9 @@ let SubtargetPredicate = UseRealTrue16Insts in
 def : ClampPat<V_MAX_F16_t16_e64, f16>;
 let SubtargetPredicate = UseFakeTrue16Insts in
 def : ClampPat<V_MAX_F16_fake16_e64, f16>;
+// FIXME-TRUE16: Pseudo expansion of this won't work with True16.
+let True16Predicate = UseFakeTrue16Insts in
+def : ClampPat<V_MAX_BF16_PSEUDO_e64, bf16>;
 
 let SubtargetPredicate = HasVOP3PInsts in {
 def : GCNPat <
@@ -1903,6 +1906,13 @@ def : GCNPat <
 >;
 }
 
+let SubtargetPredicate = HasBF16PackedInsts in {
+def : GCNPat <
+  (v2bf16 (AMDGPUclamp (VOP3PMods v2bf16:$src0, i32:$src0_modifiers))),
+  (V_PK_MAX_NUM_BF16 $src0_modifiers, $src0,
+                     $src0_modifiers, $src0, DSTCLAMP.ENABLE)
+>;
+} // End SubtargetPredicate = HasBF16PackedInsts
 
 /********** ================================ **********/
 /********** Floating point absolute/negative **********/
diff --git a/llvm/lib/Target/AMDGPU/SILowerSGPRSpills.cpp b/llvm/lib/Target/AMDGPU/SILowerSGPRSpills.cpp
index 9f61bf8..9509199 100644
--- a/llvm/lib/Target/AMDGPU/SILowerSGPRSpills.cpp
+++ b/llvm/lib/Target/AMDGPU/SILowerSGPRSpills.cpp
@@ -351,6 +351,7 @@ void SILowerSGPRSpills::determineRegsForWWMAllocation(MachineFunction &MF,
   MachineRegisterInfo &MRI = MF.getRegInfo();
   BitVector ReservedRegs = TRI->getReservedRegs(MF);
   BitVector NonWwmAllocMask(TRI->getNumRegs());
+  const GCNSubtarget &ST = MF.getSubtarget<GCNSubtarget>();
 
   // FIXME: MaxNumVGPRsForWwmAllocation might need to be adjusted in the future
   // to have a balanced allocation between WWM values and per-thread vector
@@ -359,7 +360,7 @@ void SILowerSGPRSpills::determineRegsForWWMAllocation(MachineFunction &MF,
   NumRegs =
       std::min(static_cast<unsigned>(MFI->getSGPRSpillVGPRs().size()), NumRegs);
 
-  auto [MaxNumVGPRs, MaxNumAGPRs] = TRI->getMaxNumVectorRegs(MF);
+  auto [MaxNumVGPRs, MaxNumAGPRs] = ST.getMaxNumVectorRegs(MF.getFunction());
   // Try to use the highest available registers for now. Later after
   // vgpr-regalloc, they can be shifted to the lowest range.
   unsigned I = 0;
@@ -376,7 +377,7 @@ void SILowerSGPRSpills::determineRegsForWWMAllocation(MachineFunction &MF,
     // Reserve an arbitrary register and report the error.
     TRI->markSuperRegs(RegMask, AMDGPU::VGPR0);
     MF.getFunction().getContext().emitError(
-        "can't find enough VGPRs for wwm-regalloc");
+        "cannot find enough VGPRs for wwm-regalloc");
   }
 }
 
diff --git a/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp b/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp
index 0e8a420..607825e 100644
--- a/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp
+++ b/llvm/lib/Target/AMDGPU/SIMemoryLegalizer.cpp
@@ -321,7 +321,8 @@ public:
                                               bool IsNonTemporal,
                                               bool IsLastUse = false) const = 0;
 
-  virtual bool expandSystemScopeStore(MachineBasicBlock::iterator &MI) const {
+  virtual bool finalizeStore(MachineBasicBlock::iterator &MI,
+                             bool Atomic) const {
     return false;
   };
 
@@ -602,7 +603,8 @@ public:
                                       bool IsVolatile, bool IsNonTemporal,
                                       bool IsLastUse) const override;
 
-  bool expandSystemScopeStore(MachineBasicBlock::iterator &MI) const override;
+  bool finalizeStore(MachineBasicBlock::iterator &MI,
+                     bool Atomic) const override;
 
   bool insertRelease(MachineBasicBlock::iterator &MI, SIAtomicScope Scope,
                      SIAtomicAddrSpace AddrSpace, bool IsCrossAddrSpaceOrdering,
@@ -2551,11 +2553,25 @@ bool SIGfx12CacheControl::enableVolatileAndOrNonTemporal(
   return Changed;
 }
 
-bool SIGfx12CacheControl::expandSystemScopeStore(
-    MachineBasicBlock::iterator &MI) const {
+bool SIGfx12CacheControl::finalizeStore(MachineBasicBlock::iterator &MI,
+                                        bool Atomic) const {
   MachineOperand *CPol = TII->getNamedOperand(*MI, OpName::cpol);
-  if (CPol && ((CPol->getImm() & CPol::SCOPE) == CPol::SCOPE_SYS))
-    return insertWaitsBeforeSystemScopeStore(MI);
+  if (!CPol)
+    return false;
+
+  const unsigned Scope = CPol->getImm() & CPol::SCOPE;
+
+  // GFX12.0 only: Extra waits needed before system scope stores.
+  if (!ST.hasGFX1250Insts()) {
+    if (!Atomic && Scope == CPol::SCOPE_SYS)
+      return insertWaitsBeforeSystemScopeStore(MI);
+    return false;
+  }
+
+  // GFX12.5 only: Require SCOPE_SE on stores that may hit the scratch address
+  // space.
+  if (TII->mayAccessScratchThroughFlat(*MI) && Scope == CPol::SCOPE_CU)
+    return setScope(MI, CPol::SCOPE_SE);
 
   return false;
 }
@@ -2674,6 +2690,7 @@ bool SIMemoryLegalizer::expandStore(const SIMemOpInfo &MOI,
                                    MOI.getIsCrossAddressSpaceOrdering(),
                                    Position::BEFORE);
 
+    Changed |= CC->finalizeStore(MI, /*Atomic=*/true);
     return Changed;
   }
 
@@ -2686,7 +2703,7 @@ bool SIMemoryLegalizer::expandStore(const SIMemOpInfo &MOI,
 
   // GFX12 specific, scope(desired coherence domain in cache hierarchy) is
   // instruction field, do not confuse it with atomic scope.
-  Changed |= CC->expandSystemScopeStore(MI);
+  Changed |= CC->finalizeStore(MI, /*Atomic=*/false);
   return Changed;
 }
 
diff --git a/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp b/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
index 84cfa87..f3acc5c 100644
--- a/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
+++ b/llvm/lib/Target/AMDGPU/SIRegisterInfo.cpp
@@ -572,65 +572,6 @@ MCRegister SIRegisterInfo::reservedPrivateSegmentBufferReg(
   return getAlignedHighSGPRForRC(MF, /*Align=*/4, &AMDGPU::SGPR_128RegClass);
 }
 
-std::pair<unsigned, unsigned>
-SIRegisterInfo::getMaxNumVectorRegs(const MachineFunction &MF) const {
-  const unsigned MaxVectorRegs = ST.getMaxNumVGPRs(MF);
-
-  unsigned MaxNumVGPRs = MaxVectorRegs;
-  unsigned MaxNumAGPRs = 0;
-
-  // On GFX90A, the number of VGPRs and AGPRs need not be equal. Theoretically,
-  // a wave may have up to 512 total vector registers combining together both
-  // VGPRs and AGPRs. Hence, in an entry function without calls and without
-  // AGPRs used within it, it is possible to use the whole vector register
-  // budget for VGPRs.
-  //
-  // TODO: it shall be possible to estimate maximum AGPR/VGPR pressure and split
-  //       register file accordingly.
-  if (ST.hasGFX90AInsts()) {
-    unsigned MinNumAGPRs = 0;
-    const unsigned TotalNumAGPRs = AMDGPU::AGPR_32RegClass.getNumRegs();
-    const unsigned TotalNumVGPRs = AMDGPU::VGPR_32RegClass.getNumRegs();
-
-    const std::pair<unsigned, unsigned> DefaultNumAGPR = {~0u, ~0u};
-
-    // TODO: Move this logic into subtarget on IR function
-    //
-    // TODO: The lower bound should probably force the number of required
-    // registers up, overriding amdgpu-waves-per-eu.
-    std::tie(MinNumAGPRs, MaxNumAGPRs) = AMDGPU::getIntegerPairAttribute(
-        MF.getFunction(), "amdgpu-agpr-alloc", DefaultNumAGPR,
-        /*OnlyFirstRequired=*/true);
-
-    if (MinNumAGPRs == DefaultNumAGPR.first) {
-      // Default to splitting half the registers if AGPRs are required.
-      MinNumAGPRs = MaxNumAGPRs = MaxVectorRegs / 2;
-    } else {
-      // Align to accum_offset's allocation granularity.
-      MinNumAGPRs = alignTo(MinNumAGPRs, 4);
-
-      MinNumAGPRs = std::min(MinNumAGPRs, TotalNumAGPRs);
-    }
-
-    // Clamp values to be inbounds of our limits, and ensure min <= max.
-
-    MaxNumAGPRs = std::min(std::max(MinNumAGPRs, MaxNumAGPRs), MaxVectorRegs);
-    MinNumAGPRs = std::min(std::min(MinNumAGPRs, TotalNumAGPRs), MaxNumAGPRs);
-
-    MaxNumVGPRs = std::min(MaxVectorRegs - MinNumAGPRs, TotalNumVGPRs);
-    MaxNumAGPRs = std::min(MaxVectorRegs - MaxNumVGPRs, MaxNumAGPRs);
-
-    assert(MaxNumVGPRs + MaxNumAGPRs <= MaxVectorRegs &&
-           MaxNumAGPRs <= TotalNumAGPRs && MaxNumVGPRs <= TotalNumVGPRs &&
-           "invalid register counts");
-  } else if (ST.hasMAIInsts()) {
-    // On gfx908 the number of AGPRs always equals the number of VGPRs.
-    MaxNumAGPRs = MaxNumVGPRs = MaxVectorRegs;
-  }
-
-  return std::pair(MaxNumVGPRs, MaxNumAGPRs);
-}
-
 BitVector SIRegisterInfo::getReservedRegs(const MachineFunction &MF) const {
   BitVector Reserved(getNumRegs());
   Reserved.set(AMDGPU::MODE);
@@ -742,7 +683,7 @@ BitVector SIRegisterInfo::getReservedRegs(const MachineFunction &MF) const {
 
   // Reserve VGPRs/AGPRs.
   //
-  auto [MaxNumVGPRs, MaxNumAGPRs] = getMaxNumVectorRegs(MF);
+  auto [MaxNumVGPRs, MaxNumAGPRs] = ST.getMaxNumVectorRegs(MF.getFunction());
 
   for (const TargetRegisterClass *RC : regclasses()) {
     if (RC->isBaseClass() && isVGPRClass(RC)) {
diff --git a/llvm/lib/Target/AMDGPU/SIRegisterInfo.h b/llvm/lib/Target/AMDGPU/SIRegisterInfo.h
index 0008e5f..5508f07 100644
--- a/llvm/lib/Target/AMDGPU/SIRegisterInfo.h
+++ b/llvm/lib/Target/AMDGPU/SIRegisterInfo.h
@@ -90,11 +90,6 @@ public:
   /// spilling is needed.
   MCRegister reservedPrivateSegmentBufferReg(const MachineFunction &MF) const;
 
-  /// Return a pair of maximum numbers of VGPRs and AGPRs that meet the number
-  /// of waves per execution unit required for the function \p MF.
-  std::pair<unsigned, unsigned>
-  getMaxNumVectorRegs(const MachineFunction &MF) const;
-
   BitVector getReservedRegs(const MachineFunction &MF) const override;
   bool isAsmClobberable(const MachineFunction &MF,
                         MCRegister PhysReg) const override;
diff --git a/llvm/lib/Target/AMDGPU/SIRegisterInfo.td b/llvm/lib/Target/AMDGPU/SIRegisterInfo.td
index 0039d2f..218841d 100644
--- a/llvm/lib/Target/AMDGPU/SIRegisterInfo.td
+++ b/llvm/lib/Target/AMDGPU/SIRegisterInfo.td
@@ -109,6 +109,23 @@ class SIRegisterClass <string n, list<ValueType> rTypes, int Align, dag rList>
   let TSFlags{2} = HasVGPR;
   let TSFlags{3} = HasAGPR;
   let TSFlags{4} = HasSGPR;
+
+  // RA will use RegisterClass AllocationPriority amongst other info (e.g. ordering in the basic block) 
+  // to decide which registers to try to assign first. Usually, this RegisterClass priority is given
+  // very high priority, if not the highest priority, when considering which VirtReg to allocate next.
+  //
+  // We have 5 bits to assign AllocationPriorities to RegisterClasses. Generally, it is beneficial to 
+  // assign more constrained RegisterClasses first. As a result, we prioritize register classes with 
+  // more 32 bit tuples (e.g. VReg_512) over registers with fewer tuples (e.g. VGPR_32). 
+  // 
+  // The interesting case is the vector register case on architectures which have ARegs, VRegs, AVRegs.
+  // In this case, we would like to assign ARegs and VRegs before AVRegs, as AVRegs are less constrained
+  // and can be assigned to both AGPRs and VGPRs. We use the 5th bit to encode this into the 
+  // RegisterClass AllocationPriority. BaseClassPriority is used to turn the bit on, and BaseClassScaleFactor
+  // is used for scaling of the bit (i.e. 1 << 4).
+  field int BaseClassPriority = 1;
+  field int BaseClassScaleFactor = 16;
+
 }
 
 multiclass SIRegLoHi16 <string n, bits<8> regIdx, bit ArtificialHigh = 1,
@@ -575,7 +592,7 @@ let HasVGPR = 1 in {
 def VGPR_16 : SIRegisterClass<"AMDGPU",  Reg16Types.types, 16,
                             (add (interleave (sequence "VGPR%u_LO16", 0, 255),
                                              (sequence "VGPR%u_HI16", 0, 255)))> {
-  let AllocationPriority = 2;
+  let AllocationPriority = !add(2, !mul(BaseClassPriority, BaseClassScaleFactor));
   let Size = 16;
   let GeneratePressureSet = 0;
 
@@ -601,7 +618,7 @@ def VGPR_16_Lo128 : SIRegisterClass<"AMDGPU",  Reg16Types.types, 16,
 // i16/f16 only on VI+
 def VGPR_32 : SIRegisterClass<"AMDGPU", !listconcat(Reg32Types.types, Reg16Types.types), 32,
                             (add (sequence "VGPR%u", 0, 255))> {
-  let AllocationPriority = 0;
+  let AllocationPriority = !add(0, !mul(BaseClassPriority, BaseClassScaleFactor));
   let Size = 32;
   let Weight = 1;
   let BaseClassOrder = 32;
@@ -610,7 +627,7 @@ def VGPR_32 : SIRegisterClass<"AMDGPU", !listconcat(Reg32Types.types, Reg16Types
 // Identical to VGPR_32 except it only contains the low 128 (Lo128) registers.
 def VGPR_32_Lo128 : SIRegisterClass<"AMDGPU", !listconcat(Reg32Types.types, Reg16Types.types), 32,
                             (add (sequence "VGPR%u", 0, 127))> {
-  let AllocationPriority = 0;
+  let AllocationPriority = !add(0, !mul(BaseClassPriority, BaseClassScaleFactor));
   let GeneratePressureSet = 0;
   let Size = 32;
   let Weight = 1;
@@ -668,7 +685,7 @@ def AGPR_LO16 : SIRegisterClass<"AMDGPU", Reg16Types.types, 16,
 // AccVGPR 32-bit registers
 def AGPR_32 : SIRegisterClass<"AMDGPU", [i32, f32, i16, f16, bf16, v2i16, v2f16, v2bf16], 32,
                             (add (sequence "AGPR%u", 0, 255))> {
-  let AllocationPriority = 0;
+  let AllocationPriority = !add(0, !mul(BaseClassPriority, BaseClassScaleFactor));
   let Size = 32;
   let Weight = 1;
   let BaseClassOrder = 32;
@@ -940,14 +957,23 @@ class VRegClassBase<int numRegs, list<ValueType> regTypes, dag regList> :
 
   // Requires n v_mov_b32 to copy
   let CopyCost = numRegs;
-  let AllocationPriority = !sub(numRegs, 1);
+
+  // Since we only have 5 bits for the RegisterClass Allocation Priorty, and since we use the 
+  // 5th bit for BaseClassPriority, we need to encode the SizePriority into 4 bits. As a result 
+  // of this encoding, for registers with numRegs 15 or 16, we give SizePriority of 14, and for 
+  // regsters with numRegs 17+ we give SizePriority of 15. In  practice, there is only one 
+  // RegClass per Vector Register type in each of these groups (i.e. numRegs = 15,16 : {VReg_512}, 
+  // and numRegs = 17+ : {VReg_1024}). Therefore, we have not lost any info by compressing. 
+  defvar SizePrioriity = !if(!le(numRegs, 14), !sub(numRegs, 1), !if(!le(numRegs, 16), 14, 15));
+  
+  let AllocationPriority = !add(SizePrioriity, !mul(BaseClassPriority, BaseClassScaleFactor));
   let Weight = numRegs;
 }
 
 // Define a register tuple class, along with one requiring an even
 // aligned base register.
 multiclass VRegClass<int numRegs, list<ValueType> regTypes, dag regList> {
-  let HasVGPR = 1 in {
+  let HasVGPR = 1, BaseClassPriority = 1 in {
     // Define the regular class.
     def "" : VRegClassBase<numRegs, regTypes, regList> {
       let BaseClassOrder = !mul(numRegs, 32);
@@ -981,7 +1007,7 @@ defm VReg_1024 : VRegClass<32, Reg1024Types.types, (add VGPR_1024)>;
 }
 
 multiclass ARegClass<int numRegs, list<ValueType> regTypes, dag regList> {
-  let CopyCost = !add(numRegs, numRegs, 1), HasAGPR = 1 in {
+  let CopyCost = !add(numRegs, numRegs, 1), HasAGPR = 1, BaseClassPriority = 1 in {
     // Define the regular class.
     def "" : VRegClassBase<numRegs, regTypes, regList> {
       let BaseClassOrder = !mul(numRegs, 32);
@@ -1066,6 +1092,7 @@ def VS_64 : SIRegisterClass<"AMDGPU", VReg_64.RegTypes, 32, (add VReg_64, SReg_6
 def AV_32 : SIRegisterClass<"AMDGPU", VGPR_32.RegTypes, 32, (add VGPR_32, AGPR_32)> {
   let HasVGPR = 1;
   let HasAGPR = 1;
+  let BaseClassPriority = 0;
   let Size = 32;
 }
 } // End GeneratePressureSet = 0
@@ -1074,7 +1101,7 @@ def AV_32 : SIRegisterClass<"AMDGPU", VGPR_32.RegTypes, 32, (add VGPR_32, AGPR_3
 // aligned base register.
 multiclass AVRegClass<int numRegs, list<ValueType> regTypes,
                       dag vregList,  dag aregList> {
-  let HasVGPR = 1, HasAGPR = 1 in {
+  let HasVGPR = 1, HasAGPR = 1, BaseClassPriority = 0 in {
     // Define the regular class.
     def "" : VRegClassBase<numRegs, regTypes, (add vregList, aregList)>;
 
diff --git a/llvm/lib/Target/AMDGPU/VOP3PInstructions.td b/llvm/lib/Target/AMDGPU/VOP3PInstructions.td
index c812dc9..95fcd4a 100644
--- a/llvm/lib/Target/AMDGPU/VOP3PInstructions.td
+++ b/llvm/lib/Target/AMDGPU/VOP3PInstructions.td
@@ -1236,6 +1236,12 @@ let isCommutable = 1, isReMaterializable = 1 in {
     defm V_PK_MIN_NUM_BF16 : VOP3PInst<"v_pk_min_num_bf16", VOP3P_Profile<VOP_V2BF16_V2BF16_V2BF16, VOP3_PACKED>, fminnum_like>;
     defm V_PK_MAX_NUM_BF16 : VOP3PInst<"v_pk_max_num_bf16", VOP3P_Profile<VOP_V2BF16_V2BF16_V2BF16, VOP3_PACKED>, fmaxnum_like>;
     defm V_PK_FMA_BF16     : VOP3PInst<"v_pk_fma_bf16", VOP3P_Profile<VOP_V2BF16_V2BF16_V2BF16_V2BF16, VOP3_PACKED>, any_fma>;
+
+    // Scalar pseudo used to emulate AMDGPUClamp.
+    // Expanded to V_PK_MAX_NUM_BF16 with unused high half.
+    // FIXME-TRUE16: Pseudo expansion of this won't work with True16.
+    let True16Predicate = UseFakeTrue16Insts in
+    defm V_MAX_BF16_PSEUDO : VOP3Inst <"v_max_bf16", VOP_BF16_BF16_BF16>;
   }
 } // End isCommutable = 1, isReMaterializable = 1
 
diff --git a/llvm/lib/Target/ARM/ARMISelLowering.cpp b/llvm/lib/Target/ARM/ARMISelLowering.cpp
index fca5dff..066b392 100644
--- a/llvm/lib/Target/ARM/ARMISelLowering.cpp
+++ b/llvm/lib/Target/ARM/ARMISelLowering.cpp
@@ -370,6 +370,11 @@ void ARMTargetLowering::addMVEVectorTypes(bool HasMVEFP) {
       setOperationAction(ISD::FMINNUM, VT, Legal);
       setOperationAction(ISD::FMAXNUM, VT, Legal);
       setOperationAction(ISD::FROUND, VT, Legal);
+      setOperationAction(ISD::FROUNDEVEN, VT, Legal);
+      setOperationAction(ISD::FRINT, VT, Legal);
+      setOperationAction(ISD::FTRUNC, VT, Legal);
+      setOperationAction(ISD::FFLOOR, VT, Legal);
+      setOperationAction(ISD::FCEIL, VT, Legal);
       setOperationAction(ISD::VECREDUCE_FADD, VT, Custom);
       setOperationAction(ISD::VECREDUCE_FMUL, VT, Custom);
       setOperationAction(ISD::VECREDUCE_FMIN, VT, Custom);
@@ -1507,6 +1512,12 @@ ARMTargetLowering::ARMTargetLowering(const TargetMachine &TM_,
     setOperationAction(ISD::FLOG2, MVT::f16, Promote);
 
     setOperationAction(ISD::FROUND, MVT::f16, Legal);
+    setOperationAction(ISD::FROUNDEVEN, MVT::f16, Legal);
+    setOperationAction(ISD::FTRUNC, MVT::f16, Legal);
+    setOperationAction(ISD::FNEARBYINT, MVT::f16, Legal);
+    setOperationAction(ISD::FRINT, MVT::f16, Legal);
+    setOperationAction(ISD::FFLOOR, MVT::f16, Legal);
+    setOperationAction(ISD::FCEIL, MVT::f16, Legal);
   }
 
   if (Subtarget->hasNEON()) {
diff --git a/llvm/lib/Target/ARM/AsmParser/ARMAsmParser.cpp b/llvm/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
index ec6f4e2..ece6c10 100644
--- a/llvm/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
+++ b/llvm/lib/Target/ARM/AsmParser/ARMAsmParser.cpp
@@ -12327,7 +12327,7 @@ bool ARMAsmParser::parseDirectiveEven(SMLoc L) {
   }
 
   assert(Section && "must have section to emit alignment");
-  if (Section->useCodeAlign())
+  if (getContext().getAsmInfo()->useCodeAlign(*Section))
     getStreamer().emitCodeAlignment(Align(2), &getSTI());
   else
     getStreamer().emitValueToAlignment(Align(2));
@@ -12525,7 +12525,7 @@ bool ARMAsmParser::parseDirectiveAlign(SMLoc L) {
     // '.align' is target specifically handled to mean 2**2 byte alignment.
     const MCSection *Section = getStreamer().getCurrentSectionOnly();
     assert(Section && "must have section to emit alignment");
-    if (Section->useCodeAlign())
+    if (getContext().getAsmInfo()->useCodeAlign(*Section))
       getStreamer().emitCodeAlignment(Align(4), &getSTI(), 0);
     else
       getStreamer().emitValueToAlignment(Align(4), 0, 1, 0);
diff --git a/llvm/lib/Target/ARM/MCTargetDesc/ARMELFStreamer.cpp b/llvm/lib/Target/ARM/MCTargetDesc/ARMELFStreamer.cpp
index f64a223..868556b 100644
--- a/llvm/lib/Target/ARM/MCTargetDesc/ARMELFStreamer.cpp
+++ b/llvm/lib/Target/ARM/MCTargetDesc/ARMELFStreamer.cpp
@@ -1138,7 +1138,8 @@ void ARMTargetELFStreamer::finish() {
   MCContext &Ctx = getContext();
   auto &Asm = getStreamer().getAssembler();
   if (any_of(Asm, [](const MCSection &Sec) {
-        return cast<MCSectionELF>(Sec).getFlags() & ELF::SHF_ARM_PURECODE;
+        return static_cast<const MCSectionELF &>(Sec).getFlags() &
+               ELF::SHF_ARM_PURECODE;
       })) {
     auto *Text =
         static_cast<MCSectionELF *>(Ctx.getObjectFileInfo()->getTextSection());
diff --git a/llvm/lib/Target/AVR/AVRAsmPrinter.cpp b/llvm/lib/Target/AVR/AVRAsmPrinter.cpp
index ad8aa571..0fb33cd 100644
--- a/llvm/lib/Target/AVR/AVRAsmPrinter.cpp
+++ b/llvm/lib/Target/AVR/AVRAsmPrinter.cpp
@@ -260,7 +260,7 @@ bool AVRAsmPrinter::doFinalization(Module &M) {
       continue;
     }
 
-    auto *Section = cast<MCSectionELF>(TLOF.SectionForGlobal(&GO, TM));
+    auto *Section = static_cast<MCSectionELF *>(TLOF.SectionForGlobal(&GO, TM));
     if (Section->getName().starts_with(".data"))
       NeedsCopyData = true;
     else if (Section->getName().starts_with(".rodata") && SubTM->hasLPM())
diff --git a/llvm/lib/Target/AVR/MCTargetDesc/AVRMCAsmInfo.cpp b/llvm/lib/Target/AVR/MCTargetDesc/AVRMCAsmInfo.cpp
index c2c1bb4..0615ec9 100644
--- a/llvm/lib/Target/AVR/MCTargetDesc/AVRMCAsmInfo.cpp
+++ b/llvm/lib/Target/AVR/MCTargetDesc/AVRMCAsmInfo.cpp
@@ -24,8 +24,6 @@ AVRMCAsmInfo::AVRMCAsmInfo(const Triple &TT, const MCTargetOptions &Options) {
   CalleeSaveStackSlotSize = 2;
   CommentString = ";";
   SeparatorString = "$";
-  PrivateGlobalPrefix = ".L";
-  PrivateLabelPrefix = ".L";
   UsesELFSectionDirectiveForBSS = true;
   SupportsDebugInformation = true;
 }
diff --git a/llvm/lib/Target/AVR/MCTargetDesc/AVRMCAsmInfo.h b/llvm/lib/Target/AVR/MCTargetDesc/AVRMCAsmInfo.h
index fab2713..1915fa8 100644
--- a/llvm/lib/Target/AVR/MCTargetDesc/AVRMCAsmInfo.h
+++ b/llvm/lib/Target/AVR/MCTargetDesc/AVRMCAsmInfo.h
@@ -14,7 +14,7 @@
 #define LLVM_AVR_ASM_INFO_H
 
 #include "MCTargetDesc/AVRMCExpr.h"
-#include "llvm/MC/MCAsmInfo.h"
+#include "llvm/MC/MCAsmInfoELF.h"
 #include "llvm/MC/MCExpr.h"
 
 namespace llvm {
@@ -22,7 +22,7 @@ namespace llvm {
 class Triple;
 
 /// Specifies the format of AVR assembly files.
-class AVRMCAsmInfo : public MCAsmInfo {
+class AVRMCAsmInfo : public MCAsmInfoELF {
 public:
   explicit AVRMCAsmInfo(const Triple &TT, const MCTargetOptions &Options);
   void printSpecifierExpr(raw_ostream &OS,
diff --git a/llvm/lib/Target/BPF/BTFDebug.cpp b/llvm/lib/Target/BPF/BTFDebug.cpp
index 1e29a0f..a87b9a2 100644
--- a/llvm/lib/Target/BPF/BTFDebug.cpp
+++ b/llvm/lib/Target/BPF/BTFDebug.cpp
@@ -1255,10 +1255,8 @@ void BTFDebug::beginFunctionImpl(const MachineFunction *MF) {
   FuncInfo.Label = FuncLabel;
   FuncInfo.TypeId = FuncTypeId;
   if (FuncLabel->isInSection()) {
-    MCSection &Section = FuncLabel->getSection();
-    const MCSectionELF *SectionELF = dyn_cast<MCSectionELF>(&Section);
-    assert(SectionELF && "Null section for Function Label");
-    SecNameOff = addString(SectionELF->getName());
+    auto &Sec = static_cast<const MCSectionELF &>(FuncLabel->getSection());
+    SecNameOff = addString(Sec.getName());
   } else {
     SecNameOff = addString(".text");
   }
diff --git a/llvm/lib/Target/BPF/MCTargetDesc/BPFELFObjectWriter.cpp b/llvm/lib/Target/BPF/MCTargetDesc/BPFELFObjectWriter.cpp
index 827e928..bb74f6a 100644
--- a/llvm/lib/Target/BPF/MCTargetDesc/BPFELFObjectWriter.cpp
+++ b/llvm/lib/Target/BPF/MCTargetDesc/BPFELFObjectWriter.cpp
@@ -54,11 +54,8 @@ unsigned BPFELFObjectWriter::getRelocType(const MCFixup &Fixup,
       const MCSymbol &Sym = *A;
 
       if (Sym.isDefined()) {
-        MCSection &Section = Sym.getSection();
-        const MCSectionELF *SectionELF = dyn_cast<MCSectionELF>(&Section);
-        assert(SectionELF && "Null section for reloc symbol");
-
-        unsigned Flags = SectionELF->getFlags();
+        auto &Section = static_cast<const MCSectionELF &>(Sym.getSection());
+        unsigned Flags = Section.getFlags();
 
         if (Sym.isTemporary()) {
           // .BTF.ext generates FK_Data_4 relocations for
diff --git a/llvm/lib/Target/BPF/MCTargetDesc/BPFMCAsmInfo.h b/llvm/lib/Target/BPF/MCTargetDesc/BPFMCAsmInfo.h
index 7b21684..63d6e6f 100644
--- a/llvm/lib/Target/BPF/MCTargetDesc/BPFMCAsmInfo.h
+++ b/llvm/lib/Target/BPF/MCTargetDesc/BPFMCAsmInfo.h
@@ -13,18 +13,19 @@
 #ifndef LLVM_LIB_TARGET_BPF_MCTARGETDESC_BPFMCASMINFO_H
 #define LLVM_LIB_TARGET_BPF_MCTARGETDESC_BPFMCASMINFO_H
 
-#include "llvm/MC/MCAsmInfo.h"
+#include "llvm/MC/MCAsmInfoELF.h"
 #include "llvm/TargetParser/Triple.h"
 
 namespace llvm {
 
-class BPFMCAsmInfo : public MCAsmInfo {
+class BPFMCAsmInfo : public MCAsmInfoELF {
 public:
   explicit BPFMCAsmInfo(const Triple &TT, const MCTargetOptions &Options) {
     if (TT.getArch() == Triple::bpfeb)
       IsLittleEndian = false;
 
     PrivateGlobalPrefix = ".L";
+    PrivateLabelPrefix = "L";
     WeakRefDirective = "\t.weak\t";
 
     UsesELFSectionDirectiveForBSS = true;
diff --git a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
index e915a3c4..d96136c 100644
--- a/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
+++ b/llvm/lib/Target/LoongArch/LoongArchISelLowering.cpp
@@ -2385,23 +2385,6 @@ SDValue LoongArchTargetLowering::lowerBF16_TO_FP(SDValue Op,
   return Res;
 }
 
-static bool isConstantOrUndef(const SDValue Op) {
-  if (Op->isUndef())
-    return true;
-  if (isa<ConstantSDNode>(Op))
-    return true;
-  if (isa<ConstantFPSDNode>(Op))
-    return true;
-  return false;
-}
-
-static bool isConstantOrUndefBUILD_VECTOR(const BuildVectorSDNode *Op) {
-  for (unsigned i = 0; i < Op->getNumOperands(); ++i)
-    if (isConstantOrUndef(Op->getOperand(i)))
-      return true;
-  return false;
-}
-
 // Lower BUILD_VECTOR as broadcast load (if possible).
 // For example:
 //   %a = load i8, ptr %ptr
@@ -2451,10 +2434,14 @@ SDValue LoongArchTargetLowering::lowerBUILD_VECTOR(SDValue Op,
                                                    SelectionDAG &DAG) const {
   BuildVectorSDNode *Node = cast<BuildVectorSDNode>(Op);
   EVT ResTy = Op->getValueType(0);
+  unsigned NumElts = ResTy.getVectorNumElements();
   SDLoc DL(Op);
   APInt SplatValue, SplatUndef;
   unsigned SplatBitSize;
   bool HasAnyUndefs;
+  bool IsConstant = false;
+  bool UseSameConstant = true;
+  SDValue ConstantValue;
   bool Is128Vec = ResTy.is128BitVector();
   bool Is256Vec = ResTy.is256BitVector();
 
@@ -2505,20 +2492,45 @@ SDValue LoongArchTargetLowering::lowerBUILD_VECTOR(SDValue Op,
   if (DAG.isSplatValue(Op, /*AllowUndefs=*/false))
     return Op;
 
-  if (!isConstantOrUndefBUILD_VECTOR(Node)) {
+  for (unsigned i = 0; i < NumElts; ++i) {
+    SDValue Opi = Node->getOperand(i);
+    if (isIntOrFPConstant(Opi)) {
+      IsConstant = true;
+      if (!ConstantValue.getNode())
+        ConstantValue = Opi;
+      else if (ConstantValue != Opi)
+        UseSameConstant = false;
+    }
+  }
+
+  // If the type of BUILD_VECTOR is v2f64, custom legalizing it has no benefits.
+  if (IsConstant && UseSameConstant && ResTy != MVT::v2f64) {
+    SDValue Result = DAG.getSplatBuildVector(ResTy, DL, ConstantValue);
+    for (unsigned i = 0; i < NumElts; ++i) {
+      SDValue Opi = Node->getOperand(i);
+      if (!isIntOrFPConstant(Opi))
+        Result = DAG.getNode(ISD::INSERT_VECTOR_ELT, DL, ResTy, Result, Opi,
+                             DAG.getConstant(i, DL, Subtarget.getGRLenVT()));
+    }
+    return Result;
+  }
+
+  if (!IsConstant) {
     // Use INSERT_VECTOR_ELT operations rather than expand to stores.
     // The resulting code is the same length as the expansion, but it doesn't
     // use memory operations.
-    EVT ResTy = Node->getValueType(0);
-
     assert(ResTy.isVector());
 
-    unsigned NumElts = ResTy.getVectorNumElements();
-    SDValue Vector =
-        DAG.getNode(ISD::SCALAR_TO_VECTOR, DL, ResTy, Node->getOperand(0));
+    SDValue Op0 = Node->getOperand(0);
+    SDValue Vector = DAG.getUNDEF(ResTy);
+
+    if (!Op0.isUndef())
+      Vector = DAG.getNode(ISD::SCALAR_TO_VECTOR, DL, ResTy, Op0);
     for (unsigned i = 1; i < NumElts; ++i) {
-      Vector = DAG.getNode(ISD::INSERT_VECTOR_ELT, DL, ResTy, Vector,
-                           Node->getOperand(i),
+      SDValue Opi = Node->getOperand(i);
+      if (Opi.isUndef())
+        continue;
+      Vector = DAG.getNode(ISD::INSERT_VECTOR_ELT, DL, ResTy, Vector, Opi,
                            DAG.getConstant(i, DL, Subtarget.getGRLenVT()));
     }
     return Vector;
diff --git a/llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchAsmBackend.cpp b/llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchAsmBackend.cpp
index 8fa72bc..d9ea88c 100644
--- a/llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchAsmBackend.cpp
+++ b/llvm/lib/Target/LoongArch/MCTargetDesc/LoongArchAsmBackend.cpp
@@ -254,6 +254,7 @@ bool LoongArchAsmBackend::relaxAlign(MCFragment &F, unsigned &Size) {
   MCFixup Fixup =
       MCFixup::create(0, Expr, FirstLiteralRelocationKind + ELF::R_LARCH_ALIGN);
   F.setVarFixups({Fixup});
+  F.setLinkerRelaxable();
   F.getParent()->setLinkerRelaxable();
   return true;
 }
diff --git a/llvm/lib/Target/Mips/MCTargetDesc/MipsTargetStreamer.cpp b/llvm/lib/Target/Mips/MCTargetDesc/MipsTargetStreamer.cpp
index feb4eb3..d9680c7 100644
--- a/llvm/lib/Target/Mips/MCTargetDesc/MipsTargetStreamer.cpp
+++ b/llvm/lib/Target/Mips/MCTargetDesc/MipsTargetStreamer.cpp
@@ -969,7 +969,7 @@ void MipsTargetELFStreamer::finish() {
 
       Align Alignment = Section.getAlign();
       S.switchSection(&Section);
-      if (Section.useCodeAlign())
+      if (getContext().getAsmInfo()->useCodeAlign(Section))
         S.emitCodeAlignment(Alignment, &STI, Alignment.value());
       else
         S.emitValueToAlignment(Alignment, 0, 1, Alignment.value());
diff --git a/llvm/lib/Target/Mips/MipsAsmPrinter.cpp b/llvm/lib/Target/Mips/MipsAsmPrinter.cpp
index ca03310..a2e48ab 100644
--- a/llvm/lib/Target/Mips/MipsAsmPrinter.cpp
+++ b/llvm/lib/Target/Mips/MipsAsmPrinter.cpp
@@ -737,14 +737,18 @@ void MipsAsmPrinter::emitStartOfAsmFile(Module &M) {
     if (FS.empty() && M.size() && F->hasFnAttribute("target-features"))
       FS = F->getFnAttribute("target-features").getValueAsString();
 
+    std::string strFS = FS.str();
+    if (M.size() && F->getFnAttribute("use-soft-float").getValueAsBool())
+      strFS += strFS.empty() ? "+soft-float" : ",+soft-float";
+
     // Compute MIPS architecture attributes based on the default subtarget
     // that we'd have constructed.
     // FIXME: For ifunc related functions we could iterate over and look
     // for a feature string that doesn't match the default one.
     StringRef CPU = MIPS_MC::selectMipsCPU(TT, TM.getTargetCPU());
     const MipsTargetMachine &MTM = static_cast<const MipsTargetMachine &>(TM);
-    const MipsSubtarget STI(TT, CPU, FS, MTM.isLittleEndian(), MTM,
-                            std::nullopt);
+    const MipsSubtarget STI(TT, CPU, StringRef(strFS), MTM.isLittleEndian(),
+                            MTM, std::nullopt);
 
     bool IsABICalls = STI.isABICalls();
     const MipsABIInfo &ABI = MTM.getABI();
diff --git a/llvm/lib/Target/Mips/MipsISelLowering.cpp b/llvm/lib/Target/Mips/MipsISelLowering.cpp
index 0e581a7..ec6b382 100644
--- a/llvm/lib/Target/Mips/MipsISelLowering.cpp
+++ b/llvm/lib/Target/Mips/MipsISelLowering.cpp
@@ -522,9 +522,6 @@ MipsTargetLowering::MipsTargetLowering(const MipsTargetMachine &TM,
 
   setOperationAction(ISD::TRAP, MVT::Other, Legal);
 
-  setOperationAction(ISD::ConstantFP, MVT::f32, Custom);
-  setOperationAction(ISD::ConstantFP, MVT::f64, Custom);
-
   setTargetDAGCombine({ISD::SDIVREM, ISD::UDIVREM, ISD::SELECT, ISD::AND,
                        ISD::OR, ISD::ADD, ISD::SUB, ISD::AssertZext, ISD::SHL,
                        ISD::SIGN_EXTEND});
@@ -1360,8 +1357,6 @@ LowerOperation(SDValue Op, SelectionDAG &DAG) const
   case ISD::FP_TO_SINT:         return lowerFP_TO_SINT(Op, DAG);
   case ISD::READCYCLECOUNTER:
     return lowerREADCYCLECOUNTER(Op, DAG);
-  case ISD::ConstantFP:
-    return lowerConstantFP(Op, DAG);
   }
   return SDValue();
 }
@@ -3019,30 +3014,6 @@ SDValue MipsTargetLowering::lowerFP_TO_SINT(SDValue Op,
   return DAG.getNode(ISD::BITCAST, SDLoc(Op), Op.getValueType(), Trunc);
 }
 
-SDValue MipsTargetLowering::lowerConstantFP(SDValue Op,
-                                            SelectionDAG &DAG) const {
-  SDLoc DL(Op);
-  EVT VT = Op.getSimpleValueType();
-  SDNode *N = Op.getNode();
-  ConstantFPSDNode *CFP = cast<ConstantFPSDNode>(N);
-
-  if (!CFP->isNaN() || Subtarget.isNaN2008()) {
-    return SDValue();
-  }
-
-  APFloat NaNValue = CFP->getValueAPF();
-  auto &Sem = NaNValue.getSemantics();
-
-  // The MSB of the mantissa should be zero for QNaNs in the MIPS legacy NaN
-  // encodings, and one for sNaNs. Check every NaN constants and make sure
-  // they are correctly encoded for legacy encodings.
-  if (!NaNValue.isSignaling()) {
-    APFloat RealQNaN = NaNValue.getSNaN(Sem);
-    return DAG.getConstantFP(RealQNaN, DL, VT);
-  }
-  return SDValue();
-}
-
 //===----------------------------------------------------------------------===//
 //                      Calling Convention Implementation
 //===----------------------------------------------------------------------===//
diff --git a/llvm/lib/Target/Mips/MipsISelLowering.h b/llvm/lib/Target/Mips/MipsISelLowering.h
index 31ac5d4..c65c76c 100644
--- a/llvm/lib/Target/Mips/MipsISelLowering.h
+++ b/llvm/lib/Target/Mips/MipsISelLowering.h
@@ -592,7 +592,6 @@ class TargetRegisterClass;
     SDValue lowerEH_DWARF_CFA(SDValue Op, SelectionDAG &DAG) const;
     SDValue lowerFP_TO_SINT(SDValue Op, SelectionDAG &DAG) const;
     SDValue lowerREADCYCLECOUNTER(SDValue Op, SelectionDAG &DAG) const;
-    SDValue lowerConstantFP(SDValue Op, SelectionDAG &DAG) const;
 
     /// isEligibleForTailCallOptimization - Check whether the call is eligible
     /// for tail call optimization.
diff --git a/llvm/lib/Target/NVPTX/MCTargetDesc/NVPTXMCAsmInfo.cpp b/llvm/lib/Target/NVPTX/MCTargetDesc/NVPTXMCAsmInfo.cpp
index 614b321..ce9cd12 100644
--- a/llvm/lib/Target/NVPTX/MCTargetDesc/NVPTXMCAsmInfo.cpp
+++ b/llvm/lib/Target/NVPTX/MCTargetDesc/NVPTXMCAsmInfo.cpp
@@ -15,8 +15,6 @@
 
 using namespace llvm;
 
-void NVPTXMCAsmInfo::anchor() {}
-
 NVPTXMCAsmInfo::NVPTXMCAsmInfo(const Triple &TheTriple,
                                const MCTargetOptions &Options) {
   if (TheTriple.getArch() == Triple::nvptx64) {
diff --git a/llvm/lib/Target/NVPTX/MCTargetDesc/NVPTXMCAsmInfo.h b/llvm/lib/Target/NVPTX/MCTargetDesc/NVPTXMCAsmInfo.h
index 77c4dae..f071406 100644
--- a/llvm/lib/Target/NVPTX/MCTargetDesc/NVPTXMCAsmInfo.h
+++ b/llvm/lib/Target/NVPTX/MCTargetDesc/NVPTXMCAsmInfo.h
@@ -19,8 +19,6 @@ namespace llvm {
 class Triple;
 
 class NVPTXMCAsmInfo : public MCAsmInfo {
-  virtual void anchor();
-
 public:
   explicit NVPTXMCAsmInfo(const Triple &TheTriple,
                           const MCTargetOptions &Options);
diff --git a/llvm/lib/Target/NVPTX/MCTargetDesc/NVPTXTargetStreamer.cpp b/llvm/lib/Target/NVPTX/MCTargetDesc/NVPTXTargetStreamer.cpp
index 9f91143..329e3b5 100644
--- a/llvm/lib/Target/NVPTX/MCTargetDesc/NVPTXTargetStreamer.cpp
+++ b/llvm/lib/Target/NVPTX/MCTargetDesc/NVPTXTargetStreamer.cpp
@@ -97,10 +97,7 @@ void NVPTXTargetStreamer::changeSection(const MCSection *CurSection,
   if (isDwarfSection(FI, Section)) {
     // Emit DWARF .file directives in the outermost scope.
     outputDwarfFileDirectives();
-    OS << "\t.section";
-    Section->printSwitchToSection(*getStreamer().getContext().getAsmInfo(),
-                                  getStreamer().getContext().getTargetTriple(),
-                                  OS, SubSection);
+    OS << "\t.section\t" << Section->getName() << '\n';
     // DWARF sections are enclosed into braces - emit the open one.
     OS << "\t{\n";
     HasSections = true;
diff --git a/llvm/lib/Target/PowerPC/MCTargetDesc/PPCMCAsmInfo.cpp b/llvm/lib/Target/PowerPC/MCTargetDesc/PPCMCAsmInfo.cpp
index 8baf866..1af2f9c 100644
--- a/llvm/lib/Target/PowerPC/MCTargetDesc/PPCMCAsmInfo.cpp
+++ b/llvm/lib/Target/PowerPC/MCTargetDesc/PPCMCAsmInfo.cpp
@@ -220,8 +220,6 @@ bool PPCELFMCAsmInfo::evaluateAsRelocatableImpl(const MCSpecifierExpr &Expr,
   return evaluateAsRelocatable(Expr, Res, Asm);
 }
 
-void PPCXCOFFMCAsmInfo::anchor() {}
-
 PPCXCOFFMCAsmInfo::PPCXCOFFMCAsmInfo(bool Is64Bit, const Triple &T) {
   if (T.getArch() == Triple::ppc64le || T.getArch() == Triple::ppcle)
     report_fatal_error("XCOFF is not supported for little-endian targets");
diff --git a/llvm/lib/Target/PowerPC/MCTargetDesc/PPCMCAsmInfo.h b/llvm/lib/Target/PowerPC/MCTargetDesc/PPCMCAsmInfo.h
index 0f945b3..6af1bd7 100644
--- a/llvm/lib/Target/PowerPC/MCTargetDesc/PPCMCAsmInfo.h
+++ b/llvm/lib/Target/PowerPC/MCTargetDesc/PPCMCAsmInfo.h
@@ -33,8 +33,6 @@ public:
 };
 
 class PPCXCOFFMCAsmInfo : public MCAsmInfoXCOFF {
-  void anchor() override;
-
 public:
   explicit PPCXCOFFMCAsmInfo(bool is64Bit, const Triple &);
   void printSpecifierExpr(raw_ostream &OS,
diff --git a/llvm/lib/Target/PowerPC/MCTargetDesc/PPCMCTargetDesc.cpp b/llvm/lib/Target/PowerPC/MCTargetDesc/PPCMCTargetDesc.cpp
index 54497d9..3dad0e8 100644
--- a/llvm/lib/Target/PowerPC/MCTargetDesc/PPCMCTargetDesc.cpp
+++ b/llvm/lib/Target/PowerPC/MCTargetDesc/PPCMCTargetDesc.cpp
@@ -213,7 +213,7 @@ public:
   void emitTCEntry(const MCSymbol &S, PPCMCExpr::Specifier Kind) override {
     if (const MCSymbolXCOFF *XSym = dyn_cast<MCSymbolXCOFF>(&S)) {
       MCSymbolXCOFF *TCSym =
-          cast<MCSectionXCOFF>(Streamer.getCurrentSectionOnly())
+          static_cast<const MCSectionXCOFF *>(Streamer.getCurrentSectionOnly())
               ->getQualNameSymbol();
       // On AIX, we have TLS variable offsets (symbol@({gd|ie|le|ld}) depending
       // on the TLS access method (or model). For the general-dynamic access
diff --git a/llvm/lib/Target/PowerPC/PPCAsmPrinter.cpp b/llvm/lib/Target/PowerPC/PPCAsmPrinter.cpp
index a091b21..ce1d51a 100644
--- a/llvm/lib/Target/PowerPC/PPCAsmPrinter.cpp
+++ b/llvm/lib/Target/PowerPC/PPCAsmPrinter.cpp
@@ -2274,9 +2274,9 @@ void PPCAIXAsmPrinter::emitLinkage(const GlobalValue *GV,
 
 void PPCAIXAsmPrinter::SetupMachineFunction(MachineFunction &MF) {
   // Setup CurrentFnDescSym and its containing csect.
-  MCSectionXCOFF *FnDescSec =
-      cast<MCSectionXCOFF>(getObjFileLowering().getSectionForFunctionDescriptor(
-          &MF.getFunction(), TM));
+  auto *FnDescSec = static_cast<MCSectionXCOFF *>(
+      getObjFileLowering().getSectionForFunctionDescriptor(&MF.getFunction(),
+                                                           TM));
   FnDescSec->setAlignment(Align(Subtarget->isPPC64() ? 8 : 4));
 
   CurrentFnDescSym = FnDescSec->getQualNameSymbol();
@@ -2669,9 +2669,9 @@ void PPCAIXAsmPrinter::emitTracebackTable() {
     MCSymbol *EHInfoSym =
         TargetLoweringObjectFileXCOFF::getEHInfoTableSymbol(MF);
     MCSymbol *TOCEntry = lookUpOrCreateTOCEntry(EHInfoSym, TOCType_EHBlock);
-    const MCSymbol *TOCBaseSym =
-        cast<MCSectionXCOFF>(getObjFileLowering().getTOCBaseSection())
-            ->getQualNameSymbol();
+    const MCSymbol *TOCBaseSym = static_cast<const MCSectionXCOFF *>(
+                                     getObjFileLowering().getTOCBaseSection())
+                                     ->getQualNameSymbol();
     const MCExpr *Exp =
         MCBinaryExpr::createSub(MCSymbolRefExpr::create(TOCEntry, Ctx),
                                 MCSymbolRefExpr::create(TOCBaseSym, Ctx), Ctx);
@@ -2788,7 +2788,7 @@ void PPCAIXAsmPrinter::emitGlobalVariableHelper(const GlobalVariable *GV) {
     }
   }
 
-  MCSectionXCOFF *Csect = cast<MCSectionXCOFF>(
+  auto *Csect = static_cast<MCSectionXCOFF *>(
       getObjFileLowering().SectionForGlobal(GV, GVKind, TM));
 
   // Switch to the containing csect.
@@ -2869,9 +2869,9 @@ void PPCAIXAsmPrinter::emitFunctionDescriptor() {
   OutStreamer->emitValue(MCSymbolRefExpr::create(CurrentFnSym, OutContext),
                          PointerSize);
   // Emit TOC base address.
-  const MCSymbol *TOCBaseSym =
-      cast<MCSectionXCOFF>(getObjFileLowering().getTOCBaseSection())
-          ->getQualNameSymbol();
+  const MCSymbol *TOCBaseSym = static_cast<const MCSectionXCOFF *>(
+                                   getObjFileLowering().getTOCBaseSection())
+                                   ->getQualNameSymbol();
   OutStreamer->emitValue(MCSymbolRefExpr::create(TOCBaseSym, OutContext),
                          PointerSize);
   // Emit a null environment pointer.
@@ -2996,10 +2996,10 @@ void PPCAIXAsmPrinter::emitEndOfAsmFile(Module &M) {
       Name += Prefix;
       Name += cast<MCSymbolXCOFF>(I.first.first)->getSymbolTableName();
       MCSymbol *S = OutContext.getOrCreateSymbol(Name);
-      TCEntry = cast<MCSectionXCOFF>(
+      TCEntry = static_cast<MCSectionXCOFF *>(
           getObjFileLowering().getSectionForTOCEntry(S, TM));
     } else {
-      TCEntry = cast<MCSectionXCOFF>(
+      TCEntry = static_cast<MCSectionXCOFF *>(
           getObjFileLowering().getSectionForTOCEntry(I.first.first, TM));
     }
     OutStreamer->switchSection(TCEntry);
@@ -3054,7 +3054,7 @@ bool PPCAIXAsmPrinter::doInitialization(Module &M) {
       return;
 
     SectionKind GOKind = getObjFileLowering().getKindForGlobal(GO, TM);
-    MCSectionXCOFF *Csect = cast<MCSectionXCOFF>(
+    auto *Csect = static_cast<MCSectionXCOFF *>(
         getObjFileLowering().SectionForGlobal(GO, GOKind, TM));
 
     Align GOAlign = getGVAlignment(GO, GO->getDataLayout());
@@ -3316,9 +3316,9 @@ void PPCAIXAsmPrinter::emitTTypeReference(const GlobalValue *GV,
       GlobalType = TOCType_GlobalExternal;
     MCSymbol *TypeInfoSym = TM.getSymbol(GV);
     MCSymbol *TOCEntry = lookUpOrCreateTOCEntry(TypeInfoSym, GlobalType);
-    const MCSymbol *TOCBaseSym =
-        cast<MCSectionXCOFF>(getObjFileLowering().getTOCBaseSection())
-            ->getQualNameSymbol();
+    const MCSymbol *TOCBaseSym = static_cast<const MCSectionXCOFF *>(
+                                     getObjFileLowering().getTOCBaseSection())
+                                     ->getQualNameSymbol();
     auto &Ctx = OutStreamer->getContext();
     const MCExpr *Exp =
         MCBinaryExpr::createSub(MCSymbolRefExpr::create(TOCEntry, Ctx),
diff --git a/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp b/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp
index a143d85..d71c42c 100644
--- a/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp
+++ b/llvm/lib/Target/RISCV/AsmParser/RISCVAsmParser.cpp
@@ -3849,9 +3849,14 @@ bool RISCVAsmParser::processInstruction(MCInst &Inst, SMLoc IDLoc,
   switch (Inst.getOpcode()) {
   default:
     break;
-  case RISCV::PseudoC_ADDI_NOP:
-    emitToStreamer(Out, MCInstBuilder(RISCV::C_NOP));
+  case RISCV::PseudoC_ADDI_NOP: {
+    if (Inst.getOperand(2).getImm() == 0)
+      emitToStreamer(Out, MCInstBuilder(RISCV::C_NOP));
+    else
+      emitToStreamer(
+          Out, MCInstBuilder(RISCV::C_NOP_HINT).addOperand(Inst.getOperand(2)));
     return false;
+  }
   case RISCV::PseudoLLAImm:
   case RISCV::PseudoLAImm:
   case RISCV::PseudoLI: {
diff --git a/llvm/lib/Target/RISCV/Disassembler/RISCVDisassembler.cpp b/llvm/lib/Target/RISCV/Disassembler/RISCVDisassembler.cpp
index fa7bcfa..5e54b82 100644
--- a/llvm/lib/Target/RISCV/Disassembler/RISCVDisassembler.cpp
+++ b/llvm/lib/Target/RISCV/Disassembler/RISCVDisassembler.cpp
@@ -193,21 +193,19 @@ static DecodeStatus DecodeFPR128RegisterClass(MCInst &Inst, uint32_t RegNo,
 static DecodeStatus DecodeGPRNoX0RegisterClass(MCInst &Inst, uint32_t RegNo,
                                                uint64_t Address,
                                                const MCDisassembler *Decoder) {
-  if (RegNo == 0) {
+  if (RegNo == 0)
     return MCDisassembler::Fail;
-  }
 
   return DecodeGPRRegisterClass(Inst, RegNo, Address, Decoder);
 }
 
-static DecodeStatus
-DecodeGPRNoX0X2RegisterClass(MCInst &Inst, uint64_t RegNo, uint32_t Address,
-                             const MCDisassembler *Decoder) {
-  if (RegNo == 2) {
+static DecodeStatus DecodeGPRNoX2RegisterClass(MCInst &Inst, uint64_t RegNo,
+                                               uint32_t Address,
+                                               const MCDisassembler *Decoder) {
+  if (RegNo == 2)
     return MCDisassembler::Fail;
-  }
 
-  return DecodeGPRNoX0RegisterClass(Inst, RegNo, Address, Decoder);
+  return DecodeGPRRegisterClass(Inst, RegNo, Address, Decoder);
 }
 
 static DecodeStatus DecodeGPRNoX31RegisterClass(MCInst &Inst, uint32_t RegNo,
@@ -536,31 +534,6 @@ static DecodeStatus decodeRTZArg(MCInst &Inst, uint32_t Imm, int64_t Address,
   return MCDisassembler::Success;
 }
 
-static DecodeStatus decodeRVCInstrRdRs1ImmZero(MCInst &Inst, uint32_t Insn,
-                                               uint64_t Address,
-                                               const MCDisassembler *Decoder);
-
-static DecodeStatus decodeRVCInstrRdSImm6(MCInst &Inst, uint32_t Insn,
-                                          uint64_t Address,
-                                          const MCDisassembler *Decoder);
-
-static DecodeStatus decodeRVCInstrRdCLUIImm(MCInst &Inst, uint32_t Insn,
-                                            uint64_t Address,
-                                            const MCDisassembler *Decoder);
-
-static DecodeStatus
-decodeRVCInstrRdRs1UImmLog2XLenNonZero(MCInst &Inst, uint32_t Insn,
-                                       uint64_t Address,
-                                       const MCDisassembler *Decoder);
-
-static DecodeStatus decodeRVCInstrRdRs2(MCInst &Inst, uint32_t Insn,
-                                        uint64_t Address,
-                                        const MCDisassembler *Decoder);
-
-static DecodeStatus decodeRVCInstrRdRs1Rs2(MCInst &Inst, uint32_t Insn,
-                                           uint64_t Address,
-                                           const MCDisassembler *Decoder);
-
 static DecodeStatus decodeXTHeadMemPair(MCInst &Inst, uint32_t Insn,
                                         uint64_t Address,
                                         const MCDisassembler *Decoder);
@@ -579,18 +552,6 @@ static DecodeStatus decodeCSSPushPopchk(MCInst &Inst, uint32_t Insn,
 
 #include "RISCVGenDisassemblerTables.inc"
 
-static DecodeStatus decodeRVCInstrRdRs1ImmZero(MCInst &Inst, uint32_t Insn,
-                                               uint64_t Address,
-                                               const MCDisassembler *Decoder) {
-  DecodeStatus S = MCDisassembler::Success;
-  uint32_t Rd = fieldFromInstruction(Insn, 7, 5);
-  if (!Check(S, DecodeGPRNoX0RegisterClass(Inst, Rd, Address, Decoder)))
-    return MCDisassembler::Fail;
-  Inst.addOperand(Inst.getOperand(0));
-  Inst.addOperand(MCOperand::createImm(0));
-  return S;
-}
-
 static DecodeStatus decodeCSSPushPopchk(MCInst &Inst, uint32_t Insn,
                                         uint64_t Address,
                                         const MCDisassembler *Decoder) {
@@ -601,66 +562,6 @@ static DecodeStatus decodeCSSPushPopchk(MCInst &Inst, uint32_t Insn,
   return MCDisassembler::Success;
 }
 
-static DecodeStatus decodeRVCInstrRdSImm6(MCInst &Inst, uint32_t Insn,
-                                          uint64_t Address,
-                                          const MCDisassembler *Decoder) {
-  Inst.addOperand(MCOperand::createReg(RISCV::X0));
-  uint32_t Imm =
-      fieldFromInstruction(Insn, 12, 1) << 5 | fieldFromInstruction(Insn, 2, 5);
-  [[maybe_unused]] DecodeStatus Result =
-      decodeSImmOperand<6>(Inst, Imm, Address, Decoder);
-  assert(Result == MCDisassembler::Success && "Invalid immediate");
-  return MCDisassembler::Success;
-}
-
-static DecodeStatus decodeRVCInstrRdCLUIImm(MCInst &Inst, uint32_t Insn,
-                                            uint64_t Address,
-                                            const MCDisassembler *Decoder) {
-  Inst.addOperand(MCOperand::createReg(RISCV::X0));
-  uint32_t Imm =
-      fieldFromInstruction(Insn, 12, 1) << 5 | fieldFromInstruction(Insn, 2, 5);
-  return decodeCLUIImmOperand(Inst, Imm, Address, Decoder);
-}
-
-static DecodeStatus
-decodeRVCInstrRdRs1UImmLog2XLenNonZero(MCInst &Inst, uint32_t Insn,
-                                       uint64_t Address,
-                                       const MCDisassembler *Decoder) {
-  Inst.addOperand(MCOperand::createReg(RISCV::X0));
-  Inst.addOperand(Inst.getOperand(0));
-
-  uint32_t UImm6 =
-      fieldFromInstruction(Insn, 12, 1) << 5 | fieldFromInstruction(Insn, 2, 5);
-  return decodeUImmLog2XLenNonZeroOperand(Inst, UImm6, Address, Decoder);
-}
-
-static DecodeStatus decodeRVCInstrRdRs2(MCInst &Inst, uint32_t Insn,
-                                        uint64_t Address,
-                                        const MCDisassembler *Decoder) {
-  DecodeStatus S = MCDisassembler::Success;
-  uint32_t Rd = fieldFromInstruction(Insn, 7, 5);
-  uint32_t Rs2 = fieldFromInstruction(Insn, 2, 5);
-  if (!Check(S, DecodeGPRRegisterClass(Inst, Rd, Address, Decoder)))
-    return MCDisassembler::Fail;
-  if (!Check(S, DecodeGPRRegisterClass(Inst, Rs2, Address, Decoder)))
-    return MCDisassembler::Fail;
-  return S;
-}
-
-static DecodeStatus decodeRVCInstrRdRs1Rs2(MCInst &Inst, uint32_t Insn,
-                                           uint64_t Address,
-                                           const MCDisassembler *Decoder) {
-  DecodeStatus S = MCDisassembler::Success;
-  uint32_t Rd = fieldFromInstruction(Insn, 7, 5);
-  uint32_t Rs2 = fieldFromInstruction(Insn, 2, 5);
-  if (!Check(S, DecodeGPRRegisterClass(Inst, Rd, Address, Decoder)))
-    return MCDisassembler::Fail;
-  Inst.addOperand(Inst.getOperand(0));
-  if (!Check(S, DecodeGPRRegisterClass(Inst, Rs2, Address, Decoder)))
-    return MCDisassembler::Fail;
-  return S;
-}
-
 static DecodeStatus decodeXTHeadMemPair(MCInst &Inst, uint32_t Insn,
                                         uint64_t Address,
                                         const MCDisassembler *Decoder) {
diff --git a/llvm/lib/Target/RISCV/MCTargetDesc/RISCVAsmBackend.cpp b/llvm/lib/Target/RISCV/MCTargetDesc/RISCVAsmBackend.cpp
index 2c37c3b..82e3b5c 100644
--- a/llvm/lib/Target/RISCV/MCTargetDesc/RISCVAsmBackend.cpp
+++ b/llvm/lib/Target/RISCV/MCTargetDesc/RISCVAsmBackend.cpp
@@ -320,6 +320,7 @@ bool RISCVAsmBackend::relaxAlign(MCFragment &F, unsigned &Size) {
   MCFixup Fixup =
       MCFixup::create(0, Expr, FirstLiteralRelocationKind + ELF::R_RISCV_ALIGN);
   F.setVarFixups({Fixup});
+  F.setLinkerRelaxable();
   F.getParent()->setLinkerRelaxable();
   return true;
 }
diff --git a/llvm/lib/Target/RISCV/MCTargetDesc/RISCVBaseInfo.h b/llvm/lib/Target/RISCV/MCTargetDesc/RISCVBaseInfo.h
index 7ad5d5f..bddea43 100644
--- a/llvm/lib/Target/RISCV/MCTargetDesc/RISCVBaseInfo.h
+++ b/llvm/lib/Target/RISCV/MCTargetDesc/RISCVBaseInfo.h
@@ -330,7 +330,6 @@ enum OperandType : unsigned {
   OPERAND_UIMM32,
   OPERAND_UIMM48,
   OPERAND_UIMM64,
-  OPERAND_ZERO,
   OPERAND_THREE,
   OPERAND_FOUR,
   OPERAND_SIMM5,
diff --git a/llvm/lib/Target/RISCV/RISCVAsmPrinter.cpp b/llvm/lib/Target/RISCV/RISCVAsmPrinter.cpp
index d4f5d8f..2f32e2a 100644
--- a/llvm/lib/Target/RISCV/RISCVAsmPrinter.cpp
+++ b/llvm/lib/Target/RISCV/RISCVAsmPrinter.cpp
@@ -293,7 +293,7 @@ void RISCVAsmPrinter::emitNTLHint(const MachineInstr *MI) {
 
   MCInst Hint;
   if (STI->hasStdExtZca())
-    Hint.setOpcode(RISCV::C_ADD_HINT);
+    Hint.setOpcode(RISCV::C_ADD);
   else
     Hint.setOpcode(RISCV::ADD);
 
diff --git a/llvm/lib/Target/RISCV/RISCVFrameLowering.cpp b/llvm/lib/Target/RISCV/RISCVFrameLowering.cpp
index b1ab76a..9fc0d81 100644
--- a/llvm/lib/Target/RISCV/RISCVFrameLowering.cpp
+++ b/llvm/lib/Target/RISCV/RISCVFrameLowering.cpp
@@ -1581,7 +1581,8 @@ void RISCVFrameLowering::determineCalleeSaves(MachineFunction &MF,
     // Set the register and all its subregisters.
     if (!MRI.def_empty(CSReg) || MRI.getUsedPhysRegsMask().test(CSReg)) {
       SavedRegs.set(CSReg);
-      llvm::for_each(SubRegs, [&](unsigned Reg) { return SavedRegs.set(Reg); });
+      for (unsigned Reg : SubRegs)
+        SavedRegs.set(Reg);
     }
 
     // Combine to super register if all of its subregisters are marked.
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
index 64f9e3e..085064e 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfo.cpp
@@ -2859,9 +2859,6 @@ bool RISCVInstrInfo::verifyInstruction(const MachineInstr &MI,
         case RISCVOp::OPERAND_UIMM16_NONZERO:
           Ok = isUInt<16>(Imm) && (Imm != 0);
           break;
-        case RISCVOp::OPERAND_ZERO:
-          Ok = Imm == 0;
-          break;
         case RISCVOp::OPERAND_THREE:
           Ok = Imm == 3;
           break;
diff --git a/llvm/lib/Target/RISCV/RISCVInstrInfoC.td b/llvm/lib/Target/RISCV/RISCVInstrInfoC.td
index 8252a9b..c5551fb 100644
--- a/llvm/lib/Target/RISCV/RISCVInstrInfoC.td
+++ b/llvm/lib/Target/RISCV/RISCVInstrInfoC.td
@@ -57,12 +57,6 @@ def simm6nonzero : RISCVOp,
   }];
 }
 
-def immzero : RISCVOp,
-              ImmLeaf<XLenVT, [{return (Imm == 0);}]> {
-  let ParserMatchClass = ImmZeroAsmOperand;
-  let OperandType = "OPERAND_ZERO";
-}
-
 def CLUIImmAsmOperand : AsmOperandClass {
   let Name = "CLUIImm";
   let RenderMethod = "addImmOperands";
@@ -272,7 +266,7 @@ class Bcz<bits<3> funct3, string OpcodeStr>
 let hasSideEffects = 0, mayLoad = 0, mayStore = 0 in
 class Shift_right<bits<2> funct2, string OpcodeStr>
     : RVInst16CB<0b100, 0b01, (outs GPRC:$rd),
-                 (ins GPRC:$rs1, uimmlog2xlennonzero:$imm),
+                 (ins GPRC:$rs1, uimmlog2xlen:$imm),
                  OpcodeStr, "$rs1, $imm"> {
   let Constraints = "$rs1 = $rd";
   let Inst{12} = imm{5};
@@ -402,17 +396,19 @@ def C_NOP : RVInst16CI<0b000, 0b01, (outs), (ins), "c.nop", "">,
 
 let hasSideEffects = 0, mayLoad = 0, mayStore = 0 in
 def C_ADDI : RVInst16CI<0b000, 0b01, (outs GPRNoX0:$rd_wb),
-                        (ins GPRNoX0:$rd, simm6nonzero:$imm),
+                        (ins GPRNoX0:$rd, simm6:$imm),
                         "c.addi", "$rd, $imm">,
              Sched<[WriteIALU, ReadIALU]> {
   let Constraints = "$rd = $rd_wb";
 }
 
-// Alternate syntax for c.nop. Converted to C_NOP by the assembler.
+// Alternate syntax for c.nop. Converted to C_NOP/C_NOP_HINT by the assembler.
 let hasSideEffects = 0, mayLoad = 0, mayStore = 0, isCodeGenOnly = 0,
     isAsmParserOnly = 1 in
-def PseudoC_ADDI_NOP : Pseudo<(outs GPRX0:$rd), (ins GPRX0:$rs1, immzero:$imm),
-                              [], "c.addi", "$rd, $imm">;
+def PseudoC_ADDI_NOP : Pseudo<(outs GPRX0:$rd), (ins GPRX0:$rs1, simm6:$imm),
+                              [], "c.addi", "$rd, $imm"> {
+  let Constraints = "$rs1 = $rd";
+}
 
 let hasSideEffects = 0, mayLoad = 0, mayStore = 0, isCall = 1,
     DecoderNamespace = "RV32Only", Defs = [X1],
@@ -430,7 +426,7 @@ def C_ADDIW : RVInst16CI<0b001, 0b01, (outs GPRNoX0:$rd_wb),
 }
 
 let hasSideEffects = 0, mayLoad = 0, mayStore = 0 in
-def C_LI : RVInst16CI<0b010, 0b01, (outs GPRNoX0:$rd), (ins simm6:$imm),
+def C_LI : RVInst16CI<0b010, 0b01, (outs GPR:$rd), (ins simm6:$imm),
                       "c.li", "$rd, $imm">,
            Sched<[WriteIALU]>;
 
@@ -449,7 +445,7 @@ def C_ADDI16SP : RVInst16CI<0b011, 0b01, (outs SP:$rd_wb),
 }
 
 let hasSideEffects = 0, mayLoad = 0, mayStore = 0 in
-def C_LUI : RVInst16CI<0b011, 0b01, (outs GPRNoX0X2:$rd),
+def C_LUI : RVInst16CI<0b011, 0b01, (outs GPRNoX2:$rd),
                        (ins c_lui_imm:$imm),
                        "c.lui", "$rd, $imm">,
             Sched<[WriteIALU]>;
@@ -497,8 +493,8 @@ def C_BEQZ : Bcz<0b110, "c.beqz">, Sched<[WriteJmp, ReadJmp]>;
 def C_BNEZ : Bcz<0b111, "c.bnez">, Sched<[WriteJmp, ReadJmp]>;
 
 let hasSideEffects = 0, mayLoad = 0, mayStore = 0 in
-def C_SLLI : RVInst16CI<0b000, 0b10, (outs GPRNoX0:$rd_wb),
-                        (ins GPRNoX0:$rd, uimmlog2xlennonzero:$imm),
+def C_SLLI : RVInst16CI<0b000, 0b10, (outs GPR:$rd_wb),
+                        (ins GPR:$rd, uimmlog2xlen:$imm),
                         "c.slli", "$rd, $imm">,
              Sched<[WriteShiftImm, ReadShiftImm]> {
   let Constraints = "$rd = $rd_wb";
@@ -544,7 +540,7 @@ def C_JR : RVInst16CR<0b1000, 0b10, (outs), (ins GPRNoX0:$rs1),
 
 let hasSideEffects = 0, mayLoad = 0, mayStore = 0, isMoveReg = 1,
     isAsCheapAsAMove = 1 in
-def C_MV : RVInst16CR<0b1000, 0b10, (outs GPRNoX0:$rs1), (ins GPRNoX0:$rs2),
+def C_MV : RVInst16CR<0b1000, 0b10, (outs GPR:$rs1), (ins GPRNoX0:$rs2),
                       "c.mv", "$rs1, $rs2">,
            Sched<[WriteIALU, ReadIALU]>;
 
@@ -557,8 +553,8 @@ def C_JALR : RVInst16CR<0b1001, 0b10, (outs), (ins GPRNoX0:$rs1),
                         "c.jalr", "$rs1">, Sched<[WriteJalr, ReadJalr]>;
 
 let hasSideEffects = 0, mayLoad = 0, mayStore = 0 in
-def C_ADD : RVInst16CR<0b1001, 0b10, (outs GPRNoX0:$rd),
-                       (ins GPRNoX0:$rs1, GPRNoX0:$rs2),
+def C_ADD : RVInst16CR<0b1001, 0b10, (outs GPR:$rd),
+                       (ins GPR:$rs1, GPRNoX0:$rs2),
                        "c.add", "$rs1, $rs2">,
             Sched<[WriteIALU, ReadIALU, ReadIALU]> {
   let Constraints = "$rs1 = $rd";
@@ -616,81 +612,6 @@ def C_NOP_HINT : RVInst16CI<0b000, 0b01, (outs), (ins simm6nonzero:$imm),
   let rd = 0;
 }
 
-def C_ADDI_HINT_IMM_ZERO : RVInst16CI<0b000, 0b01, (outs GPRNoX0:$rd_wb),
-                                      (ins GPRNoX0:$rd, immzero:$imm),
-                                      "c.addi", "$rd, $imm">,
-                           Sched<[WriteIALU, ReadIALU]> {
-  let Constraints = "$rd = $rd_wb";
-  let imm = 0;
-  let DecoderMethod = "decodeRVCInstrRdRs1ImmZero";
-}
-
-def C_LI_HINT : RVInst16CI<0b010, 0b01, (outs GPRX0:$rd), (ins simm6:$imm),
-                           "c.li", "$rd, $imm">,
-                Sched<[WriteIALU]> {
-  let Inst{11-7} = 0;
-  let DecoderMethod = "decodeRVCInstrRdSImm6";
-}
-
-def C_LUI_HINT : RVInst16CI<0b011, 0b01, (outs GPRX0:$rd),
-                            (ins c_lui_imm:$imm),
-                            "c.lui", "$rd, $imm">,
-                 Sched<[WriteIALU]> {
-  let Inst{11-7} = 0;
-  let DecoderMethod = "decodeRVCInstrRdCLUIImm";
-}
-
-def C_MV_HINT : RVInst16CR<0b1000, 0b10, (outs GPRX0:$rs1), (ins GPRNoX0:$rs2),
-                           "c.mv", "$rs1, $rs2">, Sched<[WriteIALU, ReadIALU]> {
-  let Inst{11-7} = 0;
-  let DecoderMethod = "decodeRVCInstrRdRs2";
-}
-
-def C_ADD_HINT : RVInst16CR<0b1001, 0b10, (outs GPRX0:$rd),
-                            (ins GPRX0:$rs1, GPRNoX0:$rs2),
-                            "c.add", "$rs1, $rs2">,
-                 Sched<[WriteIALU, ReadIALU, ReadIALU]> {
-  let Constraints = "$rs1 = $rd";
-  let Inst{11-7} = 0;
-  let DecoderMethod = "decodeRVCInstrRdRs1Rs2";
-}
-
-def C_SLLI_HINT : RVInst16CI<0b000, 0b10, (outs GPRX0:$rd_wb),
-                             (ins GPRX0:$rd, uimmlog2xlennonzero:$imm),
-                             "c.slli", "$rd, $imm">,
-                  Sched<[WriteShiftImm, ReadShiftImm]> {
-  let Constraints = "$rd = $rd_wb";
-  let Inst{11-7} = 0;
-  let DecoderMethod = "decodeRVCInstrRdRs1UImmLog2XLenNonZero";
-}
-
-def C_SLLI64_HINT : RVInst16CI<0b000, 0b10, (outs GPR:$rd_wb), (ins GPR:$rd),
-                               "c.slli64", "$rd">,
-                    Sched<[WriteShiftImm, ReadShiftImm]> {
-  let Constraints = "$rd = $rd_wb";
-  let imm = 0;
-}
-
-def C_SRLI64_HINT : RVInst16CB<0b100, 0b01, (outs GPRC:$rd),
-                               (ins GPRC:$rs1),
-                               "c.srli64", "$rs1">,
-                    Sched<[WriteShiftImm, ReadShiftImm]> {
-  let Constraints = "$rs1 = $rd";
-  let Inst{6-2} = 0;
-  let Inst{11-10} = 0b00;
-  let Inst{12} = 0;
-}
-
-def C_SRAI64_HINT : RVInst16CB<0b100, 0b01, (outs GPRC:$rd),
-                               (ins GPRC:$rs1),
-                               "c.srai64", "$rs1">,
-                    Sched<[WriteShiftImm, ReadShiftImm]> {
-  let Constraints = "$rs1 = $rd";
-  let Inst{6-2} = 0;
-  let Inst{11-10} = 0b01;
-  let Inst{12} = 0;
-}
-
 } // Predicates = [HasStdExtZca], hasSideEffects = 0, mayLoad = 0,
   // mayStore = 0
 
@@ -699,15 +620,17 @@ def C_SRAI64_HINT : RVInst16CB<0b100, 0b01, (outs GPRC:$rd),
 //===----------------------------------------------------------------------===//
 
 let Predicates = [HasStdExtZca] in {
-// Just a different syntax for the c.nop hint: c.addi x0, simm6 vs c.nop simm6.
-def : InstAlias<"c.addi x0, $imm", (C_NOP_HINT simm6nonzero:$imm), 0>;
+// Legacy aliases.
+def : InstAlias<"c.slli64 $rd", (C_SLLI GPR:$rd, 0), 0>;
+def : InstAlias<"c.srli64 $rs1", (C_SRLI GPRC:$rs1, 0), 0>;
+def : InstAlias<"c.srai64 $rs1", (C_SRAI GPRC:$rs1, 0), 0>;
 }
 
 let Predicates = [HasStdExtC, HasStdExtZihintntl] in {
-def : InstAlias<"c.ntl.p1", (C_ADD_HINT X0, X2)>;
-def : InstAlias<"c.ntl.pall", (C_ADD_HINT X0, X3)>;
-def : InstAlias<"c.ntl.s1", (C_ADD_HINT X0, X4)>;
-def : InstAlias<"c.ntl.all", (C_ADD_HINT X0, X5)>;
+def : InstAlias<"c.ntl.p1", (C_ADD X0, X2)>;
+def : InstAlias<"c.ntl.pall", (C_ADD X0, X3)>;
+def : InstAlias<"c.ntl.s1", (C_ADD X0, X4)>;
+def : InstAlias<"c.ntl.all", (C_ADD X0, X5)>;
 } // Predicates = [HasStdExtC, HasStdExtZihintntl]
 
 let EmitPriority = 0 in {
diff --git a/llvm/lib/Target/RISCV/RISCVRegisterInfo.td b/llvm/lib/Target/RISCV/RISCVRegisterInfo.td
index e87f452..ccb39e8 100644
--- a/llvm/lib/Target/RISCV/RISCVRegisterInfo.td
+++ b/llvm/lib/Target/RISCV/RISCVRegisterInfo.td
@@ -268,6 +268,11 @@ def GPRNoX0 : GPRRegisterClass<(sub GPR, X0)> {
   let DiagnosticString = "register must be a GPR excluding zero (x0)";
 }
 
+def GPRNoX2 : GPRRegisterClass<(sub GPR, X2)> {
+  let DiagnosticType = "InvalidRegClassGPRNoX2";
+  let DiagnosticString = "register must be a GPR excluding sp (x2)";
+}
+
 def GPRNoX0X2 : GPRRegisterClass<(sub GPR, X0, X2)> {
   let DiagnosticType = "InvalidRegClassGPRNoX0X2";
   let DiagnosticString = "register must be a GPR excluding zero (x0) and sp (x2)";
diff --git a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
index 1624f12a..fd634b5 100644
--- a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
+++ b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.cpp
@@ -979,12 +979,10 @@ InstructionCost RISCVTTIImpl::getInterleavedMemoryOpCost(
     Align Alignment, unsigned AddressSpace, TTI::TargetCostKind CostKind,
     bool UseMaskForCond, bool UseMaskForGaps) const {
 
-  // The interleaved memory access pass will lower (de)interleave ops combined
-  // with an adjacent appropriate memory to vlseg/vsseg intrinsics. vlseg/vsseg
-  // only support masking per-iteration (i.e. condition), not per-segment (i.e.
-  // gap).
-  // TODO: Support masked interleaved access for fixed length vector.
-  if ((isa<ScalableVectorType>(VecTy) || !UseMaskForCond) && !UseMaskForGaps &&
+  // The interleaved memory access pass will lower interleaved memory ops (i.e
+  // a load and store followed by a specific shuffle) to vlseg/vsseg
+  // intrinsics.
+  if (!UseMaskForCond && !UseMaskForGaps &&
       Factor <= TLI->getMaxSupportedInterleaveFactor()) {
     auto *VTy = cast<VectorType>(VecTy);
     std::pair<InstructionCost, MVT> LT = getTypeLegalizationCost(VTy);
diff --git a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.h b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.h
index 05d504c..d62d99c 100644
--- a/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.h
+++ b/llvm/lib/Target/RISCV/RISCVTargetTransformInfo.h
@@ -398,10 +398,6 @@ public:
 
   bool enableInterleavedAccessVectorization() const override { return true; }
 
-  bool enableMaskedInterleavedAccessVectorization() const override {
-    return ST->hasVInstructions();
-  }
-
   unsigned getMinTripCountTailFoldingThreshold() const override;
 
   enum RISCVRegisterClass { GPRRC, FPRRC, VRRC };
diff --git a/llvm/lib/Target/SystemZ/MCTargetDesc/SystemZHLASMAsmStreamer.cpp b/llvm/lib/Target/SystemZ/MCTargetDesc/SystemZHLASMAsmStreamer.cpp
index 3ef6030..72bb372 100644
--- a/llvm/lib/Target/SystemZ/MCTargetDesc/SystemZHLASMAsmStreamer.cpp
+++ b/llvm/lib/Target/SystemZ/MCTargetDesc/SystemZHLASMAsmStreamer.cpp
@@ -69,8 +69,8 @@ void SystemZHLASMAsmStreamer::EmitEOL() {
 
 void SystemZHLASMAsmStreamer::changeSection(MCSection *Section,
                                             uint32_t Subsection) {
-  Section->printSwitchToSection(*MAI, getContext().getTargetTriple(), OS,
-                                Subsection);
+  MAI->printSwitchToSection(*Section, Subsection,
+                            getContext().getTargetTriple(), OS);
   MCStreamer::changeSection(Section, Subsection);
 }
 
diff --git a/llvm/lib/Target/WebAssembly/AsmParser/WebAssemblyAsmParser.cpp b/llvm/lib/Target/WebAssembly/AsmParser/WebAssemblyAsmParser.cpp
index 19c9e9c..6ae69a4 100644
--- a/llvm/lib/Target/WebAssembly/AsmParser/WebAssemblyAsmParser.cpp
+++ b/llvm/lib/Target/WebAssembly/AsmParser/WebAssemblyAsmParser.cpp
@@ -900,7 +900,8 @@ public:
 
   bool checkDataSection() {
     if (CurrentState != DataSection) {
-      auto *WS = cast<MCSectionWasm>(getStreamer().getCurrentSectionOnly());
+      auto *WS = static_cast<const MCSectionWasm *>(
+          getStreamer().getCurrentSectionOnly());
       if (WS && WS->isText())
         return error("data directive must occur in a data segment: ",
                      Lexer.getTok());
@@ -1218,7 +1219,8 @@ public:
 
   void doBeforeLabelEmit(MCSymbol *Symbol, SMLoc IDLoc) override {
     // Code below only applies to labels in text sections.
-    auto *CWS = cast<MCSectionWasm>(getStreamer().getCurrentSectionOnly());
+    auto *CWS = static_cast<const MCSectionWasm *>(
+        getStreamer().getCurrentSectionOnly());
     if (!CWS->isText())
       return;
 
diff --git a/llvm/lib/Target/WebAssembly/WebAssembly.td b/llvm/lib/Target/WebAssembly/WebAssembly.td
index 13603f8..a606209 100644
--- a/llvm/lib/Target/WebAssembly/WebAssembly.td
+++ b/llvm/lib/Target/WebAssembly/WebAssembly.td
@@ -71,6 +71,7 @@ def FeatureReferenceTypes :
       SubtargetFeature<"reference-types", "HasReferenceTypes", "true",
                        "Enable reference types">;
 
+def FeatureGC : SubtargetFeature<"gc", "HasGC", "true", "Enable wasm gc">;
 def FeatureRelaxedSIMD :
       SubtargetFeature<"relaxed-simd", "SIMDLevel", "RelaxedSIMD",
                        "Enable relaxed-simd instructions">;
@@ -136,13 +137,13 @@ def : ProcessorModel<"lime1", NoSchedModel,
 
 // Latest and greatest experimental version of WebAssembly. Bugs included!
 def : ProcessorModel<"bleeding-edge", NoSchedModel,
-                      [FeatureAtomics, FeatureBulkMemory, FeatureBulkMemoryOpt,
-                       FeatureCallIndirectOverlong, FeatureExceptionHandling,
-                       FeatureExtendedConst, FeatureFP16, FeatureMultiMemory,
-                       FeatureMultivalue, FeatureMutableGlobals,
-                       FeatureNontrappingFPToInt, FeatureRelaxedSIMD,
-                       FeatureReferenceTypes, FeatureSIMD128, FeatureSignExt,
-                       FeatureTailCall]>;
+                     [FeatureAtomics, FeatureBulkMemory, FeatureBulkMemoryOpt,
+                      FeatureCallIndirectOverlong, FeatureExceptionHandling,
+                      FeatureExtendedConst, FeatureFP16, FeatureMultiMemory,
+                      FeatureMultivalue, FeatureMutableGlobals,
+                      FeatureNontrappingFPToInt, FeatureRelaxedSIMD,
+                      FeatureReferenceTypes, FeatureGC, FeatureSIMD128,
+                      FeatureSignExt, FeatureTailCall]>;
 
 //===----------------------------------------------------------------------===//
 // Target Declaration
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp b/llvm/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp
index 11936a3..cd434f7 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyISelLowering.cpp
@@ -288,7 +288,7 @@ WebAssemblyTargetLowering::WebAssemblyTargetLowering(
 
     // Expand float operations supported for scalars but not SIMD
     for (auto Op : {ISD::FCOPYSIGN, ISD::FLOG, ISD::FLOG2, ISD::FLOG10,
-                    ISD::FEXP, ISD::FEXP2})
+                    ISD::FEXP, ISD::FEXP2, ISD::FEXP10})
       for (auto T : {MVT::v4f32, MVT::v2f64})
         setOperationAction(Op, T, Expand);
 
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyInstrInfo.td b/llvm/lib/Target/WebAssembly/WebAssemblyInstrInfo.td
index b5e723e..2b632fd 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyInstrInfo.td
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyInstrInfo.td
@@ -76,6 +76,9 @@ def HasReferenceTypes :
     Predicate<"Subtarget->hasReferenceTypes()">,
     AssemblerPredicate<(all_of FeatureReferenceTypes), "reference-types">;
 
+def HasGC : Predicate<"Subtarget->hasGC()">,
+            AssemblerPredicate<(all_of FeatureGC), "gc">;
+
 def HasRelaxedSIMD :
     Predicate<"Subtarget->hasRelaxedSIMD()">,
     AssemblerPredicate<(all_of FeatureRelaxedSIMD), "relaxed-simd">;
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblyInstrRef.td b/llvm/lib/Target/WebAssembly/WebAssemblyInstrRef.td
index 40b87a0..fc82e5b 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblyInstrRef.td
+++ b/llvm/lib/Target/WebAssembly/WebAssemblyInstrRef.td
@@ -36,13 +36,10 @@ multiclass REF_I<WebAssemblyRegClass rc, ValueType vt, string ht> {
         Requires<[HasReferenceTypes]>;
 }
 
-defm REF_TEST_FUNCREF :
-  I<(outs I32: $res),
-    (ins TypeIndex:$type, FUNCREF: $ref),
-    (outs),
-    (ins TypeIndex:$type),
-    [],
-    "ref.test\t$type, $ref", "ref.test $type", 0xfb14>;
+defm REF_TEST_FUNCREF : I<(outs I32:$res), (ins TypeIndex:$type, FUNCREF:$ref),
+                          (outs), (ins TypeIndex:$type), [],
+                          "ref.test\t$type, $ref", "ref.test $type", 0xfb14>,
+                        Requires<[HasGC]>;
 
 defm "" : REF_I<FUNCREF, funcref, "func">;
 defm "" : REF_I<EXTERNREF, externref, "extern">;
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblySubtarget.cpp b/llvm/lib/Target/WebAssembly/WebAssemblySubtarget.cpp
index 40ea48a..a3ce40f 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblySubtarget.cpp
+++ b/llvm/lib/Target/WebAssembly/WebAssemblySubtarget.cpp
@@ -43,6 +43,11 @@ WebAssemblySubtarget::initializeSubtargetDependencies(StringRef CPU,
     Bits.set(WebAssembly::FeatureBulkMemoryOpt);
   }
 
+  // gc implies reference-types
+  if (HasGC) {
+    HasReferenceTypes = true;
+  }
+
   // reference-types implies call-indirect-overlong
   if (HasReferenceTypes) {
     HasCallIndirectOverlong = true;
diff --git a/llvm/lib/Target/WebAssembly/WebAssemblySubtarget.h b/llvm/lib/Target/WebAssembly/WebAssemblySubtarget.h
index 591ce256..f814274 100644
--- a/llvm/lib/Target/WebAssembly/WebAssemblySubtarget.h
+++ b/llvm/lib/Target/WebAssembly/WebAssemblySubtarget.h
@@ -51,6 +51,7 @@ class WebAssemblySubtarget final : public WebAssemblyGenSubtargetInfo {
   bool HasMutableGlobals = false;
   bool HasNontrappingFPToInt = false;
   bool HasReferenceTypes = false;
+  bool HasGC = false;
   bool HasSignExt = false;
   bool HasTailCall = false;
   bool HasWideArithmetic = false;
@@ -107,6 +108,7 @@ public:
   bool hasMutableGlobals() const { return HasMutableGlobals; }
   bool hasNontrappingFPToInt() const { return HasNontrappingFPToInt; }
   bool hasReferenceTypes() const { return HasReferenceTypes; }
+  bool hasGC() const { return HasGC; }
   bool hasRelaxedSIMD() const { return SIMDLevel >= RelaxedSIMD; }
   bool hasSignExt() const { return HasSignExt; }
   bool hasSIMD128() const { return SIMDLevel >= SIMD128; }
diff --git a/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp b/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp
index 8213e51..d7671ed 100644
--- a/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp
+++ b/llvm/lib/Target/X86/AsmParser/X86AsmParser.cpp
@@ -4803,7 +4803,7 @@ bool X86AsmParser::parseDirectiveEven(SMLoc L) {
     getStreamer().initSections(false, getSTI());
     Section = getStreamer().getCurrentSectionOnly();
   }
-  if (Section->useCodeAlign())
+  if (getContext().getAsmInfo()->useCodeAlign(*Section))
     getStreamer().emitCodeAlignment(Align(2), &getSTI(), 0);
   else
     getStreamer().emitValueToAlignment(Align(2), 0, 1, 0);
diff --git a/llvm/lib/Target/X86/X86PassRegistry.def b/llvm/lib/Target/X86/X86PassRegistry.def
index 620526ff..3f2a433 100644
--- a/llvm/lib/Target/X86/X86PassRegistry.def
+++ b/llvm/lib/Target/X86/X86PassRegistry.def
@@ -12,8 +12,52 @@
 
 // NOTE: NO INCLUDE GUARD DESIRED!
 
+#ifndef DUMMY_FUNCTION_PASS
+#define DUMMY_FUNCTION_PASS(NAME, CREATE_PASS)
+#endif
+DUMMY_FUNCTION_PASS("lower-amx-intrinsics", X86LowerAMXIntrinsics(*this))
+DUMMY_FUNCTION_PASS("lower-amx-type", X86LowerAMXTypePass(*this))
+DUMMY_FUNCTION_PASS("x86-partial-reduction", X86PartialReduction())
+DUMMY_FUNCTION_PASS("x86-winehstate", WinEHStatePass())
+#undef DUMMY_FUNCTION_PASS
+
 #ifndef MACHINE_FUNCTION_PASS
 #define MACHINE_FUNCTION_PASS(NAME, CREATE_PASS)
 #endif
 MACHINE_FUNCTION_PASS("x86-isel", X86ISelDAGToDAGPass(*this))
 #undef MACHINE_FUNCTION_PASS
+
+#ifndef DUMMY_MACHINE_FUNCTION_PASS
+#define DUMMY_MACHINE_FUNCTION_PASS(NAME, PASS_NAME)
+#endif
+DUMMY_MACHINE_FUNCTION_PASS("x86-avoid-SFB", X86AvoidSFBPass())
+DUMMY_MACHINE_FUNCTION_PASS("x86-avoid-trailing-call", X86AvoidTrailingCallPass())
+DUMMY_MACHINE_FUNCTION_PASS("x86-cf-opt", X86CallFrameOptimization())
+DUMMY_MACHINE_FUNCTION_PASS("x86-cmov-conversion", X86CmovConverterPass())
+DUMMY_MACHINE_FUNCTION_PASS("x86-codege", FPS())
+DUMMY_MACHINE_FUNCTION_PASS("x86-compress-evex", CompressEVEXPass())
+DUMMY_MACHINE_FUNCTION_PASS("x86-domain-reassignment", X86DomainReassignment())
+DUMMY_MACHINE_FUNCTION_PASS("x86-dyn-alloca-expander", X86DynAllocaExpander())
+DUMMY_MACHINE_FUNCTION_PASS("x86-execution-domain-fix", X86ExecutionDomainFix())
+DUMMY_MACHINE_FUNCTION_PASS("fastpretileconfig", X86FastPreTileConfig())
+DUMMY_MACHINE_FUNCTION_PASS("fasttileconfig", X86FastTileConfig())
+DUMMY_MACHINE_FUNCTION_PASS("x86-fixup-LEAs", FixupLEAPass())
+DUMMY_MACHINE_FUNCTION_PASS("x86-fixup-bw-inst", FixupBWInstPass())
+DUMMY_MACHINE_FUNCTION_PASS("x86-fixup-inst-tuning", X86FixupInstTuningPass())
+DUMMY_MACHINE_FUNCTION_PASS("x86-fixup-setcc", X86FixupSetCCPass())
+DUMMY_MACHINE_FUNCTION_PASS("x86-fixup-vector-constants", X86FixupVectorConstantsPass())
+DUMMY_MACHINE_FUNCTION_PASS("x86-flags-copy-lowering", X86FlagsCopyLoweringPass())
+DUMMY_MACHINE_FUNCTION_PASS("x86-lower-tile-copy", X86LowerTileCopy())
+DUMMY_MACHINE_FUNCTION_PASS("x86-lvi-load", X86LoadValueInjectionLoadHardeningPass())
+DUMMY_MACHINE_FUNCTION_PASS("x86-lvi-ret", X86LoadValueInjectionRetHardeningPass())
+DUMMY_MACHINE_FUNCTION_PASS("x86-optimize-LEAs", X86OptimizeLEAPass())
+DUMMY_MACHINE_FUNCTION_PASS("x86-pseudo", X86ExpandPseudo())
+DUMMY_MACHINE_FUNCTION_PASS("x86-return-thunks", X86ReturnThunks())
+DUMMY_MACHINE_FUNCTION_PASS("x86-seses", X86SpeculativeExecutionSideEffectSuppression())
+DUMMY_MACHINE_FUNCTION_PASS("x86-slh", X86SpeculativeLoadHardeningPass())
+DUMMY_MACHINE_FUNCTION_PASS("x86-suppress-apx-for-relocation", X86SuppressAPXForRelocationPass())
+DUMMY_MACHINE_FUNCTION_PASS("tile-pre-config", X86PreTileConfig())
+DUMMY_MACHINE_FUNCTION_PASS("tileconfig", X86TileConfig())
+DUMMY_MACHINE_FUNCTION_PASS("x86-wineh-unwindv2", X86WinEHUnwindV2())
+DUMMY_MACHINE_FUNCTION_PASS("x86argumentstackrebase", X86ArgumentStackSlotPass())
+#undef DUMMY_MACHINE_FUNCTION_PASS
diff --git a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
index 37a7b37..90791fc 100644
--- a/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
+++ b/llvm/lib/Target/X86/X86TargetTransformInfo.cpp
@@ -1838,14 +1838,15 @@ InstructionCost X86TTIImpl::getShuffleCost(TTI::ShuffleKind Kind,
         return LT.first * *KindCost;
 
   static const CostKindTblEntry AVX512BWShuffleTbl[] = {
-    { TTI::SK_Broadcast, MVT::v32i16, { 1, 1, 1, 1 } }, // vpbroadcastw
-    { TTI::SK_Broadcast, MVT::v32f16, { 1, 1, 1, 1 } }, // vpbroadcastw
-    { TTI::SK_Broadcast, MVT::v64i8,  { 1, 1, 1, 1 } }, // vpbroadcastb
+    { TTI::SK_Broadcast, MVT::v32i16, { 1, 3, 1, 1 } }, // vpbroadcastw
+    { TTI::SK_Broadcast, MVT::v32f16, { 1, 3, 1, 1 } }, // vpbroadcastw
+    { TTI::SK_Broadcast, MVT::v64i8,  { 1, 3, 1, 1 } }, // vpbroadcastb
 
-    { TTI::SK_Reverse, MVT::v32i16,   { 2, 2, 2, 2 } }, // vpermw
-    { TTI::SK_Reverse, MVT::v32f16,   { 2, 2, 2, 2 } }, // vpermw
+    { TTI::SK_Reverse, MVT::v32i16,   { 2, 6, 2, 4 } }, // vpermw
+    { TTI::SK_Reverse, MVT::v32f16,   { 2, 6, 2, 4 } }, // vpermw
     { TTI::SK_Reverse, MVT::v16i16,   { 2, 2, 2, 2 } }, // vpermw
-    { TTI::SK_Reverse, MVT::v64i8,    { 2, 2, 2, 2 } }, // pshufb + vshufi64x2
+    { TTI::SK_Reverse, MVT::v16f16,   { 2, 2, 2, 2 } }, // vpermw
+    { TTI::SK_Reverse, MVT::v64i8,    { 2, 9, 2, 3 } }, // pshufb + vshufi64x2
 
     { TTI::SK_PermuteSingleSrc, MVT::v32i16, { 2, 2, 2, 2 } }, // vpermw
     { TTI::SK_PermuteSingleSrc, MVT::v32f16, { 2, 2, 2, 2 } }, // vpermw
@@ -1874,18 +1875,25 @@ InstructionCost X86TTIImpl::getShuffleCost(TTI::ShuffleKind Kind,
         return LT.first * *KindCost;
 
   static const CostKindTblEntry AVX512ShuffleTbl[] = {
-      {TTI::SK_Broadcast, MVT::v8f64,  { 1, 1, 1, 1 } }, // vbroadcastsd
-      {TTI::SK_Broadcast, MVT::v16f32, { 1, 1, 1, 1 } }, // vbroadcastss
-      {TTI::SK_Broadcast, MVT::v8i64,  { 1, 1, 1, 1 } }, // vpbroadcastq
-      {TTI::SK_Broadcast, MVT::v16i32, { 1, 1, 1, 1 } }, // vpbroadcastd
-      {TTI::SK_Broadcast, MVT::v32i16, { 1, 1, 1, 1 } }, // vpbroadcastw
-      {TTI::SK_Broadcast, MVT::v32f16, { 1, 1, 1, 1 } }, // vpbroadcastw
-      {TTI::SK_Broadcast, MVT::v64i8,  { 1, 1, 1, 1 } }, // vpbroadcastb
-
-      {TTI::SK_Reverse, MVT::v8f64,  { 1, 3, 1, 1 } }, // vpermpd
-      {TTI::SK_Reverse, MVT::v16f32, { 1, 3, 1, 1 } }, // vpermps
-      {TTI::SK_Reverse, MVT::v8i64,  { 1, 3, 1, 1 } }, // vpermq
-      {TTI::SK_Reverse, MVT::v16i32, { 1, 3, 1, 1 } }, // vpermd
+      {TTI::SK_Broadcast, MVT::v8f64,  { 1, 3, 1, 1 } }, // vbroadcastsd
+      {TTI::SK_Broadcast, MVT::v4f64,  { 1, 3, 1, 1 } }, // vbroadcastsd
+      {TTI::SK_Broadcast, MVT::v16f32, { 1, 3, 1, 1 } }, // vbroadcastss
+      {TTI::SK_Broadcast, MVT::v8f32,  { 1, 3, 1, 1 } }, // vbroadcastss
+      {TTI::SK_Broadcast, MVT::v8i64,  { 1, 3, 1, 1 } }, // vpbroadcastq
+      {TTI::SK_Broadcast, MVT::v4i64,  { 1, 3, 1, 1 } }, // vpbroadcastq
+      {TTI::SK_Broadcast, MVT::v16i32, { 1, 3, 1, 1 } }, // vpbroadcastd
+      {TTI::SK_Broadcast, MVT::v8i32,  { 1, 3, 1, 1 } }, // vpbroadcastd
+      {TTI::SK_Broadcast, MVT::v32i16, { 1, 3, 1, 1 } }, // vpbroadcastw
+      {TTI::SK_Broadcast, MVT::v16i16, { 1, 3, 1, 1 } }, // vpbroadcastw
+      {TTI::SK_Broadcast, MVT::v32f16, { 1, 3, 1, 1 } }, // vpbroadcastw
+      {TTI::SK_Broadcast, MVT::v16f16, { 1, 3, 1, 1 } }, // vpbroadcastw
+      {TTI::SK_Broadcast, MVT::v64i8,  { 1, 3, 1, 1 } }, // vpbroadcastb
+      {TTI::SK_Broadcast, MVT::v32i8,  { 1, 3, 1, 1 }},  // vpbroadcastb
+
+      {TTI::SK_Reverse, MVT::v8f64,  { 1, 5, 2, 3 } }, // vpermpd
+      {TTI::SK_Reverse, MVT::v16f32, { 1, 3, 2, 3 } }, // vpermps
+      {TTI::SK_Reverse, MVT::v8i64,  { 1, 5, 2, 3 } }, // vpermq
+      {TTI::SK_Reverse, MVT::v16i32, { 1, 3, 2, 3 } }, // vpermd
       {TTI::SK_Reverse, MVT::v32i16, { 7, 7, 7, 7 } }, // per mca
       {TTI::SK_Reverse, MVT::v32f16, { 7, 7, 7, 7 } }, // per mca
       {TTI::SK_Reverse, MVT::v64i8,  { 7, 7, 7, 7 } }, // per mca
@@ -1973,21 +1981,24 @@ InstructionCost X86TTIImpl::getShuffleCost(TTI::ShuffleKind Kind,
         return LT.first * *KindCost;
 
   static const CostKindTblEntry AVX2ShuffleTbl[] = {
-    { TTI::SK_Broadcast, MVT::v4f64,  { 1, 1, 1, 1 } }, // vbroadcastpd
-    { TTI::SK_Broadcast, MVT::v8f32,  { 1, 1, 1, 1 } }, // vbroadcastps
-    { TTI::SK_Broadcast, MVT::v4i64,  { 1, 1, 1, 1 } }, // vpbroadcastq
-    { TTI::SK_Broadcast, MVT::v8i32,  { 1, 1, 1, 1 } }, // vpbroadcastd
-    { TTI::SK_Broadcast, MVT::v16i16, { 1, 1, 1, 1 } }, // vpbroadcastw
-    { TTI::SK_Broadcast, MVT::v16f16, { 1, 1, 1, 1 } }, // vpbroadcastw
-    { TTI::SK_Broadcast, MVT::v32i8,  { 1, 1, 1, 1 } }, // vpbroadcastb
-
-    { TTI::SK_Reverse, MVT::v4f64,    { 1, 1, 1, 1 } }, // vpermpd
-    { TTI::SK_Reverse, MVT::v8f32,    { 1, 1, 1, 1 } }, // vpermps
-    { TTI::SK_Reverse, MVT::v4i64,    { 1, 1, 1, 1 } }, // vpermq
-    { TTI::SK_Reverse, MVT::v8i32,    { 1, 1, 1, 1 } }, // vpermd
-    { TTI::SK_Reverse, MVT::v16i16,   { 2, 2, 2, 2 } }, // vperm2i128 + pshufb
-    { TTI::SK_Reverse, MVT::v16f16,   { 2, 2, 2, 2 } }, // vperm2i128 + pshufb
-    { TTI::SK_Reverse, MVT::v32i8,    { 2, 2, 2, 2 } }, // vperm2i128 + pshufb
+    { TTI::SK_Broadcast, MVT::v4f64,  { 1, 3, 1, 2 } }, // vbroadcastpd
+    { TTI::SK_Broadcast, MVT::v8f32,  { 1, 3, 1, 2 } }, // vbroadcastps
+    { TTI::SK_Broadcast, MVT::v4i64,  { 1, 3, 1, 2 } }, // vpbroadcastq
+    { TTI::SK_Broadcast, MVT::v8i32,  { 1, 3, 1, 2 } }, // vpbroadcastd
+    { TTI::SK_Broadcast, MVT::v16i16, { 1, 3, 1, 2 } }, // vpbroadcastw
+    { TTI::SK_Broadcast, MVT::v8i16,  { 1, 3, 1, 1 } }, // vpbroadcastw
+    { TTI::SK_Broadcast, MVT::v16f16, { 1, 3, 1, 2 } }, // vpbroadcastw
+    { TTI::SK_Broadcast, MVT::v8f16,  { 1, 3, 1, 1 } }, // vpbroadcastw
+    { TTI::SK_Broadcast, MVT::v32i8,  { 1, 3, 1, 2 } }, // vpbroadcastb
+    { TTI::SK_Broadcast, MVT::v16i8,  { 1, 3, 1, 1 } }, // vpbroadcastb
+
+    { TTI::SK_Reverse, MVT::v4f64,    { 1, 6, 1, 2 } }, // vpermpd
+    { TTI::SK_Reverse, MVT::v8f32,    { 2, 7, 2, 4 } }, // vpermps
+    { TTI::SK_Reverse, MVT::v4i64,    { 1, 6, 1, 2 } }, // vpermq
+    { TTI::SK_Reverse, MVT::v8i32,    { 2, 7, 2, 4 } }, // vpermd
+    { TTI::SK_Reverse, MVT::v16i16,   { 2, 9, 2, 4 } }, // vperm2i128 + pshufb
+    { TTI::SK_Reverse, MVT::v16f16,   { 2, 9, 2, 4 } }, // vperm2i128 + pshufb
+    { TTI::SK_Reverse, MVT::v32i8,    { 2, 9, 2, 4 } }, // vperm2i128 + pshufb
 
     { TTI::SK_Select, MVT::v16i16,    { 1, 1, 1, 1 } }, // vpblendvb
     { TTI::SK_Select, MVT::v16f16,    { 1, 1, 1, 1 } }, // vpblendvb
@@ -2077,23 +2088,23 @@ InstructionCost X86TTIImpl::getShuffleCost(TTI::ShuffleKind Kind,
         return LT.first * *KindCost;
 
   static const CostKindTblEntry AVX1ShuffleTbl[] = {
-      {TTI::SK_Broadcast, MVT::v4f64,  {2,2,2,2}}, // vperm2f128 + vpermilpd
-      {TTI::SK_Broadcast, MVT::v8f32,  {2,2,2,2}}, // vperm2f128 + vpermilps
-      {TTI::SK_Broadcast, MVT::v4i64,  {2,2,2,2}}, // vperm2f128 + vpermilpd
-      {TTI::SK_Broadcast, MVT::v8i32,  {2,2,2,2}}, // vperm2f128 + vpermilps
-      {TTI::SK_Broadcast, MVT::v16i16, {3,3,3,3}}, // vpshuflw + vpshufd + vinsertf128
-      {TTI::SK_Broadcast, MVT::v16f16, {3,3,3,3}}, // vpshuflw + vpshufd + vinsertf128
-      {TTI::SK_Broadcast, MVT::v32i8,  {2,2,2,2}}, // vpshufb + vinsertf128
-
-      {TTI::SK_Reverse, MVT::v4f64,  {2,2,2,2}}, // vperm2f128 + vpermilpd
-      {TTI::SK_Reverse, MVT::v8f32,  {2,2,2,2}}, // vperm2f128 + vpermilps
-      {TTI::SK_Reverse, MVT::v4i64,  {2,2,2,2}}, // vperm2f128 + vpermilpd
-      {TTI::SK_Reverse, MVT::v8i32,  {2,2,2,2}}, // vperm2f128 + vpermilps
-      {TTI::SK_Reverse, MVT::v16i16, {4,4,4,4}}, // vextractf128 + 2*pshufb
+      {TTI::SK_Broadcast, MVT::v4f64,  {2,3,2,3}}, // vperm2f128 + vpermilpd
+      {TTI::SK_Broadcast, MVT::v8f32,  {2,3,2,3}}, // vperm2f128 + vpermilps
+      {TTI::SK_Broadcast, MVT::v4i64,  {2,3,2,3}}, // vperm2f128 + vpermilpd
+      {TTI::SK_Broadcast, MVT::v8i32,  {2,3,2,3}}, // vperm2f128 + vpermilps
+      {TTI::SK_Broadcast, MVT::v16i16, {2,3,3,4}}, // vpshuflw + vpshufd + vinsertf128
+      {TTI::SK_Broadcast, MVT::v16f16, {2,3,3,4}}, // vpshuflw + vpshufd + vinsertf128
+      {TTI::SK_Broadcast, MVT::v32i8,  {3,4,3,6}}, // vpshufb + vinsertf128
+
+      {TTI::SK_Reverse, MVT::v4f64,  {2,6,2,2}}, // vperm2f128 + vpermilpd
+      {TTI::SK_Reverse, MVT::v8f32,  {2,7,2,4}}, // vperm2f128 + vpermilps
+      {TTI::SK_Reverse, MVT::v4i64,  {2,6,2,2}}, // vperm2f128 + vpermilpd
+      {TTI::SK_Reverse, MVT::v8i32,  {2,7,2,4}}, // vperm2f128 + vpermilps
+      {TTI::SK_Reverse, MVT::v16i16, {2,9,5,5}}, // vextractf128 + 2*pshufb
                                                  // + vinsertf128
-      {TTI::SK_Reverse, MVT::v16f16, {4,4,4,4}}, // vextractf128 + 2*pshufb
+      {TTI::SK_Reverse, MVT::v16f16, {2,9,5,5}}, // vextractf128 + 2*pshufb
                                                  // + vinsertf128
-      {TTI::SK_Reverse, MVT::v32i8,  {4,4,4,4}}, // vextractf128 + 2*pshufb
+      {TTI::SK_Reverse, MVT::v32i8,  {2,9,5,5}}, // vextractf128 + 2*pshufb
                                                  // + vinsertf128
 
       {TTI::SK_Select, MVT::v4i64,  {1,1,1,1}}, // vblendpd
@@ -2156,13 +2167,13 @@ InstructionCost X86TTIImpl::getShuffleCost(TTI::ShuffleKind Kind,
         return LT.first * *KindCost;
 
   static const CostKindTblEntry SSSE3ShuffleTbl[] = {
-      {TTI::SK_Broadcast, MVT::v8i16, {1, 1, 1, 1}}, // pshufb
-      {TTI::SK_Broadcast, MVT::v8f16, {1, 1, 1, 1}}, // pshufb
-      {TTI::SK_Broadcast, MVT::v16i8, {1, 1, 1, 1}}, // pshufb
+      {TTI::SK_Broadcast, MVT::v8i16, {1, 3, 2, 2}}, // pshufb
+      {TTI::SK_Broadcast, MVT::v8f16, {1, 3, 2, 2}}, // pshufb
+      {TTI::SK_Broadcast, MVT::v16i8, {1, 3, 2, 2}}, // pshufb
 
-      {TTI::SK_Reverse, MVT::v8i16, {1, 1, 1, 1}}, // pshufb
-      {TTI::SK_Reverse, MVT::v8f16, {1, 1, 1, 1}}, // pshufb
-      {TTI::SK_Reverse, MVT::v16i8, {1, 1, 1, 1}}, // pshufb
+      {TTI::SK_Reverse, MVT::v8i16, {1, 2, 1, 2}}, // pshufb
+      {TTI::SK_Reverse, MVT::v8f16, {1, 2, 1, 2}}, // pshufb
+      {TTI::SK_Reverse, MVT::v16i8, {1, 2, 1, 2}}, // pshufb
 
       {TTI::SK_Select, MVT::v8i16, {3, 3, 3, 3}}, // 2*pshufb + por
       {TTI::SK_Select, MVT::v8f16, {3, 3, 3, 3}}, // 2*pshufb + por
@@ -2192,16 +2203,16 @@ InstructionCost X86TTIImpl::getShuffleCost(TTI::ShuffleKind Kind,
       {TTI::SK_Broadcast, MVT::v2f64, {1, 1, 1, 1}}, // shufpd
       {TTI::SK_Broadcast, MVT::v2i64, {1, 1, 1, 1}}, // pshufd
       {TTI::SK_Broadcast, MVT::v4i32, {1, 1, 1, 1}}, // pshufd
-      {TTI::SK_Broadcast, MVT::v8i16, {2, 2, 2, 2}}, // pshuflw + pshufd
-      {TTI::SK_Broadcast, MVT::v8f16, {2, 2, 2, 2}}, // pshuflw + pshufd
-      {TTI::SK_Broadcast, MVT::v16i8, {3, 3, 3, 3}}, // unpck + pshuflw + pshufd
+      {TTI::SK_Broadcast, MVT::v8i16, {1, 2, 2, 2}}, // pshuflw + pshufd
+      {TTI::SK_Broadcast, MVT::v8f16, {1, 2, 2, 2}}, // pshuflw + pshufd
+      {TTI::SK_Broadcast, MVT::v16i8, {2, 3, 3, 4}}, // unpck + pshuflw + pshufd
 
       {TTI::SK_Reverse, MVT::v2f64, {1, 1, 1, 1}}, // shufpd
       {TTI::SK_Reverse, MVT::v2i64, {1, 1, 1, 1}}, // pshufd
       {TTI::SK_Reverse, MVT::v4i32, {1, 1, 1, 1}}, // pshufd
-      {TTI::SK_Reverse, MVT::v8i16, {3, 3, 3, 3}}, // pshuflw + pshufhw + pshufd
-      {TTI::SK_Reverse, MVT::v8f16, {3, 3, 3, 3}}, // pshuflw + pshufhw + pshufd
-      {TTI::SK_Reverse, MVT::v16i8, {9, 9, 9, 9}}, // 2*pshuflw + 2*pshufhw
+      {TTI::SK_Reverse, MVT::v8i16, {2, 3, 3, 3}}, // pshuflw + pshufhw + pshufd
+      {TTI::SK_Reverse, MVT::v8f16, {2, 3, 3, 3}}, // pshuflw + pshufhw + pshufd
+      {TTI::SK_Reverse, MVT::v16i8, {5, 6,11,11}}, // 2*pshuflw + 2*pshufhw
                                                    // + 2*pshufd + 2*unpck + packus
 
       {TTI::SK_Select, MVT::v2i64, {1, 1, 1, 1}}, // movsd
diff --git a/llvm/lib/TargetParser/RISCVISAInfo.cpp b/llvm/lib/TargetParser/RISCVISAInfo.cpp
index 17c9833..d6afb8a 100644
--- a/llvm/lib/TargetParser/RISCVISAInfo.cpp
+++ b/llvm/lib/TargetParser/RISCVISAInfo.cpp
@@ -858,16 +858,15 @@ void RISCVISAInfo::updateImplication() {
     StringRef ExtName = WorkList.pop_back_val();
     auto Range = std::equal_range(std::begin(ImpliedExts),
                                   std::end(ImpliedExts), ExtName);
-    std::for_each(Range.first, Range.second,
-                  [&](const ImpliedExtsEntry &Implied) {
-                    const char *ImpliedExt = Implied.ImpliedExt;
-                    auto [It, Inserted] = Exts.try_emplace(ImpliedExt);
-                    if (!Inserted)
-                      return;
-                    auto Version = findDefaultVersion(ImpliedExt);
-                    It->second = *Version;
-                    WorkList.push_back(ImpliedExt);
-                  });
+    for (const ImpliedExtsEntry &Implied : llvm::make_range(Range)) {
+      const char *ImpliedExt = Implied.ImpliedExt;
+      auto [It, Inserted] = Exts.try_emplace(ImpliedExt);
+      if (!Inserted)
+        continue;
+      auto Version = findDefaultVersion(ImpliedExt);
+      It->second = *Version;
+      WorkList.push_back(ImpliedExt);
+    }
   }
 
   // Add Zcd if C and D are enabled.
diff --git a/llvm/lib/TextAPI/SymbolSet.cpp b/llvm/lib/TextAPI/SymbolSet.cpp
index 2e0b416..f21a061 100644
--- a/llvm/lib/TextAPI/SymbolSet.cpp
+++ b/llvm/lib/TextAPI/SymbolSet.cpp
@@ -11,6 +11,11 @@
 using namespace llvm;
 using namespace llvm::MachO;
 
+SymbolSet::~SymbolSet() {
+  for (auto &[Key, Sym] : Symbols)
+    Sym->~Symbol();
+}
+
 Symbol *SymbolSet::addGlobalImpl(EncodeKind Kind, StringRef Name,
                                  SymbolFlags Flags) {
   Name = copyString(Name);
diff --git a/llvm/lib/Transforms/IPO/MemProfContextDisambiguation.cpp b/llvm/lib/Transforms/IPO/MemProfContextDisambiguation.cpp
index b803c97..0164fcd 100644
--- a/llvm/lib/Transforms/IPO/MemProfContextDisambiguation.cpp
+++ b/llvm/lib/Transforms/IPO/MemProfContextDisambiguation.cpp
@@ -2073,14 +2073,14 @@ std::string IndexCallsiteContextGraph::getLabel(const FunctionSummary *Func,
                                                 unsigned CloneNo) const {
   auto VI = FSToVIMap.find(Func);
   assert(VI != FSToVIMap.end());
+  std::string CallerName = getMemProfFuncName(VI->second.name(), CloneNo);
   if (isa<AllocInfo *>(Call))
-    return (VI->second.name() + " -> alloc").str();
+    return CallerName + " -> alloc";
   else {
     auto *Callsite = dyn_cast_if_present<CallsiteInfo *>(Call);
-    return (VI->second.name() + " -> " +
-            getMemProfFuncName(Callsite->Callee.name(),
-                               Callsite->Clones[CloneNo]))
-        .str();
+    return CallerName + " -> " +
+           getMemProfFuncName(Callsite->Callee.name(),
+                              Callsite->Clones[CloneNo]);
   }
 }
 
diff --git a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
index 033ef8b..a43a6ee 100644
--- a/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
+++ b/llvm/lib/Transforms/InstCombine/InstCombineCasts.cpp
@@ -708,10 +708,14 @@ static Instruction *shrinkSplatShuffle(TruncInst &Trunc,
   auto *Shuf = dyn_cast<ShuffleVectorInst>(Trunc.getOperand(0));
   if (Shuf && Shuf->hasOneUse() && match(Shuf->getOperand(1), m_Undef()) &&
       all_equal(Shuf->getShuffleMask()) &&
-      Shuf->getType() == Shuf->getOperand(0)->getType()) {
+      ElementCount::isKnownGE(Shuf->getType()->getElementCount(),
+                              cast<VectorType>(Shuf->getOperand(0)->getType())
+                                  ->getElementCount())) {
     // trunc (shuf X, Undef, SplatMask) --> shuf (trunc X), Poison, SplatMask
     // trunc (shuf X, Poison, SplatMask) --> shuf (trunc X), Poison, SplatMask
-    Value *NarrowOp = Builder.CreateTrunc(Shuf->getOperand(0), Trunc.getType());
+    Type *NewTruncTy = Shuf->getOperand(0)->getType()->getWithNewType(
+        Trunc.getType()->getScalarType());
+    Value *NarrowOp = Builder.CreateTrunc(Shuf->getOperand(0), NewTruncTy);
     return new ShuffleVectorInst(NarrowOp, Shuf->getShuffleMask());
   }
 
diff --git a/llvm/lib/Transforms/Instrumentation/HWAddressSanitizer.cpp b/llvm/lib/Transforms/Instrumentation/HWAddressSanitizer.cpp
index fe3315e..4e5a8d1 100644
--- a/llvm/lib/Transforms/Instrumentation/HWAddressSanitizer.cpp
+++ b/llvm/lib/Transforms/Instrumentation/HWAddressSanitizer.cpp
@@ -1459,8 +1459,6 @@ void HWAddressSanitizer::instrumentStack(memtag::StackInfo &SInfo,
     size_t Size = memtag::getAllocaSizeInBytes(*AI);
     size_t AlignedSize = alignTo(Size, Mapping.getObjectAlignment());
 
-    Value *AICast = IRB.CreatePointerCast(AI, PtrTy);
-
     auto HandleLifetime = [&](IntrinsicInst *II) {
       // Set the lifetime intrinsic to cover the whole alloca. This reduces the
       // set of assumptions we need to make about the lifetime. Without this we
@@ -1473,14 +1471,13 @@ void HWAddressSanitizer::instrumentStack(memtag::StackInfo &SInfo,
       // one set of start / end in any execution (i.e. the ends are not
       // reachable from each other), so this will not cause any problems.
       II->setArgOperand(0, ConstantInt::get(Int64Ty, AlignedSize));
-      II->setArgOperand(1, AICast);
     };
     llvm::for_each(Info.LifetimeStart, HandleLifetime);
     llvm::for_each(Info.LifetimeEnd, HandleLifetime);
 
-    AI->replaceUsesWithIf(Replacement, [AICast, AILong](const Use &U) {
+    AI->replaceUsesWithIf(Replacement, [AILong](const Use &U) {
       auto *User = U.getUser();
-      return User != AILong && User != AICast && !isa<LifetimeIntrinsic>(User);
+      return User != AILong && !isa<LifetimeIntrinsic>(User);
     });
 
     memtag::annotateDebugRecords(Info, retagMask(N));
diff --git a/llvm/lib/Transforms/ObjCARC/CMakeLists.txt b/llvm/lib/Transforms/ObjCARC/CMakeLists.txt
index 80867db..4274667 100644
--- a/llvm/lib/Transforms/ObjCARC/CMakeLists.txt
+++ b/llvm/lib/Transforms/ObjCARC/CMakeLists.txt
@@ -2,7 +2,6 @@ add_llvm_component_library(LLVMObjCARCOpts
   ObjCARC.cpp
   ObjCARCOpts.cpp
   ObjCARCExpand.cpp
-  ObjCARCAPElim.cpp
   ObjCARCContract.cpp
   DependencyAnalysis.cpp
   ProvenanceAnalysis.cpp
diff --git a/llvm/lib/Transforms/ObjCARC/ObjCARCAPElim.cpp b/llvm/lib/Transforms/ObjCARC/ObjCARCAPElim.cpp
deleted file mode 100644
index dceb2eb..0000000
--- a/llvm/lib/Transforms/ObjCARC/ObjCARCAPElim.cpp
+++ /dev/null
@@ -1,156 +0,0 @@
-//===- ObjCARCAPElim.cpp - ObjC ARC Optimization --------------------------===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-/// \file
-///
-/// This file defines ObjC ARC optimizations. ARC stands for Automatic
-/// Reference Counting and is a system for managing reference counts for objects
-/// in Objective C.
-///
-/// This specific file implements optimizations which remove extraneous
-/// autorelease pools.
-///
-/// WARNING: This file knows about certain library functions. It recognizes them
-/// by name, and hardwires knowledge of their semantics.
-///
-/// WARNING: This file knows about how certain Objective-C library functions are
-/// used. Naive LLVM IR transformations which would otherwise be
-/// behavior-preserving may break these assumptions.
-///
-//===----------------------------------------------------------------------===//
-
-#include "llvm/ADT/STLExtras.h"
-#include "llvm/Analysis/ObjCARCAnalysisUtils.h"
-#include "llvm/Analysis/ObjCARCInstKind.h"
-#include "llvm/IR/Constants.h"
-#include "llvm/IR/InstrTypes.h"
-#include "llvm/IR/PassManager.h"
-#include "llvm/Support/Debug.h"
-#include "llvm/Support/raw_ostream.h"
-#include "llvm/Transforms/ObjCARC.h"
-
-using namespace llvm;
-using namespace llvm::objcarc;
-
-#define DEBUG_TYPE "objc-arc-ap-elim"
-
-namespace {
-
-/// Interprocedurally determine if calls made by the given call site can
-/// possibly produce autoreleases.
-bool MayAutorelease(const CallBase &CB, unsigned Depth = 0) {
-  if (const Function *Callee = CB.getCalledFunction()) {
-    if (!Callee->hasExactDefinition())
-      return true;
-    for (const BasicBlock &BB : *Callee) {
-      for (const Instruction &I : BB)
-        if (const CallBase *JCB = dyn_cast<CallBase>(&I))
-          // This recursion depth limit is arbitrary. It's just great
-          // enough to cover known interesting testcases.
-          if (Depth < 3 && !JCB->onlyReadsMemory() &&
-              MayAutorelease(*JCB, Depth + 1))
-            return true;
-    }
-    return false;
-  }
-
-  return true;
-}
-
-bool OptimizeBB(BasicBlock *BB) {
-  bool Changed = false;
-
-  Instruction *Push = nullptr;
-  for (Instruction &Inst : llvm::make_early_inc_range(*BB)) {
-    switch (GetBasicARCInstKind(&Inst)) {
-    case ARCInstKind::AutoreleasepoolPush:
-      Push = &Inst;
-      break;
-    case ARCInstKind::AutoreleasepoolPop:
-      // If this pop matches a push and nothing in between can autorelease,
-      // zap the pair.
-      if (Push && cast<CallInst>(&Inst)->getArgOperand(0) == Push) {
-        Changed = true;
-        LLVM_DEBUG(dbgs() << "ObjCARCAPElim::OptimizeBB: Zapping push pop "
-                             "autorelease pair:\n"
-                             "                           Pop: "
-                          << Inst << "\n"
-                          << "                           Push: " << *Push
-                          << "\n");
-        Inst.eraseFromParent();
-        Push->eraseFromParent();
-      }
-      Push = nullptr;
-      break;
-    case ARCInstKind::CallOrUser:
-      if (MayAutorelease(cast<CallBase>(Inst)))
-        Push = nullptr;
-      break;
-    default:
-      break;
-    }
-  }
-
-  return Changed;
-}
-
-bool runImpl(Module &M) {
-  if (!EnableARCOpts)
-    return false;
-
-  // If nothing in the Module uses ARC, don't do anything.
-  if (!ModuleHasARC(M))
-    return false;
-  // Find the llvm.global_ctors variable, as the first step in
-  // identifying the global constructors. In theory, unnecessary autorelease
-  // pools could occur anywhere, but in practice it's pretty rare. Global
-  // ctors are a place where autorelease pools get inserted automatically,
-  // so it's pretty common for them to be unnecessary, and it's pretty
-  // profitable to eliminate them.
-  GlobalVariable *GV = M.getGlobalVariable("llvm.global_ctors");
-  if (!GV)
-    return false;
-
-  assert(GV->hasDefinitiveInitializer() &&
-         "llvm.global_ctors is uncooperative!");
-
-  bool Changed = false;
-
-  // Dig the constructor functions out of GV's initializer.
-  ConstantArray *Init = cast<ConstantArray>(GV->getInitializer());
-  for (User::op_iterator OI = Init->op_begin(), OE = Init->op_end();
-       OI != OE; ++OI) {
-    Value *Op = *OI;
-    // llvm.global_ctors is an array of three-field structs where the second
-    // members are constructor functions.
-    Function *F = dyn_cast<Function>(cast<ConstantStruct>(Op)->getOperand(1));
-    // If the user used a constructor function with the wrong signature and
-    // it got bitcasted or whatever, look the other way.
-    if (!F)
-      continue;
-    // Only look at function definitions.
-    if (F->isDeclaration())
-      continue;
-    // Only look at functions with one basic block.
-    if (std::next(F->begin()) != F->end())
-      continue;
-    // Ok, a single-block constructor function definition. Try to optimize it.
-    Changed |= OptimizeBB(&F->front());
-  }
-
-  return Changed;
-}
-
-} // namespace
-
-PreservedAnalyses ObjCARCAPElimPass::run(Module &M, ModuleAnalysisManager &AM) {
-  if (!runImpl(M))
-    return PreservedAnalyses::all();
-  PreservedAnalyses PA;
-  PA.preserveSet<CFGAnalyses>();
-  return PA;
-}
diff --git a/llvm/lib/Transforms/Scalar/LoopFuse.cpp b/llvm/lib/Transforms/Scalar/LoopFuse.cpp
index d6bd92d..b5eb647 100644
--- a/llvm/lib/Transforms/Scalar/LoopFuse.cpp
+++ b/llvm/lib/Transforms/Scalar/LoopFuse.cpp
@@ -1176,6 +1176,28 @@ private:
     return true;
   }
 
+  /// This function fixes PHI nodes after fusion in \p SafeToSink.
+  /// \p SafeToSink instructions are the instructions that are to be moved past
+  /// the fused loop. Thus, the PHI nodes in \p SafeToSink should be updated to
+  /// receive values from the fused loop if they are currently taking values
+  /// from the first loop (i.e. FC0)'s latch.
+  void fixPHINodes(ArrayRef<Instruction *> SafeToSink,
+                   const FusionCandidate &FC0,
+                   const FusionCandidate &FC1) const {
+    for (Instruction *Inst : SafeToSink) {
+      // No update needed for non-PHI nodes.
+      PHINode *Phi = dyn_cast<PHINode>(Inst);
+      if (!Phi)
+        continue;
+      for (unsigned I = 0; I < Phi->getNumIncomingValues(); I++) {
+        if (Phi->getIncomingBlock(I) != FC0.Latch)
+          continue;
+        assert(FC1.Latch && "FC1 latch is not set");
+        Phi->setIncomingBlock(I, FC1.Latch);
+      }
+    }
+  }
+
   /// Collect instructions in the \p FC1 Preheader that can be hoisted
   /// to the \p FC0 Preheader or sunk into the \p FC1 Body
   bool collectMovablePreheaderInsts(
@@ -1481,6 +1503,9 @@ private:
       assert(I->getParent() == FC1.Preheader);
       I->moveBefore(*FC1.ExitBlock, FC1.ExitBlock->getFirstInsertionPt());
     }
+    // PHI nodes in SinkInsts need to be updated to receive values from the
+    // fused loop.
+    fixPHINodes(SinkInsts, FC0, FC1);
   }
 
   /// Determine if two fusion candidates have identical guards
diff --git a/llvm/lib/Transforms/Scalar/LoopInterchange.cpp b/llvm/lib/Transforms/Scalar/LoopInterchange.cpp
index 70e9eee..08446cc 100644
--- a/llvm/lib/Transforms/Scalar/LoopInterchange.cpp
+++ b/llvm/lib/Transforms/Scalar/LoopInterchange.cpp
@@ -17,8 +17,8 @@
 #include "llvm/ADT/SmallSet.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/Statistic.h"
+#include "llvm/ADT/StringMap.h"
 #include "llvm/ADT/StringRef.h"
-#include "llvm/ADT/StringSet.h"
 #include "llvm/Analysis/DependenceAnalysis.h"
 #include "llvm/Analysis/LoopCacheAnalysis.h"
 #include "llvm/Analysis/LoopInfo.h"
@@ -70,6 +70,13 @@ namespace {
 
 using LoopVector = SmallVector<Loop *, 8>;
 
+/// A list of direction vectors. Each entry represents a direction vector
+/// corresponding to one or more dependencies existing in the loop nest. The
+/// length of all direction vectors is equal and is N + 1, where N is the depth
+/// of the loop nest. The first N elements correspond to the dependency
+/// direction of each N loops. The last one indicates whether this entry is
+/// forward dependency ('<') or not ('*'). The term "forward" aligns with what
+/// is defined in LoopAccessAnalysis.
 // TODO: Check if we can use a sparse matrix here.
 using CharMatrix = std::vector<std::vector<char>>;
 
@@ -126,11 +133,33 @@ static bool noDuplicateRulesAndIgnore(ArrayRef<RuleTy> Rules) {
 
 static void printDepMatrix(CharMatrix &DepMatrix) {
   for (auto &Row : DepMatrix) {
-    for (auto D : Row)
+    // Drop the last element because it is a flag indicating whether this is
+    // forward dependency or not, which doesn't affect the legality check.
+    for (char D : drop_end(Row))
       LLVM_DEBUG(dbgs() << D << " ");
     LLVM_DEBUG(dbgs() << "\n");
   }
 }
+
+/// Return true if \p Src appears before \p Dst in the same basic block.
+/// Precondition: \p Src and \Dst are distinct instructions within the same
+/// basic block.
+static bool inThisOrder(const Instruction *Src, const Instruction *Dst) {
+  assert(Src->getParent() == Dst->getParent() && Src != Dst &&
+         "Expected Src and Dst to be different instructions in the same BB");
+
+  bool FoundSrc = false;
+  for (const Instruction &I : *(Src->getParent())) {
+    if (&I == Src) {
+      FoundSrc = true;
+      continue;
+    }
+    if (&I == Dst)
+      return FoundSrc;
+  }
+
+  llvm_unreachable("Dst not found");
+}
 #endif
 
 static bool populateDependencyMatrix(CharMatrix &DepMatrix, unsigned Level,
@@ -174,7 +203,10 @@ static bool populateDependencyMatrix(CharMatrix &DepMatrix, unsigned Level,
     return false;
   }
   ValueVector::iterator I, IE, J, JE;
-  StringSet<> Seen;
+
+  // Manage direction vectors that are already seen. Map each direction vector
+  // to an index of DepMatrix at which it is stored.
+  StringMap<unsigned> Seen;
 
   for (I = MemInstr.begin(), IE = MemInstr.end(); I != IE; ++I) {
     for (J = I, JE = MemInstr.end(); J != JE; ++J) {
@@ -228,9 +260,49 @@ static bool populateDependencyMatrix(CharMatrix &DepMatrix, unsigned Level,
           Dep.push_back('I');
         }
 
+        // Test whether the dependency is forward or not.
+        bool IsKnownForward = true;
+        if (Src->getParent() != Dst->getParent()) {
+          // In general, when Src and Dst are in different BBs, the execution
+          // order of them within a single iteration is not guaranteed. Treat
+          // conservatively as not-forward dependency in this case.
+          IsKnownForward = false;
+        } else {
+          // Src and Dst are in the same BB. If they are the different
+          // instructions, Src should appear before Dst in the BB as they are
+          // stored to MemInstr in that order.
+          assert((Src == Dst || inThisOrder(Src, Dst)) &&
+                 "Unexpected instructions");
+
+          // If the Dependence object is reversed (due to normalization), it
+          // represents the dependency from Dst to Src, meaning it is a backward
+          // dependency. Otherwise it should be a forward dependency.
+          bool IsReversed = D->getSrc() != Src;
+          if (IsReversed)
+            IsKnownForward = false;
+        }
+
+        // Initialize the last element. Assume forward dependencies only; it
+        // will be updated later if there is any non-forward dependency.
+        Dep.push_back('<');
+
+        // The last element should express the "summary" among one or more
+        // direction vectors whose first N elements are the same (where N is
+        // the depth of the loop nest). Hence we exclude the last element from
+        // the Seen map.
+        auto [Ite, Inserted] = Seen.try_emplace(
+            StringRef(Dep.data(), Dep.size() - 1), DepMatrix.size());
+
         // Make sure we only add unique entries to the dependency matrix.
-        if (Seen.insert(StringRef(Dep.data(), Dep.size())).second)
+        if (Inserted)
           DepMatrix.push_back(Dep);
+
+        // If we cannot prove that this dependency is forward, change the last
+        // element of the corresponding entry. Since a `[... *]` dependency
+        // includes a `[... <]` dependency, we do not need to keep both and
+        // change the existing entry instead.
+        if (!IsKnownForward)
+          DepMatrix[Ite->second].back() = '*';
       }
     }
   }
@@ -281,11 +353,12 @@ static bool isLegalToInterChangeLoops(CharMatrix &DepMatrix,
       continue;
 
     // Check if the direction vector is lexicographically positive (or zero)
-    // for both before/after exchanged.
-    if (isLexicographicallyPositive(Cur, OuterLoopId, Cur.size()) == false)
+    // for both before/after exchanged. Ignore the last element because it
+    // doesn't affect the legality.
+    if (isLexicographicallyPositive(Cur, OuterLoopId, Cur.size() - 1) == false)
       return false;
     std::swap(Cur[InnerLoopId], Cur[OuterLoopId]);
-    if (isLexicographicallyPositive(Cur, OuterLoopId, Cur.size()) == false)
+    if (isLexicographicallyPositive(Cur, OuterLoopId, Cur.size() - 1) == false)
       return false;
   }
   return true;
@@ -1334,22 +1407,35 @@ LoopInterchangeProfitability::isProfitablePerInstrOrderCost() {
 static bool canVectorize(const CharMatrix &DepMatrix, unsigned LoopId) {
   for (const auto &Dep : DepMatrix) {
     char Dir = Dep[LoopId];
-    if (Dir != 'I' && Dir != '=')
-      return false;
+    char DepType = Dep.back();
+    assert((DepType == '<' || DepType == '*') &&
+           "Unexpected element in dependency vector");
+
+    // There are no loop-carried dependencies.
+    if (Dir == '=' || Dir == 'I')
+      continue;
+
+    // DepType being '<' means that this direction vector represents a forward
+    // dependency. In principle, a loop with '<' direction can be vectorized in
+    // this case.
+    if (Dir == '<' && DepType == '<')
+      continue;
+
+    // We cannot prove that the loop is vectorizable.
+    return false;
   }
   return true;
 }
 
 std::optional<bool> LoopInterchangeProfitability::isProfitableForVectorization(
     unsigned InnerLoopId, unsigned OuterLoopId, CharMatrix &DepMatrix) {
-  // If the outer loop is not loop independent it is not profitable to move
-  // this to inner position, since doing so would not enable inner loop
-  // parallelism.
+  // If the outer loop cannot be vectorized, it is not profitable to move this
+  // to inner position.
   if (!canVectorize(DepMatrix, OuterLoopId))
     return false;
 
-  // If inner loop has dependence and outer loop is loop independent then it is
-  // profitable to interchange to enable inner loop parallelism.
+  // If the inner loop cannot be vectorized but the outer loop can be, then it
+  // is profitable to interchange to enable inner loop parallelism.
   if (!canVectorize(DepMatrix, InnerLoopId))
     return true;
 
diff --git a/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp b/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp
index 84d1c0b..9220abb 100644
--- a/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp
+++ b/llvm/lib/Transforms/Scalar/MemCpyOptimizer.cpp
@@ -1593,11 +1593,8 @@ bool MemCpyOptPass::performStackMoveOptzn(Instruction *Load, Instruction *Store,
             // since both llvm.lifetime.start and llvm.lifetime.end intrinsics
             // practically fill all the bytes of the alloca with an undefined
             // value, although conceptually marked as alive/dead.
-            int64_t Size = cast<ConstantInt>(UI->getOperand(0))->getSExtValue();
-            if (Size < 0 || Size == DestSize) {
-              LifetimeMarkers.push_back(UI);
-              continue;
-            }
+            LifetimeMarkers.push_back(UI);
+            continue;
           }
           AAMetadataInstrs.insert(UI);
 
@@ -1614,9 +1611,8 @@ bool MemCpyOptPass::performStackMoveOptzn(Instruction *Load, Instruction *Store,
     return true;
   };
 
-  // Check that dest has no Mod/Ref, from the alloca to the Store, except full
-  // size lifetime intrinsics. And collect modref inst for the reachability
-  // check.
+  // Check that dest has no Mod/Ref, from the alloca to the Store. And collect
+  // modref inst for the reachability check.
   ModRefInfo DestModRef = ModRefInfo::NoModRef;
   MemoryLocation DestLoc(DestAlloca, LocationSize::precise(Size));
   SmallVector<BasicBlock *, 8> ReachabilityWorklist;
diff --git a/llvm/lib/Transforms/Scalar/StructurizeCFG.cpp b/llvm/lib/Transforms/Scalar/StructurizeCFG.cpp
index a69d649..44e63a0 100644
--- a/llvm/lib/Transforms/Scalar/StructurizeCFG.cpp
+++ b/llvm/lib/Transforms/Scalar/StructurizeCFG.cpp
@@ -19,6 +19,7 @@
 #include "llvm/Analysis/RegionInfo.h"
 #include "llvm/Analysis/RegionIterator.h"
 #include "llvm/Analysis/RegionPass.h"
+#include "llvm/Analysis/TargetTransformInfo.h"
 #include "llvm/Analysis/UniformityAnalysis.h"
 #include "llvm/IR/BasicBlock.h"
 #include "llvm/IR/CFG.h"
@@ -128,6 +129,7 @@ struct PredInfo {
 using BBPredicates = DenseMap<BasicBlock *, PredInfo>;
 using PredMap = DenseMap<BasicBlock *, BBPredicates>;
 using BB2BBMap = DenseMap<BasicBlock *, BasicBlock *>;
+using Val2BBMap = DenseMap<Value *, BasicBlock *>;
 
 // A traits type that is intended to be used in graph algorithms. The graph
 // traits starts at an entry node, and traverses the RegionNodes that are in
@@ -279,7 +281,7 @@ class StructurizeCFG {
   ConstantInt *BoolTrue;
   ConstantInt *BoolFalse;
   Value *BoolPoison;
-
+  const TargetTransformInfo *TTI;
   Function *Func;
   Region *ParentRegion;
 
@@ -301,8 +303,12 @@ class StructurizeCFG {
   PredMap LoopPreds;
   BranchVector LoopConds;
 
+  Val2BBMap HoistedValues;
+
   RegionNode *PrevNode;
 
+  void hoistZeroCostElseBlockPhiValues(BasicBlock *ElseBB, BasicBlock *ThenBB);
+
   void orderNodes();
 
   void analyzeLoops(RegionNode *N);
@@ -332,6 +338,8 @@ class StructurizeCFG {
 
   void simplifyAffectedPhis();
 
+  void simplifyHoistedPhis();
+
   DebugLoc killTerminator(BasicBlock *BB);
 
   void changeExit(RegionNode *Node, BasicBlock *NewExit,
@@ -359,7 +367,7 @@ class StructurizeCFG {
 
 public:
   void init(Region *R);
-  bool run(Region *R, DominatorTree *DT);
+  bool run(Region *R, DominatorTree *DT, const TargetTransformInfo *TTI);
   bool makeUniformRegion(Region *R, UniformityInfo &UA);
 };
 
@@ -385,8 +393,11 @@ public:
       if (SCFG.makeUniformRegion(R, UA))
         return false;
     }
+    Function *F = R->getEntry()->getParent();
+    const TargetTransformInfo *TTI =
+        &getAnalysis<TargetTransformInfoWrapperPass>().getTTI(*F);
     DominatorTree *DT = &getAnalysis<DominatorTreeWrapperPass>().getDomTree();
-    return SCFG.run(R, DT);
+    return SCFG.run(R, DT, TTI);
   }
 
   StringRef getPassName() const override { return "Structurize control flow"; }
@@ -394,7 +405,9 @@ public:
   void getAnalysisUsage(AnalysisUsage &AU) const override {
     if (SkipUniformRegions)
       AU.addRequired<UniformityInfoWrapperPass>();
+    AU.addRequired<TargetTransformInfoWrapperPass>();
     AU.addRequired<DominatorTreeWrapperPass>();
+    AU.addRequired<TargetTransformInfoWrapperPass>();
 
     AU.addPreserved<DominatorTreeWrapperPass>();
     RegionPass::getAnalysisUsage(AU);
@@ -403,6 +416,34 @@ public:
 
 } // end anonymous namespace
 
+/// Checks whether an instruction is zero cost instruction and checks if the
+/// operands are from different BB. If so, this instruction can be coalesced
+/// if its hoisted to predecessor block. So, this returns true.
+static bool isHoistableInstruction(Instruction *I, BasicBlock *BB,
+                                   const TargetTransformInfo *TTI) {
+  if (I->getParent() != BB || isa<PHINode>(I))
+    return false;
+
+  // If the instruction is not a zero cost instruction, return false.
+  auto Cost = TTI->getInstructionCost(I, TargetTransformInfo::TCK_Latency);
+  InstructionCost::CostType CostVal =
+      Cost.isValid()
+          ? Cost.getValue()
+          : (InstructionCost::CostType)TargetTransformInfo::TCC_Expensive;
+  if (CostVal != 0)
+    return false;
+
+  // Check if any operands are instructions defined in the same block.
+  for (auto &Op : I->operands()) {
+    if (auto *OpI = dyn_cast<Instruction>(Op)) {
+      if (OpI->getParent() == BB)
+        return false;
+    }
+  }
+
+  return true;
+}
+
 char StructurizeCFGLegacyPass::ID = 0;
 
 INITIALIZE_PASS_BEGIN(StructurizeCFGLegacyPass, "structurizecfg",
@@ -413,6 +454,39 @@ INITIALIZE_PASS_DEPENDENCY(RegionInfoPass)
 INITIALIZE_PASS_END(StructurizeCFGLegacyPass, "structurizecfg",
                     "Structurize the CFG", false, false)
 
+/// Structurization can introduce unnecessary VGPR copies due to register
+/// coalescing interference. For example, if the Else block has a zero-cost
+/// instruction and the Then block modifies the VGPR value, only one value is
+/// live at a time in merge block before structurization. After structurization,
+/// the coalescer may incorrectly treat the Then value as live in the Else block
+/// (via the path Then → Flow → Else), leading to unnecessary VGPR copies.
+///
+/// This function examines phi nodes whose incoming values are zero-cost
+/// instructions in the Else block. It identifies such values that can be safely
+/// hoisted and moves them to the nearest common dominator of Then and Else
+/// blocks. A follow-up function after setting PhiNodes assigns the hoisted
+/// value to poison phi nodes along the if→flow edge, aiding register coalescing
+/// and minimizing unnecessary live ranges.
+void StructurizeCFG::hoistZeroCostElseBlockPhiValues(BasicBlock *ElseBB,
+                                                     BasicBlock *ThenBB) {
+
+  BasicBlock *ElseSucc = ElseBB->getSingleSuccessor();
+  BasicBlock *CommonDominator = DT->findNearestCommonDominator(ElseBB, ThenBB);
+
+  if (!ElseSucc || !CommonDominator)
+    return;
+  Instruction *Term = CommonDominator->getTerminator();
+  for (PHINode &Phi : ElseSucc->phis()) {
+    Value *ElseVal = Phi.getIncomingValueForBlock(ElseBB);
+    auto *Inst = dyn_cast<Instruction>(ElseVal);
+    if (!Inst || !isHoistableInstruction(Inst, ElseBB, TTI))
+      continue;
+    Inst->removeFromParent();
+    Inst->insertInto(CommonDominator, Term->getIterator());
+    HoistedValues[Inst] = CommonDominator;
+  }
+}
+
 /// Build up the general order of nodes, by performing a topological sort of the
 /// parent region's nodes, while ensuring that there is no outer cycle node
 /// between any two inner cycle nodes.
@@ -535,7 +609,7 @@ void StructurizeCFG::gatherPredicates(RegionNode *N) {
             BasicBlock *Other = Term->getSuccessor(!i);
             if (Visited.count(Other) && !Loops.count(Other) &&
                 !Pred.count(Other) && !Pred.count(P)) {
-
+              hoistZeroCostElseBlockPhiValues(Succ, Other);
               Pred[Other] = {BoolFalse, std::nullopt};
               Pred[P] = {BoolTrue, std::nullopt};
               continue;
@@ -891,6 +965,44 @@ void StructurizeCFG::setPhiValues() {
   AffectedPhis.append(InsertedPhis.begin(), InsertedPhis.end());
 }
 
+/// Updates PHI nodes after hoisted zero cost instructions by replacing poison
+/// entries on Flow nodes with the appropriate hoisted values
+void StructurizeCFG::simplifyHoistedPhis() {
+  for (WeakVH VH : AffectedPhis) {
+    PHINode *Phi = dyn_cast_or_null<PHINode>(VH);
+    if (!Phi || Phi->getNumIncomingValues() != 2)
+      continue;
+
+    for (int i = 0; i < 2; i++) {
+      Value *V = Phi->getIncomingValue(i);
+      auto BBIt = HoistedValues.find(V);
+
+      if (BBIt == HoistedValues.end())
+        continue;
+
+      Value *OtherV = Phi->getIncomingValue(!i);
+      PHINode *OtherPhi = dyn_cast<PHINode>(OtherV);
+      if (!OtherPhi)
+        continue;
+
+      int PoisonValBBIdx = -1;
+      for (size_t i = 0; i < OtherPhi->getNumIncomingValues(); i++) {
+        if (!isa<PoisonValue>(OtherPhi->getIncomingValue(i)))
+          continue;
+        PoisonValBBIdx = i;
+        break;
+      }
+      if (PoisonValBBIdx == -1 ||
+          !DT->dominates(BBIt->second,
+                         OtherPhi->getIncomingBlock(PoisonValBBIdx)))
+        continue;
+
+      OtherPhi->setIncomingValue(PoisonValBBIdx, V);
+      Phi->setIncomingValue(i, OtherV);
+    }
+  }
+}
+
 void StructurizeCFG::simplifyAffectedPhis() {
   bool Changed;
   do {
@@ -1283,12 +1395,13 @@ bool StructurizeCFG::makeUniformRegion(Region *R, UniformityInfo &UA) {
 }
 
 /// Run the transformation for each region found
-bool StructurizeCFG::run(Region *R, DominatorTree *DT) {
+bool StructurizeCFG::run(Region *R, DominatorTree *DT,
+                         const TargetTransformInfo *TTI) {
   if (R->isTopLevelRegion())
     return false;
 
   this->DT = DT;
-
+  this->TTI = TTI;
   Func = R->getEntry()->getParent();
   assert(hasOnlySimpleTerminator(*Func) && "Unsupported block terminator.");
 
@@ -1300,6 +1413,7 @@ bool StructurizeCFG::run(Region *R, DominatorTree *DT) {
   insertConditions(false);
   insertConditions(true);
   setPhiValues();
+  simplifyHoistedPhis();
   simplifyConditions();
   simplifyAffectedPhis();
   rebuildSSA();
@@ -1349,7 +1463,7 @@ PreservedAnalyses StructurizeCFGPass::run(Function &F,
   bool Changed = false;
   DominatorTree *DT = &AM.getResult<DominatorTreeAnalysis>(F);
   auto &RI = AM.getResult<RegionInfoAnalysis>(F);
-
+  TargetTransformInfo *TTI = &AM.getResult<TargetIRAnalysis>(F);
   UniformityInfo *UI = nullptr;
   if (SkipUniformRegions)
     UI = &AM.getResult<UniformityInfoAnalysis>(F);
@@ -1368,7 +1482,7 @@ PreservedAnalyses StructurizeCFGPass::run(Function &F,
       continue;
     }
 
-    Changed |= SCFG.run(R, DT);
+    Changed |= SCFG.run(R, DT, TTI);
   }
   if (!Changed)
     return PreservedAnalyses::all();
diff --git a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
index ed08c0b..ddb062b 100644
--- a/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
+++ b/llvm/lib/Transforms/Utils/ScalarEvolutionExpander.cpp
@@ -18,6 +18,7 @@
 #include "llvm/ADT/SmallSet.h"
 #include "llvm/Analysis/InstructionSimplify.h"
 #include "llvm/Analysis/LoopInfo.h"
+#include "llvm/Analysis/ScalarEvolutionPatternMatch.h"
 #include "llvm/Analysis/TargetTransformInfo.h"
 #include "llvm/Analysis/ValueTracking.h"
 #include "llvm/IR/DataLayout.h"
@@ -42,6 +43,7 @@ cl::opt<unsigned> llvm::SCEVCheapExpansionBudget(
              "controls the budget that is considered cheap (default = 4)"));
 
 using namespace PatternMatch;
+using namespace SCEVPatternMatch;
 
 PoisonFlags::PoisonFlags(const Instruction *I) {
   NUW = false;
@@ -1224,6 +1226,7 @@ Value *SCEVExpander::expandAddRecExprLiterally(const SCEVAddRecExpr *S) {
 }
 
 Value *SCEVExpander::tryToReuseLCSSAPhi(const SCEVAddRecExpr *S) {
+  Type *STy = S->getType();
   const Loop *L = S->getLoop();
   BasicBlock *EB = L->getExitBlock();
   if (!EB || !EB->getSinglePredecessor() ||
@@ -1231,11 +1234,36 @@ Value *SCEVExpander::tryToReuseLCSSAPhi(const SCEVAddRecExpr *S) {
     return nullptr;
 
   for (auto &PN : EB->phis()) {
-    if (!SE.isSCEVable(PN.getType()) || PN.getType() != S->getType())
+    if (!SE.isSCEVable(PN.getType()))
       continue;
-    auto *ExitV = SE.getSCEV(&PN);
-    if (S == ExitV)
-      return &PN;
+    auto *ExitSCEV = SE.getSCEV(&PN);
+    if (!isa<SCEVAddRecExpr>(ExitSCEV))
+      continue;
+    Type *PhiTy = PN.getType();
+    if (STy->isIntegerTy() && PhiTy->isPointerTy())
+      ExitSCEV = SE.getPtrToIntExpr(ExitSCEV, STy);
+    else if (S->getType() != PN.getType())
+      continue;
+
+    // Check if we can re-use the existing PN, by adjusting it with an expanded
+    // offset, if the offset is simpler.
+    const SCEV *Diff = SE.getMinusSCEV(S, ExitSCEV);
+    const SCEV *Op = Diff;
+    match(Diff, m_scev_Mul(m_scev_AllOnes(), m_SCEV(Op)));
+    match(Op, m_scev_PtrToInt(m_SCEV(Op)));
+    if (!isa<SCEVConstant, SCEVUnknown>(Op))
+      continue;
+
+    assert(Diff->getType()->isIntegerTy() &&
+           "difference must be of integer type");
+    Value *DiffV = expand(Diff);
+    Value *BaseV = &PN;
+    if (PhiTy->isPointerTy()) {
+      if (STy->isPointerTy())
+        return Builder.CreatePtrAdd(BaseV, DiffV);
+      BaseV = Builder.CreatePtrToInt(BaseV, DiffV->getType());
+    }
+    return Builder.CreateAdd(BaseV, DiffV);
   }
 
   return nullptr;
diff --git a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
index 9c9201c..1c2de75 100644
--- a/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
+++ b/llvm/lib/Transforms/Vectorize/LoopVectorize.cpp
@@ -1359,9 +1359,7 @@ public:
       return;
     // Override EVL styles if needed.
     // FIXME: Investigate opportunity for fixed vector factor.
-    // FIXME: Support interleave accesses.
     bool EVLIsLegal = UserIC <= 1 && IsScalableVF &&
-                      !InterleaveInfo.hasGroups() &&
                       TTI.hasActiveVectorLength() && !EnableVPlanNativePath;
     if (EVLIsLegal)
       return;
@@ -2023,6 +2021,9 @@ public:
   /// Retrieves the MemCheckCond and MemCheckBlock that were generated as IR
   /// outside VPlan.
   std::pair<Value *, BasicBlock *> getMemRuntimeChecks() {
+    using namespace llvm::PatternMatch;
+    if (MemRuntimeCheckCond && match(MemRuntimeCheckCond, m_ZeroInt()))
+      return {nullptr, nullptr};
     return {MemRuntimeCheckCond, MemCheckBlock};
   }
 
@@ -4499,19 +4500,17 @@ VectorizationFactor LoopVectorizationPlanner::selectEpilogueVectorizationFactor(
   Type *TCType = Legal->getWidestInductionType();
   const SCEV *RemainingIterations = nullptr;
   unsigned MaxTripCount = 0;
-  if (MainLoopVF.isFixed()) {
-    // TODO: extend to support scalable VFs.
-    const SCEV *TC = vputils::getSCEVExprForVPValue(
-        getPlanFor(MainLoopVF).getTripCount(), SE);
-    assert(!isa<SCEVCouldNotCompute>(TC) &&
-           "Trip count SCEV must be computable");
-    RemainingIterations = SE.getURemExpr(
-        TC, SE.getConstant(TCType, MainLoopVF.getFixedValue() * IC));
-
-    // No iterations left to process in the epilogue.
-    if (RemainingIterations->isZero())
-      return Result;
+  const SCEV *TC =
+      vputils::getSCEVExprForVPValue(getPlanFor(MainLoopVF).getTripCount(), SE);
+  assert(!isa<SCEVCouldNotCompute>(TC) && "Trip count SCEV must be computable");
+  RemainingIterations =
+      SE.getURemExpr(TC, SE.getElementCount(TCType, MainLoopVF * IC));
+
+  // No iterations left to process in the epilogue.
+  if (RemainingIterations->isZero())
+    return Result;
 
+  if (MainLoopVF.isFixed()) {
     MaxTripCount = MainLoopVF.getFixedValue() * IC - 1;
     if (SE.isKnownPredicate(CmpInst::ICMP_ULT, RemainingIterations,
                             SE.getConstant(TCType, MaxTripCount))) {
@@ -7278,6 +7277,7 @@ DenseMap<const SCEV *, Value *> LoopVectorizationPlanner::executePlan(
   VPBasicBlock *VectorPH = cast<VPBasicBlock>(BestVPlan.getVectorPreheader());
   VPlanTransforms::optimizeForVFAndUF(BestVPlan, BestVF, BestUF, PSE);
   VPlanTransforms::simplifyRecipes(BestVPlan, *Legal->getWidestInductionType());
+  VPlanTransforms::removeBranchOnConst(BestVPlan);
   VPlanTransforms::narrowInterleaveGroups(
       BestVPlan, BestVF,
       TTI.getRegisterBitWidth(TargetTransformInfo::RGK_FixedWidthVector));
@@ -10074,12 +10074,6 @@ bool LoopVectorizePass::processLoop(Loop *L) {
   // Get user vectorization factor and interleave count.
   ElementCount UserVF = Hints.getWidth();
   unsigned UserIC = Hints.getInterleave();
-  if (LVL.hasUncountableEarlyExit() && UserIC != 1) {
-    UserIC = 1;
-    reportVectorizationInfo("Interleaving not supported for loops "
-                            "with uncountable early exits",
-                            "InterleaveEarlyExitDisabled", ORE, L);
-  }
 
   // Plan how to best vectorize.
   LVP.plan(UserVF, UserIC);
@@ -10097,9 +10091,20 @@ bool LoopVectorizePass::processLoop(Loop *L) {
     unsigned SelectedIC = std::max(IC, UserIC);
     //  Optimistically generate runtime checks if they are needed. Drop them if
     //  they turn out to not be profitable.
-    if (VF.Width.isVector() || SelectedIC > 1)
+    if (VF.Width.isVector() || SelectedIC > 1) {
       Checks.create(L, *LVL.getLAI(), PSE.getPredicate(), VF.Width, SelectedIC);
 
+      // Bail out early if either the SCEV or memory runtime checks are known to
+      // fail. In that case, the vector loop would never execute.
+      using namespace llvm::PatternMatch;
+      if (Checks.getSCEVChecks().first &&
+          match(Checks.getSCEVChecks().first, m_One()))
+        return false;
+      if (Checks.getMemRuntimeChecks().first &&
+          match(Checks.getMemRuntimeChecks().first, m_One()))
+        return false;
+    }
+
     // Check if it is profitable to vectorize with runtime checks.
     bool ForceVectorization =
         Hints.getForce() == LoopVectorizeHints::FK_Enabled;
@@ -10230,6 +10235,11 @@ bool LoopVectorizePass::processLoop(Loop *L) {
           L, PSE, LI, DT, TLI, TTI, AC, ORE, ElementCount::getFixed(1),
           ElementCount::getFixed(1), IC, &CM, BFI, PSI, Checks, BestPlan);
 
+      // TODO: Move to general VPlan pipeline once epilogue loops are also
+      // supported.
+      VPlanTransforms::runPass(VPlanTransforms::materializeVectorTripCount,
+                               BestPlan, VF.Width, IC, PSE);
+
       LVP.executePlan(VF.Width, IC, BestPlan, Unroller, DT, false);
 
       ORE->emit([&]() {
@@ -10297,6 +10307,11 @@ bool LoopVectorizePass::processLoop(Loop *L) {
         InnerLoopVectorizer LB(L, PSE, LI, DT, TLI, TTI, AC, ORE, VF.Width,
                                VF.MinProfitableTripCount, IC, &CM, BFI, PSI,
                                Checks, BestPlan);
+        // TODO: Move to general VPlan pipeline once epilogue loops are also
+        // supported.
+        VPlanTransforms::runPass(VPlanTransforms::materializeVectorTripCount,
+                                 BestPlan, VF.Width, IC, PSE);
+
         LVP.executePlan(VF.Width, IC, BestPlan, LB, DT, false);
         ++LoopsVectorized;
 
diff --git a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
index 0d0b342..593868f 100644
--- a/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
+++ b/llvm/lib/Transforms/Vectorize/SLPVectorizer.cpp
@@ -206,6 +206,12 @@ static cl::opt<bool> VectorizeNonPowerOf2(
     "slp-vectorize-non-power-of-2", cl::init(false), cl::Hidden,
     cl::desc("Try to vectorize with non-power-of-2 number of elements."));
 
+/// Enables vectorization of copyable elements.
+static cl::opt<bool> VectorizeCopyableElements(
+    "slp-copyable-elements", cl::init(true), cl::Hidden,
+    cl::desc("Try to replace values with the idempotent instructions for "
+             "better vectorization."));
+
 // Limit the number of alias checks. The limit is chosen so that
 // it has no negative effect on the llvm benchmarks.
 static const unsigned AliasedCheckLimit = 10;
@@ -855,6 +861,13 @@ static std::optional<unsigned> getExtractIndex(const Instruction *E) {
   return *EI->idx_begin();
 }
 
+namespace llvm {
+/// Checks if the specified value does not require scheduling. It does not
+/// require scheduling if all operands and all users do not need to be scheduled
+/// in the current basic block.
+static bool doesNotNeedToBeScheduled(Value *V);
+} // namespace llvm
+
 namespace {
 /// \returns true if \p Opcode is allowed as part of the main/alternate
 /// instruction for SLP vectorization.
@@ -957,6 +970,33 @@ class BinOpSameOpcodeHelper {
         return Instruction::Xor;
       llvm_unreachable("Cannot find interchangeable instruction.");
     }
+
+    /// Return true if the instruction can be converted to \p Opcode.
+    bool hasCandidateOpcode(unsigned Opcode) const {
+      MaskType Candidate = Mask & SeenBefore;
+      switch (Opcode) {
+      case Instruction::Shl:
+        return Candidate & ShlBIT;
+      case Instruction::AShr:
+        return Candidate & AShrBIT;
+      case Instruction::Mul:
+        return Candidate & MulBIT;
+      case Instruction::Add:
+        return Candidate & AddBIT;
+      case Instruction::Sub:
+        return Candidate & SubBIT;
+      case Instruction::And:
+        return Candidate & AndBIT;
+      case Instruction::Or:
+        return Candidate & OrBIT;
+      case Instruction::Xor:
+        return Candidate & XorBIT;
+      default:
+        break;
+      }
+      llvm_unreachable("Cannot find interchangeable instruction.");
+    }
+
     SmallVector<Value *> getOperand(const Instruction *To) const {
       unsigned ToOpcode = To->getOpcode();
       unsigned FromOpcode = I->getOpcode();
@@ -1117,6 +1157,10 @@ public:
             AltOp.trySet(OpcodeInMaskForm, InterchangeableMask));
   }
   unsigned getMainOpcode() const { return MainOp.getOpcode(); }
+  /// Checks if the list of potential opcodes includes \p Opcode.
+  bool hasCandidateOpcode(unsigned Opcode) const {
+    return MainOp.hasCandidateOpcode(Opcode);
+  }
   bool hasAltOp() const { return AltOp.I; }
   unsigned getAltOpcode() const {
     return hasAltOp() ? AltOp.getOpcode() : getMainOpcode();
@@ -1152,6 +1196,8 @@ class InstructionsState {
   /// GetVectorCost.
   Instruction *MainOp = nullptr;
   Instruction *AltOp = nullptr;
+  /// Wether the instruction state represents copyable instructions.
+  bool HasCopyables = false;
 
 public:
   Instruction *getMainOp() const {
@@ -1190,9 +1236,11 @@ public:
     if (!I->isBinaryOp())
       return nullptr;
     BinOpSameOpcodeHelper Converter(MainOp);
-    if (Converter.add(I) && Converter.add(MainOp) && !Converter.hasAltOp())
-      return MainOp;
-    return AltOp;
+    if (!Converter.add(I) || !Converter.add(MainOp))
+      return nullptr;
+    if (Converter.hasAltOp() && !isAltShuffle())
+      return nullptr;
+    return Converter.hasAltOp() ? AltOp : MainOp;
   }
 
   /// Checks if main/alt instructions are shift operations.
@@ -1237,9 +1285,63 @@ public:
   explicit operator bool() const { return valid(); }
 
   InstructionsState() = delete;
-  InstructionsState(Instruction *MainOp, Instruction *AltOp)
-      : MainOp(MainOp), AltOp(AltOp) {}
+  InstructionsState(Instruction *MainOp, Instruction *AltOp,
+                    bool HasCopyables = false)
+      : MainOp(MainOp), AltOp(AltOp), HasCopyables(HasCopyables) {}
   static InstructionsState invalid() { return {nullptr, nullptr}; }
+
+  bool isCopyableElement(Value *V) const {
+    assert(valid() && "InstructionsState is invalid.");
+    if (!HasCopyables)
+      return false;
+    if (isAltShuffle() || getOpcode() == Instruction::GetElementPtr)
+      return false;
+    auto *I = dyn_cast<Instruction>(V);
+    if (!I)
+      return !isa<PoisonValue>(V);
+    if (I->getParent() != MainOp->getParent() &&
+        (!isVectorLikeInstWithConstOps(I) ||
+         !isVectorLikeInstWithConstOps(MainOp)))
+      return true;
+    if (I->getOpcode() == MainOp->getOpcode())
+      return false;
+    if (!I->isBinaryOp())
+      return true;
+    BinOpSameOpcodeHelper Converter(MainOp);
+    return !Converter.add(I) || !Converter.add(MainOp) ||
+           Converter.hasAltOp() || !Converter.hasCandidateOpcode(getOpcode());
+  }
+
+  /// Checks if the value is non-schedulable.
+  bool isNonSchedulable(Value *V) const {
+    assert(valid() && "InstructionsState is invalid.");
+    auto *I = dyn_cast<Instruction>(V);
+    if (!HasCopyables)
+      return !I || isa<PHINode>(I) || isVectorLikeInstWithConstOps(I) ||
+             doesNotNeedToBeScheduled(V);
+    // MainOp for copyables always schedulable to correctly identify
+    // non-schedulable copyables.
+    if (isCopyableElement(V)) {
+      auto IsNonSchedulableCopyableElement = [this](Value *V) {
+        auto *I = dyn_cast<Instruction>(V);
+        return !I || isa<PHINode>(I) || I->getParent() != MainOp->getParent() ||
+               (doesNotNeedToBeScheduled(I) &&
+                // If the copyable instructions comes after MainOp
+                // (non-schedulable, but used in the block) - cannot vectorize
+                // it, will possibly generate use before def.
+                (isVectorLikeInstWithConstOps(I) || !MainOp->comesBefore(I)));
+      };
+
+      return IsNonSchedulableCopyableElement(V);
+    }
+    return !I || isa<PHINode>(I) || isVectorLikeInstWithConstOps(I) ||
+           doesNotNeedToBeScheduled(V);
+  }
+
+  bool areInstructionsWithCopyableElements() const {
+    assert(valid() && "InstructionsState is invalid.");
+    return HasCopyables;
+  }
 };
 
 std::pair<Instruction *, SmallVector<Value *>>
@@ -1917,6 +2019,7 @@ public:
     CompressEntryToData.clear();
     ExternalUses.clear();
     ExternalUsesAsOriginalScalar.clear();
+    ExternalUsesWithNonUsers.clear();
     for (auto &Iter : BlocksSchedules) {
       BlockScheduling *BS = Iter.second.get();
       BS->clear();
@@ -2899,9 +3002,6 @@ public:
       for (OperandDataVec &Ops : OpsVec)
         Ops.resize(NumLanes);
       for (unsigned Lane : seq<unsigned>(NumLanes)) {
-        Value *V = VL[Lane];
-        assert((isa<Instruction>(V) || isa<PoisonValue>(V)) &&
-               "Expected instruction or poison value");
         // Our tree has just 3 nodes: the root and two operands.
         // It is therefore trivial to get the APO. We only need to check the
         // opcode of V and whether the operand at OpIdx is the LHS or RHS
@@ -2912,17 +3012,24 @@ public:
         // Since operand reordering is performed on groups of commutative
         // operations or alternating sequences (e.g., +, -), we can safely tell
         // the inverse operations by checking commutativity.
-        if (isa<PoisonValue>(V)) {
+        auto *I = dyn_cast<Instruction>(VL[Lane]);
+        if (!I && isa<PoisonValue>(VL[Lane])) {
           for (unsigned OpIdx : seq<unsigned>(NumOperands))
             OpsVec[OpIdx][Lane] = {Operands[OpIdx][Lane], true, false};
           continue;
         }
-        auto [SelectedOp, Ops] = convertTo(cast<Instruction>(V), S);
-        // We cannot check commutativity by the converted instruction
-        // (SelectedOp) because isCommutative also examines def-use
-        // relationships.
-        bool IsInverseOperation =
-            !isCommutative(SelectedOp, cast<Instruction>(V));
+        bool IsInverseOperation = false;
+        if (S.isCopyableElement(VL[Lane])) {
+          // The value is a copyable element.
+          IsInverseOperation = !isCommutative(MainOp);
+        } else {
+          assert(I && "Expected instruction");
+          auto [SelectedOp, Ops] = convertTo(I, S);
+          // We cannot check commutativity by the converted instruction
+          // (SelectedOp) because isCommutative also examines def-use
+          // relationships.
+          IsInverseOperation = !isCommutative(SelectedOp, I);
+        }
         for (unsigned OpIdx : seq<unsigned>(ArgSize)) {
           bool APO = (OpIdx == 0) ? false : IsInverseOperation;
           OpsVec[OpIdx][Lane] = {Operands[OpIdx][Lane], APO, false};
@@ -3792,6 +3899,9 @@ private:
     /// reordering of operands during buildTreeRec() and vectorizeTree().
     SmallVector<ValueList, 2> Operands;
 
+    /// Copyable elements of the entry node.
+    SmallPtrSet<const Value *, 4> CopyableElements;
+
     /// MainOp and AltOp are recorded inside. S should be obtained from
     /// newTreeEntry.
     InstructionsState S = InstructionsState::invalid();
@@ -3820,11 +3930,7 @@ private:
     void setInterleave(unsigned Factor) { InterleaveFactor = Factor; }
 
     /// Marks the node as one that does not require scheduling.
-    void setDoesNotNeedToSchedule() {
-      assert(::doesNotNeedToSchedule(Scalars) &&
-             "Expected to not need scheduling");
-      DoesNotNeedToSchedule = true;
-    }
+    void setDoesNotNeedToSchedule() { DoesNotNeedToSchedule = true; }
     /// Returns true if the node is marked as one that does not require
     /// scheduling.
     bool doesNotNeedToSchedule() const { return DoesNotNeedToSchedule; }
@@ -3896,6 +4002,20 @@ private:
 
     bool hasState() const { return S.valid(); }
 
+    /// Add \p V to the list of copyable elements.
+    void addCopyableElement(Value *V) {
+      assert(S.isCopyableElement(V) && "Not a copyable element.");
+      CopyableElements.insert(V);
+    }
+
+    /// Returns true if \p V is a copyable element.
+    bool isCopyableElement(Value *V) const {
+      return CopyableElements.contains(V);
+    }
+
+    /// Returns true if any scalar in the list is a copyable element.
+    bool hasCopyableElements() const { return !CopyableElements.empty(); }
+
     /// When ReuseReorderShuffleIndices is empty it just returns position of \p
     /// V within vector of Scalars. Otherwise, try to remap on its reuse index.
     unsigned findLaneForValue(Value *V) const {
@@ -3968,6 +4088,8 @@ private:
       for (Value *V : Scalars)
         dbgs().indent(2) << *V << "\n";
       dbgs() << "State: ";
+      if (S && hasCopyableElements())
+        dbgs() << "[[Copyable]] ";
       switch (State) {
       case Vectorize:
         if (InterleaveFactor > 0) {
@@ -4145,12 +4267,20 @@ private:
         }
       }
     } else if (!Last->isGather()) {
-      if (doesNotNeedToSchedule(VL))
+      if (isa<PHINode>(S.getMainOp()) ||
+          isVectorLikeInstWithConstOps(S.getMainOp()) ||
+          (!S.areInstructionsWithCopyableElements() &&
+           doesNotNeedToSchedule(VL)) ||
+          all_of(VL, [&](Value *V) { return S.isNonSchedulable(V); }))
         Last->setDoesNotNeedToSchedule();
       SmallPtrSet<Value *, 4> Processed;
       for (Value *V : VL) {
         if (isa<PoisonValue>(V))
           continue;
+        if (S.isCopyableElement(V)) {
+          Last->addCopyableElement(V);
+          continue;
+        }
         auto It = ScalarToTreeEntries.find(V);
         if (It == ScalarToTreeEntries.end()) {
           ScalarToTreeEntries.try_emplace(V).first->getSecond().push_back(Last);
@@ -4162,16 +4292,14 @@ private:
         }
       }
       // Update the scheduler bundle to point to this TreeEntry.
-      assert((!Bundle.getBundle().empty() || isa<PHINode>(S.getMainOp()) ||
-              isVectorLikeInstWithConstOps(S.getMainOp()) ||
-              Last->doesNotNeedToSchedule()) &&
+      assert((!Bundle.getBundle().empty() || Last->doesNotNeedToSchedule()) &&
              "Bundle and VL out of sync");
       if (!Bundle.getBundle().empty()) {
 #if !defined(NDEBUG) || defined(EXPENSIVE_CHECKS)
         auto *BundleMember = Bundle.getBundle().begin();
         SmallPtrSet<Value *, 4> Processed;
         for (Value *V : VL) {
-          if (doesNotNeedToBeScheduled(V) || !Processed.insert(V).second)
+          if (S.isNonSchedulable(V) || !Processed.insert(V).second)
             continue;
           ++BundleMember;
         }
@@ -4280,7 +4408,8 @@ private:
   /// in general.
   ScalarsVectorizationLegality
   getScalarsVectorizationLegality(ArrayRef<Value *> VL, unsigned Depth,
-                                  const EdgeInfo &UserTreeIdx) const;
+                                  const EdgeInfo &UserTreeIdx,
+                                  bool TryCopyableElementsVectorization) const;
 
   /// Checks if the specified list of the instructions/values can be vectorized
   /// and fills required data before actual scheduling of the instructions.
@@ -4420,6 +4549,10 @@ private:
   /// extractelement instructions.
   SmallPtrSet<Value *, 4> ExternalUsesAsOriginalScalar;
 
+  /// A list of scalar to be extracted without specific user necause of too many
+  /// uses.
+  SmallPtrSet<Value *, 4> ExternalUsesWithNonUsers;
+
   /// Values used only by @llvm.assume calls.
   SmallPtrSet<const Value *, 32> EphValues;
 
@@ -4996,7 +5129,8 @@ private:
 
     /// Build a bundle from the ScheduleData nodes corresponding to the
     /// scalar instruction for each lane.
-    ScheduleBundle &buildBundle(ArrayRef<Value *> VL);
+    ScheduleBundle &buildBundle(ArrayRef<Value *> VL,
+                                const InstructionsState &S);
 
     /// Checks if a bundle of instructions can be scheduled, i.e. has no
     /// cyclic dependencies. This is only a dry-run, no instructions are
@@ -6727,7 +6861,7 @@ BoUpSLP::getReorderingData(const TreeEntry &TE, bool TopToBottom,
     return std::move(ResOrder);
   }
   if (TE.State == TreeEntry::StridedVectorize && !TopToBottom &&
-      (!TE.UserTreeIndex ||
+      (!TE.UserTreeIndex || !TE.UserTreeIndex.UserTE->hasState() ||
        !Instruction::isBinaryOp(TE.UserTreeIndex.UserTE->getOpcode())) &&
       (TE.ReorderIndices.empty() || isReverseOrder(TE.ReorderIndices)))
     return std::nullopt;
@@ -7038,10 +7172,11 @@ bool BoUpSLP::isProfitableToReorder() const {
          VectorizableTree.front()->getOpcode() == Instruction::ICmp))) &&
       VectorizableTree.front()->ReorderIndices.empty()) {
     // Check if the tree has only single store and single (unordered) load node,
-    // other nodes are phis or geps/binops, combined with phis, and/orsingle
+    // other nodes are phis or geps/binops, combined with phis, and/or single
     // gather load node
     bool HasPhis = false;
-    if (VectorizableTree.front()->getOpcode() == Instruction::PHI &&
+    if (VectorizableTree.front()->hasState() &&
+        VectorizableTree.front()->getOpcode() == Instruction::PHI &&
         VectorizableTree.front()->Scalars.size() == TinyVF &&
         VectorizableTree.front()->getNumOperands() > PhiOpsLimit)
       return false;
@@ -7049,6 +7184,8 @@ bool BoUpSLP::isProfitableToReorder() const {
     unsigned GatherLoads = 0;
     for (const std::unique_ptr<TreeEntry> &TE :
          ArrayRef(VectorizableTree).drop_front()) {
+      if (TE->State == TreeEntry::SplitVectorize)
+        continue;
       if (!TE->hasState()) {
         if (all_of(TE->Scalars, IsaPred<Constant, PHINode>) ||
             all_of(TE->Scalars, IsaPred<BinaryOperator, PHINode>))
@@ -7072,7 +7209,10 @@ bool BoUpSLP::isProfitableToReorder() const {
       if (TE->getOpcode() == Instruction::GetElementPtr ||
           Instruction::isBinaryOp(TE->getOpcode()))
         continue;
-      if (TE->getOpcode() != Instruction::PHI)
+      if (TE->getOpcode() != Instruction::PHI &&
+          (!TE->hasCopyableElements() ||
+           static_cast<unsigned>(count_if(TE->Scalars, IsaPred<PHINode>)) <
+               TE->Scalars.size() / 2))
         return true;
       if (VectorizableTree.front()->Scalars.size() == TinyVF &&
           TE->getNumOperands() > PhiOpsLimit)
@@ -7860,7 +8000,8 @@ void BoUpSLP::reorderBottomToTop(bool IgnoreReorder) {
 }
 
 Instruction *BoUpSLP::getRootEntryInstruction(const TreeEntry &Entry) const {
-  if ((Entry.getOpcode() == Instruction::Store ||
+  if (Entry.hasState() &&
+      (Entry.getOpcode() == Instruction::Store ||
        Entry.getOpcode() == Instruction::Load) &&
       Entry.State == TreeEntry::StridedVectorize &&
       !Entry.ReorderIndices.empty() && isReverseOrder(Entry.ReorderIndices))
@@ -7870,7 +8011,9 @@ Instruction *BoUpSLP::getRootEntryInstruction(const TreeEntry &Entry) const {
 
 void BoUpSLP::buildExternalUses(
     const ExtraValueToDebugLocsMap &ExternallyUsedValues) {
+  const size_t NumVectScalars = ScalarToTreeEntries.size() + 1;
   DenseMap<Value *, unsigned> ScalarToExtUses;
+  SmallPtrSet<Value *, 4> ExternalUsers;
   // Collect the values that we need to extract from the tree.
   for (auto &TEPtr : VectorizableTree) {
     TreeEntry *Entry = TEPtr.get();
@@ -7882,13 +8025,24 @@ void BoUpSLP::buildExternalUses(
     // For each lane:
     for (int Lane = 0, LE = Entry->Scalars.size(); Lane != LE; ++Lane) {
       Value *Scalar = Entry->Scalars[Lane];
-      if (!isa<Instruction>(Scalar))
+      if (!isa<Instruction>(Scalar) || Entry->isCopyableElement(Scalar))
         continue;
+
       // All uses must be replaced already? No need to do it again.
       auto It = ScalarToExtUses.find(Scalar);
       if (It != ScalarToExtUses.end() && !ExternalUses[It->second].User)
         continue;
 
+      if (Scalar->hasNUsesOrMore(NumVectScalars)) {
+        unsigned FoundLane = Entry->findLaneForValue(Scalar);
+        LLVM_DEBUG(dbgs() << "SLP: Need to extract from lane " << FoundLane
+                          << " from " << *Scalar << "for many users.\n");
+        It = ScalarToExtUses.try_emplace(Scalar, ExternalUses.size()).first;
+        ExternalUses.emplace_back(Scalar, nullptr, *Entry, FoundLane);
+        ExternalUsesWithNonUsers.insert(Scalar);
+        continue;
+      }
+
       // Check if the scalar is externally used as an extra arg.
       const auto ExtI = ExternallyUsedValues.find(Scalar);
       if (ExtI != ExternallyUsedValues.end()) {
@@ -7916,7 +8070,10 @@ void BoUpSLP::buildExternalUses(
           // Some in-tree scalars will remain as scalar in vectorized
           // instructions. If that is the case, the one in FoundLane will
           // be used.
-          if (all_of(UseEntries, [&](TreeEntry *UseEntry) {
+          if (!((Scalar->getType()->getScalarType()->isPointerTy() &&
+                 isa<LoadInst, StoreInst>(UserInst)) ||
+                isa<CallInst>(UserInst)) ||
+              all_of(UseEntries, [&](TreeEntry *UseEntry) {
                 return UseEntry->State == TreeEntry::ScatterVectorize ||
                        !doesInTreeUserNeedToExtract(
                            Scalar, getRootEntryInstruction(*UseEntry), TLI,
@@ -7946,6 +8103,7 @@ void BoUpSLP::buildExternalUses(
                           << ".\n");
         It = ScalarToExtUses.try_emplace(Scalar, ExternalUses.size()).first;
         ExternalUses.emplace_back(Scalar, U, *Entry, FoundLane);
+        ExternalUsesWithNonUsers.insert(Scalar);
         if (!U)
           break;
       }
@@ -9612,7 +9770,8 @@ static bool tryToFindDuplicates(SmallVectorImpl<Value *> &VL,
             PoisonValue::get(UniqueValues.front()->getType()));
         // Check that extended with poisons operations are still valid for
         // vectorization (div/rem are not allowed).
-        if (!getSameOpcode(PaddedUniqueValues, TLI).valid()) {
+        if (!S.areInstructionsWithCopyableElements() &&
+            !getSameOpcode(PaddedUniqueValues, TLI).valid()) {
           LLVM_DEBUG(dbgs() << "SLP: Scalar used twice in bundle.\n");
           ReuseShuffleIndices.clear();
           return false;
@@ -9761,13 +9920,95 @@ bool BoUpSLP::canBuildSplitNode(ArrayRef<Value *> VL,
 }
 
 namespace {
-/// Class accepts incoming list of values and generates the list of values
-/// for scheduling and list of operands for the new nodes.
+/// Class accepts incoming list of values, checks if it is able to model
+/// "copyable" values as compatible operations, and generates the list of values
+/// for scheduling and list of operands doe the new nodes.
 class InstructionsCompatibilityAnalysis {
   DominatorTree &DT;
   const DataLayout &DL;
   const TargetTransformInfo &TTI;
   const TargetLibraryInfo &TLI;
+  unsigned MainOpcode = 0;
+  Instruction *MainOp = nullptr;
+
+  /// Identifies the best candidate value, which represents main opcode
+  /// operation.
+  /// Currently the best candidate is the Add instruction with the parent
+  /// block with the highest DFS incoming number (block, that dominates other).
+  void findAndSetMainInstruction(ArrayRef<Value *> VL) {
+    BasicBlock *Parent = nullptr;
+    // Checks if the instruction has supported opcode.
+    auto IsSupportedOpcode = [](Instruction *I) {
+      return I && I->getOpcode() == Instruction::Add;
+    };
+    SmallDenseSet<Value *, 8> Operands;
+    for (Value *V : VL) {
+      auto *I = dyn_cast<Instruction>(V);
+      if (!I)
+        continue;
+      if (!DT.isReachableFromEntry(I->getParent()))
+        continue;
+      if (!MainOp) {
+        MainOp = I;
+        Parent = I->getParent();
+        Operands.insert(I->op_begin(), I->op_end());
+        continue;
+      }
+      if (Parent == I->getParent()) {
+        if (!IsSupportedOpcode(MainOp))
+          MainOp = I;
+        if (MainOp->getOpcode() == I->getOpcode() &&
+            doesNotNeedToBeScheduled(MainOp) && !doesNotNeedToBeScheduled(I))
+          MainOp = I;
+        Operands.insert(I->op_begin(), I->op_end());
+        continue;
+      }
+      auto *NodeA = DT.getNode(Parent);
+      auto *NodeB = DT.getNode(I->getParent());
+      assert(NodeA && "Should only process reachable instructions");
+      assert(NodeB && "Should only process reachable instructions");
+      assert((NodeA == NodeB) ==
+                 (NodeA->getDFSNumIn() == NodeB->getDFSNumIn()) &&
+             "Different nodes should have different DFS numbers");
+      if (NodeA->getDFSNumIn() < NodeB->getDFSNumIn()) {
+        MainOp = I;
+        Parent = I->getParent();
+        Operands.clear();
+        Operands.insert(I->op_begin(), I->op_end());
+      }
+    }
+    if (!IsSupportedOpcode(MainOp) || Operands.contains(MainOp)) {
+      MainOp = nullptr;
+      return;
+    }
+    MainOpcode = MainOp->getOpcode();
+  }
+
+  /// Returns the idempotent value for the \p MainOp with the detected \p
+  /// MainOpcode. For Add, returns 0. For Or, it should choose between false and
+  /// the operand itself, since V or V == V.
+  Value *selectBestIdempotentValue() const {
+    assert(MainOpcode == Instruction::Add && "Unsupported opcode");
+    return ConstantExpr::getBinOpIdentity(MainOpcode, MainOp->getType(),
+                                          !MainOp->isCommutative());
+  }
+
+  /// Returns the value and operands for the \p V, considering if it is original
+  /// instruction and its actual operands should be returned, or it is a
+  /// copyable element and its should be represented as idempotent instruction.
+  SmallVector<Value *> getOperands(const InstructionsState &S, Value *V) const {
+    if (isa<PoisonValue>(V))
+      return {V, V};
+    if (!S.isCopyableElement(V))
+      return convertTo(cast<Instruction>(V), S).second;
+    switch (MainOpcode) {
+    case Instruction::Add:
+      return {V, selectBestIdempotentValue()};
+    default:
+      break;
+    }
+    llvm_unreachable("Unsupported opcode");
+  }
 
   /// Builds operands for the original instructions.
   void
@@ -9928,22 +10169,165 @@ public:
                                     const TargetLibraryInfo &TLI)
       : DT(DT), DL(DL), TTI(TTI), TLI(TLI) {}
 
+  InstructionsState
+  buildInstructionsState(ArrayRef<Value *> VL, const BoUpSLP &R,
+                         bool TryCopyableElementsVectorization,
+                         bool WithProfitabilityCheck = false,
+                         bool SkipSameCodeCheck = false) {
+    InstructionsState S = (SkipSameCodeCheck || !allSameBlock(VL))
+                              ? InstructionsState::invalid()
+                              : getSameOpcode(VL, TLI);
+    if (S)
+      return S;
+    if (!VectorizeCopyableElements || !TryCopyableElementsVectorization)
+      return S;
+    findAndSetMainInstruction(VL);
+    if (!MainOp)
+      return InstructionsState::invalid();
+    S = InstructionsState(MainOp, MainOp, /*HasCopyables=*/true);
+    // TODO: Remove this check once support for schulable copyables is landed.
+    if (any_of(VL, [&](Value *V) {
+          return S.isCopyableElement(V) && !S.isNonSchedulable(V);
+        }))
+      return InstructionsState::invalid();
+
+    if (!WithProfitabilityCheck)
+      return S;
+    // Check if it is profitable to vectorize the instruction.
+    SmallVector<BoUpSLP::ValueList> Operands = buildOperands(S, VL);
+    auto BuildCandidates =
+        [](SmallVectorImpl<std::pair<Value *, Value *>> &Candidates, Value *V1,
+           Value *V2) {
+          if (V1 != V2 && isa<PHINode>(V1))
+            return;
+          auto *I1 = dyn_cast<Instruction>(V1);
+          auto *I2 = dyn_cast<Instruction>(V2);
+          if (I1 && I2 && I1->getOpcode() == I2->getOpcode() &&
+              I1->getParent() != I2->getParent())
+            return;
+          Candidates.emplace_back(V1, (I1 || I2) ? V2 : V1);
+        };
+    if (VL.size() == 2) {
+      // Check if the operands allow better vectorization.
+      SmallVector<std::pair<Value *, Value *>, 4> Candidates1, Candidates2;
+      BuildCandidates(Candidates1, Operands[0][0], Operands[0][1]);
+      BuildCandidates(Candidates2, Operands[1][0], Operands[1][1]);
+      bool Res = !Candidates1.empty() && !Candidates2.empty() &&
+                 R.findBestRootPair(Candidates1) &&
+                 R.findBestRootPair(Candidates2);
+      if (!Res && isCommutative(MainOp)) {
+        Candidates1.clear();
+        Candidates2.clear();
+        BuildCandidates(Candidates1, Operands[0][0], Operands[1][1]);
+        BuildCandidates(Candidates2, Operands[1][0], Operands[0][1]);
+        Res = !Candidates1.empty() && !Candidates2.empty() &&
+              R.findBestRootPair(Candidates1) &&
+              R.findBestRootPair(Candidates2);
+      }
+      if (!Res)
+        return InstructionsState::invalid();
+      return S;
+    }
+    assert(Operands.size() == 2 && "Unexpected number of operands!");
+    unsigned CopyableNum =
+        count_if(VL, [&](Value *V) { return S.isCopyableElement(V); });
+    if (CopyableNum < VL.size() / 2)
+      return S;
+    // Too many phi copyables - exit.
+    const unsigned Limit = VL.size() / 24;
+    if ((CopyableNum >= VL.size() - Limit ||
+         (CopyableNum >= VL.size() - 1 && VL.size() > 4) ||
+         CopyableNum >= MaxPHINumOperands) &&
+        all_of(VL, [&](Value *V) {
+          return isa<PHINode>(V) || !S.isCopyableElement(V);
+        }))
+      return InstructionsState::invalid();
+    // Check profitability if number of copyables > VL.size() / 2.
+    // 1. Reorder operands for better matching.
+    if (isCommutative(MainOp)) {
+      for (auto &Ops : Operands) {
+        // Make instructions the first operands.
+        if (!isa<Instruction>(Ops.front()) && isa<Instruction>(Ops.back())) {
+          std::swap(Ops.front(), Ops.back());
+          continue;
+        }
+        // Make constants the second operands.
+        if (isa<Constant>(Ops.front())) {
+          std::swap(Ops.front(), Ops.back());
+          continue;
+        }
+      }
+    }
+    // 2. Check, if operands can be vectorized.
+    if (count_if(Operands.back(), IsaPred<Instruction>) > 1)
+      return InstructionsState::invalid();
+    auto CheckOperand = [&](ArrayRef<Value *> Ops) {
+      if (allConstant(Ops) || isSplat(Ops))
+        return true;
+      // Check if it is "almost" splat, i.e. has >= 4 elements and only single
+      // one is different.
+      constexpr unsigned Limit = 4;
+      if (Operands.front().size() >= Limit) {
+        SmallDenseMap<const Value *, unsigned> Counters;
+        for (Value *V : Ops) {
+          if (isa<UndefValue>(V))
+            continue;
+          ++Counters[V];
+        }
+        if (Counters.size() == 2 &&
+            any_of(Counters, [&](const std::pair<const Value *, unsigned> &C) {
+              return C.second == 1;
+            }))
+          return true;
+      }
+      // First operand not a constant or splat? Last attempt - check for
+      // potential vectorization.
+      InstructionsCompatibilityAnalysis Analysis(DT, DL, TTI, TLI);
+      InstructionsState OpS = Analysis.buildInstructionsState(
+          Ops, R, /*TryCopyableElementsVectorization=*/true);
+      if (!OpS || (OpS.getOpcode() == Instruction::PHI && !allSameBlock(Ops)))
+        return false;
+      unsigned CopyableNum =
+          count_if(Ops, [&](Value *V) { return OpS.isCopyableElement(V); });
+      return CopyableNum <= VL.size() / 2;
+    };
+    if (!CheckOperand(Operands.front()))
+      return InstructionsState::invalid();
+
+    return S;
+  }
+
   SmallVector<BoUpSLP::ValueList> buildOperands(const InstructionsState &S,
                                                 ArrayRef<Value *> VL) {
     assert(S && "Invalid state!");
     SmallVector<BoUpSLP::ValueList> Operands;
-    buildOriginalOperands(S, VL, Operands);
+    if (S.areInstructionsWithCopyableElements()) {
+      MainOp = S.getMainOp();
+      MainOpcode = S.getOpcode();
+      Operands.assign(MainOp->getNumOperands(),
+                      BoUpSLP::ValueList(VL.size(), nullptr));
+      for (auto [Idx, V] : enumerate(VL)) {
+        SmallVector<Value *> OperandsForValue = getOperands(S, V);
+        for (auto [OperandIdx, Operand] : enumerate(OperandsForValue))
+          Operands[OperandIdx][Idx] = Operand;
+      }
+    } else {
+      buildOriginalOperands(S, VL, Operands);
+    }
     return Operands;
   }
 };
 } // namespace
 
-BoUpSLP::ScalarsVectorizationLegality
-BoUpSLP::getScalarsVectorizationLegality(ArrayRef<Value *> VL, unsigned Depth,
-                                         const EdgeInfo &UserTreeIdx) const {
+BoUpSLP::ScalarsVectorizationLegality BoUpSLP::getScalarsVectorizationLegality(
+    ArrayRef<Value *> VL, unsigned Depth, const EdgeInfo &UserTreeIdx,
+    bool TryCopyableElementsVectorization) const {
   assert((allConstant(VL) || allSameType(VL)) && "Invalid types!");
 
-  InstructionsState S = getSameOpcode(VL, *TLI);
+  InstructionsCompatibilityAnalysis Analysis(*DT, *DL, *TTI, *TLI);
+  InstructionsState S = Analysis.buildInstructionsState(
+      VL, *this, TryCopyableElementsVectorization,
+      /*WithProfitabilityCheck=*/true, TryCopyableElementsVectorization);
 
   // Don't go into catchswitch blocks, which can happen with PHIs.
   // Such blocks can only have PHIs and the catchswitch.  There is no
@@ -10066,7 +10450,7 @@ BoUpSLP::getScalarsVectorizationLegality(ArrayRef<Value *> VL, unsigned Depth,
   bool IsScatterVectorizeUserTE =
       UserTreeIdx.UserTE &&
       UserTreeIdx.UserTE->State == TreeEntry::ScatterVectorize;
-  bool AreAllSameBlock = S && allSameBlock(VL);
+  bool AreAllSameBlock = S.valid();
   bool AreScatterAllGEPSameBlock =
       (IsScatterVectorizeUserTE && VL.front()->getType()->isPointerTy() &&
        VL.size() > 2 &&
@@ -10091,12 +10475,18 @@ BoUpSLP::getScalarsVectorizationLegality(ArrayRef<Value *> VL, unsigned Depth,
       NotProfitableForVectorization(VL)) {
     if (!S) {
       LLVM_DEBUG(dbgs() << "SLP: Try split and if failed, gathering due to "
-                           "C,S,B,O, small shuffle. \n");
+                           "C,S,B,O, small shuffle. \n";
+                 dbgs() << "[";
+                 interleaveComma(VL, dbgs(), [&](Value *V) { dbgs() << *V; });
+                 dbgs() << "]\n");
       return ScalarsVectorizationLegality(S, /*IsLegal=*/false,
                                           /*TryToFindDuplicates=*/true,
                                           /*TrySplitVectorize=*/true);
     }
-    LLVM_DEBUG(dbgs() << "SLP: Gathering due to C,S,B,O, small shuffle. \n");
+    LLVM_DEBUG(dbgs() << "SLP: Gathering due to C,S,B,O, small shuffle. \n";
+               dbgs() << "[";
+               interleaveComma(VL, dbgs(), [&](Value *V) { dbgs() << *V; });
+               dbgs() << "]\n");
     return ScalarsVectorizationLegality(S, /*IsLegal=*/false);
   }
 
@@ -10242,9 +10632,29 @@ void BoUpSLP::buildTreeRec(ArrayRef<Value *> VLRef, unsigned Depth,
     return true;
   };
 
-  ScalarsVectorizationLegality Legality =
-      getScalarsVectorizationLegality(VL, Depth, UserTreeIdx);
-  const InstructionsState &S = Legality.getInstructionsState();
+  auto AreOnlyConstsWithPHIs = [](ArrayRef<Value *> VL) {
+    bool AreConsts = false;
+    for (Value *V : VL) {
+      if (isa<PoisonValue>(V))
+        continue;
+      if (isa<Constant>(V)) {
+        AreConsts = true;
+        continue;
+      }
+      if (!isa<PHINode>(V))
+        return false;
+    }
+    return AreConsts;
+  };
+  if (AreOnlyConstsWithPHIs(VL)) {
+    LLVM_DEBUG(dbgs() << "SLP: Gathering due to all constants and PHIs.\n");
+    newGatherTreeEntry(VL, InstructionsState::invalid(), UserTreeIdx);
+    return;
+  }
+
+  ScalarsVectorizationLegality Legality = getScalarsVectorizationLegality(
+      VL, Depth, UserTreeIdx, /*TryCopyableElementsVectorization=*/false);
+  InstructionsState S = Legality.getInstructionsState();
   if (!Legality.isLegal()) {
     if (Legality.trySplitVectorize()) {
       auto [MainOp, AltOp] = getMainAltOpsNoStateVL(VL);
@@ -10252,11 +10662,18 @@ void BoUpSLP::buildTreeRec(ArrayRef<Value *> VLRef, unsigned Depth,
       if (MainOp && AltOp && TrySplitNode(InstructionsState(MainOp, AltOp)))
         return;
     }
-    if (Legality.tryToFindDuplicates())
-      tryToFindDuplicates(VL, ReuseShuffleIndices, *TTI, *TLI, S, UserTreeIdx);
+    if (!S)
+      Legality = getScalarsVectorizationLegality(
+          VL, Depth, UserTreeIdx, /*TryCopyableElementsVectorization=*/true);
+    if (!Legality.isLegal()) {
+      if (Legality.tryToFindDuplicates())
+        tryToFindDuplicates(VL, ReuseShuffleIndices, *TTI, *TLI, S,
+                            UserTreeIdx);
 
-    newGatherTreeEntry(VL, S, UserTreeIdx, ReuseShuffleIndices);
-    return;
+      newGatherTreeEntry(VL, S, UserTreeIdx, ReuseShuffleIndices);
+      return;
+    }
+    S = Legality.getInstructionsState();
   }
 
   // FIXME: investigate if there are profitable cases for VL.size() <= 4.
@@ -13024,7 +13441,8 @@ BoUpSLP::getEntryCost(const TreeEntry *E, ArrayRef<Value *> VectorizedVals,
   assert(E->getOpcode() &&
          ((allSameType(VL) && allSameBlock(VL)) ||
           (E->getOpcode() == Instruction::GetElementPtr &&
-           E->getMainOp()->getType()->isPointerTy())) &&
+           E->getMainOp()->getType()->isPointerTy()) ||
+          E->hasCopyableElements()) &&
          "Invalid VL");
   Instruction *VL0 = E->getMainOp();
   unsigned ShuffleOrOp =
@@ -13036,6 +13454,7 @@ BoUpSLP::getEntryCost(const TreeEntry *E, ArrayRef<Value *> VectorizedVals,
   SmallBitVector UsedScalars(Sz, false);
   for (unsigned I = 0; I < Sz; ++I) {
     if (isa<Instruction>(UniqueValues[I]) &&
+        !E->isCopyableElement(UniqueValues[I]) &&
         getTreeEntries(UniqueValues[I]).front() == E)
       continue;
     UsedScalars.set(I);
@@ -14075,15 +14494,45 @@ bool BoUpSLP::isTreeTinyAndNotFullyVectorizable(bool ForReduction) const {
 
   // If the tree contains only phis, buildvectors, split nodes and
   // small nodes with reuses, we can skip it.
+  SmallVector<const TreeEntry *> StoreLoadNodes;
+  unsigned NumGathers = 0;
+  constexpr int LimitTreeSize = 36;
   if (!ForReduction && !SLPCostThreshold.getNumOccurrences() &&
-      all_of(VectorizableTree, [](const std::unique_ptr<TreeEntry> &TE) {
-        return TE->State == TreeEntry::SplitVectorize ||
-               (TE->isGather() &&
-                none_of(TE->Scalars, IsaPred<ExtractElementInst>)) ||
-               (TE->hasState() && (TE->getOpcode() == Instruction::PHI ||
-                                   (!TE->ReuseShuffleIndices.empty() &&
-                                    TE->Scalars.size() == 2)));
-      }))
+      all_of(VectorizableTree,
+             [&](const std::unique_ptr<TreeEntry> &TE) {
+               if (!TE->isGather() && TE->hasState() &&
+                   (TE->getOpcode() == Instruction::Load ||
+                    TE->getOpcode() == Instruction::Store)) {
+                 StoreLoadNodes.push_back(TE.get());
+                 return true;
+               }
+               if (TE->isGather())
+                 ++NumGathers;
+               return TE->State == TreeEntry::SplitVectorize ||
+                      (TE->Idx == 0 && TE->Scalars.size() == 2 &&
+                       TE->hasState() && TE->getOpcode() == Instruction::ICmp &&
+                       VectorizableTree.size() > LimitTreeSize) ||
+                      (TE->isGather() &&
+                       none_of(TE->Scalars, IsaPred<ExtractElementInst>)) ||
+                      (TE->hasState() &&
+                       (TE->getOpcode() == Instruction::PHI ||
+                        (TE->hasCopyableElements() &&
+                         static_cast<unsigned>(count_if(
+                             TE->Scalars, IsaPred<PHINode, Constant>)) >=
+                             TE->Scalars.size() / 2) ||
+                        ((!TE->ReuseShuffleIndices.empty() ||
+                          !TE->ReorderIndices.empty() || TE->isAltShuffle()) &&
+                         TE->Scalars.size() == 2)));
+             }) &&
+      (StoreLoadNodes.empty() ||
+       (VectorizableTree.size() > LimitTreeSize * StoreLoadNodes.size() &&
+        (NumGathers > 0 || none_of(StoreLoadNodes, [&](const TreeEntry *TE) {
+           return TE->getOpcode() == Instruction::Store ||
+                  all_of(TE->Scalars, [&](Value *V) {
+                    return !isa<LoadInst>(V) ||
+                           areAllUsersVectorized(cast<Instruction>(V));
+                  });
+         })))))
     return true;
 
   // We can vectorize the tree if its size is greater than or equal to the
@@ -14826,6 +15275,7 @@ InstructionCost BoUpSLP::getTreeCost(ArrayRef<Value *> VectorizedVals,
         bool IsProfitablePHIUser =
             (KeepScalar || (ScalarCost - ExtraCost <= TTI::TCC_Basic &&
                             VectorizableTree.front()->Scalars.size() > 2)) &&
+            VectorizableTree.front()->hasState() &&
             VectorizableTree.front()->getOpcode() == Instruction::PHI &&
             !Inst->hasNUsesOrMore(UsesLimit) &&
             none_of(Inst->users(),
@@ -15276,7 +15726,8 @@ BoUpSLP::isGatherShuffledSingleRegisterEntry(
   const BasicBlock *TEInsertBlock = nullptr;
   // Main node of PHI entries keeps the correct order of operands/incoming
   // blocks.
-  if (auto *PHI = dyn_cast<PHINode>(TEUseEI.UserTE->getMainOp());
+  if (auto *PHI = dyn_cast_or_null<PHINode>(
+          TEUseEI.UserTE->hasState() ? TEUseEI.UserTE->getMainOp() : nullptr);
       PHI && TEUseEI.UserTE->State != TreeEntry::SplitVectorize) {
     TEInsertBlock = PHI->getIncomingBlock(TEUseEI.EdgeIdx);
     TEInsertPt = TEInsertBlock->getTerminator();
@@ -15375,7 +15826,8 @@ BoUpSLP::isGatherShuffledSingleRegisterEntry(
              "Expected only single user of a gather node.");
       const EdgeInfo &UseEI = TEPtr->UserTreeIndex;
 
-      PHINode *UserPHI = UseEI.UserTE->State != TreeEntry::SplitVectorize
+      PHINode *UserPHI = (UseEI.UserTE->State != TreeEntry::SplitVectorize &&
+                          UseEI.UserTE->hasState())
                              ? dyn_cast<PHINode>(UseEI.UserTE->getMainOp())
                              : nullptr;
       Instruction *InsertPt =
@@ -15388,7 +15840,8 @@ BoUpSLP::isGatherShuffledSingleRegisterEntry(
              TEUseEI.UserTE->isAltShuffle()) &&
             all_of(TEUseEI.UserTE->Scalars, isUsedOutsideBlock)) {
           if (UseEI.UserTE->State != TreeEntry::Vectorize ||
-              (UseEI.UserTE->getOpcode() == Instruction::PHI &&
+              (UseEI.UserTE->hasState() &&
+               UseEI.UserTE->getOpcode() == Instruction::PHI &&
                !UseEI.UserTE->isAltShuffle()) ||
               !all_of(UseEI.UserTE->Scalars, isUsedOutsideBlock))
             continue;
@@ -16009,25 +16462,32 @@ Instruction &BoUpSLP::getLastInstructionInBundle(const TreeEntry *E) {
   Instruction *Res = nullptr;
   // Get the basic block this bundle is in. All instructions in the bundle
   // should be in this block (except for extractelement-like instructions with
-  // constant indices or gathered loads).
-  auto *Front = E->getMainOp();
+  // constant indices or gathered loads or copyables).
+  Instruction *Front;
+  unsigned Opcode;
+  if (E->hasState()) {
+    Front = E->getMainOp();
+    Opcode = E->getOpcode();
+  } else {
+    Front = cast<Instruction>(*find_if(E->Scalars, IsaPred<Instruction>));
+    Opcode = Front->getOpcode();
+  }
   auto *BB = Front->getParent();
-  assert(((GatheredLoadsEntriesFirst.has_value() &&
-           E->getOpcode() == Instruction::Load && E->isGather() &&
-           E->Idx < *GatheredLoadsEntriesFirst) ||
-          E->State == TreeEntry::SplitVectorize ||
-          all_of(E->Scalars,
-                 [=](Value *V) -> bool {
-                   if (E->getOpcode() == Instruction::GetElementPtr &&
-                       !isa<GetElementPtrInst>(V))
-                     return true;
-                   auto *I = dyn_cast<Instruction>(V);
-                   return !I || !E->getMatchingMainOpOrAltOp(I) ||
-                          I->getParent() == BB ||
-                          isVectorLikeInstWithConstOps(I);
-                 })) &&
-         "Expected gathered loads or GEPs or instructions from same basic "
-         "block.");
+  assert(
+      ((GatheredLoadsEntriesFirst.has_value() && Opcode == Instruction::Load &&
+        E->isGather() && E->Idx < *GatheredLoadsEntriesFirst) ||
+       E->State == TreeEntry::SplitVectorize || E->hasCopyableElements() ||
+       all_of(E->Scalars,
+              [=](Value *V) -> bool {
+                if (Opcode == Instruction::GetElementPtr &&
+                    !isa<GetElementPtrInst>(V))
+                  return true;
+                auto *I = dyn_cast<Instruction>(V);
+                return !I || !E->getMatchingMainOpOrAltOp(I) ||
+                       I->getParent() == BB || isVectorLikeInstWithConstOps(I);
+              })) &&
+      "Expected gathered loads or GEPs or instructions from same basic "
+      "block.");
 
   auto FindLastInst = [&]() {
     Instruction *LastInst = Front;
@@ -16035,18 +16495,20 @@ Instruction &BoUpSLP::getLastInstructionInBundle(const TreeEntry *E) {
       auto *I = dyn_cast<Instruction>(V);
       if (!I)
         continue;
+      if (E->isCopyableElement(I))
+        continue;
       if (LastInst->getParent() == I->getParent()) {
         if (LastInst->comesBefore(I))
           LastInst = I;
         continue;
       }
-      assert(((E->getOpcode() == Instruction::GetElementPtr &&
+      assert(((Opcode == Instruction::GetElementPtr &&
                !isa<GetElementPtrInst>(I)) ||
               E->State == TreeEntry::SplitVectorize ||
               (isVectorLikeInstWithConstOps(LastInst) &&
                isVectorLikeInstWithConstOps(I)) ||
               (GatheredLoadsEntriesFirst.has_value() &&
-               E->getOpcode() == Instruction::Load && E->isGather() &&
+               Opcode == Instruction::Load && E->isGather() &&
                E->Idx < *GatheredLoadsEntriesFirst)) &&
              "Expected vector-like or non-GEP in GEP node insts only.");
       if (!DT->isReachableFromEntry(LastInst->getParent())) {
@@ -16075,16 +16537,18 @@ Instruction &BoUpSLP::getLastInstructionInBundle(const TreeEntry *E) {
       auto *I = dyn_cast<Instruction>(V);
       if (!I)
         continue;
+      if (E->isCopyableElement(I))
+        continue;
       if (FirstInst->getParent() == I->getParent()) {
         if (I->comesBefore(FirstInst))
           FirstInst = I;
         continue;
       }
-      assert(((E->getOpcode() == Instruction::GetElementPtr &&
-              !isa<GetElementPtrInst>(I)) ||
-             (isVectorLikeInstWithConstOps(FirstInst) &&
-              isVectorLikeInstWithConstOps(I))) &&
-                 "Expected vector-like or non-GEP in GEP node insts only.");
+      assert(((Opcode == Instruction::GetElementPtr &&
+               !isa<GetElementPtrInst>(I)) ||
+              (isVectorLikeInstWithConstOps(FirstInst) &&
+               isVectorLikeInstWithConstOps(I))) &&
+             "Expected vector-like or non-GEP in GEP node insts only.");
       if (!DT->isReachableFromEntry(FirstInst->getParent())) {
         FirstInst = I;
         continue;
@@ -16122,7 +16586,7 @@ Instruction &BoUpSLP::getLastInstructionInBundle(const TreeEntry *E) {
   // Set insertpoint for gathered loads to the very first load.
   if (GatheredLoadsEntriesFirst.has_value() &&
       E->Idx >= *GatheredLoadsEntriesFirst && !E->isGather() &&
-      E->getOpcode() == Instruction::Load) {
+      Opcode == Instruction::Load) {
     Res = FindFirstInst();
     EntryToLastInstruction.try_emplace(E, Res);
     return *Res;
@@ -16139,7 +16603,8 @@ Instruction &BoUpSLP::getLastInstructionInBundle(const TreeEntry *E) {
       return nullptr;
     for (Value *V : E->Scalars) {
       auto *I = dyn_cast<Instruction>(V);
-      if (!I || isa<PHINode>(I) || doesNotNeedToBeScheduled(I))
+      if (!I || isa<PHINode>(I) ||
+          (!E->isCopyableElement(I) && doesNotNeedToBeScheduled(I)))
         continue;
       ArrayRef<ScheduleBundle *> Bundles = It->second->getScheduleBundles(I);
       if (Bundles.empty())
@@ -16153,13 +16618,13 @@ Instruction &BoUpSLP::getLastInstructionInBundle(const TreeEntry *E) {
   };
   const ScheduleBundle *Bundle = FindScheduleBundle(E);
   if (!E->isGather() && !Bundle) {
-    if ((E->getOpcode() == Instruction::GetElementPtr &&
+    if ((Opcode == Instruction::GetElementPtr &&
          any_of(E->Scalars,
                 [](Value *V) {
                   return !isa<GetElementPtrInst>(V) && isa<Instruction>(V);
                 })) ||
-        all_of(E->Scalars, [](Value *V) {
-          return isa<PoisonValue>(V) ||
+        all_of(E->Scalars, [&](Value *V) {
+          return isa<PoisonValue>(V) || E->isCopyableElement(V) ||
                  (!isVectorLikeInstWithConstOps(V) && isUsedOutsideBlock(V));
         }))
       Res = FindLastInst();
@@ -18640,6 +19105,7 @@ Value *BoUpSLP::vectorizeTree(
         TE->UserTreeIndex.UserTE->State == TreeEntry::Vectorize &&
         (TE->UserTreeIndex.UserTE->getOpcode() != Instruction::PHI ||
          TE->UserTreeIndex.UserTE->isAltShuffle()) &&
+        !TE->UserTreeIndex.UserTE->hasCopyableElements() &&
         all_of(TE->UserTreeIndex.UserTE->Scalars,
                [](Value *V) { return isUsedOutsideBlock(V); })) {
       Instruction &LastInst =
@@ -18903,7 +19369,7 @@ Value *BoUpSLP::vectorizeTree(
         continue;
       assert(
           (ExternallyUsedValues.count(Scalar) ||
-           Scalar->hasNUsesOrMore(UsesLimit) ||
+          ExternalUsesWithNonUsers.count(Scalar) ||
            ExternalUsesAsOriginalScalar.contains(Scalar) ||
            any_of(
                Scalar->users(),
@@ -19182,7 +19648,7 @@ Value *BoUpSLP::vectorizeTree(
       if (auto *EE = dyn_cast<ExtractElementInst>(Scalar);
           EE && IgnoredExtracts.contains(EE))
         continue;
-      if (isa<PoisonValue>(Scalar))
+      if (!isa<Instruction>(Scalar) || Entry->isCopyableElement(Scalar))
         continue;
 #ifndef NDEBUG
       Type *Ty = Scalar->getType();
@@ -19424,12 +19890,15 @@ void BoUpSLP::optimizeGatherSequence() {
 }
 
 BoUpSLP::ScheduleBundle &
-BoUpSLP::BlockScheduling::buildBundle(ArrayRef<Value *> VL) {
+BoUpSLP::BlockScheduling::buildBundle(ArrayRef<Value *> VL,
+                                      const InstructionsState &S) {
   auto &BundlePtr =
       ScheduledBundlesList.emplace_back(std::make_unique<ScheduleBundle>());
   for (Value *V : VL) {
     if (doesNotNeedToBeScheduled(V))
       continue;
+    if (S.isCopyableElement(V))
+      continue;
     ScheduleData *BundleMember = getScheduleData(V);
     assert(BundleMember && "no ScheduleData for bundle member "
                            "(maybe not in same basic block)");
@@ -19450,10 +19919,19 @@ BoUpSLP::BlockScheduling::tryScheduleBundle(ArrayRef<Value *> VL, BoUpSLP *SLP,
                                             const InstructionsState &S) {
   // No need to schedule PHIs, insertelement, extractelement and extractvalue
   // instructions.
+  bool HasCopyables = S.areInstructionsWithCopyableElements();
   if (isa<PHINode>(S.getMainOp()) ||
-      isVectorLikeInstWithConstOps(S.getMainOp()) || doesNotNeedToSchedule(VL))
+      isVectorLikeInstWithConstOps(S.getMainOp()) ||
+      (!HasCopyables && doesNotNeedToSchedule(VL)) ||
+      all_of(VL, [&](Value *V) { return S.isNonSchedulable(V); }))
     return nullptr;
 
+  // TODO Remove once full support for copyables is landed.
+  assert(all_of(VL,
+                [&](Value *V) {
+                  return !S.isCopyableElement(V) || S.isNonSchedulable(V);
+                }) &&
+         "Copyable elements should not be schedulable");
   // Initialize the instruction bundle.
   Instruction *OldScheduleEnd = ScheduleEnd;
   LLVM_DEBUG(dbgs() << "SLP:  bundle: " << *S.getMainOp() << "\n");
@@ -19499,7 +19977,7 @@ BoUpSLP::BlockScheduling::tryScheduleBundle(ArrayRef<Value *> VL, BoUpSLP *SLP,
   // Make sure that the scheduling region contains all
   // instructions of the bundle.
   for (Value *V : VL) {
-    if (doesNotNeedToBeScheduled(V))
+    if (doesNotNeedToBeScheduled(V) || S.isCopyableElement(V))
       continue;
     if (!extendSchedulingRegion(V, S)) {
       // If the scheduling region got new instructions at the lower end (or it
@@ -19516,7 +19994,7 @@ BoUpSLP::BlockScheduling::tryScheduleBundle(ArrayRef<Value *> VL, BoUpSLP *SLP,
 
   bool ReSchedule = false;
   for (Value *V : VL) {
-    if (doesNotNeedToBeScheduled(V))
+    if (doesNotNeedToBeScheduled(V) || S.isCopyableElement(V))
       continue;
     ScheduleData *BundleMember = getScheduleData(V);
     assert(BundleMember &&
@@ -19541,7 +20019,7 @@ BoUpSLP::BlockScheduling::tryScheduleBundle(ArrayRef<Value *> VL, BoUpSLP *SLP,
     ReSchedule = true;
   }
 
-  ScheduleBundle &Bundle = buildBundle(VL);
+  ScheduleBundle &Bundle = buildBundle(VL, S);
   TryScheduleBundleImpl(ReSchedule, Bundle);
   if (!Bundle.isReady()) {
     for (ScheduleData *BD : Bundle.getBundle()) {
@@ -19558,7 +20036,7 @@ BoUpSLP::BlockScheduling::tryScheduleBundle(ArrayRef<Value *> VL, BoUpSLP *SLP,
     }
     ScheduledBundlesList.pop_back();
     for (Value *V : VL) {
-      if (doesNotNeedToBeScheduled(V))
+      if (doesNotNeedToBeScheduled(V) || S.isCopyableElement(V))
         continue;
       ScheduledBundles.find(cast<Instruction>(V))->getSecond().pop_back();
     }
@@ -20187,7 +20665,7 @@ bool BoUpSLP::collectValuesToDemote(
   };
   if (E.isGather() || !Visited.insert(&E).second ||
       any_of(E.Scalars, [&](Value *V) {
-        return !isa<PoisonValue>(V) && all_of(V->users(), [&](User *U) {
+        return !isa<Constant>(V) && all_of(V->users(), [&](User *U) {
           return isa<InsertElementInst>(U) && !isVectorized(U);
         });
       }))
@@ -20555,9 +21033,10 @@ void BoUpSLP::computeMinimumValueSizes() {
                     if (!isa<CastInst, BinaryOperator, FreezeInst, PHINode,
                              SelectInst>(U) ||
                         isa<SIToFPInst, UIToFPInst>(U) ||
-                        !isa<CastInst, BinaryOperator, FreezeInst, PHINode,
-                             SelectInst>(UserTE->getMainOp()) ||
-                        isa<SIToFPInst, UIToFPInst>(UserTE->getMainOp()))
+                        (UserTE->hasState() &&
+                         (!isa<CastInst, BinaryOperator, FreezeInst, PHINode,
+                               SelectInst>(UserTE->getMainOp()) ||
+                          isa<SIToFPInst, UIToFPInst>(UserTE->getMainOp()))))
                       return true;
                     unsigned UserTESz = DL->getTypeSizeInBits(
                         UserTE->Scalars.front()->getType());
@@ -20653,7 +21132,12 @@ void BoUpSLP::computeMinimumValueSizes() {
       if (!IsKnownPositive)
         ++BitWidth1;
 
-      APInt Mask = DB->getDemandedBits(cast<Instruction>(Root));
+      auto *I = dyn_cast<Instruction>(Root);
+      if (!I) {
+        MaxBitWidth = std::max(BitWidth1, MaxBitWidth);
+        continue;
+      }
+      APInt Mask = DB->getDemandedBits(I);
       unsigned BitWidth2 = Mask.getBitWidth() - Mask.countl_zero();
       MaxBitWidth =
           std::max<unsigned>(std::min(BitWidth1, BitWidth2), MaxBitWidth);
@@ -20802,6 +21286,7 @@ void BoUpSLP::computeMinimumValueSizes() {
           NodeIdx < VectorizableTree.size() &&
           VectorizableTree[NodeIdx]->UserTreeIndex &&
           VectorizableTree[NodeIdx]->UserTreeIndex.EdgeIdx == 0 &&
+          VectorizableTree[NodeIdx]->UserTreeIndex.UserTE->hasState() &&
           VectorizableTree[NodeIdx]->UserTreeIndex.UserTE->getOpcode() ==
               Instruction::Trunc &&
           !VectorizableTree[NodeIdx]->UserTreeIndex.UserTE->isAltShuffle();
@@ -20982,7 +21467,9 @@ SLPVectorizerPass::vectorizeStoreChain(ArrayRef<Value *> Chain, BoUpSLP &R,
   for (Value *V : Chain)
     ValOps.insert(cast<StoreInst>(V)->getValueOperand());
   // Operands are not same/alt opcodes or non-power-of-2 uniques - exit.
-  InstructionsState S = getSameOpcode(ValOps.getArrayRef(), *TLI);
+  InstructionsCompatibilityAnalysis Analysis(*DT, *DL, *TTI, *TLI);
+  InstructionsState S = Analysis.buildInstructionsState(
+      ValOps.getArrayRef(), R, /*TryCopyableElementsVectorization=*/true);
   if (all_of(ValOps, IsaPred<Instruction>) && ValOps.size() > 1) {
     DenseSet<Value *> Stores(Chain.begin(), Chain.end());
     bool IsAllowedSize =
diff --git a/llvm/lib/Transforms/Vectorize/VPlan.cpp b/llvm/lib/Transforms/Vectorize/VPlan.cpp
index 40a5565..25b9616 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlan.cpp
@@ -962,7 +962,11 @@ void VPlan::prepareToExecute(Value *TripCountV, Value *VectorTripCountV,
     BackedgeTakenCount->setUnderlyingValue(TCMO);
   }
 
-  VectorTripCount.setUnderlyingValue(VectorTripCountV);
+  if (!VectorTripCount.getUnderlyingValue())
+    VectorTripCount.setUnderlyingValue(VectorTripCountV);
+  else
+    assert(VectorTripCount.getUnderlyingValue() == VectorTripCountV &&
+           "VectorTripCount set earlier must much VectorTripCountV");
 
   IRBuilder<> Builder(State.CFG.PrevBB->getTerminator());
   // FIXME: Model VF * UF computation completely in VPlan.
diff --git a/llvm/lib/Transforms/Vectorize/VPlan.h b/llvm/lib/Transforms/Vectorize/VPlan.h
index 6655149..a5de593 100644
--- a/llvm/lib/Transforms/Vectorize/VPlan.h
+++ b/llvm/lib/Transforms/Vectorize/VPlan.h
@@ -1012,6 +1012,10 @@ public:
     ReductionStartVector,
     // Creates a step vector starting from 0 to VF with a step of 1.
     StepVector,
+    /// Extracts a single lane (first operand) from a set of vector operands.
+    /// The lane specifies an index into a vector formed by combining all vector
+    /// operands (all operands after the first one).
+    ExtractLane,
 
   };
 
@@ -1837,6 +1841,10 @@ public:
                                      getGEPNoWrapFlags(), getDebugLoc());
   }
 
+  /// Return true if this VPVectorPointerRecipe corresponds to part 0. Note that
+  /// this is only accurate after the VPlan has been unrolled.
+  bool isFirstPart() const { return getUnrollPart(*this) == 0; }
+
   /// Return the cost of this VPHeaderPHIRecipe.
   InstructionCost computeCost(ElementCount VF,
                               VPCostContext &Ctx) const override {
@@ -2304,14 +2312,15 @@ public:
   /// respective masks, ordered [I0, M0, I1, M1, I2, M2, ...]. Note that M0 can
   /// be omitted (implied by passing an odd number of operands) in which case
   /// all other incoming values are merged into it.
-  VPBlendRecipe(PHINode *Phi, ArrayRef<VPValue *> Operands)
-      : VPSingleDefRecipe(VPDef::VPBlendSC, Operands, Phi, Phi->getDebugLoc()) {
+  VPBlendRecipe(PHINode *Phi, ArrayRef<VPValue *> Operands, DebugLoc DL)
+      : VPSingleDefRecipe(VPDef::VPBlendSC, Operands, Phi, DL) {
     assert(Operands.size() > 0 && "Expected at least one operand!");
   }
 
   VPBlendRecipe *clone() override {
     SmallVector<VPValue *> Ops(operands());
-    return new VPBlendRecipe(cast<PHINode>(getUnderlyingValue()), Ops);
+    return new VPBlendRecipe(cast_or_null<PHINode>(getUnderlyingValue()), Ops,
+                             getDebugLoc());
   }
 
   VP_CLASSOF_IMPL(VPDef::VPBlendSC)
@@ -4056,6 +4065,10 @@ public:
   /// Returns VF * UF of the vector loop region.
   VPValue &getVFxUF() { return VFxUF; }
 
+  LLVMContext &getContext() const {
+    return getScalarHeader()->getIRBasicBlock()->getContext();
+  }
+
   void addVF(ElementCount VF) { VFs.insert(VF); }
 
   void setVF(ElementCount VF) {
diff --git a/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp b/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp
index 3499e65..16072f2 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanAnalysis.cpp
@@ -110,6 +110,8 @@ Type *VPTypeAnalysis::inferScalarTypeForRecipe(const VPInstruction *R) {
   case VPInstruction::BuildStructVector:
   case VPInstruction::BuildVector:
     return SetResultTyFromOp();
+  case VPInstruction::ExtractLane:
+    return inferScalarType(R->getOperand(1));
   case VPInstruction::FirstActiveLane:
     return Type::getIntNTy(Ctx, 64);
   case VPInstruction::ExtractLastElement:
diff --git a/llvm/lib/Transforms/Vectorize/VPlanConstruction.cpp b/llvm/lib/Transforms/Vectorize/VPlanConstruction.cpp
index 194874a..6c1f53b 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanConstruction.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanConstruction.cpp
@@ -437,9 +437,12 @@ static void addCanonicalIVRecipes(VPlan &Plan, VPBasicBlock *HeaderVPBB,
 
   // We are about to replace the branch to exit the region. Remove the original
   // BranchOnCond, if there is any.
+  DebugLoc LatchDL = DL;
   if (!LatchVPBB->empty() &&
-      match(&LatchVPBB->back(), m_BranchOnCond(m_VPValue())))
+      match(&LatchVPBB->back(), m_BranchOnCond(m_VPValue()))) {
+    LatchDL = LatchVPBB->getTerminator()->getDebugLoc();
     LatchVPBB->getTerminator()->eraseFromParent();
+  }
 
   VPBuilder Builder(LatchVPBB);
   // Add a VPInstruction to increment the scalar canonical IV by VF * UF.
@@ -452,7 +455,8 @@ static void addCanonicalIVRecipes(VPlan &Plan, VPBasicBlock *HeaderVPBB,
 
   // Add the BranchOnCount VPInstruction to the latch.
   Builder.createNaryOp(VPInstruction::BranchOnCount,
-                       {CanonicalIVIncrement, &Plan.getVectorTripCount()}, DL);
+                       {CanonicalIVIncrement, &Plan.getVectorTripCount()},
+                       LatchDL);
 }
 
 void VPlanTransforms::prepareForVectorization(
@@ -462,28 +466,27 @@ void VPlanTransforms::prepareForVectorization(
   VPDominatorTree VPDT;
   VPDT.recalculate(Plan);
 
-  VPBlockBase *HeaderVPB = Plan.getEntry()->getSingleSuccessor();
-  canonicalHeaderAndLatch(HeaderVPB, VPDT);
-  VPBlockBase *LatchVPB = HeaderVPB->getPredecessors()[1];
+  auto *HeaderVPBB = cast<VPBasicBlock>(Plan.getEntry()->getSingleSuccessor());
+  canonicalHeaderAndLatch(HeaderVPBB, VPDT);
+  auto *LatchVPBB = cast<VPBasicBlock>(HeaderVPBB->getPredecessors()[1]);
 
   VPBasicBlock *VecPreheader = Plan.createVPBasicBlock("vector.ph");
   VPBlockUtils::insertBlockAfter(VecPreheader, Plan.getEntry());
 
   VPBasicBlock *MiddleVPBB = Plan.createVPBasicBlock("middle.block");
-  // The canonical LatchVPB has the header block as last successor. If it has
+  // The canonical LatchVPBB has the header block as last successor. If it has
   // another successor, this successor is an exit block - insert middle block on
   // its edge. Otherwise, add middle block as another successor retaining header
   // as last.
-  if (LatchVPB->getNumSuccessors() == 2) {
-    VPBlockBase *LatchExitVPB = LatchVPB->getSuccessors()[0];
-    VPBlockUtils::insertOnEdge(LatchVPB, LatchExitVPB, MiddleVPBB);
+  if (LatchVPBB->getNumSuccessors() == 2) {
+    VPBlockBase *LatchExitVPB = LatchVPBB->getSuccessors()[0];
+    VPBlockUtils::insertOnEdge(LatchVPBB, LatchExitVPB, MiddleVPBB);
   } else {
-    VPBlockUtils::connectBlocks(LatchVPB, MiddleVPBB);
-    LatchVPB->swapSuccessors();
+    VPBlockUtils::connectBlocks(LatchVPBB, MiddleVPBB);
+    LatchVPBB->swapSuccessors();
   }
 
-  addCanonicalIVRecipes(Plan, cast<VPBasicBlock>(HeaderVPB),
-                        cast<VPBasicBlock>(LatchVPB), InductionTy, IVDL);
+  addCanonicalIVRecipes(Plan, HeaderVPBB, LatchVPBB, InductionTy, IVDL);
 
   [[maybe_unused]] bool HandledUncountableEarlyExit = false;
   // Disconnect all early exits from the loop leaving it with a single exit from
@@ -499,8 +502,7 @@ void VPlanTransforms::prepareForVectorization(
         assert(!HandledUncountableEarlyExit &&
                "can handle exactly one uncountable early exit");
         handleUncountableEarlyExit(cast<VPBasicBlock>(Pred), EB, Plan,
-                                   cast<VPBasicBlock>(HeaderVPB),
-                                   cast<VPBasicBlock>(LatchVPB), Range);
+                                   HeaderVPBB, LatchVPBB, Range);
         HandledUncountableEarlyExit = true;
       } else {
         for (VPRecipeBase &R : EB->phis())
@@ -564,15 +566,15 @@ void VPlanTransforms::prepareForVectorization(
   // the corresponding compare because they may have ended up with different
   // line numbers and we want to avoid awkward line stepping while debugging.
   // E.g., if the compare has got a line number inside the loop.
-  DebugLoc LatchDL = TheLoop->getLoopLatch()->getTerminator()->getDebugLoc();
+  DebugLoc LatchDL = LatchVPBB->getTerminator()->getDebugLoc();
   VPBuilder Builder(MiddleVPBB);
   VPValue *Cmp;
   if (!RequiresScalarEpilogueCheck)
-    Cmp = Plan.getOrAddLiveIn(ConstantInt::getFalse(
-        IntegerType::getInt1Ty(TripCount->getType()->getContext())));
+    Cmp = Plan.getOrAddLiveIn(
+        ConstantInt::getFalse(IntegerType::getInt1Ty(Plan.getContext())));
   else if (TailFolded)
-    Cmp = Plan.getOrAddLiveIn(ConstantInt::getTrue(
-        IntegerType::getInt1Ty(TripCount->getType()->getContext())));
+    Cmp = Plan.getOrAddLiveIn(
+        ConstantInt::getTrue(IntegerType::getInt1Ty(Plan.getContext())));
   else
     Cmp = Builder.createICmp(CmpInst::ICMP_EQ, Plan.getTripCount(),
                              &Plan.getVectorTripCount(), LatchDL, "cmp.n");
@@ -646,7 +648,7 @@ void VPlanTransforms::attachCheckBlock(VPlan &Plan, Value *Cond,
                    .createNaryOp(VPInstruction::BranchOnCond, {CondVPV},
                                  Plan.getCanonicalIV()->getDebugLoc());
   if (AddBranchWeights) {
-    MDBuilder MDB(Plan.getScalarHeader()->getIRBasicBlock()->getContext());
+    MDBuilder MDB(Plan.getContext());
     MDNode *BranchWeights =
         MDB.createBranchWeights(CheckBypassWeights, /*IsExpected=*/false);
     Term->addMetadata(LLVMContext::MD_prof, BranchWeights);
diff --git a/llvm/lib/Transforms/Vectorize/VPlanPredicator.cpp b/llvm/lib/Transforms/Vectorize/VPlanPredicator.cpp
index fc8458c..3b3bbc3 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanPredicator.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanPredicator.cpp
@@ -251,8 +251,9 @@ void VPPredicator::convertPhisToBlends(VPBasicBlock *VPBB) {
       }
       OperandsWithMask.push_back(EdgeMask);
     }
-    PHINode *IRPhi = cast<PHINode>(PhiR->getUnderlyingValue());
-    auto *Blend = new VPBlendRecipe(IRPhi, OperandsWithMask);
+    PHINode *IRPhi = cast_or_null<PHINode>(PhiR->getUnderlyingValue());
+    auto *Blend =
+        new VPBlendRecipe(IRPhi, OperandsWithMask, PhiR->getDebugLoc());
     Builder.insert(Blend);
     PhiR->replaceAllUsesWith(Blend);
     PhiR->eraseFromParent();
diff --git a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
index 0d6152b..225658b 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanRecipes.cpp
@@ -427,6 +427,7 @@ unsigned VPUnrollPartAccessor<PartOpIdx>::getUnrollPart(const VPUser &U) const {
 }
 
 namespace llvm {
+template class VPUnrollPartAccessor<1>;
 template class VPUnrollPartAccessor<2>;
 template class VPUnrollPartAccessor<3>;
 }
@@ -863,6 +864,31 @@ Value *VPInstruction::generate(VPTransformState &State) {
       Res = Builder.CreateOr(Res, State.get(Op));
     return State.VF.isScalar() ? Res : Builder.CreateOrReduce(Res);
   }
+  case VPInstruction::ExtractLane: {
+    Value *LaneToExtract = State.get(getOperand(0), true);
+    Type *IdxTy = State.TypeAnalysis.inferScalarType(getOperand(0));
+    Value *Res = nullptr;
+    Value *RuntimeVF = getRuntimeVF(State.Builder, IdxTy, State.VF);
+
+    for (unsigned Idx = 1; Idx != getNumOperands(); ++Idx) {
+      Value *VectorStart =
+          Builder.CreateMul(RuntimeVF, ConstantInt::get(IdxTy, Idx - 1));
+      Value *VectorIdx = Idx == 1
+                             ? LaneToExtract
+                             : Builder.CreateSub(LaneToExtract, VectorStart);
+      Value *Ext = State.VF.isScalar()
+                       ? State.get(getOperand(Idx))
+                       : Builder.CreateExtractElement(
+                             State.get(getOperand(Idx)), VectorIdx);
+      if (Res) {
+        Value *Cmp = Builder.CreateICmpUGE(LaneToExtract, VectorStart);
+        Res = Builder.CreateSelect(Cmp, Ext, Res);
+      } else {
+        Res = Ext;
+      }
+    }
+    return Res;
+  }
   case VPInstruction::FirstActiveLane: {
     if (getNumOperands() == 1) {
       Value *Mask = State.get(getOperand(0));
@@ -921,7 +947,8 @@ InstructionCost VPInstruction::computeCost(ElementCount VF,
   }
 
   switch (getOpcode()) {
-  case Instruction::ExtractElement: {
+  case Instruction::ExtractElement:
+  case VPInstruction::ExtractLane: {
     // Add on the cost of extracting the element.
     auto *VecTy = toVectorTy(Ctx.Types.inferScalarType(getOperand(0)), VF);
     return Ctx.TTI.getVectorInstrCost(Instruction::ExtractElement, VecTy,
@@ -983,6 +1010,7 @@ bool VPInstruction::isVectorToScalar() const {
   return getOpcode() == VPInstruction::ExtractLastElement ||
          getOpcode() == VPInstruction::ExtractPenultimateElement ||
          getOpcode() == Instruction::ExtractElement ||
+         getOpcode() == VPInstruction::ExtractLane ||
          getOpcode() == VPInstruction::FirstActiveLane ||
          getOpcode() == VPInstruction::ComputeAnyOfResult ||
          getOpcode() == VPInstruction::ComputeFindIVResult ||
@@ -1048,6 +1076,7 @@ bool VPInstruction::opcodeMayReadOrWriteFromMemory() const {
   case VPInstruction::BuildVector:
   case VPInstruction::CalculateTripCountMinusVF:
   case VPInstruction::CanonicalIVIncrementForPart:
+  case VPInstruction::ExtractLane:
   case VPInstruction::ExtractLastElement:
   case VPInstruction::ExtractPenultimateElement:
   case VPInstruction::FirstActiveLane:
@@ -1097,6 +1126,8 @@ bool VPInstruction::onlyFirstLaneUsed(const VPValue *Op) const {
   case VPInstruction::ComputeAnyOfResult:
   case VPInstruction::ComputeFindIVResult:
     return Op == getOperand(1);
+  case VPInstruction::ExtractLane:
+    return Op == getOperand(0);
   };
   llvm_unreachable("switch should return");
 }
@@ -1176,6 +1207,9 @@ void VPInstruction::print(raw_ostream &O, const Twine &Indent,
   case VPInstruction::BuildVector:
     O << "buildvector";
     break;
+  case VPInstruction::ExtractLane:
+    O << "extract-lane";
+    break;
   case VPInstruction::ExtractLastElement:
     O << "extract-last-element";
     break;
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
index 3372bcc..935a4e4 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.cpp
@@ -774,10 +774,10 @@ static VPValue *optimizeEarlyExitInductionUser(VPlan &Plan,
   using namespace VPlanPatternMatch;
 
   VPValue *Incoming, *Mask;
-  if (!match(Op, m_VPInstruction<Instruction::ExtractElement>(
-                     m_VPValue(Incoming),
+  if (!match(Op, m_VPInstruction<VPInstruction::ExtractLane>(
                      m_VPInstruction<VPInstruction::FirstActiveLane>(
-                         m_VPValue(Mask)))))
+                         m_VPValue(Mask)),
+                     m_VPValue(Incoming))))
     return nullptr;
 
   auto *WideIV = getOptimizableIVOf(Incoming);
@@ -1172,6 +1172,14 @@ static void simplifyRecipe(VPRecipeBase &R, VPTypeAnalysis &TypeInfo) {
   if (!Plan->isUnrolled())
     return;
 
+  // VPVectorPointer for part 0 can be replaced by their start pointer.
+  if (auto *VecPtr = dyn_cast<VPVectorPointerRecipe>(&R)) {
+    if (VecPtr->isFirstPart()) {
+      VecPtr->replaceAllUsesWith(VecPtr->getOperand(0));
+      return;
+    }
+  }
+
   // VPScalarIVSteps for part 0 can be replaced by their start value, if only
   // the first lane is demanded.
   if (auto *Steps = dyn_cast<VPScalarIVStepsRecipe>(Def)) {
@@ -1307,8 +1315,9 @@ static void simplifyBlends(VPlan &Plan) {
         OperandsWithMask.push_back(Blend->getMask(I));
       }
 
-      auto *NewBlend = new VPBlendRecipe(
-          cast<PHINode>(Blend->getUnderlyingValue()), OperandsWithMask);
+      auto *NewBlend =
+          new VPBlendRecipe(cast_or_null<PHINode>(Blend->getUnderlyingValue()),
+                            OperandsWithMask, Blend->getDebugLoc());
       NewBlend->insertBefore(&R);
 
       VPValue *DeadMask = Blend->getMask(StartIndex);
@@ -1361,7 +1370,7 @@ static bool optimizeVectorInductionWidthForTCAndVFUF(VPlan &Plan,
   unsigned NewBitWidth =
       ComputeBitWidth(TC->getValue(), BestVF.getKnownMinValue() * BestUF);
 
-  LLVMContext &Ctx = Plan.getCanonicalIV()->getScalarType()->getContext();
+  LLVMContext &Ctx = Plan.getContext();
   auto *NewIVTy = IntegerType::get(Ctx, NewBitWidth);
 
   bool MadeChange = false;
@@ -1883,9 +1892,7 @@ void VPlanTransforms::truncateToMinimalBitwidths(
   }
 }
 
-/// Remove BranchOnCond recipes with true or false conditions together with
-/// removing dead edges to their successors.
-static void removeBranchOnConst(VPlan &Plan) {
+void VPlanTransforms::removeBranchOnConst(VPlan &Plan) {
   using namespace llvm::VPlanPatternMatch;
   for (VPBasicBlock *VPBB : VPBlockUtils::blocksOnly<VPBasicBlock>(
            vp_depth_first_shallow(Plan.getEntry()))) {
@@ -1908,12 +1915,9 @@ static void removeBranchOnConst(VPlan &Plan) {
            "There must be a single edge between VPBB and its successor");
     // Values coming from VPBB into phi recipes of RemoveSucc are removed from
     // these recipes.
-    for (VPRecipeBase &R : RemovedSucc->phis()) {
-      auto *Phi = cast<VPPhiAccessors>(&R);
-      assert((!isa<VPIRPhi>(&R) || RemovedSucc->getNumPredecessors() == 1) &&
-             "VPIRPhis must have a single predecessor");
-      Phi->removeIncomingValueFor(VPBB);
-    }
+    for (VPRecipeBase &R : RemovedSucc->phis())
+      cast<VPPhiAccessors>(&R)->removeIncomingValueFor(VPBB);
+
     // Disconnect blocks and remove the terminator. RemovedSucc will be deleted
     // automatically on VPlan destruction if it becomes unreachable.
     VPBlockUtils::disconnectBlocks(VPBB, RemovedSucc);
@@ -2515,8 +2519,8 @@ void VPlanTransforms::createInterleaveGroups(
                    DL.getTypeAllocSize(getLoadStoreType(IRInsertPos)) *
                        IG->getIndex(IRInsertPos),
                    /*IsSigned=*/true);
-      VPValue *OffsetVPV = Plan.getOrAddLiveIn(
-          ConstantInt::get(IRInsertPos->getParent()->getContext(), -Offset));
+      VPValue *OffsetVPV =
+          Plan.getOrAddLiveIn(ConstantInt::get(Plan.getContext(), -Offset));
       VPBuilder B(InsertPos);
       Addr = InBounds ? B.createInBoundsPtrAdd(InsertPos->getAddr(), OffsetVPV)
                       : B.createPtrAdd(InsertPos->getAddr(), OffsetVPV);
@@ -2842,7 +2846,7 @@ void VPlanTransforms::handleUncountableEarlyExit(
           VPInstruction::FirstActiveLane, {CondToEarlyExit}, nullptr,
           "first.active.lane");
       IncomingFromEarlyExit = EarlyExitB.createNaryOp(
-          Instruction::ExtractElement, {IncomingFromEarlyExit, FirstActiveLane},
+          VPInstruction::ExtractLane, {FirstActiveLane, IncomingFromEarlyExit},
           nullptr, "early.exit.value");
       ExitIRI->setOperand(EarlyExitIdx, IncomingFromEarlyExit);
     }
@@ -3093,6 +3097,29 @@ void VPlanTransforms::materializeBroadcasts(VPlan &Plan) {
   }
 }
 
+void VPlanTransforms::materializeVectorTripCount(
+    VPlan &Plan, ElementCount BestVF, unsigned BestUF,
+    PredicatedScalarEvolution &PSE) {
+  assert(Plan.hasVF(BestVF) && "BestVF is not available in Plan");
+  assert(Plan.hasUF(BestUF) && "BestUF is not available in Plan");
+
+  VPValue *TC = Plan.getTripCount();
+  // Skip cases for which the trip count may be non-trivial to materialize.
+  if (!Plan.hasScalarTail() ||
+      Plan.getMiddleBlock()->getSingleSuccessor() ==
+          Plan.getScalarPreheader() ||
+      !TC->isLiveIn())
+    return;
+  // Materialize vector trip counts for constants early if it can simply
+  // be computed as (Original TC / VF * UF) * VF * UF.
+  ScalarEvolution &SE = *PSE.getSE();
+  auto *TCScev = SE.getSCEV(TC->getLiveInIRValue());
+  const SCEV *VFxUF = SE.getElementCount(TCScev->getType(), BestVF * BestUF);
+  auto VecTCScev = SE.getMulExpr(SE.getUDivExpr(TCScev, VFxUF), VFxUF);
+  if (auto *NewC = dyn_cast<SCEVConstant>(VecTCScev))
+    Plan.getVectorTripCount().setUnderlyingValue(NewC->getValue());
+}
+
 /// Returns true if \p V is VPWidenLoadRecipe or VPInterleaveRecipe that can be
 /// converted to a narrower recipe. \p V is used by a wide recipe that feeds a
 /// store interleave group at index \p Idx, \p WideMember0 is the recipe feeding
@@ -3350,7 +3377,7 @@ void VPlanTransforms::addBranchWeightToMiddleTerminator(
   if (VF.isScalable() && VScaleForTuning.has_value())
     VectorStep *= *VScaleForTuning;
   assert(VectorStep > 0 && "trip count should not be zero");
-  MDBuilder MDB(Plan.getScalarHeader()->getIRBasicBlock()->getContext());
+  MDBuilder MDB(Plan.getContext());
   MDNode *BranchWeights =
       MDB.createBranchWeights({1, VectorStep - 1}, /*IsExpected=*/false);
   MiddleTerm->addMetadata(LLVMContext::MD_prof, BranchWeights);
diff --git a/llvm/lib/Transforms/Vectorize/VPlanTransforms.h b/llvm/lib/Transforms/Vectorize/VPlanTransforms.h
index ab189f6..d5af6cd 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanTransforms.h
+++ b/llvm/lib/Transforms/Vectorize/VPlanTransforms.h
@@ -224,6 +224,10 @@ struct VPlanTransforms {
   /// CanonicalIVTy as type for all un-typed live-ins in VPTypeAnalysis.
   static void simplifyRecipes(VPlan &Plan, Type &CanonicalIVTy);
 
+  /// Remove BranchOnCond recipes with true or false conditions together with
+  /// removing dead edges to their successors.
+  static void removeBranchOnConst(VPlan &Plan);
+
   /// If there's a single exit block, optimize its phi recipes that use exiting
   /// IV values by feeding them precomputed end values instead, possibly taken
   /// one step backwards.
@@ -234,6 +238,12 @@ struct VPlanTransforms {
   /// Add explicit broadcasts for live-ins and VPValues defined in \p Plan's entry block if they are used as vectors.
   static void materializeBroadcasts(VPlan &Plan);
 
+  // Materialize vector trip counts for constants early if it can simply be
+  // computed as (Original TC / VF * UF) * VF * UF.
+  static void materializeVectorTripCount(VPlan &Plan, ElementCount BestVF,
+                                         unsigned BestUF,
+                                         PredicatedScalarEvolution &PSE);
+
   /// Try to convert a plan with interleave groups with VF elements to a plan
   /// with the interleave groups replaced by wide loads and stores processing VF
   /// elements, if all transformed interleave groups access the full vector
diff --git a/llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp b/llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp
index b89cd21..871e37e 100644
--- a/llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp
+++ b/llvm/lib/Transforms/Vectorize/VPlanUnroll.cpp
@@ -363,6 +363,13 @@ void UnrollState::unrollBlock(VPBlockBase *VPB) {
       continue;
     }
     VPValue *Op0;
+    if (match(&R, m_VPInstruction<VPInstruction::ExtractLane>(
+                      m_VPValue(Op0), m_VPValue(Op1)))) {
+      addUniformForAllParts(cast<VPInstruction>(&R));
+      for (unsigned Part = 1; Part != UF; ++Part)
+        R.addOperand(getValueForPart(Op1, Part));
+      continue;
+    }
     if (match(&R, m_VPInstruction<VPInstruction::ExtractLastElement>(
                       m_VPValue(Op0))) ||
         match(&R, m_VPInstruction<VPInstruction::ExtractPenultimateElement>(
diff --git a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
index 82adc34..6252f4f 100644
--- a/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
+++ b/llvm/lib/Transforms/Vectorize/VectorCombine.cpp
@@ -3174,6 +3174,55 @@ bool VectorCombine::foldCastFromReductions(Instruction &I) {
   return true;
 }
 
+/// Returns true if this ShuffleVectorInst eventually feeds into a
+/// vector reduction intrinsic (e.g., vector_reduce_add) by only following
+/// chains of shuffles and binary operators (in any combination/order).
+/// The search does not go deeper than the given Depth.
+static bool feedsIntoVectorReduction(ShuffleVectorInst *SVI) {
+  constexpr unsigned MaxVisited = 32;
+  SmallPtrSet<Instruction *, 8> Visited;
+  SmallVector<Instruction *, 4> WorkList;
+  bool FoundReduction = false;
+
+  WorkList.push_back(SVI);
+  while (!WorkList.empty()) {
+    Instruction *I = WorkList.pop_back_val();
+    for (User *U : I->users()) {
+      auto *UI = cast<Instruction>(U);
+      if (!UI || !Visited.insert(UI).second)
+        continue;
+      if (Visited.size() > MaxVisited)
+        return false;
+      if (auto *II = dyn_cast<IntrinsicInst>(UI)) {
+        // More than one reduction reached
+        if (FoundReduction)
+          return false;
+        switch (II->getIntrinsicID()) {
+        case Intrinsic::vector_reduce_add:
+        case Intrinsic::vector_reduce_mul:
+        case Intrinsic::vector_reduce_and:
+        case Intrinsic::vector_reduce_or:
+        case Intrinsic::vector_reduce_xor:
+        case Intrinsic::vector_reduce_smin:
+        case Intrinsic::vector_reduce_smax:
+        case Intrinsic::vector_reduce_umin:
+        case Intrinsic::vector_reduce_umax:
+          FoundReduction = true;
+          continue;
+        default:
+          return false;
+        }
+      }
+
+      if (!isa<BinaryOperator>(UI) && !isa<ShuffleVectorInst>(UI))
+        return false;
+
+      WorkList.emplace_back(UI);
+    }
+  }
+  return FoundReduction;
+}
+
 /// This method looks for groups of shuffles acting on binops, of the form:
 ///  %x = shuffle ...
 ///  %y = shuffle ...
@@ -3416,6 +3465,65 @@ bool VectorCombine::foldSelectShuffle(Instruction &I, bool FromReduction) {
            TTI.getShuffleCost(TTI::SK_PermuteTwoSrc, VT, VT, Mask, CostKind);
   };
 
+  unsigned ElementSize = VT->getElementType()->getPrimitiveSizeInBits();
+  unsigned MaxVectorSize =
+      TTI.getRegisterBitWidth(TargetTransformInfo::RGK_FixedWidthVector);
+  unsigned MaxElementsInVector = MaxVectorSize / ElementSize;
+  // When there are multiple shufflevector operations on the same input,
+  // especially when the vector length is larger than the register size,
+  // identical shuffle patterns may occur across different groups of elements.
+  // To avoid overestimating the cost by counting these repeated shuffles more
+  // than once, we only account for unique shuffle patterns. This adjustment
+  // prevents inflated costs in the cost model for wide vectors split into
+  // several register-sized groups.
+  std::set<SmallVector<int, 4>> UniqueShuffles;
+  auto AddShuffleMaskAdjustedCost = [&](InstructionCost C, ArrayRef<int> Mask) {
+    // Compute the cost for performing the shuffle over the full vector.
+    auto ShuffleCost =
+        TTI.getShuffleCost(TTI::SK_PermuteTwoSrc, VT, VT, Mask, CostKind);
+    unsigned NumFullVectors = Mask.size() / MaxElementsInVector;
+    if (NumFullVectors < 2)
+      return C + ShuffleCost;
+    SmallVector<int, 4> SubShuffle(MaxElementsInVector);
+    unsigned NumUniqueGroups = 0;
+    unsigned NumGroups = Mask.size() / MaxElementsInVector;
+    // For each group of MaxElementsInVector contiguous elements,
+    // collect their shuffle pattern and insert into the set of unique patterns.
+    for (unsigned I = 0; I < NumFullVectors; ++I) {
+      for (unsigned J = 0; J < MaxElementsInVector; ++J)
+        SubShuffle[J] = Mask[MaxElementsInVector * I + J];
+      if (UniqueShuffles.insert(SubShuffle).second)
+        NumUniqueGroups += 1;
+    }
+    return C + ShuffleCost * NumUniqueGroups / NumGroups;
+  };
+  auto AddShuffleAdjustedCost = [&](InstructionCost C, Instruction *I) {
+    auto *SV = dyn_cast<ShuffleVectorInst>(I);
+    if (!SV)
+      return C;
+    SmallVector<int, 16> Mask;
+    SV->getShuffleMask(Mask);
+    return AddShuffleMaskAdjustedCost(C, Mask);
+  };
+  // Check that input consists of ShuffleVectors applied to the same input
+  auto AllShufflesHaveSameOperands =
+      [](SmallPtrSetImpl<Instruction *> &InputShuffles) {
+        if (InputShuffles.size() < 2)
+          return false;
+        ShuffleVectorInst *FirstSV =
+            dyn_cast<ShuffleVectorInst>(*InputShuffles.begin());
+        if (!FirstSV)
+          return false;
+
+        Value *In0 = FirstSV->getOperand(0), *In1 = FirstSV->getOperand(1);
+        return std::all_of(
+            std::next(InputShuffles.begin()), InputShuffles.end(),
+            [&](Instruction *I) {
+              ShuffleVectorInst *SV = dyn_cast<ShuffleVectorInst>(I);
+              return SV && SV->getOperand(0) == In0 && SV->getOperand(1) == In1;
+            });
+      };
+
   // Get the costs of the shuffles + binops before and after with the new
   // shuffle masks.
   InstructionCost CostBefore =
@@ -3423,8 +3531,14 @@ bool VectorCombine::foldSelectShuffle(Instruction &I, bool FromReduction) {
       TTI.getArithmeticInstrCost(Op1->getOpcode(), VT, CostKind);
   CostBefore += std::accumulate(Shuffles.begin(), Shuffles.end(),
                                 InstructionCost(0), AddShuffleCost);
-  CostBefore += std::accumulate(InputShuffles.begin(), InputShuffles.end(),
-                                InstructionCost(0), AddShuffleCost);
+  if (AllShufflesHaveSameOperands(InputShuffles)) {
+    UniqueShuffles.clear();
+    CostBefore += std::accumulate(InputShuffles.begin(), InputShuffles.end(),
+                                  InstructionCost(0), AddShuffleAdjustedCost);
+  } else {
+    CostBefore += std::accumulate(InputShuffles.begin(), InputShuffles.end(),
+                                  InstructionCost(0), AddShuffleCost);
+  }
 
   // The new binops will be unused for lanes past the used shuffle lengths.
   // These types attempt to get the correct cost for that from the target.
@@ -3435,8 +3549,9 @@ bool VectorCombine::foldSelectShuffle(Instruction &I, bool FromReduction) {
   InstructionCost CostAfter =
       TTI.getArithmeticInstrCost(Op0->getOpcode(), Op0SmallVT, CostKind) +
       TTI.getArithmeticInstrCost(Op1->getOpcode(), Op1SmallVT, CostKind);
+  UniqueShuffles.clear();
   CostAfter += std::accumulate(ReconstructMasks.begin(), ReconstructMasks.end(),
-                               InstructionCost(0), AddShuffleMaskCost);
+                               InstructionCost(0), AddShuffleMaskAdjustedCost);
   std::set<SmallVector<int>> OutputShuffleMasks({V1A, V1B, V2A, V2B});
   CostAfter +=
       std::accumulate(OutputShuffleMasks.begin(), OutputShuffleMasks.end(),
@@ -3445,7 +3560,8 @@ bool VectorCombine::foldSelectShuffle(Instruction &I, bool FromReduction) {
   LLVM_DEBUG(dbgs() << "Found a binop select shuffle pattern: " << I << "\n");
   LLVM_DEBUG(dbgs() << "  CostBefore: " << CostBefore
                     << " vs CostAfter: " << CostAfter << "\n");
-  if (CostBefore <= CostAfter)
+  if (CostBefore < CostAfter ||
+      (CostBefore == CostAfter && !feedsIntoVectorReduction(SVI)))
     return false;
 
   // The cost model has passed, create the new instructions.
diff --git a/llvm/test/Analysis/CostModel/RISCV/arith-fp.ll b/llvm/test/Analysis/CostModel/RISCV/arith-fp.ll
index 6c974af..673bf38 100644
--- a/llvm/test/Analysis/CostModel/RISCV/arith-fp.ll
+++ b/llvm/test/Analysis/CostModel/RISCV/arith-fp.ll
@@ -1,1579 +1,2413 @@
 ; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py
 ; RUN: opt -passes="print<cost-model>" 2>&1 -disable-output -mtriple=riscv64 -mattr=+v,+f,+d,+zfh,+zvfh,+zvfbfmin < %s | FileCheck %s --check-prefixes=CHECK,ZVFH
 ; RUN: opt -passes="print<cost-model>" 2>&1 -disable-output -mtriple=riscv64 -mattr=+v,+f,+d,+zfh,+zvfhmin,+zvfbfmin < %s | FileCheck %s --check-prefixes=CHECK,ZVFHMIN
+; RUN: opt -passes="print<cost-model>" 2>&1 -disable-output -mtriple=riscv64 -mattr=+v,+f,+d < %s | FileCheck %s --check-prefixes=CHECK,NO-ZFHMIN
 ; Check that we don't crash querying costs when vectors are not enabled.
 ; RUN: opt -passes="print<cost-model>" 2>&1 -disable-output -mtriple=riscv64
 
 define void @fadd() {
 ; CHECK-LABEL: 'fadd'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = fadd bfloat undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = fadd float undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = fadd double undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1BF16 = fadd <1 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2BF16 = fadd <2 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4BF16 = fadd <4 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8BF16 = fadd <8 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16BF16 = fadd <16 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1BF16 = fadd <vscale x 1 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2BF16 = fadd <vscale x 2 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4BF16 = fadd <vscale x 4 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8BF16 = fadd <vscale x 8 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16BF16 = fadd <vscale x 16 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32 = fadd <1 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32 = fadd <2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fadd <4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fadd <8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fadd <16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F32 = fadd <vscale x 1 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F32 = fadd <vscale x 2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F32 = fadd <vscale x 4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F32 = fadd <vscale x 8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16F32 = fadd <vscale x 16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64 = fadd <1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fadd <2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fadd <4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fadd <8 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F64 = fadd <vscale x 1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2F64 = fadd <vscale x 2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4F64 = fadd <vscale x 4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8F64 = fadd <vscale x 8 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1F16_VP = call <1 x bfloat> @llvm.vp.fadd.v1bf16(<1 x bfloat> undef, <1 x bfloat> undef, <1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2F16_VP = call <2 x bfloat> @llvm.vp.fadd.v2bf16(<2 x bfloat> undef, <2 x bfloat> undef, <2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4F16_VP = call <4 x bfloat> @llvm.vp.fadd.v4bf16(<4 x bfloat> undef, <4 x bfloat> undef, <4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8F16_VP = call <8 x bfloat> @llvm.vp.fadd.v8bf16(<8 x bfloat> undef, <8 x bfloat> undef, <8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16F16_VP = call <16 x bfloat> @llvm.vp.fadd.v16bf16(<16 x bfloat> undef, <16 x bfloat> undef, <16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32_VP = call <1 x float> @llvm.vp.fadd.v1f32(<1 x float> undef, <1 x float> undef, <1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32_VP = call <2 x float> @llvm.vp.fadd.v2f32(<2 x float> undef, <2 x float> undef, <2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32_VP = call <4 x float> @llvm.vp.fadd.v4f32(<4 x float> undef, <4 x float> undef, <4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32_VP = call <8 x float> @llvm.vp.fadd.v8f32(<8 x float> undef, <8 x float> undef, <8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32_VP = call <16 x float> @llvm.vp.fadd.v16f32(<16 x float> undef, <16 x float> undef, <16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64_VP = call <1 x double> @llvm.vp.fadd.v1f64(<1 x double> undef, <1 x double> undef, <1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64_VP = call <2 x double> @llvm.vp.fadd.v2f64(<2 x double> undef, <2 x double> undef, <2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64_VP = call <4 x double> @llvm.vp.fadd.v4f64(<4 x double> undef, <4 x double> undef, <4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64_VP = call <8 x double> @llvm.vp.fadd.v8f64(<8 x double> undef, <8 x double> undef, <8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1F16_VP = call <vscale x 1 x bfloat> @llvm.vp.fadd.nxv1bf16(<vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef, <vscale x 1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2F16_VP = call <vscale x 2 x bfloat> @llvm.vp.fadd.nxv2bf16(<vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef, <vscale x 2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4F16_VP = call <vscale x 4 x bfloat> @llvm.vp.fadd.nxv4bf16(<vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef, <vscale x 4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8F16_VP = call <vscale x 8 x bfloat> @llvm.vp.fadd.nxv8bf16(<vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef, <vscale x 8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16F16_VP = call <vscale x 16 x bfloat> @llvm.vp.fadd.nxv16bf16(<vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef, <vscale x 16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.fadd.nxv1f32(<vscale x 1 x float> undef, <vscale x 1 x float> undef, <vscale x 1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.fadd.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> undef, <vscale x 2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.fadd.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x float> undef, <vscale x 4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.fadd.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x float> undef, <vscale x 8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.fadd.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x float> undef, <vscale x 16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.fadd.nxv1f64(<vscale x 1 x double> undef, <vscale x 1 x double> undef, <vscale x 1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.fadd.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x double> undef, <vscale x 2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.fadd.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x double> undef, <vscale x 4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.fadd.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x double> undef, <vscale x 8 x i1> undef, i32 undef)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = fadd float poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = fadd double poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32 = fadd <1 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32 = fadd <2 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fadd <4 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fadd <8 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fadd <16 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F32 = fadd <vscale x 1 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F32 = fadd <vscale x 2 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F32 = fadd <vscale x 4 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F32 = fadd <vscale x 8 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16F32 = fadd <vscale x 16 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64 = fadd <1 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fadd <2 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fadd <4 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fadd <8 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F64 = fadd <vscale x 1 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2F64 = fadd <vscale x 2 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4F64 = fadd <vscale x 4 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8F64 = fadd <vscale x 8 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32_VP = call <1 x float> @llvm.vp.fadd.v1f32(<1 x float> poison, <1 x float> poison, <1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32_VP = call <2 x float> @llvm.vp.fadd.v2f32(<2 x float> poison, <2 x float> poison, <2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32_VP = call <4 x float> @llvm.vp.fadd.v4f32(<4 x float> poison, <4 x float> poison, <4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32_VP = call <8 x float> @llvm.vp.fadd.v8f32(<8 x float> poison, <8 x float> poison, <8 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32_VP = call <16 x float> @llvm.vp.fadd.v16f32(<16 x float> poison, <16 x float> poison, <16 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64_VP = call <1 x double> @llvm.vp.fadd.v1f64(<1 x double> poison, <1 x double> poison, <1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64_VP = call <2 x double> @llvm.vp.fadd.v2f64(<2 x double> poison, <2 x double> poison, <2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64_VP = call <4 x double> @llvm.vp.fadd.v4f64(<4 x double> poison, <4 x double> poison, <4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64_VP = call <8 x double> @llvm.vp.fadd.v8f64(<8 x double> poison, <8 x double> poison, <8 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.fadd.nxv1f32(<vscale x 1 x float> poison, <vscale x 1 x float> poison, <vscale x 1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.fadd.nxv2f32(<vscale x 2 x float> poison, <vscale x 2 x float> poison, <vscale x 2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.fadd.nxv4f32(<vscale x 4 x float> poison, <vscale x 4 x float> poison, <vscale x 4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.fadd.nxv8f32(<vscale x 8 x float> poison, <vscale x 8 x float> poison, <vscale x 8 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.fadd.nxv16f32(<vscale x 16 x float> poison, <vscale x 16 x float> poison, <vscale x 16 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.fadd.nxv1f64(<vscale x 1 x double> poison, <vscale x 1 x double> poison, <vscale x 1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.fadd.nxv2f64(<vscale x 2 x double> poison, <vscale x 2 x double> poison, <vscale x 2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.fadd.nxv4f64(<vscale x 4 x double> poison, <vscale x 4 x double> poison, <vscale x 4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.fadd.nxv8f64(<vscale x 8 x double> poison, <vscale x 8 x double> poison, <vscale x 8 x i1> poison, i32 poison)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
-  %BF16 = fadd bfloat undef, undef
-  %F32 = fadd float undef, undef
-  %F64 = fadd double undef, undef
-
-  %V1BF16 = fadd <1 x bfloat> undef, undef
-  %V2BF16 = fadd <2 x bfloat> undef, undef
-  %V4BF16 = fadd <4 x bfloat> undef, undef
-  %V8BF16 = fadd <8 x bfloat> undef, undef
-  %V16BF16 = fadd <16 x bfloat> undef, undef
-
-  %NXV1BF16 = fadd <vscale x 1 x bfloat> undef, undef
-  %NXV2BF16 = fadd <vscale x 2 x bfloat> undef, undef
-  %NXV4BF16 = fadd <vscale x 4 x bfloat> undef, undef
-  %NXV8BF16 = fadd <vscale x 8 x bfloat> undef, undef
-  %NXV16BF16 = fadd <vscale x 16 x bfloat> undef, undef
-
-  %V1F32 = fadd <1 x float> undef, undef
-  %V2F32 = fadd <2 x float> undef, undef
-  %V4F32 = fadd <4 x float> undef, undef
-  %V8F32 = fadd <8 x float> undef, undef
-  %V16F32 = fadd <16 x float> undef, undef
-
-  %NXV1F32 = fadd <vscale x 1 x float> undef, undef
-  %NXV2F32 = fadd <vscale x 2 x float> undef, undef
-  %NXV4F32 = fadd <vscale x 4 x float> undef, undef
-  %NXV8F32 = fadd <vscale x 8 x float> undef, undef
-  %NXV16F32 = fadd <vscale x 16 x float> undef, undef
-
-  %V1F64 = fadd <1 x double> undef, undef
-  %V2F64 = fadd <2 x double> undef, undef
-  %V4F64 = fadd <4 x double> undef, undef
-  %V8F64 = fadd <8 x double> undef, undef
-
-  %NXV1F64 = fadd <vscale x 1 x double> undef, undef
-  %NXV2F64 = fadd <vscale x 2 x double> undef, undef
-  %NXV4F64 = fadd <vscale x 4 x double> undef, undef
-  %NXV8F64 = fadd <vscale x 8 x double> undef, undef
-
-  %V1F16_VP = call <1 x bfloat> @llvm.vp.fadd.v1f16(<1 x bfloat> undef, <1 x bfloat> undef, <1 x i1> undef, i32 undef)
-  %V2F16_VP = call <2 x bfloat> @llvm.vp.fadd.v2f16(<2 x bfloat> undef, <2 x bfloat> undef, <2 x i1> undef, i32 undef)
-  %V4F16_VP = call <4 x bfloat> @llvm.vp.fadd.v4f16(<4 x bfloat> undef, <4 x bfloat> undef, <4 x i1> undef, i32 undef)
-  %V8F16_VP = call <8 x bfloat> @llvm.vp.fadd.v8f16(<8 x bfloat> undef, <8 x bfloat> undef, <8 x i1> undef, i32 undef)
-  %V16F16_VP = call <16 x bfloat> @llvm.vp.fadd.v16f16(<16 x bfloat> undef, <16 x bfloat> undef, <16 x i1> undef, i32 undef)
-
-  %V1F32_VP = call <1 x float> @llvm.vp.fadd.v1f32(<1 x float> undef, <1 x float> undef, <1 x i1> undef, i32 undef)
-  %V2F32_VP = call <2 x float> @llvm.vp.fadd.v2f32(<2 x float> undef, <2 x float> undef, <2 x i1> undef, i32 undef)
-  %V4F32_VP = call <4 x float> @llvm.vp.fadd.v4f32(<4 x float> undef, <4 x float> undef, <4 x i1> undef, i32 undef)
-  %V8F32_VP = call <8 x float> @llvm.vp.fadd.v8f32(<8 x float> undef, <8 x float> undef, <8 x i1> undef, i32 undef)
-  %V16F32_VP = call <16 x float> @llvm.vp.fadd.v16f32(<16 x float> undef, <16 x float> undef, <16 x i1> undef, i32 undef)
-
-  %V1F64_VP = call <1 x double> @llvm.vp.fadd.v1f64(<1 x double> undef, <1 x double> undef, <1 x i1> undef, i32 undef)
-  %V2F64_VP = call <2 x double> @llvm.vp.fadd.v2f64(<2 x double> undef, <2 x double> undef, <2 x i1> undef, i32 undef)
-  %V4F64_VP = call <4 x double> @llvm.vp.fadd.v4f64(<4 x double> undef, <4 x double> undef, <4 x i1> undef, i32 undef)
-  %V8F64_VP = call <8 x double> @llvm.vp.fadd.v8f64(<8 x double> undef, <8 x double> undef, <8 x i1> undef, i32 undef)
-
-  %NXV1F16_VP = call <vscale x 1 x bfloat> @llvm.vp.fadd.nxv1f16(<vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2F16_VP = call <vscale x 2 x bfloat> @llvm.vp.fadd.nxv2f16(<vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4F16_VP = call <vscale x 4 x bfloat> @llvm.vp.fadd.nxv4f16(<vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8F16_VP = call <vscale x 8 x bfloat> @llvm.vp.fadd.nxv8f16(<vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef, <vscale x 8 x i1> undef, i32 undef)
-  %NXV16F16_VP = call <vscale x 16 x bfloat> @llvm.vp.fadd.nxv16f16(<vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef, <vscale x 16 x i1> undef, i32 undef)
-
-  %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.fadd.nxv1f32(<vscale x 1 x float> undef, <vscale x 1 x float> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.fadd.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.fadd.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x float> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.fadd.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x float> undef, <vscale x 8 x i1> undef, i32 undef)
-  %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.fadd.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x float> undef, <vscale x 16 x i1> undef, i32 undef)
-
-  %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.fadd.nxv1f64(<vscale x 1 x double> undef, <vscale x 1 x double> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.fadd.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x double> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.fadd.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x double> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.fadd.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x double> undef, <vscale x 8 x i1> undef, i32 undef)
+  %F32 = fadd float poison, poison
+  %F64 = fadd double poison, poison
+
+  %V1F32 = fadd <1 x float> poison, poison
+  %V2F32 = fadd <2 x float> poison, poison
+  %V4F32 = fadd <4 x float> poison, poison
+  %V8F32 = fadd <8 x float> poison, poison
+  %V16F32 = fadd <16 x float> poison, poison
+
+  %NXV1F32 = fadd <vscale x 1 x float> poison, poison
+  %NXV2F32 = fadd <vscale x 2 x float> poison, poison
+  %NXV4F32 = fadd <vscale x 4 x float> poison, poison
+  %NXV8F32 = fadd <vscale x 8 x float> poison, poison
+  %NXV16F32 = fadd <vscale x 16 x float> poison, poison
+
+  %V1F64 = fadd <1 x double> poison, poison
+  %V2F64 = fadd <2 x double> poison, poison
+  %V4F64 = fadd <4 x double> poison, poison
+  %V8F64 = fadd <8 x double> poison, poison
+
+  %NXV1F64 = fadd <vscale x 1 x double> poison, poison
+  %NXV2F64 = fadd <vscale x 2 x double> poison, poison
+  %NXV4F64 = fadd <vscale x 4 x double> poison, poison
+  %NXV8F64 = fadd <vscale x 8 x double> poison, poison
+
+  %V1F32_VP = call <1 x float> @llvm.vp.fadd(<1 x float> poison, <1 x float> poison, <1 x i1> poison, i32 poison)
+  %V2F32_VP = call <2 x float> @llvm.vp.fadd(<2 x float> poison, <2 x float> poison, <2 x i1> poison, i32 poison)
+  %V4F32_VP = call <4 x float> @llvm.vp.fadd(<4 x float> poison, <4 x float> poison, <4 x i1> poison, i32 poison)
+  %V8F32_VP = call <8 x float> @llvm.vp.fadd(<8 x float> poison, <8 x float> poison, <8 x i1> poison, i32 poison)
+  %V16F32_VP = call <16 x float> @llvm.vp.fadd(<16 x float> poison, <16 x float> poison, <16 x i1> poison, i32 poison)
+
+  %V1F64_VP = call <1 x double> @llvm.vp.fadd(<1 x double> poison, <1 x double> poison, <1 x i1> poison, i32 poison)
+  %V2F64_VP = call <2 x double> @llvm.vp.fadd(<2 x double> poison, <2 x double> poison, <2 x i1> poison, i32 poison)
+  %V4F64_VP = call <4 x double> @llvm.vp.fadd(<4 x double> poison, <4 x double> poison, <4 x i1> poison, i32 poison)
+  %V8F64_VP = call <8 x double> @llvm.vp.fadd(<8 x double> poison, <8 x double> poison, <8 x i1> poison, i32 poison)
+
+  %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.fadd(<vscale x 1 x float> poison, <vscale x 1 x float> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.fadd(<vscale x 2 x float> poison, <vscale x 2 x float> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.fadd(<vscale x 4 x float> poison, <vscale x 4 x float> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.fadd(<vscale x 8 x float> poison, <vscale x 8 x float> poison, <vscale x 8 x i1> poison, i32 poison)
+  %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.fadd(<vscale x 16 x float> poison, <vscale x 16 x float> poison, <vscale x 16 x i1> poison, i32 poison)
+
+  %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.fadd(<vscale x 1 x double> poison, <vscale x 1 x double> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.fadd(<vscale x 2 x double> poison, <vscale x 2 x double> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.fadd(<vscale x 4 x double> poison, <vscale x 4 x double> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.fadd(<vscale x 8 x double> poison, <vscale x 8 x double> poison, <vscale x 8 x i1> poison, i32 poison)
+
+  ret void
+}
+
+define void @fadd_bf16() {
+; ZVFH-LABEL: 'fadd_bf16'
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = fadd bfloat poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1BF16 = fadd <1 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2BF16 = fadd <2 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4BF16 = fadd <4 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8BF16 = fadd <8 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16BF16 = fadd <16 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1BF16 = fadd <vscale x 1 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2BF16 = fadd <vscale x 2 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4BF16 = fadd <vscale x 4 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8BF16 = fadd <vscale x 8 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16BF16 = fadd <vscale x 16 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1BF16_VP = call <1 x bfloat> @llvm.vp.fadd.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2BF16_VP = call <2 x bfloat> @llvm.vp.fadd.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4BF16_VP = call <4 x bfloat> @llvm.vp.fadd.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8BF16_VP = call <8 x bfloat> @llvm.vp.fadd.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16BF16_VP = call <16 x bfloat> @llvm.vp.fadd.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fadd.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fadd.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fadd.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fadd.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fadd.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; ZVFHMIN-LABEL: 'fadd_bf16'
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = fadd bfloat poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1BF16 = fadd <1 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2BF16 = fadd <2 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4BF16 = fadd <4 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8BF16 = fadd <8 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16BF16 = fadd <16 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1BF16 = fadd <vscale x 1 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2BF16 = fadd <vscale x 2 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4BF16 = fadd <vscale x 4 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8BF16 = fadd <vscale x 8 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16BF16 = fadd <vscale x 16 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1BF16_VP = call <1 x bfloat> @llvm.vp.fadd.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2BF16_VP = call <2 x bfloat> @llvm.vp.fadd.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4BF16_VP = call <4 x bfloat> @llvm.vp.fadd.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8BF16_VP = call <8 x bfloat> @llvm.vp.fadd.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16BF16_VP = call <16 x bfloat> @llvm.vp.fadd.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fadd.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fadd.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fadd.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fadd.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fadd.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; NO-ZFHMIN-LABEL: 'fadd_bf16'
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = fadd bfloat poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1BF16 = fadd <1 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2BF16 = fadd <2 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4BF16 = fadd <4 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8BF16 = fadd <8 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16BF16 = fadd <16 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16 = fadd <vscale x 1 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16 = fadd <vscale x 2 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16 = fadd <vscale x 4 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16 = fadd <vscale x 8 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16 = fadd <vscale x 16 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1BF16_VP = call <1 x bfloat> @llvm.vp.fadd.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2BF16_VP = call <2 x bfloat> @llvm.vp.fadd.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4BF16_VP = call <4 x bfloat> @llvm.vp.fadd.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8BF16_VP = call <8 x bfloat> @llvm.vp.fadd.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16BF16_VP = call <16 x bfloat> @llvm.vp.fadd.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fadd.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fadd.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fadd.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fadd.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fadd.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+  %BF16 = fadd bfloat poison, poison
+
+  %V1BF16 = fadd <1 x bfloat> poison, poison
+  %V2BF16 = fadd <2 x bfloat> poison, poison
+  %V4BF16 = fadd <4 x bfloat> poison, poison
+  %V8BF16 = fadd <8 x bfloat> poison, poison
+  %V16BF16 = fadd <16 x bfloat> poison, poison
+
+  %NXV1BF16 = fadd <vscale x 1 x bfloat> poison, poison
+  %NXV2BF16 = fadd <vscale x 2 x bfloat> poison, poison
+  %NXV4BF16 = fadd <vscale x 4 x bfloat> poison, poison
+  %NXV8BF16 = fadd <vscale x 8 x bfloat> poison, poison
+  %NXV16BF16 = fadd <vscale x 16 x bfloat> poison, poison
+
+  %V1BF16_VP = call <1 x bfloat> @llvm.vp.fadd(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+  %V2BF16_VP = call <2 x bfloat> @llvm.vp.fadd(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+  %V4BF16_VP = call <4 x bfloat> @llvm.vp.fadd(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+  %V8BF16_VP = call <8 x bfloat> @llvm.vp.fadd(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+  %V16BF16_VP = call <16 x bfloat> @llvm.vp.fadd(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+
+  %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fadd(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fadd(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fadd(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fadd(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+  %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fadd(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
 
   ret void
 }
 
 define void @fadd_f16() {
 ; ZVFH-LABEL: 'fadd_f16'
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fadd half undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16 = fadd <1 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16 = fadd <2 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16 = fadd <4 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16 = fadd <8 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16 = fadd <16 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32F16 = fadd <32 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16 = fadd <vscale x 1 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16 = fadd <vscale x 2 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16 = fadd <vscale x 4 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16 = fadd <vscale x 8 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16 = fadd <vscale x 16 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32F16 = fadd <vscale x 32 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fadd.v1f16(<1 x half> undef, <1 x half> undef, <1 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fadd.v2f16(<2 x half> undef, <2 x half> undef, <2 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fadd.v4f16(<4 x half> undef, <4 x half> undef, <4 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fadd.v8f16(<8 x half> undef, <8 x half> undef, <8 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fadd.v16f16(<16 x half> undef, <16 x half> undef, <16 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fadd.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef, <vscale x 1 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fadd.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef, <vscale x 2 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fadd.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef, <vscale x 4 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fadd.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef, <vscale x 8 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fadd.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef, <vscale x 16 x i1> undef, i32 undef)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fadd half poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16 = fadd <1 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16 = fadd <2 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16 = fadd <4 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16 = fadd <8 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16 = fadd <16 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32F16 = fadd <32 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16 = fadd <vscale x 1 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16 = fadd <vscale x 2 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16 = fadd <vscale x 4 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16 = fadd <vscale x 8 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16 = fadd <vscale x 16 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32F16 = fadd <vscale x 32 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fadd.v1f16(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fadd.v2f16(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fadd.v4f16(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fadd.v8f16(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fadd.v16f16(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fadd.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fadd.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fadd.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fadd.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fadd.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
 ; ZVFH-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; ZVFHMIN-LABEL: 'fadd_f16'
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fadd half undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1F16 = fadd <1 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2F16 = fadd <2 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4F16 = fadd <4 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8F16 = fadd <8 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16F16 = fadd <16 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32F16 = fadd <32 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1F16 = fadd <vscale x 1 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2F16 = fadd <vscale x 2 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4F16 = fadd <vscale x 4 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8F16 = fadd <vscale x 8 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16F16 = fadd <vscale x 16 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32F16 = fadd <vscale x 32 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fadd.v1f16(<1 x half> undef, <1 x half> undef, <1 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fadd.v2f16(<2 x half> undef, <2 x half> undef, <2 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fadd.v4f16(<4 x half> undef, <4 x half> undef, <4 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fadd.v8f16(<8 x half> undef, <8 x half> undef, <8 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fadd.v16f16(<16 x half> undef, <16 x half> undef, <16 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fadd.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef, <vscale x 1 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fadd.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef, <vscale x 2 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fadd.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef, <vscale x 4 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fadd.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef, <vscale x 8 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fadd.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef, <vscale x 16 x i1> undef, i32 undef)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fadd half poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1F16 = fadd <1 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2F16 = fadd <2 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4F16 = fadd <4 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8F16 = fadd <8 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16F16 = fadd <16 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32F16 = fadd <32 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1F16 = fadd <vscale x 1 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2F16 = fadd <vscale x 2 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4F16 = fadd <vscale x 4 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8F16 = fadd <vscale x 8 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16F16 = fadd <vscale x 16 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32F16 = fadd <vscale x 32 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fadd.v1f16(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fadd.v2f16(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fadd.v4f16(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fadd.v8f16(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fadd.v16f16(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fadd.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fadd.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fadd.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fadd.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fadd.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
 ; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
-  %F16 = fadd half undef, undef
-
-  %V1F16 = fadd <1 x half> undef, undef
-  %V2F16 = fadd <2 x half> undef, undef
-  %V4F16 = fadd <4 x half> undef, undef
-  %V8F16 = fadd <8 x half> undef, undef
-  %V16F16 = fadd <16 x half> undef, undef
-  %V32F16 = fadd <32 x half> undef, undef
-
-  %NXV1F16 = fadd <vscale x 1 x half> undef, undef
-  %NXV2F16 = fadd <vscale x 2 x half> undef, undef
-  %NXV4F16 = fadd <vscale x 4 x half> undef, undef
-  %NXV8F16 = fadd <vscale x 8 x half> undef, undef
-  %NXV16F16 = fadd <vscale x 16 x half> undef, undef
-  %NXV32F16 = fadd <vscale x 32 x half> undef, undef
-
-  %V1F16_VP = call <1 x half> @llvm.vp.fadd.v1f16(<1 x half> undef, <1 x half> undef, <1 x i1> undef, i32 undef)
-  %V2F16_VP = call <2 x half> @llvm.vp.fadd.v2f16(<2 x half> undef, <2 x half> undef, <2 x i1> undef, i32 undef)
-  %V4F16_VP = call <4 x half> @llvm.vp.fadd.v4f16(<4 x half> undef, <4 x half> undef, <4 x i1> undef, i32 undef)
-  %V8F16_VP = call <8 x half> @llvm.vp.fadd.v8f16(<8 x half> undef, <8 x half> undef, <8 x i1> undef, i32 undef)
-  %V16F16_VP = call <16 x half> @llvm.vp.fadd.v16f16(<16 x half> undef, <16 x half> undef, <16 x i1> undef, i32 undef)
-
-  %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fadd.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fadd.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fadd.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fadd.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef, <vscale x 8 x i1> undef, i32 undef)
-  %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fadd.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef, <vscale x 16 x i1> undef, i32 undef)
+; NO-ZFHMIN-LABEL: 'fadd_f16'
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fadd half poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16 = fadd <1 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2F16 = fadd <2 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4F16 = fadd <4 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8F16 = fadd <8 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16F16 = fadd <16 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V32F16 = fadd <32 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16 = fadd <vscale x 1 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16 = fadd <vscale x 2 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16 = fadd <vscale x 4 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16 = fadd <vscale x 8 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16 = fadd <vscale x 16 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32F16 = fadd <vscale x 32 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fadd.v1f16(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fadd.v2f16(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fadd.v4f16(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fadd.v8f16(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fadd.v16f16(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fadd.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fadd.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fadd.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fadd.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fadd.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+  %F16 = fadd half poison, poison
+
+  %V1F16 = fadd <1 x half> poison, poison
+  %V2F16 = fadd <2 x half> poison, poison
+  %V4F16 = fadd <4 x half> poison, poison
+  %V8F16 = fadd <8 x half> poison, poison
+  %V16F16 = fadd <16 x half> poison, poison
+  %V32F16 = fadd <32 x half> poison, poison
+
+  %NXV1F16 = fadd <vscale x 1 x half> poison, poison
+  %NXV2F16 = fadd <vscale x 2 x half> poison, poison
+  %NXV4F16 = fadd <vscale x 4 x half> poison, poison
+  %NXV8F16 = fadd <vscale x 8 x half> poison, poison
+  %NXV16F16 = fadd <vscale x 16 x half> poison, poison
+  %NXV32F16 = fadd <vscale x 32 x half> poison, poison
+
+  %V1F16_VP = call <1 x half> @llvm.vp.fadd(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+  %V2F16_VP = call <2 x half> @llvm.vp.fadd(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+  %V4F16_VP = call <4 x half> @llvm.vp.fadd(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+  %V8F16_VP = call <8 x half> @llvm.vp.fadd(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+  %V16F16_VP = call <16 x half> @llvm.vp.fadd(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+
+  %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fadd(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fadd(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fadd(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fadd(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+  %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fadd(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
 
   ret void
 }
 
 define void @fsub() {
 ; CHECK-LABEL: 'fsub'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fsub half undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = fsub float undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = fsub double undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1BF16 = fsub <1 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2BF16 = fsub <2 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4BF16 = fsub <4 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8BF16 = fsub <8 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16BF16 = fsub <16 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1BF16 = fsub <vscale x 1 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2BF16 = fsub <vscale x 2 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4BF16 = fsub <vscale x 4 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8BF16 = fsub <vscale x 8 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16BF16 = fsub <vscale x 16 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32 = fsub <1 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32 = fsub <2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fsub <4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fsub <8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fsub <16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F32 = fsub <vscale x 1 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F32 = fsub <vscale x 2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F32 = fsub <vscale x 4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F32 = fsub <vscale x 8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16F32 = fsub <vscale x 16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64 = fsub <1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fsub <2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fsub <4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fsub <8 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F64 = fsub <vscale x 1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2F64 = fsub <vscale x 2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4F64 = fsub <vscale x 4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8F64 = fsub <vscale x 8 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1F16_VP = call <1 x bfloat> @llvm.vp.fsub.v1bf16(<1 x bfloat> undef, <1 x bfloat> undef, <1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2F16_VP = call <2 x bfloat> @llvm.vp.fsub.v2bf16(<2 x bfloat> undef, <2 x bfloat> undef, <2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4F16_VP = call <4 x bfloat> @llvm.vp.fsub.v4bf16(<4 x bfloat> undef, <4 x bfloat> undef, <4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8F16_VP = call <8 x bfloat> @llvm.vp.fsub.v8bf16(<8 x bfloat> undef, <8 x bfloat> undef, <8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16F16_VP = call <16 x bfloat> @llvm.vp.fsub.v16bf16(<16 x bfloat> undef, <16 x bfloat> undef, <16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32_VP = call <1 x float> @llvm.vp.fsub.v1f32(<1 x float> undef, <1 x float> undef, <1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32_VP = call <2 x float> @llvm.vp.fsub.v2f32(<2 x float> undef, <2 x float> undef, <2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32_VP = call <4 x float> @llvm.vp.fsub.v4f32(<4 x float> undef, <4 x float> undef, <4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32_VP = call <8 x float> @llvm.vp.fsub.v8f32(<8 x float> undef, <8 x float> undef, <8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32_VP = call <16 x float> @llvm.vp.fsub.v16f32(<16 x float> undef, <16 x float> undef, <16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64_VP = call <1 x double> @llvm.vp.fsub.v1f64(<1 x double> undef, <1 x double> undef, <1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64_VP = call <2 x double> @llvm.vp.fsub.v2f64(<2 x double> undef, <2 x double> undef, <2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64_VP = call <4 x double> @llvm.vp.fsub.v4f64(<4 x double> undef, <4 x double> undef, <4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64_VP = call <8 x double> @llvm.vp.fsub.v8f64(<8 x double> undef, <8 x double> undef, <8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1F16_VP = call <vscale x 1 x bfloat> @llvm.vp.fsub.nxv1bf16(<vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef, <vscale x 1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2F16_VP = call <vscale x 2 x bfloat> @llvm.vp.fsub.nxv2bf16(<vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef, <vscale x 2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4F16_VP = call <vscale x 4 x bfloat> @llvm.vp.fsub.nxv4bf16(<vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef, <vscale x 4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8F16_VP = call <vscale x 8 x bfloat> @llvm.vp.fsub.nxv8bf16(<vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef, <vscale x 8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16F16_VP = call <vscale x 16 x bfloat> @llvm.vp.fsub.nxv16bf16(<vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef, <vscale x 16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.fsub.nxv1f32(<vscale x 1 x float> undef, <vscale x 1 x float> undef, <vscale x 1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.fsub.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> undef, <vscale x 2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.fsub.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x float> undef, <vscale x 4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.fsub.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x float> undef, <vscale x 8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.fsub.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x float> undef, <vscale x 16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.fsub.nxv1f64(<vscale x 1 x double> undef, <vscale x 1 x double> undef, <vscale x 1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.fsub.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x double> undef, <vscale x 2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.fsub.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x double> undef, <vscale x 4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.fsub.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x double> undef, <vscale x 8 x i1> undef, i32 undef)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = fsub float poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = fsub double poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32 = fsub <1 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32 = fsub <2 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fsub <4 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fsub <8 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fsub <16 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F32 = fsub <vscale x 1 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F32 = fsub <vscale x 2 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F32 = fsub <vscale x 4 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F32 = fsub <vscale x 8 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16F32 = fsub <vscale x 16 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64 = fsub <1 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fsub <2 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fsub <4 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fsub <8 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F64 = fsub <vscale x 1 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2F64 = fsub <vscale x 2 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4F64 = fsub <vscale x 4 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8F64 = fsub <vscale x 8 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32_VP = call <1 x float> @llvm.vp.fsub.v1f32(<1 x float> poison, <1 x float> poison, <1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32_VP = call <2 x float> @llvm.vp.fsub.v2f32(<2 x float> poison, <2 x float> poison, <2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32_VP = call <4 x float> @llvm.vp.fsub.v4f32(<4 x float> poison, <4 x float> poison, <4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32_VP = call <8 x float> @llvm.vp.fsub.v8f32(<8 x float> poison, <8 x float> poison, <8 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32_VP = call <16 x float> @llvm.vp.fsub.v16f32(<16 x float> poison, <16 x float> poison, <16 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64_VP = call <1 x double> @llvm.vp.fsub.v1f64(<1 x double> poison, <1 x double> poison, <1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64_VP = call <2 x double> @llvm.vp.fsub.v2f64(<2 x double> poison, <2 x double> poison, <2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64_VP = call <4 x double> @llvm.vp.fsub.v4f64(<4 x double> poison, <4 x double> poison, <4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64_VP = call <8 x double> @llvm.vp.fsub.v8f64(<8 x double> poison, <8 x double> poison, <8 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.fsub.nxv1f32(<vscale x 1 x float> poison, <vscale x 1 x float> poison, <vscale x 1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.fsub.nxv2f32(<vscale x 2 x float> poison, <vscale x 2 x float> poison, <vscale x 2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.fsub.nxv4f32(<vscale x 4 x float> poison, <vscale x 4 x float> poison, <vscale x 4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.fsub.nxv8f32(<vscale x 8 x float> poison, <vscale x 8 x float> poison, <vscale x 8 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.fsub.nxv16f32(<vscale x 16 x float> poison, <vscale x 16 x float> poison, <vscale x 16 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.fsub.nxv1f64(<vscale x 1 x double> poison, <vscale x 1 x double> poison, <vscale x 1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.fsub.nxv2f64(<vscale x 2 x double> poison, <vscale x 2 x double> poison, <vscale x 2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.fsub.nxv4f64(<vscale x 4 x double> poison, <vscale x 4 x double> poison, <vscale x 4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.fsub.nxv8f64(<vscale x 8 x double> poison, <vscale x 8 x double> poison, <vscale x 8 x i1> poison, i32 poison)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
-  %F16 = fsub half undef, undef
-  %F32 = fsub float undef, undef
-  %F64 = fsub double undef, undef
-
-  %V1BF16 = fsub <1 x bfloat> undef, undef
-  %V2BF16 = fsub <2 x bfloat> undef, undef
-  %V4BF16 = fsub <4 x bfloat> undef, undef
-  %V8BF16 = fsub <8 x bfloat> undef, undef
-  %V16BF16 = fsub <16 x bfloat> undef, undef
-
-  %NXV1BF16 = fsub <vscale x 1 x bfloat> undef, undef
-  %NXV2BF16 = fsub <vscale x 2 x bfloat> undef, undef
-  %NXV4BF16 = fsub <vscale x 4 x bfloat> undef, undef
-  %NXV8BF16 = fsub <vscale x 8 x bfloat> undef, undef
-  %NXV16BF16 = fsub <vscale x 16 x bfloat> undef, undef
-
-  %V1F32 = fsub <1 x float> undef, undef
-  %V2F32 = fsub <2 x float> undef, undef
-  %V4F32 = fsub <4 x float> undef, undef
-  %V8F32 = fsub <8 x float> undef, undef
-  %V16F32 = fsub <16 x float> undef, undef
-
-  %NXV1F32 = fsub <vscale x 1 x float> undef, undef
-  %NXV2F32 = fsub <vscale x 2 x float> undef, undef
-  %NXV4F32 = fsub <vscale x 4 x float> undef, undef
-  %NXV8F32 = fsub <vscale x 8 x float> undef, undef
-  %NXV16F32 = fsub <vscale x 16 x float> undef, undef
-
-  %V1F64 = fsub <1 x double> undef, undef
-  %V2F64 = fsub <2 x double> undef, undef
-  %V4F64 = fsub <4 x double> undef, undef
-  %V8F64 = fsub <8 x double> undef, undef
-
-  %NXV1F64 = fsub <vscale x 1 x double> undef, undef
-  %NXV2F64 = fsub <vscale x 2 x double> undef, undef
-  %NXV4F64 = fsub <vscale x 4 x double> undef, undef
-  %NXV8F64 = fsub <vscale x 8 x double> undef, undef
-
-  %V1F16_VP = call <1 x bfloat> @llvm.vp.fsub.v1f16(<1 x bfloat> undef, <1 x bfloat> undef, <1 x i1> undef, i32 undef)
-  %V2F16_VP = call <2 x bfloat> @llvm.vp.fsub.v2f16(<2 x bfloat> undef, <2 x bfloat> undef, <2 x i1> undef, i32 undef)
-  %V4F16_VP = call <4 x bfloat> @llvm.vp.fsub.v4f16(<4 x bfloat> undef, <4 x bfloat> undef, <4 x i1> undef, i32 undef)
-  %V8F16_VP = call <8 x bfloat> @llvm.vp.fsub.v8f16(<8 x bfloat> undef, <8 x bfloat> undef, <8 x i1> undef, i32 undef)
-  %V16F16_VP = call <16 x bfloat> @llvm.vp.fsub.v16f16(<16 x bfloat> undef, <16 x bfloat> undef, <16 x i1> undef, i32 undef)
-
-  %V1F32_VP = call <1 x float> @llvm.vp.fsub.v1f32(<1 x float> undef, <1 x float> undef, <1 x i1> undef, i32 undef)
-  %V2F32_VP = call <2 x float> @llvm.vp.fsub.v2f32(<2 x float> undef, <2 x float> undef, <2 x i1> undef, i32 undef)
-  %V4F32_VP = call <4 x float> @llvm.vp.fsub.v4f32(<4 x float> undef, <4 x float> undef, <4 x i1> undef, i32 undef)
-  %V8F32_VP = call <8 x float> @llvm.vp.fsub.v8f32(<8 x float> undef, <8 x float> undef, <8 x i1> undef, i32 undef)
-  %V16F32_VP = call <16 x float> @llvm.vp.fsub.v16f32(<16 x float> undef, <16 x float> undef, <16 x i1> undef, i32 undef)
-
-  %V1F64_VP = call <1 x double> @llvm.vp.fsub.v1f64(<1 x double> undef, <1 x double> undef, <1 x i1> undef, i32 undef)
-  %V2F64_VP = call <2 x double> @llvm.vp.fsub.v2f64(<2 x double> undef, <2 x double> undef, <2 x i1> undef, i32 undef)
-  %V4F64_VP = call <4 x double> @llvm.vp.fsub.v4f64(<4 x double> undef, <4 x double> undef, <4 x i1> undef, i32 undef)
-  %V8F64_VP = call <8 x double> @llvm.vp.fsub.v8f64(<8 x double> undef, <8 x double> undef, <8 x i1> undef, i32 undef)
-
-  %NXV1F16_VP = call <vscale x 1 x bfloat> @llvm.vp.fsub.nxv1f16(<vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2F16_VP = call <vscale x 2 x bfloat> @llvm.vp.fsub.nxv2f16(<vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4F16_VP = call <vscale x 4 x bfloat> @llvm.vp.fsub.nxv4f16(<vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8F16_VP = call <vscale x 8 x bfloat> @llvm.vp.fsub.nxv8f16(<vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef, <vscale x 8 x i1> undef, i32 undef)
-  %NXV16F16_VP = call <vscale x 16 x bfloat> @llvm.vp.fsub.nxv16f16(<vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef, <vscale x 16 x i1> undef, i32 undef)
-
-  %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.fsub.nxv1f32(<vscale x 1 x float> undef, <vscale x 1 x float> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.fsub.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.fsub.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x float> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.fsub.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x float> undef, <vscale x 8 x i1> undef, i32 undef)
-  %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.fsub.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x float> undef, <vscale x 16 x i1> undef, i32 undef)
-
-  %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.fsub.nxv1f64(<vscale x 1 x double> undef, <vscale x 1 x double> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.fsub.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x double> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.fsub.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x double> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.fsub.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x double> undef, <vscale x 8 x i1> undef, i32 undef)
+  %F32 = fsub float poison, poison
+  %F64 = fsub double poison, poison
+
+  %V1F32 = fsub <1 x float> poison, poison
+  %V2F32 = fsub <2 x float> poison, poison
+  %V4F32 = fsub <4 x float> poison, poison
+  %V8F32 = fsub <8 x float> poison, poison
+  %V16F32 = fsub <16 x float> poison, poison
+
+  %NXV1F32 = fsub <vscale x 1 x float> poison, poison
+  %NXV2F32 = fsub <vscale x 2 x float> poison, poison
+  %NXV4F32 = fsub <vscale x 4 x float> poison, poison
+  %NXV8F32 = fsub <vscale x 8 x float> poison, poison
+  %NXV16F32 = fsub <vscale x 16 x float> poison, poison
+
+  %V1F64 = fsub <1 x double> poison, poison
+  %V2F64 = fsub <2 x double> poison, poison
+  %V4F64 = fsub <4 x double> poison, poison
+  %V8F64 = fsub <8 x double> poison, poison
+
+  %NXV1F64 = fsub <vscale x 1 x double> poison, poison
+  %NXV2F64 = fsub <vscale x 2 x double> poison, poison
+  %NXV4F64 = fsub <vscale x 4 x double> poison, poison
+  %NXV8F64 = fsub <vscale x 8 x double> poison, poison
+
+  %V1F32_VP = call <1 x float> @llvm.vp.fsub(<1 x float> poison, <1 x float> poison, <1 x i1> poison, i32 poison)
+  %V2F32_VP = call <2 x float> @llvm.vp.fsub(<2 x float> poison, <2 x float> poison, <2 x i1> poison, i32 poison)
+  %V4F32_VP = call <4 x float> @llvm.vp.fsub(<4 x float> poison, <4 x float> poison, <4 x i1> poison, i32 poison)
+  %V8F32_VP = call <8 x float> @llvm.vp.fsub(<8 x float> poison, <8 x float> poison, <8 x i1> poison, i32 poison)
+  %V16F32_VP = call <16 x float> @llvm.vp.fsub(<16 x float> poison, <16 x float> poison, <16 x i1> poison, i32 poison)
+
+  %V1F64_VP = call <1 x double> @llvm.vp.fsub(<1 x double> poison, <1 x double> poison, <1 x i1> poison, i32 poison)
+  %V2F64_VP = call <2 x double> @llvm.vp.fsub(<2 x double> poison, <2 x double> poison, <2 x i1> poison, i32 poison)
+  %V4F64_VP = call <4 x double> @llvm.vp.fsub(<4 x double> poison, <4 x double> poison, <4 x i1> poison, i32 poison)
+  %V8F64_VP = call <8 x double> @llvm.vp.fsub(<8 x double> poison, <8 x double> poison, <8 x i1> poison, i32 poison)
+
+  %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.fsub(<vscale x 1 x float> poison, <vscale x 1 x float> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.fsub(<vscale x 2 x float> poison, <vscale x 2 x float> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.fsub(<vscale x 4 x float> poison, <vscale x 4 x float> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.fsub(<vscale x 8 x float> poison, <vscale x 8 x float> poison, <vscale x 8 x i1> poison, i32 poison)
+  %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.fsub(<vscale x 16 x float> poison, <vscale x 16 x float> poison, <vscale x 16 x i1> poison, i32 poison)
+
+  %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.fsub(<vscale x 1 x double> poison, <vscale x 1 x double> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.fsub(<vscale x 2 x double> poison, <vscale x 2 x double> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.fsub(<vscale x 4 x double> poison, <vscale x 4 x double> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.fsub(<vscale x 8 x double> poison, <vscale x 8 x double> poison, <vscale x 8 x i1> poison, i32 poison)
+
+  ret void
+}
+
+define void @fsub_bf16() {
+; ZVFH-LABEL: 'fsub_bf16'
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = fsub bfloat poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1BF16 = fsub <1 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2BF16 = fsub <2 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4BF16 = fsub <4 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8BF16 = fsub <8 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16BF16 = fsub <16 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32BF16 = fsub <32 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1BF16 = fsub <vscale x 1 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2BF16 = fsub <vscale x 2 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4BF16 = fsub <vscale x 4 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8BF16 = fsub <vscale x 8 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16BF16 = fsub <vscale x 16 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32BF16 = fsub <vscale x 32 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1BF16_VP = call <1 x bfloat> @llvm.vp.fsub.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2BF16_VP = call <2 x bfloat> @llvm.vp.fsub.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4BF16_VP = call <4 x bfloat> @llvm.vp.fsub.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8BF16_VP = call <8 x bfloat> @llvm.vp.fsub.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16BF16_VP = call <16 x bfloat> @llvm.vp.fsub.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fsub.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fsub.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fsub.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fsub.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fsub.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; ZVFHMIN-LABEL: 'fsub_bf16'
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = fsub bfloat poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1BF16 = fsub <1 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2BF16 = fsub <2 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4BF16 = fsub <4 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8BF16 = fsub <8 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16BF16 = fsub <16 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32BF16 = fsub <32 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1BF16 = fsub <vscale x 1 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2BF16 = fsub <vscale x 2 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4BF16 = fsub <vscale x 4 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8BF16 = fsub <vscale x 8 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16BF16 = fsub <vscale x 16 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32BF16 = fsub <vscale x 32 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1BF16_VP = call <1 x bfloat> @llvm.vp.fsub.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2BF16_VP = call <2 x bfloat> @llvm.vp.fsub.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4BF16_VP = call <4 x bfloat> @llvm.vp.fsub.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8BF16_VP = call <8 x bfloat> @llvm.vp.fsub.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16BF16_VP = call <16 x bfloat> @llvm.vp.fsub.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fsub.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fsub.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fsub.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fsub.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fsub.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; NO-ZFHMIN-LABEL: 'fsub_bf16'
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = fsub bfloat poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1BF16 = fsub <1 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2BF16 = fsub <2 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4BF16 = fsub <4 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8BF16 = fsub <8 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16BF16 = fsub <16 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V32BF16 = fsub <32 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16 = fsub <vscale x 1 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16 = fsub <vscale x 2 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16 = fsub <vscale x 4 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16 = fsub <vscale x 8 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16 = fsub <vscale x 16 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32BF16 = fsub <vscale x 32 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1BF16_VP = call <1 x bfloat> @llvm.vp.fsub.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2BF16_VP = call <2 x bfloat> @llvm.vp.fsub.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4BF16_VP = call <4 x bfloat> @llvm.vp.fsub.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8BF16_VP = call <8 x bfloat> @llvm.vp.fsub.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16BF16_VP = call <16 x bfloat> @llvm.vp.fsub.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fsub.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fsub.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fsub.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fsub.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fsub.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+  %BF16 = fsub bfloat poison, poison
+
+  %V1BF16 = fsub <1 x bfloat> poison, poison
+  %V2BF16 = fsub <2 x bfloat> poison, poison
+  %V4BF16 = fsub <4 x bfloat> poison, poison
+  %V8BF16 = fsub <8 x bfloat> poison, poison
+  %V16BF16 = fsub <16 x bfloat> poison, poison
+  %V32BF16 = fsub <32 x bfloat> poison, poison
+
+  %NXV1BF16 = fsub <vscale x 1 x bfloat> poison, poison
+  %NXV2BF16 = fsub <vscale x 2 x bfloat> poison, poison
+  %NXV4BF16 = fsub <vscale x 4 x bfloat> poison, poison
+  %NXV8BF16 = fsub <vscale x 8 x bfloat> poison, poison
+  %NXV16BF16 = fsub <vscale x 16 x bfloat> poison, poison
+  %NXV32BF16 = fsub <vscale x 32 x bfloat> poison, poison
+
+  %V1BF16_VP = call <1 x bfloat> @llvm.vp.fsub(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+  %V2BF16_VP = call <2 x bfloat> @llvm.vp.fsub(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+  %V4BF16_VP = call <4 x bfloat> @llvm.vp.fsub(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+  %V8BF16_VP = call <8 x bfloat> @llvm.vp.fsub(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+  %V16BF16_VP = call <16 x bfloat> @llvm.vp.fsub(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+
+  %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fsub(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fsub(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fsub(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fsub(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+  %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fsub(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
 
   ret void
 }
 
 define void @fsub_f16() {
 ; ZVFH-LABEL: 'fsub_f16'
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fsub half undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16 = fsub <1 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16 = fsub <2 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16 = fsub <4 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16 = fsub <8 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16 = fsub <16 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32F16 = fsub <32 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16 = fsub <vscale x 1 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16 = fsub <vscale x 2 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16 = fsub <vscale x 4 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16 = fsub <vscale x 8 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16 = fsub <vscale x 16 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32F16 = fsub <vscale x 32 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fsub.v1f16(<1 x half> undef, <1 x half> undef, <1 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fsub.v2f16(<2 x half> undef, <2 x half> undef, <2 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fsub.v4f16(<4 x half> undef, <4 x half> undef, <4 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fsub.v8f16(<8 x half> undef, <8 x half> undef, <8 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fsub.v16f16(<16 x half> undef, <16 x half> undef, <16 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fsub.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef, <vscale x 1 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fsub.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef, <vscale x 2 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fsub.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef, <vscale x 4 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fsub.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef, <vscale x 8 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fsub.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef, <vscale x 16 x i1> undef, i32 undef)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fsub half poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16 = fsub <1 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16 = fsub <2 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16 = fsub <4 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16 = fsub <8 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16 = fsub <16 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32F16 = fsub <32 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16 = fsub <vscale x 1 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16 = fsub <vscale x 2 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16 = fsub <vscale x 4 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16 = fsub <vscale x 8 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16 = fsub <vscale x 16 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32F16 = fsub <vscale x 32 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fsub.v1f16(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fsub.v2f16(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fsub.v4f16(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fsub.v8f16(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fsub.v16f16(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fsub.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fsub.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fsub.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fsub.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fsub.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
 ; ZVFH-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; ZVFHMIN-LABEL: 'fsub_f16'
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fsub half undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1F16 = fsub <1 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2F16 = fsub <2 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4F16 = fsub <4 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8F16 = fsub <8 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16F16 = fsub <16 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32F16 = fsub <32 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1F16 = fsub <vscale x 1 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2F16 = fsub <vscale x 2 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4F16 = fsub <vscale x 4 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8F16 = fsub <vscale x 8 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16F16 = fsub <vscale x 16 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32F16 = fsub <vscale x 32 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fsub.v1f16(<1 x half> undef, <1 x half> undef, <1 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fsub.v2f16(<2 x half> undef, <2 x half> undef, <2 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fsub.v4f16(<4 x half> undef, <4 x half> undef, <4 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fsub.v8f16(<8 x half> undef, <8 x half> undef, <8 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fsub.v16f16(<16 x half> undef, <16 x half> undef, <16 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fsub.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef, <vscale x 1 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fsub.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef, <vscale x 2 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fsub.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef, <vscale x 4 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fsub.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef, <vscale x 8 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fsub.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef, <vscale x 16 x i1> undef, i32 undef)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fsub half poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1F16 = fsub <1 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2F16 = fsub <2 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4F16 = fsub <4 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8F16 = fsub <8 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16F16 = fsub <16 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32F16 = fsub <32 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1F16 = fsub <vscale x 1 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2F16 = fsub <vscale x 2 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4F16 = fsub <vscale x 4 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8F16 = fsub <vscale x 8 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16F16 = fsub <vscale x 16 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32F16 = fsub <vscale x 32 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fsub.v1f16(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fsub.v2f16(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fsub.v4f16(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fsub.v8f16(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fsub.v16f16(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fsub.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fsub.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fsub.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fsub.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fsub.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
 ; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
-  %F16 = fsub half undef, undef
-
-  %V1F16 = fsub <1 x half> undef, undef
-  %V2F16 = fsub <2 x half> undef, undef
-  %V4F16 = fsub <4 x half> undef, undef
-  %V8F16 = fsub <8 x half> undef, undef
-  %V16F16 = fsub <16 x half> undef, undef
-  %V32F16 = fsub <32 x half> undef, undef
-
-  %NXV1F16 = fsub <vscale x 1 x half> undef, undef
-  %NXV2F16 = fsub <vscale x 2 x half> undef, undef
-  %NXV4F16 = fsub <vscale x 4 x half> undef, undef
-  %NXV8F16 = fsub <vscale x 8 x half> undef, undef
-  %NXV16F16 = fsub <vscale x 16 x half> undef, undef
-  %NXV32F16 = fsub <vscale x 32 x half> undef, undef
-
-  %V1F16_VP = call <1 x half> @llvm.vp.fsub.v1f16(<1 x half> undef, <1 x half> undef, <1 x i1> undef, i32 undef)
-  %V2F16_VP = call <2 x half> @llvm.vp.fsub.v2f16(<2 x half> undef, <2 x half> undef, <2 x i1> undef, i32 undef)
-  %V4F16_VP = call <4 x half> @llvm.vp.fsub.v4f16(<4 x half> undef, <4 x half> undef, <4 x i1> undef, i32 undef)
-  %V8F16_VP = call <8 x half> @llvm.vp.fsub.v8f16(<8 x half> undef, <8 x half> undef, <8 x i1> undef, i32 undef)
-  %V16F16_VP = call <16 x half> @llvm.vp.fsub.v16f16(<16 x half> undef, <16 x half> undef, <16 x i1> undef, i32 undef)
-
-  %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fsub.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fsub.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fsub.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fsub.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef, <vscale x 8 x i1> undef, i32 undef)
-  %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fsub.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef, <vscale x 16 x i1> undef, i32 undef)
+; NO-ZFHMIN-LABEL: 'fsub_f16'
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fsub half poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16 = fsub <1 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2F16 = fsub <2 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4F16 = fsub <4 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8F16 = fsub <8 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16F16 = fsub <16 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V32F16 = fsub <32 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16 = fsub <vscale x 1 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16 = fsub <vscale x 2 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16 = fsub <vscale x 4 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16 = fsub <vscale x 8 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16 = fsub <vscale x 16 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32F16 = fsub <vscale x 32 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fsub.v1f16(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fsub.v2f16(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fsub.v4f16(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fsub.v8f16(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fsub.v16f16(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fsub.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fsub.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fsub.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fsub.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fsub.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+  %F16 = fsub half poison, poison
+
+  %V1F16 = fsub <1 x half> poison, poison
+  %V2F16 = fsub <2 x half> poison, poison
+  %V4F16 = fsub <4 x half> poison, poison
+  %V8F16 = fsub <8 x half> poison, poison
+  %V16F16 = fsub <16 x half> poison, poison
+  %V32F16 = fsub <32 x half> poison, poison
+
+  %NXV1F16 = fsub <vscale x 1 x half> poison, poison
+  %NXV2F16 = fsub <vscale x 2 x half> poison, poison
+  %NXV4F16 = fsub <vscale x 4 x half> poison, poison
+  %NXV8F16 = fsub <vscale x 8 x half> poison, poison
+  %NXV16F16 = fsub <vscale x 16 x half> poison, poison
+  %NXV32F16 = fsub <vscale x 32 x half> poison, poison
+
+  %V1F16_VP = call <1 x half> @llvm.vp.fsub(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+  %V2F16_VP = call <2 x half> @llvm.vp.fsub(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+  %V4F16_VP = call <4 x half> @llvm.vp.fsub(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+  %V8F16_VP = call <8 x half> @llvm.vp.fsub(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+  %V16F16_VP = call <16 x half> @llvm.vp.fsub(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+
+  %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fsub(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fsub(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fsub(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fsub(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+  %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fsub(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
 
   ret void
 }
 
 define void @fmul() {
 ; CHECK-LABEL: 'fmul'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = fmul bfloat undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = fmul float undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = fmul double undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1BF16 = fmul <1 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2BF16 = fmul <2 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4BF16 = fmul <4 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8BF16 = fmul <8 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16BF16 = fmul <16 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1BF16 = fmul <vscale x 1 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2BF16 = fmul <vscale x 2 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4BF16 = fmul <vscale x 4 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8BF16 = fmul <vscale x 8 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16BF16 = fmul <vscale x 16 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32 = fmul <1 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32 = fmul <2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fmul <4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fmul <8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fmul <16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F32 = fmul <vscale x 1 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F32 = fmul <vscale x 2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F32 = fmul <vscale x 4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F32 = fmul <vscale x 8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16F32 = fmul <vscale x 16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64 = fmul <1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fmul <2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fmul <4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fmul <8 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F64 = fmul <vscale x 1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2F64 = fmul <vscale x 2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4F64 = fmul <vscale x 4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8F64 = fmul <vscale x 8 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1F16_VP = call <1 x bfloat> @llvm.vp.fmul.v1bf16(<1 x bfloat> undef, <1 x bfloat> undef, <1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2F16_VP = call <2 x bfloat> @llvm.vp.fmul.v2bf16(<2 x bfloat> undef, <2 x bfloat> undef, <2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4F16_VP = call <4 x bfloat> @llvm.vp.fmul.v4bf16(<4 x bfloat> undef, <4 x bfloat> undef, <4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8F16_VP = call <8 x bfloat> @llvm.vp.fmul.v8bf16(<8 x bfloat> undef, <8 x bfloat> undef, <8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16F16_VP = call <16 x bfloat> @llvm.vp.fmul.v16bf16(<16 x bfloat> undef, <16 x bfloat> undef, <16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32_VP = call <1 x float> @llvm.vp.fmul.v1f32(<1 x float> undef, <1 x float> undef, <1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32_VP = call <2 x float> @llvm.vp.fmul.v2f32(<2 x float> undef, <2 x float> undef, <2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32_VP = call <4 x float> @llvm.vp.fmul.v4f32(<4 x float> undef, <4 x float> undef, <4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32_VP = call <8 x float> @llvm.vp.fmul.v8f32(<8 x float> undef, <8 x float> undef, <8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32_VP = call <16 x float> @llvm.vp.fmul.v16f32(<16 x float> undef, <16 x float> undef, <16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64_VP = call <1 x double> @llvm.vp.fmul.v1f64(<1 x double> undef, <1 x double> undef, <1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64_VP = call <2 x double> @llvm.vp.fmul.v2f64(<2 x double> undef, <2 x double> undef, <2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64_VP = call <4 x double> @llvm.vp.fmul.v4f64(<4 x double> undef, <4 x double> undef, <4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64_VP = call <8 x double> @llvm.vp.fmul.v8f64(<8 x double> undef, <8 x double> undef, <8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1F16_VP = call <vscale x 1 x bfloat> @llvm.vp.fmul.nxv1bf16(<vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef, <vscale x 1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2F16_VP = call <vscale x 2 x bfloat> @llvm.vp.fmul.nxv2bf16(<vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef, <vscale x 2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4F16_VP = call <vscale x 4 x bfloat> @llvm.vp.fmul.nxv4bf16(<vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef, <vscale x 4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8F16_VP = call <vscale x 8 x bfloat> @llvm.vp.fmul.nxv8bf16(<vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef, <vscale x 8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16F16_VP = call <vscale x 16 x bfloat> @llvm.vp.fmul.nxv16bf16(<vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef, <vscale x 16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.fmul.nxv1f32(<vscale x 1 x float> undef, <vscale x 1 x float> undef, <vscale x 1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.fmul.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> undef, <vscale x 2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.fmul.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x float> undef, <vscale x 4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.fmul.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x float> undef, <vscale x 8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.fmul.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x float> undef, <vscale x 16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.fmul.nxv1f64(<vscale x 1 x double> undef, <vscale x 1 x double> undef, <vscale x 1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.fmul.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x double> undef, <vscale x 2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.fmul.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x double> undef, <vscale x 4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.fmul.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x double> undef, <vscale x 8 x i1> undef, i32 undef)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = fmul float poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = fmul double poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32 = fmul <1 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32 = fmul <2 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fmul <4 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fmul <8 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fmul <16 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F32 = fmul <vscale x 1 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F32 = fmul <vscale x 2 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F32 = fmul <vscale x 4 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F32 = fmul <vscale x 8 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16F32 = fmul <vscale x 16 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64 = fmul <1 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fmul <2 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fmul <4 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fmul <8 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F64 = fmul <vscale x 1 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2F64 = fmul <vscale x 2 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4F64 = fmul <vscale x 4 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8F64 = fmul <vscale x 8 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32_VP = call <1 x float> @llvm.vp.fmul.v1f32(<1 x float> poison, <1 x float> poison, <1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32_VP = call <2 x float> @llvm.vp.fmul.v2f32(<2 x float> poison, <2 x float> poison, <2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32_VP = call <4 x float> @llvm.vp.fmul.v4f32(<4 x float> poison, <4 x float> poison, <4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32_VP = call <8 x float> @llvm.vp.fmul.v8f32(<8 x float> poison, <8 x float> poison, <8 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32_VP = call <16 x float> @llvm.vp.fmul.v16f32(<16 x float> poison, <16 x float> poison, <16 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64_VP = call <1 x double> @llvm.vp.fmul.v1f64(<1 x double> poison, <1 x double> poison, <1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64_VP = call <2 x double> @llvm.vp.fmul.v2f64(<2 x double> poison, <2 x double> poison, <2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64_VP = call <4 x double> @llvm.vp.fmul.v4f64(<4 x double> poison, <4 x double> poison, <4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64_VP = call <8 x double> @llvm.vp.fmul.v8f64(<8 x double> poison, <8 x double> poison, <8 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.fmul.nxv1f32(<vscale x 1 x float> poison, <vscale x 1 x float> poison, <vscale x 1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.fmul.nxv2f32(<vscale x 2 x float> poison, <vscale x 2 x float> poison, <vscale x 2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.fmul.nxv4f32(<vscale x 4 x float> poison, <vscale x 4 x float> poison, <vscale x 4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.fmul.nxv8f32(<vscale x 8 x float> poison, <vscale x 8 x float> poison, <vscale x 8 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.fmul.nxv16f32(<vscale x 16 x float> poison, <vscale x 16 x float> poison, <vscale x 16 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.fmul.nxv1f64(<vscale x 1 x double> poison, <vscale x 1 x double> poison, <vscale x 1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.fmul.nxv2f64(<vscale x 2 x double> poison, <vscale x 2 x double> poison, <vscale x 2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.fmul.nxv4f64(<vscale x 4 x double> poison, <vscale x 4 x double> poison, <vscale x 4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.fmul.nxv8f64(<vscale x 8 x double> poison, <vscale x 8 x double> poison, <vscale x 8 x i1> poison, i32 poison)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
-  %BF16 = fmul bfloat undef, undef
-  %F32 = fmul float undef, undef
-  %F64 = fmul double undef, undef
-
-  %V1BF16 = fmul <1 x bfloat> undef, undef
-  %V2BF16 = fmul <2 x bfloat> undef, undef
-  %V4BF16 = fmul <4 x bfloat> undef, undef
-  %V8BF16 = fmul <8 x bfloat> undef, undef
-  %V16BF16 = fmul <16 x bfloat> undef, undef
-
-  %NXV1BF16 = fmul <vscale x 1 x bfloat> undef, undef
-  %NXV2BF16 = fmul <vscale x 2 x bfloat> undef, undef
-  %NXV4BF16 = fmul <vscale x 4 x bfloat> undef, undef
-  %NXV8BF16 = fmul <vscale x 8 x bfloat> undef, undef
-  %NXV16BF16 = fmul <vscale x 16 x bfloat> undef, undef
-
-  %V1F32 = fmul <1 x float> undef, undef
-  %V2F32 = fmul <2 x float> undef, undef
-  %V4F32 = fmul <4 x float> undef, undef
-  %V8F32 = fmul <8 x float> undef, undef
-  %V16F32 = fmul <16 x float> undef, undef
-
-  %NXV1F32 = fmul <vscale x 1 x float> undef, undef
-  %NXV2F32 = fmul <vscale x 2 x float> undef, undef
-  %NXV4F32 = fmul <vscale x 4 x float> undef, undef
-  %NXV8F32 = fmul <vscale x 8 x float> undef, undef
-  %NXV16F32 = fmul <vscale x 16 x float> undef, undef
-
-  %V1F64 = fmul <1 x double> undef, undef
-  %V2F64 = fmul <2 x double> undef, undef
-  %V4F64 = fmul <4 x double> undef, undef
-  %V8F64 = fmul <8 x double> undef, undef
-
-  %NXV1F64 = fmul <vscale x 1 x double> undef, undef
-  %NXV2F64 = fmul <vscale x 2 x double> undef, undef
-  %NXV4F64 = fmul <vscale x 4 x double> undef, undef
-  %NXV8F64 = fmul <vscale x 8 x double> undef, undef
-
-  %V1F16_VP = call <1 x bfloat> @llvm.vp.fmul.v1f16(<1 x bfloat> undef, <1 x bfloat> undef, <1 x i1> undef, i32 undef)
-  %V2F16_VP = call <2 x bfloat> @llvm.vp.fmul.v2f16(<2 x bfloat> undef, <2 x bfloat> undef, <2 x i1> undef, i32 undef)
-  %V4F16_VP = call <4 x bfloat> @llvm.vp.fmul.v4f16(<4 x bfloat> undef, <4 x bfloat> undef, <4 x i1> undef, i32 undef)
-  %V8F16_VP = call <8 x bfloat> @llvm.vp.fmul.v8f16(<8 x bfloat> undef, <8 x bfloat> undef, <8 x i1> undef, i32 undef)
-  %V16F16_VP = call <16 x bfloat> @llvm.vp.fmul.v16f16(<16 x bfloat> undef, <16 x bfloat> undef, <16 x i1> undef, i32 undef)
-
-  %V1F32_VP = call <1 x float> @llvm.vp.fmul.v1f32(<1 x float> undef, <1 x float> undef, <1 x i1> undef, i32 undef)
-  %V2F32_VP = call <2 x float> @llvm.vp.fmul.v2f32(<2 x float> undef, <2 x float> undef, <2 x i1> undef, i32 undef)
-  %V4F32_VP = call <4 x float> @llvm.vp.fmul.v4f32(<4 x float> undef, <4 x float> undef, <4 x i1> undef, i32 undef)
-  %V8F32_VP = call <8 x float> @llvm.vp.fmul.v8f32(<8 x float> undef, <8 x float> undef, <8 x i1> undef, i32 undef)
-  %V16F32_VP = call <16 x float> @llvm.vp.fmul.v16f32(<16 x float> undef, <16 x float> undef, <16 x i1> undef, i32 undef)
-
-  %V1F64_VP = call <1 x double> @llvm.vp.fmul.v1f64(<1 x double> undef, <1 x double> undef, <1 x i1> undef, i32 undef)
-  %V2F64_VP = call <2 x double> @llvm.vp.fmul.v2f64(<2 x double> undef, <2 x double> undef, <2 x i1> undef, i32 undef)
-  %V4F64_VP = call <4 x double> @llvm.vp.fmul.v4f64(<4 x double> undef, <4 x double> undef, <4 x i1> undef, i32 undef)
-  %V8F64_VP = call <8 x double> @llvm.vp.fmul.v8f64(<8 x double> undef, <8 x double> undef, <8 x i1> undef, i32 undef)
-
-  %NXV1F16_VP = call <vscale x 1 x bfloat> @llvm.vp.fmul.nxv1f16(<vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2F16_VP = call <vscale x 2 x bfloat> @llvm.vp.fmul.nxv2f16(<vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4F16_VP = call <vscale x 4 x bfloat> @llvm.vp.fmul.nxv4f16(<vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8F16_VP = call <vscale x 8 x bfloat> @llvm.vp.fmul.nxv8f16(<vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef, <vscale x 8 x i1> undef, i32 undef)
-  %NXV16F16_VP = call <vscale x 16 x bfloat> @llvm.vp.fmul.nxv16f16(<vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef, <vscale x 16 x i1> undef, i32 undef)
-
-  %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.fmul.nxv1f32(<vscale x 1 x float> undef, <vscale x 1 x float> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.fmul.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.fmul.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x float> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.fmul.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x float> undef, <vscale x 8 x i1> undef, i32 undef)
-  %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.fmul.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x float> undef, <vscale x 16 x i1> undef, i32 undef)
-
-  %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.fmul.nxv1f64(<vscale x 1 x double> undef, <vscale x 1 x double> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.fmul.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x double> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.fmul.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x double> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.fmul.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x double> undef, <vscale x 8 x i1> undef, i32 undef)
+  %F32 = fmul float poison, poison
+  %F64 = fmul double poison, poison
+
+  %V1F32 = fmul <1 x float> poison, poison
+  %V2F32 = fmul <2 x float> poison, poison
+  %V4F32 = fmul <4 x float> poison, poison
+  %V8F32 = fmul <8 x float> poison, poison
+  %V16F32 = fmul <16 x float> poison, poison
+
+  %NXV1F32 = fmul <vscale x 1 x float> poison, poison
+  %NXV2F32 = fmul <vscale x 2 x float> poison, poison
+  %NXV4F32 = fmul <vscale x 4 x float> poison, poison
+  %NXV8F32 = fmul <vscale x 8 x float> poison, poison
+  %NXV16F32 = fmul <vscale x 16 x float> poison, poison
+
+  %V1F64 = fmul <1 x double> poison, poison
+  %V2F64 = fmul <2 x double> poison, poison
+  %V4F64 = fmul <4 x double> poison, poison
+  %V8F64 = fmul <8 x double> poison, poison
+
+  %NXV1F64 = fmul <vscale x 1 x double> poison, poison
+  %NXV2F64 = fmul <vscale x 2 x double> poison, poison
+  %NXV4F64 = fmul <vscale x 4 x double> poison, poison
+  %NXV8F64 = fmul <vscale x 8 x double> poison, poison
+
+  %V1F32_VP = call <1 x float> @llvm.vp.fmul(<1 x float> poison, <1 x float> poison, <1 x i1> poison, i32 poison)
+  %V2F32_VP = call <2 x float> @llvm.vp.fmul(<2 x float> poison, <2 x float> poison, <2 x i1> poison, i32 poison)
+  %V4F32_VP = call <4 x float> @llvm.vp.fmul(<4 x float> poison, <4 x float> poison, <4 x i1> poison, i32 poison)
+  %V8F32_VP = call <8 x float> @llvm.vp.fmul(<8 x float> poison, <8 x float> poison, <8 x i1> poison, i32 poison)
+  %V16F32_VP = call <16 x float> @llvm.vp.fmul(<16 x float> poison, <16 x float> poison, <16 x i1> poison, i32 poison)
+
+  %V1F64_VP = call <1 x double> @llvm.vp.fmul(<1 x double> poison, <1 x double> poison, <1 x i1> poison, i32 poison)
+  %V2F64_VP = call <2 x double> @llvm.vp.fmul(<2 x double> poison, <2 x double> poison, <2 x i1> poison, i32 poison)
+  %V4F64_VP = call <4 x double> @llvm.vp.fmul(<4 x double> poison, <4 x double> poison, <4 x i1> poison, i32 poison)
+  %V8F64_VP = call <8 x double> @llvm.vp.fmul(<8 x double> poison, <8 x double> poison, <8 x i1> poison, i32 poison)
+
+  %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.fmul(<vscale x 1 x float> poison, <vscale x 1 x float> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.fmul(<vscale x 2 x float> poison, <vscale x 2 x float> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.fmul(<vscale x 4 x float> poison, <vscale x 4 x float> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.fmul(<vscale x 8 x float> poison, <vscale x 8 x float> poison, <vscale x 8 x i1> poison, i32 poison)
+  %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.fmul(<vscale x 16 x float> poison, <vscale x 16 x float> poison, <vscale x 16 x i1> poison, i32 poison)
+
+  %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.fmul(<vscale x 1 x double> poison, <vscale x 1 x double> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.fmul(<vscale x 2 x double> poison, <vscale x 2 x double> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.fmul(<vscale x 4 x double> poison, <vscale x 4 x double> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.fmul(<vscale x 8 x double> poison, <vscale x 8 x double> poison, <vscale x 8 x i1> poison, i32 poison)
+
+  ret void
+}
+
+define void @fmul_bf16() {
+; ZVFH-LABEL: 'fmul_bf16'
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = fmul bfloat poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1BF16 = fmul <1 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2BF16 = fmul <2 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4BF16 = fmul <4 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8BF16 = fmul <8 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16BF16 = fmul <16 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32BF16 = fmul <32 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1BF16 = fmul <vscale x 1 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2BF16 = fmul <vscale x 2 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4BF16 = fmul <vscale x 4 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8BF16 = fmul <vscale x 8 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16BF16 = fmul <vscale x 16 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32BF16 = fmul <vscale x 32 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1BF16_VP = call <1 x bfloat> @llvm.vp.fmul.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2BF16_VP = call <2 x bfloat> @llvm.vp.fmul.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4BF16_VP = call <4 x bfloat> @llvm.vp.fmul.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8BF16_VP = call <8 x bfloat> @llvm.vp.fmul.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16BF16_VP = call <16 x bfloat> @llvm.vp.fmul.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fmul.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fmul.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fmul.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fmul.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fmul.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; ZVFHMIN-LABEL: 'fmul_bf16'
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = fmul bfloat poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1BF16 = fmul <1 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2BF16 = fmul <2 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4BF16 = fmul <4 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8BF16 = fmul <8 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16BF16 = fmul <16 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32BF16 = fmul <32 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1BF16 = fmul <vscale x 1 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2BF16 = fmul <vscale x 2 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4BF16 = fmul <vscale x 4 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8BF16 = fmul <vscale x 8 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16BF16 = fmul <vscale x 16 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32BF16 = fmul <vscale x 32 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1BF16_VP = call <1 x bfloat> @llvm.vp.fmul.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2BF16_VP = call <2 x bfloat> @llvm.vp.fmul.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4BF16_VP = call <4 x bfloat> @llvm.vp.fmul.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8BF16_VP = call <8 x bfloat> @llvm.vp.fmul.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16BF16_VP = call <16 x bfloat> @llvm.vp.fmul.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fmul.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fmul.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fmul.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fmul.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fmul.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; NO-ZFHMIN-LABEL: 'fmul_bf16'
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = fmul bfloat poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1BF16 = fmul <1 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2BF16 = fmul <2 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4BF16 = fmul <4 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8BF16 = fmul <8 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16BF16 = fmul <16 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V32BF16 = fmul <32 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16 = fmul <vscale x 1 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16 = fmul <vscale x 2 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16 = fmul <vscale x 4 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16 = fmul <vscale x 8 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16 = fmul <vscale x 16 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32BF16 = fmul <vscale x 32 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1BF16_VP = call <1 x bfloat> @llvm.vp.fmul.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2BF16_VP = call <2 x bfloat> @llvm.vp.fmul.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4BF16_VP = call <4 x bfloat> @llvm.vp.fmul.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8BF16_VP = call <8 x bfloat> @llvm.vp.fmul.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16BF16_VP = call <16 x bfloat> @llvm.vp.fmul.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fmul.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fmul.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fmul.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fmul.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fmul.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+  %BF16 = fmul bfloat poison, poison
+
+  %V1BF16 = fmul <1 x bfloat> poison, poison
+  %V2BF16 = fmul <2 x bfloat> poison, poison
+  %V4BF16 = fmul <4 x bfloat> poison, poison
+  %V8BF16 = fmul <8 x bfloat> poison, poison
+  %V16BF16 = fmul <16 x bfloat> poison, poison
+  %V32BF16 = fmul <32 x bfloat> poison, poison
+
+  %NXV1BF16 = fmul <vscale x 1 x bfloat> poison, poison
+  %NXV2BF16 = fmul <vscale x 2 x bfloat> poison, poison
+  %NXV4BF16 = fmul <vscale x 4 x bfloat> poison, poison
+  %NXV8BF16 = fmul <vscale x 8 x bfloat> poison, poison
+  %NXV16BF16 = fmul <vscale x 16 x bfloat> poison, poison
+  %NXV32BF16 = fmul <vscale x 32 x bfloat> poison, poison
+
+  %V1BF16_VP = call <1 x bfloat> @llvm.vp.fmul(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+  %V2BF16_VP = call <2 x bfloat> @llvm.vp.fmul(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+  %V4BF16_VP = call <4 x bfloat> @llvm.vp.fmul(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+  %V8BF16_VP = call <8 x bfloat> @llvm.vp.fmul(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+  %V16BF16_VP = call <16 x bfloat> @llvm.vp.fmul(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+
+  %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fmul(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fmul(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fmul(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fmul(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+  %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fmul(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
 
   ret void
 }
 
 define void @fmul_f16() {
 ; ZVFH-LABEL: 'fmul_f16'
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fmul half undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16 = fmul <1 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16 = fmul <2 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16 = fmul <4 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16 = fmul <8 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16 = fmul <16 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32F16 = fmul <32 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16 = fmul <vscale x 1 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16 = fmul <vscale x 2 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16 = fmul <vscale x 4 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16 = fmul <vscale x 8 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16 = fmul <vscale x 16 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32F16 = fmul <vscale x 32 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fmul.v1f16(<1 x half> undef, <1 x half> undef, <1 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fmul.v2f16(<2 x half> undef, <2 x half> undef, <2 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fmul.v4f16(<4 x half> undef, <4 x half> undef, <4 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fmul.v8f16(<8 x half> undef, <8 x half> undef, <8 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fmul.v16f16(<16 x half> undef, <16 x half> undef, <16 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fmul.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef, <vscale x 1 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fmul.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef, <vscale x 2 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fmul.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef, <vscale x 4 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fmul.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef, <vscale x 8 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fmul.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef, <vscale x 16 x i1> undef, i32 undef)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fmul half poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16 = fmul <1 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16 = fmul <2 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16 = fmul <4 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16 = fmul <8 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16 = fmul <16 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32F16 = fmul <32 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16 = fmul <vscale x 1 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16 = fmul <vscale x 2 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16 = fmul <vscale x 4 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16 = fmul <vscale x 8 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16 = fmul <vscale x 16 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32F16 = fmul <vscale x 32 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fmul.v1f16(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fmul.v2f16(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fmul.v4f16(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fmul.v8f16(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fmul.v16f16(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fmul.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fmul.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fmul.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fmul.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fmul.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
 ; ZVFH-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; ZVFHMIN-LABEL: 'fmul_f16'
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fmul half undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1F16 = fmul <1 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2F16 = fmul <2 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4F16 = fmul <4 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8F16 = fmul <8 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16F16 = fmul <16 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32F16 = fmul <32 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1F16 = fmul <vscale x 1 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2F16 = fmul <vscale x 2 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4F16 = fmul <vscale x 4 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8F16 = fmul <vscale x 8 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16F16 = fmul <vscale x 16 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32F16 = fmul <vscale x 32 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fmul.v1f16(<1 x half> undef, <1 x half> undef, <1 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fmul.v2f16(<2 x half> undef, <2 x half> undef, <2 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fmul.v4f16(<4 x half> undef, <4 x half> undef, <4 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fmul.v8f16(<8 x half> undef, <8 x half> undef, <8 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fmul.v16f16(<16 x half> undef, <16 x half> undef, <16 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fmul.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef, <vscale x 1 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fmul.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef, <vscale x 2 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fmul.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef, <vscale x 4 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fmul.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef, <vscale x 8 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fmul.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef, <vscale x 16 x i1> undef, i32 undef)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fmul half poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1F16 = fmul <1 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2F16 = fmul <2 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4F16 = fmul <4 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8F16 = fmul <8 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16F16 = fmul <16 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32F16 = fmul <32 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1F16 = fmul <vscale x 1 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2F16 = fmul <vscale x 2 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4F16 = fmul <vscale x 4 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8F16 = fmul <vscale x 8 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16F16 = fmul <vscale x 16 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32F16 = fmul <vscale x 32 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fmul.v1f16(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fmul.v2f16(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fmul.v4f16(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fmul.v8f16(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fmul.v16f16(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fmul.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fmul.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fmul.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fmul.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fmul.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
 ; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
-  %F16 = fmul half undef, undef
-
-  %V1F16 = fmul <1 x half> undef, undef
-  %V2F16 = fmul <2 x half> undef, undef
-  %V4F16 = fmul <4 x half> undef, undef
-  %V8F16 = fmul <8 x half> undef, undef
-  %V16F16 = fmul <16 x half> undef, undef
-  %V32F16 = fmul <32 x half> undef, undef
-
-  %NXV1F16 = fmul <vscale x 1 x half> undef, undef
-  %NXV2F16 = fmul <vscale x 2 x half> undef, undef
-  %NXV4F16 = fmul <vscale x 4 x half> undef, undef
-  %NXV8F16 = fmul <vscale x 8 x half> undef, undef
-  %NXV16F16 = fmul <vscale x 16 x half> undef, undef
-  %NXV32F16 = fmul <vscale x 32 x half> undef, undef
-
-  %V1F16_VP = call <1 x half> @llvm.vp.fmul.v1f16(<1 x half> undef, <1 x half> undef, <1 x i1> undef, i32 undef)
-  %V2F16_VP = call <2 x half> @llvm.vp.fmul.v2f16(<2 x half> undef, <2 x half> undef, <2 x i1> undef, i32 undef)
-  %V4F16_VP = call <4 x half> @llvm.vp.fmul.v4f16(<4 x half> undef, <4 x half> undef, <4 x i1> undef, i32 undef)
-  %V8F16_VP = call <8 x half> @llvm.vp.fmul.v8f16(<8 x half> undef, <8 x half> undef, <8 x i1> undef, i32 undef)
-  %V16F16_VP = call <16 x half> @llvm.vp.fmul.v16f16(<16 x half> undef, <16 x half> undef, <16 x i1> undef, i32 undef)
-
-  %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fmul.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fmul.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fmul.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fmul.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef, <vscale x 8 x i1> undef, i32 undef)
-  %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fmul.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef, <vscale x 16 x i1> undef, i32 undef)
+; NO-ZFHMIN-LABEL: 'fmul_f16'
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fmul half poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16 = fmul <1 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2F16 = fmul <2 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4F16 = fmul <4 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8F16 = fmul <8 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16F16 = fmul <16 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V32F16 = fmul <32 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16 = fmul <vscale x 1 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16 = fmul <vscale x 2 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16 = fmul <vscale x 4 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16 = fmul <vscale x 8 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16 = fmul <vscale x 16 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32F16 = fmul <vscale x 32 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fmul.v1f16(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fmul.v2f16(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fmul.v4f16(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fmul.v8f16(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fmul.v16f16(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fmul.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fmul.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fmul.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fmul.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fmul.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+  %F16 = fmul half poison, poison
+
+  %V1F16 = fmul <1 x half> poison, poison
+  %V2F16 = fmul <2 x half> poison, poison
+  %V4F16 = fmul <4 x half> poison, poison
+  %V8F16 = fmul <8 x half> poison, poison
+  %V16F16 = fmul <16 x half> poison, poison
+  %V32F16 = fmul <32 x half> poison, poison
+
+  %NXV1F16 = fmul <vscale x 1 x half> poison, poison
+  %NXV2F16 = fmul <vscale x 2 x half> poison, poison
+  %NXV4F16 = fmul <vscale x 4 x half> poison, poison
+  %NXV8F16 = fmul <vscale x 8 x half> poison, poison
+  %NXV16F16 = fmul <vscale x 16 x half> poison, poison
+  %NXV32F16 = fmul <vscale x 32 x half> poison, poison
+
+  %V1F16_VP = call <1 x half> @llvm.vp.fmul(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+  %V2F16_VP = call <2 x half> @llvm.vp.fmul(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+  %V4F16_VP = call <4 x half> @llvm.vp.fmul(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+  %V8F16_VP = call <8 x half> @llvm.vp.fmul(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+  %V16F16_VP = call <16 x half> @llvm.vp.fmul(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+
+  %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fmul(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fmul(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fmul(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fmul(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+  %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fmul(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
 
   ret void
 }
 
 define void @fdiv() {
 ; CHECK-LABEL: 'fdiv'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = fdiv bfloat undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = fdiv float undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = fdiv double undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1BF16 = fdiv <1 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2BF16 = fdiv <2 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4BF16 = fdiv <4 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8BF16 = fdiv <8 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16BF16 = fdiv <16 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1BF16 = fdiv <vscale x 1 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2BF16 = fdiv <vscale x 2 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4BF16 = fdiv <vscale x 4 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8BF16 = fdiv <vscale x 8 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16BF16 = fdiv <vscale x 16 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32 = fdiv <1 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32 = fdiv <2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fdiv <4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fdiv <8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fdiv <16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F32 = fdiv <vscale x 1 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F32 = fdiv <vscale x 2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F32 = fdiv <vscale x 4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F32 = fdiv <vscale x 8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16F32 = fdiv <vscale x 16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64 = fdiv <1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fdiv <2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fdiv <4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fdiv <8 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F64 = fdiv <vscale x 1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2F64 = fdiv <vscale x 2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4F64 = fdiv <vscale x 4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8F64 = fdiv <vscale x 8 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1F16_VP = call <1 x bfloat> @llvm.vp.fdiv.v1bf16(<1 x bfloat> undef, <1 x bfloat> undef, <1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2F16_VP = call <2 x bfloat> @llvm.vp.fdiv.v2bf16(<2 x bfloat> undef, <2 x bfloat> undef, <2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4F16_VP = call <4 x bfloat> @llvm.vp.fdiv.v4bf16(<4 x bfloat> undef, <4 x bfloat> undef, <4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8F16_VP = call <8 x bfloat> @llvm.vp.fdiv.v8bf16(<8 x bfloat> undef, <8 x bfloat> undef, <8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16F16_VP = call <16 x bfloat> @llvm.vp.fdiv.v16bf16(<16 x bfloat> undef, <16 x bfloat> undef, <16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32_VP = call <1 x float> @llvm.vp.fdiv.v1f32(<1 x float> undef, <1 x float> undef, <1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32_VP = call <2 x float> @llvm.vp.fdiv.v2f32(<2 x float> undef, <2 x float> undef, <2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32_VP = call <4 x float> @llvm.vp.fdiv.v4f32(<4 x float> undef, <4 x float> undef, <4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32_VP = call <8 x float> @llvm.vp.fdiv.v8f32(<8 x float> undef, <8 x float> undef, <8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32_VP = call <16 x float> @llvm.vp.fdiv.v16f32(<16 x float> undef, <16 x float> undef, <16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64_VP = call <1 x double> @llvm.vp.fdiv.v1f64(<1 x double> undef, <1 x double> undef, <1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64_VP = call <2 x double> @llvm.vp.fdiv.v2f64(<2 x double> undef, <2 x double> undef, <2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64_VP = call <4 x double> @llvm.vp.fdiv.v4f64(<4 x double> undef, <4 x double> undef, <4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64_VP = call <8 x double> @llvm.vp.fdiv.v8f64(<8 x double> undef, <8 x double> undef, <8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1F16_VP = call <vscale x 1 x bfloat> @llvm.vp.fdiv.nxv1bf16(<vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef, <vscale x 1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2F16_VP = call <vscale x 2 x bfloat> @llvm.vp.fdiv.nxv2bf16(<vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef, <vscale x 2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4F16_VP = call <vscale x 4 x bfloat> @llvm.vp.fdiv.nxv4bf16(<vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef, <vscale x 4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8F16_VP = call <vscale x 8 x bfloat> @llvm.vp.fdiv.nxv8bf16(<vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef, <vscale x 8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16F16_VP = call <vscale x 16 x bfloat> @llvm.vp.fdiv.nxv16bf16(<vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef, <vscale x 16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.fdiv.nxv1f32(<vscale x 1 x float> undef, <vscale x 1 x float> undef, <vscale x 1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.fdiv.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> undef, <vscale x 2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.fdiv.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x float> undef, <vscale x 4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.fdiv.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x float> undef, <vscale x 8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.fdiv.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x float> undef, <vscale x 16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.fdiv.nxv1f64(<vscale x 1 x double> undef, <vscale x 1 x double> undef, <vscale x 1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.fdiv.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x double> undef, <vscale x 2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.fdiv.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x double> undef, <vscale x 4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.fdiv.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x double> undef, <vscale x 8 x i1> undef, i32 undef)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = fdiv float poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = fdiv double poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32 = fdiv <1 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32 = fdiv <2 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fdiv <4 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fdiv <8 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fdiv <16 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F32 = fdiv <vscale x 1 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F32 = fdiv <vscale x 2 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F32 = fdiv <vscale x 4 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F32 = fdiv <vscale x 8 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16F32 = fdiv <vscale x 16 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64 = fdiv <1 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fdiv <2 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fdiv <4 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fdiv <8 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F64 = fdiv <vscale x 1 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2F64 = fdiv <vscale x 2 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4F64 = fdiv <vscale x 4 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8F64 = fdiv <vscale x 8 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32_VP = call <1 x float> @llvm.vp.fdiv.v1f32(<1 x float> poison, <1 x float> poison, <1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32_VP = call <2 x float> @llvm.vp.fdiv.v2f32(<2 x float> poison, <2 x float> poison, <2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32_VP = call <4 x float> @llvm.vp.fdiv.v4f32(<4 x float> poison, <4 x float> poison, <4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32_VP = call <8 x float> @llvm.vp.fdiv.v8f32(<8 x float> poison, <8 x float> poison, <8 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32_VP = call <16 x float> @llvm.vp.fdiv.v16f32(<16 x float> poison, <16 x float> poison, <16 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64_VP = call <1 x double> @llvm.vp.fdiv.v1f64(<1 x double> poison, <1 x double> poison, <1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64_VP = call <2 x double> @llvm.vp.fdiv.v2f64(<2 x double> poison, <2 x double> poison, <2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64_VP = call <4 x double> @llvm.vp.fdiv.v4f64(<4 x double> poison, <4 x double> poison, <4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64_VP = call <8 x double> @llvm.vp.fdiv.v8f64(<8 x double> poison, <8 x double> poison, <8 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.fdiv.nxv1f32(<vscale x 1 x float> poison, <vscale x 1 x float> poison, <vscale x 1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.fdiv.nxv2f32(<vscale x 2 x float> poison, <vscale x 2 x float> poison, <vscale x 2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.fdiv.nxv4f32(<vscale x 4 x float> poison, <vscale x 4 x float> poison, <vscale x 4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.fdiv.nxv8f32(<vscale x 8 x float> poison, <vscale x 8 x float> poison, <vscale x 8 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.fdiv.nxv16f32(<vscale x 16 x float> poison, <vscale x 16 x float> poison, <vscale x 16 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.fdiv.nxv1f64(<vscale x 1 x double> poison, <vscale x 1 x double> poison, <vscale x 1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.fdiv.nxv2f64(<vscale x 2 x double> poison, <vscale x 2 x double> poison, <vscale x 2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.fdiv.nxv4f64(<vscale x 4 x double> poison, <vscale x 4 x double> poison, <vscale x 4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.fdiv.nxv8f64(<vscale x 8 x double> poison, <vscale x 8 x double> poison, <vscale x 8 x i1> poison, i32 poison)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
-  %BF16 = fdiv bfloat undef, undef
-  %F32 = fdiv float undef, undef
-  %F64 = fdiv double undef, undef
-
-  %V1BF16 = fdiv <1 x bfloat> undef, undef
-  %V2BF16 = fdiv <2 x bfloat> undef, undef
-  %V4BF16 = fdiv <4 x bfloat> undef, undef
-  %V8BF16 = fdiv <8 x bfloat> undef, undef
-  %V16BF16 = fdiv <16 x bfloat> undef, undef
-
-  %NXV1BF16 = fdiv <vscale x 1 x bfloat> undef, undef
-  %NXV2BF16 = fdiv <vscale x 2 x bfloat> undef, undef
-  %NXV4BF16 = fdiv <vscale x 4 x bfloat> undef, undef
-  %NXV8BF16 = fdiv <vscale x 8 x bfloat> undef, undef
-  %NXV16BF16 = fdiv <vscale x 16 x bfloat> undef, undef
-
-  %V1F32 = fdiv <1 x float> undef, undef
-  %V2F32 = fdiv <2 x float> undef, undef
-  %V4F32 = fdiv <4 x float> undef, undef
-  %V8F32 = fdiv <8 x float> undef, undef
-  %V16F32 = fdiv <16 x float> undef, undef
-
-  %NXV1F32 = fdiv <vscale x 1 x float> undef, undef
-  %NXV2F32 = fdiv <vscale x 2 x float> undef, undef
-  %NXV4F32 = fdiv <vscale x 4 x float> undef, undef
-  %NXV8F32 = fdiv <vscale x 8 x float> undef, undef
-  %NXV16F32 = fdiv <vscale x 16 x float> undef, undef
-
-  %V1F64 = fdiv <1 x double> undef, undef
-  %V2F64 = fdiv <2 x double> undef, undef
-  %V4F64 = fdiv <4 x double> undef, undef
-  %V8F64 = fdiv <8 x double> undef, undef
-
-  %NXV1F64 = fdiv <vscale x 1 x double> undef, undef
-  %NXV2F64 = fdiv <vscale x 2 x double> undef, undef
-  %NXV4F64 = fdiv <vscale x 4 x double> undef, undef
-  %NXV8F64 = fdiv <vscale x 8 x double> undef, undef
-
-  %V1F16_VP = call <1 x bfloat> @llvm.vp.fdiv.v1f16(<1 x bfloat> undef, <1 x bfloat> undef, <1 x i1> undef, i32 undef)
-  %V2F16_VP = call <2 x bfloat> @llvm.vp.fdiv.v2f16(<2 x bfloat> undef, <2 x bfloat> undef, <2 x i1> undef, i32 undef)
-  %V4F16_VP = call <4 x bfloat> @llvm.vp.fdiv.v4f16(<4 x bfloat> undef, <4 x bfloat> undef, <4 x i1> undef, i32 undef)
-  %V8F16_VP = call <8 x bfloat> @llvm.vp.fdiv.v8f16(<8 x bfloat> undef, <8 x bfloat> undef, <8 x i1> undef, i32 undef)
-  %V16F16_VP = call <16 x bfloat> @llvm.vp.fdiv.v16f16(<16 x bfloat> undef, <16 x bfloat> undef, <16 x i1> undef, i32 undef)
-
-  %V1F32_VP = call <1 x float> @llvm.vp.fdiv.v1f32(<1 x float> undef, <1 x float> undef, <1 x i1> undef, i32 undef)
-  %V2F32_VP = call <2 x float> @llvm.vp.fdiv.v2f32(<2 x float> undef, <2 x float> undef, <2 x i1> undef, i32 undef)
-  %V4F32_VP = call <4 x float> @llvm.vp.fdiv.v4f32(<4 x float> undef, <4 x float> undef, <4 x i1> undef, i32 undef)
-  %V8F32_VP = call <8 x float> @llvm.vp.fdiv.v8f32(<8 x float> undef, <8 x float> undef, <8 x i1> undef, i32 undef)
-  %V16F32_VP = call <16 x float> @llvm.vp.fdiv.v16f32(<16 x float> undef, <16 x float> undef, <16 x i1> undef, i32 undef)
-
-  %V1F64_VP = call <1 x double> @llvm.vp.fdiv.v1f64(<1 x double> undef, <1 x double> undef, <1 x i1> undef, i32 undef)
-  %V2F64_VP = call <2 x double> @llvm.vp.fdiv.v2f64(<2 x double> undef, <2 x double> undef, <2 x i1> undef, i32 undef)
-  %V4F64_VP = call <4 x double> @llvm.vp.fdiv.v4f64(<4 x double> undef, <4 x double> undef, <4 x i1> undef, i32 undef)
-  %V8F64_VP = call <8 x double> @llvm.vp.fdiv.v8f64(<8 x double> undef, <8 x double> undef, <8 x i1> undef, i32 undef)
-
-  %NXV1F16_VP = call <vscale x 1 x bfloat> @llvm.vp.fdiv.nxv1f16(<vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2F16_VP = call <vscale x 2 x bfloat> @llvm.vp.fdiv.nxv2f16(<vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4F16_VP = call <vscale x 4 x bfloat> @llvm.vp.fdiv.nxv4f16(<vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8F16_VP = call <vscale x 8 x bfloat> @llvm.vp.fdiv.nxv8f16(<vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef, <vscale x 8 x i1> undef, i32 undef)
-  %NXV16F16_VP = call <vscale x 16 x bfloat> @llvm.vp.fdiv.nxv16f16(<vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef, <vscale x 16 x i1> undef, i32 undef)
-
-  %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.fdiv.nxv1f32(<vscale x 1 x float> undef, <vscale x 1 x float> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.fdiv.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.fdiv.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x float> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.fdiv.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x float> undef, <vscale x 8 x i1> undef, i32 undef)
-  %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.fdiv.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x float> undef, <vscale x 16 x i1> undef, i32 undef)
-
-  %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.fdiv.nxv1f64(<vscale x 1 x double> undef, <vscale x 1 x double> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.fdiv.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x double> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.fdiv.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x double> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.fdiv.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x double> undef, <vscale x 8 x i1> undef, i32 undef)
+  %F32 = fdiv float poison, poison
+  %F64 = fdiv double poison, poison
+
+  %V1F32 = fdiv <1 x float> poison, poison
+  %V2F32 = fdiv <2 x float> poison, poison
+  %V4F32 = fdiv <4 x float> poison, poison
+  %V8F32 = fdiv <8 x float> poison, poison
+  %V16F32 = fdiv <16 x float> poison, poison
+
+  %NXV1F32 = fdiv <vscale x 1 x float> poison, poison
+  %NXV2F32 = fdiv <vscale x 2 x float> poison, poison
+  %NXV4F32 = fdiv <vscale x 4 x float> poison, poison
+  %NXV8F32 = fdiv <vscale x 8 x float> poison, poison
+  %NXV16F32 = fdiv <vscale x 16 x float> poison, poison
+
+  %V1F64 = fdiv <1 x double> poison, poison
+  %V2F64 = fdiv <2 x double> poison, poison
+  %V4F64 = fdiv <4 x double> poison, poison
+  %V8F64 = fdiv <8 x double> poison, poison
+
+  %NXV1F64 = fdiv <vscale x 1 x double> poison, poison
+  %NXV2F64 = fdiv <vscale x 2 x double> poison, poison
+  %NXV4F64 = fdiv <vscale x 4 x double> poison, poison
+  %NXV8F64 = fdiv <vscale x 8 x double> poison, poison
+
+  %V1F32_VP = call <1 x float> @llvm.vp.fdiv(<1 x float> poison, <1 x float> poison, <1 x i1> poison, i32 poison)
+  %V2F32_VP = call <2 x float> @llvm.vp.fdiv(<2 x float> poison, <2 x float> poison, <2 x i1> poison, i32 poison)
+  %V4F32_VP = call <4 x float> @llvm.vp.fdiv(<4 x float> poison, <4 x float> poison, <4 x i1> poison, i32 poison)
+  %V8F32_VP = call <8 x float> @llvm.vp.fdiv(<8 x float> poison, <8 x float> poison, <8 x i1> poison, i32 poison)
+  %V16F32_VP = call <16 x float> @llvm.vp.fdiv(<16 x float> poison, <16 x float> poison, <16 x i1> poison, i32 poison)
+
+  %V1F64_VP = call <1 x double> @llvm.vp.fdiv(<1 x double> poison, <1 x double> poison, <1 x i1> poison, i32 poison)
+  %V2F64_VP = call <2 x double> @llvm.vp.fdiv(<2 x double> poison, <2 x double> poison, <2 x i1> poison, i32 poison)
+  %V4F64_VP = call <4 x double> @llvm.vp.fdiv(<4 x double> poison, <4 x double> poison, <4 x i1> poison, i32 poison)
+  %V8F64_VP = call <8 x double> @llvm.vp.fdiv(<8 x double> poison, <8 x double> poison, <8 x i1> poison, i32 poison)
+
+
+  %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.fdiv(<vscale x 1 x float> poison, <vscale x 1 x float> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.fdiv(<vscale x 2 x float> poison, <vscale x 2 x float> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.fdiv(<vscale x 4 x float> poison, <vscale x 4 x float> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.fdiv(<vscale x 8 x float> poison, <vscale x 8 x float> poison, <vscale x 8 x i1> poison, i32 poison)
+  %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.fdiv(<vscale x 16 x float> poison, <vscale x 16 x float> poison, <vscale x 16 x i1> poison, i32 poison)
+
+  %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.fdiv(<vscale x 1 x double> poison, <vscale x 1 x double> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.fdiv(<vscale x 2 x double> poison, <vscale x 2 x double> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.fdiv(<vscale x 4 x double> poison, <vscale x 4 x double> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.fdiv(<vscale x 8 x double> poison, <vscale x 8 x double> poison, <vscale x 8 x i1> poison, i32 poison)
+
+  ret void
+}
+
+define void @fdiv_bf16() {
+; ZVFH-LABEL: 'fdiv_bf16'
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = fdiv bfloat poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1BF16 = fdiv <1 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2BF16 = fdiv <2 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4BF16 = fdiv <4 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8BF16 = fdiv <8 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16BF16 = fdiv <16 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32BF16 = fdiv <32 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1BF16 = fdiv <vscale x 1 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2BF16 = fdiv <vscale x 2 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4BF16 = fdiv <vscale x 4 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8BF16 = fdiv <vscale x 8 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16BF16 = fdiv <vscale x 16 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32BF16 = fdiv <vscale x 32 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1BF16_VP = call <1 x bfloat> @llvm.vp.fdiv.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2BF16_VP = call <2 x bfloat> @llvm.vp.fdiv.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4BF16_VP = call <4 x bfloat> @llvm.vp.fdiv.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8BF16_VP = call <8 x bfloat> @llvm.vp.fdiv.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16BF16_VP = call <16 x bfloat> @llvm.vp.fdiv.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fdiv.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fdiv.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fdiv.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fdiv.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fdiv.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; ZVFHMIN-LABEL: 'fdiv_bf16'
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = fdiv bfloat poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1BF16 = fdiv <1 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2BF16 = fdiv <2 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4BF16 = fdiv <4 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8BF16 = fdiv <8 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16BF16 = fdiv <16 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32BF16 = fdiv <32 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1BF16 = fdiv <vscale x 1 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2BF16 = fdiv <vscale x 2 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4BF16 = fdiv <vscale x 4 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8BF16 = fdiv <vscale x 8 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16BF16 = fdiv <vscale x 16 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32BF16 = fdiv <vscale x 32 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1BF16_VP = call <1 x bfloat> @llvm.vp.fdiv.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2BF16_VP = call <2 x bfloat> @llvm.vp.fdiv.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4BF16_VP = call <4 x bfloat> @llvm.vp.fdiv.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8BF16_VP = call <8 x bfloat> @llvm.vp.fdiv.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16BF16_VP = call <16 x bfloat> @llvm.vp.fdiv.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fdiv.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fdiv.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fdiv.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fdiv.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fdiv.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; NO-ZFHMIN-LABEL: 'fdiv_bf16'
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = fdiv bfloat poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1BF16 = fdiv <1 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2BF16 = fdiv <2 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4BF16 = fdiv <4 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8BF16 = fdiv <8 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16BF16 = fdiv <16 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V32BF16 = fdiv <32 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16 = fdiv <vscale x 1 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16 = fdiv <vscale x 2 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16 = fdiv <vscale x 4 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16 = fdiv <vscale x 8 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16 = fdiv <vscale x 16 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32BF16 = fdiv <vscale x 32 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1BF16_VP = call <1 x bfloat> @llvm.vp.fdiv.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2BF16_VP = call <2 x bfloat> @llvm.vp.fdiv.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4BF16_VP = call <4 x bfloat> @llvm.vp.fdiv.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8BF16_VP = call <8 x bfloat> @llvm.vp.fdiv.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16BF16_VP = call <16 x bfloat> @llvm.vp.fdiv.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fdiv.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fdiv.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fdiv.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fdiv.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fdiv.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+  %BF16 = fdiv bfloat poison, poison
+
+  %V1BF16 = fdiv <1 x bfloat> poison, poison
+  %V2BF16 = fdiv <2 x bfloat> poison, poison
+  %V4BF16 = fdiv <4 x bfloat> poison, poison
+  %V8BF16 = fdiv <8 x bfloat> poison, poison
+  %V16BF16 = fdiv <16 x bfloat> poison, poison
+  %V32BF16 = fdiv <32 x bfloat> poison, poison
+
+  %NXV1BF16 = fdiv <vscale x 1 x bfloat> poison, poison
+  %NXV2BF16 = fdiv <vscale x 2 x bfloat> poison, poison
+  %NXV4BF16 = fdiv <vscale x 4 x bfloat> poison, poison
+  %NXV8BF16 = fdiv <vscale x 8 x bfloat> poison, poison
+  %NXV16BF16 = fdiv <vscale x 16 x bfloat> poison, poison
+  %NXV32BF16 = fdiv <vscale x 32 x bfloat> poison, poison
+
+  %V1BF16_VP = call <1 x bfloat> @llvm.vp.fdiv(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+  %V2BF16_VP = call <2 x bfloat> @llvm.vp.fdiv(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+  %V4BF16_VP = call <4 x bfloat> @llvm.vp.fdiv(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+  %V8BF16_VP = call <8 x bfloat> @llvm.vp.fdiv(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+  %V16BF16_VP = call <16 x bfloat> @llvm.vp.fdiv(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+
+  %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fdiv(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fdiv(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fdiv(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fdiv(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+  %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fdiv(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
 
   ret void
 }
 
 define void @fdiv_f16() {
 ; ZVFH-LABEL: 'fdiv_f16'
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fdiv half undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16 = fdiv <1 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16 = fdiv <2 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16 = fdiv <4 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16 = fdiv <8 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16 = fdiv <16 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32F16 = fdiv <32 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16 = fdiv <vscale x 1 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16 = fdiv <vscale x 2 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16 = fdiv <vscale x 4 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16 = fdiv <vscale x 8 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16 = fdiv <vscale x 16 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32F16 = fdiv <vscale x 32 x half> undef, undef
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fdiv.v1f16(<1 x half> undef, <1 x half> undef, <1 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fdiv.v2f16(<2 x half> undef, <2 x half> undef, <2 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fdiv.v4f16(<4 x half> undef, <4 x half> undef, <4 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fdiv.v8f16(<8 x half> undef, <8 x half> undef, <8 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fdiv.v16f16(<16 x half> undef, <16 x half> undef, <16 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fdiv.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef, <vscale x 1 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fdiv.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef, <vscale x 2 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fdiv.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef, <vscale x 4 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fdiv.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef, <vscale x 8 x i1> undef, i32 undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fdiv.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef, <vscale x 16 x i1> undef, i32 undef)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fdiv half poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16 = fdiv <1 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16 = fdiv <2 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16 = fdiv <4 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16 = fdiv <8 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16 = fdiv <16 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32F16 = fdiv <32 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16 = fdiv <vscale x 1 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16 = fdiv <vscale x 2 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16 = fdiv <vscale x 4 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16 = fdiv <vscale x 8 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16 = fdiv <vscale x 16 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32F16 = fdiv <vscale x 32 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fdiv.v1f16(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fdiv.v2f16(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fdiv.v4f16(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fdiv.v8f16(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fdiv.v16f16(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fdiv.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fdiv.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fdiv.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fdiv.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fdiv.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
 ; ZVFH-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; ZVFHMIN-LABEL: 'fdiv_f16'
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fdiv half undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1F16 = fdiv <1 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2F16 = fdiv <2 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4F16 = fdiv <4 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8F16 = fdiv <8 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16F16 = fdiv <16 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32F16 = fdiv <32 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1F16 = fdiv <vscale x 1 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2F16 = fdiv <vscale x 2 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4F16 = fdiv <vscale x 4 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8F16 = fdiv <vscale x 8 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16F16 = fdiv <vscale x 16 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32F16 = fdiv <vscale x 32 x half> undef, undef
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fdiv.v1f16(<1 x half> undef, <1 x half> undef, <1 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fdiv.v2f16(<2 x half> undef, <2 x half> undef, <2 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fdiv.v4f16(<4 x half> undef, <4 x half> undef, <4 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fdiv.v8f16(<8 x half> undef, <8 x half> undef, <8 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fdiv.v16f16(<16 x half> undef, <16 x half> undef, <16 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fdiv.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef, <vscale x 1 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fdiv.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef, <vscale x 2 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fdiv.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef, <vscale x 4 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fdiv.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef, <vscale x 8 x i1> undef, i32 undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fdiv.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef, <vscale x 16 x i1> undef, i32 undef)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fdiv half poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V1F16 = fdiv <1 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V2F16 = fdiv <2 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V4F16 = fdiv <4 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V8F16 = fdiv <8 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %V16F16 = fdiv <16 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %V32F16 = fdiv <32 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV1F16 = fdiv <vscale x 1 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV2F16 = fdiv <vscale x 2 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %NXV4F16 = fdiv <vscale x 4 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 18 for instruction: %NXV8F16 = fdiv <vscale x 8 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 36 for instruction: %NXV16F16 = fdiv <vscale x 16 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32F16 = fdiv <vscale x 32 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fdiv.v1f16(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fdiv.v2f16(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fdiv.v4f16(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fdiv.v8f16(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fdiv.v16f16(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fdiv.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fdiv.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 5 for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fdiv.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 10 for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fdiv.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 20 for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fdiv.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
 ; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
-  %F16 = fdiv half undef, undef
-
-  %V1F16 = fdiv <1 x half> undef, undef
-  %V2F16 = fdiv <2 x half> undef, undef
-  %V4F16 = fdiv <4 x half> undef, undef
-  %V8F16 = fdiv <8 x half> undef, undef
-  %V16F16 = fdiv <16 x half> undef, undef
-  %V32F16 = fdiv <32 x half> undef, undef
-
-  %NXV1F16 = fdiv <vscale x 1 x half> undef, undef
-  %NXV2F16 = fdiv <vscale x 2 x half> undef, undef
-  %NXV4F16 = fdiv <vscale x 4 x half> undef, undef
-  %NXV8F16 = fdiv <vscale x 8 x half> undef, undef
-  %NXV16F16 = fdiv <vscale x 16 x half> undef, undef
-  %NXV32F16 = fdiv <vscale x 32 x half> undef, undef
-
-  %V1F16_VP = call <1 x half> @llvm.vp.fdiv.v1f16(<1 x half> undef, <1 x half> undef, <1 x i1> undef, i32 undef)
-  %V2F16_VP = call <2 x half> @llvm.vp.fdiv.v2f16(<2 x half> undef, <2 x half> undef, <2 x i1> undef, i32 undef)
-  %V4F16_VP = call <4 x half> @llvm.vp.fdiv.v4f16(<4 x half> undef, <4 x half> undef, <4 x i1> undef, i32 undef)
-  %V8F16_VP = call <8 x half> @llvm.vp.fdiv.v8f16(<8 x half> undef, <8 x half> undef, <8 x i1> undef, i32 undef)
-  %V16F16_VP = call <16 x half> @llvm.vp.fdiv.v16f16(<16 x half> undef, <16 x half> undef, <16 x i1> undef, i32 undef)
-
-  %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fdiv.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fdiv.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fdiv.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fdiv.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef, <vscale x 8 x i1> undef, i32 undef)
-  %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fdiv.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef, <vscale x 16 x i1> undef, i32 undef)
+; NO-ZFHMIN-LABEL: 'fdiv_f16'
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fdiv half poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16 = fdiv <1 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2F16 = fdiv <2 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4F16 = fdiv <4 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8F16 = fdiv <8 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16F16 = fdiv <16 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V32F16 = fdiv <32 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16 = fdiv <vscale x 1 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16 = fdiv <vscale x 2 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16 = fdiv <vscale x 4 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16 = fdiv <vscale x 8 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16 = fdiv <vscale x 16 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32F16 = fdiv <vscale x 32 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fdiv.v1f16(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fdiv.v2f16(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fdiv.v4f16(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fdiv.v8f16(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fdiv.v16f16(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fdiv.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fdiv.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fdiv.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fdiv.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fdiv.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+  %F16 = fdiv half poison, poison
+
+  %V1F16 = fdiv <1 x half> poison, poison
+  %V2F16 = fdiv <2 x half> poison, poison
+  %V4F16 = fdiv <4 x half> poison, poison
+  %V8F16 = fdiv <8 x half> poison, poison
+  %V16F16 = fdiv <16 x half> poison, poison
+  %V32F16 = fdiv <32 x half> poison, poison
+
+  %NXV1F16 = fdiv <vscale x 1 x half> poison, poison
+  %NXV2F16 = fdiv <vscale x 2 x half> poison, poison
+  %NXV4F16 = fdiv <vscale x 4 x half> poison, poison
+  %NXV8F16 = fdiv <vscale x 8 x half> poison, poison
+  %NXV16F16 = fdiv <vscale x 16 x half> poison, poison
+  %NXV32F16 = fdiv <vscale x 32 x half> poison, poison
+
+  %V1F16_VP = call <1 x half> @llvm.vp.fdiv(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+  %V2F16_VP = call <2 x half> @llvm.vp.fdiv(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+  %V4F16_VP = call <4 x half> @llvm.vp.fdiv(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+  %V8F16_VP = call <8 x half> @llvm.vp.fdiv(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+  %V16F16_VP = call <16 x half> @llvm.vp.fdiv(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+
+  %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fdiv(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fdiv(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fdiv(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fdiv(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+  %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fdiv(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
 
   ret void
 }
 
 define void @frem() {
 ; CHECK-LABEL: 'frem'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = frem bfloat undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1BF16 = frem <1 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2BF16 = frem <2 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4BF16 = frem <4 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8BF16 = frem <8 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V16BF16 = frem <16 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16 = frem <vscale x 1 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16 = frem <vscale x 2 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16 = frem <vscale x 4 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16 = frem <vscale x 8 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16 = frem <vscale x 16 x bfloat> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1F32 = frem <1 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2F32 = frem <2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4F32 = frem <4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V8F32 = frem <8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V16F32 = frem <16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F32 = frem <vscale x 1 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F32 = frem <vscale x 2 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F32 = frem <vscale x 4 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F32 = frem <vscale x 8 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F32 = frem <vscale x 16 x float> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1F64 = frem <1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4F64 = frem <4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V8F64 = frem <8 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F64 = frem <vscale x 1 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F64 = frem <vscale x 2 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F64 = frem <vscale x 4 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F64 = frem <vscale x 8 x double> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V1F16_VP = call <1 x bfloat> @llvm.vp.frem.v1bf16(<1 x bfloat> undef, <1 x bfloat> undef, <1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2F16_VP = call <2 x bfloat> @llvm.vp.frem.v2bf16(<2 x bfloat> undef, <2 x bfloat> undef, <2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4F16_VP = call <4 x bfloat> @llvm.vp.frem.v4bf16(<4 x bfloat> undef, <4 x bfloat> undef, <4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V8F16_VP = call <8 x bfloat> @llvm.vp.frem.v8bf16(<8 x bfloat> undef, <8 x bfloat> undef, <8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 94 for instruction: %V16F16_VP = call <16 x bfloat> @llvm.vp.frem.v16bf16(<16 x bfloat> undef, <16 x bfloat> undef, <16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V1F32_VP = call <1 x float> @llvm.vp.frem.v1f32(<1 x float> undef, <1 x float> undef, <1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2F32_VP = call <2 x float> @llvm.vp.frem.v2f32(<2 x float> undef, <2 x float> undef, <2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4F32_VP = call <4 x float> @llvm.vp.frem.v4f32(<4 x float> undef, <4 x float> undef, <4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8F32_VP = call <8 x float> @llvm.vp.frem.v8f32(<8 x float> undef, <8 x float> undef, <8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 94 for instruction: %V16F32_VP = call <16 x float> @llvm.vp.frem.v16f32(<16 x float> undef, <16 x float> undef, <16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V1F64_VP = call <1 x double> @llvm.vp.frem.v1f64(<1 x double> undef, <1 x double> undef, <1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2F64_VP = call <2 x double> @llvm.vp.frem.v2f64(<2 x double> undef, <2 x double> undef, <2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4F64_VP = call <4 x double> @llvm.vp.frem.v4f64(<4 x double> undef, <4 x double> undef, <4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8F64_VP = call <8 x double> @llvm.vp.frem.v8f64(<8 x double> undef, <8 x double> undef, <8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16_VP = call <vscale x 1 x bfloat> @llvm.vp.frem.nxv1bf16(<vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef, <vscale x 1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16_VP = call <vscale x 2 x bfloat> @llvm.vp.frem.nxv2bf16(<vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef, <vscale x 2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16_VP = call <vscale x 4 x bfloat> @llvm.vp.frem.nxv4bf16(<vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef, <vscale x 4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16_VP = call <vscale x 8 x bfloat> @llvm.vp.frem.nxv8bf16(<vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef, <vscale x 8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16_VP = call <vscale x 16 x bfloat> @llvm.vp.frem.nxv16bf16(<vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef, <vscale x 16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.frem.nxv1f32(<vscale x 1 x float> undef, <vscale x 1 x float> undef, <vscale x 1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.frem.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> undef, <vscale x 2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.frem.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x float> undef, <vscale x 4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.frem.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x float> undef, <vscale x 8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.frem.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x float> undef, <vscale x 16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.frem.nxv1f64(<vscale x 1 x double> undef, <vscale x 1 x double> undef, <vscale x 1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.frem.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x double> undef, <vscale x 2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.frem.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x double> undef, <vscale x 4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.frem.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x double> undef, <vscale x 8 x i1> undef, i32 undef)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = frem float poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = frem double poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1F32 = frem <1 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2F32 = frem <2 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4F32 = frem <4 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V8F32 = frem <8 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V16F32 = frem <16 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F32 = frem <vscale x 1 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F32 = frem <vscale x 2 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F32 = frem <vscale x 4 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F32 = frem <vscale x 8 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F32 = frem <vscale x 16 x float> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1F64 = frem <1 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2F64 = frem <2 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 15 for instruction: %V4F64 = frem <4 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 31 for instruction: %V8F64 = frem <8 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F64 = frem <vscale x 1 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F64 = frem <vscale x 2 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F64 = frem <vscale x 4 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F64 = frem <vscale x 8 x double> poison, poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V1F32_VP = call <1 x float> @llvm.vp.frem.v1f32(<1 x float> poison, <1 x float> poison, <1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2F32_VP = call <2 x float> @llvm.vp.frem.v2f32(<2 x float> poison, <2 x float> poison, <2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4F32_VP = call <4 x float> @llvm.vp.frem.v4f32(<4 x float> poison, <4 x float> poison, <4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8F32_VP = call <8 x float> @llvm.vp.frem.v8f32(<8 x float> poison, <8 x float> poison, <8 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 94 for instruction: %V16F32_VP = call <16 x float> @llvm.vp.frem.v16f32(<16 x float> poison, <16 x float> poison, <16 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V1F64_VP = call <1 x double> @llvm.vp.frem.v1f64(<1 x double> poison, <1 x double> poison, <1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2F64_VP = call <2 x double> @llvm.vp.frem.v2f64(<2 x double> poison, <2 x double> poison, <2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 22 for instruction: %V4F64_VP = call <4 x double> @llvm.vp.frem.v4f64(<4 x double> poison, <4 x double> poison, <4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 46 for instruction: %V8F64_VP = call <8 x double> @llvm.vp.frem.v8f64(<8 x double> poison, <8 x double> poison, <8 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.frem.nxv1f32(<vscale x 1 x float> poison, <vscale x 1 x float> poison, <vscale x 1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.frem.nxv2f32(<vscale x 2 x float> poison, <vscale x 2 x float> poison, <vscale x 2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.frem.nxv4f32(<vscale x 4 x float> poison, <vscale x 4 x float> poison, <vscale x 4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.frem.nxv8f32(<vscale x 8 x float> poison, <vscale x 8 x float> poison, <vscale x 8 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.frem.nxv16f32(<vscale x 16 x float> poison, <vscale x 16 x float> poison, <vscale x 16 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.frem.nxv1f64(<vscale x 1 x double> poison, <vscale x 1 x double> poison, <vscale x 1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.frem.nxv2f64(<vscale x 2 x double> poison, <vscale x 2 x double> poison, <vscale x 2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.frem.nxv4f64(<vscale x 4 x double> poison, <vscale x 4 x double> poison, <vscale x 4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.frem.nxv8f64(<vscale x 8 x double> poison, <vscale x 8 x double> poison, <vscale x 8 x i1> poison, i32 poison)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
-  %BF16 = frem bfloat undef, undef
-  %F32 = frem float undef, undef
-  %F64 = frem double undef, undef
-
-  %V1BF16 = frem <1 x bfloat> undef, undef
-  %V2BF16 = frem <2 x bfloat> undef, undef
-  %V4BF16 = frem <4 x bfloat> undef, undef
-  %V8BF16 = frem <8 x bfloat> undef, undef
-  %V16BF16 = frem <16 x bfloat> undef, undef
-
-  %NXV1BF16 = frem <vscale x 1 x bfloat> undef, undef
-  %NXV2BF16 = frem <vscale x 2 x bfloat> undef, undef
-  %NXV4BF16 = frem <vscale x 4 x bfloat> undef, undef
-  %NXV8BF16 = frem <vscale x 8 x bfloat> undef, undef
-  %NXV16BF16 = frem <vscale x 16 x bfloat> undef, undef
-
-  %V1F32 = frem <1 x float> undef, undef
-  %V2F32 = frem <2 x float> undef, undef
-  %V4F32 = frem <4 x float> undef, undef
-  %V8F32 = frem <8 x float> undef, undef
-  %V16F32 = frem <16 x float> undef, undef
-
-  %NXV1F32 = frem <vscale x 1 x float> undef, undef
-  %NXV2F32 = frem <vscale x 2 x float> undef, undef
-  %NXV4F32 = frem <vscale x 4 x float> undef, undef
-  %NXV8F32 = frem <vscale x 8 x float> undef, undef
-  %NXV16F32 = frem <vscale x 16 x float> undef, undef
-
-  %V1F64 = frem <1 x double> undef, undef
-  %V2F64 = frem <2 x double> undef, undef
-  %V4F64 = frem <4 x double> undef, undef
-  %V8F64 = frem <8 x double> undef, undef
-
-  %NXV1F64 = frem <vscale x 1 x double> undef, undef
-  %NXV2F64 = frem <vscale x 2 x double> undef, undef
-  %NXV4F64 = frem <vscale x 4 x double> undef, undef
-  %NXV8F64 = frem <vscale x 8 x double> undef, undef
-
-  %V1F16_VP = call <1 x bfloat> @llvm.vp.frem.v1f16(<1 x bfloat> undef, <1 x bfloat> undef, <1 x i1> undef, i32 undef)
-  %V2F16_VP = call <2 x bfloat> @llvm.vp.frem.v2f16(<2 x bfloat> undef, <2 x bfloat> undef, <2 x i1> undef, i32 undef)
-  %V4F16_VP = call <4 x bfloat> @llvm.vp.frem.v4f16(<4 x bfloat> undef, <4 x bfloat> undef, <4 x i1> undef, i32 undef)
-  %V8F16_VP = call <8 x bfloat> @llvm.vp.frem.v8f16(<8 x bfloat> undef, <8 x bfloat> undef, <8 x i1> undef, i32 undef)
-  %V16F16_VP = call <16 x bfloat> @llvm.vp.frem.v16f16(<16 x bfloat> undef, <16 x bfloat> undef, <16 x i1> undef, i32 undef)
-
-  %V1F32_VP = call <1 x float> @llvm.vp.frem.v1f32(<1 x float> undef, <1 x float> undef, <1 x i1> undef, i32 undef)
-  %V2F32_VP = call <2 x float> @llvm.vp.frem.v2f32(<2 x float> undef, <2 x float> undef, <2 x i1> undef, i32 undef)
-  %V4F32_VP = call <4 x float> @llvm.vp.frem.v4f32(<4 x float> undef, <4 x float> undef, <4 x i1> undef, i32 undef)
-  %V8F32_VP = call <8 x float> @llvm.vp.frem.v8f32(<8 x float> undef, <8 x float> undef, <8 x i1> undef, i32 undef)
-  %V16F32_VP = call <16 x float> @llvm.vp.frem.v16f32(<16 x float> undef, <16 x float> undef, <16 x i1> undef, i32 undef)
-
-  %V1F64_VP = call <1 x double> @llvm.vp.frem.v1f64(<1 x double> undef, <1 x double> undef, <1 x i1> undef, i32 undef)
-  %V2F64_VP = call <2 x double> @llvm.vp.frem.v2f64(<2 x double> undef, <2 x double> undef, <2 x i1> undef, i32 undef)
-  %V4F64_VP = call <4 x double> @llvm.vp.frem.v4f64(<4 x double> undef, <4 x double> undef, <4 x i1> undef, i32 undef)
-  %V8F64_VP = call <8 x double> @llvm.vp.frem.v8f64(<8 x double> undef, <8 x double> undef, <8 x i1> undef, i32 undef)
-
-  %NXV1F16_VP = call <vscale x 1 x bfloat> @llvm.vp.frem.nxv1f16(<vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2F16_VP = call <vscale x 2 x bfloat> @llvm.vp.frem.nxv2f16(<vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4F16_VP = call <vscale x 4 x bfloat> @llvm.vp.frem.nxv4f16(<vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8F16_VP = call <vscale x 8 x bfloat> @llvm.vp.frem.nxv8f16(<vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef, <vscale x 8 x i1> undef, i32 undef)
-  %NXV16F16_VP = call <vscale x 16 x bfloat> @llvm.vp.frem.nxv16f16(<vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef, <vscale x 16 x i1> undef, i32 undef)
-
-  %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.frem.nxv1f32(<vscale x 1 x float> undef, <vscale x 1 x float> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.frem.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.frem.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x float> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.frem.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x float> undef, <vscale x 8 x i1> undef, i32 undef)
-  %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.frem.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x float> undef, <vscale x 16 x i1> undef, i32 undef)
-
-  %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.frem.nxv1f64(<vscale x 1 x double> undef, <vscale x 1 x double> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.frem.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x double> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.frem.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x double> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.frem.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x double> undef, <vscale x 8 x i1> undef, i32 undef)
+  %F32 = frem float poison, poison
+  %F64 = frem double poison, poison
+
+  %V1F32 = frem <1 x float> poison, poison
+  %V2F32 = frem <2 x float> poison, poison
+  %V4F32 = frem <4 x float> poison, poison
+  %V8F32 = frem <8 x float> poison, poison
+  %V16F32 = frem <16 x float> poison, poison
+
+  %NXV1F32 = frem <vscale x 1 x float> poison, poison
+  %NXV2F32 = frem <vscale x 2 x float> poison, poison
+  %NXV4F32 = frem <vscale x 4 x float> poison, poison
+  %NXV8F32 = frem <vscale x 8 x float> poison, poison
+  %NXV16F32 = frem <vscale x 16 x float> poison, poison
+
+  %V1F64 = frem <1 x double> poison, poison
+  %V2F64 = frem <2 x double> poison, poison
+  %V4F64 = frem <4 x double> poison, poison
+  %V8F64 = frem <8 x double> poison, poison
+
+  %NXV1F64 = frem <vscale x 1 x double> poison, poison
+  %NXV2F64 = frem <vscale x 2 x double> poison, poison
+  %NXV4F64 = frem <vscale x 4 x double> poison, poison
+  %NXV8F64 = frem <vscale x 8 x double> poison, poison
+
+  %V1F32_VP = call <1 x float> @llvm.vp.frem(<1 x float> poison, <1 x float> poison, <1 x i1> poison, i32 poison)
+  %V2F32_VP = call <2 x float> @llvm.vp.frem(<2 x float> poison, <2 x float> poison, <2 x i1> poison, i32 poison)
+  %V4F32_VP = call <4 x float> @llvm.vp.frem(<4 x float> poison, <4 x float> poison, <4 x i1> poison, i32 poison)
+  %V8F32_VP = call <8 x float> @llvm.vp.frem(<8 x float> poison, <8 x float> poison, <8 x i1> poison, i32 poison)
+  %V16F32_VP = call <16 x float> @llvm.vp.frem(<16 x float> poison, <16 x float> poison, <16 x i1> poison, i32 poison)
+
+  %V1F64_VP = call <1 x double> @llvm.vp.frem(<1 x double> poison, <1 x double> poison, <1 x i1> poison, i32 poison)
+  %V2F64_VP = call <2 x double> @llvm.vp.frem(<2 x double> poison, <2 x double> poison, <2 x i1> poison, i32 poison)
+  %V4F64_VP = call <4 x double> @llvm.vp.frem(<4 x double> poison, <4 x double> poison, <4 x i1> poison, i32 poison)
+  %V8F64_VP = call <8 x double> @llvm.vp.frem(<8 x double> poison, <8 x double> poison, <8 x i1> poison, i32 poison)
+
+  %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.frem(<vscale x 1 x float> poison, <vscale x 1 x float> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.frem(<vscale x 2 x float> poison, <vscale x 2 x float> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.frem(<vscale x 4 x float> poison, <vscale x 4 x float> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.frem(<vscale x 8 x float> poison, <vscale x 8 x float> poison, <vscale x 8 x i1> poison, i32 poison)
+  %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.frem(<vscale x 16 x float> poison, <vscale x 16 x float> poison, <vscale x 16 x i1> poison, i32 poison)
+
+  %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.frem(<vscale x 1 x double> poison, <vscale x 1 x double> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.frem(<vscale x 2 x double> poison, <vscale x 2 x double> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.frem(<vscale x 4 x double> poison, <vscale x 4 x double> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.frem(<vscale x 8 x double> poison, <vscale x 8 x double> poison, <vscale x 8 x i1> poison, i32 poison)
+
+  ret void
+}
+
+define void @frem_bf16() {
+; ZVFH-LABEL: 'frem_bf16'
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = frem bfloat poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1BF16 = frem <1 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2BF16 = frem <2 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4BF16 = frem <4 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8BF16 = frem <8 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V16BF16 = frem <16 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 127 for instruction: %V32BF16 = frem <32 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16 = frem <vscale x 1 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16 = frem <vscale x 2 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16 = frem <vscale x 4 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16 = frem <vscale x 8 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16 = frem <vscale x 16 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV32BF16 = frem <vscale x 32 x bfloat> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V1BF16_VP = call <1 x bfloat> @llvm.vp.frem.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2BF16_VP = call <2 x bfloat> @llvm.vp.frem.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4BF16_VP = call <4 x bfloat> @llvm.vp.frem.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V8BF16_VP = call <8 x bfloat> @llvm.vp.frem.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 94 for instruction: %V16BF16_VP = call <16 x bfloat> @llvm.vp.frem.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.frem.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.frem.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.frem.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.frem.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.frem.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; ZVFHMIN-LABEL: 'frem_bf16'
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = frem bfloat poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1BF16 = frem <1 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2BF16 = frem <2 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4BF16 = frem <4 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8BF16 = frem <8 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V16BF16 = frem <16 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 127 for instruction: %V32BF16 = frem <32 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16 = frem <vscale x 1 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16 = frem <vscale x 2 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16 = frem <vscale x 4 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16 = frem <vscale x 8 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16 = frem <vscale x 16 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32BF16 = frem <vscale x 32 x bfloat> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V1BF16_VP = call <1 x bfloat> @llvm.vp.frem.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2BF16_VP = call <2 x bfloat> @llvm.vp.frem.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4BF16_VP = call <4 x bfloat> @llvm.vp.frem.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V8BF16_VP = call <8 x bfloat> @llvm.vp.frem.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 94 for instruction: %V16BF16_VP = call <16 x bfloat> @llvm.vp.frem.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.frem.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.frem.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.frem.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.frem.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.frem.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; NO-ZFHMIN-LABEL: 'frem_bf16'
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = frem bfloat poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1BF16 = frem <1 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2BF16 = frem <2 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4BF16 = frem <4 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8BF16 = frem <8 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16BF16 = frem <16 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V32BF16 = frem <32 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16 = frem <vscale x 1 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16 = frem <vscale x 2 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16 = frem <vscale x 4 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16 = frem <vscale x 8 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16 = frem <vscale x 16 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32BF16 = frem <vscale x 32 x bfloat> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1BF16_VP = call <1 x bfloat> @llvm.vp.frem.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2BF16_VP = call <2 x bfloat> @llvm.vp.frem.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4BF16_VP = call <4 x bfloat> @llvm.vp.frem.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8BF16_VP = call <8 x bfloat> @llvm.vp.frem.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16BF16_VP = call <16 x bfloat> @llvm.vp.frem.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.frem.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.frem.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.frem.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.frem.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.frem.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+  %BF16 = frem bfloat poison, poison
+
+  %V1BF16 = frem <1 x bfloat> poison, poison
+  %V2BF16 = frem <2 x bfloat> poison, poison
+  %V4BF16 = frem <4 x bfloat> poison, poison
+  %V8BF16 = frem <8 x bfloat> poison, poison
+  %V16BF16 = frem <16 x bfloat> poison, poison
+  %V32BF16 = frem <32 x bfloat> poison, poison
+
+  %NXV1BF16 = frem <vscale x 1 x bfloat> poison, poison
+  %NXV2BF16 = frem <vscale x 2 x bfloat> poison, poison
+  %NXV4BF16 = frem <vscale x 4 x bfloat> poison, poison
+  %NXV8BF16 = frem <vscale x 8 x bfloat> poison, poison
+  %NXV16BF16 = frem <vscale x 16 x bfloat> poison, poison
+  %NXV32BF16 = frem <vscale x 32 x bfloat> poison, poison
+
+  %V1BF16_VP = call <1 x bfloat> @llvm.vp.frem(<1 x bfloat> poison, <1 x bfloat> poison, <1 x i1> poison, i32 poison)
+  %V2BF16_VP = call <2 x bfloat> @llvm.vp.frem(<2 x bfloat> poison, <2 x bfloat> poison, <2 x i1> poison, i32 poison)
+  %V4BF16_VP = call <4 x bfloat> @llvm.vp.frem(<4 x bfloat> poison, <4 x bfloat> poison, <4 x i1> poison, i32 poison)
+  %V8BF16_VP = call <8 x bfloat> @llvm.vp.frem(<8 x bfloat> poison, <8 x bfloat> poison, <8 x i1> poison, i32 poison)
+  %V16BF16_VP = call <16 x bfloat> @llvm.vp.frem(<16 x bfloat> poison, <16 x bfloat> poison, <16 x i1> poison, i32 poison)
+
+  %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.frem(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.frem(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.frem(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.frem(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+  %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.frem(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
 
   ret void
 }
 
 define void @frem_f16() {
-; CHECK-LABEL: 'frem_f16'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = frem half undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1F16 = frem <1 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2F16 = frem <2 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4F16 = frem <4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8F16 = frem <8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V16F16 = frem <16 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 127 for instruction: %V32F16 = frem <32 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16 = frem <vscale x 1 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16 = frem <vscale x 2 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16 = frem <vscale x 4 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16 = frem <vscale x 8 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16 = frem <vscale x 16 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV32F16 = frem <vscale x 32 x half> undef, undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+; ZVFH-LABEL: 'frem_f16'
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = frem half poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1F16 = frem <1 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2F16 = frem <2 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4F16 = frem <4 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8F16 = frem <8 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V16F16 = frem <16 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 127 for instruction: %V32F16 = frem <32 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16 = frem <vscale x 1 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16 = frem <vscale x 2 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16 = frem <vscale x 4 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16 = frem <vscale x 8 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16 = frem <vscale x 16 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV32F16 = frem <vscale x 32 x half> poison, poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.frem.v1f16(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.frem.v2f16(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.frem.v4f16(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.frem.v8f16(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 94 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.frem.v16f16(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.frem.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.frem.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.frem.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.frem.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.frem.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; ZVFHMIN-LABEL: 'frem_f16'
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = frem half poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %V1F16 = frem <1 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 6 for instruction: %V2F16 = frem <2 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 12 for instruction: %V4F16 = frem <4 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 24 for instruction: %V8F16 = frem <8 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 63 for instruction: %V16F16 = frem <16 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 127 for instruction: %V32F16 = frem <32 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16 = frem <vscale x 1 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16 = frem <vscale x 2 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16 = frem <vscale x 4 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16 = frem <vscale x 8 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16 = frem <vscale x 16 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32F16 = frem <vscale x 32 x half> poison, poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.frem.v1f16(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 9 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.frem.v2f16(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 19 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.frem.v4f16(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 39 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.frem.v8f16(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 94 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.frem.v16f16(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.frem.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.frem.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.frem.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.frem.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.frem.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
-  %F16 = frem half undef, undef
-
-  %V1F16 = frem <1 x half> undef, undef
-  %V2F16 = frem <2 x half> undef, undef
-  %V4F16 = frem <4 x half> undef, undef
-  %V8F16 = frem <8 x half> undef, undef
-  %V16F16 = frem <16 x half> undef, undef
-  %V32F16 = frem <32 x half> undef, undef
-
-  %NXV1F16 = frem <vscale x 1 x half> undef, undef
-  %NXV2F16 = frem <vscale x 2 x half> undef, undef
-  %NXV4F16 = frem <vscale x 4 x half> undef, undef
-  %NXV8F16 = frem <vscale x 8 x half> undef, undef
-  %NXV16F16 = frem <vscale x 16 x half> undef, undef
-  %NXV32F16 = frem <vscale x 32 x half> undef, undef
+; NO-ZFHMIN-LABEL: 'frem_f16'
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = frem half poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16 = frem <1 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2F16 = frem <2 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4F16 = frem <4 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8F16 = frem <8 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16F16 = frem <16 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V32F16 = frem <32 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16 = frem <vscale x 1 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16 = frem <vscale x 2 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16 = frem <vscale x 4 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16 = frem <vscale x 8 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16 = frem <vscale x 16 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32F16 = frem <vscale x 32 x half> poison, poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.frem.v1f16(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.frem.v2f16(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.frem.v4f16(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.frem.v8f16(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.frem.v16f16(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.frem.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.frem.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.frem.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.frem.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.frem.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+  %F16 = frem half poison, poison
+
+  %V1F16 = frem <1 x half> poison, poison
+  %V2F16 = frem <2 x half> poison, poison
+  %V4F16 = frem <4 x half> poison, poison
+  %V8F16 = frem <8 x half> poison, poison
+  %V16F16 = frem <16 x half> poison, poison
+  %V32F16 = frem <32 x half> poison, poison
+
+  %NXV1F16 = frem <vscale x 1 x half> poison, poison
+  %NXV2F16 = frem <vscale x 2 x half> poison, poison
+  %NXV4F16 = frem <vscale x 4 x half> poison, poison
+  %NXV8F16 = frem <vscale x 8 x half> poison, poison
+  %NXV16F16 = frem <vscale x 16 x half> poison, poison
+  %NXV32F16 = frem <vscale x 32 x half> poison, poison
+
+  %V1F16_VP = call <1 x half> @llvm.vp.frem(<1 x half> poison, <1 x half> poison, <1 x i1> poison, i32 poison)
+  %V2F16_VP = call <2 x half> @llvm.vp.frem(<2 x half> poison, <2 x half> poison, <2 x i1> poison, i32 poison)
+  %V4F16_VP = call <4 x half> @llvm.vp.frem(<4 x half> poison, <4 x half> poison, <4 x i1> poison, i32 poison)
+  %V8F16_VP = call <8 x half> @llvm.vp.frem(<8 x half> poison, <8 x half> poison, <8 x i1> poison, i32 poison)
+  %V16F16_VP = call <16 x half> @llvm.vp.frem(<16 x half> poison, <16 x half> poison, <16 x i1> poison, i32 poison)
+
+  %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.frem(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.frem(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.frem(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.frem(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+  %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.frem(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
 
   ret void
 }
 
 define void @fneg() {
 ; CHECK-LABEL: 'fneg'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = fneg half undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = fneg float undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = fneg double undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1BF16 = fneg <1 x bfloat> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2BF16 = fneg <2 x bfloat> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4BF16 = fneg <4 x bfloat> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8BF16 = fneg <8 x bfloat> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16BF16 = fneg <16 x bfloat> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1BF16 = fneg <vscale x 1 x bfloat> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2BF16 = fneg <vscale x 2 x bfloat> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4BF16 = fneg <vscale x 4 x bfloat> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8BF16 = fneg <vscale x 8 x bfloat> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16BF16 = fneg <vscale x 16 x bfloat> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32 = fneg <1 x float> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32 = fneg <2 x float> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fneg <4 x float> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fneg <8 x float> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fneg <16 x float> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F32 = fneg <vscale x 1 x float> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F32 = fneg <vscale x 2 x float> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F32 = fneg <vscale x 4 x float> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F32 = fneg <vscale x 8 x float> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16F32 = fneg <vscale x 16 x float> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64 = fneg <1 x double> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fneg <2 x double> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fneg <4 x double> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fneg <8 x double> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F64 = fneg <vscale x 1 x double> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2F64 = fneg <vscale x 2 x double> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4F64 = fneg <vscale x 4 x double> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8F64 = fneg <vscale x 8 x double> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1BF16_VP = call <1 x bfloat> @llvm.vp.fneg.v1bf16(<1 x bfloat> undef, <1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2BF16_VP = call <2 x bfloat> @llvm.vp.fneg.v2bf16(<2 x bfloat> undef, <2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4BF16_VP = call <4 x bfloat> @llvm.vp.fneg.v4bf16(<4 x bfloat> undef, <4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8BF16_VP = call <8 x bfloat> @llvm.vp.fneg.v8bf16(<8 x bfloat> undef, <8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16BF16_VP = call <16 x bfloat> @llvm.vp.fneg.v16bf16(<16 x bfloat> undef, <16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32_VP = call <1 x float> @llvm.vp.fneg.v1f32(<1 x float> undef, <1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32_VP = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> undef, <2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32_VP = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> undef, <4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32_VP = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> undef, <8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32_VP = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> undef, <16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64_VP = call <1 x double> @llvm.vp.fneg.v1f64(<1 x double> undef, <1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64_VP = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> undef, <2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64_VP = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> undef, <4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64_VP = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> undef, <8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16F64_VP = call <16 x double> @llvm.vp.fneg.v16f64(<16 x double> undef, <16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fneg.nxv1bf16(<vscale x 1 x bfloat> undef, <vscale x 1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fneg.nxv2bf16(<vscale x 2 x bfloat> undef, <vscale x 2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fneg.nxv4bf16(<vscale x 4 x bfloat> undef, <vscale x 4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fneg.nxv8bf16(<vscale x 8 x bfloat> undef, <vscale x 8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fneg.nxv16bf16(<vscale x 16 x bfloat> undef, <vscale x 16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1BF32_VP = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> undef, <vscale x 1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2BF32_VP = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4BF32_VP = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8BF32_VP = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16BF32_VP = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1BF64_VP = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> undef, <vscale x 1 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2BF64_VP = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4BF64_VP = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8BF64_VP = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x i1> undef, i32 undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %NXV16BF64_VP = call <vscale x 16 x double> @llvm.vp.fneg.nxv16f64(<vscale x 16 x double> undef, <vscale x 16 x i1> undef, i32 undef)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F32 = fneg float poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F64 = fneg double poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32 = fneg <1 x float> poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32 = fneg <2 x float> poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = fneg <4 x float> poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32 = fneg <8 x float> poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32 = fneg <16 x float> poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F32 = fneg <vscale x 1 x float> poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F32 = fneg <vscale x 2 x float> poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F32 = fneg <vscale x 4 x float> poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F32 = fneg <vscale x 8 x float> poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16F32 = fneg <vscale x 16 x float> poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64 = fneg <1 x double> poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = fneg <2 x double> poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64 = fneg <4 x double> poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64 = fneg <8 x double> poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F64 = fneg <vscale x 1 x double> poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2F64 = fneg <vscale x 2 x double> poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4F64 = fneg <vscale x 4 x double> poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8F64 = fneg <vscale x 8 x double> poison
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32_VP = call <1 x float> @llvm.vp.fneg.v1f32(<1 x float> poison, <1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32_VP = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> poison, <2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32_VP = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> poison, <4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F32_VP = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> poison, <8 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V16F32_VP = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> poison, <16 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64_VP = call <1 x double> @llvm.vp.fneg.v1f64(<1 x double> poison, <1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64_VP = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> poison, <2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V4F64_VP = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> poison, <4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V8F64_VP = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> poison, <8 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V16F64_VP = call <16 x double> @llvm.vp.fneg.v16f64(<16 x double> poison, <16 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> poison, <vscale x 1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> poison, <vscale x 2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> poison, <vscale x 4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> poison, <vscale x 8 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> poison, <vscale x 16 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> poison, <vscale x 1 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> poison, <vscale x 2 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> poison, <vscale x 4 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> poison, <vscale x 8 x i1> poison, i32 poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %NXV16F64_VP = call <vscale x 16 x double> @llvm.vp.fneg.nxv16f64(<vscale x 16 x double> poison, <vscale x 16 x i1> poison, i32 poison)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
-  %BF16 = fneg half undef
-  %F32 = fneg float undef
-  %F64 = fneg double undef
-
-  %V1BF16 = fneg <1 x bfloat> undef
-  %V2BF16 = fneg <2 x bfloat> undef
-  %V4BF16 = fneg <4 x bfloat> undef
-  %V8BF16 = fneg <8 x bfloat> undef
-  %V16BF16 = fneg <16 x bfloat> undef
-
-  %NXV1BF16 = fneg <vscale x 1 x bfloat> undef
-  %NXV2BF16 = fneg <vscale x 2 x bfloat> undef
-  %NXV4BF16 = fneg <vscale x 4 x bfloat> undef
-  %NXV8BF16 = fneg <vscale x 8 x bfloat> undef
-  %NXV16BF16 = fneg <vscale x 16 x bfloat> undef
-
-  %V1F32 = fneg <1 x float> undef
-  %V2F32 = fneg <2 x float> undef
-  %V4F32 = fneg <4 x float> undef
-  %V8F32 = fneg <8 x float> undef
-  %V16F32 = fneg <16 x float> undef
-
-  %NXV1F32 = fneg <vscale x 1 x float> undef
-  %NXV2F32 = fneg <vscale x 2 x float> undef
-  %NXV4F32 = fneg <vscale x 4 x float> undef
-  %NXV8F32 = fneg <vscale x 8 x float> undef
-  %NXV16F32 = fneg <vscale x 16 x float> undef
-
-  %V1F64 = fneg <1 x double> undef
-  %V2F64 = fneg <2 x double> undef
-  %V4F64 = fneg <4 x double> undef
-  %V8F64 = fneg <8 x double> undef
-
-  %NXV1F64 = fneg <vscale x 1 x double> undef
-  %NXV2F64 = fneg <vscale x 2 x double> undef
-  %NXV4F64 = fneg <vscale x 4 x double> undef
-  %NXV8F64 = fneg <vscale x 8 x double> undef
-
-  %V1BF16_VP = call <1 x bfloat> @llvm.vp.fneg.v1f16(<1 x bfloat> undef, <1 x i1> undef, i32 undef)
-  %V2BF16_VP = call <2 x bfloat> @llvm.vp.fneg.v2f16(<2 x bfloat> undef, <2 x i1> undef, i32 undef)
-  %V4BF16_VP = call <4 x bfloat> @llvm.vp.fneg.v4f16(<4 x bfloat> undef, <4 x i1> undef, i32 undef)
-  %V8BF16_VP = call <8 x bfloat> @llvm.vp.fneg.v8f16(<8 x bfloat> undef, <8 x i1> undef, i32 undef)
-  %V16BF16_VP = call <16 x bfloat> @llvm.vp.fneg.v16f16(<16 x bfloat> undef, <16 x i1> undef, i32 undef)
-
-  %V1F32_VP = call <1 x float> @llvm.vp.fneg.v1f32(<1 x float> undef, <1 x i1> undef, i32 undef)
-  %V2F32_VP = call <2 x float> @llvm.vp.fneg.v2f32(<2 x float> undef, <2 x i1> undef, i32 undef)
-  %V4F32_VP = call <4 x float> @llvm.vp.fneg.v4f32(<4 x float> undef, <4 x i1> undef, i32 undef)
-  %V8F32_VP = call <8 x float> @llvm.vp.fneg.v8f32(<8 x float> undef, <8 x i1> undef, i32 undef)
-  %V16F32_VP = call <16 x float> @llvm.vp.fneg.v16f32(<16 x float> undef, <16 x i1> undef, i32 undef)
-
-  %V1F64_VP = call <1 x double> @llvm.vp.fneg.v1f64(<1 x double> undef, <1 x i1> undef, i32 undef)
-  %V2F64_VP = call <2 x double> @llvm.vp.fneg.v2f64(<2 x double> undef, <2 x i1> undef, i32 undef)
-  %V4F64_VP = call <4 x double> @llvm.vp.fneg.v4f64(<4 x double> undef, <4 x i1> undef, i32 undef)
-  %V8F64_VP = call <8 x double> @llvm.vp.fneg.v8f64(<8 x double> undef, <8 x i1> undef, i32 undef)
-  %V16F64_VP = call <16 x double> @llvm.vp.fneg.v16f64(<16 x double> undef, <16 x i1> undef, i32 undef)
-
-  %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fneg.nxv1f16(<vscale x 1 x bfloat> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fneg.nxv2f16(<vscale x 2 x bfloat> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fneg.nxv4f16(<vscale x 4 x bfloat> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fneg.nxv8f16(<vscale x 8 x bfloat> undef, <vscale x 8 x i1> undef, i32 undef)
-  %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fneg.nxv16f16(<vscale x 16 x bfloat> undef, <vscale x 16 x i1> undef, i32 undef)
-
-  %NXV1BF32_VP = call <vscale x 1 x float> @llvm.vp.fneg.nxv1f32(<vscale x 1 x float> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2BF32_VP = call <vscale x 2 x float> @llvm.vp.fneg.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4BF32_VP = call <vscale x 4 x float> @llvm.vp.fneg.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8BF32_VP = call <vscale x 8 x float> @llvm.vp.fneg.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x i1> undef, i32 undef)
-  %NXV16BF32_VP = call <vscale x 16 x float> @llvm.vp.fneg.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x i1> undef, i32 undef)
-
-  %NXV1BF64_VP = call <vscale x 1 x double> @llvm.vp.fneg.nxv1f64(<vscale x 1 x double> undef, <vscale x 1 x i1> undef, i32 undef)
-  %NXV2BF64_VP = call <vscale x 2 x double> @llvm.vp.fneg.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x i1> undef, i32 undef)
-  %NXV4BF64_VP = call <vscale x 4 x double> @llvm.vp.fneg.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x i1> undef, i32 undef)
-  %NXV8BF64_VP = call <vscale x 8 x double> @llvm.vp.fneg.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x i1> undef, i32 undef)
-  %NXV16BF64_VP = call <vscale x 16 x double> @llvm.vp.fneg.nxv16f64(<vscale x 16 x double> undef, <vscale x 16 x i1> undef, i32 undef)
+  %F32 = fneg float poison
+  %F64 = fneg double poison
+
+  %V1F32 = fneg <1 x float> poison
+  %V2F32 = fneg <2 x float> poison
+  %V4F32 = fneg <4 x float> poison
+  %V8F32 = fneg <8 x float> poison
+  %V16F32 = fneg <16 x float> poison
+
+  %NXV1F32 = fneg <vscale x 1 x float> poison
+  %NXV2F32 = fneg <vscale x 2 x float> poison
+  %NXV4F32 = fneg <vscale x 4 x float> poison
+  %NXV8F32 = fneg <vscale x 8 x float> poison
+  %NXV16F32 = fneg <vscale x 16 x float> poison
+
+  %V1F64 = fneg <1 x double> poison
+  %V2F64 = fneg <2 x double> poison
+  %V4F64 = fneg <4 x double> poison
+  %V8F64 = fneg <8 x double> poison
+
+  %NXV1F64 = fneg <vscale x 1 x double> poison
+  %NXV2F64 = fneg <vscale x 2 x double> poison
+  %NXV4F64 = fneg <vscale x 4 x double> poison
+  %NXV8F64 = fneg <vscale x 8 x double> poison
+
+  %V1F32_VP = call <1 x float> @llvm.vp.fneg(<1 x float> poison, <1 x i1> poison, i32 poison)
+  %V2F32_VP = call <2 x float> @llvm.vp.fneg(<2 x float> poison, <2 x i1> poison, i32 poison)
+  %V4F32_VP = call <4 x float> @llvm.vp.fneg(<4 x float> poison, <4 x i1> poison, i32 poison)
+  %V8F32_VP = call <8 x float> @llvm.vp.fneg(<8 x float> poison, <8 x i1> poison, i32 poison)
+  %V16F32_VP = call <16 x float> @llvm.vp.fneg(<16 x float> poison, <16 x i1> poison, i32 poison)
+
+  %V1F64_VP = call <1 x double> @llvm.vp.fneg(<1 x double> poison, <1 x i1> poison, i32 poison)
+  %V2F64_VP = call <2 x double> @llvm.vp.fneg(<2 x double> poison, <2 x i1> poison, i32 poison)
+  %V4F64_VP = call <4 x double> @llvm.vp.fneg(<4 x double> poison, <4 x i1> poison, i32 poison)
+  %V8F64_VP = call <8 x double> @llvm.vp.fneg(<8 x double> poison, <8 x i1> poison, i32 poison)
+  %V16F64_VP = call <16 x double> @llvm.vp.fneg(<16 x double> poison, <16 x i1> poison, i32 poison)
+
+  %NXV1F32_VP = call <vscale x 1 x float> @llvm.vp.fneg(<vscale x 1 x float> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2F32_VP = call <vscale x 2 x float> @llvm.vp.fneg(<vscale x 2 x float> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4F32_VP = call <vscale x 4 x float> @llvm.vp.fneg(<vscale x 4 x float> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8F32_VP = call <vscale x 8 x float> @llvm.vp.fneg(<vscale x 8 x float> poison, <vscale x 8 x i1> poison, i32 poison)
+  %NXV16F32_VP = call <vscale x 16 x float> @llvm.vp.fneg(<vscale x 16 x float> poison, <vscale x 16 x i1> poison, i32 poison)
+
+  %NXV1F64_VP = call <vscale x 1 x double> @llvm.vp.fneg(<vscale x 1 x double> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2F64_VP = call <vscale x 2 x double> @llvm.vp.fneg(<vscale x 2 x double> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4F64_VP = call <vscale x 4 x double> @llvm.vp.fneg(<vscale x 4 x double> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8F64_VP = call <vscale x 8 x double> @llvm.vp.fneg(<vscale x 8 x double> poison, <vscale x 8 x i1> poison, i32 poison)
+  %NXV16F64_VP = call <vscale x 16 x double> @llvm.vp.fneg(<vscale x 16 x double> poison, <vscale x 16 x i1> poison, i32 poison)
+
+  ret void
+}
+
+define void @fneg_bf16() {
+; ZVFH-LABEL: 'fneg_bf16'
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = fneg bfloat poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1BF16 = fneg <1 x bfloat> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2BF16 = fneg <2 x bfloat> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4BF16 = fneg <4 x bfloat> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8BF16 = fneg <8 x bfloat> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16BF16 = fneg <16 x bfloat> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32BF16 = fneg <32 x bfloat> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1BF16 = fneg <vscale x 1 x bfloat> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2BF16 = fneg <vscale x 2 x bfloat> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4BF16 = fneg <vscale x 4 x bfloat> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8BF16 = fneg <vscale x 8 x bfloat> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16BF16 = fneg <vscale x 16 x bfloat> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32BF16 = fneg <vscale x 32 x bfloat> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1BF16_VP = call <1 x bfloat> @llvm.vp.fneg.v1bf16(<1 x bfloat> poison, <1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2BF16_VP = call <2 x bfloat> @llvm.vp.fneg.v2bf16(<2 x bfloat> poison, <2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4BF16_VP = call <4 x bfloat> @llvm.vp.fneg.v4bf16(<4 x bfloat> poison, <4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8BF16_VP = call <8 x bfloat> @llvm.vp.fneg.v8bf16(<8 x bfloat> poison, <8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16BF16_VP = call <16 x bfloat> @llvm.vp.fneg.v16bf16(<16 x bfloat> poison, <16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fneg.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fneg.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fneg.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fneg.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fneg.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; ZVFHMIN-LABEL: 'fneg_bf16'
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = fneg bfloat poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1BF16 = fneg <1 x bfloat> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2BF16 = fneg <2 x bfloat> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4BF16 = fneg <4 x bfloat> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8BF16 = fneg <8 x bfloat> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16BF16 = fneg <16 x bfloat> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32BF16 = fneg <32 x bfloat> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1BF16 = fneg <vscale x 1 x bfloat> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2BF16 = fneg <vscale x 2 x bfloat> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4BF16 = fneg <vscale x 4 x bfloat> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8BF16 = fneg <vscale x 8 x bfloat> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16BF16 = fneg <vscale x 16 x bfloat> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32BF16 = fneg <vscale x 32 x bfloat> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1BF16_VP = call <1 x bfloat> @llvm.vp.fneg.v1bf16(<1 x bfloat> poison, <1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2BF16_VP = call <2 x bfloat> @llvm.vp.fneg.v2bf16(<2 x bfloat> poison, <2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4BF16_VP = call <4 x bfloat> @llvm.vp.fneg.v4bf16(<4 x bfloat> poison, <4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8BF16_VP = call <8 x bfloat> @llvm.vp.fneg.v8bf16(<8 x bfloat> poison, <8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16BF16_VP = call <16 x bfloat> @llvm.vp.fneg.v16bf16(<16 x bfloat> poison, <16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fneg.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fneg.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fneg.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fneg.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fneg.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; NO-ZFHMIN-LABEL: 'fneg_bf16'
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %BF16 = fneg bfloat poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1BF16 = fneg <1 x bfloat> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2BF16 = fneg <2 x bfloat> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4BF16 = fneg <4 x bfloat> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8BF16 = fneg <8 x bfloat> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16BF16 = fneg <16 x bfloat> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V32BF16 = fneg <32 x bfloat> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16 = fneg <vscale x 1 x bfloat> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16 = fneg <vscale x 2 x bfloat> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16 = fneg <vscale x 4 x bfloat> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16 = fneg <vscale x 8 x bfloat> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16 = fneg <vscale x 16 x bfloat> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32BF16 = fneg <vscale x 32 x bfloat> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1BF16_VP = call <1 x bfloat> @llvm.vp.fneg.v1bf16(<1 x bfloat> poison, <1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2BF16_VP = call <2 x bfloat> @llvm.vp.fneg.v2bf16(<2 x bfloat> poison, <2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4BF16_VP = call <4 x bfloat> @llvm.vp.fneg.v4bf16(<4 x bfloat> poison, <4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8BF16_VP = call <8 x bfloat> @llvm.vp.fneg.v8bf16(<8 x bfloat> poison, <8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16BF16_VP = call <16 x bfloat> @llvm.vp.fneg.v16bf16(<16 x bfloat> poison, <16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fneg.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fneg.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fneg.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fneg.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fneg.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+  %BF16 = fneg bfloat poison
+
+  %V1BF16 = fneg <1 x bfloat> poison
+  %V2BF16 = fneg <2 x bfloat> poison
+  %V4BF16 = fneg <4 x bfloat> poison
+  %V8BF16 = fneg <8 x bfloat> poison
+  %V16BF16 = fneg <16 x bfloat> poison
+  %V32BF16 = fneg <32 x bfloat> poison
+
+  %NXV1BF16 = fneg <vscale x 1 x bfloat> poison
+  %NXV2BF16 = fneg <vscale x 2 x bfloat> poison
+  %NXV4BF16 = fneg <vscale x 4 x bfloat> poison
+  %NXV8BF16 = fneg <vscale x 8 x bfloat> poison
+  %NXV16BF16 = fneg <vscale x 16 x bfloat> poison
+  %NXV32BF16 = fneg <vscale x 32 x bfloat> poison
+
+  %V1BF16_VP = call <1 x bfloat> @llvm.vp.fneg(<1 x bfloat> poison, <1 x i1> poison, i32 poison)
+  %V2BF16_VP = call <2 x bfloat> @llvm.vp.fneg(<2 x bfloat> poison, <2 x i1> poison, i32 poison)
+  %V4BF16_VP = call <4 x bfloat> @llvm.vp.fneg(<4 x bfloat> poison, <4 x i1> poison, i32 poison)
+  %V8BF16_VP = call <8 x bfloat> @llvm.vp.fneg(<8 x bfloat> poison, <8 x i1> poison, i32 poison)
+  %V16BF16_VP = call <16 x bfloat> @llvm.vp.fneg(<16 x bfloat> poison, <16 x i1> poison, i32 poison)
+
+  %NXV1BF16_VP = call <vscale x 1 x bfloat> @llvm.vp.fneg(<vscale x 1 x bfloat> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2BF16_VP = call <vscale x 2 x bfloat> @llvm.vp.fneg(<vscale x 2 x bfloat> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4BF16_VP = call <vscale x 4 x bfloat> @llvm.vp.fneg(<vscale x 4 x bfloat> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8BF16_VP = call <vscale x 8 x bfloat> @llvm.vp.fneg(<vscale x 8 x bfloat> poison, <vscale x 8 x i1> poison, i32 poison)
+  %NXV16BF16_VP = call <vscale x 16 x bfloat> @llvm.vp.fneg(<vscale x 16 x bfloat> poison, <vscale x 16 x i1> poison, i32 poison)
 
   ret void
 }
 
 define void @fneg_f16() {
-; CHECK-LABEL: 'fneg_f16'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fneg half undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16 = fneg <1 x half> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16 = fneg <2 x half> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16 = fneg <4 x half> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16 = fneg <8 x half> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16 = fneg <16 x half> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32F16 = fneg <32 x half> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16 = fneg <vscale x 1 x half> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16 = fneg <vscale x 2 x half> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16 = fneg <vscale x 4 x half> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16 = fneg <vscale x 8 x half> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16 = fneg <vscale x 16 x half> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32F16 = fneg <vscale x 32 x half> undef
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+; ZVFH-LABEL: 'fneg_f16'
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fneg half poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16 = fneg <1 x half> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16 = fneg <2 x half> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16 = fneg <4 x half> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16 = fneg <8 x half> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16 = fneg <16 x half> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32F16 = fneg <32 x half> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16 = fneg <vscale x 1 x half> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16 = fneg <vscale x 2 x half> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16 = fneg <vscale x 4 x half> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16 = fneg <vscale x 8 x half> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16 = fneg <vscale x 16 x half> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32F16 = fneg <vscale x 32 x half> poison
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fneg.v1f16(<1 x half> poison, <1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> poison, <2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> poison, <4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> poison, <8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> poison, <16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; ZVFHMIN-LABEL: 'fneg_f16'
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fneg half poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16 = fneg <1 x half> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16 = fneg <2 x half> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16 = fneg <4 x half> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16 = fneg <8 x half> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16 = fneg <16 x half> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V32F16 = fneg <32 x half> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16 = fneg <vscale x 1 x half> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16 = fneg <vscale x 2 x half> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16 = fneg <vscale x 4 x half> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16 = fneg <vscale x 8 x half> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16 = fneg <vscale x 16 x half> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %NXV32F16 = fneg <vscale x 32 x half> poison
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fneg.v1f16(<1 x half> poison, <1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> poison, <2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> poison, <4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> poison, <8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> poison, <16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
-  %F16 = fneg half undef
-
-  %V1F16 = fneg <1 x half> undef
-  %V2F16 = fneg <2 x half> undef
-  %V4F16 = fneg <4 x half> undef
-  %V8F16 = fneg <8 x half> undef
-  %V16F16 = fneg <16 x half> undef
-  %V32F16 = fneg <32 x half> undef
-
-  %NXV1F16 = fneg <vscale x 1 x half> undef
-  %NXV2F16 = fneg <vscale x 2 x half> undef
-  %NXV4F16 = fneg <vscale x 4 x half> undef
-  %NXV8F16 = fneg <vscale x 8 x half> undef
-  %NXV16F16 = fneg <vscale x 16 x half> undef
-  %NXV32F16 = fneg <vscale x 32 x half> undef
+; NO-ZFHMIN-LABEL: 'fneg_f16'
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %F16 = fneg half poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16 = fneg <1 x half> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2F16 = fneg <2 x half> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4F16 = fneg <4 x half> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8F16 = fneg <8 x half> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16F16 = fneg <16 x half> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V32F16 = fneg <32 x half> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16 = fneg <vscale x 1 x half> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16 = fneg <vscale x 2 x half> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16 = fneg <vscale x 4 x half> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16 = fneg <vscale x 8 x half> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16 = fneg <vscale x 16 x half> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32F16 = fneg <vscale x 32 x half> poison
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16_VP = call <1 x half> @llvm.vp.fneg.v1f16(<1 x half> poison, <1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2F16_VP = call <2 x half> @llvm.vp.fneg.v2f16(<2 x half> poison, <2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4F16_VP = call <4 x half> @llvm.vp.fneg.v4f16(<4 x half> poison, <4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8F16_VP = call <8 x half> @llvm.vp.fneg.v8f16(<8 x half> poison, <8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16F16_VP = call <16 x half> @llvm.vp.fneg.v16f16(<16 x half> poison, <16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fneg.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fneg.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fneg.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fneg.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fneg.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+  %F16 = fneg half poison
+
+  %V1F16 = fneg <1 x half> poison
+  %V2F16 = fneg <2 x half> poison
+  %V4F16 = fneg <4 x half> poison
+  %V8F16 = fneg <8 x half> poison
+  %V16F16 = fneg <16 x half> poison
+  %V32F16 = fneg <32 x half> poison
+
+  %NXV1F16 = fneg <vscale x 1 x half> poison
+  %NXV2F16 = fneg <vscale x 2 x half> poison
+  %NXV4F16 = fneg <vscale x 4 x half> poison
+  %NXV8F16 = fneg <vscale x 8 x half> poison
+  %NXV16F16 = fneg <vscale x 16 x half> poison
+  %NXV32F16 = fneg <vscale x 32 x half> poison
+
+  %V1F16_VP = call <1 x half> @llvm.vp.fneg(<1 x half> poison, <1 x i1> poison, i32 poison)
+  %V2F16_VP = call <2 x half> @llvm.vp.fneg(<2 x half> poison, <2 x i1> poison, i32 poison)
+  %V4F16_VP = call <4 x half> @llvm.vp.fneg(<4 x half> poison, <4 x i1> poison, i32 poison)
+  %V8F16_VP = call <8 x half> @llvm.vp.fneg(<8 x half> poison, <8 x i1> poison, i32 poison)
+  %V16F16_VP = call <16 x half> @llvm.vp.fneg(<16 x half> poison, <16 x i1> poison, i32 poison)
+
+  %NXV1F16_VP = call <vscale x 1 x half> @llvm.vp.fneg(<vscale x 1 x half> poison, <vscale x 1 x i1> poison, i32 poison)
+  %NXV2F16_VP = call <vscale x 2 x half> @llvm.vp.fneg(<vscale x 2 x half> poison, <vscale x 2 x i1> poison, i32 poison)
+  %NXV4F16_VP = call <vscale x 4 x half> @llvm.vp.fneg(<vscale x 4 x half> poison, <vscale x 4 x i1> poison, i32 poison)
+  %NXV8F16_VP = call <vscale x 8 x half> @llvm.vp.fneg(<vscale x 8 x half> poison, <vscale x 8 x i1> poison, i32 poison)
+  %NXV16F16_VP = call <vscale x 16 x half> @llvm.vp.fneg(<vscale x 16 x half> poison, <vscale x 16 x i1> poison, i32 poison)
 
   ret void
 }
 
 define void @fcopysign() {
 ; CHECK-LABEL: 'fcopysign'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %BF16 = call bfloat @llvm.copysign.bf16(bfloat undef, bfloat undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = call float @llvm.copysign.f32(float undef, float undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = call double @llvm.copysign.f64(double undef, double undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V1BF16 = call <1 x bfloat> @llvm.copysign.v1bf16(<1 x bfloat> undef, <1 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2BF16 = call <2 x bfloat> @llvm.copysign.v2bf16(<2 x bfloat> undef, <2 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4BF16 = call <4 x bfloat> @llvm.copysign.v4bf16(<4 x bfloat> undef, <4 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8BF16 = call <8 x bfloat> @llvm.copysign.v8bf16(<8 x bfloat> undef, <8 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 47 for instruction: %V16BF16 = call <16 x bfloat> @llvm.copysign.v16bf16(<16 x bfloat> undef, <16 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16 = call <vscale x 1 x bfloat> @llvm.copysign.nxv1bf16(<vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16 = call <vscale x 2 x bfloat> @llvm.copysign.nxv2bf16(<vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16 = call <vscale x 4 x bfloat> @llvm.copysign.nxv4bf16(<vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16 = call <vscale x 8 x bfloat> @llvm.copysign.nxv8bf16(<vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16 = call <vscale x 16 x bfloat> @llvm.copysign.nxv16bf16(<vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32 = call <1 x float> @llvm.copysign.v1f32(<1 x float> undef, <1 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32 = call <2 x float> @llvm.copysign.v2f32(<2 x float> undef, <2 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.copysign.v4f32(<4 x float> undef, <4 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = call <8 x float> @llvm.copysign.v8f32(<8 x float> undef, <8 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = call <16 x float> @llvm.copysign.v16f32(<16 x float> undef, <16 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV1F32 = call <vscale x 1 x float> @llvm.copysign.nxv1f32(<vscale x 1 x float> undef, <vscale x 1 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV2F32 = call <vscale x 2 x float> @llvm.copysign.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV4F32 = call <vscale x 4 x float> @llvm.copysign.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV8F32 = call <vscale x 8 x float> @llvm.copysign.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV16F32 = call <vscale x 16 x float> @llvm.copysign.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64 = call <1 x double> @llvm.copysign.v1f64(<1 x double> undef, <1 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.copysign.v2f64(<2 x double> undef, <2 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = call <4 x double> @llvm.copysign.v4f64(<4 x double> undef, <4 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = call <8 x double> @llvm.copysign.v8f64(<8 x double> undef, <8 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV1F64 = call <vscale x 1 x double> @llvm.copysign.nxv1f64(<vscale x 1 x double> undef, <vscale x 1 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV2F64 = call <vscale x 2 x double> @llvm.copysign.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV4F64 = call <vscale x 4 x double> @llvm.copysign.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV8F64 = call <vscale x 8 x double> @llvm.copysign.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x double> undef)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = call float @llvm.copysign.f32(float poison, float poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = call double @llvm.copysign.f64(double poison, double poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F32 = call <1 x float> @llvm.copysign.v1f32(<1 x float> poison, <1 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F32 = call <2 x float> @llvm.copysign.v2f32(<2 x float> poison, <2 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F32 = call <4 x float> @llvm.copysign.v4f32(<4 x float> poison, <4 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = call <8 x float> @llvm.copysign.v8f32(<8 x float> poison, <8 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16F32 = call <16 x float> @llvm.copysign.v16f32(<16 x float> poison, <16 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV1F32 = call <vscale x 1 x float> @llvm.copysign.nxv1f32(<vscale x 1 x float> poison, <vscale x 1 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV2F32 = call <vscale x 2 x float> @llvm.copysign.nxv2f32(<vscale x 2 x float> poison, <vscale x 2 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV4F32 = call <vscale x 4 x float> @llvm.copysign.nxv4f32(<vscale x 4 x float> poison, <vscale x 4 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV8F32 = call <vscale x 8 x float> @llvm.copysign.nxv8f32(<vscale x 8 x float> poison, <vscale x 8 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV16F32 = call <vscale x 16 x float> @llvm.copysign.nxv16f32(<vscale x 16 x float> poison, <vscale x 16 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F64 = call <1 x double> @llvm.copysign.v1f64(<1 x double> poison, <1 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F64 = call <2 x double> @llvm.copysign.v2f64(<2 x double> poison, <2 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = call <4 x double> @llvm.copysign.v4f64(<4 x double> poison, <4 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F64 = call <8 x double> @llvm.copysign.v8f64(<8 x double> poison, <8 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV1F64 = call <vscale x 1 x double> @llvm.copysign.nxv1f64(<vscale x 1 x double> poison, <vscale x 1 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV2F64 = call <vscale x 2 x double> @llvm.copysign.nxv2f64(<vscale x 2 x double> poison, <vscale x 2 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV4F64 = call <vscale x 4 x double> @llvm.copysign.nxv4f64(<vscale x 4 x double> poison, <vscale x 4 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV8F64 = call <vscale x 8 x double> @llvm.copysign.nxv8f64(<vscale x 8 x double> poison, <vscale x 8 x double> poison)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
-  %BF16 = call bfloat @llvm.copysign.bf16(bfloat undef, bfloat undef)
-  %F32 = call float @llvm.copysign.f32(float undef, float undef)
-  %F64 = call double @llvm.copysign.f64(double undef, double undef)
-
-  %V1BF16 = call <1 x bfloat> @llvm.copysign.v1bf16(<1 x bfloat> undef, <1 x bfloat> undef)
-  %V2BF16 = call <2 x bfloat> @llvm.copysign.v2bf16(<2 x bfloat> undef, <2 x bfloat> undef)
-  %V4BF16 = call <4 x bfloat> @llvm.copysign.v4bf16(<4 x bfloat> undef, <4 x bfloat> undef)
-  %V8BF16 = call <8 x bfloat> @llvm.copysign.v8bf16(<8 x bfloat> undef, <8 x bfloat> undef)
-  %V16BF16 = call <16 x bfloat> @llvm.copysign.v16bf16(<16 x bfloat> undef, <16 x bfloat> undef)
-
-  %NXV1BF16 = call <vscale x 1 x bfloat> @llvm.copysign.nxv1bf16(<vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef)
-  %NXV2BF16 = call <vscale x 2 x bfloat> @llvm.copysign.nxv2bf16(<vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef)
-  %NXV4BF16 = call <vscale x 4 x bfloat> @llvm.copysign.nxv4bf16(<vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef)
-  %NXV8BF16 = call <vscale x 8 x bfloat> @llvm.copysign.nxv8bf16(<vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef)
-  %NXV16BF16 = call <vscale x 16 x bfloat> @llvm.copysign.nxv16bf16(<vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef)
-
-  %V1F32 = call <1 x float> @llvm.copysign.v1f32(<1 x float> undef, <1 x float> undef)
-  %V2F32 = call <2 x float> @llvm.copysign.v2f32(<2 x float> undef, <2 x float> undef)
-  %V4F32 = call <4 x float> @llvm.copysign.v4f32(<4 x float> undef, <4 x float> undef)
-  %V8F32 = call <8 x float> @llvm.copysign.v8f32(<8 x float> undef, <8 x float> undef)
-  %V16F32 = call <16 x float> @llvm.copysign.v16f32(<16 x float> undef, <16 x float> undef)
-
-  %NXV1F32 = call <vscale x 1 x float> @llvm.copysign.nxv1f32(<vscale x 1 x float> undef, <vscale x 1 x float> undef)
-  %NXV2F32 = call <vscale x 2 x float> @llvm.copysign.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> undef)
-  %NXV4F32 = call <vscale x 4 x float> @llvm.copysign.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x float> undef)
-  %NXV8F32 = call <vscale x 8 x float> @llvm.copysign.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x float> undef)
-  %NXV16F32 = call <vscale x 16 x float> @llvm.copysign.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x float> undef)
-
-  %V1F64 = call <1 x double> @llvm.copysign.v1f64(<1 x double> undef, <1 x double> undef)
-  %V2F64 = call <2 x double> @llvm.copysign.v2f64(<2 x double> undef, <2 x double> undef)
-  %V4F64 = call <4 x double> @llvm.copysign.v4f64(<4 x double> undef, <4 x double> undef)
-  %V8F64 = call <8 x double> @llvm.copysign.v8f64(<8 x double> undef, <8 x double> undef)
-
-  %NXV1F64 = call <vscale x 1 x double> @llvm.copysign.nxv1f64(<vscale x 1 x double> undef, <vscale x 1 x double> undef)
-  %NXV2F64 = call <vscale x 2 x double> @llvm.copysign.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x double> undef)
-  %NXV4F64 = call <vscale x 4 x double> @llvm.copysign.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x double> undef)
-  %NXV8F64 = call <vscale x 8 x double> @llvm.copysign.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x double> undef)
+  %F32 = call float @llvm.copysign.f32(float poison, float poison)
+  %F64 = call double @llvm.copysign.f64(double poison, double poison)
+
+  %V1F32 = call <1 x float> @llvm.copysign(<1 x float> poison, <1 x float> poison)
+  %V2F32 = call <2 x float> @llvm.copysign(<2 x float> poison, <2 x float> poison)
+  %V4F32 = call <4 x float> @llvm.copysign(<4 x float> poison, <4 x float> poison)
+  %V8F32 = call <8 x float> @llvm.copysign(<8 x float> poison, <8 x float> poison)
+  %V16F32 = call <16 x float> @llvm.copysign(<16 x float> poison, <16 x float> poison)
+
+  %NXV1F32 = call <vscale x 1 x float> @llvm.copysign(<vscale x 1 x float> poison, <vscale x 1 x float> poison)
+  %NXV2F32 = call <vscale x 2 x float> @llvm.copysign(<vscale x 2 x float> poison, <vscale x 2 x float> poison)
+  %NXV4F32 = call <vscale x 4 x float> @llvm.copysign(<vscale x 4 x float> poison, <vscale x 4 x float> poison)
+  %NXV8F32 = call <vscale x 8 x float> @llvm.copysign(<vscale x 8 x float> poison, <vscale x 8 x float> poison)
+  %NXV16F32 = call <vscale x 16 x float> @llvm.copysign(<vscale x 16 x float> poison, <vscale x 16 x float> poison)
+
+  %V1F64 = call <1 x double> @llvm.copysign(<1 x double> poison, <1 x double> poison)
+  %V2F64 = call <2 x double> @llvm.copysign(<2 x double> poison, <2 x double> poison)
+  %V4F64 = call <4 x double> @llvm.copysign(<4 x double> poison, <4 x double> poison)
+  %V8F64 = call <8 x double> @llvm.copysign(<8 x double> poison, <8 x double> poison)
+
+  %NXV1F64 = call <vscale x 1 x double> @llvm.copysign(<vscale x 1 x double> poison, <vscale x 1 x double> poison)
+  %NXV2F64 = call <vscale x 2 x double> @llvm.copysign(<vscale x 2 x double> poison, <vscale x 2 x double> poison)
+  %NXV4F64 = call <vscale x 4 x double> @llvm.copysign(<vscale x 4 x double> poison, <vscale x 4 x double> poison)
+  %NXV8F64 = call <vscale x 8 x double> @llvm.copysign(<vscale x 8 x double> poison, <vscale x 8 x double> poison)
+
+  ret void
+}
+
+define void @fcopysign_bf16() {
+; ZVFH-LABEL: 'fcopysign_bf16'
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %BF16 = call bfloat @llvm.copysign.bf16(bfloat poison, bfloat poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V1BF16 = call <1 x bfloat> @llvm.copysign.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2BF16 = call <2 x bfloat> @llvm.copysign.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4BF16 = call <4 x bfloat> @llvm.copysign.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8BF16 = call <8 x bfloat> @llvm.copysign.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 47 for instruction: %V16BF16 = call <16 x bfloat> @llvm.copysign.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 95 for instruction: %V32BF16 = call <32 x bfloat> @llvm.copysign.v32bf16(<32 x bfloat> poison, <32 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16 = call <vscale x 1 x bfloat> @llvm.copysign.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16 = call <vscale x 2 x bfloat> @llvm.copysign.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16 = call <vscale x 4 x bfloat> @llvm.copysign.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16 = call <vscale x 8 x bfloat> @llvm.copysign.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16 = call <vscale x 16 x bfloat> @llvm.copysign.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Invalid cost for instruction: %NXV32BF16 = call <vscale x 32 x bfloat> @llvm.copysign.nxv32bf16(<vscale x 32 x bfloat> poison, <vscale x 32 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; ZVFHMIN-LABEL: 'fcopysign_bf16'
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %BF16 = call bfloat @llvm.copysign.bf16(bfloat poison, bfloat poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V1BF16 = call <1 x bfloat> @llvm.copysign.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2BF16 = call <2 x bfloat> @llvm.copysign.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4BF16 = call <4 x bfloat> @llvm.copysign.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8BF16 = call <8 x bfloat> @llvm.copysign.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 47 for instruction: %V16BF16 = call <16 x bfloat> @llvm.copysign.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 95 for instruction: %V32BF16 = call <32 x bfloat> @llvm.copysign.v32bf16(<32 x bfloat> poison, <32 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16 = call <vscale x 1 x bfloat> @llvm.copysign.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16 = call <vscale x 2 x bfloat> @llvm.copysign.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16 = call <vscale x 4 x bfloat> @llvm.copysign.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16 = call <vscale x 8 x bfloat> @llvm.copysign.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16 = call <vscale x 16 x bfloat> @llvm.copysign.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32BF16 = call <vscale x 32 x bfloat> @llvm.copysign.nxv32bf16(<vscale x 32 x bfloat> poison, <vscale x 32 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; NO-ZFHMIN-LABEL: 'fcopysign_bf16'
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %BF16 = call bfloat @llvm.copysign.bf16(bfloat poison, bfloat poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1BF16 = call <1 x bfloat> @llvm.copysign.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2BF16 = call <2 x bfloat> @llvm.copysign.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4BF16 = call <4 x bfloat> @llvm.copysign.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8BF16 = call <8 x bfloat> @llvm.copysign.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16BF16 = call <16 x bfloat> @llvm.copysign.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V32BF16 = call <32 x bfloat> @llvm.copysign.v32bf16(<32 x bfloat> poison, <32 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16 = call <vscale x 1 x bfloat> @llvm.copysign.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16 = call <vscale x 2 x bfloat> @llvm.copysign.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16 = call <vscale x 4 x bfloat> @llvm.copysign.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16 = call <vscale x 8 x bfloat> @llvm.copysign.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16 = call <vscale x 16 x bfloat> @llvm.copysign.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32BF16 = call <vscale x 32 x bfloat> @llvm.copysign.nxv32bf16(<vscale x 32 x bfloat> poison, <vscale x 32 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+  %BF16 = call bfloat @llvm.copysign(bfloat poison, bfloat poison)
+
+  %V1BF16 = call <1 x bfloat> @llvm.copysign(<1 x bfloat> poison, <1 x bfloat> poison)
+  %V2BF16 = call <2 x bfloat> @llvm.copysign(<2 x bfloat> poison, <2 x bfloat> poison)
+  %V4BF16 = call <4 x bfloat> @llvm.copysign(<4 x bfloat> poison, <4 x bfloat> poison)
+  %V8BF16 = call <8 x bfloat> @llvm.copysign(<8 x bfloat> poison, <8 x bfloat> poison)
+  %V16BF16 = call <16 x bfloat> @llvm.copysign(<16 x bfloat> poison, <16 x bfloat> poison)
+  %V32BF16 = call <32 x bfloat> @llvm.copysign(<32 x bfloat> poison, <32 x bfloat> poison)
+
+  %NXV1BF16 = call <vscale x 1 x bfloat> @llvm.copysign(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison)
+  %NXV2BF16 = call <vscale x 2 x bfloat> @llvm.copysign(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison)
+  %NXV4BF16 = call <vscale x 4 x bfloat> @llvm.copysign(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison)
+  %NXV8BF16 = call <vscale x 8 x bfloat> @llvm.copysign(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison)
+  %NXV16BF16 = call <vscale x 16 x bfloat> @llvm.copysign(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison)
+  %NXV32BF16 = call <vscale x 32 x bfloat> @llvm.copysign(<vscale x 32 x bfloat> poison, <vscale x 32 x bfloat> poison)
 
   ret void
 }
 
 define void @fcopysign_f16() {
 ; ZVFH-LABEL: 'fcopysign_f16'
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F16 = call half @llvm.copysign.f16(half undef, half undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16 = call <1 x half> @llvm.copysign.v1f16(<1 x half> undef, <1 x half> undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16 = call <2 x half> @llvm.copysign.v2f16(<2 x half> undef, <2 x half> undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16 = call <4 x half> @llvm.copysign.v4f16(<4 x half> undef, <4 x half> undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16 = call <8 x half> @llvm.copysign.v8f16(<8 x half> undef, <8 x half> undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16F16 = call <16 x half> @llvm.copysign.v16f16(<16 x half> undef, <16 x half> undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32F16 = call <32 x half> @llvm.copysign.v32f16(<32 x half> undef, <32 x half> undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV1F16 = call <vscale x 1 x half> @llvm.copysign.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV2F16 = call <vscale x 2 x half> @llvm.copysign.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV4F16 = call <vscale x 4 x half> @llvm.copysign.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV8F16 = call <vscale x 8 x half> @llvm.copysign.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV16F16 = call <vscale x 16 x half> @llvm.copysign.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef)
-; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV32F16 = call <vscale x 32 x half> @llvm.copysign.nxv32f16(<vscale x 32 x half> undef, <vscale x 32 x half> undef)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F16 = call half @llvm.copysign.f16(half poison, half poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V1F16 = call <1 x half> @llvm.copysign.v1f16(<1 x half> poison, <1 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V2F16 = call <2 x half> @llvm.copysign.v2f16(<2 x half> poison, <2 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F16 = call <4 x half> @llvm.copysign.v4f16(<4 x half> poison, <4 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F16 = call <8 x half> @llvm.copysign.v8f16(<8 x half> poison, <8 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16F16 = call <16 x half> @llvm.copysign.v16f16(<16 x half> poison, <16 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V32F16 = call <32 x half> @llvm.copysign.v32f16(<32 x half> poison, <32 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV1F16 = call <vscale x 1 x half> @llvm.copysign.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV2F16 = call <vscale x 2 x half> @llvm.copysign.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV4F16 = call <vscale x 4 x half> @llvm.copysign.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV8F16 = call <vscale x 8 x half> @llvm.copysign.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV16F16 = call <vscale x 16 x half> @llvm.copysign.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV32F16 = call <vscale x 32 x half> @llvm.copysign.nxv32f16(<vscale x 32 x half> poison, <vscale x 32 x half> poison)
 ; ZVFH-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
 ; ZVFHMIN-LABEL: 'fcopysign_f16'
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F16 = call half @llvm.copysign.f16(half undef, half undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V1F16 = call <1 x half> @llvm.copysign.v1f16(<1 x half> undef, <1 x half> undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2F16 = call <2 x half> @llvm.copysign.v2f16(<2 x half> undef, <2 x half> undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4F16 = call <4 x half> @llvm.copysign.v4f16(<4 x half> undef, <4 x half> undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8F16 = call <8 x half> @llvm.copysign.v8f16(<8 x half> undef, <8 x half> undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 47 for instruction: %V16F16 = call <16 x half> @llvm.copysign.v16f16(<16 x half> undef, <16 x half> undef)
-; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 95 for instruction: %V32F16 = call <32 x half> @llvm.copysign.v32f16(<32 x half> undef, <32 x half> undef)
-; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16 = call <vscale x 1 x half> @llvm.copysign.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef)
-; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16 = call <vscale x 2 x half> @llvm.copysign.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef)
-; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16 = call <vscale x 4 x half> @llvm.copysign.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef)
-; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16 = call <vscale x 8 x half> @llvm.copysign.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef)
-; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16 = call <vscale x 16 x half> @llvm.copysign.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef)
-; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32F16 = call <vscale x 32 x half> @llvm.copysign.nxv32f16(<vscale x 32 x half> undef, <vscale x 32 x half> undef)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F16 = call half @llvm.copysign.f16(half poison, half poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V1F16 = call <1 x half> @llvm.copysign.v1f16(<1 x half> poison, <1 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2F16 = call <2 x half> @llvm.copysign.v2f16(<2 x half> poison, <2 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4F16 = call <4 x half> @llvm.copysign.v4f16(<4 x half> poison, <4 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8F16 = call <8 x half> @llvm.copysign.v8f16(<8 x half> poison, <8 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 47 for instruction: %V16F16 = call <16 x half> @llvm.copysign.v16f16(<16 x half> poison, <16 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 95 for instruction: %V32F16 = call <32 x half> @llvm.copysign.v32f16(<32 x half> poison, <32 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16 = call <vscale x 1 x half> @llvm.copysign.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16 = call <vscale x 2 x half> @llvm.copysign.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16 = call <vscale x 4 x half> @llvm.copysign.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16 = call <vscale x 8 x half> @llvm.copysign.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16 = call <vscale x 16 x half> @llvm.copysign.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32F16 = call <vscale x 32 x half> @llvm.copysign.nxv32f16(<vscale x 32 x half> poison, <vscale x 32 x half> poison)
 ; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
-  %F16 = call half @llvm.copysign.f16(half undef, half undef)
-
-  %V1F16 = call <1 x half> @llvm.copysign.v1f16(<1 x half> undef, <1 x half> undef)
-  %V2F16 = call <2 x half> @llvm.copysign.v2f16(<2 x half> undef, <2 x half> undef)
-  %V4F16 = call <4 x half> @llvm.copysign.v4f16(<4 x half> undef, <4 x half> undef)
-  %V8F16 = call <8 x half> @llvm.copysign.v8f16(<8 x half> undef, <8 x half> undef)
-  %V16F16 = call <16 x half> @llvm.copysign.v16f16(<16 x half> undef, <16 x half> undef)
-  %V32F16 = call <32 x half> @llvm.copysign.v32f16(<32 x half> undef, <32 x half> undef)
-
-  %NXV1F16 = call <vscale x 1 x half> @llvm.copysign.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef)
-  %NXV2F16 = call <vscale x 2 x half> @llvm.copysign.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef)
-  %NXV4F16 = call <vscale x 4 x half> @llvm.copysign.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef)
-  %NXV8F16 = call <vscale x 8 x half> @llvm.copysign.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef)
-  %NXV16F16 = call <vscale x 16 x half> @llvm.copysign.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef)
-  %NXV32F16 = call <vscale x 32 x half> @llvm.copysign.nxv32f16(<vscale x 32 x half> undef, <vscale x 32 x half> undef)
+; NO-ZFHMIN-LABEL: 'fcopysign_f16'
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F16 = call half @llvm.copysign.f16(half poison, half poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1F16 = call <1 x half> @llvm.copysign.v1f16(<1 x half> poison, <1 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2F16 = call <2 x half> @llvm.copysign.v2f16(<2 x half> poison, <2 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4F16 = call <4 x half> @llvm.copysign.v4f16(<4 x half> poison, <4 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8F16 = call <8 x half> @llvm.copysign.v8f16(<8 x half> poison, <8 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16F16 = call <16 x half> @llvm.copysign.v16f16(<16 x half> poison, <16 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V32F16 = call <32 x half> @llvm.copysign.v32f16(<32 x half> poison, <32 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16 = call <vscale x 1 x half> @llvm.copysign.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16 = call <vscale x 2 x half> @llvm.copysign.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16 = call <vscale x 4 x half> @llvm.copysign.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16 = call <vscale x 8 x half> @llvm.copysign.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16 = call <vscale x 16 x half> @llvm.copysign.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32F16 = call <vscale x 32 x half> @llvm.copysign.nxv32f16(<vscale x 32 x half> poison, <vscale x 32 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+  %F16 = call half @llvm.copysign(half poison, half poison)
+
+  %V1F16 = call <1 x half> @llvm.copysign(<1 x half> poison, <1 x half> poison)
+  %V2F16 = call <2 x half> @llvm.copysign(<2 x half> poison, <2 x half> poison)
+  %V4F16 = call <4 x half> @llvm.copysign(<4 x half> poison, <4 x half> poison)
+  %V8F16 = call <8 x half> @llvm.copysign(<8 x half> poison, <8 x half> poison)
+  %V16F16 = call <16 x half> @llvm.copysign(<16 x half> poison, <16 x half> poison)
+  %V32F16 = call <32 x half> @llvm.copysign(<32 x half> poison, <32 x half> poison)
+
+  %NXV1F16 = call <vscale x 1 x half> @llvm.copysign(<vscale x 1 x half> poison, <vscale x 1 x half> poison)
+  %NXV2F16 = call <vscale x 2 x half> @llvm.copysign(<vscale x 2 x half> poison, <vscale x 2 x half> poison)
+  %NXV4F16 = call <vscale x 4 x half> @llvm.copysign(<vscale x 4 x half> poison, <vscale x 4 x half> poison)
+  %NXV8F16 = call <vscale x 8 x half> @llvm.copysign(<vscale x 8 x half> poison, <vscale x 8 x half> poison)
+  %NXV16F16 = call <vscale x 16 x half> @llvm.copysign(<vscale x 16 x half> poison, <vscale x 16 x half> poison)
+  %NXV32F16 = call <vscale x 32 x half> @llvm.copysign(<vscale x 32 x half> poison, <vscale x 32 x half> poison)
 
   ret void
 }
 
 define void @fma() {
 ; CHECK-LABEL: 'fma'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %BF16 = call bfloat @llvm.fma.bf16(bfloat undef, bfloat undef, bfloat undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1BF16 = call <1 x bfloat> @llvm.fma.v1bf16(<1 x bfloat> undef, <1 x bfloat> undef, <1 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2BF16 = call <2 x bfloat> @llvm.fma.v2bf16(<2 x bfloat> undef, <2 x bfloat> undef, <2 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4BF16 = call <4 x bfloat> @llvm.fma.v4bf16(<4 x bfloat> undef, <4 x bfloat> undef, <4 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8BF16 = call <8 x bfloat> @llvm.fma.v8bf16(<8 x bfloat> undef, <8 x bfloat> undef, <8 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16BF16 = call <16 x bfloat> @llvm.fma.v16bf16(<16 x bfloat> undef, <16 x bfloat> undef, <16 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV1BF16 = call <vscale x 1 x bfloat> @llvm.fma.nxv1bf16(<vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV2BF16 = call <vscale x 2 x bfloat> @llvm.fma.nxv2bf16(<vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV4BF16 = call <vscale x 4 x bfloat> @llvm.fma.nxv4bf16(<vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV8BF16 = call <vscale x 8 x bfloat> @llvm.fma.nxv8bf16(<vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV16BF16 = call <vscale x 16 x bfloat> @llvm.fma.nxv16bf16(<vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1F32 = call <1 x float> @llvm.fma.v1f32(<1 x float> undef, <1 x float> undef, <1 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F32 = call <2 x float> @llvm.fma.v2f32(<2 x float> undef, <2 x float> undef, <2 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = call <4 x float> @llvm.fma.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV1F32 = call <vscale x 1 x float> @llvm.fma.nxv1f32(<vscale x 1 x float> undef, <vscale x 1 x float> undef, <vscale x 1 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV2F32 = call <vscale x 2 x float> @llvm.fma.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> undef, <vscale x 2 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F32 = call <vscale x 4 x float> @llvm.fma.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x float> undef, <vscale x 4 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F32 = call <vscale x 8 x float> @llvm.fma.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x float> undef, <vscale x 8 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F32 = call <vscale x 16 x float> @llvm.fma.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x float> undef, <vscale x 16 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1F64 = call <1 x double> @llvm.fma.v1f64(<1 x double> undef, <1 x double> undef, <1 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = call <2 x double> @llvm.fma.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = call <4 x double> @llvm.fma.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = call <8 x double> @llvm.fma.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV1F64 = call <vscale x 1 x double> @llvm.fma.nxv1f64(<vscale x 1 x double> undef, <vscale x 1 x double> undef, <vscale x 1 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F64 = call <vscale x 2 x double> @llvm.fma.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x double> undef, <vscale x 2 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F64 = call <vscale x 4 x double> @llvm.fma.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x double> undef, <vscale x 4 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F64 = call <vscale x 8 x double> @llvm.fma.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x double> undef, <vscale x 8 x double> undef)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F32 = call float @llvm.fma.f32(float poison, float poison, float poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F64 = call double @llvm.fma.f64(double poison, double poison, double poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1F32 = call <1 x float> @llvm.fma.v1f32(<1 x float> poison, <1 x float> poison, <1 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F32 = call <2 x float> @llvm.fma.v2f32(<2 x float> poison, <2 x float> poison, <2 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F32 = call <4 x float> @llvm.fma.v4f32(<4 x float> poison, <4 x float> poison, <4 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> poison, <8 x float> poison, <8 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> poison, <16 x float> poison, <16 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV1F32 = call <vscale x 1 x float> @llvm.fma.nxv1f32(<vscale x 1 x float> poison, <vscale x 1 x float> poison, <vscale x 1 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV2F32 = call <vscale x 2 x float> @llvm.fma.nxv2f32(<vscale x 2 x float> poison, <vscale x 2 x float> poison, <vscale x 2 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV4F32 = call <vscale x 4 x float> @llvm.fma.nxv4f32(<vscale x 4 x float> poison, <vscale x 4 x float> poison, <vscale x 4 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV8F32 = call <vscale x 8 x float> @llvm.fma.nxv8f32(<vscale x 8 x float> poison, <vscale x 8 x float> poison, <vscale x 8 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV16F32 = call <vscale x 16 x float> @llvm.fma.nxv16f32(<vscale x 16 x float> poison, <vscale x 16 x float> poison, <vscale x 16 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1F64 = call <1 x double> @llvm.fma.v1f64(<1 x double> poison, <1 x double> poison, <1 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F64 = call <2 x double> @llvm.fma.v2f64(<2 x double> poison, <2 x double> poison, <2 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V4F64 = call <4 x double> @llvm.fma.v4f64(<4 x double> poison, <4 x double> poison, <4 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V8F64 = call <8 x double> @llvm.fma.v8f64(<8 x double> poison, <8 x double> poison, <8 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV1F64 = call <vscale x 1 x double> @llvm.fma.nxv1f64(<vscale x 1 x double> poison, <vscale x 1 x double> poison, <vscale x 1 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV2F64 = call <vscale x 2 x double> @llvm.fma.nxv2f64(<vscale x 2 x double> poison, <vscale x 2 x double> poison, <vscale x 2 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV4F64 = call <vscale x 4 x double> @llvm.fma.nxv4f64(<vscale x 4 x double> poison, <vscale x 4 x double> poison, <vscale x 4 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV8F64 = call <vscale x 8 x double> @llvm.fma.nxv8f64(<vscale x 8 x double> poison, <vscale x 8 x double> poison, <vscale x 8 x double> poison)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
-  %BF16 = call bfloat @llvm.fma.bf16(bfloat undef, bfloat undef, bfloat undef)
-  %F32 = call float @llvm.fma.f32(float undef, float undef, float undef)
-  %F64 = call double @llvm.fma.f64(double undef, double undef, double undef)
-
-  %V1BF16 = call <1 x bfloat> @llvm.fma.v1bf16(<1 x bfloat> undef, <1 x bfloat> undef, <1 x bfloat> undef)
-  %V2BF16 = call <2 x bfloat> @llvm.fma.v2bf16(<2 x bfloat> undef, <2 x bfloat> undef, <2 x bfloat> undef)
-  %V4BF16 = call <4 x bfloat> @llvm.fma.v4bf16(<4 x bfloat> undef, <4 x bfloat> undef, <4 x bfloat> undef)
-  %V8BF16 = call <8 x bfloat> @llvm.fma.v8bf16(<8 x bfloat> undef, <8 x bfloat> undef, <8 x bfloat> undef)
-  %V16BF16 = call <16 x bfloat> @llvm.fma.v16bf16(<16 x bfloat> undef, <16 x bfloat> undef, <16 x bfloat> undef)
-
-  %NXV1BF16 = call <vscale x 1 x bfloat> @llvm.fma.nxv1bf16(<vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef)
-  %NXV2BF16 = call <vscale x 2 x bfloat> @llvm.fma.nxv2bf16(<vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef)
-  %NXV4BF16 = call <vscale x 4 x bfloat> @llvm.fma.nxv4bf16(<vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef)
-  %NXV8BF16 = call <vscale x 8 x bfloat> @llvm.fma.nxv8bf16(<vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef)
-  %NXV16BF16 = call <vscale x 16 x bfloat> @llvm.fma.nxv16bf16(<vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef)
-
-  %V1F32 = call <1 x float> @llvm.fma.v1f32(<1 x float> undef, <1 x float> undef, <1 x float> undef)
-  %V2F32 = call <2 x float> @llvm.fma.v2f32(<2 x float> undef, <2 x float> undef, <2 x float> undef)
-  %V4F32 = call <4 x float> @llvm.fma.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
-  %V8F32 = call <8 x float> @llvm.fma.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
-  %V16F32 = call <16 x float> @llvm.fma.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
-
-  %NXV1F32 = call <vscale x 1 x float> @llvm.fma.nxv1f32(<vscale x 1 x float> undef, <vscale x 1 x float> undef, <vscale x 1 x float> undef)
-  %NXV2F32 = call <vscale x 2 x float> @llvm.fma.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> undef, <vscale x 2 x float> undef)
-  %NXV4F32 = call <vscale x 4 x float> @llvm.fma.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x float> undef, <vscale x 4 x float> undef)
-  %NXV8F32 = call <vscale x 8 x float> @llvm.fma.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x float> undef, <vscale x 8 x float> undef)
-  %NXV16F32 = call <vscale x 16 x float> @llvm.fma.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x float> undef, <vscale x 16 x float> undef)
-
-  %V1F64 = call <1 x double> @llvm.fma.v1f64(<1 x double> undef, <1 x double> undef, <1 x double> undef)
-  %V2F64 = call <2 x double> @llvm.fma.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
-  %V4F64 = call <4 x double> @llvm.fma.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
-  %V8F64 = call <8 x double> @llvm.fma.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
-
-  %NXV1F64 = call <vscale x 1 x double> @llvm.fma.nxv1f64(<vscale x 1 x double> undef, <vscale x 1 x double> undef, <vscale x 1 x double> undef)
-  %NXV2F64 = call <vscale x 2 x double> @llvm.fma.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x double> undef, <vscale x 2 x double> undef)
-  %NXV4F64 = call <vscale x 4 x double> @llvm.fma.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x double> undef, <vscale x 4 x double> undef)
-  %NXV8F64 = call <vscale x 8 x double> @llvm.fma.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x double> undef, <vscale x 8 x double> undef)
+  %F32 = call float @llvm.fma(float poison, float poison, float poison)
+  %F64 = call double @llvm.fma(double poison, double poison, double poison)
+
+  %V1F32 = call <1 x float> @llvm.fma(<1 x float> poison, <1 x float> poison, <1 x float> poison)
+  %V2F32 = call <2 x float> @llvm.fma(<2 x float> poison, <2 x float> poison, <2 x float> poison)
+  %V4F32 = call <4 x float> @llvm.fma(<4 x float> poison, <4 x float> poison, <4 x float> poison)
+  %V8F32 = call <8 x float> @llvm.fma(<8 x float> poison, <8 x float> poison, <8 x float> poison)
+  %V16F32 = call <16 x float> @llvm.fma(<16 x float> poison, <16 x float> poison, <16 x float> poison)
+
+  %NXV1F32 = call <vscale x 1 x float> @llvm.fma(<vscale x 1 x float> poison, <vscale x 1 x float> poison, <vscale x 1 x float> poison)
+  %NXV2F32 = call <vscale x 2 x float> @llvm.fma(<vscale x 2 x float> poison, <vscale x 2 x float> poison, <vscale x 2 x float> poison)
+  %NXV4F32 = call <vscale x 4 x float> @llvm.fma(<vscale x 4 x float> poison, <vscale x 4 x float> poison, <vscale x 4 x float> poison)
+  %NXV8F32 = call <vscale x 8 x float> @llvm.fma(<vscale x 8 x float> poison, <vscale x 8 x float> poison, <vscale x 8 x float> poison)
+  %NXV16F32 = call <vscale x 16 x float> @llvm.fma(<vscale x 16 x float> poison, <vscale x 16 x float> poison, <vscale x 16 x float> poison)
+
+  %V1F64 = call <1 x double> @llvm.fma(<1 x double> poison, <1 x double> poison, <1 x double> poison)
+  %V2F64 = call <2 x double> @llvm.fma(<2 x double> poison, <2 x double> poison, <2 x double> poison)
+  %V4F64 = call <4 x double> @llvm.fma(<4 x double> poison, <4 x double> poison, <4 x double> poison)
+  %V8F64 = call <8 x double> @llvm.fma(<8 x double> poison, <8 x double> poison, <8 x double> poison)
+
+  %NXV1F64 = call <vscale x 1 x double> @llvm.fma(<vscale x 1 x double> poison, <vscale x 1 x double> poison, <vscale x 1 x double> poison)
+  %NXV2F64 = call <vscale x 2 x double> @llvm.fma(<vscale x 2 x double> poison, <vscale x 2 x double> poison, <vscale x 2 x double> poison)
+  %NXV4F64 = call <vscale x 4 x double> @llvm.fma(<vscale x 4 x double> poison, <vscale x 4 x double> poison, <vscale x 4 x double> poison)
+  %NXV8F64 = call <vscale x 8 x double> @llvm.fma(<vscale x 8 x double> poison, <vscale x 8 x double> poison, <vscale x 8 x double> poison)
+
+  ret void
+}
+
+define void @fma_bf16() {
+; ZVFH-LABEL: 'fma_bf16'
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %BF16 = call bfloat @llvm.fma.bf16(bfloat poison, bfloat poison, bfloat poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1BF16 = call <1 x bfloat> @llvm.fma.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison, <1 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2BF16 = call <2 x bfloat> @llvm.fma.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4BF16 = call <4 x bfloat> @llvm.fma.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8BF16 = call <8 x bfloat> @llvm.fma.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16BF16 = call <16 x bfloat> @llvm.fma.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32BF16 = call <32 x bfloat> @llvm.fma.v32bf16(<32 x bfloat> poison, <32 x bfloat> poison, <32 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV1BF16 = call <vscale x 1 x bfloat> @llvm.fma.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV2BF16 = call <vscale x 2 x bfloat> @llvm.fma.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV4BF16 = call <vscale x 4 x bfloat> @llvm.fma.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV8BF16 = call <vscale x 8 x bfloat> @llvm.fma.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV16BF16 = call <vscale x 16 x bfloat> @llvm.fma.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV32BF16 = call <vscale x 32 x bfloat> @llvm.fma.nxv32bf16(<vscale x 32 x bfloat> poison, <vscale x 32 x bfloat> poison, <vscale x 32 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; ZVFHMIN-LABEL: 'fma_bf16'
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %BF16 = call bfloat @llvm.fma.bf16(bfloat poison, bfloat poison, bfloat poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1BF16 = call <1 x bfloat> @llvm.fma.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison, <1 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2BF16 = call <2 x bfloat> @llvm.fma.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4BF16 = call <4 x bfloat> @llvm.fma.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8BF16 = call <8 x bfloat> @llvm.fma.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16BF16 = call <16 x bfloat> @llvm.fma.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32BF16 = call <32 x bfloat> @llvm.fma.v32bf16(<32 x bfloat> poison, <32 x bfloat> poison, <32 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV1BF16 = call <vscale x 1 x bfloat> @llvm.fma.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV2BF16 = call <vscale x 2 x bfloat> @llvm.fma.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV4BF16 = call <vscale x 4 x bfloat> @llvm.fma.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV8BF16 = call <vscale x 8 x bfloat> @llvm.fma.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV16BF16 = call <vscale x 16 x bfloat> @llvm.fma.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV32BF16 = call <vscale x 32 x bfloat> @llvm.fma.nxv32bf16(<vscale x 32 x bfloat> poison, <vscale x 32 x bfloat> poison, <vscale x 32 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; NO-ZFHMIN-LABEL: 'fma_bf16'
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %BF16 = call bfloat @llvm.fma.bf16(bfloat poison, bfloat poison, bfloat poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1BF16 = call <1 x bfloat> @llvm.fma.v1bf16(<1 x bfloat> poison, <1 x bfloat> poison, <1 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2BF16 = call <2 x bfloat> @llvm.fma.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4BF16 = call <4 x bfloat> @llvm.fma.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8BF16 = call <8 x bfloat> @llvm.fma.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16BF16 = call <16 x bfloat> @llvm.fma.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V32BF16 = call <32 x bfloat> @llvm.fma.v32bf16(<32 x bfloat> poison, <32 x bfloat> poison, <32 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1BF16 = call <vscale x 1 x bfloat> @llvm.fma.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2BF16 = call <vscale x 2 x bfloat> @llvm.fma.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4BF16 = call <vscale x 4 x bfloat> @llvm.fma.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8BF16 = call <vscale x 8 x bfloat> @llvm.fma.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16BF16 = call <vscale x 16 x bfloat> @llvm.fma.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32BF16 = call <vscale x 32 x bfloat> @llvm.fma.nxv32bf16(<vscale x 32 x bfloat> poison, <vscale x 32 x bfloat> poison, <vscale x 32 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+  %BF16 = call bfloat @llvm.fma(bfloat poison, bfloat poison, bfloat poison)
+
+  %V1BF16 = call <1 x bfloat> @llvm.fma(<1 x bfloat> poison, <1 x bfloat> poison, <1 x bfloat> poison)
+  %V2BF16 = call <2 x bfloat> @llvm.fma(<2 x bfloat> poison, <2 x bfloat> poison, <2 x bfloat> poison)
+  %V4BF16 = call <4 x bfloat> @llvm.fma(<4 x bfloat> poison, <4 x bfloat> poison, <4 x bfloat> poison)
+  %V8BF16 = call <8 x bfloat> @llvm.fma(<8 x bfloat> poison, <8 x bfloat> poison, <8 x bfloat> poison)
+  %V16BF16 = call <16 x bfloat> @llvm.fma(<16 x bfloat> poison, <16 x bfloat> poison, <16 x bfloat> poison)
+  %V32BF16 = call <32 x bfloat> @llvm.fma(<32 x bfloat> poison, <32 x bfloat> poison, <32 x bfloat> poison)
+
+  %NXV1BF16 = call <vscale x 1 x bfloat> @llvm.fma(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison)
+  %NXV2BF16 = call <vscale x 2 x bfloat> @llvm.fma(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison)
+  %NXV4BF16 = call <vscale x 4 x bfloat> @llvm.fma(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison)
+  %NXV8BF16 = call <vscale x 8 x bfloat> @llvm.fma(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison)
+  %NXV16BF16 = call <vscale x 16 x bfloat> @llvm.fma(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison)
+  %NXV32BF16 = call <vscale x 32 x bfloat> @llvm.fma(<vscale x 32 x bfloat> poison, <vscale x 32 x bfloat> poison, <vscale x 32 x bfloat> poison)
 
   ret void
 }
 
 define void @fma_f16() {
-; CHECK-LABEL: 'fma_f16'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F16 = call half @llvm.fma.f16(half undef, half undef, half undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1F16 = call <1 x half> @llvm.fma.v1f16(<1 x half> undef, <1 x half> undef, <1 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F16 = call <2 x half> @llvm.fma.v2f16(<2 x half> undef, <2 x half> undef, <2 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F16 = call <4 x half> @llvm.fma.v4f16(<4 x half> undef, <4 x half> undef, <4 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8F16 = call <8 x half> @llvm.fma.v8f16(<8 x half> undef, <8 x half> undef, <8 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16F16 = call <16 x half> @llvm.fma.v16f16(<16 x half> undef, <16 x half> undef, <16 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32F16 = call <32 x half> @llvm.fma.v32f16(<32 x half> undef, <32 x half> undef, <32 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV1F16 = call <vscale x 1 x half> @llvm.fma.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef, <vscale x 1 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV2F16 = call <vscale x 2 x half> @llvm.fma.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef, <vscale x 2 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV4F16 = call <vscale x 4 x half> @llvm.fma.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef, <vscale x 4 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV8F16 = call <vscale x 8 x half> @llvm.fma.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef, <vscale x 8 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV16F16 = call <vscale x 16 x half> @llvm.fma.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef, <vscale x 16 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV32F16 = call <vscale x 32 x half> @llvm.fma.nxv32f16(<vscale x 32 x half> undef, <vscale x 32 x half> undef, <vscale x 32 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+; ZVFH-LABEL: 'fma_f16'
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F16 = call half @llvm.fma.f16(half poison, half poison, half poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1F16 = call <1 x half> @llvm.fma.v1f16(<1 x half> poison, <1 x half> poison, <1 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F16 = call <2 x half> @llvm.fma.v2f16(<2 x half> poison, <2 x half> poison, <2 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F16 = call <4 x half> @llvm.fma.v4f16(<4 x half> poison, <4 x half> poison, <4 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8F16 = call <8 x half> @llvm.fma.v8f16(<8 x half> poison, <8 x half> poison, <8 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16F16 = call <16 x half> @llvm.fma.v16f16(<16 x half> poison, <16 x half> poison, <16 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32F16 = call <32 x half> @llvm.fma.v32f16(<32 x half> poison, <32 x half> poison, <32 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV1F16 = call <vscale x 1 x half> @llvm.fma.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV2F16 = call <vscale x 2 x half> @llvm.fma.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV4F16 = call <vscale x 4 x half> @llvm.fma.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV8F16 = call <vscale x 8 x half> @llvm.fma.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV16F16 = call <vscale x 16 x half> @llvm.fma.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV32F16 = call <vscale x 32 x half> @llvm.fma.nxv32f16(<vscale x 32 x half> poison, <vscale x 32 x half> poison, <vscale x 32 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; ZVFHMIN-LABEL: 'fma_f16'
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F16 = call half @llvm.fma.f16(half poison, half poison, half poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1F16 = call <1 x half> @llvm.fma.v1f16(<1 x half> poison, <1 x half> poison, <1 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V2F16 = call <2 x half> @llvm.fma.v2f16(<2 x half> poison, <2 x half> poison, <2 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V4F16 = call <4 x half> @llvm.fma.v4f16(<4 x half> poison, <4 x half> poison, <4 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V8F16 = call <8 x half> @llvm.fma.v8f16(<8 x half> poison, <8 x half> poison, <8 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %V16F16 = call <16 x half> @llvm.fma.v16f16(<16 x half> poison, <16 x half> poison, <16 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V32F16 = call <32 x half> @llvm.fma.v32f16(<32 x half> poison, <32 x half> poison, <32 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV1F16 = call <vscale x 1 x half> @llvm.fma.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV2F16 = call <vscale x 2 x half> @llvm.fma.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %NXV4F16 = call <vscale x 4 x half> @llvm.fma.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %NXV8F16 = call <vscale x 8 x half> @llvm.fma.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %NXV16F16 = call <vscale x 16 x half> @llvm.fma.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %NXV32F16 = call <vscale x 32 x half> @llvm.fma.nxv32f16(<vscale x 32 x half> poison, <vscale x 32 x half> poison, <vscale x 32 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; NO-ZFHMIN-LABEL: 'fma_f16'
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %F16 = call half @llvm.fma.f16(half poison, half poison, half poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %V1F16 = call <1 x half> @llvm.fma.v1f16(<1 x half> poison, <1 x half> poison, <1 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %V2F16 = call <2 x half> @llvm.fma.v2f16(<2 x half> poison, <2 x half> poison, <2 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %V4F16 = call <4 x half> @llvm.fma.v4f16(<4 x half> poison, <4 x half> poison, <4 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %V8F16 = call <8 x half> @llvm.fma.v8f16(<8 x half> poison, <8 x half> poison, <8 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %V16F16 = call <16 x half> @llvm.fma.v16f16(<16 x half> poison, <16 x half> poison, <16 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 64 for instruction: %V32F16 = call <32 x half> @llvm.fma.v32f16(<32 x half> poison, <32 x half> poison, <32 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV1F16 = call <vscale x 1 x half> @llvm.fma.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV2F16 = call <vscale x 2 x half> @llvm.fma.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV4F16 = call <vscale x 4 x half> @llvm.fma.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV8F16 = call <vscale x 8 x half> @llvm.fma.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV16F16 = call <vscale x 16 x half> @llvm.fma.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %NXV32F16 = call <vscale x 32 x half> @llvm.fma.nxv32f16(<vscale x 32 x half> poison, <vscale x 32 x half> poison, <vscale x 32 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
-  %F16 = call half @llvm.fma.f16(half undef, half undef, half undef)
-
-  %V1F16 = call <1 x half> @llvm.fma.v1f16(<1 x half> undef, <1 x half> undef, <1 x half> undef)
-  %V2F16 = call <2 x half> @llvm.fma.v2f16(<2 x half> undef, <2 x half> undef, <2 x half> undef)
-  %V4F16 = call <4 x half> @llvm.fma.v4f16(<4 x half> undef, <4 x half> undef, <4 x half> undef)
-  %V8F16 = call <8 x half> @llvm.fma.v8f16(<8 x half> undef, <8 x half> undef, <8 x half> undef)
-  %V16F16 = call <16 x half> @llvm.fma.v16f16(<16 x half> undef, <16 x half> undef, <16 x half> undef)
-  %V32F16 = call <32 x half> @llvm.fma.v32f16(<32 x half> undef, <32 x half> undef, <32 x half> undef)
-
-  %NXV1F16 = call <vscale x 1 x half> @llvm.fma.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef, <vscale x 1 x half> undef)
-  %NXV2F16 = call <vscale x 2 x half> @llvm.fma.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef, <vscale x 2 x half> undef)
-  %NXV4F16 = call <vscale x 4 x half> @llvm.fma.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef, <vscale x 4 x half> undef)
-  %NXV8F16 = call <vscale x 8 x half> @llvm.fma.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef, <vscale x 8 x half> undef)
-  %NXV16F16 = call <vscale x 16 x half> @llvm.fma.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef, <vscale x 16 x half> undef)
-  %NXV32F16 = call <vscale x 32 x half> @llvm.fma.nxv32f16(<vscale x 32 x half> undef, <vscale x 32 x half> undef, <vscale x 32 x half> undef)
+  %F16 = call half @llvm.fma(half poison, half poison, half poison)
+
+  %V1F16 = call <1 x half> @llvm.fma(<1 x half> poison, <1 x half> poison, <1 x half> poison)
+  %V2F16 = call <2 x half> @llvm.fma(<2 x half> poison, <2 x half> poison, <2 x half> poison)
+  %V4F16 = call <4 x half> @llvm.fma(<4 x half> poison, <4 x half> poison, <4 x half> poison)
+  %V8F16 = call <8 x half> @llvm.fma(<8 x half> poison, <8 x half> poison, <8 x half> poison)
+  %V16F16 = call <16 x half> @llvm.fma(<16 x half> poison, <16 x half> poison, <16 x half> poison)
+  %V32F16 = call <32 x half> @llvm.fma(<32 x half> poison, <32 x half> poison, <32 x half> poison)
+
+  %NXV1F16 = call <vscale x 1 x half> @llvm.fma(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x half> poison)
+  %NXV2F16 = call <vscale x 2 x half> @llvm.fma(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x half> poison)
+  %NXV4F16 = call <vscale x 4 x half> @llvm.fma(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x half> poison)
+  %NXV8F16 = call <vscale x 8 x half> @llvm.fma(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x half> poison)
+  %NXV16F16 = call <vscale x 16 x half> @llvm.fma(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x half> poison)
+  %NXV32F16 = call <vscale x 32 x half> @llvm.fma(<vscale x 32 x half> poison, <vscale x 32 x half> poison, <vscale x 32 x half> poison)
 
   ret void
 }
 
 define void @fmuladd() {
 ; CHECK-LABEL: 'fmuladd'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %1 = call bfloat @llvm.fmuladd.bf16(bfloat undef, bfloat undef, bfloat undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %2 = call float @llvm.fmuladd.f32(float undef, float undef, float undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %3 = call double @llvm.fmuladd.f64(double undef, double undef, double undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %4 = call <2 x bfloat> @llvm.fmuladd.v2bf16(<2 x bfloat> undef, <2 x bfloat> undef, <2 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %5 = call <4 x bfloat> @llvm.fmuladd.v4bf16(<4 x bfloat> undef, <4 x bfloat> undef, <4 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %6 = call <8 x bfloat> @llvm.fmuladd.v8bf16(<8 x bfloat> undef, <8 x bfloat> undef, <8 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %7 = call <16 x bfloat> @llvm.fmuladd.v16bf16(<16 x bfloat> undef, <16 x bfloat> undef, <16 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %8 = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> undef, <2 x float> undef, <2 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %9 = call <4 x float> @llvm.fmuladd.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %10 = call <8 x float> @llvm.fmuladd.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %11 = call <16 x float> @llvm.fmuladd.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %12 = call <2 x double> @llvm.fmuladd.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %13 = call <4 x double> @llvm.fmuladd.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %14 = call <8 x double> @llvm.fmuladd.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %15 = call <16 x double> @llvm.fmuladd.v16f64(<16 x double> undef, <16 x double> undef, <16 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %16 = call <vscale x 1 x bfloat> @llvm.fmuladd.nxv1bf16(<vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %17 = call <vscale x 2 x bfloat> @llvm.fmuladd.nxv2bf16(<vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %18 = call <vscale x 4 x bfloat> @llvm.fmuladd.nxv4bf16(<vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %19 = call <vscale x 8 x bfloat> @llvm.fmuladd.nxv8bf16(<vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %20 = call <vscale x 16 x bfloat> @llvm.fmuladd.nxv16bf16(<vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %21 = call <vscale x 1 x float> @llvm.fmuladd.nxv1f32(<vscale x 1 x float> undef, <vscale x 1 x float> undef, <vscale x 1 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %22 = call <vscale x 2 x float> @llvm.fmuladd.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> undef, <vscale x 2 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %23 = call <vscale x 4 x float> @llvm.fmuladd.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x float> undef, <vscale x 4 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %24 = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x float> undef, <vscale x 8 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %25 = call <vscale x 16 x float> @llvm.fmuladd.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x float> undef, <vscale x 16 x float> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %26 = call <vscale x 1 x double> @llvm.fmuladd.nxv1f64(<vscale x 1 x double> undef, <vscale x 1 x double> undef, <vscale x 1 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %27 = call <vscale x 2 x double> @llvm.fmuladd.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x double> undef, <vscale x 2 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %28 = call <vscale x 4 x double> @llvm.fmuladd.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x double> undef, <vscale x 4 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %29 = call <vscale x 8 x double> @llvm.fmuladd.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x double> undef, <vscale x 8 x double> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %30 = call <vscale x 16 x double> @llvm.fmuladd.nxv16f64(<vscale x 16 x double> undef, <vscale x 16 x double> undef, <vscale x 16 x double> undef)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %1 = call float @llvm.fmuladd.f32(float poison, float poison, float poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %2 = call double @llvm.fmuladd.f64(double poison, double poison, double poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %3 = call <2 x float> @llvm.fmuladd.v2f32(<2 x float> poison, <2 x float> poison, <2 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %4 = call <4 x float> @llvm.fmuladd.v4f32(<4 x float> poison, <4 x float> poison, <4 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %5 = call <8 x float> @llvm.fmuladd.v8f32(<8 x float> poison, <8 x float> poison, <8 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %6 = call <16 x float> @llvm.fmuladd.v16f32(<16 x float> poison, <16 x float> poison, <16 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %7 = call <2 x double> @llvm.fmuladd.v2f64(<2 x double> poison, <2 x double> poison, <2 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %8 = call <4 x double> @llvm.fmuladd.v4f64(<4 x double> poison, <4 x double> poison, <4 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %9 = call <8 x double> @llvm.fmuladd.v8f64(<8 x double> poison, <8 x double> poison, <8 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %10 = call <16 x double> @llvm.fmuladd.v16f64(<16 x double> poison, <16 x double> poison, <16 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %11 = call <vscale x 1 x float> @llvm.fmuladd.nxv1f32(<vscale x 1 x float> poison, <vscale x 1 x float> poison, <vscale x 1 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %12 = call <vscale x 2 x float> @llvm.fmuladd.nxv2f32(<vscale x 2 x float> poison, <vscale x 2 x float> poison, <vscale x 2 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %13 = call <vscale x 4 x float> @llvm.fmuladd.nxv4f32(<vscale x 4 x float> poison, <vscale x 4 x float> poison, <vscale x 4 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %14 = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> poison, <vscale x 8 x float> poison, <vscale x 8 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %15 = call <vscale x 16 x float> @llvm.fmuladd.nxv16f32(<vscale x 16 x float> poison, <vscale x 16 x float> poison, <vscale x 16 x float> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %16 = call <vscale x 1 x double> @llvm.fmuladd.nxv1f64(<vscale x 1 x double> poison, <vscale x 1 x double> poison, <vscale x 1 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %17 = call <vscale x 2 x double> @llvm.fmuladd.nxv2f64(<vscale x 2 x double> poison, <vscale x 2 x double> poison, <vscale x 2 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %18 = call <vscale x 4 x double> @llvm.fmuladd.nxv4f64(<vscale x 4 x double> poison, <vscale x 4 x double> poison, <vscale x 4 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %19 = call <vscale x 8 x double> @llvm.fmuladd.nxv8f64(<vscale x 8 x double> poison, <vscale x 8 x double> poison, <vscale x 8 x double> poison)
+; CHECK-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %20 = call <vscale x 16 x double> @llvm.fmuladd.nxv16f64(<vscale x 16 x double> poison, <vscale x 16 x double> poison, <vscale x 16 x double> poison)
 ; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
-  call bfloat @llvm.fmuladd.bf16(bfloat undef, bfloat undef, bfloat undef)
-  call float @llvm.fmuladd.f32(float undef, float undef, float undef)
-  call double @llvm.fmuladd.f64(double undef, double undef, double undef)
-  call <2 x bfloat> @llvm.fmuladd.v2bf16(<2 x bfloat> undef, <2 x bfloat> undef, <2 x bfloat> undef)
-  call <4 x bfloat> @llvm.fmuladd.v4bf16(<4 x bfloat> undef, <4 x bfloat> undef, <4 x bfloat> undef)
-  call <8 x bfloat> @llvm.fmuladd.v8bf16(<8 x bfloat> undef, <8 x bfloat> undef, <8 x bfloat> undef)
-  call <16 x bfloat> @llvm.fmuladd.v16bf16(<16 x bfloat> undef, <16 x bfloat> undef, <16 x bfloat> undef)
-  call <2 x float> @llvm.fmuladd.v2f32(<2 x float> undef, <2 x float> undef, <2 x float> undef)
-  call <4 x float> @llvm.fmuladd.v4f32(<4 x float> undef, <4 x float> undef, <4 x float> undef)
-  call <8 x float> @llvm.fmuladd.v8f32(<8 x float> undef, <8 x float> undef, <8 x float> undef)
-  call <16 x float> @llvm.fmuladd.v16f32(<16 x float> undef, <16 x float> undef, <16 x float> undef)
-  call <2 x double> @llvm.fmuladd.v2f64(<2 x double> undef, <2 x double> undef, <2 x double> undef)
-  call <4 x double> @llvm.fmuladd.v4f64(<4 x double> undef, <4 x double> undef, <4 x double> undef)
-  call <8 x double> @llvm.fmuladd.v8f64(<8 x double> undef, <8 x double> undef, <8 x double> undef)
-  call <16 x double> @llvm.fmuladd.v16f64(<16 x double> undef, <16 x double> undef, <16 x double> undef)
-  call <vscale x 1 x bfloat> @llvm.fmuladd.nxv1bf16(<vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef, <vscale x 1 x bfloat> undef)
-  call <vscale x 2 x bfloat> @llvm.fmuladd.nxv2bf16(<vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef, <vscale x 2 x bfloat> undef)
-  call <vscale x 4 x bfloat> @llvm.fmuladd.nxv4bf16(<vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef, <vscale x 4 x bfloat> undef)
-  call <vscale x 8 x bfloat> @llvm.fmuladd.nxv8bf16(<vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef, <vscale x 8 x bfloat> undef)
-  call <vscale x 16 x bfloat> @llvm.fmuladd.nxv16bf16(<vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef, <vscale x 16 x bfloat> undef)
-  call <vscale x 1 x float> @llvm.fmuladd.nxv1f32(<vscale x 1 x float> undef, <vscale x 1 x float> undef, <vscale x 1 x float> undef)
-  call <vscale x 2 x float> @llvm.fmuladd.nxv2f32(<vscale x 2 x float> undef, <vscale x 2 x float> undef, <vscale x 2 x float> undef)
-  call <vscale x 4 x float> @llvm.fmuladd.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x float> undef, <vscale x 4 x float> undef)
-  call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> undef, <vscale x 8 x float> undef, <vscale x 8 x float> undef)
-  call <vscale x 16 x float> @llvm.fmuladd.nxv16f32(<vscale x 16 x float> undef, <vscale x 16 x float> undef, <vscale x 16 x float> undef)
-  call <vscale x 1 x double> @llvm.fmuladd.nxv1f64(<vscale x 1 x double> undef, <vscale x 1 x double> undef, <vscale x 1 x double> undef)
-  call <vscale x 2 x double> @llvm.fmuladd.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x double> undef, <vscale x 2 x double> undef)
-  call <vscale x 4 x double> @llvm.fmuladd.nxv4f64(<vscale x 4 x double> undef, <vscale x 4 x double> undef, <vscale x 4 x double> undef)
-  call <vscale x 8 x double> @llvm.fmuladd.nxv8f64(<vscale x 8 x double> undef, <vscale x 8 x double> undef, <vscale x 8 x double> undef)
-  call <vscale x 16 x double> @llvm.fmuladd.nxv16f64(<vscale x 16 x double> undef, <vscale x 16 x double> undef, <vscale x 16 x double> undef)
+  call float @llvm.fmuladd(float poison, float poison, float poison)
+  call double @llvm.fmuladd(double poison, double poison, double poison)
+
+  call <2 x float> @llvm.fmuladd(<2 x float> poison, <2 x float> poison, <2 x float> poison)
+  call <4 x float> @llvm.fmuladd(<4 x float> poison, <4 x float> poison, <4 x float> poison)
+  call <8 x float> @llvm.fmuladd(<8 x float> poison, <8 x float> poison, <8 x float> poison)
+  call <16 x float> @llvm.fmuladd(<16 x float> poison, <16 x float> poison, <16 x float> poison)
+
+  call <2 x double> @llvm.fmuladd(<2 x double> poison, <2 x double> poison, <2 x double> poison)
+  call <4 x double> @llvm.fmuladd(<4 x double> poison, <4 x double> poison, <4 x double> poison)
+  call <8 x double> @llvm.fmuladd(<8 x double> poison, <8 x double> poison, <8 x double> poison)
+  call <16 x double> @llvm.fmuladd(<16 x double> poison, <16 x double> poison, <16 x double> poison)
+
+  call <vscale x 1 x float> @llvm.fmuladd(<vscale x 1 x float> poison, <vscale x 1 x float> poison, <vscale x 1 x float> poison)
+  call <vscale x 2 x float> @llvm.fmuladd(<vscale x 2 x float> poison, <vscale x 2 x float> poison, <vscale x 2 x float> poison)
+  call <vscale x 4 x float> @llvm.fmuladd(<vscale x 4 x float> poison, <vscale x 4 x float> poison, <vscale x 4 x float> poison)
+  call <vscale x 8 x float> @llvm.fmuladd(<vscale x 8 x float> poison, <vscale x 8 x float> poison, <vscale x 8 x float> poison)
+  call <vscale x 16 x float> @llvm.fmuladd(<vscale x 16 x float> poison, <vscale x 16 x float> poison, <vscale x 16 x float> poison)
+
+  call <vscale x 1 x double> @llvm.fmuladd(<vscale x 1 x double> poison, <vscale x 1 x double> poison, <vscale x 1 x double> poison)
+  call <vscale x 2 x double> @llvm.fmuladd(<vscale x 2 x double> poison, <vscale x 2 x double> poison, <vscale x 2 x double> poison)
+  call <vscale x 4 x double> @llvm.fmuladd(<vscale x 4 x double> poison, <vscale x 4 x double> poison, <vscale x 4 x double> poison)
+  call <vscale x 8 x double> @llvm.fmuladd(<vscale x 8 x double> poison, <vscale x 8 x double> poison, <vscale x 8 x double> poison)
+  call <vscale x 16 x double> @llvm.fmuladd(<vscale x 16 x double> poison, <vscale x 16 x double> poison, <vscale x 16 x double> poison)
+
+  ret void
+}
+
+define void @fmuladd_bf16() {
+; ZVFH-LABEL: 'fmuladd_bf16'
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %1 = call bfloat @llvm.fmuladd.bf16(bfloat poison, bfloat poison, bfloat poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %2 = call <2 x bfloat> @llvm.fmuladd.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %3 = call <4 x bfloat> @llvm.fmuladd.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %4 = call <8 x bfloat> @llvm.fmuladd.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %5 = call <16 x bfloat> @llvm.fmuladd.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %6 = call <vscale x 1 x bfloat> @llvm.fmuladd.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %7 = call <vscale x 2 x bfloat> @llvm.fmuladd.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %8 = call <vscale x 4 x bfloat> @llvm.fmuladd.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %9 = call <vscale x 8 x bfloat> @llvm.fmuladd.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %10 = call <vscale x 16 x bfloat> @llvm.fmuladd.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; ZVFHMIN-LABEL: 'fmuladd_bf16'
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %1 = call bfloat @llvm.fmuladd.bf16(bfloat poison, bfloat poison, bfloat poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %2 = call <2 x bfloat> @llvm.fmuladd.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %3 = call <4 x bfloat> @llvm.fmuladd.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %4 = call <8 x bfloat> @llvm.fmuladd.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %5 = call <16 x bfloat> @llvm.fmuladd.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %6 = call <vscale x 1 x bfloat> @llvm.fmuladd.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %7 = call <vscale x 2 x bfloat> @llvm.fmuladd.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %8 = call <vscale x 4 x bfloat> @llvm.fmuladd.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %9 = call <vscale x 8 x bfloat> @llvm.fmuladd.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %10 = call <vscale x 16 x bfloat> @llvm.fmuladd.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; NO-ZFHMIN-LABEL: 'fmuladd_bf16'
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %1 = call bfloat @llvm.fmuladd.bf16(bfloat poison, bfloat poison, bfloat poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %2 = call <2 x bfloat> @llvm.fmuladd.v2bf16(<2 x bfloat> poison, <2 x bfloat> poison, <2 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %3 = call <4 x bfloat> @llvm.fmuladd.v4bf16(<4 x bfloat> poison, <4 x bfloat> poison, <4 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %4 = call <8 x bfloat> @llvm.fmuladd.v8bf16(<8 x bfloat> poison, <8 x bfloat> poison, <8 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %5 = call <16 x bfloat> @llvm.fmuladd.v16bf16(<16 x bfloat> poison, <16 x bfloat> poison, <16 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %6 = call <vscale x 1 x bfloat> @llvm.fmuladd.nxv1bf16(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %7 = call <vscale x 2 x bfloat> @llvm.fmuladd.nxv2bf16(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %8 = call <vscale x 4 x bfloat> @llvm.fmuladd.nxv4bf16(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %9 = call <vscale x 8 x bfloat> @llvm.fmuladd.nxv8bf16(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %10 = call <vscale x 16 x bfloat> @llvm.fmuladd.nxv16bf16(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+  call bfloat @llvm.fmuladd(bfloat poison, bfloat poison, bfloat poison)
+
+  call <2 x bfloat> @llvm.fmuladd(<2 x bfloat> poison, <2 x bfloat> poison, <2 x bfloat> poison)
+  call <4 x bfloat> @llvm.fmuladd(<4 x bfloat> poison, <4 x bfloat> poison, <4 x bfloat> poison)
+  call <8 x bfloat> @llvm.fmuladd(<8 x bfloat> poison, <8 x bfloat> poison, <8 x bfloat> poison)
+  call <16 x bfloat> @llvm.fmuladd(<16 x bfloat> poison, <16 x bfloat> poison, <16 x bfloat> poison)
+
+  call <vscale x 1 x bfloat> @llvm.fmuladd(<vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison, <vscale x 1 x bfloat> poison)
+  call <vscale x 2 x bfloat> @llvm.fmuladd(<vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison, <vscale x 2 x bfloat> poison)
+  call <vscale x 4 x bfloat> @llvm.fmuladd(<vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison, <vscale x 4 x bfloat> poison)
+  call <vscale x 8 x bfloat> @llvm.fmuladd(<vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison, <vscale x 8 x bfloat> poison)
+  call <vscale x 16 x bfloat> @llvm.fmuladd(<vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison, <vscale x 16 x bfloat> poison)
   ret void
 }
 
 define void @fmuladd_f16() {
-; CHECK-LABEL: 'fmuladd_f16'
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %1 = call half @llvm.fmuladd.f16(half undef, half undef, half undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %2 = call <2 x half> @llvm.fmuladd.v2f16(<2 x half> undef, <2 x half> undef, <2 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %3 = call <4 x half> @llvm.fmuladd.v4f16(<4 x half> undef, <4 x half> undef, <4 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %4 = call <8 x half> @llvm.fmuladd.v8f16(<8 x half> undef, <8 x half> undef, <8 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %5 = call <16 x half> @llvm.fmuladd.v16f16(<16 x half> undef, <16 x half> undef, <16 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %6 = call <vscale x 1 x half> @llvm.fmuladd.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef, <vscale x 1 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %7 = call <vscale x 2 x half> @llvm.fmuladd.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef, <vscale x 2 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %8 = call <vscale x 4 x half> @llvm.fmuladd.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef, <vscale x 4 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %9 = call <vscale x 8 x half> @llvm.fmuladd.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef, <vscale x 8 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %10 = call <vscale x 16 x half> @llvm.fmuladd.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef, <vscale x 16 x half> undef)
-; CHECK-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+; ZVFH-LABEL: 'fmuladd_f16'
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %1 = call half @llvm.fmuladd.f16(half poison, half poison, half poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %2 = call <2 x half> @llvm.fmuladd.v2f16(<2 x half> poison, <2 x half> poison, <2 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %3 = call <4 x half> @llvm.fmuladd.v4f16(<4 x half> poison, <4 x half> poison, <4 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %4 = call <8 x half> @llvm.fmuladd.v8f16(<8 x half> poison, <8 x half> poison, <8 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %5 = call <16 x half> @llvm.fmuladd.v16f16(<16 x half> poison, <16 x half> poison, <16 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %6 = call <vscale x 1 x half> @llvm.fmuladd.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %7 = call <vscale x 2 x half> @llvm.fmuladd.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %8 = call <vscale x 4 x half> @llvm.fmuladd.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %9 = call <vscale x 8 x half> @llvm.fmuladd.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %10 = call <vscale x 16 x half> @llvm.fmuladd.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x half> poison)
+; ZVFH-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; ZVFHMIN-LABEL: 'fmuladd_f16'
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %1 = call half @llvm.fmuladd.f16(half poison, half poison, half poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %2 = call <2 x half> @llvm.fmuladd.v2f16(<2 x half> poison, <2 x half> poison, <2 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %3 = call <4 x half> @llvm.fmuladd.v4f16(<4 x half> poison, <4 x half> poison, <4 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %4 = call <8 x half> @llvm.fmuladd.v8f16(<8 x half> poison, <8 x half> poison, <8 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %5 = call <16 x half> @llvm.fmuladd.v16f16(<16 x half> poison, <16 x half> poison, <16 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %6 = call <vscale x 1 x half> @llvm.fmuladd.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %7 = call <vscale x 2 x half> @llvm.fmuladd.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %8 = call <vscale x 4 x half> @llvm.fmuladd.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %9 = call <vscale x 8 x half> @llvm.fmuladd.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %10 = call <vscale x 16 x half> @llvm.fmuladd.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x half> poison)
+; ZVFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
+;
+; NO-ZFHMIN-LABEL: 'fmuladd_f16'
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %1 = call half @llvm.fmuladd.f16(half poison, half poison, half poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 4 for instruction: %2 = call <2 x half> @llvm.fmuladd.v2f16(<2 x half> poison, <2 x half> poison, <2 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 8 for instruction: %3 = call <4 x half> @llvm.fmuladd.v4f16(<4 x half> poison, <4 x half> poison, <4 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 16 for instruction: %4 = call <8 x half> @llvm.fmuladd.v8f16(<8 x half> poison, <8 x half> poison, <8 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 32 for instruction: %5 = call <16 x half> @llvm.fmuladd.v16f16(<16 x half> poison, <16 x half> poison, <16 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %6 = call <vscale x 1 x half> @llvm.fmuladd.nxv1f16(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %7 = call <vscale x 2 x half> @llvm.fmuladd.nxv2f16(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %8 = call <vscale x 4 x half> @llvm.fmuladd.nxv4f16(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %9 = call <vscale x 8 x half> @llvm.fmuladd.nxv8f16(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Invalid cost for instruction: %10 = call <vscale x 16 x half> @llvm.fmuladd.nxv16f16(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x half> poison)
+; NO-ZFHMIN-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret void
 ;
-  call half @llvm.fmuladd.f16(half undef, half undef, half undef)
-  call <2 x half> @llvm.fmuladd.v2f16(<2 x half> undef, <2 x half> undef, <2 x half> undef)
-  call <4 x half> @llvm.fmuladd.v4f16(<4 x half> undef, <4 x half> undef, <4 x half> undef)
-  call <8 x half> @llvm.fmuladd.v8f16(<8 x half> undef, <8 x half> undef, <8 x half> undef)
-  call <16 x half> @llvm.fmuladd.v16f16(<16 x half> undef, <16 x half> undef, <16 x half> undef)
-  call <vscale x 1 x half> @llvm.fmuladd.nxv1f16(<vscale x 1 x half> undef, <vscale x 1 x half> undef, <vscale x 1 x half> undef)
-  call <vscale x 2 x half> @llvm.fmuladd.nxv2f16(<vscale x 2 x half> undef, <vscale x 2 x half> undef, <vscale x 2 x half> undef)
-  call <vscale x 4 x half> @llvm.fmuladd.nxv4f16(<vscale x 4 x half> undef, <vscale x 4 x half> undef, <vscale x 4 x half> undef)
-  call <vscale x 8 x half> @llvm.fmuladd.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x half> undef, <vscale x 8 x half> undef)
-  call <vscale x 16 x half> @llvm.fmuladd.nxv16f16(<vscale x 16 x half> undef, <vscale x 16 x half> undef, <vscale x 16 x half> undef)
+  call half @llvm.fmuladd(half poison, half poison, half poison)
+
+  call <2 x half> @llvm.fmuladd(<2 x half> poison, <2 x half> poison, <2 x half> poison)
+  call <4 x half> @llvm.fmuladd(<4 x half> poison, <4 x half> poison, <4 x half> poison)
+  call <8 x half> @llvm.fmuladd(<8 x half> poison, <8 x half> poison, <8 x half> poison)
+  call <16 x half> @llvm.fmuladd(<16 x half> poison, <16 x half> poison, <16 x half> poison)
+
+  call <vscale x 1 x half> @llvm.fmuladd(<vscale x 1 x half> poison, <vscale x 1 x half> poison, <vscale x 1 x half> poison)
+  call <vscale x 2 x half> @llvm.fmuladd(<vscale x 2 x half> poison, <vscale x 2 x half> poison, <vscale x 2 x half> poison)
+  call <vscale x 4 x half> @llvm.fmuladd(<vscale x 4 x half> poison, <vscale x 4 x half> poison, <vscale x 4 x half> poison)
+  call <vscale x 8 x half> @llvm.fmuladd(<vscale x 8 x half> poison, <vscale x 8 x half> poison, <vscale x 8 x half> poison)
+  call <vscale x 16 x half> @llvm.fmuladd(<vscale x 16 x half> poison, <vscale x 16 x half> poison, <vscale x 16 x half> poison)
   ret void
 }
diff --git a/llvm/test/Analysis/CostModel/X86/fshl.ll b/llvm/test/Analysis/CostModel/X86/fshl.ll
index 485e7f6..c53a507 100644
--- a/llvm/test/Analysis/CostModel/X86/fshl.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshl.ll
@@ -414,8 +414,8 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX1-LABEL: 'splatvar_funnel_i64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:13 SizeLat:11 for: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:21 Lat:23 SizeLat:30 for: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:38 CodeSize:42 Lat:46 SizeLat:60 for: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
@@ -423,8 +423,8 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX2-LABEL: 'splatvar_funnel_i64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:11 SizeLat:11 for: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:7 Lat:14 SizeLat:14 for: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:14 Lat:28 SizeLat:28 for: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
@@ -432,8 +432,8 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX512F-LABEL: 'splatvar_funnel_i64'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:13 SizeLat:10 for: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:13 SizeLat:9 for: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
@@ -441,8 +441,8 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX512BW-LABEL: 'splatvar_funnel_i64'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:13 SizeLat:9 for: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:13 SizeLat:9 for: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
@@ -450,8 +450,8 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX512DQ-LABEL: 'splatvar_funnel_i64'
 ; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:13 SizeLat:10 for: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:13 SizeLat:9 for: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
@@ -459,8 +459,8 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX512VBMI2-LABEL: 'splatvar_funnel_i64'
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
@@ -486,8 +486,8 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; XOP-LABEL: 'splatvar_funnel_i64'
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:12 SizeLat:10 for: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:21 Lat:23 SizeLat:30 for: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:38 CodeSize:42 Lat:46 SizeLat:60 for: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
@@ -495,8 +495,8 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX512GFNI-LABEL: 'splatvar_funnel_i64'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
@@ -532,8 +532,8 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX1-LABEL: 'splatvar_funnel_i32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:12 SizeLat:10 for: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:23 Lat:23 SizeLat:32 for: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:38 CodeSize:46 Lat:46 SizeLat:64 for: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
@@ -541,8 +541,8 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX2-LABEL: 'splatvar_funnel_i32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:10 SizeLat:10 for: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:9 Lat:14 SizeLat:16 for: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:18 Lat:28 SizeLat:32 for: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
@@ -550,8 +550,8 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX512F-LABEL: 'splatvar_funnel_i32'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:12 for: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:11 for: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
@@ -559,8 +559,8 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX512BW-LABEL: 'splatvar_funnel_i32'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:11 for: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:11 for: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
@@ -568,8 +568,8 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX512DQ-LABEL: 'splatvar_funnel_i32'
 ; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:12 for: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:11 for: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
@@ -577,8 +577,8 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX512VBMI2-LABEL: 'splatvar_funnel_i32'
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
@@ -604,8 +604,8 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; XOP-LABEL: 'splatvar_funnel_i32'
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:12 SizeLat:10 for: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:23 Lat:23 SizeLat:32 for: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:38 CodeSize:46 Lat:46 SizeLat:64 for: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
@@ -613,8 +613,8 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX512GFNI-LABEL: 'splatvar_funnel_i32'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
@@ -631,108 +631,108 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 
 define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512, i16 %b16, <8 x i16> %b128, <16 x i16> %b256, <32 x i16> %b512, i16 %c16, <8 x i16> %c128, <16 x i16> %c256, <32 x i16> %c512) {
 ; SSSE3-LABEL: 'splatvar_funnel_i16'
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:9 Lat:10 SizeLat:11 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:17 Lat:19 SizeLat:21 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:33 Lat:37 SizeLat:41 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSE42-LABEL: 'splatvar_funnel_i16'
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:7 Lat:10 SizeLat:10 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:13 Lat:19 SizeLat:19 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:25 Lat:37 SizeLat:37 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'splatvar_funnel_i16'
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:12 SizeLat:10 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:25 Lat:23 SizeLat:33 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:38 CodeSize:50 Lat:46 SizeLat:66 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'splatvar_funnel_i16'
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:10 SizeLat:10 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:9 Lat:14 SizeLat:16 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:18 Lat:28 SizeLat:32 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512F-LABEL: 'splatvar_funnel_i16'
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:13 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:26 Lat:38 SizeLat:30 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512BW-LABEL: 'splatvar_funnel_i16'
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:11 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:11 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512DQ-LABEL: 'splatvar_funnel_i16'
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:13 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:26 Lat:38 SizeLat:30 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512VBMI2-LABEL: 'splatvar_funnel_i16'
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SLM-LABEL: 'splatvar_funnel_i16'
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:7 Lat:12 SizeLat:11 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:13 Lat:23 SizeLat:21 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:48 CodeSize:25 Lat:45 SizeLat:41 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; GLM-LABEL: 'splatvar_funnel_i16'
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; GLM-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:7 Lat:10 SizeLat:10 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:13 Lat:19 SizeLat:19 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:25 Lat:37 SizeLat:37 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'splatvar_funnel_i16'
-; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 3 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 3 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:12 SizeLat:10 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:25 Lat:23 SizeLat:33 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:38 CodeSize:50 Lat:46 SizeLat:66 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512GFNI-LABEL: 'splatvar_funnel_i16'
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
@@ -749,108 +749,108 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 
 define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %a512, i8 %b8, <16 x i8> %b128, <32 x i8> %b256, <64 x i8> %b512, i8 %c8, <16 x i8> %c128, <32 x i8> %c256, <64 x i8> %c512) {
 ; SSSE3-LABEL: 'splatvar_funnel_i8'
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:18 Lat:29 SizeLat:25 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:48 CodeSize:35 Lat:57 SizeLat:49 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:96 CodeSize:69 Lat:113 SizeLat:97 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSE42-LABEL: 'splatvar_funnel_i8'
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:16 Lat:29 SizeLat:24 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:48 CodeSize:31 Lat:57 SizeLat:47 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:96 CodeSize:61 Lat:113 SizeLat:93 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'splatvar_funnel_i8'
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:16 Lat:18 SizeLat:22 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:27 CodeSize:36 Lat:26 SizeLat:47 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:54 CodeSize:72 Lat:52 SizeLat:94 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'splatvar_funnel_i8'
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:15 Lat:20 SizeLat:21 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:18 Lat:21 SizeLat:27 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:36 Lat:42 SizeLat:54 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512F-LABEL: 'splatvar_funnel_i8'
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:15 Lat:19 SizeLat:20 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:24 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:73 Lat:56 SizeLat:85 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512BW-LABEL: 'splatvar_funnel_i8'
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:15 Lat:20 SizeLat:20 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:22 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:23 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512DQ-LABEL: 'splatvar_funnel_i8'
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:15 Lat:19 SizeLat:20 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:24 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:73 Lat:56 SizeLat:85 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512VBMI2-LABEL: 'splatvar_funnel_i8'
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:15 Lat:20 SizeLat:20 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:22 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:23 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SLM-LABEL: 'splatvar_funnel_i8'
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:26 CodeSize:16 Lat:31 SizeLat:25 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:52 CodeSize:31 Lat:61 SizeLat:49 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:104 CodeSize:61 Lat:121 SizeLat:97 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; GLM-LABEL: 'splatvar_funnel_i8'
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; GLM-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:16 Lat:29 SizeLat:24 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:48 CodeSize:31 Lat:57 SizeLat:47 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:96 CodeSize:61 Lat:113 SizeLat:93 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'splatvar_funnel_i8'
-; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; XOP-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:7 Lat:12 SizeLat:8 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:23 CodeSize:25 Lat:23 SizeLat:31 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:46 CodeSize:50 Lat:46 SizeLat:62 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512GFNI-LABEL: 'splatvar_funnel_i8'
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:15 Lat:20 SizeLat:20 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:22 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:23 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
@@ -1934,8 +1934,8 @@ define void @splatvar_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX1-LABEL: 'splatvar_rotate_i64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:9 SizeLat:7 for: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:15 Lat:18 SizeLat:22 for: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:30 Lat:36 SizeLat:44 for: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
@@ -1943,8 +1943,8 @@ define void @splatvar_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX2-LABEL: 'splatvar_rotate_i64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:7 SizeLat:7 for: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:5 Lat:11 SizeLat:10 for: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:10 Lat:22 SizeLat:20 for: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
@@ -1952,8 +1952,8 @@ define void @splatvar_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX512-LABEL: 'splatvar_rotate_i64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
@@ -1979,8 +1979,8 @@ define void @splatvar_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; XOP-LABEL: 'splatvar_rotate_i64'
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V2I64 = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:5 Lat:7 SizeLat:6 for: %V4I64 = call <4 x i64> @llvm.fshl.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:10 Lat:14 SizeLat:12 for: %V8I64 = call <8 x i64> @llvm.fshl.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
@@ -2007,8 +2007,8 @@ define void @splatvar_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX1-LABEL: 'splatvar_rotate_i32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:9 SizeLat:7 for: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:17 Lat:18 SizeLat:24 for: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:34 Lat:36 SizeLat:48 for: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
@@ -2016,8 +2016,8 @@ define void @splatvar_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX2-LABEL: 'splatvar_rotate_i32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:7 SizeLat:7 for: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:11 SizeLat:12 for: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:14 Lat:22 SizeLat:24 for: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
@@ -2025,8 +2025,8 @@ define void @splatvar_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX512-LABEL: 'splatvar_rotate_i32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
@@ -2052,8 +2052,8 @@ define void @splatvar_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; XOP-LABEL: 'splatvar_rotate_i32'
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V2I32 = call <4 x i32> @llvm.fshl.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:5 Lat:7 SizeLat:6 for: %V4I32 = call <8 x i32> @llvm.fshl.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:10 Lat:14 SizeLat:12 for: %V8I32 = call <16 x i32> @llvm.fshl.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
@@ -2070,99 +2070,99 @@ define void @splatvar_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 
 define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512, i16 %c16, <8 x i16> %c128, <16 x i16> %c256, <32 x i16> %c512) {
 ; SSE-LABEL: 'splatvar_rotate_i16'
-; SSE-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SSE-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:5 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; SSE-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:9 Lat:13 SizeLat:13 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; SSE-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:17 Lat:25 SizeLat:25 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'splatvar_rotate_i16'
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:9 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:17 Lat:18 SizeLat:24 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:34 Lat:36 SizeLat:48 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'splatvar_rotate_i16'
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:11 SizeLat:12 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:14 Lat:22 SizeLat:24 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512F-LABEL: 'splatvar_rotate_i16'
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:11 SizeLat:10 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:17 Lat:29 SizeLat:21 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512BW-LABEL: 'splatvar_rotate_i16'
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:6 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:6 Lat:8 SizeLat:7 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:6 Lat:8 SizeLat:8 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512DQ-LABEL: 'splatvar_rotate_i16'
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:11 SizeLat:10 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:17 Lat:29 SizeLat:21 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512VBMI2-LABEL: 'splatvar_rotate_i16'
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SLM-LABEL: 'splatvar_rotate_i16'
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:5 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:9 Lat:13 SizeLat:13 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:17 Lat:25 SizeLat:25 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; GLM-LABEL: 'splatvar_rotate_i16'
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; GLM-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:5 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:9 Lat:13 SizeLat:13 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:17 Lat:25 SizeLat:25 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'splatvar_rotate_i16'
-; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 3 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 3 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:5 Lat:7 SizeLat:6 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:10 Lat:14 SizeLat:12 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512GFNI-LABEL: 'splatvar_rotate_i16'
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshl.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshl.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshl.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
@@ -2179,99 +2179,99 @@ define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 
 define void @splatvar_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %a512, i8 %c8, <16 x i8> %c128, <32 x i8> %c256, <64 x i8> %c512) {
 ; SSE-LABEL: 'splatvar_rotate_i8'
-; SSE-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SSE-NEXT:  Cost Model: Found costs of RThru:21 CodeSize:14 Lat:26 SizeLat:21 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; SSE-NEXT:  Cost Model: Found costs of RThru:42 CodeSize:27 Lat:51 SizeLat:41 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; SSE-NEXT:  Cost Model: Found costs of RThru:84 CodeSize:53 Lat:101 SizeLat:81 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'splatvar_rotate_i8'
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:14 Lat:15 SizeLat:19 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:28 Lat:21 SizeLat:38 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:56 Lat:42 SizeLat:76 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'splatvar_rotate_i8'
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:13 Lat:17 SizeLat:18 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:16 Lat:18 SizeLat:23 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:22 CodeSize:32 Lat:36 SizeLat:46 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512F-LABEL: 'splatvar_rotate_i8'
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:13 Lat:17 SizeLat:18 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:16 Lat:18 SizeLat:21 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:35 CodeSize:64 Lat:47 SizeLat:76 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512BW-LABEL: 'splatvar_rotate_i8'
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:7 Lat:15 SizeLat:10 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:7 Lat:15 SizeLat:10 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:11 Lat:6 SizeLat:12 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512DQ-LABEL: 'splatvar_rotate_i8'
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:13 Lat:17 SizeLat:18 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:16 Lat:18 SizeLat:21 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:35 CodeSize:64 Lat:47 SizeLat:76 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512VBMI2-LABEL: 'splatvar_rotate_i8'
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:7 Lat:15 SizeLat:10 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:7 Lat:15 SizeLat:10 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:11 Lat:6 SizeLat:12 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SLM-LABEL: 'splatvar_rotate_i8'
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:21 CodeSize:14 Lat:26 SizeLat:21 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:42 CodeSize:27 Lat:51 SizeLat:41 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:84 CodeSize:53 Lat:101 SizeLat:81 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; GLM-LABEL: 'splatvar_rotate_i8'
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; GLM-NEXT:  Cost Model: Found costs of RThru:21 CodeSize:14 Lat:26 SizeLat:21 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:42 CodeSize:27 Lat:51 SizeLat:41 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:84 CodeSize:53 Lat:101 SizeLat:81 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'splatvar_rotate_i8'
-; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:5 Lat:7 SizeLat:6 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:10 Lat:14 SizeLat:12 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512GFNI-LABEL: 'splatvar_rotate_i8'
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:7 Lat:15 SizeLat:10 for: %V16I8 = call <16 x i8> @llvm.fshl.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:7 Lat:15 SizeLat:10 for: %V32I8 = call <32 x i8> @llvm.fshl.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:11 Lat:6 SizeLat:12 for: %V64I8 = call <64 x i8> @llvm.fshl.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
diff --git a/llvm/test/Analysis/CostModel/X86/fshr.ll b/llvm/test/Analysis/CostModel/X86/fshr.ll
index af9a91b..1990605 100644
--- a/llvm/test/Analysis/CostModel/X86/fshr.ll
+++ b/llvm/test/Analysis/CostModel/X86/fshr.ll
@@ -414,8 +414,8 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX1-LABEL: 'splatvar_funnel_i64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:13 SizeLat:11 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:21 Lat:23 SizeLat:30 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:38 CodeSize:42 Lat:46 SizeLat:60 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
@@ -423,8 +423,8 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX2-LABEL: 'splatvar_funnel_i64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:11 SizeLat:11 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:7 Lat:14 SizeLat:14 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:14 Lat:28 SizeLat:28 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
@@ -432,8 +432,8 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX512F-LABEL: 'splatvar_funnel_i64'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:13 SizeLat:10 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:13 SizeLat:9 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
@@ -441,8 +441,8 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX512BW-LABEL: 'splatvar_funnel_i64'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:13 SizeLat:9 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:13 SizeLat:9 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
@@ -450,8 +450,8 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX512DQ-LABEL: 'splatvar_funnel_i64'
 ; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:13 SizeLat:10 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:13 SizeLat:9 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
@@ -459,8 +459,8 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX512VBMI2-LABEL: 'splatvar_funnel_i64'
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
@@ -486,8 +486,8 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; XOP-LABEL: 'splatvar_funnel_i64'
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:12 SizeLat:10 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:21 Lat:23 SizeLat:30 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:38 CodeSize:42 Lat:46 SizeLat:60 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
@@ -495,8 +495,8 @@ define void @splatvar_funnel_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX512GFNI-LABEL: 'splatvar_funnel_i64'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %b128, <2 x i64> %u128)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %b256, <4 x i64> %u256)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %b512, <8 x i64> %u512)
@@ -532,8 +532,8 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX1-LABEL: 'splatvar_funnel_i32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:12 SizeLat:10 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:23 Lat:23 SizeLat:32 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:38 CodeSize:46 Lat:46 SizeLat:64 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
@@ -541,8 +541,8 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX2-LABEL: 'splatvar_funnel_i32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:10 SizeLat:10 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:9 Lat:14 SizeLat:16 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:18 Lat:28 SizeLat:32 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
@@ -550,8 +550,8 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX512F-LABEL: 'splatvar_funnel_i32'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:12 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:11 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
@@ -559,8 +559,8 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX512BW-LABEL: 'splatvar_funnel_i32'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:11 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:11 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
@@ -568,8 +568,8 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX512DQ-LABEL: 'splatvar_funnel_i32'
 ; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:12 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:11 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
@@ -577,8 +577,8 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX512VBMI2-LABEL: 'splatvar_funnel_i32'
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
@@ -604,8 +604,8 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; XOP-LABEL: 'splatvar_funnel_i32'
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:12 SizeLat:10 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:23 Lat:23 SizeLat:32 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:38 CodeSize:46 Lat:46 SizeLat:64 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
@@ -613,8 +613,8 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX512GFNI-LABEL: 'splatvar_funnel_i32'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %b128, <4 x i32> %u128)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %b256, <8 x i32> %u256)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %b512, <16 x i32> %u512)
@@ -631,108 +631,108 @@ define void @splatvar_funnel_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 
 define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512, i16 %b16, <8 x i16> %b128, <16 x i16> %b256, <32 x i16> %b512, i16 %c16, <8 x i16> %c128, <16 x i16> %c256, <32 x i16> %c512) {
 ; SSSE3-LABEL: 'splatvar_funnel_i16'
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:9 Lat:10 SizeLat:11 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:17 Lat:19 SizeLat:21 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:33 Lat:37 SizeLat:41 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSE42-LABEL: 'splatvar_funnel_i16'
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:7 Lat:10 SizeLat:10 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:13 Lat:19 SizeLat:19 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:25 Lat:37 SizeLat:37 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'splatvar_funnel_i16'
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:12 SizeLat:10 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:25 Lat:23 SizeLat:33 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:38 CodeSize:50 Lat:46 SizeLat:66 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'splatvar_funnel_i16'
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:10 SizeLat:10 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:9 Lat:14 SizeLat:16 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:18 Lat:28 SizeLat:32 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512F-LABEL: 'splatvar_funnel_i16'
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:13 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:26 Lat:38 SizeLat:30 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512BW-LABEL: 'splatvar_funnel_i16'
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:11 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:11 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512DQ-LABEL: 'splatvar_funnel_i16'
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:9 SizeLat:9 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:9 Lat:13 SizeLat:13 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:26 Lat:38 SizeLat:30 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512VBMI2-LABEL: 'splatvar_funnel_i16'
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SLM-LABEL: 'splatvar_funnel_i16'
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:7 Lat:12 SizeLat:11 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:13 Lat:23 SizeLat:21 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:48 CodeSize:25 Lat:45 SizeLat:41 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; GLM-LABEL: 'splatvar_funnel_i16'
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; GLM-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:7 Lat:10 SizeLat:10 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:13 Lat:19 SizeLat:19 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:25 Lat:37 SizeLat:37 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'splatvar_funnel_i16'
-; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 3 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 3 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:7 Lat:12 SizeLat:10 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:25 Lat:23 SizeLat:33 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:38 CodeSize:50 Lat:46 SizeLat:66 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512GFNI-LABEL: 'splatvar_funnel_i16'
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %b128, <8 x i16> %u128)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %b256, <16 x i16> %u256)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %b512, <32 x i16> %u512)
@@ -749,108 +749,108 @@ define void @splatvar_funnel_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 
 define void @splatvar_funnel_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %a512, i8 %b8, <16 x i8> %b128, <32 x i8> %b256, <64 x i8> %b512, i8 %c8, <16 x i8> %c128, <32 x i8> %c256, <64 x i8> %c512) {
 ; SSSE3-LABEL: 'splatvar_funnel_i8'
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:18 Lat:29 SizeLat:25 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:48 CodeSize:35 Lat:57 SizeLat:49 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:96 CodeSize:69 Lat:113 SizeLat:97 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSE42-LABEL: 'splatvar_funnel_i8'
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:16 Lat:29 SizeLat:24 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:48 CodeSize:31 Lat:57 SizeLat:47 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:96 CodeSize:61 Lat:113 SizeLat:93 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'splatvar_funnel_i8'
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:16 Lat:18 SizeLat:22 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:27 CodeSize:36 Lat:26 SizeLat:47 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:54 CodeSize:72 Lat:52 SizeLat:94 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'splatvar_funnel_i8'
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:15 Lat:20 SizeLat:21 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:18 Lat:21 SizeLat:27 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:36 Lat:42 SizeLat:54 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512F-LABEL: 'splatvar_funnel_i8'
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:15 Lat:19 SizeLat:20 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:24 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:73 Lat:56 SizeLat:85 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512BW-LABEL: 'splatvar_funnel_i8'
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:15 Lat:20 SizeLat:20 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:22 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:23 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512DQ-LABEL: 'splatvar_funnel_i8'
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:15 Lat:19 SizeLat:20 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:24 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:73 Lat:56 SizeLat:85 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512VBMI2-LABEL: 'splatvar_funnel_i8'
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:15 Lat:20 SizeLat:20 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:22 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:23 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SLM-LABEL: 'splatvar_funnel_i8'
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:26 CodeSize:16 Lat:31 SizeLat:25 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:52 CodeSize:31 Lat:61 SizeLat:49 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:104 CodeSize:61 Lat:121 SizeLat:97 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; GLM-LABEL: 'splatvar_funnel_i8'
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; GLM-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:16 Lat:29 SizeLat:24 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:48 CodeSize:31 Lat:57 SizeLat:47 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:96 CodeSize:61 Lat:113 SizeLat:93 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'splatvar_funnel_i8'
-; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; XOP-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:7 Lat:12 SizeLat:8 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:23 CodeSize:25 Lat:23 SizeLat:31 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:46 CodeSize:50 Lat:46 SizeLat:62 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512GFNI-LABEL: 'splatvar_funnel_i8'
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:15 Lat:20 SizeLat:20 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %b128, <16 x i8> %u128)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:22 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %b256, <32 x i8> %u256)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:13 CodeSize:18 Lat:20 SizeLat:23 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %b512, <64 x i8> %u512)
@@ -1934,8 +1934,8 @@ define void @splatvar_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX1-LABEL: 'splatvar_rotate_i64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:9 SizeLat:7 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:15 Lat:18 SizeLat:22 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:30 Lat:36 SizeLat:44 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
@@ -1943,8 +1943,8 @@ define void @splatvar_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX2-LABEL: 'splatvar_rotate_i64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:7 SizeLat:7 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:5 Lat:11 SizeLat:10 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:10 Lat:22 SizeLat:20 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
@@ -1952,8 +1952,8 @@ define void @splatvar_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; AVX512-LABEL: 'splatvar_rotate_i64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
@@ -1979,8 +1979,8 @@ define void @splatvar_rotate_i64(i64 %a64, <2 x i64> %a128, <4 x i64> %a256, <8
 ;
 ; XOP-LABEL: 'splatvar_rotate_i64'
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <2 x i64> %c128, <2 x i64> undef, <2 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <4 x i64> %c256, <4 x i64> undef, <4 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <8 x i64> %c512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:3 SizeLat:3 for: %V2I64 = call <2 x i64> @llvm.fshr.v2i64(<2 x i64> %a128, <2 x i64> %a128, <2 x i64> %u128)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:7 SizeLat:9 for: %V4I64 = call <4 x i64> @llvm.fshr.v4i64(<4 x i64> %a256, <4 x i64> %a256, <4 x i64> %u256)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:16 Lat:14 SizeLat:18 for: %V8I64 = call <8 x i64> @llvm.fshr.v8i64(<8 x i64> %a512, <8 x i64> %a512, <8 x i64> %u512)
@@ -2007,8 +2007,8 @@ define void @splatvar_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX1-LABEL: 'splatvar_rotate_i32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:9 SizeLat:7 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:17 Lat:18 SizeLat:24 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:34 Lat:36 SizeLat:48 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
@@ -2016,8 +2016,8 @@ define void @splatvar_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX2-LABEL: 'splatvar_rotate_i32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:7 SizeLat:7 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:11 SizeLat:12 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:14 Lat:22 SizeLat:24 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
@@ -2025,8 +2025,8 @@ define void @splatvar_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; AVX512-LABEL: 'splatvar_rotate_i32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
@@ -2052,8 +2052,8 @@ define void @splatvar_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 ;
 ; XOP-LABEL: 'splatvar_rotate_i32'
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <4 x i32> %c128, <4 x i32> undef, <4 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u256 = shufflevector <8 x i32> %c256, <8 x i32> undef, <8 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %u512 = shufflevector <16 x i32> %c512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:3 SizeLat:3 for: %V2I32 = call <4 x i32> @llvm.fshr.v4i32(<4 x i32> %a128, <4 x i32> %a128, <4 x i32> %u128)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:7 SizeLat:9 for: %V4I32 = call <8 x i32> @llvm.fshr.v8i32(<8 x i32> %a256, <8 x i32> %a256, <8 x i32> %u256)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:16 Lat:14 SizeLat:18 for: %V8I32 = call <16 x i32> @llvm.fshr.v16i32(<16 x i32> %a512, <16 x i32> %a512, <16 x i32> %u512)
@@ -2070,99 +2070,99 @@ define void @splatvar_rotate_i32(i32 %a32, <4 x i32> %a128, <8 x i32> %a256, <16
 
 define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <32 x i16> %a512, i16 %c16, <8 x i16> %c128, <16 x i16> %c256, <32 x i16> %c512) {
 ; SSE-LABEL: 'splatvar_rotate_i16'
-; SSE-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SSE-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:5 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; SSE-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:9 Lat:13 SizeLat:13 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; SSE-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:17 Lat:25 SizeLat:25 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'splatvar_rotate_i16'
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:9 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:17 Lat:18 SizeLat:24 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:34 Lat:36 SizeLat:48 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'splatvar_rotate_i16'
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:11 SizeLat:12 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:14 Lat:22 SizeLat:24 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512F-LABEL: 'splatvar_rotate_i16'
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:11 SizeLat:10 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:17 Lat:29 SizeLat:21 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512BW-LABEL: 'splatvar_rotate_i16'
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:6 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:6 Lat:8 SizeLat:7 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:6 Lat:8 SizeLat:8 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512DQ-LABEL: 'splatvar_rotate_i16'
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:7 Lat:11 SizeLat:10 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:17 Lat:29 SizeLat:21 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512VBMI2-LABEL: 'splatvar_rotate_i16'
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SLM-LABEL: 'splatvar_rotate_i16'
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:5 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:9 Lat:13 SizeLat:13 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:17 Lat:25 SizeLat:25 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; GLM-LABEL: 'splatvar_rotate_i16'
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; GLM-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:5 Lat:7 SizeLat:7 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:9 Lat:13 SizeLat:13 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:28 CodeSize:17 Lat:25 SizeLat:25 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'splatvar_rotate_i16'
-; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 3 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 3 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:3 SizeLat:3 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:7 SizeLat:9 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:16 Lat:14 SizeLat:18 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512GFNI-LABEL: 'splatvar_rotate_i16'
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <8 x i16> %c128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <16 x i16> %c256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <32 x i16> %c512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V8I16 = call <8 x i16> @llvm.fshr.v8i16(<8 x i16> %a128, <8 x i16> %a128, <8 x i16> %u128)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V16I16 = call <16 x i16> @llvm.fshr.v16i16(<16 x i16> %a256, <16 x i16> %a256, <16 x i16> %u256)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %V32I16 = call <32 x i16> @llvm.fshr.v32i16(<32 x i16> %a512, <32 x i16> %a512, <32 x i16> %u512)
@@ -2179,99 +2179,99 @@ define void @splatvar_rotate_i16(i16 %a16, <8 x i16> %a128, <16 x i16> %a256, <3
 
 define void @splatvar_rotate_i8(i8 %a8, <16 x i8> %a128, <32 x i8> %a256, <64 x i8> %a512, i8 %c8, <16 x i8> %c128, <32 x i8> %c256, <64 x i8> %c512) {
 ; SSE-LABEL: 'splatvar_rotate_i8'
-; SSE-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SSE-NEXT:  Cost Model: Found costs of RThru:21 CodeSize:14 Lat:26 SizeLat:21 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; SSE-NEXT:  Cost Model: Found costs of RThru:42 CodeSize:27 Lat:51 SizeLat:41 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; SSE-NEXT:  Cost Model: Found costs of RThru:84 CodeSize:53 Lat:101 SizeLat:81 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'splatvar_rotate_i8'
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:14 Lat:15 SizeLat:19 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:28 Lat:21 SizeLat:38 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:40 CodeSize:56 Lat:42 SizeLat:76 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'splatvar_rotate_i8'
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:13 Lat:17 SizeLat:18 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:16 Lat:18 SizeLat:23 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:22 CodeSize:32 Lat:36 SizeLat:46 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512F-LABEL: 'splatvar_rotate_i8'
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:13 Lat:17 SizeLat:18 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:16 Lat:18 SizeLat:21 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:35 CodeSize:64 Lat:47 SizeLat:76 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512BW-LABEL: 'splatvar_rotate_i8'
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:12 Lat:6 SizeLat:14 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512DQ-LABEL: 'splatvar_rotate_i8'
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512DQ-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:13 Lat:17 SizeLat:18 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:16 Lat:18 SizeLat:21 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:35 CodeSize:64 Lat:47 SizeLat:76 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; AVX512DQ-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512VBMI2-LABEL: 'splatvar_rotate_i8'
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512VBMI2-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:12 Lat:6 SizeLat:14 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; AVX512VBMI2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SLM-LABEL: 'splatvar_rotate_i8'
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:21 CodeSize:14 Lat:26 SizeLat:21 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:42 CodeSize:27 Lat:51 SizeLat:41 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:84 CodeSize:53 Lat:101 SizeLat:81 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; GLM-LABEL: 'splatvar_rotate_i8'
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; GLM-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; GLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; GLM-NEXT:  Cost Model: Found costs of RThru:21 CodeSize:14 Lat:26 SizeLat:21 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:42 CodeSize:27 Lat:51 SizeLat:41 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:84 CodeSize:53 Lat:101 SizeLat:81 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; GLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'splatvar_rotate_i8'
-; XOP-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; XOP-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:3 Lat:3 SizeLat:3 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:7 SizeLat:9 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:16 Lat:14 SizeLat:18 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512GFNI-LABEL: 'splatvar_rotate_i8'
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u128 = shufflevector <16 x i8> %c128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u256 = shufflevector <32 x i8> %c256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %u512 = shufflevector <64 x i8> %c512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V16I8 = call <16 x i8> @llvm.fshr.v16i8(<16 x i8> %a128, <16 x i8> %a128, <16 x i8> %u128)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:6 Lat:14 SizeLat:9 for: %V32I8 = call <32 x i8> @llvm.fshr.v32i8(<32 x i8> %a256, <32 x i8> %a256, <32 x i8> %u256)
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:12 Lat:6 SizeLat:14 for: %V64I8 = call <64 x i8> @llvm.fshr.v64i8(<64 x i8> %a512, <64 x i8> %a512, <64 x i8> %u512)
diff --git a/llvm/test/Analysis/CostModel/X86/load-broadcast.ll b/llvm/test/Analysis/CostModel/X86/load-broadcast.ll
index 4cd8e5b..ecf54ce 100644
--- a/llvm/test/Analysis/CostModel/X86/load-broadcast.ll
+++ b/llvm/test/Analysis/CostModel/X86/load-broadcast.ll
@@ -1,16 +1,16 @@
 ; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 4
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mattr=+sse2 | FileCheck %s -check-prefixes=SSE,SSE2
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mattr=+ssse3 | FileCheck %s -check-prefixes=SSE,SSSE3
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mattr=+sse4.2 | FileCheck %s -check-prefixes=SSE,SSE42
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mattr=+avx | FileCheck %s -check-prefixes=AVX,AVX1
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mattr=+avx2 | FileCheck %s -check-prefixes=AVX,AVX2
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mattr=+avx512f | FileCheck %s --check-prefixes=AVX,AVX512
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=AVX,AVX512
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mattr=+avx512f,+avx512bw,+avx512vbmi | FileCheck %s --check-prefixes=AVX,AVX512
-;
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=slm | FileCheck %s --check-prefixes=SSE,SSE42
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=goldmont | FileCheck %s --check-prefixes=SSE,SSE42
-; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -mcpu=btver2 | FileCheck %s --check-prefixes=AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mattr=+sse2 | FileCheck %s -check-prefixes=SSE,SSE2
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mattr=+ssse3 | FileCheck %s -check-prefixes=SSE,SSSE3
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mattr=+sse4.2 | FileCheck %s -check-prefixes=SSE,SSE42
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mattr=+avx | FileCheck %s -check-prefixes=AVX,AVX1
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mattr=+avx2 | FileCheck %s -check-prefixes=AVX,AVX2
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mattr=+avx512f | FileCheck %s --check-prefixes=AVX,AVX512
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mattr=+avx512f,+avx512bw | FileCheck %s --check-prefixes=AVX,AVX512
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mattr=+avx512f,+avx512bw,+avx512vbmi | FileCheck %s --check-prefixes=AVX,AVX512
+;
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mcpu=slm | FileCheck %s --check-prefixes=SSE,SSE42
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mcpu=goldmont | FileCheck %s --check-prefixes=SSE,SSE42
+; RUN: opt < %s -mtriple=x86_64-unknown-linux-gnu -passes="print<cost-model>" 2>&1 -disable-output -cost-kind=all -mcpu=btver2 | FileCheck %s --check-prefixes=AVX,AVX1
 
 ;
 ; vXf64
@@ -18,24 +18,24 @@
 
 define <2 x double> @broadcast_load_v2f64_v2f64(ptr %src) {
 ; SSE2-LABEL: 'broadcast_load_v2f64_v2f64'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <2 x double>, ptr %src, align 16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <2 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %bcst
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <2 x double>, ptr %src, align 16
+; SSE2-NEXT:  Cost Model: Found costs of 1 for: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <2 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <2 x double> %bcst
 ;
 ; SSSE3-LABEL: 'broadcast_load_v2f64_v2f64'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <2 x double>, ptr %src, align 16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <2 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %bcst
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <2 x double>, ptr %src, align 16
+; SSSE3-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <2 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <2 x double> %bcst
 ;
 ; SSE42-LABEL: 'broadcast_load_v2f64_v2f64'
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <2 x double>, ptr %src, align 16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <2 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %bcst
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <2 x double>, ptr %src, align 16
+; SSE42-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <2 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <2 x double> %bcst
 ;
 ; AVX-LABEL: 'broadcast_load_v2f64_v2f64'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <2 x double>, ptr %src, align 16
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <2 x i32> zeroinitializer
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %bcst
+; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <2 x double>, ptr %src, align 16
+; AVX-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <2 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <2 x double> %bcst
 ;
   %load = load <2 x double>, ptr %src
   %bcst = shufflevector <2 x double> %load, <2 x double> poison, <2 x i32> zeroinitializer
@@ -44,24 +44,24 @@ define <2 x double> @broadcast_load_v2f64_v2f64(ptr %src) {
 
 define <4 x double> @broadcast_load_v2f64_v4f64(ptr %src) {
 ; SSE2-LABEL: 'broadcast_load_v2f64_v4f64'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <2 x double>, ptr %src, align 16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <4 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x double> %bcst
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <2 x double>, ptr %src, align 16
+; SSE2-NEXT:  Cost Model: Found costs of 1 for: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <4 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x double> %bcst
 ;
 ; SSSE3-LABEL: 'broadcast_load_v2f64_v4f64'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <2 x double>, ptr %src, align 16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <4 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x double> %bcst
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <2 x double>, ptr %src, align 16
+; SSSE3-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <4 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x double> %bcst
 ;
 ; SSE42-LABEL: 'broadcast_load_v2f64_v4f64'
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <2 x double>, ptr %src, align 16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <4 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x double> %bcst
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <2 x double>, ptr %src, align 16
+; SSE42-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <4 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x double> %bcst
 ;
 ; AVX-LABEL: 'broadcast_load_v2f64_v4f64'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <2 x double>, ptr %src, align 16
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <4 x i32> zeroinitializer
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x double> %bcst
+; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <2 x double>, ptr %src, align 16
+; AVX-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <4 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x double> %bcst
 ;
   %load = load <2 x double>, ptr %src
   %bcst = shufflevector <2 x double> %load, <2 x double> poison, <4 x i32> zeroinitializer
@@ -70,24 +70,24 @@ define <4 x double> @broadcast_load_v2f64_v4f64(ptr %src) {
 
 define <8 x double> @broadcast_load_v2f64_v8f64(ptr %src) {
 ; SSE2-LABEL: 'broadcast_load_v2f64_v8f64'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <2 x double>, ptr %src, align 16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <8 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x double> %bcst
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <2 x double>, ptr %src, align 16
+; SSE2-NEXT:  Cost Model: Found costs of 1 for: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <8 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x double> %bcst
 ;
 ; SSSE3-LABEL: 'broadcast_load_v2f64_v8f64'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <2 x double>, ptr %src, align 16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <8 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x double> %bcst
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <2 x double>, ptr %src, align 16
+; SSSE3-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <8 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x double> %bcst
 ;
 ; SSE42-LABEL: 'broadcast_load_v2f64_v8f64'
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <2 x double>, ptr %src, align 16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <8 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x double> %bcst
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <2 x double>, ptr %src, align 16
+; SSE42-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <8 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x double> %bcst
 ;
 ; AVX-LABEL: 'broadcast_load_v2f64_v8f64'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <2 x double>, ptr %src, align 16
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <8 x i32> zeroinitializer
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x double> %bcst
+; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <2 x double>, ptr %src, align 16
+; AVX-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <2 x double> %load, <2 x double> poison, <8 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x double> %bcst
 ;
   %load = load <2 x double>, ptr %src
   %bcst = shufflevector <2 x double> %load, <2 x double> poison, <8 x i32> zeroinitializer
@@ -96,16 +96,16 @@ define <8 x double> @broadcast_load_v2f64_v8f64(ptr %src) {
 
 define <2 x double> @broadcast_load_f64_v2f64(ptr %src) {
 ; SSE-LABEL: 'broadcast_load_f64_v2f64'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load double, ptr %src, align 8
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %insert = insertelement <2 x double> poison, double %load, i32 0
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <2 x double> %insert, <2 x double> poison, <2 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %bcst
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load double, ptr %src, align 8
+; SSE-NEXT:  Cost Model: Found costs of 0 for: %insert = insertelement <2 x double> poison, double %load, i32 0
+; SSE-NEXT:  Cost Model: Found costs of 1 for: %bcst = shufflevector <2 x double> %insert, <2 x double> poison, <2 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <2 x double> %bcst
 ;
 ; AVX-LABEL: 'broadcast_load_f64_v2f64'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load double, ptr %src, align 8
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %insert = insertelement <2 x double> poison, double %load, i32 0
-; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <2 x double> %insert, <2 x double> poison, <2 x i32> zeroinitializer
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x double> %bcst
+; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load double, ptr %src, align 8
+; AVX-NEXT:  Cost Model: Found costs of 0 for: %insert = insertelement <2 x double> poison, double %load, i32 0
+; AVX-NEXT:  Cost Model: Found costs of 1 for: %bcst = shufflevector <2 x double> %insert, <2 x double> poison, <2 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <2 x double> %bcst
 ;
   %load = load double, ptr %src
   %insert = insertelement <2 x double> poison, double %load, i32 0
@@ -115,28 +115,28 @@ define <2 x double> @broadcast_load_f64_v2f64(ptr %src) {
 
 define <4 x double> @broadcast_load_f64_v4f64(ptr %src) {
 ; SSE-LABEL: 'broadcast_load_f64_v4f64'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load double, ptr %src, align 8
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %insert = insertelement <4 x double> poison, double %load, i32 0
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <4 x double> %insert, <4 x double> poison, <4 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x double> %bcst
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load double, ptr %src, align 8
+; SSE-NEXT:  Cost Model: Found costs of 0 for: %insert = insertelement <4 x double> poison, double %load, i32 0
+; SSE-NEXT:  Cost Model: Found costs of 1 for: %bcst = shufflevector <4 x double> %insert, <4 x double> poison, <4 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x double> %bcst
 ;
 ; AVX1-LABEL: 'broadcast_load_f64_v4f64'
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load double, ptr %src, align 8
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %insert = insertelement <4 x double> poison, double %load, i32 0
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bcst = shufflevector <4 x double> %insert, <4 x double> poison, <4 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x double> %bcst
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load double, ptr %src, align 8
+; AVX1-NEXT:  Cost Model: Found costs of 0 for: %insert = insertelement <4 x double> poison, double %load, i32 0
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %bcst = shufflevector <4 x double> %insert, <4 x double> poison, <4 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x double> %bcst
 ;
 ; AVX2-LABEL: 'broadcast_load_f64_v4f64'
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load double, ptr %src, align 8
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %insert = insertelement <4 x double> poison, double %load, i32 0
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <4 x double> %insert, <4 x double> poison, <4 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x double> %bcst
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load double, ptr %src, align 8
+; AVX2-NEXT:  Cost Model: Found costs of 0 for: %insert = insertelement <4 x double> poison, double %load, i32 0
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %bcst = shufflevector <4 x double> %insert, <4 x double> poison, <4 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x double> %bcst
 ;
 ; AVX512-LABEL: 'broadcast_load_f64_v4f64'
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load double, ptr %src, align 8
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %insert = insertelement <4 x double> poison, double %load, i32 0
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <4 x double> %insert, <4 x double> poison, <4 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x double> %bcst
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load double, ptr %src, align 8
+; AVX512-NEXT:  Cost Model: Found costs of 0 for: %insert = insertelement <4 x double> poison, double %load, i32 0
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %bcst = shufflevector <4 x double> %insert, <4 x double> poison, <4 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x double> %bcst
 ;
   %load = load double, ptr %src
   %insert = insertelement <4 x double> poison, double %load, i32 0
@@ -146,28 +146,28 @@ define <4 x double> @broadcast_load_f64_v4f64(ptr %src) {
 
 define <8 x double> @broadcast_load_f64_v8f64(ptr %src) {
 ; SSE-LABEL: 'broadcast_load_f64_v8f64'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load double, ptr %src, align 8
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %insert = insertelement <8 x double> poison, double %load, i32 0
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <8 x double> %insert, <8 x double> poison, <8 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x double> %bcst
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load double, ptr %src, align 8
+; SSE-NEXT:  Cost Model: Found costs of 0 for: %insert = insertelement <8 x double> poison, double %load, i32 0
+; SSE-NEXT:  Cost Model: Found costs of 1 for: %bcst = shufflevector <8 x double> %insert, <8 x double> poison, <8 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x double> %bcst
 ;
 ; AVX1-LABEL: 'broadcast_load_f64_v8f64'
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load double, ptr %src, align 8
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %insert = insertelement <8 x double> poison, double %load, i32 0
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bcst = shufflevector <8 x double> %insert, <8 x double> poison, <8 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x double> %bcst
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load double, ptr %src, align 8
+; AVX1-NEXT:  Cost Model: Found costs of 0 for: %insert = insertelement <8 x double> poison, double %load, i32 0
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %bcst = shufflevector <8 x double> %insert, <8 x double> poison, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x double> %bcst
 ;
 ; AVX2-LABEL: 'broadcast_load_f64_v8f64'
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load double, ptr %src, align 8
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %insert = insertelement <8 x double> poison, double %load, i32 0
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <8 x double> %insert, <8 x double> poison, <8 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x double> %bcst
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load double, ptr %src, align 8
+; AVX2-NEXT:  Cost Model: Found costs of 0 for: %insert = insertelement <8 x double> poison, double %load, i32 0
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %bcst = shufflevector <8 x double> %insert, <8 x double> poison, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x double> %bcst
 ;
 ; AVX512-LABEL: 'broadcast_load_f64_v8f64'
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load double, ptr %src, align 8
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %insert = insertelement <8 x double> poison, double %load, i32 0
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <8 x double> %insert, <8 x double> poison, <8 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x double> %bcst
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load double, ptr %src, align 8
+; AVX512-NEXT:  Cost Model: Found costs of 0 for: %insert = insertelement <8 x double> poison, double %load, i32 0
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %bcst = shufflevector <8 x double> %insert, <8 x double> poison, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x double> %bcst
 ;
   %load = load double, ptr %src
   %insert = insertelement <8 x double> poison, double %load, i32 0
@@ -181,14 +181,14 @@ define <8 x double> @broadcast_load_f64_v8f64(ptr %src) {
 
 define <4 x float> @broadcast_load_v4f32_v4f32(ptr %src) {
 ; SSE-LABEL: 'broadcast_load_v4f32_v4f32'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <4 x float>, ptr %src, align 16
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <4 x float> %load, <4 x float> poison, <4 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %bcst
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <4 x float>, ptr %src, align 16
+; SSE-NEXT:  Cost Model: Found costs of 1 for: %bcst = shufflevector <4 x float> %load, <4 x float> poison, <4 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x float> %bcst
 ;
 ; AVX-LABEL: 'broadcast_load_v4f32_v4f32'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <4 x float>, ptr %src, align 16
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <4 x float> %load, <4 x float> poison, <4 x i32> zeroinitializer
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x float> %bcst
+; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <4 x float>, ptr %src, align 16
+; AVX-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <4 x float> %load, <4 x float> poison, <4 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x float> %bcst
 ;
   %load = load <4 x float>, ptr %src
   %bcst = shufflevector <4 x float> %load, <4 x float> poison, <4 x i32> zeroinitializer
@@ -197,14 +197,14 @@ define <4 x float> @broadcast_load_v4f32_v4f32(ptr %src) {
 
 define <8 x float> @broadcast_load_v4f32_v8f32(ptr %src) {
 ; SSE-LABEL: 'broadcast_load_v4f32_v8f32'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <4 x float>, ptr %src, align 16
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <4 x float> %load, <4 x float> poison, <8 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x float> %bcst
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <4 x float>, ptr %src, align 16
+; SSE-NEXT:  Cost Model: Found costs of 1 for: %bcst = shufflevector <4 x float> %load, <4 x float> poison, <8 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x float> %bcst
 ;
 ; AVX-LABEL: 'broadcast_load_v4f32_v8f32'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <4 x float>, ptr %src, align 16
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <4 x float> %load, <4 x float> poison, <8 x i32> zeroinitializer
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x float> %bcst
+; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <4 x float>, ptr %src, align 16
+; AVX-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <4 x float> %load, <4 x float> poison, <8 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x float> %bcst
 ;
   %load = load <4 x float>, ptr %src
   %bcst = shufflevector <4 x float> %load, <4 x float> poison, <8 x i32> zeroinitializer
@@ -213,14 +213,14 @@ define <8 x float> @broadcast_load_v4f32_v8f32(ptr %src) {
 
 define <16 x float> @broadcast_load_v4f32_v16f32(ptr %src) {
 ; SSE-LABEL: 'broadcast_load_v4f32_v16f32'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <4 x float>, ptr %src, align 16
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <4 x float> %load, <4 x float> poison, <16 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %bcst
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <4 x float>, ptr %src, align 16
+; SSE-NEXT:  Cost Model: Found costs of 1 for: %bcst = shufflevector <4 x float> %load, <4 x float> poison, <16 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x float> %bcst
 ;
 ; AVX-LABEL: 'broadcast_load_v4f32_v16f32'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <4 x float>, ptr %src, align 16
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <4 x float> %load, <4 x float> poison, <16 x i32> zeroinitializer
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x float> %bcst
+; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <4 x float>, ptr %src, align 16
+; AVX-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <4 x float> %load, <4 x float> poison, <16 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x float> %bcst
 ;
   %load = load <4 x float>, ptr %src
   %bcst = shufflevector <4 x float> %load, <4 x float> poison, <16 x i32> zeroinitializer
@@ -229,14 +229,14 @@ define <16 x float> @broadcast_load_v4f32_v16f32(ptr %src) {
 
 define <2 x i64> @broadcast_load_v2i64_v2i64(ptr %src) {
 ; SSE-LABEL: 'broadcast_load_v2i64_v2i64'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <2 x i64>, ptr %src, align 16
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <2 x i64> %load, <2 x i64> poison, <2 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %bcst
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <2 x i64>, ptr %src, align 16
+; SSE-NEXT:  Cost Model: Found costs of 1 for: %bcst = shufflevector <2 x i64> %load, <2 x i64> poison, <2 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <2 x i64> %bcst
 ;
 ; AVX-LABEL: 'broadcast_load_v2i64_v2i64'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <2 x i64>, ptr %src, align 16
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <2 x i64> %load, <2 x i64> poison, <2 x i32> zeroinitializer
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <2 x i64> %bcst
+; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <2 x i64>, ptr %src, align 16
+; AVX-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <2 x i64> %load, <2 x i64> poison, <2 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <2 x i64> %bcst
 ;
   %load = load <2 x i64>, ptr %src
   %bcst = shufflevector <2 x i64> %load, <2 x i64> poison, <2 x i32> zeroinitializer
@@ -249,14 +249,14 @@ define <2 x i64> @broadcast_load_v2i64_v2i64(ptr %src) {
 
 define <4 x i64> @broadcast_load_v2i64_v4i64(ptr %src) {
 ; SSE-LABEL: 'broadcast_load_v2i64_v4i64'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <2 x i64>, ptr %src, align 16
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <2 x i64> %load, <2 x i64> poison, <4 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %bcst
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <2 x i64>, ptr %src, align 16
+; SSE-NEXT:  Cost Model: Found costs of 1 for: %bcst = shufflevector <2 x i64> %load, <2 x i64> poison, <4 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %bcst
 ;
 ; AVX-LABEL: 'broadcast_load_v2i64_v4i64'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <2 x i64>, ptr %src, align 16
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <2 x i64> %load, <2 x i64> poison, <4 x i32> zeroinitializer
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i64> %bcst
+; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <2 x i64>, ptr %src, align 16
+; AVX-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <2 x i64> %load, <2 x i64> poison, <4 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %bcst
 ;
   %load = load <2 x i64>, ptr %src
   %bcst = shufflevector <2 x i64> %load, <2 x i64> poison, <4 x i32> zeroinitializer
@@ -265,14 +265,14 @@ define <4 x i64> @broadcast_load_v2i64_v4i64(ptr %src) {
 
 define <8 x i64> @broadcast_load_v2i64_v8i64(ptr %src) {
 ; SSE-LABEL: 'broadcast_load_v2i64_v8i64'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <2 x i64>, ptr %src, align 16
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <2 x i64> %load, <2 x i64> poison, <8 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %bcst
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <2 x i64>, ptr %src, align 16
+; SSE-NEXT:  Cost Model: Found costs of 1 for: %bcst = shufflevector <2 x i64> %load, <2 x i64> poison, <8 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %bcst
 ;
 ; AVX-LABEL: 'broadcast_load_v2i64_v8i64'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <2 x i64>, ptr %src, align 16
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <2 x i64> %load, <2 x i64> poison, <8 x i32> zeroinitializer
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i64> %bcst
+; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <2 x i64>, ptr %src, align 16
+; AVX-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <2 x i64> %load, <2 x i64> poison, <8 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %bcst
 ;
   %load = load <2 x i64>, ptr %src
   %bcst = shufflevector <2 x i64> %load, <2 x i64> poison, <8 x i32> zeroinitializer
@@ -285,14 +285,14 @@ define <8 x i64> @broadcast_load_v2i64_v8i64(ptr %src) {
 
 define <4 x i32> @broadcast_load_v4i32_v4i32(ptr %src) {
 ; SSE-LABEL: 'broadcast_load_v4i32_v4i32'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <4 x i32>, ptr %src, align 16
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <4 x i32> %load, <4 x i32> poison, <4 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %bcst
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <4 x i32>, ptr %src, align 16
+; SSE-NEXT:  Cost Model: Found costs of 1 for: %bcst = shufflevector <4 x i32> %load, <4 x i32> poison, <4 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i32> %bcst
 ;
 ; AVX-LABEL: 'broadcast_load_v4i32_v4i32'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <4 x i32>, ptr %src, align 16
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <4 x i32> %load, <4 x i32> poison, <4 x i32> zeroinitializer
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <4 x i32> %bcst
+; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <4 x i32>, ptr %src, align 16
+; AVX-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <4 x i32> %load, <4 x i32> poison, <4 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i32> %bcst
 ;
   %load = load <4 x i32>, ptr %src
   %bcst = shufflevector <4 x i32> %load, <4 x i32> poison, <4 x i32> zeroinitializer
@@ -301,14 +301,14 @@ define <4 x i32> @broadcast_load_v4i32_v4i32(ptr %src) {
 
 define <8 x i32> @broadcast_load_v4i32_v8i32(ptr %src) {
 ; SSE-LABEL: 'broadcast_load_v4i32_v8i32'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <4 x i32>, ptr %src, align 16
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <4 x i32> %load, <4 x i32> poison, <8 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %bcst
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <4 x i32>, ptr %src, align 16
+; SSE-NEXT:  Cost Model: Found costs of 1 for: %bcst = shufflevector <4 x i32> %load, <4 x i32> poison, <8 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %bcst
 ;
 ; AVX-LABEL: 'broadcast_load_v4i32_v8i32'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <4 x i32>, ptr %src, align 16
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <4 x i32> %load, <4 x i32> poison, <8 x i32> zeroinitializer
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i32> %bcst
+; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <4 x i32>, ptr %src, align 16
+; AVX-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <4 x i32> %load, <4 x i32> poison, <8 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %bcst
 ;
   %load = load <4 x i32>, ptr %src
   %bcst = shufflevector <4 x i32> %load, <4 x i32> poison, <8 x i32> zeroinitializer
@@ -317,14 +317,14 @@ define <8 x i32> @broadcast_load_v4i32_v8i32(ptr %src) {
 
 define <16 x i32> @broadcast_load_v4i32_v16i32(ptr %src) {
 ; SSE-LABEL: 'broadcast_load_v4i32_v16i32'
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <4 x i32>, ptr %src, align 16
-; SSE-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <4 x i32> %load, <4 x i32> poison, <16 x i32> zeroinitializer
-; SSE-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %bcst
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <4 x i32>, ptr %src, align 16
+; SSE-NEXT:  Cost Model: Found costs of 1 for: %bcst = shufflevector <4 x i32> %load, <4 x i32> poison, <16 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %bcst
 ;
 ; AVX-LABEL: 'broadcast_load_v4i32_v16i32'
-; AVX-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <4 x i32>, ptr %src, align 16
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <4 x i32> %load, <4 x i32> poison, <16 x i32> zeroinitializer
-; AVX-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i32> %bcst
+; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <4 x i32>, ptr %src, align 16
+; AVX-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <4 x i32> %load, <4 x i32> poison, <16 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %bcst
 ;
   %load = load <4 x i32>, ptr %src
   %bcst = shufflevector <4 x i32> %load, <4 x i32> poison, <16 x i32> zeroinitializer
@@ -337,34 +337,34 @@ define <16 x i32> @broadcast_load_v4i32_v16i32(ptr %src) {
 
 define <8 x i16> @broadcast_load_v8i16_v8i16(ptr %src) {
 ; SSE2-LABEL: 'broadcast_load_v8i16_v8i16'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <8 x i16>, ptr %src, align 16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <8 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %bcst
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <8 x i16>, ptr %src, align 16
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <8 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %bcst
 ;
 ; SSSE3-LABEL: 'broadcast_load_v8i16_v8i16'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <8 x i16>, ptr %src, align 16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <8 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %bcst
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <8 x i16>, ptr %src, align 16
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <8 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %bcst
 ;
 ; SSE42-LABEL: 'broadcast_load_v8i16_v8i16'
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <8 x i16>, ptr %src, align 16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <8 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %bcst
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <8 x i16>, ptr %src, align 16
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <8 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %bcst
 ;
 ; AVX1-LABEL: 'broadcast_load_v8i16_v8i16'
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <8 x i16>, ptr %src, align 16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <8 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %bcst
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <8 x i16>, ptr %src, align 16
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %bcst
 ;
 ; AVX2-LABEL: 'broadcast_load_v8i16_v8i16'
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <8 x i16>, ptr %src, align 16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <8 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %bcst
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <8 x i16>, ptr %src, align 16
+; AVX2-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %bcst
 ;
 ; AVX512-LABEL: 'broadcast_load_v8i16_v8i16'
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <8 x i16>, ptr %src, align 16
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <8 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <8 x i16> %bcst
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <8 x i16>, ptr %src, align 16
+; AVX512-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %bcst
 ;
   %load = load <8 x i16>, ptr %src
   %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <8 x i32> zeroinitializer
@@ -373,34 +373,34 @@ define <8 x i16> @broadcast_load_v8i16_v8i16(ptr %src) {
 
 define <16 x i16> @broadcast_load_v8i16_v16i16(ptr %src) {
 ; SSE2-LABEL: 'broadcast_load_v8i16_v16i16'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <8 x i16>, ptr %src, align 16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <16 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %bcst
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <8 x i16>, ptr %src, align 16
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %bcst
 ;
 ; SSSE3-LABEL: 'broadcast_load_v8i16_v16i16'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <8 x i16>, ptr %src, align 16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <16 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %bcst
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <8 x i16>, ptr %src, align 16
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <16 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %bcst
 ;
 ; SSE42-LABEL: 'broadcast_load_v8i16_v16i16'
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <8 x i16>, ptr %src, align 16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <16 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %bcst
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <8 x i16>, ptr %src, align 16
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %bcst
 ;
 ; AVX1-LABEL: 'broadcast_load_v8i16_v16i16'
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <8 x i16>, ptr %src, align 16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <16 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %bcst
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <8 x i16>, ptr %src, align 16
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %bcst
 ;
 ; AVX2-LABEL: 'broadcast_load_v8i16_v16i16'
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <8 x i16>, ptr %src, align 16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <16 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %bcst
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <8 x i16>, ptr %src, align 16
+; AVX2-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %bcst
 ;
 ; AVX512-LABEL: 'broadcast_load_v8i16_v16i16'
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <8 x i16>, ptr %src, align 16
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <16 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i16> %bcst
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <8 x i16>, ptr %src, align 16
+; AVX512-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %bcst
 ;
   %load = load <8 x i16>, ptr %src
   %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <16 x i32> zeroinitializer
@@ -409,34 +409,34 @@ define <16 x i16> @broadcast_load_v8i16_v16i16(ptr %src) {
 
 define <32 x i16> @broadcast_load_v8i16_v32i16(ptr %src) {
 ; SSE2-LABEL: 'broadcast_load_v8i16_v32i16'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <8 x i16>, ptr %src, align 16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <32 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %bcst
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <8 x i16>, ptr %src, align 16
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %bcst
 ;
 ; SSSE3-LABEL: 'broadcast_load_v8i16_v32i16'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <8 x i16>, ptr %src, align 16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <32 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %bcst
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <8 x i16>, ptr %src, align 16
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <32 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %bcst
 ;
 ; SSE42-LABEL: 'broadcast_load_v8i16_v32i16'
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <8 x i16>, ptr %src, align 16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <32 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %bcst
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <8 x i16>, ptr %src, align 16
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %bcst
 ;
 ; AVX1-LABEL: 'broadcast_load_v8i16_v32i16'
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <8 x i16>, ptr %src, align 16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <32 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %bcst
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <8 x i16>, ptr %src, align 16
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %bcst
 ;
 ; AVX2-LABEL: 'broadcast_load_v8i16_v32i16'
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <8 x i16>, ptr %src, align 16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <32 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %bcst
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <8 x i16>, ptr %src, align 16
+; AVX2-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %bcst
 ;
 ; AVX512-LABEL: 'broadcast_load_v8i16_v32i16'
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <8 x i16>, ptr %src, align 16
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <32 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i16> %bcst
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <8 x i16>, ptr %src, align 16
+; AVX512-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <32 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %bcst
 ;
   %load = load <8 x i16>, ptr %src
   %bcst = shufflevector <8 x i16> %load, <8 x i16> poison, <32 x i32> zeroinitializer
@@ -449,34 +449,34 @@ define <32 x i16> @broadcast_load_v8i16_v32i16(ptr %src) {
 
 define <16 x i8> @broadcast_load_v16i8_v16i8(ptr %src) {
 ; SSE2-LABEL: 'broadcast_load_v16i8_v16i8'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <16 x i8>, ptr %src, align 16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <16 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %bcst
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <16 x i8>, ptr %src, align 16
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %bcst
 ;
 ; SSSE3-LABEL: 'broadcast_load_v16i8_v16i8'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <16 x i8>, ptr %src, align 16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <16 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %bcst
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <16 x i8>, ptr %src, align 16
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <16 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %bcst
 ;
 ; SSE42-LABEL: 'broadcast_load_v16i8_v16i8'
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <16 x i8>, ptr %src, align 16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <16 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %bcst
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <16 x i8>, ptr %src, align 16
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %bcst
 ;
 ; AVX1-LABEL: 'broadcast_load_v16i8_v16i8'
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <16 x i8>, ptr %src, align 16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <16 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %bcst
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <16 x i8>, ptr %src, align 16
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %bcst
 ;
 ; AVX2-LABEL: 'broadcast_load_v16i8_v16i8'
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <16 x i8>, ptr %src, align 16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <16 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %bcst
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <16 x i8>, ptr %src, align 16
+; AVX2-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %bcst
 ;
 ; AVX512-LABEL: 'broadcast_load_v16i8_v16i8'
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <16 x i8>, ptr %src, align 16
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <16 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <16 x i8> %bcst
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <16 x i8>, ptr %src, align 16
+; AVX512-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %bcst
 ;
   %load = load <16 x i8>, ptr %src
   %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <16 x i32> zeroinitializer
@@ -485,34 +485,34 @@ define <16 x i8> @broadcast_load_v16i8_v16i8(ptr %src) {
 
 define <32 x i8> @broadcast_load_v16i8_v32i8(ptr %src) {
 ; SSE2-LABEL: 'broadcast_load_v16i8_v32i8'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <16 x i8>, ptr %src, align 16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <32 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %bcst
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <16 x i8>, ptr %src, align 16
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %bcst
 ;
 ; SSSE3-LABEL: 'broadcast_load_v16i8_v32i8'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <16 x i8>, ptr %src, align 16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <32 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %bcst
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <16 x i8>, ptr %src, align 16
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <32 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %bcst
 ;
 ; SSE42-LABEL: 'broadcast_load_v16i8_v32i8'
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <16 x i8>, ptr %src, align 16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <32 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %bcst
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <16 x i8>, ptr %src, align 16
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %bcst
 ;
 ; AVX1-LABEL: 'broadcast_load_v16i8_v32i8'
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <16 x i8>, ptr %src, align 16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <32 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %bcst
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <16 x i8>, ptr %src, align 16
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %bcst
 ;
 ; AVX2-LABEL: 'broadcast_load_v16i8_v32i8'
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <16 x i8>, ptr %src, align 16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <32 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %bcst
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <16 x i8>, ptr %src, align 16
+; AVX2-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %bcst
 ;
 ; AVX512-LABEL: 'broadcast_load_v16i8_v32i8'
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <16 x i8>, ptr %src, align 16
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <32 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <32 x i8> %bcst
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <16 x i8>, ptr %src, align 16
+; AVX512-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <32 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %bcst
 ;
   %load = load <16 x i8>, ptr %src
   %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <32 x i32> zeroinitializer
@@ -521,34 +521,34 @@ define <32 x i8> @broadcast_load_v16i8_v32i8(ptr %src) {
 
 define <64 x i8> @broadcast_load_v16i8_v64i8(ptr %src) {
 ; SSE2-LABEL: 'broadcast_load_v16i8_v64i8'
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <16 x i8>, ptr %src, align 16
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 3 for instruction: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <64 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %bcst
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <16 x i8>, ptr %src, align 16
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <64 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %bcst
 ;
 ; SSSE3-LABEL: 'broadcast_load_v16i8_v64i8'
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <16 x i8>, ptr %src, align 16
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <64 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %bcst
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <16 x i8>, ptr %src, align 16
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <64 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %bcst
 ;
 ; SSE42-LABEL: 'broadcast_load_v16i8_v64i8'
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <16 x i8>, ptr %src, align 16
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <64 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %bcst
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <16 x i8>, ptr %src, align 16
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <64 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %bcst
 ;
 ; AVX1-LABEL: 'broadcast_load_v16i8_v64i8'
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <16 x i8>, ptr %src, align 16
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 2 for instruction: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <64 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %bcst
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <16 x i8>, ptr %src, align 16
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <64 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %bcst
 ;
 ; AVX2-LABEL: 'broadcast_load_v16i8_v64i8'
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <16 x i8>, ptr %src, align 16
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <64 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %bcst
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <16 x i8>, ptr %src, align 16
+; AVX2-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <64 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %bcst
 ;
 ; AVX512-LABEL: 'broadcast_load_v16i8_v64i8'
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 1 for instruction: %load = load <16 x i8>, ptr %src, align 16
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <64 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found an estimated cost of 0 for instruction: ret <64 x i8> %bcst
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %load = load <16 x i8>, ptr %src, align 16
+; AVX512-NEXT:  Cost Model: Found costs of 0 for: %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <64 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %bcst
 ;
   %load = load <16 x i8>, ptr %src
   %bcst = shufflevector <16 x i8> %load, <16 x i8> poison, <64 x i32> zeroinitializer
diff --git a/llvm/test/Analysis/CostModel/X86/masked-intrinsic-cost-inseltpoison.ll b/llvm/test/Analysis/CostModel/X86/masked-intrinsic-cost-inseltpoison.ll
index 3031629..aea7cc8 100644
--- a/llvm/test/Analysis/CostModel/X86/masked-intrinsic-cost-inseltpoison.ll
+++ b/llvm/test/Analysis/CostModel/X86/masked-intrinsic-cost-inseltpoison.ll
@@ -2032,7 +2032,7 @@ define <16 x float> @test_gather_16f32_const_mask2(ptr %base, <16 x i32> %ind) {
 ;
 ; AVX1-LABEL: 'test_gather_16f32_const_mask2'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splatinsert = insertelement <16 x ptr> poison, ptr %base, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> poison, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> poison, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:1 Lat:1 SizeLat:1 for: %sext_ind = sext <16 x i32> %ind to <16 x i64>
 ; AVX1-NEXT:  Cost Model: Found costs of 0 for: %gep.random = getelementptr float, <16 x ptr> %broadcast.splat, <16 x i64> %sext_ind
 ; AVX1-NEXT:  Cost Model: Found costs of 50 for: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0(<16 x ptr> %gep.random, i32 4, <16 x i1> splat (i1 true), <16 x float> undef)
@@ -2040,7 +2040,7 @@ define <16 x float> @test_gather_16f32_const_mask2(ptr %base, <16 x i32> %ind) {
 ;
 ; AVX2-LABEL: 'test_gather_16f32_const_mask2'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splatinsert = insertelement <16 x ptr> poison, ptr %base, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> poison, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> poison, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:1 Lat:1 SizeLat:1 for: %sext_ind = sext <16 x i32> %ind to <16 x i64>
 ; AVX2-NEXT:  Cost Model: Found costs of 0 for: %gep.random = getelementptr float, <16 x ptr> %broadcast.splat, <16 x i64> %sext_ind
 ; AVX2-NEXT:  Cost Model: Found costs of 50 for: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0(<16 x ptr> %gep.random, i32 4, <16 x i1> splat (i1 true), <16 x float> undef)
@@ -2048,7 +2048,7 @@ define <16 x float> @test_gather_16f32_const_mask2(ptr %base, <16 x i32> %ind) {
 ;
 ; SKL-LABEL: 'test_gather_16f32_const_mask2'
 ; SKL-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splatinsert = insertelement <16 x ptr> poison, ptr %base, i32 0
-; SKL-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> poison, <16 x i32> zeroinitializer
+; SKL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> poison, <16 x i32> zeroinitializer
 ; SKL-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:1 Lat:1 SizeLat:1 for: %sext_ind = sext <16 x i32> %ind to <16 x i64>
 ; SKL-NEXT:  Cost Model: Found costs of 0 for: %gep.random = getelementptr float, <16 x ptr> %broadcast.splat, <16 x i64> %sext_ind
 ; SKL-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:4 Lat:24 SizeLat:24 for: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0(<16 x ptr> %gep.random, i32 4, <16 x i1> splat (i1 true), <16 x float> undef)
@@ -2056,7 +2056,7 @@ define <16 x float> @test_gather_16f32_const_mask2(ptr %base, <16 x i32> %ind) {
 ;
 ; AVX512-LABEL: 'test_gather_16f32_const_mask2'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splatinsert = insertelement <16 x ptr> poison, ptr %base, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> poison, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> poison, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:1 Lat:1 SizeLat:1 for: %sext_ind = sext <16 x i32> %ind to <16 x i64>
 ; AVX512-NEXT:  Cost Model: Found costs of 0 for: %gep.random = getelementptr float, <16 x ptr> %broadcast.splat, <16 x i64> %sext_ind
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:1 Lat:18 SizeLat:18 for: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0(<16 x ptr> %gep.random, i32 4, <16 x i1> splat (i1 true), <16 x float> undef)
@@ -2091,7 +2091,7 @@ define void @test_scatter_16i32(ptr %base, <16 x i32> %ind, i16 %mask, <16 x i32
 ;
 ; AVX1-LABEL: 'test_scatter_16i32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splatinsert = insertelement <16 x ptr> poison, ptr %base, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> poison, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> poison, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of 0 for: %gep.random = getelementptr i32, <16 x ptr> %broadcast.splat, <16 x i32> %ind
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %imask = bitcast i16 %mask to <16 x i1>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:55 CodeSize:71 Lat:71 SizeLat:71 for: call void @llvm.masked.scatter.v16i32.v16p0(<16 x i32> %val, <16 x ptr> %gep.random, i32 4, <16 x i1> %imask)
@@ -2099,7 +2099,7 @@ define void @test_scatter_16i32(ptr %base, <16 x i32> %ind, i16 %mask, <16 x i32
 ;
 ; AVX2-LABEL: 'test_scatter_16i32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splatinsert = insertelement <16 x ptr> poison, ptr %base, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> poison, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> poison, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of 0 for: %gep.random = getelementptr i32, <16 x ptr> %broadcast.splat, <16 x i32> %ind
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %imask = bitcast i16 %mask to <16 x i1>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:55 CodeSize:71 Lat:71 SizeLat:71 for: call void @llvm.masked.scatter.v16i32.v16p0(<16 x i32> %val, <16 x ptr> %gep.random, i32 4, <16 x i1> %imask)
@@ -2107,7 +2107,7 @@ define void @test_scatter_16i32(ptr %base, <16 x i32> %ind, i16 %mask, <16 x i32
 ;
 ; SKL-LABEL: 'test_scatter_16i32'
 ; SKL-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splatinsert = insertelement <16 x ptr> poison, ptr %base, i32 0
-; SKL-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> poison, <16 x i32> zeroinitializer
+; SKL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> poison, <16 x i32> zeroinitializer
 ; SKL-NEXT:  Cost Model: Found costs of 0 for: %gep.random = getelementptr i32, <16 x ptr> %broadcast.splat, <16 x i32> %ind
 ; SKL-NEXT:  Cost Model: Found costs of 1 for: %imask = bitcast i16 %mask to <16 x i1>
 ; SKL-NEXT:  Cost Model: Found costs of RThru:55 CodeSize:71 Lat:71 SizeLat:71 for: call void @llvm.masked.scatter.v16i32.v16p0(<16 x i32> %val, <16 x ptr> %gep.random, i32 4, <16 x i1> %imask)
@@ -2115,7 +2115,7 @@ define void @test_scatter_16i32(ptr %base, <16 x i32> %ind, i16 %mask, <16 x i32
 ;
 ; AVX512-LABEL: 'test_scatter_16i32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splatinsert = insertelement <16 x ptr> poison, ptr %base, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> poison, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> poison, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of 0 for: %gep.random = getelementptr i32, <16 x ptr> %broadcast.splat, <16 x i32> %ind
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %imask = bitcast i16 %mask to <16 x i1>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:1 Lat:18 SizeLat:18 for: call void @llvm.masked.scatter.v16i32.v16p0(<16 x i32> %val, <16 x ptr> %gep.random, i32 4, <16 x i1> %imask)
diff --git a/llvm/test/Analysis/CostModel/X86/masked-intrinsic-cost.ll b/llvm/test/Analysis/CostModel/X86/masked-intrinsic-cost.ll
index 8fae37b3..742b5b2 100644
--- a/llvm/test/Analysis/CostModel/X86/masked-intrinsic-cost.ll
+++ b/llvm/test/Analysis/CostModel/X86/masked-intrinsic-cost.ll
@@ -2032,7 +2032,7 @@ define <16 x float> @test_gather_16f32_const_mask2(ptr %base, <16 x i32> %ind) {
 ;
 ; AVX1-LABEL: 'test_gather_16f32_const_mask2'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splatinsert = insertelement <16 x ptr> undef, ptr %base, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> undef, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:1 Lat:1 SizeLat:1 for: %sext_ind = sext <16 x i32> %ind to <16 x i64>
 ; AVX1-NEXT:  Cost Model: Found costs of 0 for: %gep.random = getelementptr float, <16 x ptr> %broadcast.splat, <16 x i64> %sext_ind
 ; AVX1-NEXT:  Cost Model: Found costs of 50 for: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0(<16 x ptr> %gep.random, i32 4, <16 x i1> splat (i1 true), <16 x float> undef)
@@ -2040,7 +2040,7 @@ define <16 x float> @test_gather_16f32_const_mask2(ptr %base, <16 x i32> %ind) {
 ;
 ; AVX2-LABEL: 'test_gather_16f32_const_mask2'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splatinsert = insertelement <16 x ptr> undef, ptr %base, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:1 Lat:1 SizeLat:1 for: %sext_ind = sext <16 x i32> %ind to <16 x i64>
 ; AVX2-NEXT:  Cost Model: Found costs of 0 for: %gep.random = getelementptr float, <16 x ptr> %broadcast.splat, <16 x i64> %sext_ind
 ; AVX2-NEXT:  Cost Model: Found costs of 50 for: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0(<16 x ptr> %gep.random, i32 4, <16 x i1> splat (i1 true), <16 x float> undef)
@@ -2048,7 +2048,7 @@ define <16 x float> @test_gather_16f32_const_mask2(ptr %base, <16 x i32> %ind) {
 ;
 ; SKL-LABEL: 'test_gather_16f32_const_mask2'
 ; SKL-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splatinsert = insertelement <16 x ptr> undef, ptr %base, i32 0
-; SKL-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> undef, <16 x i32> zeroinitializer
+; SKL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> undef, <16 x i32> zeroinitializer
 ; SKL-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:1 Lat:1 SizeLat:1 for: %sext_ind = sext <16 x i32> %ind to <16 x i64>
 ; SKL-NEXT:  Cost Model: Found costs of 0 for: %gep.random = getelementptr float, <16 x ptr> %broadcast.splat, <16 x i64> %sext_ind
 ; SKL-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:4 Lat:24 SizeLat:24 for: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0(<16 x ptr> %gep.random, i32 4, <16 x i1> splat (i1 true), <16 x float> undef)
@@ -2056,7 +2056,7 @@ define <16 x float> @test_gather_16f32_const_mask2(ptr %base, <16 x i32> %ind) {
 ;
 ; AVX512-LABEL: 'test_gather_16f32_const_mask2'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splatinsert = insertelement <16 x ptr> undef, ptr %base, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> undef, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:1 Lat:1 SizeLat:1 for: %sext_ind = sext <16 x i32> %ind to <16 x i64>
 ; AVX512-NEXT:  Cost Model: Found costs of 0 for: %gep.random = getelementptr float, <16 x ptr> %broadcast.splat, <16 x i64> %sext_ind
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:1 Lat:18 SizeLat:18 for: %res = call <16 x float> @llvm.masked.gather.v16f32.v16p0(<16 x ptr> %gep.random, i32 4, <16 x i1> splat (i1 true), <16 x float> undef)
@@ -2091,7 +2091,7 @@ define void @test_scatter_16i32(ptr %base, <16 x i32> %ind, i16 %mask, <16 x i32
 ;
 ; AVX1-LABEL: 'test_scatter_16i32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splatinsert = insertelement <16 x ptr> undef, ptr %base, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> undef, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of 0 for: %gep.random = getelementptr i32, <16 x ptr> %broadcast.splat, <16 x i32> %ind
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %imask = bitcast i16 %mask to <16 x i1>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:55 CodeSize:71 Lat:71 SizeLat:71 for: call void @llvm.masked.scatter.v16i32.v16p0(<16 x i32> %val, <16 x ptr> %gep.random, i32 4, <16 x i1> %imask)
@@ -2099,7 +2099,7 @@ define void @test_scatter_16i32(ptr %base, <16 x i32> %ind, i16 %mask, <16 x i32
 ;
 ; AVX2-LABEL: 'test_scatter_16i32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splatinsert = insertelement <16 x ptr> undef, ptr %base, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of 0 for: %gep.random = getelementptr i32, <16 x ptr> %broadcast.splat, <16 x i32> %ind
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %imask = bitcast i16 %mask to <16 x i1>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:55 CodeSize:71 Lat:71 SizeLat:71 for: call void @llvm.masked.scatter.v16i32.v16p0(<16 x i32> %val, <16 x ptr> %gep.random, i32 4, <16 x i1> %imask)
@@ -2107,7 +2107,7 @@ define void @test_scatter_16i32(ptr %base, <16 x i32> %ind, i16 %mask, <16 x i32
 ;
 ; SKL-LABEL: 'test_scatter_16i32'
 ; SKL-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splatinsert = insertelement <16 x ptr> undef, ptr %base, i32 0
-; SKL-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> undef, <16 x i32> zeroinitializer
+; SKL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> undef, <16 x i32> zeroinitializer
 ; SKL-NEXT:  Cost Model: Found costs of 0 for: %gep.random = getelementptr i32, <16 x ptr> %broadcast.splat, <16 x i32> %ind
 ; SKL-NEXT:  Cost Model: Found costs of 1 for: %imask = bitcast i16 %mask to <16 x i1>
 ; SKL-NEXT:  Cost Model: Found costs of RThru:55 CodeSize:71 Lat:71 SizeLat:71 for: call void @llvm.masked.scatter.v16i32.v16p0(<16 x i32> %val, <16 x ptr> %gep.random, i32 4, <16 x i1> %imask)
@@ -2115,7 +2115,7 @@ define void @test_scatter_16i32(ptr %base, <16 x i32> %ind, i16 %mask, <16 x i32
 ;
 ; AVX512-LABEL: 'test_scatter_16i32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splatinsert = insertelement <16 x ptr> undef, ptr %base, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %broadcast.splat = shufflevector <16 x ptr> %broadcast.splatinsert, <16 x ptr> undef, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of 0 for: %gep.random = getelementptr i32, <16 x ptr> %broadcast.splat, <16 x i32> %ind
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %imask = bitcast i16 %mask to <16 x i1>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:1 Lat:18 SizeLat:18 for: call void @llvm.masked.scatter.v16i32.v16p0(<16 x i32> %val, <16 x ptr> %gep.random, i32 4, <16 x i1> %imask)
diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-broadcast-fp16.ll b/llvm/test/Analysis/CostModel/X86/shuffle-broadcast-fp16.ll
index a8ceebf..c4cf8b7 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-broadcast-fp16.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-broadcast-fp16.ll
@@ -3,11 +3,11 @@
 
 define void @test_vXf16(<2 x half> %src32, <4 x half> %src64, <8 x half> %src128, <16 x half> %src256, <32 x half> %src512) {
 ; CHECK-LABEL: 'test_vXf16'
-; CHECK-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
-; CHECK-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
-; CHECK-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
-; CHECK-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
-; CHECK-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
+; CHECK-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
+; CHECK-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
+; CHECK-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
+; CHECK-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
+; CHECK-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
 ; CHECK-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V32  = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-broadcast.ll b/llvm/test/Analysis/CostModel/X86/shuffle-broadcast.ll
index 1a2d098..79e1622 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-broadcast.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-broadcast.ll
@@ -25,20 +25,20 @@ define void @test_vXf64(<2 x double> %src128, <4 x double> %src256, <8 x double>
 ;
 ; AVX1-LABEL: 'test_vXf64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXf64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXf64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> zeroinitializer
@@ -56,20 +56,20 @@ define void @test_vXi64(<2 x i64> %src128, <4 x i64> %src256, <8 x i64> %src512)
 ;
 ; AVX1-LABEL: 'test_vXi64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXi64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> zeroinitializer
@@ -89,22 +89,22 @@ define void @test_vXf32(<2 x float> %src64, <4 x float> %src128, <8 x float> %sr
 ; AVX1-LABEL: 'test_vXf32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXf32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXf32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> zeroinitializer
@@ -125,22 +125,22 @@ define void @test_vXi32(<2 x i32> %src64, <4 x i32> %src128, <8 x i32> %src256,
 ; AVX1-LABEL: 'test_vXi32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXi32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> zeroinitializer
@@ -154,41 +154,41 @@ define void @test_vXf16(<2 x half> %src32, <4 x half> %src64, <8 x half> %src128
 ; SSE2-LABEL: 'test_vXf16'
 ; SSE2-NEXT:  Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSSE3-LABEL: 'test_vXf16'
 ; SSSE3-NEXT:  Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
 ; SSSE3-NEXT:  Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSE42-LABEL: 'test_vXf16'
 ; SSE42-NEXT:  Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXf16'
 ; AVX2-NEXT:  Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXf16'
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V32  = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> zeroinitializer
@@ -244,49 +244,49 @@ define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <
 ; SSE2-LABEL: 'test_vXi16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSSE3-LABEL: 'test_vXi16'
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSE42-LABEL: 'test_vXi16'
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXi16'
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi16'
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXi16'
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V32  = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> zeroinitializer
@@ -302,54 +302,54 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16 x
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSSE3-LABEL: 'test_vXi8'
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSE42-LABEL: 'test_vXi8'
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXi8'
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi8'
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXi8'
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> zeroinitializer
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V16  = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> zeroinitializer
@@ -365,46 +365,46 @@ define void @test_vXi1(<2 x i1> %src2, <4 x i1> %src4, <8 x i1> %src8, <16 x i1>
 ; SSE2-LABEL: 'test_vXi1'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> zeroinitializer
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSSE3-LABEL: 'test_vXi1'
 ; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> zeroinitializer
 ; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> zeroinitializer
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> zeroinitializer
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> zeroinitializer
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSE42-LABEL: 'test_vXi1'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> zeroinitializer
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXi1'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> zeroinitializer
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi1'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> zeroinitializer
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXi1'
@@ -440,22 +440,22 @@ define void @test_upper_vXf32(<2 x float> %a64, <2 x float> %b64, <4 x float> %a
 ; AVX1-LABEL: 'test_upper_vXf32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %a64, <2 x float> %b64, <2 x i32> <i32 2, i32 2>
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %a128, <4 x float> %b128, <4 x i32> <i32 4, i32 4, i32 4, i32 4>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_upper_vXf32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %a64, <2 x float> %b64, <2 x i32> <i32 2, i32 2>
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %a128, <4 x float> %b128, <4 x i32> <i32 4, i32 4, i32 4, i32 4>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_upper_vXf32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %a64, <2 x float> %b64, <2 x i32> <i32 2, i32 2>
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %a128, <4 x float> %b128, <4 x i32> <i32 4, i32 4, i32 4, i32 4>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8, i32 8>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16, i32 16>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V64 = shufflevector <2 x float> %a64, <2 x float> %b64, <2 x i32> <i32 2, i32 2>
diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-load.ll b/llvm/test/Analysis/CostModel/X86/shuffle-load.ll
index ccfb543..8bb9cbc 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-load.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-load.ll
@@ -32,21 +32,21 @@ define void @shuffle_load() {
 ; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_8xi8 = load <8 x i8>, ptr undef, align 8
 ; SSE-NEXT:  Cost Model: Found costs of 2 for: %sf_8xi8 = shufflevector <8 x i8> %ld_8xi8, <8 x i8> undef, <8 x i32> zeroinitializer
 ; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_16xi8 = load <16 x i8>, ptr undef, align 16
-; SSE-NEXT:  Cost Model: Found costs of 3 for: %sf_16xi8 = shufflevector <16 x i8> %ld_16xi8, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %sf_16xi8 = shufflevector <16 x i8> %ld_16xi8, <16 x i8> undef, <16 x i32> zeroinitializer
 ; SSE-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:1 for: %ld_32xi8 = load <32 x i8>, ptr undef, align 32
-; SSE-NEXT:  Cost Model: Found costs of 3 for: %sf_32xi8 = shufflevector <32 x i8> %ld_32xi8, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %sf_32xi8 = shufflevector <32 x i8> %ld_32xi8, <32 x i8> undef, <32 x i32> zeroinitializer
 ; SSE-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_64xi8 = load <64 x i8>, ptr undef, align 64
-; SSE-NEXT:  Cost Model: Found costs of 3 for: %sf_64xi8 = shufflevector <64 x i8> %ld_64xi8, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %sf_64xi8 = shufflevector <64 x i8> %ld_64xi8, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_2xi16 = load <2 x i16>, ptr undef, align 4
 ; SSE-NEXT:  Cost Model: Found costs of 1 for: %sf_2xi16 = shufflevector <2 x i16> %ld_2xi16, <2 x i16> undef, <2 x i32> zeroinitializer
 ; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xi16 = load <4 x i16>, ptr undef, align 8
 ; SSE-NEXT:  Cost Model: Found costs of 1 for: %sf_4xi16 = shufflevector <4 x i16> %ld_4xi16, <4 x i16> undef, <4 x i32> zeroinitializer
 ; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_8xi16 = load <8 x i16>, ptr undef, align 16
-; SSE-NEXT:  Cost Model: Found costs of 2 for: %sf_8xi16 = shufflevector <8 x i16> %ld_8xi16, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_8xi16 = shufflevector <8 x i16> %ld_8xi16, <8 x i16> undef, <8 x i32> zeroinitializer
 ; SSE-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:1 for: %ld_16xi16 = load <16 x i16>, ptr undef, align 32
-; SSE-NEXT:  Cost Model: Found costs of 2 for: %sf_16xi16 = shufflevector <16 x i16> %ld_16xi16, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_16xi16 = shufflevector <16 x i16> %ld_16xi16, <16 x i16> undef, <16 x i32> zeroinitializer
 ; SSE-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_32xi16 = load <32 x i16>, ptr undef, align 64
-; SSE-NEXT:  Cost Model: Found costs of 2 for: %sf_32xi16 = shufflevector <32 x i16> %ld_32xi16, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_32xi16 = shufflevector <32 x i16> %ld_32xi16, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_2xi32 = load <2 x i32>, ptr undef, align 8
 ; SSE-NEXT:  Cost Model: Found costs of 1 for: %sf_2xi32 = shufflevector <2 x i32> %ld_2xi32, <2 x i32> undef, <2 x i32> zeroinitializer
 ; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xi32 = load <4 x i32>, ptr undef, align 16
@@ -66,11 +66,11 @@ define void @shuffle_load() {
 ; SSE-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xf16 = load <4 x half>, ptr undef, align 8
 ; SSE-NEXT:  Cost Model: Found costs of 0 for: %sf_4xf16 = shufflevector <4 x half> %ld_4xf16, <4 x half> undef, <4 x i32> zeroinitializer
 ; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_8xf16 = load <8 x half>, ptr undef, align 16
-; SSE-NEXT:  Cost Model: Found costs of 2 for: %sf_8xf16 = shufflevector <8 x half> %ld_8xf16, <8 x half> undef, <8 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_8xf16 = shufflevector <8 x half> %ld_8xf16, <8 x half> undef, <8 x i32> zeroinitializer
 ; SSE-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:1 for: %ld_16xf16 = load <16 x half>, ptr undef, align 32
-; SSE-NEXT:  Cost Model: Found costs of 2 for: %sf_16xf16 = shufflevector <16 x half> %ld_16xf16, <16 x half> undef, <16 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_16xf16 = shufflevector <16 x half> %ld_16xf16, <16 x half> undef, <16 x i32> zeroinitializer
 ; SSE-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_32xf16 = load <32 x half>, ptr undef, align 64
-; SSE-NEXT:  Cost Model: Found costs of 2 for: %sf_32xf16 = shufflevector <32 x half> %ld_32xf16, <32 x half> undef, <32 x i32> zeroinitializer
+; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_32xf16 = shufflevector <32 x half> %ld_32xf16, <32 x half> undef, <32 x i32> zeroinitializer
 ; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_2xf32 = load <2 x float>, ptr undef, align 8
 ; SSE-NEXT:  Cost Model: Found costs of 1 for: %sf_2xf32 = shufflevector <2 x float> %ld_2xf32, <2 x float> undef, <2 x i32> zeroinitializer
 ; SSE-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xf32 = load <4 x float>, ptr undef, align 16
@@ -95,21 +95,21 @@ define void @shuffle_load() {
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_8xi8 = load <8 x i8>, ptr undef, align 8
 ; SSE2-NEXT:  Cost Model: Found costs of 2 for: %sf_8xi8 = shufflevector <8 x i8> %ld_8xi8, <8 x i8> undef, <8 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_16xi8 = load <16 x i8>, ptr undef, align 16
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %sf_16xi8 = shufflevector <16 x i8> %ld_16xi8, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %sf_16xi8 = shufflevector <16 x i8> %ld_16xi8, <16 x i8> undef, <16 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:1 for: %ld_32xi8 = load <32 x i8>, ptr undef, align 32
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %sf_32xi8 = shufflevector <32 x i8> %ld_32xi8, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %sf_32xi8 = shufflevector <32 x i8> %ld_32xi8, <32 x i8> undef, <32 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_64xi8 = load <64 x i8>, ptr undef, align 64
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %sf_64xi8 = shufflevector <64 x i8> %ld_64xi8, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %sf_64xi8 = shufflevector <64 x i8> %ld_64xi8, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_2xi16 = load <2 x i16>, ptr undef, align 4
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %sf_2xi16 = shufflevector <2 x i16> %ld_2xi16, <2 x i16> undef, <2 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xi16 = load <4 x i16>, ptr undef, align 8
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %sf_4xi16 = shufflevector <4 x i16> %ld_4xi16, <4 x i16> undef, <4 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_8xi16 = load <8 x i16>, ptr undef, align 16
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %sf_8xi16 = shufflevector <8 x i16> %ld_8xi16, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_8xi16 = shufflevector <8 x i16> %ld_8xi16, <8 x i16> undef, <8 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:1 for: %ld_16xi16 = load <16 x i16>, ptr undef, align 32
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %sf_16xi16 = shufflevector <16 x i16> %ld_16xi16, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_16xi16 = shufflevector <16 x i16> %ld_16xi16, <16 x i16> undef, <16 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_32xi16 = load <32 x i16>, ptr undef, align 64
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %sf_32xi16 = shufflevector <32 x i16> %ld_32xi16, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_32xi16 = shufflevector <32 x i16> %ld_32xi16, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_2xi32 = load <2 x i32>, ptr undef, align 8
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %sf_2xi32 = shufflevector <2 x i32> %ld_2xi32, <2 x i32> undef, <2 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xi32 = load <4 x i32>, ptr undef, align 16
@@ -129,11 +129,11 @@ define void @shuffle_load() {
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xf16 = load <4 x half>, ptr undef, align 8
 ; SSE2-NEXT:  Cost Model: Found costs of 0 for: %sf_4xf16 = shufflevector <4 x half> %ld_4xf16, <4 x half> undef, <4 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_8xf16 = load <8 x half>, ptr undef, align 16
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %sf_8xf16 = shufflevector <8 x half> %ld_8xf16, <8 x half> undef, <8 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_8xf16 = shufflevector <8 x half> %ld_8xf16, <8 x half> undef, <8 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:1 for: %ld_16xf16 = load <16 x half>, ptr undef, align 32
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %sf_16xf16 = shufflevector <16 x half> %ld_16xf16, <16 x half> undef, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_16xf16 = shufflevector <16 x half> %ld_16xf16, <16 x half> undef, <16 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_32xf16 = load <32 x half>, ptr undef, align 64
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %sf_32xf16 = shufflevector <32 x half> %ld_32xf16, <32 x half> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_32xf16 = shufflevector <32 x half> %ld_32xf16, <32 x half> undef, <32 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_2xf32 = load <2 x float>, ptr undef, align 8
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %sf_2xf32 = shufflevector <2 x float> %ld_2xf32, <2 x float> undef, <2 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xf32 = load <4 x float>, ptr undef, align 16
@@ -158,21 +158,21 @@ define void @shuffle_load() {
 ; SSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_8xi8 = load <8 x i8>, ptr undef, align 8
 ; SSE3-NEXT:  Cost Model: Found costs of 2 for: %sf_8xi8 = shufflevector <8 x i8> %ld_8xi8, <8 x i8> undef, <8 x i32> zeroinitializer
 ; SSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_16xi8 = load <16 x i8>, ptr undef, align 16
-; SSE3-NEXT:  Cost Model: Found costs of 3 for: %sf_16xi8 = shufflevector <16 x i8> %ld_16xi8, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE3-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %sf_16xi8 = shufflevector <16 x i8> %ld_16xi8, <16 x i8> undef, <16 x i32> zeroinitializer
 ; SSE3-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:1 for: %ld_32xi8 = load <32 x i8>, ptr undef, align 32
-; SSE3-NEXT:  Cost Model: Found costs of 3 for: %sf_32xi8 = shufflevector <32 x i8> %ld_32xi8, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE3-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %sf_32xi8 = shufflevector <32 x i8> %ld_32xi8, <32 x i8> undef, <32 x i32> zeroinitializer
 ; SSE3-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_64xi8 = load <64 x i8>, ptr undef, align 64
-; SSE3-NEXT:  Cost Model: Found costs of 3 for: %sf_64xi8 = shufflevector <64 x i8> %ld_64xi8, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE3-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %sf_64xi8 = shufflevector <64 x i8> %ld_64xi8, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_2xi16 = load <2 x i16>, ptr undef, align 4
 ; SSE3-NEXT:  Cost Model: Found costs of 1 for: %sf_2xi16 = shufflevector <2 x i16> %ld_2xi16, <2 x i16> undef, <2 x i32> zeroinitializer
 ; SSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xi16 = load <4 x i16>, ptr undef, align 8
 ; SSE3-NEXT:  Cost Model: Found costs of 1 for: %sf_4xi16 = shufflevector <4 x i16> %ld_4xi16, <4 x i16> undef, <4 x i32> zeroinitializer
 ; SSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_8xi16 = load <8 x i16>, ptr undef, align 16
-; SSE3-NEXT:  Cost Model: Found costs of 2 for: %sf_8xi16 = shufflevector <8 x i16> %ld_8xi16, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_8xi16 = shufflevector <8 x i16> %ld_8xi16, <8 x i16> undef, <8 x i32> zeroinitializer
 ; SSE3-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:1 for: %ld_16xi16 = load <16 x i16>, ptr undef, align 32
-; SSE3-NEXT:  Cost Model: Found costs of 2 for: %sf_16xi16 = shufflevector <16 x i16> %ld_16xi16, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_16xi16 = shufflevector <16 x i16> %ld_16xi16, <16 x i16> undef, <16 x i32> zeroinitializer
 ; SSE3-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_32xi16 = load <32 x i16>, ptr undef, align 64
-; SSE3-NEXT:  Cost Model: Found costs of 2 for: %sf_32xi16 = shufflevector <32 x i16> %ld_32xi16, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_32xi16 = shufflevector <32 x i16> %ld_32xi16, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_2xi32 = load <2 x i32>, ptr undef, align 8
 ; SSE3-NEXT:  Cost Model: Found costs of 1 for: %sf_2xi32 = shufflevector <2 x i32> %ld_2xi32, <2 x i32> undef, <2 x i32> zeroinitializer
 ; SSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xi32 = load <4 x i32>, ptr undef, align 16
@@ -192,11 +192,11 @@ define void @shuffle_load() {
 ; SSE3-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xf16 = load <4 x half>, ptr undef, align 8
 ; SSE3-NEXT:  Cost Model: Found costs of 0 for: %sf_4xf16 = shufflevector <4 x half> %ld_4xf16, <4 x half> undef, <4 x i32> zeroinitializer
 ; SSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_8xf16 = load <8 x half>, ptr undef, align 16
-; SSE3-NEXT:  Cost Model: Found costs of 2 for: %sf_8xf16 = shufflevector <8 x half> %ld_8xf16, <8 x half> undef, <8 x i32> zeroinitializer
+; SSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_8xf16 = shufflevector <8 x half> %ld_8xf16, <8 x half> undef, <8 x i32> zeroinitializer
 ; SSE3-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:1 for: %ld_16xf16 = load <16 x half>, ptr undef, align 32
-; SSE3-NEXT:  Cost Model: Found costs of 2 for: %sf_16xf16 = shufflevector <16 x half> %ld_16xf16, <16 x half> undef, <16 x i32> zeroinitializer
+; SSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_16xf16 = shufflevector <16 x half> %ld_16xf16, <16 x half> undef, <16 x i32> zeroinitializer
 ; SSE3-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_32xf16 = load <32 x half>, ptr undef, align 64
-; SSE3-NEXT:  Cost Model: Found costs of 2 for: %sf_32xf16 = shufflevector <32 x half> %ld_32xf16, <32 x half> undef, <32 x i32> zeroinitializer
+; SSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_32xf16 = shufflevector <32 x half> %ld_32xf16, <32 x half> undef, <32 x i32> zeroinitializer
 ; SSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_2xf32 = load <2 x float>, ptr undef, align 8
 ; SSE3-NEXT:  Cost Model: Found costs of 1 for: %sf_2xf32 = shufflevector <2 x float> %ld_2xf32, <2 x float> undef, <2 x i32> zeroinitializer
 ; SSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xf32 = load <4 x float>, ptr undef, align 16
@@ -221,21 +221,21 @@ define void @shuffle_load() {
 ; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_8xi8 = load <8 x i8>, ptr undef, align 8
 ; AVX-NEXT:  Cost Model: Found costs of 2 for: %sf_8xi8 = shufflevector <8 x i8> %ld_8xi8, <8 x i8> undef, <8 x i32> zeroinitializer
 ; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_16xi8 = load <16 x i8>, ptr undef, align 16
-; AVX-NEXT:  Cost Model: Found costs of 3 for: %sf_16xi8 = shufflevector <16 x i8> %ld_16xi8, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %sf_16xi8 = shufflevector <16 x i8> %ld_16xi8, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:1 for: %ld_32xi8 = load <32 x i8>, ptr undef, align 32
-; AVX-NEXT:  Cost Model: Found costs of 3 for: %sf_32xi8 = shufflevector <32 x i8> %ld_32xi8, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %sf_32xi8 = shufflevector <32 x i8> %ld_32xi8, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_64xi8 = load <64 x i8>, ptr undef, align 64
-; AVX-NEXT:  Cost Model: Found costs of 3 for: %sf_64xi8 = shufflevector <64 x i8> %ld_64xi8, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %sf_64xi8 = shufflevector <64 x i8> %ld_64xi8, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_2xi16 = load <2 x i16>, ptr undef, align 4
 ; AVX-NEXT:  Cost Model: Found costs of 1 for: %sf_2xi16 = shufflevector <2 x i16> %ld_2xi16, <2 x i16> undef, <2 x i32> zeroinitializer
 ; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xi16 = load <4 x i16>, ptr undef, align 8
 ; AVX-NEXT:  Cost Model: Found costs of 1 for: %sf_4xi16 = shufflevector <4 x i16> %ld_4xi16, <4 x i16> undef, <4 x i32> zeroinitializer
 ; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_8xi16 = load <8 x i16>, ptr undef, align 16
-; AVX-NEXT:  Cost Model: Found costs of 2 for: %sf_8xi16 = shufflevector <8 x i16> %ld_8xi16, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_8xi16 = shufflevector <8 x i16> %ld_8xi16, <8 x i16> undef, <8 x i32> zeroinitializer
 ; AVX-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:1 for: %ld_16xi16 = load <16 x i16>, ptr undef, align 32
-; AVX-NEXT:  Cost Model: Found costs of 2 for: %sf_16xi16 = shufflevector <16 x i16> %ld_16xi16, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_16xi16 = shufflevector <16 x i16> %ld_16xi16, <16 x i16> undef, <16 x i32> zeroinitializer
 ; AVX-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_32xi16 = load <32 x i16>, ptr undef, align 64
-; AVX-NEXT:  Cost Model: Found costs of 2 for: %sf_32xi16 = shufflevector <32 x i16> %ld_32xi16, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_32xi16 = shufflevector <32 x i16> %ld_32xi16, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_2xi32 = load <2 x i32>, ptr undef, align 8
 ; AVX-NEXT:  Cost Model: Found costs of 1 for: %sf_2xi32 = shufflevector <2 x i32> %ld_2xi32, <2 x i32> undef, <2 x i32> zeroinitializer
 ; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xi32 = load <4 x i32>, ptr undef, align 16
@@ -255,11 +255,11 @@ define void @shuffle_load() {
 ; AVX-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xf16 = load <4 x half>, ptr undef, align 8
 ; AVX-NEXT:  Cost Model: Found costs of 0 for: %sf_4xf16 = shufflevector <4 x half> %ld_4xf16, <4 x half> undef, <4 x i32> zeroinitializer
 ; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_8xf16 = load <8 x half>, ptr undef, align 16
-; AVX-NEXT:  Cost Model: Found costs of 2 for: %sf_8xf16 = shufflevector <8 x half> %ld_8xf16, <8 x half> undef, <8 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_8xf16 = shufflevector <8 x half> %ld_8xf16, <8 x half> undef, <8 x i32> zeroinitializer
 ; AVX-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:1 for: %ld_16xf16 = load <16 x half>, ptr undef, align 32
-; AVX-NEXT:  Cost Model: Found costs of 2 for: %sf_16xf16 = shufflevector <16 x half> %ld_16xf16, <16 x half> undef, <16 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_16xf16 = shufflevector <16 x half> %ld_16xf16, <16 x half> undef, <16 x i32> zeroinitializer
 ; AVX-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_32xf16 = load <32 x half>, ptr undef, align 64
-; AVX-NEXT:  Cost Model: Found costs of 2 for: %sf_32xf16 = shufflevector <32 x half> %ld_32xf16, <32 x half> undef, <32 x i32> zeroinitializer
+; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_32xf16 = shufflevector <32 x half> %ld_32xf16, <32 x half> undef, <32 x i32> zeroinitializer
 ; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_2xf32 = load <2 x float>, ptr undef, align 8
 ; AVX-NEXT:  Cost Model: Found costs of 1 for: %sf_2xf32 = shufflevector <2 x float> %ld_2xf32, <2 x float> undef, <2 x i32> zeroinitializer
 ; AVX-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xf32 = load <4 x float>, ptr undef, align 16
@@ -284,21 +284,21 @@ define void @shuffle_load() {
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_8xi8 = load <8 x i8>, ptr undef, align 8
 ; AVX2-NEXT:  Cost Model: Found costs of 2 for: %sf_8xi8 = shufflevector <8 x i8> %ld_8xi8, <8 x i8> undef, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_16xi8 = load <16 x i8>, ptr undef, align 16
-; AVX2-NEXT:  Cost Model: Found costs of 3 for: %sf_16xi8 = shufflevector <16 x i8> %ld_16xi8, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %sf_16xi8 = shufflevector <16 x i8> %ld_16xi8, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:1 for: %ld_32xi8 = load <32 x i8>, ptr undef, align 32
-; AVX2-NEXT:  Cost Model: Found costs of 3 for: %sf_32xi8 = shufflevector <32 x i8> %ld_32xi8, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %sf_32xi8 = shufflevector <32 x i8> %ld_32xi8, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_64xi8 = load <64 x i8>, ptr undef, align 64
-; AVX2-NEXT:  Cost Model: Found costs of 3 for: %sf_64xi8 = shufflevector <64 x i8> %ld_64xi8, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %sf_64xi8 = shufflevector <64 x i8> %ld_64xi8, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_2xi16 = load <2 x i16>, ptr undef, align 4
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %sf_2xi16 = shufflevector <2 x i16> %ld_2xi16, <2 x i16> undef, <2 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xi16 = load <4 x i16>, ptr undef, align 8
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %sf_4xi16 = shufflevector <4 x i16> %ld_4xi16, <4 x i16> undef, <4 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_8xi16 = load <8 x i16>, ptr undef, align 16
-; AVX2-NEXT:  Cost Model: Found costs of 2 for: %sf_8xi16 = shufflevector <8 x i16> %ld_8xi16, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_8xi16 = shufflevector <8 x i16> %ld_8xi16, <8 x i16> undef, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:1 for: %ld_16xi16 = load <16 x i16>, ptr undef, align 32
-; AVX2-NEXT:  Cost Model: Found costs of 2 for: %sf_16xi16 = shufflevector <16 x i16> %ld_16xi16, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_16xi16 = shufflevector <16 x i16> %ld_16xi16, <16 x i16> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_32xi16 = load <32 x i16>, ptr undef, align 64
-; AVX2-NEXT:  Cost Model: Found costs of 2 for: %sf_32xi16 = shufflevector <32 x i16> %ld_32xi16, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_32xi16 = shufflevector <32 x i16> %ld_32xi16, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_2xi32 = load <2 x i32>, ptr undef, align 8
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %sf_2xi32 = shufflevector <2 x i32> %ld_2xi32, <2 x i32> undef, <2 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xi32 = load <4 x i32>, ptr undef, align 16
@@ -318,11 +318,11 @@ define void @shuffle_load() {
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xf16 = load <4 x half>, ptr undef, align 8
 ; AVX2-NEXT:  Cost Model: Found costs of 0 for: %sf_4xf16 = shufflevector <4 x half> %ld_4xf16, <4 x half> undef, <4 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_8xf16 = load <8 x half>, ptr undef, align 16
-; AVX2-NEXT:  Cost Model: Found costs of 2 for: %sf_8xf16 = shufflevector <8 x half> %ld_8xf16, <8 x half> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_8xf16 = shufflevector <8 x half> %ld_8xf16, <8 x half> undef, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:1 for: %ld_16xf16 = load <16 x half>, ptr undef, align 32
-; AVX2-NEXT:  Cost Model: Found costs of 2 for: %sf_16xf16 = shufflevector <16 x half> %ld_16xf16, <16 x half> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_16xf16 = shufflevector <16 x half> %ld_16xf16, <16 x half> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_32xf16 = load <32 x half>, ptr undef, align 64
-; AVX2-NEXT:  Cost Model: Found costs of 2 for: %sf_32xf16 = shufflevector <32 x half> %ld_32xf16, <32 x half> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_32xf16 = shufflevector <32 x half> %ld_32xf16, <32 x half> undef, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_2xf32 = load <2 x float>, ptr undef, align 8
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %sf_2xf32 = shufflevector <2 x float> %ld_2xf32, <2 x float> undef, <2 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xf32 = load <4 x float>, ptr undef, align 16
@@ -347,21 +347,21 @@ define void @shuffle_load() {
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_8xi8 = load <8 x i8>, ptr undef, align 8
 ; AVX512-NEXT:  Cost Model: Found costs of 2 for: %sf_8xi8 = shufflevector <8 x i8> %ld_8xi8, <8 x i8> undef, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_16xi8 = load <16 x i8>, ptr undef, align 16
-; AVX512-NEXT:  Cost Model: Found costs of 3 for: %sf_16xi8 = shufflevector <16 x i8> %ld_16xi8, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %sf_16xi8 = shufflevector <16 x i8> %ld_16xi8, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:1 for: %ld_32xi8 = load <32 x i8>, ptr undef, align 32
-; AVX512-NEXT:  Cost Model: Found costs of 3 for: %sf_32xi8 = shufflevector <32 x i8> %ld_32xi8, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %sf_32xi8 = shufflevector <32 x i8> %ld_32xi8, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_64xi8 = load <64 x i8>, ptr undef, align 64
-; AVX512-NEXT:  Cost Model: Found costs of 3 for: %sf_64xi8 = shufflevector <64 x i8> %ld_64xi8, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %sf_64xi8 = shufflevector <64 x i8> %ld_64xi8, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_2xi16 = load <2 x i16>, ptr undef, align 4
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %sf_2xi16 = shufflevector <2 x i16> %ld_2xi16, <2 x i16> undef, <2 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xi16 = load <4 x i16>, ptr undef, align 8
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %sf_4xi16 = shufflevector <4 x i16> %ld_4xi16, <4 x i16> undef, <4 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_8xi16 = load <8 x i16>, ptr undef, align 16
-; AVX512-NEXT:  Cost Model: Found costs of 2 for: %sf_8xi16 = shufflevector <8 x i16> %ld_8xi16, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_8xi16 = shufflevector <8 x i16> %ld_8xi16, <8 x i16> undef, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:1 for: %ld_16xi16 = load <16 x i16>, ptr undef, align 32
-; AVX512-NEXT:  Cost Model: Found costs of 2 for: %sf_16xi16 = shufflevector <16 x i16> %ld_16xi16, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_16xi16 = shufflevector <16 x i16> %ld_16xi16, <16 x i16> undef, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_32xi16 = load <32 x i16>, ptr undef, align 64
-; AVX512-NEXT:  Cost Model: Found costs of 2 for: %sf_32xi16 = shufflevector <32 x i16> %ld_32xi16, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_32xi16 = shufflevector <32 x i16> %ld_32xi16, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_2xi32 = load <2 x i32>, ptr undef, align 8
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %sf_2xi32 = shufflevector <2 x i32> %ld_2xi32, <2 x i32> undef, <2 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xi32 = load <4 x i32>, ptr undef, align 16
@@ -381,11 +381,11 @@ define void @shuffle_load() {
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xf16 = load <4 x half>, ptr undef, align 8
 ; AVX512-NEXT:  Cost Model: Found costs of 0 for: %sf_4xf16 = shufflevector <4 x half> %ld_4xf16, <4 x half> undef, <4 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_8xf16 = load <8 x half>, ptr undef, align 16
-; AVX512-NEXT:  Cost Model: Found costs of 2 for: %sf_8xf16 = shufflevector <8 x half> %ld_8xf16, <8 x half> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_8xf16 = shufflevector <8 x half> %ld_8xf16, <8 x half> undef, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:1 for: %ld_16xf16 = load <16 x half>, ptr undef, align 32
-; AVX512-NEXT:  Cost Model: Found costs of 2 for: %sf_16xf16 = shufflevector <16 x half> %ld_16xf16, <16 x half> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_16xf16 = shufflevector <16 x half> %ld_16xf16, <16 x half> undef, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:1 Lat:4 SizeLat:1 for: %ld_32xf16 = load <32 x half>, ptr undef, align 64
-; AVX512-NEXT:  Cost Model: Found costs of 2 for: %sf_32xf16 = shufflevector <32 x half> %ld_32xf16, <32 x half> undef, <32 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %sf_32xf16 = shufflevector <32 x half> %ld_32xf16, <32 x half> undef, <32 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_2xf32 = load <2 x float>, ptr undef, align 8
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %sf_2xf32 = shufflevector <2 x float> %ld_2xf32, <2 x float> undef, <2 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:1 for: %ld_4xf32 = load <4 x float>, ptr undef, align 16
diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-reverse-fp16.ll b/llvm/test/Analysis/CostModel/X86/shuffle-reverse-fp16.ll
index 4affc7e..f47e1a3 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-reverse-fp16.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-reverse-fp16.ll
@@ -3,11 +3,11 @@
 
 define void @test_vXf16(<2 x half> %src32, <4 x half> %src64, <8 x half> %src128, <16 x half> %src256, <32 x half> %src512) {
 ; CHECK-LABEL: 'test_vXf16'
-; CHECK-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> <i32 1, i32 0>
-; CHECK-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; CHECK-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> <i32 1, i32 0>
+; CHECK-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; CHECK-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; CHECK-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:6 SizeLat:4 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; CHECK-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V32  = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> <i32 1, i32 0>
diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-reverse.ll b/llvm/test/Analysis/CostModel/X86/shuffle-reverse.ll
index 1bcf2d5..aca0ba3 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-reverse.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-reverse.ll
@@ -25,20 +25,20 @@ define void @test_vXf64(<2 x double> %src128, <4 x double> %src256, <8 x double>
 ;
 ; AVX1-LABEL: 'test_vXf64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:6 SizeLat:2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:12 SizeLat:4 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXf64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:6 SizeLat:2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:12 SizeLat:4 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXf64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 0>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:6 SizeLat:2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:5 SizeLat:3 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 0>
@@ -56,20 +56,20 @@ define void @test_vXi64(<2 x i64> %src128, <4 x i64> %src256, <8 x i64> %src512)
 ;
 ; AVX1-LABEL: 'test_vXi64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:6 SizeLat:2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:12 SizeLat:4 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:6 SizeLat:2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:12 SizeLat:4 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXi64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:6 SizeLat:2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:5 SizeLat:3 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 0>
@@ -89,22 +89,22 @@ define void @test_vXf32(<2 x float> %src64, <4 x float> %src128, <8 x float> %sr
 ; AVX1-LABEL: 'test_vXf32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:14 SizeLat:8 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXf32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:14 SizeLat:8 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXf32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 0>
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 0>
@@ -125,22 +125,22 @@ define void @test_vXi32(<2 x i32> %src64, <4 x i32> %src128, <8 x i32> %src256,
 ; AVX1-LABEL: 'test_vXi32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:14 SizeLat:8 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:14 SizeLat:8 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXi32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 0>
@@ -154,65 +154,65 @@ define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <
 ; SSE2-LABEL: 'test_vXi16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of 6 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of 12 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:3 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:6 SizeLat:6 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:12 Lat:12 SizeLat:12 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSSE3-LABEL: 'test_vXi16'
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:8 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSE42-LABEL: 'test_vXi16'
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:8 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXi16'
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:5 Lat:9 SizeLat:5 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:10 Lat:18 SizeLat:10 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi16'
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:9 SizeLat:4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:18 SizeLat:8 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512F-LABEL: 'test_vXi16'
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:9 SizeLat:4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512F-NEXT:  Cost Model: Found costs of 7 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512BW-LABEL: 'test_vXi16'
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512BW-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:6 SizeLat:4 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512VBMI-LABEL: 'test_vXi16'
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:6 SizeLat:4 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512VBMI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V32  = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 0>
@@ -227,71 +227,71 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16 x
 ; SSE2-LABEL: 'test_vXi8'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
 ; SSE2-NEXT:  Cost Model: Found costs of 3 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of 9 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of 9 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of 18 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of 36 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:11 Lat:6 SizeLat:11 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:11 Lat:6 SizeLat:11 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:22 Lat:12 SizeLat:22 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:44 Lat:24 SizeLat:44 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSSE3-LABEL: 'test_vXi8'
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:8 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSE42-LABEL: 'test_vXi8'
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:8 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXi8'
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:5 Lat:9 SizeLat:5 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:10 Lat:18 SizeLat:10 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi8'
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:9 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:18 SizeLat:8 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512F-LABEL: 'test_vXi8'
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:9 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512F-NEXT:  Cost Model: Found costs of 7 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512BW-LABEL: 'test_vXi8'
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:9 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:9 SizeLat:3 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512VBMI-LABEL: 'test_vXi8'
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 0>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512VBMI-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512VBMI-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512VBMI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
@@ -320,22 +320,22 @@ define void @test_upper_vXf32(<2 x float> %a64, <2 x float> %b64, <4 x float> %a
 ; AVX1-LABEL: 'test_upper_vXf32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %a64, <2 x float> %b64, <2 x i32> <i32 3, i32 2>
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %a128, <4 x float> %b128, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8>
-; AVX1-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:14 SizeLat:8 for: %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_upper_vXf32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %a64, <2 x float> %b64, <2 x i32> <i32 3, i32 2>
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %a128, <4 x float> %b128, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8>
-; AVX2-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:14 SizeLat:8 for: %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_upper_vXf32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %a64, <2 x float> %b64, <2 x i32> <i32 3, i32 2>
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %a128, <4 x float> %b128, <4 x i32> <i32 7, i32 6, i32 5, i32 4>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:7 SizeLat:4 for: %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V64 = shufflevector <2 x float> %a64, <2 x float> %b64, <2 x i32> <i32 3, i32 2>
diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-single-src-fp16.ll b/llvm/test/Analysis/CostModel/X86/shuffle-single-src-fp16.ll
index 007aa59bc..21d5766 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-single-src-fp16.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-single-src-fp16.ll
@@ -3,15 +3,15 @@
 
 define void @test_vXf16(<8 x half> %src128, <16 x half> %src256, <32 x half> %src512, <64 x half> %src1024) {
 ; CHECK-LABEL: 'test_vXf16'
-; CHECK-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; CHECK-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; CHECK-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 31, i32 30, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; CHECK-NEXT:  Cost Model: Found costs of 4 for: %V1024 = shufflevector <64 x half> %src1024, <64 x half> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; CHECK-NEXT:  Cost Model: Found costs of 7 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; CHECK-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:  Cost Model: Found costs of 4 for: %V1024 = shufflevector <64 x half> %src1024, <64 x half> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
 ; CHECK-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
-  %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-  %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-  %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 31, i32 30, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-  %V1024 = shufflevector <64 x half> %src1024, <64 x half> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+  %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+  %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  %V1024 = shufflevector <64 x half> %src1024, <64 x half> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
   ret void
 }
diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-single-src.ll b/llvm/test/Analysis/CostModel/X86/shuffle-single-src.ll
index 3727d7b..11f14df 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-single-src.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-single-src.ll
@@ -21,42 +21,42 @@ define void @test_vXf64(<2 x double> %src128, <4 x double> %src256, <8 x double>
 ; SSE-LABEL: 'test_vXf64'
 ; SSE-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
 ; SSE-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; SSE-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE-NEXT:  Cost Model: Found costs of 8 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
+; SSE-NEXT:  Cost Model: Found costs of 8 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> undef, <16 x i32> <i32 15, i32 0, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 13, i32 3, i32 11, i32 1, i32 0>
 ; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'test_vXf64'
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
 ; XOP-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 8 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 5 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 11 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> undef, <16 x i32> <i32 15, i32 0, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 13, i32 3, i32 11, i32 1, i32 0>
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXf64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
 ; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 8 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 5 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 11 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> undef, <16 x i32> <i32 15, i32 0, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 13, i32 3, i32 11, i32 1, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXf64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 4 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 10 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> undef, <16 x i32> <i32 15, i32 0, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 13, i32 3, i32 11, i32 1, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXf64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:6 SizeLat:2 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:6 SizeLat:2 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> undef, <16 x i32> <i32 15, i32 0, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 13, i32 3, i32 11, i32 1, i32 0>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
   %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-  %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-  %V1024 = shufflevector <16 x double> %src1024, <16 x double> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
+  %V1024 = shufflevector <16 x double> %src1024, <16 x double> undef, <16 x i32> <i32 15, i32 0, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 13, i32 3, i32 11, i32 1, i32 0>
   ret void
 }
 
@@ -64,79 +64,93 @@ define void @test_vXi64(<2 x i64> %src128, <4 x i64> %src256, <8 x i64> %src512)
 ; SSE-LABEL: 'test_vXi64'
 ; SSE-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
 ; SSE-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; SSE-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
 ; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'test_vXi64'
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
 ; XOP-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 5 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXi64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
 ; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 5 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXi64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
   %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-  %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
+  %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 5, i32 1, i32 0>
   ret void
 }
 
 define void @test_vXf32(<2 x float> %src64, <4 x float> %src128, <8 x float> %src256, <16 x float> %src512) {
-; SSE-LABEL: 'test_vXf32'
-; SSE-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
-; SSE-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; SSE-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
+; SSE2-LABEL: 'test_vXf32'
+; SSE2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
+; SSE2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of 11 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
+;
+; SSSE3-LABEL: 'test_vXf32'
+; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
+; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of 11 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
+;
+; SSE42-LABEL: 'test_vXf32'
+; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
+; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of 10 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'test_vXf32'
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
 ; XOP-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 12 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXf32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 6 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 12 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXf32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 9 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXf32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V64  = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
   %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
   %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-  %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
   ret void
 }
 
@@ -144,48 +158,48 @@ define void @test_vXi32(<2 x i32> %src64, <4 x i32> %src128, <8 x i32> %src256,
 ; SSE-LABEL: 'test_vXi32'
 ; SSE-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
 ; SSE-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; SSE-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 2, i32 1, i32 0>
-; SSE-NEXT:  Cost Model: Found costs of 6 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 13, i32 10, i32 9, i32 8, i32 8, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE-NEXT:  Cost Model: Found costs of 10 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> undef, <32 x i32> <i32 31, i32 30, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 0, i32 1, i32 0>
+; SSE-NEXT:  Cost Model: Found costs of 11 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; SSE-NEXT:  Cost Model: Found costs of 18 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'test_vXi32'
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 2, i32 1, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 6 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 13, i32 10, i32 9, i32 8, i32 8, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 12 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> undef, <32 x i32> <i32 31, i32 30, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 0, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 12 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; XOP-NEXT:  Cost Model: Found costs of 26 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXi32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 13, i32 10, i32 9, i32 8, i32 8, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 14 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> undef, <32 x i32> <i32 31, i32 30, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 0, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 12 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; AVX1-NEXT:  Cost Model: Found costs of 26 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 13, i32 10, i32 9, i32 8, i32 8, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 8 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> undef, <32 x i32> <i32 31, i32 30, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 0, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 9 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; AVX2-NEXT:  Cost Model: Found costs of 20 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXi32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 13, i32 10, i32 9, i32 8, i32 8, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:6 SizeLat:2 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> undef, <32 x i32> <i32 31, i32 30, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 0, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:6 SizeLat:2 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V64  = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
   %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 0>
-  %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 2, i32 1, i32 0>
-  %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 13, i32 10, i32 9, i32 8, i32 8, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-  %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> undef, <32 x i32> <i32 31, i32 30, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 7, i32 6, i32 5, i32 5, i32 3, i32 0, i32 1, i32 0>
+  %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+  %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
   ret void
 }
 
@@ -193,90 +207,90 @@ define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <
 ; SSE2-LABEL: 'test_vXi16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
-; SSE2-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:5 SizeLat:5 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:8 Lat:8 SizeLat:8 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:24 Lat:24 SizeLat:24 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of RThru:24 CodeSize:26 Lat:26 SizeLat:26 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:5 SizeLat:5 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:24 Lat:24 SizeLat:24 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:42 CodeSize:56 Lat:56 SizeLat:56 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:72 CodeSize:94 Lat:94 SizeLat:94 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSSE3-LABEL: 'test_vXi16'
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
 ; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 8 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of 9 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; SSSE3-NEXT:  Cost Model: Found costs of 21 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of 39 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSE42-LABEL: 'test_vXi16'
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 8 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of 9 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; SSE42-NEXT:  Cost Model: Found costs of 21 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of 35 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'test_vXi16'
-; XOP-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
-; XOP-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 16 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 9 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; XOP-NEXT:  Cost Model: Found costs of 18 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 63 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXi16'
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 8 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 16 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 20 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 15 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; AVX1-NEXT:  Cost Model: Found costs of 30 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 99 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi16'
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 10 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 7 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; AVX2-NEXT:  Cost Model: Found costs of 14 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 44 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512F-LABEL: 'test_vXi16'
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT:  Cost Model: Found costs of 14 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT:  Cost Model: Found costs of 21 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; AVX512F-NEXT:  Cost Model: Found costs of 7 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; AVX512F-NEXT:  Cost Model: Found costs of 14 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT:  Cost Model: Found costs of 84 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512BW-LABEL: 'test_vXi16'
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT:  Cost Model: Found costs of 4 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; AVX512BW-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of 4 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512VBMI-LABEL: 'test_vXi16'
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
 ; AVX512VBMI-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 4 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of 4 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
 ; AVX512VBMI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V32  = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
   %V64  = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 2, i32 2, i32 1, i32 1>
-  %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 2, i32 1, i32 0>
-  %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 13, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-  %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 31, i32 30, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-  %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 7, i32 6, i32 6, i32 4, i32 3, i32 0, i32 1, i32 0>
+  %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 4, i32 15, i32 13, i32 13, i32 11, i32 10, i32 8, i32 8, i32 3, i32 6, i32 5, i32 14, i32 7, i32 2, i32 1, i32 16>
+  %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
   ret void
 }
 
@@ -286,88 +300,88 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16 x
 ; SSE2-NEXT:  Cost Model: Found costs of 3 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
 ; SSE2-NEXT:  Cost Model: Found costs of 5 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:10 Lat:10 SizeLat:10 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of RThru:17 CodeSize:19 Lat:19 SizeLat:19 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of RThru:35 CodeSize:37 Lat:37 SizeLat:37 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:22 CodeSize:26 Lat:26 SizeLat:26 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:77 CodeSize:91 Lat:91 SizeLat:91 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSSE3-LABEL: 'test_vXi8'
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
 ; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
 ; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
 ; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of 6 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of 21 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSE42-LABEL: 'test_vXi8'
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of 6 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of 21 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'test_vXi8'
-; XOP-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
+; XOP-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 18 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXi8'
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 8 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 12 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 8 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 30 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi8'
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 6 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 14 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512F-LABEL: 'test_vXi8'
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT:  Cost Model: Found costs of 14 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT:  Cost Model: Found costs of 14 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512BW-LABEL: 'test_vXi8'
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512VBMI-LABEL: 'test_vXi8'
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
 ; AVX512VBMI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
 ; AVX512VBMI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
 ; AVX512VBMI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
 ; AVX512VBMI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V16  = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
   %V32  = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 1, i32 1>
   %V64  = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 7, i32 7, i32 5, i32 5, i32 3, i32 3, i32 1, i32 1>
   %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-  %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-  %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 63, i32 62, i32 61, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 8, i32 30, i32 20, i32 28, i32 27, i32 15, i32 13, i32 24, i32 11, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 26, i32 15, i32 14, i32 25, i32 12, i32 11, i32 23, i32 9, i32 31, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 0, i32 62, i32 61, i32 60, i32 7, i32 58, i32 57, i32 55, i32 56, i32 54, i32 53, i32 52, i32 1, i32 50, i32 47, i32 48, i32 49, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 11, i32 37, i32 36, i32 33, i32 34, i32 35, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 23, i32 24, i32 20, i32 35, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 42, i32 12, i32 34, i32 10, i32 8, i32 9, i32 24, i32 6, i32 5, i32 4, i32 29, i32 2, i32 1, i32 0>
   ret void
 }
 
diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-splat.ll b/llvm/test/Analysis/CostModel/X86/shuffle-splat.ll
index 6d22546..c98efcd 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-splat.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-splat.ll
@@ -25,20 +25,20 @@ define void @test_vXf64(<2 x double> %src128, <4 x double> %src256, <8 x double>
 ;
 ; AVX1-LABEL: 'test_vXf64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXf64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXf64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V128 = shufflevector <2 x double> %src128, <2 x double> undef, <2 x i32> <i32 1, i32 1>
@@ -56,20 +56,20 @@ define void @test_vXi64(<2 x i64> %src128, <4 x i64> %src256, <8 x i64> %src512)
 ;
 ; AVX1-LABEL: 'test_vXi64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXi64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V128 = shufflevector <2 x i64> %src128, <2 x i64> undef, <2 x i32> <i32 1, i32 1>
@@ -89,22 +89,22 @@ define void @test_vXf32(<2 x float> %src64, <4 x float> %src128, <8 x float> %sr
 ; AVX1-LABEL: 'test_vXf32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXf32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXf32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V64 = shufflevector <2 x float> %src64, <2 x float> undef, <2 x i32> <i32 1, i32 1>
@@ -125,22 +125,22 @@ define void @test_vXi32(<2 x i32> %src64, <4 x i32> %src128, <8 x i32> %src256,
 ; AVX1-LABEL: 'test_vXi32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXi32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V64 = shufflevector <2 x i32> %src64, <2 x i32> undef, <2 x i32> <i32 1, i32 1>
@@ -154,41 +154,41 @@ define void @test_vXf16(<2 x half> %src32, <4 x half> %src64, <8 x half> %src128
 ; SSE2-LABEL: 'test_vXf16'
 ; SSE2-NEXT:  Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> <i32 1, i32 1>
 ; SSE2-NEXT:  Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSSE3-LABEL: 'test_vXf16'
 ; SSSE3-NEXT:  Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> <i32 1, i32 1>
 ; SSSE3-NEXT:  Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSE42-LABEL: 'test_vXf16'
 ; SSE42-NEXT:  Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> <i32 1, i32 1>
 ; SSE42-NEXT:  Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXf16'
 ; AVX2-NEXT:  Cost Model: Found costs of 0 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> <i32 1, i32 1>
 ; AVX2-NEXT:  Cost Model: Found costs of 0 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXf16'
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> <i32 1, i32 1>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> <i32 1, i32 1>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x half> %src64, <4 x half> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x half> %src128, <8 x half> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <16 x half> %src256, <16 x half> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <32 x half> %src512, <32 x half> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V32  = shufflevector <2 x half> %src32, <2 x half> undef, <2 x i32> <i32 1, i32 1>
@@ -244,49 +244,49 @@ define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <
 ; SSE2-LABEL: 'test_vXi16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSSE3-LABEL: 'test_vXi16'
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSE42-LABEL: 'test_vXi16'
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXi16'
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi16'
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXi16'
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V32  = shufflevector <2 x i16> %src32, <2 x i16> undef, <2 x i32> <i32 1, i32 1>
@@ -302,54 +302,54 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16 x
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
 ; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
 ; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSSE3-LABEL: 'test_vXi8'
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSE42-LABEL: 'test_vXi8'
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXi8'
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi8'
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXi8'
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V16  = shufflevector <2 x i8> %src16, <2 x i8> undef, <2 x i32> <i32 1, i32 1>
@@ -365,46 +365,46 @@ define void @test_vXi1(<2 x i1> %src2, <4 x i1> %src4, <8 x i1> %src8, <16 x i1>
 ; SSE2-LABEL: 'test_vXi1'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> <i32 1, i32 1>
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSSE3-LABEL: 'test_vXi1'
 ; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> <i32 1, i32 1>
 ; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSSE3-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSSE3-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSE42-LABEL: 'test_vXi1'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> <i32 1, i32 1>
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXi1'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> <i32 1, i32 1>
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi1'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V2 = shufflevector <2 x i1> %src2, <2 x i1> undef, <2 x i32> <i32 1, i32 1>
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V4 = shufflevector <4 x i1> %src4, <4 x i1> undef, <4 x i32> <i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V8 = shufflevector <8 x i1> %src8, <8 x i1> undef, <8 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V16 = shufflevector <16 x i1> %src16, <16 x i1> undef, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V32 = shufflevector <32 x i1> %src32, <32 x i1> undef, <32 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V64 = shufflevector <64 x i1> %src64, <64 x i1> undef, <64 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXi1'
@@ -440,22 +440,22 @@ define void @test_upper_vXf32(<2 x float> %a64, <2 x float> %b64, <4 x float> %a
 ; AVX1-LABEL: 'test_upper_vXf32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %a64, <2 x float> %b64, <2 x i32> <i32 3, i32 3>
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %a128, <4 x float> %b128, <4 x i32> <i32 6, i32 6, i32 6, i32 6>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11>
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11>
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_upper_vXf32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %a64, <2 x float> %b64, <2 x i32> <i32 3, i32 3>
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %a128, <4 x float> %b128, <4 x i32> <i32 6, i32 6, i32 6, i32 6>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11>
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11>
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_upper_vXf32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <2 x float> %a64, <2 x float> %b64, <2 x i32> <i32 3, i32 3>
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <4 x float> %a128, <4 x float> %b128, <4 x i32> <i32 6, i32 6, i32 6, i32 6>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11>
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x float> %a256, <8 x float> %b256, <8 x i32> <i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11, i32 11>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x float> %a512, <16 x float> %b512, <16 x i32> <i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3, i32 3>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V64 = shufflevector <2 x float> %a64, <2 x float> %b64, <2 x i32> <i32 3, i32 3>
diff --git a/llvm/test/Analysis/CostModel/X86/shuffle-two-src.ll b/llvm/test/Analysis/CostModel/X86/shuffle-two-src.ll
index 56afc01..3016316 100644
--- a/llvm/test/Analysis/CostModel/X86/shuffle-two-src.ll
+++ b/llvm/test/Analysis/CostModel/X86/shuffle-two-src.ll
@@ -22,41 +22,41 @@ define void @test_vXf64(<2 x double> %src128, <4 x double> %src256, <8 x double>
 ; SSE-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> %src128_1, <2 x i32> <i32 3, i32 0>
 ; SSE-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x double> %src256, <4 x double> %src256_1, <4 x i32> <i32 3, i32 3, i32 7, i32 6>
 ; SSE-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x double> %src512, <8 x double> %src512_1, <8 x i32> <i32 7, i32 6, i32 12, i32 4, i32 3, i32 2, i32 1, i32 15>
-; SSE-NEXT:  Cost Model: Found costs of 8 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 12, i32 13, i32 10, i32 18, i32 8, i32 8, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE-NEXT:  Cost Model: Found costs of 8 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 28, i32 13, i32 10, i32 18, i32 8, i32 8, i32 24, i32 31, i32 4, i32 9, i32 2, i32 8, i32 0>
 ; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'test_vXf64'
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> %src128_1, <2 x i32> <i32 3, i32 0>
 ; XOP-NEXT:  Cost Model: Found costs of 3 for: %V256 = shufflevector <4 x double> %src256, <4 x double> %src256_1, <4 x i32> <i32 3, i32 3, i32 7, i32 6>
 ; XOP-NEXT:  Cost Model: Found costs of 6 for: %V512 = shufflevector <8 x double> %src512, <8 x double> %src512_1, <8 x i32> <i32 7, i32 6, i32 12, i32 4, i32 3, i32 2, i32 1, i32 15>
-; XOP-NEXT:  Cost Model: Found costs of 12 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 12, i32 13, i32 10, i32 18, i32 8, i32 8, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 16 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 28, i32 13, i32 10, i32 18, i32 8, i32 8, i32 24, i32 31, i32 4, i32 9, i32 2, i32 8, i32 0>
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXf64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> %src128_1, <2 x i32> <i32 3, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of 3 for: %V256 = shufflevector <4 x double> %src256, <4 x double> %src256_1, <4 x i32> <i32 3, i32 3, i32 7, i32 6>
 ; AVX1-NEXT:  Cost Model: Found costs of 6 for: %V512 = shufflevector <8 x double> %src512, <8 x double> %src512_1, <8 x i32> <i32 7, i32 6, i32 12, i32 4, i32 3, i32 2, i32 1, i32 15>
-; AVX1-NEXT:  Cost Model: Found costs of 12 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 12, i32 13, i32 10, i32 18, i32 8, i32 8, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 16 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 28, i32 13, i32 10, i32 18, i32 8, i32 8, i32 24, i32 31, i32 4, i32 9, i32 2, i32 8, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXf64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> %src128_1, <2 x i32> <i32 3, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of 3 for: %V256 = shufflevector <4 x double> %src256, <4 x double> %src256_1, <4 x i32> <i32 3, i32 3, i32 7, i32 6>
 ; AVX2-NEXT:  Cost Model: Found costs of 6 for: %V512 = shufflevector <8 x double> %src512, <8 x double> %src512_1, <8 x i32> <i32 7, i32 6, i32 12, i32 4, i32 3, i32 2, i32 1, i32 15>
-; AVX2-NEXT:  Cost Model: Found costs of 11 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 12, i32 13, i32 10, i32 18, i32 8, i32 8, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 16 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 28, i32 13, i32 10, i32 18, i32 8, i32 8, i32 24, i32 31, i32 4, i32 9, i32 2, i32 8, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXf64'
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <2 x double> %src128, <2 x double> %src128_1, <2 x i32> <i32 3, i32 0>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x double> %src256, <4 x double> %src256_1, <4 x i32> <i32 3, i32 3, i32 7, i32 6>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x double> %src512, <8 x double> %src512_1, <8 x i32> <i32 7, i32 6, i32 12, i32 4, i32 3, i32 2, i32 1, i32 15>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:9 SizeLat:3 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 12, i32 13, i32 10, i32 18, i32 8, i32 8, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:12 SizeLat:4 for: %V1024 = shufflevector <16 x double> %src1024, <16 x double> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 28, i32 13, i32 10, i32 18, i32 8, i32 8, i32 24, i32 31, i32 4, i32 9, i32 2, i32 8, i32 0>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V128 = shufflevector <2 x double> %src128, <2 x double> %src128_1, <2 x i32> <i32 3, i32 0>
   %V256 = shufflevector <4 x double> %src256, <4 x double> %src256_1, <4 x i32> <i32 3, i32 3, i32 7, i32 6>
   %V512 = shufflevector <8 x double> %src512, <8 x double> %src512_1, <8 x i32> <i32 7, i32 6, i32 12, i32 4, i32 3, i32 2, i32 1, i32 15>
-  %V1024 = shufflevector <16 x double> %src1024, <16 x double> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 12, i32 13, i32 10, i32 18, i32 8, i32 8, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  %V1024 = shufflevector <16 x double> %src1024, <16 x double> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 28, i32 13, i32 10, i32 18, i32 8, i32 8, i32 24, i32 31, i32 4, i32 9, i32 2, i32 8, i32 0>
   ret void
 }
 
@@ -65,41 +65,41 @@ define void @test_vXi64(<2 x i64> %src128, <4 x i64> %src256, <8 x i64> %src512,
 ; SSE-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> %src128_1, <2 x i32> <i32 3, i32 0>
 ; SSE-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> %src256_1, <4 x i32> <i32 3, i32 3, i32 7, i32 6>
 ; SSE-NEXT:  Cost Model: Found costs of 4 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> %src512_1, <8 x i32> <i32 7, i32 6, i32 12, i32 4, i32 3, i32 2, i32 1, i32 15>
-; SSE-NEXT:  Cost Model: Found costs of 8 for: %V1024 = shufflevector <16 x i64> %src1024, <16 x i64> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 12, i32 13, i32 10, i32 18, i32 8, i32 8, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE-NEXT:  Cost Model: Found costs of 8 for: %V1024 = shufflevector <16 x i64> %src1024, <16 x i64> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 28, i32 13, i32 10, i32 18, i32 8, i32 8, i32 24, i32 31, i32 4, i32 9, i32 2, i32 8, i32 0>
 ; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'test_vXi64'
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> %src128_1, <2 x i32> <i32 3, i32 0>
 ; XOP-NEXT:  Cost Model: Found costs of 3 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> %src256_1, <4 x i32> <i32 3, i32 3, i32 7, i32 6>
 ; XOP-NEXT:  Cost Model: Found costs of 6 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> %src512_1, <8 x i32> <i32 7, i32 6, i32 12, i32 4, i32 3, i32 2, i32 1, i32 15>
-; XOP-NEXT:  Cost Model: Found costs of 12 for: %V1024 = shufflevector <16 x i64> %src1024, <16 x i64> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 12, i32 13, i32 10, i32 18, i32 8, i32 8, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 16 for: %V1024 = shufflevector <16 x i64> %src1024, <16 x i64> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 28, i32 13, i32 10, i32 18, i32 8, i32 8, i32 24, i32 31, i32 4, i32 9, i32 2, i32 8, i32 0>
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXi64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> %src128_1, <2 x i32> <i32 3, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of 3 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> %src256_1, <4 x i32> <i32 3, i32 3, i32 7, i32 6>
 ; AVX1-NEXT:  Cost Model: Found costs of 6 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> %src512_1, <8 x i32> <i32 7, i32 6, i32 12, i32 4, i32 3, i32 2, i32 1, i32 15>
-; AVX1-NEXT:  Cost Model: Found costs of 12 for: %V1024 = shufflevector <16 x i64> %src1024, <16 x i64> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 12, i32 13, i32 10, i32 18, i32 8, i32 8, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 16 for: %V1024 = shufflevector <16 x i64> %src1024, <16 x i64> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 28, i32 13, i32 10, i32 18, i32 8, i32 8, i32 24, i32 31, i32 4, i32 9, i32 2, i32 8, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> %src128_1, <2 x i32> <i32 3, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of 3 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> %src256_1, <4 x i32> <i32 3, i32 3, i32 7, i32 6>
 ; AVX2-NEXT:  Cost Model: Found costs of 6 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> %src512_1, <8 x i32> <i32 7, i32 6, i32 12, i32 4, i32 3, i32 2, i32 1, i32 15>
-; AVX2-NEXT:  Cost Model: Found costs of 11 for: %V1024 = shufflevector <16 x i64> %src1024, <16 x i64> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 12, i32 13, i32 10, i32 18, i32 8, i32 8, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 16 for: %V1024 = shufflevector <16 x i64> %src1024, <16 x i64> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 28, i32 13, i32 10, i32 18, i32 8, i32 8, i32 24, i32 31, i32 4, i32 9, i32 2, i32 8, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXi64'
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <2 x i64> %src128, <2 x i64> %src128_1, <2 x i32> <i32 3, i32 0>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <4 x i64> %src256, <4 x i64> %src256_1, <4 x i32> <i32 3, i32 3, i32 7, i32 6>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <8 x i64> %src512, <8 x i64> %src512_1, <8 x i32> <i32 7, i32 6, i32 12, i32 4, i32 3, i32 2, i32 1, i32 15>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:9 SizeLat:3 for: %V1024 = shufflevector <16 x i64> %src1024, <16 x i64> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 12, i32 13, i32 10, i32 18, i32 8, i32 8, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:12 SizeLat:4 for: %V1024 = shufflevector <16 x i64> %src1024, <16 x i64> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 28, i32 13, i32 10, i32 18, i32 8, i32 8, i32 24, i32 31, i32 4, i32 9, i32 2, i32 8, i32 0>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V128 = shufflevector <2 x i64> %src128, <2 x i64> %src128_1, <2 x i32> <i32 3, i32 0>
   %V256 = shufflevector <4 x i64> %src256, <4 x i64> %src256_1, <4 x i32> <i32 3, i32 3, i32 7, i32 6>
   %V512 = shufflevector <8 x i64> %src512, <8 x i64> %src512_1, <8 x i32> <i32 7, i32 6, i32 12, i32 4, i32 3, i32 2, i32 1, i32 15>
-  %V1024 = shufflevector <16 x i64> %src1024, <16 x i64> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 12, i32 13, i32 10, i32 18, i32 8, i32 8, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  %V1024 = shufflevector <16 x i64> %src1024, <16 x i64> %src1024_1, <16 x i32> <i32 30, i32 14, i32 13, i32 28, i32 13, i32 10, i32 18, i32 8, i32 8, i32 24, i32 31, i32 4, i32 9, i32 2, i32 8, i32 0>
   ret void
 }
 
@@ -107,129 +107,113 @@ define void @test_vXf32(<2 x float> %src64, <4 x float> %src128, <8 x float> %sr
 ; SSE2-LABEL: 'test_vXf32'
 ; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <2 x float> %src64, <2 x float> %src64_1, <2 x i32> <i32 3, i32 0>
 ; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <4 x float> %src128, <4 x float> %src128_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
-; SSE2-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> <i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 11, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of 14 for: %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 12, i32 11, i32 11, i32 9, i32 45, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> <i32 11, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of 10 for: %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of 22 for: %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 20, i32 53, i32 22, i32 20, i32 19, i32 18, i32 40, i32 16, i32 15, i32 48, i32 13, i32 40, i32 11, i32 11, i32 9, i32 45, i32 33, i32 11, i32 5, i32 4, i32 35, i32 2, i32 33, i32 0>
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSSE3-LABEL: 'test_vXf32'
 ; SSSE3-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <2 x float> %src64, <2 x float> %src64_1, <2 x i32> <i32 3, i32 0>
 ; SSSE3-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <4 x float> %src128, <4 x float> %src128_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
-; SSSE3-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> <i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 11, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 14 for: %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 12, i32 11, i32 11, i32 9, i32 45, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> <i32 11, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of 10 for: %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of 22 for: %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 20, i32 53, i32 22, i32 20, i32 19, i32 18, i32 40, i32 16, i32 15, i32 48, i32 13, i32 40, i32 11, i32 11, i32 9, i32 45, i32 33, i32 11, i32 5, i32 4, i32 35, i32 2, i32 33, i32 0>
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSE42-LABEL: 'test_vXf32'
 ; SSE42-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <2 x float> %src64, <2 x float> %src64_1, <2 x i32> <i32 3, i32 0>
 ; SSE42-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <4 x float> %src128, <4 x float> %src128_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
-; SSE42-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> <i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 7 for: %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 11, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 13 for: %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 12, i32 11, i32 11, i32 9, i32 45, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> <i32 11, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of 9 for: %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of 20 for: %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 20, i32 53, i32 22, i32 20, i32 19, i32 18, i32 40, i32 16, i32 15, i32 48, i32 13, i32 40, i32 11, i32 11, i32 9, i32 45, i32 33, i32 11, i32 5, i32 4, i32 35, i32 2, i32 33, i32 0>
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'test_vXf32'
 ; XOP-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <2 x float> %src64, <2 x float> %src64_1, <2 x i32> <i32 3, i32 0>
 ; XOP-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <4 x float> %src128, <4 x float> %src128_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
-; XOP-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> <i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 11, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 17 for: %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 12, i32 11, i32 11, i32 9, i32 45, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> <i32 11, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 10 for: %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 27 for: %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 20, i32 53, i32 22, i32 20, i32 19, i32 18, i32 40, i32 16, i32 15, i32 48, i32 13, i32 40, i32 11, i32 11, i32 9, i32 45, i32 33, i32 11, i32 5, i32 4, i32 35, i32 2, i32 33, i32 0>
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXf32'
 ; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <2 x float> %src64, <2 x float> %src64_1, <2 x i32> <i32 3, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <4 x float> %src128, <4 x float> %src128_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
-; AVX1-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> <i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 11, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 17 for: %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 12, i32 11, i32 11, i32 9, i32 45, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> <i32 11, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 10 for: %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 27 for: %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 20, i32 53, i32 22, i32 20, i32 19, i32 18, i32 40, i32 16, i32 15, i32 48, i32 13, i32 40, i32 11, i32 11, i32 9, i32 45, i32 33, i32 11, i32 5, i32 4, i32 35, i32 2, i32 33, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXf32'
 ; AVX2-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <2 x float> %src64, <2 x float> %src64_1, <2 x i32> <i32 3, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <4 x float> %src128, <4 x float> %src128_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
-; AVX2-NEXT:  Cost Model: Found costs of 3 for: %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> <i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 6 for: %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 11, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 13 for: %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 12, i32 11, i32 11, i32 9, i32 45, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 3 for: %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> <i32 11, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 21 for: %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 20, i32 53, i32 22, i32 20, i32 19, i32 18, i32 40, i32 16, i32 15, i32 48, i32 13, i32 40, i32 11, i32 11, i32 9, i32 45, i32 33, i32 11, i32 5, i32 4, i32 35, i32 2, i32 33, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXf32'
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <2 x float> %src64, <2 x float> %src64_1, <2 x i32> <i32 3, i32 0>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <4 x float> %src128, <4 x float> %src128_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> <i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 11, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:9 SizeLat:3 for: %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 12, i32 11, i32 11, i32 9, i32 45, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> <i32 11, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:12 SizeLat:4 for: %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 20, i32 53, i32 22, i32 20, i32 19, i32 18, i32 40, i32 16, i32 15, i32 48, i32 13, i32 40, i32 11, i32 11, i32 9, i32 45, i32 33, i32 11, i32 5, i32 4, i32 35, i32 2, i32 33, i32 0>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V64 = shufflevector <2 x float> %src64, <2 x float> %src64_1, <2 x i32> <i32 3, i32 0>
   %V128 = shufflevector <4 x float> %src128, <4 x float> %src128_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
-  %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> <i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
-  %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 11, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-  %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 12, i32 11, i32 11, i32 9, i32 45, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  %V256 = shufflevector <8 x float> %src256, <8 x float> %src256_1, <8 x i32> <i32 11, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
+  %V512 = shufflevector <16 x float> %src512, <16 x float> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+  %V1024 = shufflevector <32 x float> %src1024, <32 x float> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 20, i32 53, i32 22, i32 20, i32 19, i32 18, i32 40, i32 16, i32 15, i32 48, i32 13, i32 40, i32 11, i32 11, i32 9, i32 45, i32 33, i32 11, i32 5, i32 4, i32 35, i32 2, i32 33, i32 0>
   ret void
 }
 
 define void @test_vXi32(<2 x i32> %src64, <4 x i32> %src128, <8 x i32> %src256, <16 x i32> %src512, <32 x i32> %src1024, <2 x i32> %src64_1, <4 x i32> %src128_1, <8 x i32> %src256_1, <16 x i32> %src512_1, <32 x i32> %src1024_1) {
-; SSE2-LABEL: 'test_vXi32'
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> %src64_1, <2 x i32> <i32 3, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> %src128_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
-; SSE2-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> <i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 11, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of 14 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 12, i32 11, i32 11, i32 9, i32 45, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
-;
-; SSSE3-LABEL: 'test_vXi32'
-; SSSE3-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> %src64_1, <2 x i32> <i32 3, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> %src128_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
-; SSSE3-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> <i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 11, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 14 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 12, i32 11, i32 11, i32 9, i32 45, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
-;
-; SSE42-LABEL: 'test_vXi32'
-; SSE42-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> %src64_1, <2 x i32> <i32 3, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> %src128_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
-; SSE42-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> <i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 11, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 13 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 12, i32 11, i32 11, i32 9, i32 45, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
+; SSE-LABEL: 'test_vXi32'
+; SSE-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> %src64_1, <2 x i32> <i32 3, i32 0>
+; SSE-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> %src128_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
+; SSE-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> <i32 11, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
+; SSE-NEXT:  Cost Model: Found costs of 10 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; SSE-NEXT:  Cost Model: Found costs of 23 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 20, i32 53, i32 22, i32 20, i32 19, i32 18, i32 40, i32 16, i32 15, i32 48, i32 13, i32 40, i32 11, i32 11, i32 9, i32 45, i32 33, i32 11, i32 5, i32 4, i32 35, i32 2, i32 33, i32 0>
+; SSE-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'test_vXi32'
 ; XOP-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> %src64_1, <2 x i32> <i32 3, i32 0>
 ; XOP-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> %src128_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
-; XOP-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> <i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 11, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 17 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 12, i32 11, i32 11, i32 9, i32 45, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> <i32 11, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 10 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 27 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 20, i32 53, i32 22, i32 20, i32 19, i32 18, i32 40, i32 16, i32 15, i32 48, i32 13, i32 40, i32 11, i32 11, i32 9, i32 45, i32 33, i32 11, i32 5, i32 4, i32 35, i32 2, i32 33, i32 0>
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXi32'
 ; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> %src64_1, <2 x i32> <i32 3, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> %src128_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
-; AVX1-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> <i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 11, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 17 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 12, i32 11, i32 11, i32 9, i32 45, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> <i32 11, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 10 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 27 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 20, i32 53, i32 22, i32 20, i32 19, i32 18, i32 40, i32 16, i32 15, i32 48, i32 13, i32 40, i32 11, i32 11, i32 9, i32 45, i32 33, i32 11, i32 5, i32 4, i32 35, i32 2, i32 33, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi32'
 ; AVX2-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> %src64_1, <2 x i32> <i32 3, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> %src128_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
-; AVX2-NEXT:  Cost Model: Found costs of 3 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> <i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 6 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 11, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 13 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 12, i32 11, i32 11, i32 9, i32 45, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 3 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> <i32 11, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 8 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 21 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 20, i32 53, i32 22, i32 20, i32 19, i32 18, i32 40, i32 16, i32 15, i32 48, i32 13, i32 40, i32 11, i32 11, i32 9, i32 45, i32 33, i32 11, i32 5, i32 4, i32 35, i32 2, i32 33, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512-LABEL: 'test_vXi32'
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V64 = shufflevector <2 x i32> %src64, <2 x i32> %src64_1, <2 x i32> <i32 3, i32 0>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V128 = shufflevector <4 x i32> %src128, <4 x i32> %src128_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> <i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 11, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:9 SizeLat:3 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 12, i32 11, i32 11, i32 9, i32 45, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> <i32 11, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; AVX512-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:12 SizeLat:4 for: %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 20, i32 53, i32 22, i32 20, i32 19, i32 18, i32 40, i32 16, i32 15, i32 48, i32 13, i32 40, i32 11, i32 11, i32 9, i32 45, i32 33, i32 11, i32 5, i32 4, i32 35, i32 2, i32 33, i32 0>
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V64 = shufflevector <2 x i32> %src64, <2 x i32> %src64_1, <2 x i32> <i32 3, i32 0>
   %V128 = shufflevector <4 x i32> %src128, <4 x i32> %src128_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
-  %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> <i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
-  %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 11, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-  %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 12, i32 11, i32 11, i32 9, i32 45, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  %V256 = shufflevector <8 x i32> %src256, <8 x i32> %src256_1, <8 x i32> <i32 11, i32 6, i32 8, i32 4, i32 3, i32 2, i32 12, i32 0>
+  %V512 = shufflevector <16 x i32> %src512, <16 x i32> %src512_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+  %V1024 = shufflevector <32 x i32> %src1024, <32 x i32> %src1024_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 20, i32 53, i32 22, i32 20, i32 19, i32 18, i32 40, i32 16, i32 15, i32 48, i32 13, i32 40, i32 11, i32 11, i32 9, i32 45, i32 33, i32 11, i32 5, i32 4, i32 35, i32 2, i32 33, i32 0>
   ret void
 }
 
@@ -238,89 +222,89 @@ define void @test_vXi16(<2 x i16> %src32, <4 x i16> %src64, <8 x i16> %src128, <
 ; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> %src32_1, <2 x i32> <i32 3, i32 0>
 ; SSE2-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> %src64_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:8 Lat:8 SizeLat:8 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> %src128_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:11 Lat:11 SizeLat:11 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 14, i32 13, i32 20, i32 21, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of RThru:21 CodeSize:27 Lat:27 SizeLat:27 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 38, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of RThru:30 CodeSize:36 Lat:36 SizeLat:36 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 63, i32 62, i32 71, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 66, i32 2, i32 1, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:24 Lat:24 SizeLat:24 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:39 CodeSize:51 Lat:51 SizeLat:51 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:69 CodeSize:91 Lat:91 SizeLat:91 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 41, i32 40, i32 39, i32 39, i32 42, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 0, i32 1, i32 5, i32 6, i32 2, i32 66, i32 2, i32 6, i32 7>
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSSE3-LABEL: 'test_vXi16'
 ; SSSE3-NEXT:  Cost Model: Found costs of 3 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> %src32_1, <2 x i32> <i32 3, i32 0>
 ; SSSE3-NEXT:  Cost Model: Found costs of 3 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> %src64_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
 ; SSSE3-NEXT:  Cost Model: Found costs of 3 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> %src128_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 14, i32 13, i32 20, i32 21, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 10 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 38, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 12 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 63, i32 62, i32 71, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 66, i32 2, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of 9 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of 21 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of 37 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 41, i32 40, i32 39, i32 39, i32 42, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 0, i32 1, i32 5, i32 6, i32 2, i32 66, i32 2, i32 6, i32 7>
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSE42-LABEL: 'test_vXi16'
 ; SSE42-NEXT:  Cost Model: Found costs of 3 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> %src32_1, <2 x i32> <i32 3, i32 0>
 ; SSE42-NEXT:  Cost Model: Found costs of 3 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> %src64_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
 ; SSE42-NEXT:  Cost Model: Found costs of 3 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> %src128_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 14, i32 13, i32 20, i32 21, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 10 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 38, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 12 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 63, i32 62, i32 71, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 66, i32 2, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of 9 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of 19 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of 33 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 41, i32 40, i32 39, i32 39, i32 42, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 0, i32 1, i32 5, i32 6, i32 2, i32 66, i32 2, i32 6, i32 7>
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'test_vXi16'
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> %src32_1, <2 x i32> <i32 3, i32 0>
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> %src64_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> %src128_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 9 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 14, i32 13, i32 20, i32 21, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 18 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 38, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 26 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 63, i32 62, i32 71, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 66, i32 2, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 9 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 30 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 54 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 41, i32 40, i32 39, i32 39, i32 42, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 0, i32 1, i32 5, i32 6, i32 2, i32 66, i32 2, i32 6, i32 7>
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXi16'
 ; AVX1-NEXT:  Cost Model: Found costs of 3 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> %src32_1, <2 x i32> <i32 3, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of 3 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> %src64_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
 ; AVX1-NEXT:  Cost Model: Found costs of 3 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> %src128_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 15 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 14, i32 13, i32 20, i32 21, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 30 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 38, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 42 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 63, i32 62, i32 71, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 66, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 15 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 42 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 90 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 41, i32 40, i32 39, i32 39, i32 42, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 0, i32 1, i32 5, i32 6, i32 2, i32 66, i32 2, i32 6, i32 7>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi16'
 ; AVX2-NEXT:  Cost Model: Found costs of 3 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> %src32_1, <2 x i32> <i32 3, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of 3 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> %src64_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
 ; AVX2-NEXT:  Cost Model: Found costs of 3 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> %src128_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 7 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 14, i32 13, i32 20, i32 21, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 14 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 38, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 20 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 63, i32 62, i32 71, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 66, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 7 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 17 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 42 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 41, i32 40, i32 39, i32 39, i32 42, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 0, i32 1, i32 5, i32 6, i32 2, i32 66, i32 2, i32 6, i32 7>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512F-LABEL: 'test_vXi16'
 ; AVX512F-NEXT:  Cost Model: Found costs of 3 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> %src32_1, <2 x i32> <i32 3, i32 0>
 ; AVX512F-NEXT:  Cost Model: Found costs of 3 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> %src64_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
 ; AVX512F-NEXT:  Cost Model: Found costs of 3 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> %src128_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
-; AVX512F-NEXT:  Cost Model: Found costs of 7 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 14, i32 13, i32 20, i32 21, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT:  Cost Model: Found costs of 42 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 38, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT:  Cost Model: Found costs of 84 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 63, i32 62, i32 71, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 66, i32 2, i32 1, i32 0>
+; AVX512F-NEXT:  Cost Model: Found costs of 7 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; AVX512F-NEXT:  Cost Model: Found costs of 42 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+; AVX512F-NEXT:  Cost Model: Found costs of 168 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 41, i32 40, i32 39, i32 39, i32 42, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 0, i32 1, i32 5, i32 6, i32 2, i32 66, i32 2, i32 6, i32 7>
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512BW-LABEL: 'test_vXi16'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> %src32_1, <2 x i32> <i32 3, i32 0>
 ; AVX512BW-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> %src64_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
 ; AVX512BW-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> %src128_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 14, i32 13, i32 20, i32 21, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 38, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT:  Cost Model: Found costs of 4 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 63, i32 62, i32 71, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 66, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of 8 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 41, i32 40, i32 39, i32 39, i32 42, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 0, i32 1, i32 5, i32 6, i32 2, i32 66, i32 2, i32 6, i32 7>
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512VBMI-LABEL: 'test_vXi16'
 ; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V32 = shufflevector <2 x i16> %src32, <2 x i16> %src32_1, <2 x i32> <i32 3, i32 0>
 ; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <4 x i16> %src64, <4 x i16> %src64_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
 ; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <8 x i16> %src128, <8 x i16> %src128_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 14, i32 13, i32 20, i32 21, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 38, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 4 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 63, i32 62, i32 71, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 66, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of 8 for: %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 41, i32 40, i32 39, i32 39, i32 42, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 0, i32 1, i32 5, i32 6, i32 2, i32 66, i32 2, i32 6, i32 7>
 ; AVX512VBMI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V32 = shufflevector <2 x i16> %src32, <2 x i16> %src32_1, <2 x i32> <i32 3, i32 0>
   %V64 = shufflevector <4 x i16> %src64, <4 x i16> %src64_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
   %V128 = shufflevector <8 x i16> %src128, <8 x i16> %src128_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
-  %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 14, i32 13, i32 20, i32 21, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-  %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 38, i32 11, i32 11, i32 9, i32 8, i32 7, i32 11, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-  %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 63, i32 62, i32 71, i32 60, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 66, i32 2, i32 1, i32 0>
+  %V256 = shufflevector <16 x i16> %src256, <16 x i16> %src256_1, <16 x i32> <i32 15, i32 17, i32 13, i32 20, i32 29, i32 10, i32 8, i32 8, i32 7, i32 22, i32 5, i32 4, i32 22, i32 2, i32 1, i32 0>
+  %V512 = shufflevector <32 x i16> %src512, <32 x i16> %src512_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+  %V1024 = shufflevector <64 x i16> %src1024, <64 x i16> %src1024_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 41, i32 40, i32 39, i32 39, i32 42, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 0, i32 1, i32 5, i32 6, i32 2, i32 66, i32 2, i32 6, i32 7>
   ret void
 }
 
@@ -330,8 +314,8 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16 x
 ; SSE2-NEXT:  Cost Model: Found costs of 4 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> %src32_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
 ; SSE2-NEXT:  Cost Model: Found costs of 7 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> %src64_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:13 Lat:13 SizeLat:13 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> %src128_1, <16 x i32> <i32 29, i32 14, i32 28, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of RThru:19 CodeSize:23 Lat:23 SizeLat:23 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE2-NEXT:  Cost Model: Found costs of RThru:37 CodeSize:41 Lat:41 SizeLat:41 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 63, i32 100, i32 61, i32 96, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:42 Lat:42 SizeLat:42 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+; SSE2-NEXT:  Cost Model: Found costs of RThru:66 CodeSize:78 Lat:78 SizeLat:78 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 42, i32 41, i32 39, i32 39, i32 40, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 5, i32 6, i32 4, i32 66, i32 2, i32 1, i32 0>
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSSE3-LABEL: 'test_vXi8'
@@ -339,8 +323,8 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16 x
 ; SSSE3-NEXT:  Cost Model: Found costs of 3 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> %src32_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
 ; SSSE3-NEXT:  Cost Model: Found costs of 3 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> %src64_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
 ; SSSE3-NEXT:  Cost Model: Found costs of 3 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> %src128_1, <16 x i32> <i32 29, i32 14, i32 28, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSSE3-NEXT:  Cost Model: Found costs of 6 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 63, i32 100, i32 61, i32 96, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of 12 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+; SSSE3-NEXT:  Cost Model: Found costs of 18 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 42, i32 41, i32 39, i32 39, i32 40, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 5, i32 6, i32 4, i32 66, i32 2, i32 1, i32 0>
 ; SSSE3-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; SSE42-LABEL: 'test_vXi8'
@@ -348,8 +332,8 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16 x
 ; SSE42-NEXT:  Cost Model: Found costs of 3 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> %src32_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
 ; SSE42-NEXT:  Cost Model: Found costs of 3 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> %src64_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
 ; SSE42-NEXT:  Cost Model: Found costs of 3 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> %src128_1, <16 x i32> <i32 29, i32 14, i32 28, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 4 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-; SSE42-NEXT:  Cost Model: Found costs of 6 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 63, i32 100, i32 61, i32 96, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of 10 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+; SSE42-NEXT:  Cost Model: Found costs of 18 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 42, i32 41, i32 39, i32 39, i32 40, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 5, i32 6, i32 4, i32 66, i32 2, i32 1, i32 0>
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; XOP-LABEL: 'test_vXi8'
@@ -357,8 +341,8 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16 x
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> %src32_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> %src64_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
 ; XOP-NEXT:  Cost Model: Found costs of 1 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> %src128_1, <16 x i32> <i32 29, i32 14, i32 28, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 9 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-; XOP-NEXT:  Cost Model: Found costs of 13 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 63, i32 100, i32 61, i32 96, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 9 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+; XOP-NEXT:  Cost Model: Found costs of 36 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 42, i32 41, i32 39, i32 39, i32 40, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 5, i32 6, i32 4, i32 66, i32 2, i32 1, i32 0>
 ; XOP-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX1-LABEL: 'test_vXi8'
@@ -366,8 +350,8 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16 x
 ; AVX1-NEXT:  Cost Model: Found costs of 3 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> %src32_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
 ; AVX1-NEXT:  Cost Model: Found costs of 3 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> %src64_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of 3 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> %src128_1, <16 x i32> <i32 29, i32 14, i32 28, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 15 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX1-NEXT:  Cost Model: Found costs of 23 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 63, i32 100, i32 61, i32 96, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 15 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+; AVX1-NEXT:  Cost Model: Found costs of 54 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 42, i32 41, i32 39, i32 39, i32 40, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 5, i32 6, i32 4, i32 66, i32 2, i32 1, i32 0>
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX2-LABEL: 'test_vXi8'
@@ -375,8 +359,8 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16 x
 ; AVX2-NEXT:  Cost Model: Found costs of 3 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> %src32_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
 ; AVX2-NEXT:  Cost Model: Found costs of 3 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> %src64_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of 3 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> %src128_1, <16 x i32> <i32 29, i32 14, i32 28, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 7 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX2-NEXT:  Cost Model: Found costs of 11 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 63, i32 100, i32 61, i32 96, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 7 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+; AVX2-NEXT:  Cost Model: Found costs of 23 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 42, i32 41, i32 39, i32 39, i32 40, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 5, i32 6, i32 4, i32 66, i32 2, i32 1, i32 0>
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512F-LABEL: 'test_vXi8'
@@ -384,8 +368,8 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16 x
 ; AVX512F-NEXT:  Cost Model: Found costs of 3 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> %src32_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
 ; AVX512F-NEXT:  Cost Model: Found costs of 3 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> %src64_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
 ; AVX512F-NEXT:  Cost Model: Found costs of 3 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> %src128_1, <16 x i32> <i32 29, i32 14, i32 28, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT:  Cost Model: Found costs of 7 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512F-NEXT:  Cost Model: Found costs of 42 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 63, i32 100, i32 61, i32 96, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512F-NEXT:  Cost Model: Found costs of 7 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+; AVX512F-NEXT:  Cost Model: Found costs of 42 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 42, i32 41, i32 39, i32 39, i32 40, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 5, i32 6, i32 4, i32 66, i32 2, i32 1, i32 0>
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512BW-LABEL: 'test_vXi8'
@@ -393,8 +377,8 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16 x
 ; AVX512BW-NEXT:  Cost Model: Found costs of 3 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> %src32_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
 ; AVX512BW-NEXT:  Cost Model: Found costs of 3 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> %src64_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
 ; AVX512BW-NEXT:  Cost Model: Found costs of 3 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> %src128_1, <16 x i32> <i32 29, i32 14, i32 28, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT:  Cost Model: Found costs of 7 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512BW-NEXT:  Cost Model: Found costs of 19 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 63, i32 100, i32 61, i32 96, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of 7 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+; AVX512BW-NEXT:  Cost Model: Found costs of 19 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 42, i32 41, i32 39, i32 39, i32 40, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 5, i32 6, i32 4, i32 66, i32 2, i32 1, i32 0>
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
 ; AVX512VBMI-LABEL: 'test_vXi8'
@@ -402,15 +386,15 @@ define void @test_vXi8(<2 x i8> %src16, <4 x i8> %src32, <8 x i8> %src64, <16 x
 ; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V32 = shufflevector <4 x i8> %src32, <4 x i8> %src32_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
 ; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V64 = shufflevector <8 x i8> %src64, <8 x i8> %src64_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
 ; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V128 = shufflevector <16 x i8> %src128, <16 x i8> %src128_1, <16 x i32> <i32 29, i32 14, i32 28, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 63, i32 100, i32 61, i32 96, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+; AVX512VBMI-NEXT:  Cost Model: Found costs of 2 for: %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 42, i32 41, i32 39, i32 39, i32 40, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 5, i32 6, i32 4, i32 66, i32 2, i32 1, i32 0>
 ; AVX512VBMI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret void
 ;
   %V16 = shufflevector <2 x i8> %src16, <2 x i8> %src16_1, <2 x i32> <i32 3, i32 0>
   %V32 = shufflevector <4 x i8> %src32, <4 x i8> %src32_1, <4 x i32> <i32 3, i32 6, i32 1, i32 5>
   %V64 = shufflevector <8 x i8> %src64, <8 x i8> %src64_1, <8 x i32> <i32 7, i32 6, i32 6, i32 8, i32 9, i32 2, i32 1, i32 0>
   %V128 = shufflevector <16 x i8> %src128, <16 x i8> %src128_1, <16 x i32> <i32 29, i32 14, i32 28, i32 12, i32 11, i32 10, i32 11, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-  %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 30, i32 45, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 8, i32 8, i32 7, i32 6, i32 8, i32 4, i32 3, i32 2, i32 1, i32 0>
-  %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 63, i32 100, i32 61, i32 96, i32 59, i32 58, i32 57, i32 56, i32 55, i32 54, i32 53, i32 52, i32 51, i32 50, i32 49, i32 48, i32 47, i32 46, i32 45, i32 44, i32 43, i32 42, i32 41, i32 40, i32 39, i32 38, i32 37, i32 36, i32 35, i32 34, i32 33, i32 32, i32 31, i32 30, i32 29, i32 28, i32 27, i32 26, i32 25, i32 24, i32 23, i32 20, i32 21, i32 20, i32 19, i32 18, i32 17, i32 16, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+  %V256 = shufflevector <32 x i8> %src256, <32 x i8> %src256_1, <32 x i32> <i32 31, i32 33, i32 20, i32 27, i32 28, i32 26, i32 25, i32 24, i32 23, i32 22, i32 21, i32 7, i32 19, i32 18, i32 17, i32 16, i32 15, i32 48, i32 13, i32 40, i32 55, i32 11, i32 9, i32 45, i32 4, i32 11, i32 4, i32 5, i32 35, i32 2, i32 33, i32 0>
+  %V512 = shufflevector <64 x i8> %src512, <64 x i8> %src512_1, <64 x i32> <i32 124, i32 62, i32 71, i32 127, i32 58, i32 60, i32 55, i32 56, i32 57, i32 54, i32 127, i32 52, i32 50, i32 51, i32 49, i32 48, i32 47, i32 45, i32 44, i32 44, i32 100, i32 42, i32 41, i32 39, i32 39, i32 40, i32 37, i32 38, i32 35, i32 34, i32 32, i32 33, i32 31, i32 27, i32 30, i32 28, i32 29, i32 26, i32 25, i32 24, i32 23, i32 99, i32 21, i32 20, i32 19, i32 18, i32 17, i32 72, i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 5, i32 6, i32 4, i32 66, i32 2, i32 1, i32 0>
   ret void
 }
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost-inseltpoison.ll b/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost-inseltpoison.ll
index bd642791..c53da90 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost-inseltpoison.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost-inseltpoison.ll
@@ -531,31 +531,31 @@ define <4 x i64> @splatvar_shift_v4i64(<4 x i64> %a, i64 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v4i64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> poison, i64 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:7 SizeLat:13 for: %shift = ashr <4 x i64> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v4i64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> poison, i64 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:5 Lat:6 SizeLat:9 for: %shift = ashr <4 x i64> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v4i64'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> poison, i64 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:7 SizeLat:13 for: %shift = ashr <4 x i64> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v4i64'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> poison, i64 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:5 Lat:6 SizeLat:9 for: %shift = ashr <4 x i64> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v4i64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> poison, i64 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:2 for: %shift = ashr <4 x i64> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
@@ -574,31 +574,31 @@ define <8 x i64> @splatvar_shift_v8i64(<8 x i64> %a, i64 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v8i64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> poison, i64 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:20 Lat:14 SizeLat:26 for: %shift = ashr <8 x i64> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v8i64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> poison, i64 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:10 Lat:12 SizeLat:18 for: %shift = ashr <8 x i64> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v8i64'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> poison, i64 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:20 Lat:14 SizeLat:26 for: %shift = ashr <8 x i64> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v8i64'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> poison, i64 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:10 Lat:12 SizeLat:18 for: %shift = ashr <8 x i64> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v8i64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> poison, i64 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:2 for: %shift = ashr <8 x i64> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
@@ -660,31 +660,31 @@ define <8 x i32> @splatvar_shift_v8i32(<8 x i32> %a, i32 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v8i32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> poison, i32 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = ashr <8 x i32> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v8i32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> poison, i32 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <8 x i32> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v8i32'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> poison, i32 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = ashr <8 x i32> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v8i32'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> poison, i32 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <8 x i32> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v8i32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> poison, i32 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <8 x i32> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
@@ -703,31 +703,31 @@ define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v16i32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> poison, i32 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = ashr <16 x i32> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v16i32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> poison, i32 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = ashr <16 x i32> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v16i32'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> poison, i32 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = ashr <16 x i32> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v16i32'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> poison, i32 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = ashr <16 x i32> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v16i32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> poison, i32 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <16 x i32> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
@@ -740,43 +740,43 @@ define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
 define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v8i16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:2 SizeLat:2 for: %shift = ashr <8 x i16> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v8i16'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:2 SizeLat:2 for: %shift = ashr <8 x i16> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v8i16'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %shift = ashr <8 x i16> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v8i16'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %shift = ashr <8 x i16> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v8i16'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %shift = ashr <8 x i16> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v8i16'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %shift = ashr <8 x i16> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v8i16'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %shift = ashr <8 x i16> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
@@ -789,43 +789,43 @@ define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
 define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v16i16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:4 for: %shift = ashr <16 x i16> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v16i16'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:4 for: %shift = ashr <16 x i16> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v16i16'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = ashr <16 x i16> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v16i16'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <16 x i16> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v16i16'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = ashr <16 x i16> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v16i16'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <16 x i16> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v16i16'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <16 x i16> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
@@ -838,67 +838,67 @@ define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
 define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v32i16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:4 Lat:8 SizeLat:8 for: %shift = ashr <32 x i16> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v32i16'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:4 Lat:8 SizeLat:8 for: %shift = ashr <32 x i16> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v32i16'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = ashr <32 x i16> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v32i16'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = ashr <32 x i16> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v32i16'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = ashr <32 x i16> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v32i16'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = ashr <32 x i16> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512F-LABEL: 'splatvar_shift_v32i16'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:10 SizeLat:7 for: %shift = ashr <32 x i16> %a, %splat
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512BW-LABEL: 'splatvar_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512VL-LABEL: 'splatvar_shift_v32i16'
 ; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:10 SizeLat:7 for: %shift = ashr <32 x i16> %a, %splat
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512BWVL-LABEL: 'splatvar_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <32 x i16> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <32 x i16> %a, %splat
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
@@ -911,67 +911,67 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v16i8'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:9 Lat:15 SizeLat:13 for: %shift = ashr <16 x i8> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v16i8'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:9 Lat:15 SizeLat:13 for: %shift = ashr <16 x i8> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v16i8'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:9 Lat:6 SizeLat:13 for: %shift = ashr <16 x i8> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v16i8'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:9 Lat:5 SizeLat:13 for: %shift = ashr <16 x i8> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v16i8'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:1 for: %shift = ashr <16 x i8> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v16i8'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:9 Lat:5 SizeLat:13 for: %shift = ashr <16 x i8> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512F-LABEL: 'splatvar_shift_v16i8'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:9 Lat:5 SizeLat:13 for: %shift = ashr <16 x i8> %a, %splat
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512BW-LABEL: 'splatvar_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:12 SizeLat:12 for: %shift = ashr <16 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512VL-LABEL: 'splatvar_shift_v16i8'
 ; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:9 Lat:5 SizeLat:13 for: %shift = ashr <16 x i8> %a, %splat
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512BWVL-LABEL: 'splatvar_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:12 SizeLat:12 for: %shift = ashr <16 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512GFNI-LABEL: 'splatvar_shift_v16i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:12 SizeLat:12 for: %shift = ashr <16 x i8> %a, %splat
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
@@ -984,67 +984,67 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v32i8'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:22 CodeSize:18 Lat:30 SizeLat:26 for: %shift = ashr <32 x i8> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v32i8'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:22 CodeSize:18 Lat:30 SizeLat:26 for: %shift = ashr <32 x i8> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v32i8'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:16 Lat:11 SizeLat:21 for: %shift = ashr <32 x i8> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v32i8'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:11 Lat:9 SizeLat:16 for: %shift = ashr <32 x i8> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v32i8'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:5 Lat:7 SizeLat:6 for: %shift = ashr <32 x i8> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v32i8'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:11 Lat:9 SizeLat:16 for: %shift = ashr <32 x i8> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX512F-LABEL: 'splatvar_shift_v32i8'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:11 Lat:9 SizeLat:16 for: %shift = ashr <32 x i8> %a, %splat
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX512BW-LABEL: 'splatvar_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:10 Lat:10 SizeLat:13 for: %shift = ashr <32 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX512VL-LABEL: 'splatvar_shift_v32i8'
 ; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:11 Lat:9 SizeLat:16 for: %shift = ashr <32 x i8> %a, %splat
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX512BWVL-LABEL: 'splatvar_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:10 Lat:10 SizeLat:13 for: %shift = ashr <32 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX512GFNI-LABEL: 'splatvar_shift_v32i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:10 Lat:10 SizeLat:13 for: %shift = ashr <32 x i8> %a, %splat
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
@@ -1057,67 +1057,67 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v64i8'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:44 CodeSize:36 Lat:60 SizeLat:52 for: %shift = ashr <64 x i8> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v64i8'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:44 CodeSize:36 Lat:60 SizeLat:52 for: %shift = ashr <64 x i8> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v64i8'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:32 Lat:22 SizeLat:42 for: %shift = ashr <64 x i8> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v64i8'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:22 Lat:18 SizeLat:32 for: %shift = ashr <64 x i8> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v64i8'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:10 Lat:14 SizeLat:12 for: %shift = ashr <64 x i8> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v64i8'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:22 Lat:18 SizeLat:32 for: %shift = ashr <64 x i8> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512F-LABEL: 'splatvar_shift_v64i8'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:37 CodeSize:51 Lat:37 SizeLat:63 for: %shift = ashr <64 x i8> %a, %splat
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512BW-LABEL: 'splatvar_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:10 Lat:10 SizeLat:15 for: %shift = ashr <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512VL-LABEL: 'splatvar_shift_v64i8'
 ; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:37 CodeSize:51 Lat:37 SizeLat:63 for: %shift = ashr <64 x i8> %a, %splat
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512BWVL-LABEL: 'splatvar_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:10 Lat:10 SizeLat:15 for: %shift = ashr <64 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:10 Lat:10 SizeLat:15 for: %shift = ashr <64 x i8> %a, %splat
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost.ll b/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost.ll
index fdf011e..7a06269 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-ashr-cost.ll
@@ -531,31 +531,31 @@ define <4 x i64> @splatvar_shift_v4i64(<4 x i64> %a, i64 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v4i64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:7 SizeLat:13 for: %shift = ashr <4 x i64> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v4i64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:5 Lat:6 SizeLat:9 for: %shift = ashr <4 x i64> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v4i64'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:7 SizeLat:13 for: %shift = ashr <4 x i64> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v4i64'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:5 Lat:6 SizeLat:9 for: %shift = ashr <4 x i64> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v4i64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:2 for: %shift = ashr <4 x i64> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
@@ -574,31 +574,31 @@ define <8 x i64> @splatvar_shift_v8i64(<8 x i64> %a, i64 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v8i64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:20 Lat:14 SizeLat:26 for: %shift = ashr <8 x i64> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v8i64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:10 Lat:12 SizeLat:18 for: %shift = ashr <8 x i64> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v8i64'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:20 Lat:14 SizeLat:26 for: %shift = ashr <8 x i64> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v8i64'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:10 Lat:12 SizeLat:18 for: %shift = ashr <8 x i64> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v8i64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:2 for: %shift = ashr <8 x i64> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
@@ -660,31 +660,31 @@ define <8 x i32> @splatvar_shift_v8i32(<8 x i32> %a, i32 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v8i32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = ashr <8 x i32> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v8i32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <8 x i32> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v8i32'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = ashr <8 x i32> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v8i32'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <8 x i32> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v8i32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <8 x i32> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
@@ -703,31 +703,31 @@ define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v16i32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = ashr <16 x i32> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v16i32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = ashr <16 x i32> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v16i32'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = ashr <16 x i32> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v16i32'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = ashr <16 x i32> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v16i32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <16 x i32> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
@@ -740,43 +740,43 @@ define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
 define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v8i16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:2 SizeLat:2 for: %shift = ashr <8 x i16> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v8i16'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:2 SizeLat:2 for: %shift = ashr <8 x i16> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v8i16'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %shift = ashr <8 x i16> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v8i16'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %shift = ashr <8 x i16> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v8i16'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %shift = ashr <8 x i16> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v8i16'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %shift = ashr <8 x i16> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v8i16'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %shift = ashr <8 x i16> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
@@ -789,43 +789,43 @@ define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
 define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v16i16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:4 for: %shift = ashr <16 x i16> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v16i16'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:4 for: %shift = ashr <16 x i16> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v16i16'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = ashr <16 x i16> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v16i16'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <16 x i16> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v16i16'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = ashr <16 x i16> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v16i16'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <16 x i16> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v16i16'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <16 x i16> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
@@ -838,67 +838,67 @@ define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
 define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v32i16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:4 Lat:8 SizeLat:8 for: %shift = ashr <32 x i16> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v32i16'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:4 Lat:8 SizeLat:8 for: %shift = ashr <32 x i16> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v32i16'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = ashr <32 x i16> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v32i16'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = ashr <32 x i16> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v32i16'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = ashr <32 x i16> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v32i16'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = ashr <32 x i16> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512F-LABEL: 'splatvar_shift_v32i16'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:10 SizeLat:7 for: %shift = ashr <32 x i16> %a, %splat
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512BW-LABEL: 'splatvar_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512VL-LABEL: 'splatvar_shift_v32i16'
 ; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:10 SizeLat:7 for: %shift = ashr <32 x i16> %a, %splat
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512BWVL-LABEL: 'splatvar_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <32 x i16> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = ashr <32 x i16> %a, %splat
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
@@ -911,67 +911,67 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v16i8'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:9 Lat:15 SizeLat:13 for: %shift = ashr <16 x i8> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v16i8'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:11 CodeSize:9 Lat:15 SizeLat:13 for: %shift = ashr <16 x i8> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v16i8'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:9 Lat:6 SizeLat:13 for: %shift = ashr <16 x i8> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v16i8'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:9 Lat:5 SizeLat:13 for: %shift = ashr <16 x i8> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v16i8'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:1 for: %shift = ashr <16 x i8> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v16i8'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:9 Lat:5 SizeLat:13 for: %shift = ashr <16 x i8> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512F-LABEL: 'splatvar_shift_v16i8'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:9 Lat:5 SizeLat:13 for: %shift = ashr <16 x i8> %a, %splat
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512BW-LABEL: 'splatvar_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:12 SizeLat:12 for: %shift = ashr <16 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512VL-LABEL: 'splatvar_shift_v16i8'
 ; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:9 Lat:5 SizeLat:13 for: %shift = ashr <16 x i8> %a, %splat
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512BWVL-LABEL: 'splatvar_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:12 SizeLat:12 for: %shift = ashr <16 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512GFNI-LABEL: 'splatvar_shift_v16i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:8 Lat:12 SizeLat:12 for: %shift = ashr <16 x i8> %a, %splat
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
@@ -984,67 +984,67 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v32i8'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:22 CodeSize:18 Lat:30 SizeLat:26 for: %shift = ashr <32 x i8> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v32i8'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:22 CodeSize:18 Lat:30 SizeLat:26 for: %shift = ashr <32 x i8> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v32i8'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:10 CodeSize:16 Lat:11 SizeLat:21 for: %shift = ashr <32 x i8> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v32i8'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:11 Lat:9 SizeLat:16 for: %shift = ashr <32 x i8> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v32i8'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:5 Lat:7 SizeLat:6 for: %shift = ashr <32 x i8> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v32i8'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:11 Lat:9 SizeLat:16 for: %shift = ashr <32 x i8> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX512F-LABEL: 'splatvar_shift_v32i8'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:11 Lat:9 SizeLat:16 for: %shift = ashr <32 x i8> %a, %splat
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX512BW-LABEL: 'splatvar_shift_v32i8'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:10 Lat:10 SizeLat:13 for: %shift = ashr <32 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX512VL-LABEL: 'splatvar_shift_v32i8'
 ; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:11 Lat:9 SizeLat:16 for: %shift = ashr <32 x i8> %a, %splat
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX512BWVL-LABEL: 'splatvar_shift_v32i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:10 Lat:10 SizeLat:13 for: %shift = ashr <32 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX512GFNI-LABEL: 'splatvar_shift_v32i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:10 Lat:10 SizeLat:13 for: %shift = ashr <32 x i8> %a, %splat
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
@@ -1057,67 +1057,67 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v64i8'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:44 CodeSize:36 Lat:60 SizeLat:52 for: %shift = ashr <64 x i8> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v64i8'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:44 CodeSize:36 Lat:60 SizeLat:52 for: %shift = ashr <64 x i8> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v64i8'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:20 CodeSize:32 Lat:22 SizeLat:42 for: %shift = ashr <64 x i8> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v64i8'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:22 Lat:18 SizeLat:32 for: %shift = ashr <64 x i8> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v64i8'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:10 Lat:14 SizeLat:12 for: %shift = ashr <64 x i8> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v64i8'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:22 Lat:18 SizeLat:32 for: %shift = ashr <64 x i8> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512F-LABEL: 'splatvar_shift_v64i8'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:37 CodeSize:51 Lat:37 SizeLat:63 for: %shift = ashr <64 x i8> %a, %splat
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512BW-LABEL: 'splatvar_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:10 Lat:10 SizeLat:15 for: %shift = ashr <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512VL-LABEL: 'splatvar_shift_v64i8'
 ; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:37 CodeSize:51 Lat:37 SizeLat:63 for: %shift = ashr <64 x i8> %a, %splat
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512BWVL-LABEL: 'splatvar_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:10 Lat:10 SizeLat:15 for: %shift = ashr <64 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:10 Lat:10 SizeLat:15 for: %shift = ashr <64 x i8> %a, %splat
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost-inseltpoison.ll b/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost-inseltpoison.ll
index 2f70a4b..ee7546f 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost-inseltpoison.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost-inseltpoison.ll
@@ -527,31 +527,31 @@ define <4 x i64> @splatvar_shift_v4i64(<4 x i64> %a, i64 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v4i64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> poison, i64 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:4 Lat:7 SizeLat:6 for: %shift = lshr <4 x i64> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v4i64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> poison, i64 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:2 for: %shift = lshr <4 x i64> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v4i64'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> poison, i64 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:4 Lat:7 SizeLat:6 for: %shift = lshr <4 x i64> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v4i64'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> poison, i64 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:2 for: %shift = lshr <4 x i64> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v4i64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> poison, i64 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:2 for: %shift = lshr <4 x i64> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
@@ -570,31 +570,31 @@ define <8 x i64> @splatvar_shift_v8i64(<8 x i64> %a, i64 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v8i64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> poison, i64 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:8 Lat:14 SizeLat:12 for: %shift = lshr <8 x i64> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v8i64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> poison, i64 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:8 SizeLat:4 for: %shift = lshr <8 x i64> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v8i64'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> poison, i64 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:8 Lat:14 SizeLat:12 for: %shift = lshr <8 x i64> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v8i64'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> poison, i64 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:8 SizeLat:4 for: %shift = lshr <8 x i64> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v8i64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> poison, i64 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:2 for: %shift = lshr <8 x i64> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
@@ -656,31 +656,31 @@ define <8 x i32> @splatvar_shift_v8i32(<8 x i32> %a, i32 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v8i32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> poison, i32 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = lshr <8 x i32> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v8i32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> poison, i32 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <8 x i32> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v8i32'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> poison, i32 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = lshr <8 x i32> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v8i32'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> poison, i32 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <8 x i32> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v8i32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> poison, i32 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <8 x i32> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
@@ -699,31 +699,31 @@ define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v16i32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> poison, i32 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = lshr <16 x i32> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v16i32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> poison, i32 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = lshr <16 x i32> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v16i32'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> poison, i32 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = lshr <16 x i32> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v16i32'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> poison, i32 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = lshr <16 x i32> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v16i32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> poison, i32 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <16 x i32> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
@@ -736,43 +736,43 @@ define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
 define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v8i16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:2 SizeLat:2 for: %shift = lshr <8 x i16> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v8i16'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:2 SizeLat:2 for: %shift = lshr <8 x i16> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v8i16'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %shift = lshr <8 x i16> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v8i16'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %shift = lshr <8 x i16> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v8i16'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %shift = lshr <8 x i16> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v8i16'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %shift = lshr <8 x i16> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v8i16'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %shift = lshr <8 x i16> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
@@ -785,43 +785,43 @@ define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
 define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v16i16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:4 for: %shift = lshr <16 x i16> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v16i16'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:4 for: %shift = lshr <16 x i16> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v16i16'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = lshr <16 x i16> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v16i16'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <16 x i16> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v16i16'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = lshr <16 x i16> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v16i16'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <16 x i16> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v16i16'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <16 x i16> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
@@ -834,67 +834,67 @@ define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
 define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v32i16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:4 Lat:8 SizeLat:8 for: %shift = lshr <32 x i16> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v32i16'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:4 Lat:8 SizeLat:8 for: %shift = lshr <32 x i16> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v32i16'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = lshr <32 x i16> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v32i16'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = lshr <32 x i16> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v32i16'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = lshr <32 x i16> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v32i16'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = lshr <32 x i16> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512F-LABEL: 'splatvar_shift_v32i16'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:10 SizeLat:7 for: %shift = lshr <32 x i16> %a, %splat
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512BW-LABEL: 'splatvar_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512VL-LABEL: 'splatvar_shift_v32i16'
 ; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:10 SizeLat:7 for: %shift = lshr <32 x i16> %a, %splat
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512BWVL-LABEL: 'splatvar_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <32 x i16> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <32 x i16> %a, %splat
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
@@ -907,67 +907,67 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v16i8'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:5 Lat:13 SizeLat:9 for: %shift = lshr <16 x i8> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v16i8'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:5 Lat:13 SizeLat:9 for: %shift = lshr <16 x i8> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v16i8'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:5 Lat:8 SizeLat:8 for: %shift = lshr <16 x i8> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v16i8'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:9 SizeLat:8 for: %shift = lshr <16 x i8> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v16i8'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:1 for: %shift = lshr <16 x i8> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v16i8'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:9 SizeLat:8 for: %shift = lshr <16 x i8> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512F-LABEL: 'splatvar_shift_v16i8'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:9 SizeLat:8 for: %shift = lshr <16 x i8> %a, %splat
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512BW-LABEL: 'splatvar_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:10 SizeLat:8 for: %shift = lshr <16 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512VL-LABEL: 'splatvar_shift_v16i8'
 ; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:9 SizeLat:8 for: %shift = lshr <16 x i8> %a, %splat
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512BWVL-LABEL: 'splatvar_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:10 SizeLat:8 for: %shift = lshr <16 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512GFNI-LABEL: 'splatvar_shift_v16i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:10 SizeLat:8 for: %shift = lshr <16 x i8> %a, %splat
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
@@ -980,43 +980,43 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v32i8'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:10 Lat:26 SizeLat:18 for: %shift = lshr <32 x i8> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v32i8'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:10 Lat:26 SizeLat:18 for: %shift = lshr <32 x i8> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v32i8'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:10 Lat:9 SizeLat:14 for: %shift = lshr <32 x i8> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v32i8'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:7 Lat:8 SizeLat:9 for: %shift = lshr <32 x i8> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v32i8'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:5 Lat:7 SizeLat:6 for: %shift = lshr <32 x i8> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v32i8'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:7 Lat:8 SizeLat:9 for: %shift = lshr <32 x i8> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v32i8'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:7 Lat:8 SizeLat:9 for: %shift = lshr <32 x i8> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
@@ -1029,67 +1029,67 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v64i8'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:20 Lat:52 SizeLat:36 for: %shift = lshr <64 x i8> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v64i8'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:20 Lat:52 SizeLat:36 for: %shift = lshr <64 x i8> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v64i8'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:20 Lat:18 SizeLat:28 for: %shift = lshr <64 x i8> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v64i8'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:14 Lat:16 SizeLat:18 for: %shift = lshr <64 x i8> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v64i8'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:10 Lat:14 SizeLat:12 for: %shift = lshr <64 x i8> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v64i8'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:14 Lat:16 SizeLat:18 for: %shift = lshr <64 x i8> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512F-LABEL: 'splatvar_shift_v64i8'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:30 Lat:19 SizeLat:36 for: %shift = lshr <64 x i8> %a, %splat
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512BW-LABEL: 'splatvar_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:7 Lat:8 SizeLat:10 for: %shift = lshr <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512VL-LABEL: 'splatvar_shift_v64i8'
 ; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:30 Lat:19 SizeLat:36 for: %shift = lshr <64 x i8> %a, %splat
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512BWVL-LABEL: 'splatvar_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:7 Lat:8 SizeLat:10 for: %shift = lshr <64 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:7 Lat:8 SizeLat:10 for: %shift = lshr <64 x i8> %a, %splat
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost.ll b/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost.ll
index 78fd9f33..a508c2a 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-lshr-cost.ll
@@ -527,31 +527,31 @@ define <4 x i64> @splatvar_shift_v4i64(<4 x i64> %a, i64 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v4i64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:4 Lat:7 SizeLat:6 for: %shift = lshr <4 x i64> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v4i64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:2 for: %shift = lshr <4 x i64> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v4i64'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:4 Lat:7 SizeLat:6 for: %shift = lshr <4 x i64> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v4i64'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:2 for: %shift = lshr <4 x i64> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v4i64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:2 for: %shift = lshr <4 x i64> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
@@ -570,31 +570,31 @@ define <8 x i64> @splatvar_shift_v8i64(<8 x i64> %a, i64 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v8i64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:8 Lat:14 SizeLat:12 for: %shift = lshr <8 x i64> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v8i64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:8 SizeLat:4 for: %shift = lshr <8 x i64> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v8i64'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:8 Lat:14 SizeLat:12 for: %shift = lshr <8 x i64> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v8i64'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:8 SizeLat:4 for: %shift = lshr <8 x i64> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v8i64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:2 for: %shift = lshr <8 x i64> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
@@ -656,31 +656,31 @@ define <8 x i32> @splatvar_shift_v8i32(<8 x i32> %a, i32 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v8i32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = lshr <8 x i32> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v8i32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <8 x i32> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v8i32'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = lshr <8 x i32> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v8i32'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <8 x i32> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v8i32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <8 x i32> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
@@ -699,31 +699,31 @@ define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v16i32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = lshr <16 x i32> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v16i32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = lshr <16 x i32> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v16i32'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = lshr <16 x i32> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v16i32'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = lshr <16 x i32> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v16i32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <16 x i32> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
@@ -736,43 +736,43 @@ define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
 define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v8i16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:2 SizeLat:2 for: %shift = lshr <8 x i16> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v8i16'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:2 SizeLat:2 for: %shift = lshr <8 x i16> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v8i16'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %shift = lshr <8 x i16> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v8i16'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %shift = lshr <8 x i16> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v8i16'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %shift = lshr <8 x i16> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v8i16'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %shift = lshr <8 x i16> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v8i16'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %shift = lshr <8 x i16> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
@@ -785,43 +785,43 @@ define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
 define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v16i16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:4 for: %shift = lshr <16 x i16> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v16i16'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:4 for: %shift = lshr <16 x i16> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v16i16'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = lshr <16 x i16> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v16i16'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <16 x i16> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v16i16'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = lshr <16 x i16> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v16i16'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <16 x i16> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v16i16'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <16 x i16> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
@@ -834,67 +834,67 @@ define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
 define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v32i16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:4 Lat:8 SizeLat:8 for: %shift = lshr <32 x i16> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v32i16'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:4 Lat:8 SizeLat:8 for: %shift = lshr <32 x i16> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v32i16'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = lshr <32 x i16> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v32i16'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = lshr <32 x i16> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v32i16'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = lshr <32 x i16> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v32i16'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = lshr <32 x i16> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512F-LABEL: 'splatvar_shift_v32i16'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:10 SizeLat:7 for: %shift = lshr <32 x i16> %a, %splat
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512BW-LABEL: 'splatvar_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512VL-LABEL: 'splatvar_shift_v32i16'
 ; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:10 SizeLat:7 for: %shift = lshr <32 x i16> %a, %splat
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512BWVL-LABEL: 'splatvar_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <32 x i16> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = lshr <32 x i16> %a, %splat
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
@@ -907,67 +907,67 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v16i8'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:5 Lat:13 SizeLat:9 for: %shift = lshr <16 x i8> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v16i8'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:5 Lat:13 SizeLat:9 for: %shift = lshr <16 x i8> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v16i8'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:5 Lat:8 SizeLat:8 for: %shift = lshr <16 x i8> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v16i8'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:9 SizeLat:8 for: %shift = lshr <16 x i8> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v16i8'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:3 SizeLat:1 for: %shift = lshr <16 x i8> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v16i8'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:9 SizeLat:8 for: %shift = lshr <16 x i8> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512F-LABEL: 'splatvar_shift_v16i8'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:9 SizeLat:8 for: %shift = lshr <16 x i8> %a, %splat
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512BW-LABEL: 'splatvar_shift_v16i8'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:10 SizeLat:8 for: %shift = lshr <16 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512VL-LABEL: 'splatvar_shift_v16i8'
 ; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:9 SizeLat:8 for: %shift = lshr <16 x i8> %a, %splat
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512BWVL-LABEL: 'splatvar_shift_v16i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:10 SizeLat:8 for: %shift = lshr <16 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512GFNI-LABEL: 'splatvar_shift_v16i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:10 SizeLat:8 for: %shift = lshr <16 x i8> %a, %splat
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
@@ -980,43 +980,43 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v32i8'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:10 Lat:26 SizeLat:18 for: %shift = lshr <32 x i8> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v32i8'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:10 Lat:26 SizeLat:18 for: %shift = lshr <32 x i8> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v32i8'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:10 Lat:9 SizeLat:14 for: %shift = lshr <32 x i8> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v32i8'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:7 Lat:8 SizeLat:9 for: %shift = lshr <32 x i8> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v32i8'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:5 Lat:7 SizeLat:6 for: %shift = lshr <32 x i8> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v32i8'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:7 Lat:8 SizeLat:9 for: %shift = lshr <32 x i8> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v32i8'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:7 Lat:8 SizeLat:9 for: %shift = lshr <32 x i8> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
@@ -1029,67 +1029,67 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v64i8'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:20 Lat:52 SizeLat:36 for: %shift = lshr <64 x i8> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v64i8'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:20 Lat:52 SizeLat:36 for: %shift = lshr <64 x i8> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v64i8'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:20 Lat:18 SizeLat:28 for: %shift = lshr <64 x i8> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v64i8'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:14 Lat:16 SizeLat:18 for: %shift = lshr <64 x i8> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v64i8'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:12 CodeSize:10 Lat:14 SizeLat:12 for: %shift = lshr <64 x i8> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v64i8'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:14 Lat:16 SizeLat:18 for: %shift = lshr <64 x i8> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512F-LABEL: 'splatvar_shift_v64i8'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:30 Lat:19 SizeLat:36 for: %shift = lshr <64 x i8> %a, %splat
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512BW-LABEL: 'splatvar_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:7 Lat:8 SizeLat:10 for: %shift = lshr <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512VL-LABEL: 'splatvar_shift_v64i8'
 ; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:30 Lat:19 SizeLat:36 for: %shift = lshr <64 x i8> %a, %splat
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512BWVL-LABEL: 'splatvar_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:7 Lat:8 SizeLat:10 for: %shift = lshr <64 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:7 Lat:8 SizeLat:10 for: %shift = lshr <64 x i8> %a, %splat
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-shl-cost-inseltpoison.ll b/llvm/test/Analysis/CostModel/X86/vshift-shl-cost-inseltpoison.ll
index adedca8..375dac5 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-shl-cost-inseltpoison.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-shl-cost-inseltpoison.ll
@@ -567,31 +567,31 @@ define <4 x i64> @splatvar_shift_v4i64(<4 x i64> %a, i64 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v4i64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> poison, i64 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:4 Lat:7 SizeLat:6 for: %shift = shl <4 x i64> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v4i64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> poison, i64 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:2 for: %shift = shl <4 x i64> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v4i64'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> poison, i64 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:4 Lat:7 SizeLat:6 for: %shift = shl <4 x i64> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v4i64'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> poison, i64 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:2 for: %shift = shl <4 x i64> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v4i64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> poison, i64 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> poison, <4 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:2 for: %shift = shl <4 x i64> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
@@ -610,31 +610,31 @@ define <8 x i64> @splatvar_shift_v8i64(<8 x i64> %a, i64 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v8i64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> poison, i64 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:8 Lat:14 SizeLat:12 for: %shift = shl <8 x i64> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v8i64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> poison, i64 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:8 SizeLat:4 for: %shift = shl <8 x i64> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v8i64'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> poison, i64 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:8 Lat:14 SizeLat:12 for: %shift = shl <8 x i64> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v8i64'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> poison, i64 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:8 SizeLat:4 for: %shift = shl <8 x i64> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v8i64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> poison, i64 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> poison, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:2 for: %shift = shl <8 x i64> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
@@ -696,31 +696,31 @@ define <8 x i32> @splatvar_shift_v8i32(<8 x i32> %a, i32 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v8i32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> poison, i32 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = shl <8 x i32> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v8i32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> poison, i32 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <8 x i32> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v8i32'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> poison, i32 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = shl <8 x i32> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v8i32'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> poison, i32 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <8 x i32> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v8i32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> poison, i32 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> poison, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <8 x i32> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
@@ -739,31 +739,31 @@ define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v16i32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> poison, i32 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = shl <16 x i32> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v16i32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> poison, i32 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = shl <16 x i32> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v16i32'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> poison, i32 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = shl <16 x i32> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v16i32'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> poison, i32 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = shl <16 x i32> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v16i32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> poison, i32 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> poison, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <16 x i32> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
@@ -776,49 +776,49 @@ define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
 define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v8i16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:2 SizeLat:2 for: %shift = shl <8 x i16> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v8i16'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:2 SizeLat:2 for: %shift = shl <8 x i16> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v8i16'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %shift = shl <8 x i16> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v8i16'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %shift = shl <8 x i16> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v8i16'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %shift = shl <8 x i16> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v8i16'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %shift = shl <8 x i16> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v8i16'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %shift = shl <8 x i16> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; SLM-LABEL: 'splatvar_shift_v8i16'
 ; SLM-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> poison, i16 %b, i32 0
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> poison, <8 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:2 SizeLat:2 for: %shift = shl <8 x i16> %a, %splat
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
@@ -831,49 +831,49 @@ define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
 define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v16i16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:4 for: %shift = shl <16 x i16> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v16i16'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:4 for: %shift = shl <16 x i16> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v16i16'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = shl <16 x i16> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v16i16'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <16 x i16> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v16i16'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = shl <16 x i16> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v16i16'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <16 x i16> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v16i16'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <16 x i16> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; SLM-LABEL: 'splatvar_shift_v16i16'
 ; SLM-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> poison, i16 %b, i32 0
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> poison, <16 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:4 for: %shift = shl <16 x i16> %a, %splat
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
@@ -886,73 +886,73 @@ define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
 define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v32i16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:4 Lat:8 SizeLat:8 for: %shift = shl <32 x i16> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v32i16'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:4 Lat:8 SizeLat:8 for: %shift = shl <32 x i16> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v32i16'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = shl <32 x i16> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v32i16'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = shl <32 x i16> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v32i16'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = shl <32 x i16> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v32i16'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = shl <32 x i16> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512F-LABEL: 'splatvar_shift_v32i16'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:10 SizeLat:7 for: %shift = shl <32 x i16> %a, %splat
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512BW-LABEL: 'splatvar_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512VL-LABEL: 'splatvar_shift_v32i16'
 ; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:10 SizeLat:7 for: %shift = shl <32 x i16> %a, %splat
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512BWVL-LABEL: 'splatvar_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <32 x i16> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; SLM-LABEL: 'splatvar_shift_v32i16'
 ; SLM-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:4 Lat:8 SizeLat:8 for: %shift = shl <32 x i16> %a, %splat
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> poison, i16 %b, i32 0
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> poison, <32 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <32 x i16> %a, %splat
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
@@ -965,49 +965,49 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v16i8'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:6 Lat:10 SizeLat:9 for: %shift = shl <16 x i8> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v16i8'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:6 Lat:10 SizeLat:9 for: %shift = shl <16 x i8> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v16i8'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:4 SizeLat:8 for: %shift = shl <16 x i8> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v16i8'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:5 SizeLat:7 for: %shift = shl <16 x i8> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v16i8'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %shift = shl <16 x i8> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v16i8'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:5 SizeLat:7 for: %shift = shl <16 x i8> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v16i8'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:5 SizeLat:7 for: %shift = shl <16 x i8> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; SLM-LABEL: 'splatvar_shift_v16i8'
 ; SLM-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> poison, i8 %b, i32 0
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> poison, <16 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:6 Lat:10 SizeLat:9 for: %shift = shl <16 x i8> %a, %splat
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
@@ -1020,49 +1020,49 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v32i8'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:12 Lat:20 SizeLat:18 for: %shift = shl <32 x i8> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v32i8'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:12 Lat:20 SizeLat:18 for: %shift = shl <32 x i8> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v32i8'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:11 Lat:8 SizeLat:14 for: %shift = shl <32 x i8> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v32i8'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:7 SizeLat:8 for: %shift = shl <32 x i8> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v32i8'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:5 Lat:7 SizeLat:6 for: %shift = shl <32 x i8> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v32i8'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:7 SizeLat:8 for: %shift = shl <32 x i8> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v32i8'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:7 SizeLat:8 for: %shift = shl <32 x i8> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; SLM-LABEL: 'splatvar_shift_v32i8'
 ; SLM-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> poison, i8 %b, i32 0
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> poison, <32 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:12 Lat:20 SizeLat:18 for: %shift = shl <32 x i8> %a, %splat
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
@@ -1075,73 +1075,73 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v64i8'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:24 Lat:40 SizeLat:36 for: %shift = shl <64 x i8> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v64i8'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:24 Lat:40 SizeLat:36 for: %shift = shl <64 x i8> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v64i8'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:22 Lat:16 SizeLat:28 for: %shift = shl <64 x i8> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v64i8'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:12 Lat:14 SizeLat:16 for: %shift = shl <64 x i8> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v64i8'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:10 Lat:14 SizeLat:12 for: %shift = shl <64 x i8> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v64i8'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:12 Lat:14 SizeLat:16 for: %shift = shl <64 x i8> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512F-LABEL: 'splatvar_shift_v64i8'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:27 Lat:19 SizeLat:33 for: %shift = shl <64 x i8> %a, %splat
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512BW-LABEL: 'splatvar_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:7 SizeLat:8 for: %shift = shl <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512VL-LABEL: 'splatvar_shift_v64i8'
 ; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:27 Lat:19 SizeLat:33 for: %shift = shl <64 x i8> %a, %splat
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512BWVL-LABEL: 'splatvar_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:7 SizeLat:8 for: %shift = shl <64 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; SLM-LABEL: 'splatvar_shift_v64i8'
 ; SLM-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:24 Lat:40 SizeLat:36 for: %shift = shl <64 x i8> %a, %splat
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> poison, i8 %b, i32 0
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> poison, <64 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:7 SizeLat:8 for: %shift = shl <64 x i8> %a, %splat
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
diff --git a/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll b/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll
index a3536ec..b2435be 100644
--- a/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll
+++ b/llvm/test/Analysis/CostModel/X86/vshift-shl-cost.ll
@@ -567,31 +567,31 @@ define <4 x i64> @splatvar_shift_v4i64(<4 x i64> %a, i64 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v4i64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:4 Lat:7 SizeLat:6 for: %shift = shl <4 x i64> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v4i64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:2 for: %shift = shl <4 x i64> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v4i64'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:4 Lat:7 SizeLat:6 for: %shift = shl <4 x i64> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v4i64'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:4 SizeLat:2 for: %shift = shl <4 x i64> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v4i64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <4 x i64> undef, i64 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <4 x i64> %insert, <4 x i64> undef, <4 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:2 for: %shift = shl <4 x i64> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <4 x i64> %shift
 ;
@@ -610,31 +610,31 @@ define <8 x i64> @splatvar_shift_v8i64(<8 x i64> %a, i64 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v8i64'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:8 Lat:14 SizeLat:12 for: %shift = shl <8 x i64> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v8i64'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:8 SizeLat:4 for: %shift = shl <8 x i64> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v8i64'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:8 Lat:14 SizeLat:12 for: %shift = shl <8 x i64> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v8i64'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:8 SizeLat:4 for: %shift = shl <8 x i64> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v8i64'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i64> undef, i64 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i64> %insert, <8 x i64> undef, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:4 SizeLat:2 for: %shift = shl <8 x i64> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i64> %shift
 ;
@@ -696,31 +696,31 @@ define <8 x i32> @splatvar_shift_v8i32(<8 x i32> %a, i32 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v8i32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = shl <8 x i32> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v8i32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <8 x i32> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v8i32'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = shl <8 x i32> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v8i32'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <8 x i32> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v8i32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i32> undef, i32 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i32> %insert, <8 x i32> undef, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <8 x i32> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i32> %shift
 ;
@@ -739,31 +739,31 @@ define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
 ;
 ; AVX1-LABEL: 'splatvar_shift_v16i32'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = shl <16 x i32> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v16i32'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = shl <16 x i32> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v16i32'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:3 SizeLat:3 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = shl <16 x i32> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v16i32'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = shl <16 x i32> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v16i32'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i32> undef, i32 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i32> %insert, <16 x i32> undef, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <16 x i32> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i32> %shift
 ;
@@ -776,49 +776,49 @@ define <16 x i32> @splatvar_shift_v16i32(<16 x i32> %a, i32 %b) {
 define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v8i16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:2 SizeLat:2 for: %shift = shl <8 x i16> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v8i16'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:2 SizeLat:2 for: %shift = shl <8 x i16> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v8i16'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %shift = shl <8 x i16> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v8i16'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %shift = shl <8 x i16> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v8i16'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %shift = shl <8 x i16> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v8i16'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %shift = shl <8 x i16> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v8i16'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:2 SizeLat:2 for: %shift = shl <8 x i16> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
 ; SLM-LABEL: 'splatvar_shift_v8i16'
 ; SLM-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <8 x i16> undef, i16 %b, i32 0
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <8 x i16> %insert, <8 x i16> undef, <8 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:1 Lat:2 SizeLat:2 for: %shift = shl <8 x i16> %a, %splat
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <8 x i16> %shift
 ;
@@ -831,49 +831,49 @@ define <8 x i16> @splatvar_shift_v8i16(<8 x i16> %a, i16 %b) {
 define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v16i16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:4 for: %shift = shl <16 x i16> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v16i16'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:4 for: %shift = shl <16 x i16> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v16i16'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = shl <16 x i16> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v16i16'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <16 x i16> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v16i16'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:7 SizeLat:7 for: %shift = shl <16 x i16> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v16i16'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <16 x i16> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v16i16'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <16 x i16> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
 ; SLM-LABEL: 'splatvar_shift_v16i16'
 ; SLM-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i16> undef, i16 %b, i32 0
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i16> %insert, <16 x i16> undef, <16 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:2 Lat:4 SizeLat:4 for: %shift = shl <16 x i16> %a, %splat
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i16> %shift
 ;
@@ -886,73 +886,73 @@ define <16 x i16> @splatvar_shift_v16i16(<16 x i16> %a, i16 %b) {
 define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v32i16'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:2 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:4 Lat:8 SizeLat:8 for: %shift = shl <32 x i16> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v32i16'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:4 Lat:8 SizeLat:8 for: %shift = shl <32 x i16> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v32i16'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = shl <32 x i16> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v32i16'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = shl <32 x i16> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v32i16'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:6 CodeSize:10 Lat:14 SizeLat:14 for: %shift = shl <32 x i16> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v32i16'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:4 Lat:8 SizeLat:6 for: %shift = shl <32 x i16> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512F-LABEL: 'splatvar_shift_v32i16'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:10 SizeLat:7 for: %shift = shl <32 x i16> %a, %splat
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512BW-LABEL: 'splatvar_shift_v32i16'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <32 x i16> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512VL-LABEL: 'splatvar_shift_v32i16'
 ; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:5 CodeSize:5 Lat:10 SizeLat:7 for: %shift = shl <32 x i16> %a, %splat
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512BWVL-LABEL: 'splatvar_shift_v32i16'
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <32 x i16> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; SLM-LABEL: 'splatvar_shift_v32i16'
 ; SLM-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:4 Lat:8 SizeLat:8 for: %shift = shl <32 x i16> %a, %splat
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
 ; AVX512GFNI-LABEL: 'splatvar_shift_v32i16'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i16> undef, i16 %b, i32 0
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i16> %insert, <32 x i16> undef, <32 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:2 Lat:4 SizeLat:3 for: %shift = shl <32 x i16> %a, %splat
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i16> %shift
 ;
@@ -965,49 +965,49 @@ define <32 x i16> @splatvar_shift_v32i16(<32 x i16> %a, i16 %b) {
 define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v16i8'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:6 Lat:10 SizeLat:9 for: %shift = shl <16 x i8> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v16i8'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:6 Lat:10 SizeLat:9 for: %shift = shl <16 x i8> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v16i8'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:4 SizeLat:8 for: %shift = shl <16 x i8> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v16i8'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:5 SizeLat:7 for: %shift = shl <16 x i8> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v16i8'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %shift = shl <16 x i8> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v16i8'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:5 SizeLat:7 for: %shift = shl <16 x i8> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v16i8'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:5 Lat:5 SizeLat:7 for: %shift = shl <16 x i8> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
 ; SLM-LABEL: 'splatvar_shift_v16i8'
 ; SLM-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <16 x i8> undef, i8 %b, i32 0
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <16 x i8> %insert, <16 x i8> undef, <16 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:9 CodeSize:6 Lat:10 SizeLat:9 for: %shift = shl <16 x i8> %a, %splat
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <16 x i8> %shift
 ;
@@ -1020,49 +1020,49 @@ define <16 x i8> @splatvar_shift_v16i8(<16 x i8> %a, i8 %b) {
 define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v32i8'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:12 Lat:20 SizeLat:18 for: %shift = shl <32 x i8> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v32i8'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:12 Lat:20 SizeLat:18 for: %shift = shl <32 x i8> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v32i8'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:7 CodeSize:11 Lat:8 SizeLat:14 for: %shift = shl <32 x i8> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v32i8'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:7 SizeLat:8 for: %shift = shl <32 x i8> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v32i8'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:5 Lat:7 SizeLat:6 for: %shift = shl <32 x i8> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v32i8'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:7 SizeLat:8 for: %shift = shl <32 x i8> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; AVX512-LABEL: 'splatvar_shift_v32i8'
 ; AVX512-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; AVX512-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; AVX512-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:7 SizeLat:8 for: %shift = shl <32 x i8> %a, %splat
 ; AVX512-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
 ; SLM-LABEL: 'splatvar_shift_v32i8'
 ; SLM-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <32 x i8> undef, i8 %b, i32 0
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <32 x i8> %insert, <32 x i8> undef, <32 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:18 CodeSize:12 Lat:20 SizeLat:18 for: %shift = shl <32 x i8> %a, %splat
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <32 x i8> %shift
 ;
@@ -1075,73 +1075,73 @@ define <32 x i8> @splatvar_shift_v32i8(<32 x i8> %a, i8 %b) {
 define <64 x i8> @splatvar_shift_v64i8(<64 x i8> %a, i8 %b) {
 ; SSE2-LABEL: 'splatvar_shift_v64i8'
 ; SSE2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; SSE2-NEXT:  Cost Model: Found costs of 3 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE2-NEXT:  Cost Model: Found costs of RThru:2 CodeSize:3 Lat:3 SizeLat:4 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:24 Lat:40 SizeLat:36 for: %shift = shl <64 x i8> %a, %splat
 ; SSE2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; SSE42-LABEL: 'splatvar_shift_v64i8'
 ; SSE42-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; SSE42-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; SSE42-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:24 Lat:40 SizeLat:36 for: %shift = shl <64 x i8> %a, %splat
 ; SSE42-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX1-LABEL: 'splatvar_shift_v64i8'
 ; AVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:14 CodeSize:22 Lat:16 SizeLat:28 for: %shift = shl <64 x i8> %a, %splat
 ; AVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX2-LABEL: 'splatvar_shift_v64i8'
 ; AVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:12 Lat:14 SizeLat:16 for: %shift = shl <64 x i8> %a, %splat
 ; AVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; XOPAVX1-LABEL: 'splatvar_shift_v64i8'
 ; XOPAVX1-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; XOPAVX1-NEXT:  Cost Model: Found costs of 2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:3 CodeSize:3 Lat:4 SizeLat:6 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:10 Lat:14 SizeLat:12 for: %shift = shl <64 x i8> %a, %splat
 ; XOPAVX1-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; XOPAVX2-LABEL: 'splatvar_shift_v64i8'
 ; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; XOPAVX2-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:8 CodeSize:12 Lat:14 SizeLat:16 for: %shift = shl <64 x i8> %a, %splat
 ; XOPAVX2-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512F-LABEL: 'splatvar_shift_v64i8'
 ; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX512F-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512F-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:27 Lat:19 SizeLat:33 for: %shift = shl <64 x i8> %a, %splat
 ; AVX512F-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512BW-LABEL: 'splatvar_shift_v64i8'
 ; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX512BW-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512BW-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:7 SizeLat:8 for: %shift = shl <64 x i8> %a, %splat
 ; AVX512BW-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512VL-LABEL: 'splatvar_shift_v64i8'
 ; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX512VL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512VL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:15 CodeSize:27 Lat:19 SizeLat:33 for: %shift = shl <64 x i8> %a, %splat
 ; AVX512VL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512BWVL-LABEL: 'splatvar_shift_v64i8'
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX512BWVL-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:7 SizeLat:8 for: %shift = shl <64 x i8> %a, %splat
 ; AVX512BWVL-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; SLM-LABEL: 'splatvar_shift_v64i8'
 ; SLM-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; SLM-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; SLM-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:2 Lat:3 SizeLat:2 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; SLM-NEXT:  Cost Model: Found costs of RThru:36 CodeSize:24 Lat:40 SizeLat:36 for: %shift = shl <64 x i8> %a, %splat
 ; SLM-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
 ; AVX512GFNI-LABEL: 'splatvar_shift_v64i8'
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %insert = insertelement <64 x i8> undef, i8 %b, i32 0
-; AVX512GFNI-NEXT:  Cost Model: Found costs of 1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
+; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:1 CodeSize:1 Lat:3 SizeLat:1 for: %splat = shufflevector <64 x i8> %insert, <64 x i8> undef, <64 x i32> zeroinitializer
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:4 CodeSize:6 Lat:7 SizeLat:8 for: %shift = shl <64 x i8> %a, %splat
 ; AVX512GFNI-NEXT:  Cost Model: Found costs of RThru:0 CodeSize:1 Lat:1 SizeLat:1 for: ret <64 x i8> %shift
 ;
diff --git a/llvm/test/Analysis/DependenceAnalysis/Banerjee.ll b/llvm/test/Analysis/DependenceAnalysis/Banerjee.ll
index d330152..e0def90 100644
--- a/llvm/test/Analysis/DependenceAnalysis/Banerjee.ll
+++ b/llvm/test/Analysis/DependenceAnalysis/Banerjee.ll
@@ -113,7 +113,7 @@ define void @banerjee1(ptr %A, ptr %B, i64 %m, i64 %n) nounwind uwtable ssp {
 ; CHECK-NEXT:  Src: %2 = load i64, ptr %arrayidx6, align 8 --> Dst: store i64 %2, ptr %B.addr.12, align 8
 ; CHECK-NEXT:    da analyze - confused!
 ; CHECK-NEXT:  Src: store i64 %2, ptr %B.addr.12, align 8 --> Dst: store i64 %2, ptr %B.addr.12, align 8
-; CHECK-NEXT:    da analyze - output [* *]!
+; CHECK-NEXT:    da analyze - confused!
 ;
 ; NORMALIZE-LABEL: 'banerjee1'
 ; NORMALIZE-NEXT:  Src: store i64 0, ptr %arrayidx, align 8 --> Dst: store i64 0, ptr %arrayidx, align 8
@@ -127,7 +127,7 @@ define void @banerjee1(ptr %A, ptr %B, i64 %m, i64 %n) nounwind uwtable ssp {
 ; NORMALIZE-NEXT:  Src: %2 = load i64, ptr %arrayidx6, align 8 --> Dst: store i64 %2, ptr %B.addr.12, align 8
 ; NORMALIZE-NEXT:    da analyze - confused!
 ; NORMALIZE-NEXT:  Src: store i64 %2, ptr %B.addr.12, align 8 --> Dst: store i64 %2, ptr %B.addr.12, align 8
-; NORMALIZE-NEXT:    da analyze - output [* *]!
+; NORMALIZE-NEXT:    da analyze - confused!
 ;
 ; DELIN-LABEL: 'banerjee1'
 ; DELIN-NEXT:  Src: store i64 0, ptr %arrayidx, align 8 --> Dst: store i64 0, ptr %arrayidx, align 8
@@ -141,7 +141,7 @@ define void @banerjee1(ptr %A, ptr %B, i64 %m, i64 %n) nounwind uwtable ssp {
 ; DELIN-NEXT:  Src: %2 = load i64, ptr %arrayidx6, align 8 --> Dst: store i64 %2, ptr %B.addr.12, align 8
 ; DELIN-NEXT:    da analyze - confused!
 ; DELIN-NEXT:  Src: store i64 %2, ptr %B.addr.12, align 8 --> Dst: store i64 %2, ptr %B.addr.12, align 8
-; DELIN-NEXT:    da analyze - output [* *]!
+; DELIN-NEXT:    da analyze - confused!
 ;
 entry:
   %cmp4 = icmp sgt i64 %n, 0
diff --git a/llvm/test/Analysis/DependenceAnalysis/FlipFlopBaseAddress.ll b/llvm/test/Analysis/DependenceAnalysis/FlipFlopBaseAddress.ll
index 3e3426a..bf2268b 100644
--- a/llvm/test/Analysis/DependenceAnalysis/FlipFlopBaseAddress.ll
+++ b/llvm/test/Analysis/DependenceAnalysis/FlipFlopBaseAddress.ll
@@ -8,11 +8,11 @@
 define float @bug41488_test1(float %f) {
 ; CHECK-LABEL: 'bug41488_test1'
 ; CHECK-NEXT:  Src: %0 = load float, ptr %p, align 4 --> Dst: %0 = load float, ptr %p, align 4
-; CHECK-NEXT:    da analyze - input [*]!
+; CHECK-NEXT:    da analyze - confused!
 ; CHECK-NEXT:  Src: %0 = load float, ptr %p, align 4 --> Dst: store float %f, ptr %q, align 4
 ; CHECK-NEXT:    da analyze - confused!
 ; CHECK-NEXT:  Src: store float %f, ptr %q, align 4 --> Dst: store float %f, ptr %q, align 4
-; CHECK-NEXT:    da analyze - output [*]!
+; CHECK-NEXT:    da analyze - confused!
 ;
 entry:
   %g = alloca float, align 4
@@ -34,11 +34,11 @@ for.cond.cleanup:
 define void @bug41488_test2(i32 %n) {
 ; CHECK-LABEL: 'bug41488_test2'
 ; CHECK-NEXT:  Src: %0 = load float, ptr %p, align 4 --> Dst: %0 = load float, ptr %p, align 4
-; CHECK-NEXT:    da analyze - input [*]!
+; CHECK-NEXT:    da analyze - confused!
 ; CHECK-NEXT:  Src: %0 = load float, ptr %p, align 4 --> Dst: store float 0.000000e+00, ptr %q, align 4
 ; CHECK-NEXT:    da analyze - confused!
 ; CHECK-NEXT:  Src: store float 0.000000e+00, ptr %q, align 4 --> Dst: store float 0.000000e+00, ptr %q, align 4
-; CHECK-NEXT:    da analyze - output [*]!
+; CHECK-NEXT:    da analyze - confused!
 ;
 entry:
   %g = alloca float, align 4
@@ -68,7 +68,7 @@ define void @bug53942_foo(i32 noundef %n, ptr noalias nocapture noundef writeonl
 ; CHECK-NEXT:  Src: %.pre = load double, ptr %B, align 8 --> Dst: store double %.pre, ptr %arrayidx2, align 8
 ; CHECK-NEXT:    da analyze - confused!
 ; CHECK-NEXT:  Src: store double %.pre, ptr %arrayidx2, align 8 --> Dst: store double %.pre, ptr %arrayidx2, align 8
-; CHECK-NEXT:    da analyze - output [*]!
+; CHECK-NEXT:    da analyze - confused!
 ;
 entry:
   %cmp8 = icmp sgt i32 %n, 1
@@ -99,11 +99,11 @@ for.body:                                         ; preds = %for.body.preheader,
 define void @bug53942_bar(i32 noundef %n, ptr noalias noundef %A, ptr noalias noundef %B) {
 ; CHECK-LABEL: 'bug53942_bar'
 ; CHECK-NEXT:  Src: %0 = load double, ptr %arrayidx, align 8 --> Dst: %0 = load double, ptr %arrayidx, align 8
-; CHECK-NEXT:    da analyze - input [*]!
+; CHECK-NEXT:    da analyze - confused!
 ; CHECK-NEXT:  Src: %0 = load double, ptr %arrayidx, align 8 --> Dst: store double %0, ptr %arrayidx8, align 8
 ; CHECK-NEXT:    da analyze - confused!
 ; CHECK-NEXT:  Src: store double %0, ptr %arrayidx8, align 8 --> Dst: store double %0, ptr %arrayidx8, align 8
-; CHECK-NEXT:    da analyze - output [*]!
+; CHECK-NEXT:    da analyze - confused!
 ;
 entry:
   br label %for.cond
@@ -166,14 +166,14 @@ for.end:                                          ; preds = %for.cond.cleanup
 ;       (j % 2 == 0 ? A[i][j] : A[i][j+1]) = 1;
 ; }
 ;
-; FIXME: There are loop-carried dependencies between the store instruction. For
+; There are loop-carried dependencies between the store instruction. For
 ; example, the value of %ptr0 when (i, j) = (0, 1) is %A+8, which is the same
 ; as when (i, j) = (0, 2).
 
 define void @non_invariant_baseptr_with_identical_obj(ptr %A) {
 ; CHECK-LABEL: 'non_invariant_baseptr_with_identical_obj'
 ; CHECK-NEXT:  Src: store i32 1, ptr %idx, align 4 --> Dst: store i32 1, ptr %idx, align 4
-; CHECK-NEXT:    da analyze - none!
+; CHECK-NEXT:    da analyze - confused!
 ;
 entry:
   br label %loop.i.header
@@ -216,13 +216,13 @@ exit:
 ; Similar to the above case, but ptr0 is loop-invariant with respsect to the
 ; k-loop.
 ;
-; FIXME: Same as the above case, there are loop-carried dependencies between
-; the store.
+; Same as the above case, there are loop-carried dependencies between the
+; store.
 
 define void @non_invariant_baseptr_with_identical_obj2(ptr %A) {
 ; CHECK-LABEL: 'non_invariant_baseptr_with_identical_obj2'
 ; CHECK-NEXT:  Src: store i32 1, ptr %idx, align 4 --> Dst: store i32 1, ptr %idx, align 4
-; CHECK-NEXT:    da analyze - none!
+; CHECK-NEXT:    da analyze - confused!
 ;
 entry:
   br label %loop.i.header
@@ -259,3 +259,58 @@ loop.i.latch:
 exit:
   ret void
 }
+
+; Pseudo-code that is approximately semantically equivalent to the below IR:
+;
+; void f(int A[][32]) {
+;   for (int i = 0; i < 100; i++)
+;     for (int j = 0; j < 15; j++) {
+;       int offset = (j % 2 == 0) ? 1 : 0;
+;       A[i][2 * j + offset + 0] = 1;
+;       A[i][2 * j + offset + 1] = 1;
+;     }
+; }
+;
+; There are loop-carried dependencies between the two stores. For example,
+; A[0][2] is accessed from both the former one when (i, j) = (0, 1) and the
+; latter one when (i, j) = (0, 0).
+;
+define void @non_invariant_baseptr_with_identical_obj3(ptr %A) {
+; CHECK-LABEL: 'non_invariant_baseptr_with_identical_obj3'
+; CHECK-NEXT:  Src: store i32 1, ptr %idx0, align 4 --> Dst: store i32 1, ptr %idx0, align 4
+; CHECK-NEXT:    da analyze - confused!
+; CHECK-NEXT:  Src: store i32 1, ptr %idx0, align 4 --> Dst: store i32 1, ptr %idx1, align 4
+; CHECK-NEXT:    da analyze - confused!
+; CHECK-NEXT:  Src: store i32 1, ptr %idx1, align 4 --> Dst: store i32 1, ptr %idx1, align 4
+; CHECK-NEXT:    da analyze - confused!
+;
+entry:
+  br label %loop.i.header
+
+loop.i.header:
+  %i = phi i32 [ 0, %entry ], [ %i.inc, %loop.i.latch ]
+  %A1 = getelementptr i32, ptr %A, i32 1
+  br label %loop.j
+
+loop.j:
+  %j = phi i32 [ 0, %loop.i.header ], [ %j.inc, %loop.j ]
+  %ptr0 = phi ptr [ %A1, %loop.i.header ], [ %ptr1, %loop.j ]
+  %ptr1 = phi ptr [ %A, %loop.i.header ], [ %ptr0, %loop.j ]
+  %j2_0 = shl i32 %j, 1
+  %j2_1 = add i32 %j2_0, 1
+  %idx0 = getelementptr [32 x i32], ptr %ptr0, i32 %i, i32 %j2_0
+  %idx1 = getelementptr [32 x i32], ptr %ptr0, i32 %i, i32 %j2_1
+  store i32 1, ptr %idx0
+  store i32 1, ptr %idx1
+  %j.inc = add i32 %j, 1
+  %cmp.j = icmp slt i32 %j.inc, 15
+  br i1 %cmp.j, label %loop.j, label %loop.i.latch
+
+loop.i.latch:
+  %i.inc = add i32 %i, 1
+  %cmp.i = icmp slt i32 %i.inc, 100
+  br i1 %cmp.i, label %loop.i.header, label %exit
+
+exit:
+  ret void
+}
diff --git a/llvm/test/Analysis/DependenceAnalysis/GCD.ll b/llvm/test/Analysis/DependenceAnalysis/GCD.ll
index c0e1362..03343e7 100644
--- a/llvm/test/Analysis/DependenceAnalysis/GCD.ll
+++ b/llvm/test/Analysis/DependenceAnalysis/GCD.ll
@@ -398,7 +398,7 @@ define void @gcd6(i64 %n, ptr %A, ptr %B) nounwind uwtable ssp {
 ; CHECK-NEXT:  Src: %2 = load i32, ptr %arrayidx9, align 4 --> Dst: store i32 %2, ptr %B.addr.12, align 4
 ; CHECK-NEXT:    da analyze - confused!
 ; CHECK-NEXT:  Src: store i32 %2, ptr %B.addr.12, align 4 --> Dst: store i32 %2, ptr %B.addr.12, align 4
-; CHECK-NEXT:    da analyze - output [* *]!
+; CHECK-NEXT:    da analyze - confused!
 ;
 entry:
   %cmp4 = icmp sgt i64 %n, 0
@@ -475,7 +475,7 @@ define void @gcd7(i32 %n, ptr %A, ptr %B) nounwind uwtable ssp {
 ; CHECK-NEXT:  Src: %11 = load i32, ptr %arrayidx12, align 4 --> Dst: store i32 %11, ptr %B.addr.12, align 4
 ; CHECK-NEXT:    da analyze - confused!
 ; CHECK-NEXT:  Src: store i32 %11, ptr %B.addr.12, align 4 --> Dst: store i32 %11, ptr %B.addr.12, align 4
-; CHECK-NEXT:    da analyze - output [* *]!
+; CHECK-NEXT:    da analyze - confused!
 ;
 entry:
   %0 = zext i32 %n to i64
@@ -566,7 +566,7 @@ define void @gcd8(i32 %n, ptr %A, ptr %B) nounwind uwtable ssp {
 ; CHECK-NEXT:  Src: %5 = load i32, ptr %arrayidx12, align 4 --> Dst: store i32 %5, ptr %B.addr.12, align 4
 ; CHECK-NEXT:    da analyze - confused!
 ; CHECK-NEXT:  Src: store i32 %5, ptr %B.addr.12, align 4 --> Dst: store i32 %5, ptr %B.addr.12, align 4
-; CHECK-NEXT:    da analyze - output [* *]!
+; CHECK-NEXT:    da analyze - confused!
 ;
 entry:
   %cmp4 = icmp sgt i32 %n, 0
@@ -650,7 +650,7 @@ define void @gcd9(i32 %n, ptr %A, ptr %B) nounwind uwtable ssp {
 ; CHECK-NEXT:  Src: %11 = load i32, ptr %arrayidx12, align 4 --> Dst: store i32 %11, ptr %B.addr.12, align 4
 ; CHECK-NEXT:    da analyze - confused!
 ; CHECK-NEXT:  Src: store i32 %11, ptr %B.addr.12, align 4 --> Dst: store i32 %11, ptr %B.addr.12, align 4
-; CHECK-NEXT:    da analyze - output [* *]!
+; CHECK-NEXT:    da analyze - confused!
 ;
 entry:
   %0 = zext i32 %n to i64
diff --git a/llvm/test/Analysis/DependenceAnalysis/NonAffineExpr.ll b/llvm/test/Analysis/DependenceAnalysis/NonAffineExpr.ll
index d983bd4..3e110ac 100644
--- a/llvm/test/Analysis/DependenceAnalysis/NonAffineExpr.ll
+++ b/llvm/test/Analysis/DependenceAnalysis/NonAffineExpr.ll
@@ -12,7 +12,7 @@ define void @f(ptr %a, i32 %n, i1 %arg) align 2 {
 ; CHECK-NEXT:  Src: %t.2 = load ptr, ptr %a, align 4 --> Dst: %t.4 = load i32, ptr %t.3, align 4
 ; CHECK-NEXT:    da analyze - confused!
 ; CHECK-NEXT:  Src: %t.4 = load i32, ptr %t.3, align 4 --> Dst: %t.4 = load i32, ptr %t.3, align 4
-; CHECK-NEXT:    da analyze - input [* *]!
+; CHECK-NEXT:    da analyze - confused!
 ;
 for.preheader:
   %t.0 = ashr exact i32 %n, 3
diff --git a/llvm/test/Analysis/DependenceAnalysis/Preliminary.ll b/llvm/test/Analysis/DependenceAnalysis/Preliminary.ll
index 4ab8777..8cb0e2a 100644
--- a/llvm/test/Analysis/DependenceAnalysis/Preliminary.ll
+++ b/llvm/test/Analysis/DependenceAnalysis/Preliminary.ll
@@ -69,7 +69,7 @@ define void @p2(i64 %n, ptr %A, ptr %B) nounwind uwtable ssp {
 ; CHECK-NEXT:  Src: %0 = load i64, ptr %arrayidx17, align 8 --> Dst: store i64 %0, ptr %B.addr.24, align 8
 ; CHECK-NEXT:    da analyze - confused!
 ; CHECK-NEXT:  Src: store i64 %0, ptr %B.addr.24, align 8 --> Dst: store i64 %0, ptr %B.addr.24, align 8
-; CHECK-NEXT:    da analyze - output [* * *]!
+; CHECK-NEXT:    da analyze - confused!
 ;
 entry:
   %cmp10 = icmp sgt i64 %n, 0
diff --git a/llvm/test/Analysis/DependenceAnalysis/PreliminaryNoValidityCheckFixedSize.ll b/llvm/test/Analysis/DependenceAnalysis/PreliminaryNoValidityCheckFixedSize.ll
index 4040187..e67cae7d 100644
--- a/llvm/test/Analysis/DependenceAnalysis/PreliminaryNoValidityCheckFixedSize.ll
+++ b/llvm/test/Analysis/DependenceAnalysis/PreliminaryNoValidityCheckFixedSize.ll
@@ -28,7 +28,7 @@ define void @p2(i64 %n, ptr %A, ptr %B) nounwind uwtable ssp {
 ; CHECK-NEXT:  Src: %0 = load i64, ptr %arrayidx17, align 8 --> Dst: store i64 %0, ptr %B.addr.24, align 8
 ; CHECK-NEXT:    da analyze - confused!
 ; CHECK-NEXT:  Src: store i64 %0, ptr %B.addr.24, align 8 --> Dst: store i64 %0, ptr %B.addr.24, align 8
-; CHECK-NEXT:    da analyze - output [* * *]!
+; CHECK-NEXT:    da analyze - confused!
 ;
 ; LIN-LABEL: 'p2'
 ; LIN-NEXT:  Src: store i64 %i.011, ptr %arrayidx8, align 8 --> Dst: store i64 %i.011, ptr %arrayidx8, align 8
@@ -42,7 +42,7 @@ define void @p2(i64 %n, ptr %A, ptr %B) nounwind uwtable ssp {
 ; LIN-NEXT:  Src: %0 = load i64, ptr %arrayidx17, align 8 --> Dst: store i64 %0, ptr %B.addr.24, align 8
 ; LIN-NEXT:    da analyze - confused!
 ; LIN-NEXT:  Src: store i64 %0, ptr %B.addr.24, align 8 --> Dst: store i64 %0, ptr %B.addr.24, align 8
-; LIN-NEXT:    da analyze - output [* * *]!
+; LIN-NEXT:    da analyze - confused!
 ;
 entry:
   %cmp10 = icmp sgt i64 %n, 0
diff --git a/llvm/test/Analysis/DependenceAnalysis/SymbolicRDIV.ll b/llvm/test/Analysis/DependenceAnalysis/SymbolicRDIV.ll
index f64a748..8b9aa25 100644
--- a/llvm/test/Analysis/DependenceAnalysis/SymbolicRDIV.ll
+++ b/llvm/test/Analysis/DependenceAnalysis/SymbolicRDIV.ll
@@ -437,7 +437,7 @@ define void @symbolicrdiv6(ptr %A, ptr %B, i64 %n1, i64 %n2) nounwind uwtable ss
 ; CHECK-NEXT:  Src: %0 = load i32, ptr %arrayidx4, align 4 --> Dst: store i32 %0, ptr %B.addr.12, align 4
 ; CHECK-NEXT:    da analyze - confused!
 ; CHECK-NEXT:  Src: store i32 %0, ptr %B.addr.12, align 4 --> Dst: store i32 %0, ptr %B.addr.12, align 4
-; CHECK-NEXT:    da analyze - output [* *]!
+; CHECK-NEXT:    da analyze - confused!
 ;
 entry:
   %cmp4 = icmp eq i64 %n1, 0
diff --git a/llvm/test/Analysis/LoopAccessAnalysis/runtime-check-known-true.ll b/llvm/test/Analysis/LoopAccessAnalysis/runtime-check-known-true.ll
new file mode 100644
index 0000000..30c8088
--- /dev/null
+++ b/llvm/test/Analysis/LoopAccessAnalysis/runtime-check-known-true.ll
@@ -0,0 +1,50 @@
+; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -passes='print<access-info>' -disable-output %s 2>&1 | FileCheck %s
+
+; TODO: Accesses are known completely before or after.
+define void @test_runtime_check_known_false_after_construction(ptr %start.1, ptr %start.2, ptr %end) {
+; CHECK-LABEL: 'test_runtime_check_known_false_after_construction'
+; CHECK-NEXT:    loop:
+; CHECK-NEXT:      Memory dependences are safe with run-time checks
+; CHECK-NEXT:      Dependences:
+; CHECK-NEXT:      Run-time memory checks:
+; CHECK-NEXT:      Check 0:
+; CHECK-NEXT:        Comparing group GRP0:
+; CHECK-NEXT:          %ptr.iv.1 = phi ptr [ %ptr.iv.1.next, %loop ], [ %start.1, %entry ]
+; CHECK-NEXT:        Against group GRP1:
+; CHECK-NEXT:          %ptr.iv.2 = phi ptr [ %ptr.iv.2.next, %loop ], [ %start.2.diff, %entry ]
+; CHECK-NEXT:      Grouped accesses:
+; CHECK-NEXT:        Group GRP0:
+; CHECK-NEXT:          (Low: ((-8 * ((2305843009213693951 * (8 + (-1 * (ptrtoint ptr %start.1 to i64)) + (ptrtoint ptr %end to i64))) /u 8)) + %start.1) High: (8 + %start.1))
+; CHECK-NEXT:            Member: {%start.1,+,-8}<%loop>
+; CHECK-NEXT:        Group GRP1:
+; CHECK-NEXT:          (Low: (-8 + (-8 * ((2305843009213693951 * (8 + (-1 * (ptrtoint ptr %start.1 to i64)) + (ptrtoint ptr %end to i64))) /u 8)) + (-1 * (ptrtoint ptr %start.2 to i64)) + (ptrtoint ptr %start.1 to i64) + %start.2) High: ((-1 * (ptrtoint ptr %start.2 to i64)) + (ptrtoint ptr %start.1 to i64) + %start.2))
+; CHECK-NEXT:            Member: {(-8 + (-1 * (ptrtoint ptr %start.2 to i64)) + (ptrtoint ptr %start.1 to i64) + %start.2),+,-8}<%loop>
+; CHECK-EMPTY:
+; CHECK-NEXT:      Non vectorizable stores to invariant address were not found in loop.
+; CHECK-NEXT:      SCEV assumptions:
+; CHECK-NEXT:      Equal predicate: (zext i3 ((trunc i64 (ptrtoint ptr %end to i64) to i3) + (-1 * (trunc i64 (ptrtoint ptr %start.1 to i64) to i3))) to i64) == 0
+; CHECK-EMPTY:
+; CHECK-NEXT:      Expressions re-written:
+;
+entry:
+  %gep.start.2 = getelementptr i8, ptr %start.2, i64 8
+  %start.1.int = ptrtoint ptr %start.1 to i64
+  %start.2.int = ptrtoint ptr %gep.start.2 to i64
+  %diff = sub i64 %start.1.int, %start.2.int
+  %start.2.diff = getelementptr i8, ptr %start.2, i64 %diff
+  br label %loop
+
+loop:
+  %ptr.iv.1 = phi ptr [ %ptr.iv.1.next, %loop ], [ %start.1, %entry ]
+  %ptr.iv.2 = phi ptr [ %ptr.iv.2.next, %loop ], [ %start.2.diff, %entry ]
+  %ptr.iv.2.next = getelementptr i8, ptr %ptr.iv.2, i64 -8
+  %ptr.iv.1.next = getelementptr i8, ptr %ptr.iv.1, i64 -8
+  %l = load i64, ptr %ptr.iv.2, align 8
+  store i64 %l, ptr %ptr.iv.1, align 8
+  %ec = icmp eq ptr %ptr.iv.2, %end
+  br i1 %ec, label %exit, label %loop
+
+exit:
+  ret void
+}
diff --git a/llvm/test/CMakeLists.txt b/llvm/test/CMakeLists.txt
index 3426b6f..3042b8f 100644
--- a/llvm/test/CMakeLists.txt
+++ b/llvm/test/CMakeLists.txt
@@ -19,6 +19,7 @@ llvm_canonicalize_cmake_booleans(
   LLVM_EXAMPLEIRTRANSFORMS_LINK_INTO_TOOLS
   LLVM_HAVE_TF_AOT
   LLVM_HAVE_TFLITE
+  LLVM_ENABLE_PROFCHECK
   LLVM_INLINER_MODEL_AUTOGENERATED
   LLVM_RAEVICT_MODEL_AUTOGENERATED
   LLVM_ENABLE_EXPENSIVE_CHECKS
diff --git a/llvm/test/CodeGen/AArch64/aarch64-combine-gather-lanes.mir b/llvm/test/CodeGen/AArch64/aarch64-combine-gather-lanes.mir
deleted file mode 100644
index 09eb18b..0000000
--- a/llvm/test/CodeGen/AArch64/aarch64-combine-gather-lanes.mir
+++ /dev/null
@@ -1,364 +0,0 @@
-# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 5
-# RUN: llc -run-pass=machine-combiner -mcpu=neoverse-n2 -mtriple=aarch64-none-linux-gnu -verify-machineinstrs %s -o - | FileCheck %s
-
----
-name:            split_loads_to_fpr128
-body:             |
-  bb.0.entry:
-    liveins: $x0, $x1, $x2, $x3, $x4
-
-    ; CHECK-LABEL: name: split_loads_to_fpr128
-    ; CHECK: [[COPY:%[0-9]+]]:gpr64common = COPY $x0
-    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr64common = COPY $x1
-    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gpr64common = COPY $x2
-    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:gpr64common = COPY $x3
-    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:gpr64common = COPY $x4
-    ; CHECK-NEXT: [[LD_i32:%[0-9]+]]:fpr32 = LDRSroX [[COPY]], killed [[COPY1]], 0, 1
-    ; CHECK-NEXT: [[FIRST_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG 0, killed [[LD_i32]], %subreg.ssub
-    ; CHECK-NEXT: [[LD0_1:%[0-9]+]]:fpr128 = LD1i32 [[FIRST_REG]], 1, killed [[COPY2]] 
-    ; CHECK-NEXT: [[LD1_0:%[0-9]+]]:fpr32 = LDRSui [[COPY3]], 0
-    ; CHECK-NEXT: [[SECOND_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG 0, killed [[LD1_0]], %subreg.ssub
-    ; CHECK-NEXT: [[LD1_1:%[0-9]+]]:fpr128 = LD1i32 [[SECOND_REG]], 1, killed [[COPY4]]
-    ; CHECK-NEXT: [[ZIP:%[0-9]+]]:fpr128 = ZIP1v2i64 [[LD0_1]], [[LD1_1]]
-    ; CHECK-NEXT: $q0 = COPY [[ZIP]]
-    ; CHECK-NEXT: RET_ReallyLR implicit $q0
-    %0:gpr64common = COPY $x0
-    %1:gpr64common = COPY $x1
-    %2:gpr64common = COPY $x2
-    %3:gpr64common = COPY $x3
-    %4:gpr64common = COPY $x4
-    %5:fpr32 = LDRSroX %0, killed %1, 0, 1
-    %6:fpr128 = SUBREG_TO_REG 0, killed %5, %subreg.ssub
-    %7:fpr128 = LD1i32 %6, 1, killed %2
-    %8:fpr128 = LD1i32 %7, 2, killed %3
-    %9:fpr128 = LD1i32 %8, 3, killed %4
-    $q0 = COPY %9
-    RET_ReallyLR implicit $q0
-
----
-name:            split_loads_to_fpr128_ui
-body:             |
-  bb.0.entry:
-    liveins: $x0, $x1, $x2, $x3, $x4
-
-    ; CHECK-LABEL: name: split_loads_to_fpr128_ui
-    ; CHECK: [[COPY:%[0-9]+]]:gpr64common = COPY $x0
-    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr64common = COPY $x1
-    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gpr64common = COPY $x2
-    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:gpr64common = COPY $x3
-    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:gpr64common = COPY $x4
-    ; CHECK-NEXT: [[LD_i32:%[0-9]+]]:fpr32 = LDRSui [[COPY]], 0
-    ; CHECK-NEXT: [[FIRST_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG 0, killed [[LD_i32]], %subreg.ssub
-    ; CHECK-NEXT: [[LD0_1:%[0-9]+]]:fpr128 = LD1i32 [[FIRST_REG]], 1, killed [[COPY1]] 
-    ; CHECK-NEXT: [[LD1_0:%[0-9]+]]:fpr32 = LDRSui [[COPY2]], 0
-    ; CHECK-NEXT: [[SECOND_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG 0, killed [[LD1_0]], %subreg.ssub
-    ; CHECK-NEXT: [[LD1_1:%[0-9]+]]:fpr128 = LD1i32 [[SECOND_REG]], 1, killed [[COPY3]]
-    ; CHECK-NEXT: [[ZIP:%[0-9]+]]:fpr128 = ZIP1v2i64 [[LD0_1]], [[LD1_1]]
-    ; CHECK-NEXT: $q0 = COPY [[ZIP]]
-    ; CHECK-NEXT: RET_ReallyLR implicit $q0
-    %0:gpr64common = COPY $x0
-    %1:gpr64common = COPY $x1
-    %2:gpr64common = COPY $x2
-    %3:gpr64common = COPY $x3
-    %4:gpr64common = COPY $x4
-    %5:fpr32 = LDRSui %0, 0
-    %6:fpr128 = SUBREG_TO_REG 0, killed %5, %subreg.ssub
-    %7:fpr128 = LD1i32 %6, 1, killed %1
-    %8:fpr128 = LD1i32 %7, 2, killed %2
-    %9:fpr128 = LD1i32 %8, 3, killed %3
-    $q0 = COPY %9
-    RET_ReallyLR implicit $q0
-
----
-name:            split_loads_to_fpr128_i16
-body:             |
-  bb.0.entry:
-    liveins: $x0, $x1, $x2, $x3, $x4, $x5, $x6, $x7, $x8
-
-    ; CHECK-LABEL: name: split_loads_to_fpr128_i16
-    ; CHECK: [[COPY:%[0-9]+]]:gpr64common = COPY $x0
-    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr64common = COPY $x1
-    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gpr64common = COPY $x2
-    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:gpr64common = COPY $x3
-    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:gpr64common = COPY $x4
-    ; CHECK-NEXT: [[COPY5:%[0-9]+]]:gpr64common = COPY $x5
-    ; CHECK-NEXT: [[COPY6:%[0-9]+]]:gpr64common = COPY $x6
-    ; CHECK-NEXT: [[COPY7:%[0-9]+]]:gpr64common = COPY $x7
-    ; CHECK-NEXT: [[COPY8:%[0-9]+]]:gpr64common = COPY $x8
-    ; CHECK-NEXT: [[LD_i16:%[0-9]+]]:fpr16 = LDRHroX [[COPY]], killed [[COPY1]], 0, 1
-    ; CHECK-NEXT: [[FIRST_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG 0, killed [[LD_i16]], %subreg.hsub
-    ; CHECK-NEXT: [[LD0_1:%[0-9]+]]:fpr128 = LD1i16 [[FIRST_REG]], 1, killed [[COPY2]]
-    ; CHECK-NEXT: [[LD0_2:%[0-9]+]]:fpr128 = LD1i16 [[LD0_1]], 2, killed [[COPY3]]
-    ; CHECK-NEXT: [[LD0_3:%[0-9]+]]:fpr128 = LD1i16 [[LD0_2]], 3, killed [[COPY4]]
-    ; CHECK-NEXT: [[LD1_0:%[0-9]+]]:fpr16 = LDRHui [[COPY5]], 0
-    ; CHECK-NEXT: [[SECOND_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG 0, killed [[LD1_0]], %subreg.hsub
-    ; CHECK-NEXT: [[LD1_1:%[0-9]+]]:fpr128 = LD1i16 [[SECOND_REG]], 1, killed [[COPY6]]
-    ; CHECK-NEXT: [[LD1_2:%[0-9]+]]:fpr128 = LD1i16 [[LD1_1]], 2, killed [[COPY7]]
-    ; CHECK-NEXT: [[LD1_3:%[0-9]+]]:fpr128 = LD1i16 [[LD1_2]], 3, killed [[COPY8]]
-    ; CHECK-NEXT: [[ZIP:%[0-9]+]]:fpr128 = ZIP1v2i64 [[LD0_3]], [[LD1_3]]
-    ; CHECK-NEXT: $q0 = COPY [[ZIP]]
-    ; CHECK-NEXT: RET_ReallyLR implicit $q0
-    %0:gpr64common = COPY $x0
-    %1:gpr64common = COPY $x1
-    %2:gpr64common = COPY $x2
-    %3:gpr64common = COPY $x3
-    %4:gpr64common = COPY $x4
-    %5:gpr64common = COPY $x5
-    %6:gpr64common = COPY $x6
-    %7:gpr64common = COPY $x7
-    %8:gpr64common = COPY $x8
-    %9:fpr16 = LDRHroX %0, killed %1, 0, 1
-    %10:fpr128 = SUBREG_TO_REG 0, killed %9, %subreg.hsub
-    %11:fpr128 = LD1i16 %10, 1, killed %2
-    %12:fpr128 = LD1i16 %11, 2, killed %3
-    %13:fpr128 = LD1i16 %12, 3, killed %4
-    %14:fpr128 = LD1i16 %13, 4, killed %5
-    %15:fpr128 = LD1i16 %14, 5, killed %6
-    %16:fpr128 = LD1i16 %15, 6, killed %7
-    %17:fpr128 = LD1i16 %16, 7, killed %8
-    $q0 = COPY %17
-    RET_ReallyLR implicit $q0
-
----
-name:            split_loads_to_fpr128_i16_ui
-body:             |
-  bb.0.entry:
-    liveins: $x0, $x1, $x2, $x3, $x4, $x5, $x6, $x7, $x8
-
-    ; CHECK-LABEL: name: split_loads_to_fpr128_i16_ui
-    ; CHECK: [[COPY:%[0-9]+]]:gpr64common = COPY $x0
-    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr64common = COPY $x1
-    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gpr64common = COPY $x2
-    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:gpr64common = COPY $x3
-    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:gpr64common = COPY $x4
-    ; CHECK-NEXT: [[COPY5:%[0-9]+]]:gpr64common = COPY $x5
-    ; CHECK-NEXT: [[COPY6:%[0-9]+]]:gpr64common = COPY $x6
-    ; CHECK-NEXT: [[COPY7:%[0-9]+]]:gpr64common = COPY $x7
-    ; CHECK-NEXT: [[COPY8:%[0-9]+]]:gpr64common = COPY $x8
-    ; CHECK-NEXT: [[LD_i16:%[0-9]+]]:fpr16 = LDRHui [[COPY]], 0
-    ; CHECK-NEXT: [[FIRST_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG 0, killed [[LD_i16]], %subreg.hsub
-    ; CHECK-NEXT: [[LD0_1:%[0-9]+]]:fpr128 = LD1i16 [[FIRST_REG]], 1, killed [[COPY1]]
-    ; CHECK-NEXT: [[LD0_2:%[0-9]+]]:fpr128 = LD1i16 [[LD0_1]], 2, killed [[COPY2]]
-    ; CHECK-NEXT: [[LD0_3:%[0-9]+]]:fpr128 = LD1i16 [[LD0_2]], 3, killed [[COPY3]]
-    ; CHECK-NEXT: [[LD1_0:%[0-9]+]]:fpr16 = LDRHui [[COPY4]], 0
-    ; CHECK-NEXT: [[SECOND_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG 0, killed [[LD1_0]], %subreg.hsub
-    ; CHECK-NEXT: [[LD1_1:%[0-9]+]]:fpr128 = LD1i16 [[SECOND_REG]], 1, killed [[COPY5]]
-    ; CHECK-NEXT: [[LD1_2:%[0-9]+]]:fpr128 = LD1i16 [[LD1_1]], 2, killed [[COPY6]]
-    ; CHECK-NEXT: [[LD1_3:%[0-9]+]]:fpr128 = LD1i16 [[LD1_2]], 3, killed [[COPY7]]
-    ; CHECK-NEXT: [[ZIP:%[0-9]+]]:fpr128 = ZIP1v2i64 [[LD0_3]], [[LD1_3]]
-    ; CHECK-NEXT: $q0 = COPY [[ZIP]]
-    ; CHECK-NEXT: RET_ReallyLR implicit $q0
-    %0:gpr64common = COPY $x0
-    %1:gpr64common = COPY $x1
-    %2:gpr64common = COPY $x2
-    %3:gpr64common = COPY $x3
-    %4:gpr64common = COPY $x4
-    %5:gpr64common = COPY $x5
-    %6:gpr64common = COPY $x6
-    %7:gpr64common = COPY $x7
-    %8:gpr64common = COPY $x8
-    %9:fpr16 = LDRHui %0, 0
-    %10:fpr128 = SUBREG_TO_REG 0, killed %9, %subreg.hsub
-    %11:fpr128 = LD1i16 %10, 1, killed %1
-    %12:fpr128 = LD1i16 %11, 2, killed %2
-    %13:fpr128 = LD1i16 %12, 3, killed %3
-    %14:fpr128 = LD1i16 %13, 4, killed %4
-    %15:fpr128 = LD1i16 %14, 5, killed %5
-    %16:fpr128 = LD1i16 %15, 6, killed %6
-    %17:fpr128 = LD1i16 %16, 7, killed %7
-    $q0 = COPY %17
-    RET_ReallyLR implicit $q0
-
----
-name:            split_loads_to_fpr128_i8
-body:             |
-  bb.0.entry:
-    liveins: $x0, $x1, $x2, $x3, $x4, $x5, $x6, $x7, $x8, $x9, $x10, $x11, $x12, $x13, $x14, $x15, $x16
-
-    ; CHECK-LABEL: name: split_loads_to_fpr128_i8
-    ; CHECK: [[COPY:%[0-9]+]]:gpr64common = COPY $x0
-    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr64common = COPY $x1
-    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gpr64common = COPY $x2
-    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:gpr64common = COPY $x3
-    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:gpr64common = COPY $x4
-    ; CHECK-NEXT: [[COPY5:%[0-9]+]]:gpr64common = COPY $x5
-    ; CHECK-NEXT: [[COPY6:%[0-9]+]]:gpr64common = COPY $x6
-    ; CHECK-NEXT: [[COPY7:%[0-9]+]]:gpr64common = COPY $x7
-    ; CHECK-NEXT: [[COPY8:%[0-9]+]]:gpr64common = COPY $x8
-    ; CHECK-NEXT: [[COPY9:%[0-9]+]]:gpr64common = COPY $x9
-    ; CHECK-NEXT: [[COPY10:%[0-9]+]]:gpr64common = COPY $x10
-    ; CHECK-NEXT: [[COPY11:%[0-9]+]]:gpr64common = COPY $x11
-    ; CHECK-NEXT: [[COPY12:%[0-9]+]]:gpr64common = COPY $x12
-    ; CHECK-NEXT: [[COPY13:%[0-9]+]]:gpr64common = COPY $x13
-    ; CHECK-NEXT: [[COPY14:%[0-9]+]]:gpr64common = COPY $x14
-    ; CHECK-NEXT: [[COPY15:%[0-9]+]]:gpr64common = COPY $x15
-    ; CHECK-NEXT: [[COPY16:%[0-9]+]]:gpr64common = COPY $x16
-    ; CHECK-NEXT: [[LD_i8:%[0-9]+]]:fpr8 = LDRBroX [[COPY]], killed [[COPY1]], 0, 0
-    ; CHECK-NEXT: [[FIRST_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG 0, killed [[LD_i8]], %subreg.bsub
-    ; CHECK-NEXT: [[LD0_1:%[0-9]+]]:fpr128 = LD1i8 [[FIRST_REG]], 1, killed [[COPY2]]
-    ; CHECK-NEXT: [[LD0_2:%[0-9]+]]:fpr128 = LD1i8 [[LD0_1]], 2, killed [[COPY3]]
-    ; CHECK-NEXT: [[LD0_3:%[0-9]+]]:fpr128 = LD1i8 [[LD0_2]], 3, killed [[COPY4]]
-    ; CHECK-NEXT: [[LD0_4:%[0-9]+]]:fpr128 = LD1i8 [[LD0_3]], 4, killed [[COPY5]]
-    ; CHECK-NEXT: [[LD0_5:%[0-9]+]]:fpr128 = LD1i8 [[LD0_4]], 5, killed [[COPY6]]
-    ; CHECK-NEXT: [[LD0_6:%[0-9]+]]:fpr128 = LD1i8 [[LD0_5]], 6, killed [[COPY7]]
-    ; CHECK-NEXT: [[LD0_7:%[0-9]+]]:fpr128 = LD1i8 [[LD0_6]], 7, killed [[COPY8]]
-    ; CHECK-NEXT: [[LD1_0:%[0-9]+]]:fpr8 = LDRBui [[COPY9]], 0
-    ; CHECK-NEXT: [[SECOND_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG 0, killed [[LD1_0]], %subreg.bsub
-    ; CHECK-NEXT: [[LD1_1:%[0-9]+]]:fpr128 = LD1i8 [[SECOND_REG]], 1, killed [[COPY10]]
-    ; CHECK-NEXT: [[LD1_2:%[0-9]+]]:fpr128 = LD1i8 [[LD1_1]], 2, killed [[COPY11]]
-    ; CHECK-NEXT: [[LD1_3:%[0-9]+]]:fpr128 = LD1i8 [[LD1_2]], 3, killed [[COPY12]]
-    ; CHECK-NEXT: [[LD1_4:%[0-9]+]]:fpr128 = LD1i8 [[LD1_3]], 4, killed [[COPY13]]
-    ; CHECK-NEXT: [[LD1_5:%[0-9]+]]:fpr128 = LD1i8 [[LD1_4]], 5, killed [[COPY14]]
-    ; CHECK-NEXT: [[LD1_6:%[0-9]+]]:fpr128 = LD1i8 [[LD1_5]], 6, killed [[COPY15]]
-    ; CHECK-NEXT: [[LD1_7:%[0-9]+]]:fpr128 = LD1i8 [[LD1_6]], 7, killed [[COPY16]]
-    ; CHECK-NEXT: [[ZIP:%[0-9]+]]:fpr128 = ZIP1v2i64 [[LD0_7]], [[LD1_7]]
-    ; CHECK-NEXT: $q0 = COPY [[ZIP]]
-    ; CHECK-NEXT: RET_ReallyLR implicit $q0
-    %0:gpr64common = COPY $x0
-    %1:gpr64common = COPY $x1
-    %2:gpr64common = COPY $x2
-    %3:gpr64common = COPY $x3
-    %4:gpr64common = COPY $x4
-    %5:gpr64common = COPY $x5
-    %6:gpr64common = COPY $x6
-    %7:gpr64common = COPY $x7
-    %8:gpr64common = COPY $x8
-    %9:gpr64common = COPY $x9
-    %10:gpr64common = COPY $x10
-    %11:gpr64common = COPY $x11
-    %12:gpr64common = COPY $x12
-    %13:gpr64common = COPY $x13
-    %14:gpr64common = COPY $x14
-    %15:gpr64common = COPY $x15
-    %16:gpr64common = COPY $x16
-    %17:fpr8 = LDRBroX %0, killed %1, 0, 0
-    %18:fpr128 = SUBREG_TO_REG 0, killed %17, %subreg.bsub
-    %19:fpr128 = LD1i8 %18, 1, killed %2
-    %20:fpr128 = LD1i8 %19, 2, killed %3
-    %21:fpr128 = LD1i8 %20, 3, killed %4
-    %22:fpr128 = LD1i8 %21, 4, killed %5
-    %23:fpr128 = LD1i8 %22, 5, killed %6
-    %24:fpr128 = LD1i8 %23, 6, killed %7
-    %25:fpr128 = LD1i8 %24, 7, killed %8
-    %26:fpr128 = LD1i8 %25, 8, killed %9
-    %27:fpr128 = LD1i8 %26, 9, killed %10
-    %28:fpr128 = LD1i8 %27, 10, killed %11
-    %29:fpr128 = LD1i8 %28, 11, killed %12
-    %30:fpr128 = LD1i8 %29, 12, killed %13
-    %31:fpr128 = LD1i8 %30, 13, killed %14
-    %32:fpr128 = LD1i8 %31, 14, killed %15
-    %33:fpr128 = LD1i8 %32, 15, killed %16
-    $q0 = COPY %33
-    RET_ReallyLR implicit $q0
-
----
-name:            negative_pattern_missing_lanes
-body:             |
-  bb.0.entry:
-    liveins: $x0, $x1
-
-    ; CHECK-LABEL: name: negative_pattern_missing_lanes
-    ; CHECK:      [[LD1:%.*]]:fpr128 = LDRQui $x1, 0
-    ; CHECK-NEXT: [[LD2:%.*]]:fpr128 = LD1i32 [[LD1]]
-  
-    %0:gpr64common = COPY $x0
-    %1:fpr128 = LDRQui $x1, 0
-    %2:fpr128 = LD1i32 %1, 3, %0
-    $q0 = COPY %2
-    RET_ReallyLR implicit $q0
-
----
-name:            out_of_order_lanes
-body:             |
-  bb.0.entry:
-    liveins: $x0, $x1, $x2, $x3, $x4
-
-    ; CHECK-LABEL: name: out_of_order_lanes
-    ; CHECK: [[COPY:%[0-9]+]]:gpr64common = COPY $x0
-    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr64common = COPY $x1
-    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gpr64common = COPY $x2
-    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:gpr64common = COPY $x3
-    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:gpr64common = COPY $x4
-    ; CHECK-NEXT: [[LD_i32:%[0-9]+]]:fpr32 = LDRSroX [[COPY]], killed [[COPY1]], 0, 1
-    ; CHECK-NEXT: [[FIRST_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG 0, killed [[LD_i32]], %subreg.ssub
-    ; CHECK-NEXT: [[LD0_1:%[0-9]+]]:fpr128 = LD1i32 [[FIRST_REG]], 1, killed [[COPY3]] 
-    ; CHECK-NEXT: [[LD1_0:%[0-9]+]]:fpr32 = LDRSui [[COPY2]], 0
-    ; CHECK-NEXT: [[SECOND_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG 0, killed [[LD1_0]], %subreg.ssub
-    ; CHECK-NEXT: [[LD1_1:%[0-9]+]]:fpr128 = LD1i32 [[SECOND_REG]], 1, killed [[COPY4]]
-    ; CHECK-NEXT: [[ZIP:%[0-9]+]]:fpr128 = ZIP1v2i64 [[LD0_1]], [[LD1_1]]
-    ; CHECK-NEXT: $q0 = COPY [[ZIP]]
-    ; CHECK-NEXT: RET_ReallyLR implicit $q0
-    %0:gpr64common = COPY $x0
-    %1:gpr64common = COPY $x1
-    %2:gpr64common = COPY $x2
-    %3:gpr64common = COPY $x3
-    %4:gpr64common = COPY $x4
-    %5:fpr32 = LDRSroX %0, killed %1, 0, 1
-    %6:fpr128 = SUBREG_TO_REG 0, killed %5, %subreg.ssub
-    %7:fpr128 = LD1i32 %6, 2, killed %2
-    %8:fpr128 = LD1i32 %7, 1, killed %3
-    %9:fpr128 = LD1i32 %8, 3, killed %4
-    $q0 = COPY %9
-    RET_ReallyLR implicit $q0
-
----
-name:            negative_pattern_no_subreg_to_reg
-body:             |
-  bb.0.entry:
-    liveins: $x0, $x1, $x2, $x3
-
-    ; CHECK-LABEL: name: negative_pattern_no_subreg_to_reg
-    ; CHECK: [[COPY:%[0-9]+]]:gpr64common = COPY $x0
-    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr64common = COPY $x1
-    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gpr64common = COPY $x2
-    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:gpr64common = COPY $x3
-    ; CHECK-NEXT: [[INITIAL_VEC:%[0-9]+]]:fpr128 = LDRQui [[COPY]], 0
-    ; CHECK-NEXT: [[LD_LANE_1:%[0-9]+]]:fpr128 = LD1i32 [[INITIAL_VEC]], 1, killed [[COPY1]]
-    ; CHECK-NEXT: [[LD_LANE_2:%[0-9]+]]:fpr128 = LD1i32 [[LD_LANE_1]], 2, killed [[COPY2]]
-    ; CHECK-NEXT: [[LD_LANE_3:%[0-9]+]]:fpr128 = LD1i32 [[LD_LANE_2]], 3, killed [[COPY3]]
-    ; CHECK-NEXT: $q0 = COPY [[LD_LANE_3]]
-    ; CHECK-NEXT: RET_ReallyLR implicit $q0
-    %0:gpr64common = COPY $x0
-    %1:gpr64common = COPY $x1
-    %2:gpr64common = COPY $x2
-    %3:gpr64common = COPY $x3
-    %4:fpr128 = LDRQui %0, 0            
-    %5:fpr128 = LD1i32 %4, 1, killed %1 
-    %6:fpr128 = LD1i32 %5, 2, killed %2
-    %7:fpr128 = LD1i32 %6, 3, killed %3
-    $q0 = COPY %7
-    RET_ReallyLR implicit $q0
-
----
-name:            negative_pattern_multiple_users
-body:             |
-  bb.0.entry:
-    liveins: $x0, $x1, $x2, $x3, $x4
-
-    ; CHECK-LABEL: name: negative_pattern_multiple_users
-    ; CHECK: [[COPY:%[0-9]+]]:gpr64common = COPY $x0
-    ; CHECK-NEXT: [[COPY1:%[0-9]+]]:gpr64common = COPY $x1
-    ; CHECK-NEXT: [[COPY2:%[0-9]+]]:gpr64common = COPY $x2
-    ; CHECK-NEXT: [[COPY3:%[0-9]+]]:gpr64common = COPY $x3
-    ; CHECK-NEXT: [[COPY4:%[0-9]+]]:gpr64common = COPY $x4
-    ; CHECK-NEXT: [[LD_i32:%[0-9]+]]:fpr32 = LDRSroX [[COPY]], killed [[COPY1]], 0, 1
-    ; CHECK-NEXT: [[FIRST_REG:%[0-9]+]]:fpr128 = SUBREG_TO_REG 0, killed [[LD_i32]], %subreg.ssub
-    ; CHECK-NEXT: [[LD_LANE_1:%[0-9]+]]:fpr128 = LD1i32 [[FIRST_REG]], 1, killed [[COPY2]]
-    ; CHECK-NEXT: [[LD_LANE_2:%[0-9]+]]:fpr128 = LD1i32 [[LD_LANE_1]], 2, killed [[COPY3]]
-    ; CHECK-NEXT: [[LD_LANE_3:%[0-9]+]]:fpr128 = LD1i32 [[LD_LANE_2]], 3, killed [[COPY4]]
-    ; CHECK-NEXT: $q0 = COPY [[LD_LANE_3]]
-    ; CHECK-NEXT: $q1 = COPY [[LD_LANE_2]]
-    ; CHECK-NEXT: RET_ReallyLR implicit $q0, implicit $q1
-    %0:gpr64common = COPY $x0
-    %1:gpr64common = COPY $x1
-    %2:gpr64common = COPY $x2
-    %3:gpr64common = COPY $x3
-    %4:gpr64common = COPY $x4
-    %5:fpr32 = LDRSroX %0, killed %1, 0, 1
-    %6:fpr128 = SUBREG_TO_REG 0, killed %5, %subreg.ssub
-    %7:fpr128 = LD1i32 %6, 1, killed %2
-    %8:fpr128 = LD1i32 %7, 2, killed %3
-    %9:fpr128 = LD1i32 %8, 3, killed %4
-    $q0 = COPY %9
-    $q1 = COPY %8
-    RET_ReallyLR implicit $q0, implicit $q1
diff --git a/llvm/test/CodeGen/AArch64/andcompare.ll b/llvm/test/CodeGen/AArch64/andcompare.ll
index cbacd17..0e15b94 100644
--- a/llvm/test/CodeGen/AArch64/andcompare.ll
+++ b/llvm/test/CodeGen/AArch64/andcompare.ll
@@ -1,23 +1,23 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -mtriple=aarch64-none-elf -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,SDISEL
-; RUN: llc -mtriple=aarch64-none-elf -global-isel -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,GISEL
+; RUN: llc -mtriple=aarch64-none-elf -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SD
+; RUN: llc -mtriple=aarch64-none-elf -global-isel -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-GI
 
 define i32 @and_eq_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_eq_eq:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, eq
-; SDISEL-NEXT:    cset w0, eq
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_eq_eq:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, eq
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, eq
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_eq_eq:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, eq
+; CHECK-SD-NEXT:    cset w0, eq
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_eq_eq:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, eq
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, eq
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp eq i32 %s0, %s1
   %c1 = icmp eq i32 %s2, %s3
@@ -27,21 +27,21 @@ entry:
 }
 
 define i32 @and_eq_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_eq_ne:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, eq
-; SDISEL-NEXT:    cset w0, ne
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_eq_ne:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, eq
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ne
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_eq_ne:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, eq
+; CHECK-SD-NEXT:    cset w0, ne
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_eq_ne:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, eq
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ne
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp eq i32 %s0, %s1
   %c1 = icmp ne i32 %s2, %s3
@@ -51,21 +51,21 @@ entry:
 }
 
 define i32 @and_eq_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_eq_ult:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, eq
-; SDISEL-NEXT:    cset w0, lo
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_eq_ult:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, eq
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lo
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_eq_ult:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, eq
+; CHECK-SD-NEXT:    cset w0, lo
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_eq_ult:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, eq
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lo
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp eq i32 %s0, %s1
   %c1 = icmp ult i32 %s2, %s3
@@ -75,21 +75,21 @@ entry:
 }
 
 define i32 @and_eq_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_eq_ule:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, eq
-; SDISEL-NEXT:    cset w0, ls
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_eq_ule:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, eq
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ls
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_eq_ule:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, eq
+; CHECK-SD-NEXT:    cset w0, ls
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_eq_ule:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, eq
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ls
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp eq i32 %s0, %s1
   %c1 = icmp ule i32 %s2, %s3
@@ -99,21 +99,21 @@ entry:
 }
 
 define i32 @and_eq_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_eq_ugt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, eq
-; SDISEL-NEXT:    cset w0, hi
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_eq_ugt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, eq
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hi
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_eq_ugt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, eq
+; CHECK-SD-NEXT:    cset w0, hi
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_eq_ugt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, eq
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hi
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp eq i32 %s0, %s1
   %c1 = icmp ugt i32 %s2, %s3
@@ -123,21 +123,21 @@ entry:
 }
 
 define i32 @and_eq_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_eq_uge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, eq
-; SDISEL-NEXT:    cset w0, hs
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_eq_uge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, eq
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hs
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_eq_uge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, eq
+; CHECK-SD-NEXT:    cset w0, hs
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_eq_uge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, eq
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hs
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp eq i32 %s0, %s1
   %c1 = icmp uge i32 %s2, %s3
@@ -147,21 +147,21 @@ entry:
 }
 
 define i32 @and_eq_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_eq_slt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, eq
-; SDISEL-NEXT:    cset w0, lt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_eq_slt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, eq
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_eq_slt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, eq
+; CHECK-SD-NEXT:    cset w0, lt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_eq_slt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, eq
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp eq i32 %s0, %s1
   %c1 = icmp slt i32 %s2, %s3
@@ -171,21 +171,21 @@ entry:
 }
 
 define i32 @and_eq_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_eq_sle:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, eq
-; SDISEL-NEXT:    cset w0, le
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_eq_sle:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, eq
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, le
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_eq_sle:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, eq
+; CHECK-SD-NEXT:    cset w0, le
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_eq_sle:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, eq
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, le
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp eq i32 %s0, %s1
   %c1 = icmp sle i32 %s2, %s3
@@ -195,21 +195,21 @@ entry:
 }
 
 define i32 @and_eq_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_eq_sgt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, eq
-; SDISEL-NEXT:    cset w0, gt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_eq_sgt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, eq
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, gt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_eq_sgt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, eq
+; CHECK-SD-NEXT:    cset w0, gt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_eq_sgt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, eq
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, gt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp eq i32 %s0, %s1
   %c1 = icmp sgt i32 %s2, %s3
@@ -219,21 +219,21 @@ entry:
 }
 
 define i32 @and_eq_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_eq_sge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #8, eq
-; SDISEL-NEXT:    cset w0, ge
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_eq_sge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, eq
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ge
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_eq_sge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #8, eq
+; CHECK-SD-NEXT:    cset w0, ge
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_eq_sge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, eq
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ge
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp eq i32 %s0, %s1
   %c1 = icmp sge i32 %s2, %s3
@@ -243,21 +243,21 @@ entry:
 }
 
 define i32 @and_ne_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ne_eq:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, ne
-; SDISEL-NEXT:    cset w0, eq
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ne_eq:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ne
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, eq
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ne_eq:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, ne
+; CHECK-SD-NEXT:    cset w0, eq
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ne_eq:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ne
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, eq
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ne i32 %s0, %s1
   %c1 = icmp eq i32 %s2, %s3
@@ -267,21 +267,21 @@ entry:
 }
 
 define i32 @and_ne_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ne_ne:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, ne
-; SDISEL-NEXT:    cset w0, ne
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ne_ne:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ne
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ne
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ne_ne:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, ne
+; CHECK-SD-NEXT:    cset w0, ne
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ne_ne:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ne
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ne
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ne i32 %s0, %s1
   %c1 = icmp ne i32 %s2, %s3
@@ -291,21 +291,21 @@ entry:
 }
 
 define i32 @and_ne_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ne_ult:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, ne
-; SDISEL-NEXT:    cset w0, lo
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ne_ult:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ne
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lo
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ne_ult:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, ne
+; CHECK-SD-NEXT:    cset w0, lo
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ne_ult:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ne
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lo
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ne i32 %s0, %s1
   %c1 = icmp ult i32 %s2, %s3
@@ -315,21 +315,21 @@ entry:
 }
 
 define i32 @and_ne_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ne_ule:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, ne
-; SDISEL-NEXT:    cset w0, ls
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ne_ule:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ne
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ls
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ne_ule:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, ne
+; CHECK-SD-NEXT:    cset w0, ls
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ne_ule:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ne
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ls
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ne i32 %s0, %s1
   %c1 = icmp ule i32 %s2, %s3
@@ -339,21 +339,21 @@ entry:
 }
 
 define i32 @and_ne_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ne_ugt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, ne
-; SDISEL-NEXT:    cset w0, hi
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ne_ugt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ne
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hi
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ne_ugt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, ne
+; CHECK-SD-NEXT:    cset w0, hi
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ne_ugt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ne
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hi
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ne i32 %s0, %s1
   %c1 = icmp ugt i32 %s2, %s3
@@ -363,21 +363,21 @@ entry:
 }
 
 define i32 @and_ne_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ne_uge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, ne
-; SDISEL-NEXT:    cset w0, hs
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ne_uge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ne
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hs
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ne_uge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, ne
+; CHECK-SD-NEXT:    cset w0, hs
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ne_uge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ne
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hs
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ne i32 %s0, %s1
   %c1 = icmp uge i32 %s2, %s3
@@ -387,21 +387,21 @@ entry:
 }
 
 define i32 @and_ne_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ne_slt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, ne
-; SDISEL-NEXT:    cset w0, lt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ne_slt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ne
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ne_slt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, ne
+; CHECK-SD-NEXT:    cset w0, lt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ne_slt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ne
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ne i32 %s0, %s1
   %c1 = icmp slt i32 %s2, %s3
@@ -411,21 +411,21 @@ entry:
 }
 
 define i32 @and_ne_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ne_sle:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, ne
-; SDISEL-NEXT:    cset w0, le
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ne_sle:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ne
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, le
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ne_sle:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, ne
+; CHECK-SD-NEXT:    cset w0, le
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ne_sle:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ne
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, le
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ne i32 %s0, %s1
   %c1 = icmp sle i32 %s2, %s3
@@ -435,21 +435,21 @@ entry:
 }
 
 define i32 @and_ne_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ne_sgt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, ne
-; SDISEL-NEXT:    cset w0, gt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ne_sgt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ne
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, gt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ne_sgt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, ne
+; CHECK-SD-NEXT:    cset w0, gt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ne_sgt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ne
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, gt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ne i32 %s0, %s1
   %c1 = icmp sgt i32 %s2, %s3
@@ -459,21 +459,21 @@ entry:
 }
 
 define i32 @and_ne_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ne_sge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #8, ne
-; SDISEL-NEXT:    cset w0, ge
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ne_sge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ne
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ge
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ne_sge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #8, ne
+; CHECK-SD-NEXT:    cset w0, ge
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ne_sge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ne
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ge
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ne i32 %s0, %s1
   %c1 = icmp sge i32 %s2, %s3
@@ -483,21 +483,21 @@ entry:
 }
 
 define i32 @and_ult_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ult_eq:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, lo
-; SDISEL-NEXT:    cset w0, eq
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ult_eq:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lo
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, eq
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ult_eq:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, lo
+; CHECK-SD-NEXT:    cset w0, eq
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ult_eq:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lo
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, eq
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ult i32 %s0, %s1
   %c1 = icmp eq i32 %s2, %s3
@@ -507,21 +507,21 @@ entry:
 }
 
 define i32 @and_ult_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ult_ne:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, lo
-; SDISEL-NEXT:    cset w0, ne
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ult_ne:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lo
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ne
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ult_ne:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, lo
+; CHECK-SD-NEXT:    cset w0, ne
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ult_ne:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lo
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ne
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ult i32 %s0, %s1
   %c1 = icmp ne i32 %s2, %s3
@@ -531,21 +531,21 @@ entry:
 }
 
 define i32 @and_ult_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ult_ult:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, lo
-; SDISEL-NEXT:    cset w0, lo
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ult_ult:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lo
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lo
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ult_ult:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, lo
+; CHECK-SD-NEXT:    cset w0, lo
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ult_ult:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lo
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lo
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ult i32 %s0, %s1
   %c1 = icmp ult i32 %s2, %s3
@@ -555,21 +555,21 @@ entry:
 }
 
 define i32 @and_ult_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ult_ule:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, lo
-; SDISEL-NEXT:    cset w0, ls
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ult_ule:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lo
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ls
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ult_ule:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, lo
+; CHECK-SD-NEXT:    cset w0, ls
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ult_ule:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lo
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ls
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ult i32 %s0, %s1
   %c1 = icmp ule i32 %s2, %s3
@@ -579,21 +579,21 @@ entry:
 }
 
 define i32 @and_ult_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ult_ugt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, lo
-; SDISEL-NEXT:    cset w0, hi
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ult_ugt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lo
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hi
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ult_ugt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, lo
+; CHECK-SD-NEXT:    cset w0, hi
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ult_ugt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lo
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hi
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ult i32 %s0, %s1
   %c1 = icmp ugt i32 %s2, %s3
@@ -603,21 +603,21 @@ entry:
 }
 
 define i32 @and_ult_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ult_uge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, lo
-; SDISEL-NEXT:    cset w0, hs
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ult_uge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lo
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hs
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ult_uge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, lo
+; CHECK-SD-NEXT:    cset w0, hs
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ult_uge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lo
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hs
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ult i32 %s0, %s1
   %c1 = icmp uge i32 %s2, %s3
@@ -627,21 +627,21 @@ entry:
 }
 
 define i32 @and_ult_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ult_slt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, lo
-; SDISEL-NEXT:    cset w0, lt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ult_slt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lo
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ult_slt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, lo
+; CHECK-SD-NEXT:    cset w0, lt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ult_slt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lo
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ult i32 %s0, %s1
   %c1 = icmp slt i32 %s2, %s3
@@ -651,21 +651,21 @@ entry:
 }
 
 define i32 @and_ult_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ult_sle:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, lo
-; SDISEL-NEXT:    cset w0, le
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ult_sle:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lo
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, le
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ult_sle:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, lo
+; CHECK-SD-NEXT:    cset w0, le
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ult_sle:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lo
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, le
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ult i32 %s0, %s1
   %c1 = icmp sle i32 %s2, %s3
@@ -675,21 +675,21 @@ entry:
 }
 
 define i32 @and_ult_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ult_sgt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, lo
-; SDISEL-NEXT:    cset w0, gt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ult_sgt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lo
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, gt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ult_sgt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, lo
+; CHECK-SD-NEXT:    cset w0, gt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ult_sgt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lo
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, gt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ult i32 %s0, %s1
   %c1 = icmp sgt i32 %s2, %s3
@@ -699,21 +699,21 @@ entry:
 }
 
 define i32 @and_ult_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ult_sge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #8, lo
-; SDISEL-NEXT:    cset w0, ge
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ult_sge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lo
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ge
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ult_sge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #8, lo
+; CHECK-SD-NEXT:    cset w0, ge
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ult_sge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lo
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ge
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ult i32 %s0, %s1
   %c1 = icmp sge i32 %s2, %s3
@@ -723,21 +723,21 @@ entry:
 }
 
 define i32 @and_ule_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ule_eq:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, ls
-; SDISEL-NEXT:    cset w0, eq
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ule_eq:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ls
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, eq
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ule_eq:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, ls
+; CHECK-SD-NEXT:    cset w0, eq
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ule_eq:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ls
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, eq
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ule i32 %s0, %s1
   %c1 = icmp eq i32 %s2, %s3
@@ -747,21 +747,21 @@ entry:
 }
 
 define i32 @and_ule_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ule_ne:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, ls
-; SDISEL-NEXT:    cset w0, ne
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ule_ne:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ls
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ne
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ule_ne:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, ls
+; CHECK-SD-NEXT:    cset w0, ne
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ule_ne:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ls
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ne
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ule i32 %s0, %s1
   %c1 = icmp ne i32 %s2, %s3
@@ -771,21 +771,21 @@ entry:
 }
 
 define i32 @and_ule_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ule_ult:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, ls
-; SDISEL-NEXT:    cset w0, lo
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ule_ult:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ls
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lo
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ule_ult:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, ls
+; CHECK-SD-NEXT:    cset w0, lo
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ule_ult:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ls
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lo
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ule i32 %s0, %s1
   %c1 = icmp ult i32 %s2, %s3
@@ -795,21 +795,21 @@ entry:
 }
 
 define i32 @and_ule_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ule_ule:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, ls
-; SDISEL-NEXT:    cset w0, ls
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ule_ule:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ls
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ls
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ule_ule:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, ls
+; CHECK-SD-NEXT:    cset w0, ls
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ule_ule:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ls
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ls
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ule i32 %s0, %s1
   %c1 = icmp ule i32 %s2, %s3
@@ -819,21 +819,21 @@ entry:
 }
 
 define i32 @and_ule_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ule_ugt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, ls
-; SDISEL-NEXT:    cset w0, hi
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ule_ugt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ls
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hi
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ule_ugt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, ls
+; CHECK-SD-NEXT:    cset w0, hi
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ule_ugt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ls
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hi
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ule i32 %s0, %s1
   %c1 = icmp ugt i32 %s2, %s3
@@ -843,21 +843,21 @@ entry:
 }
 
 define i32 @and_ule_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ule_uge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, ls
-; SDISEL-NEXT:    cset w0, hs
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ule_uge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ls
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hs
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ule_uge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, ls
+; CHECK-SD-NEXT:    cset w0, hs
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ule_uge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ls
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hs
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ule i32 %s0, %s1
   %c1 = icmp uge i32 %s2, %s3
@@ -867,21 +867,21 @@ entry:
 }
 
 define i32 @and_ule_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ule_slt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, ls
-; SDISEL-NEXT:    cset w0, lt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ule_slt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ls
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ule_slt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, ls
+; CHECK-SD-NEXT:    cset w0, lt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ule_slt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ls
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ule i32 %s0, %s1
   %c1 = icmp slt i32 %s2, %s3
@@ -891,21 +891,21 @@ entry:
 }
 
 define i32 @and_ule_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ule_sle:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, ls
-; SDISEL-NEXT:    cset w0, le
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ule_sle:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ls
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, le
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ule_sle:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, ls
+; CHECK-SD-NEXT:    cset w0, le
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ule_sle:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ls
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, le
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ule i32 %s0, %s1
   %c1 = icmp sle i32 %s2, %s3
@@ -915,21 +915,21 @@ entry:
 }
 
 define i32 @and_ule_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ule_sgt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, ls
-; SDISEL-NEXT:    cset w0, gt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ule_sgt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ls
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, gt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ule_sgt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, ls
+; CHECK-SD-NEXT:    cset w0, gt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ule_sgt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ls
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, gt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ule i32 %s0, %s1
   %c1 = icmp sgt i32 %s2, %s3
@@ -939,21 +939,21 @@ entry:
 }
 
 define i32 @and_ule_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ule_sge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #8, ls
-; SDISEL-NEXT:    cset w0, ge
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ule_sge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ls
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ge
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ule_sge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #8, ls
+; CHECK-SD-NEXT:    cset w0, ge
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ule_sge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ls
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ge
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ule i32 %s0, %s1
   %c1 = icmp sge i32 %s2, %s3
@@ -963,21 +963,21 @@ entry:
 }
 
 define i32 @and_ugt_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ugt_eq:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, hi
-; SDISEL-NEXT:    cset w0, eq
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ugt_eq:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hi
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, eq
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ugt_eq:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, hi
+; CHECK-SD-NEXT:    cset w0, eq
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ugt_eq:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hi
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, eq
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ugt i32 %s0, %s1
   %c1 = icmp eq i32 %s2, %s3
@@ -987,21 +987,21 @@ entry:
 }
 
 define i32 @and_ugt_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ugt_ne:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, hi
-; SDISEL-NEXT:    cset w0, ne
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ugt_ne:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hi
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ne
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ugt_ne:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, hi
+; CHECK-SD-NEXT:    cset w0, ne
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ugt_ne:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hi
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ne
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ugt i32 %s0, %s1
   %c1 = icmp ne i32 %s2, %s3
@@ -1011,21 +1011,21 @@ entry:
 }
 
 define i32 @and_ugt_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ugt_ult:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, hi
-; SDISEL-NEXT:    cset w0, lo
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ugt_ult:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hi
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lo
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ugt_ult:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, hi
+; CHECK-SD-NEXT:    cset w0, lo
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ugt_ult:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hi
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lo
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ugt i32 %s0, %s1
   %c1 = icmp ult i32 %s2, %s3
@@ -1035,21 +1035,21 @@ entry:
 }
 
 define i32 @and_ugt_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ugt_ule:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, hi
-; SDISEL-NEXT:    cset w0, ls
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ugt_ule:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hi
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ls
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ugt_ule:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, hi
+; CHECK-SD-NEXT:    cset w0, ls
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ugt_ule:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hi
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ls
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ugt i32 %s0, %s1
   %c1 = icmp ule i32 %s2, %s3
@@ -1059,21 +1059,21 @@ entry:
 }
 
 define i32 @and_ugt_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ugt_ugt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, hi
-; SDISEL-NEXT:    cset w0, hi
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ugt_ugt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hi
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hi
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ugt_ugt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, hi
+; CHECK-SD-NEXT:    cset w0, hi
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ugt_ugt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hi
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hi
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ugt i32 %s0, %s1
   %c1 = icmp ugt i32 %s2, %s3
@@ -1083,21 +1083,21 @@ entry:
 }
 
 define i32 @and_ugt_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ugt_uge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, hi
-; SDISEL-NEXT:    cset w0, hs
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ugt_uge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hi
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hs
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ugt_uge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, hi
+; CHECK-SD-NEXT:    cset w0, hs
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ugt_uge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hi
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hs
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ugt i32 %s0, %s1
   %c1 = icmp uge i32 %s2, %s3
@@ -1107,21 +1107,21 @@ entry:
 }
 
 define i32 @and_ugt_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ugt_slt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, hi
-; SDISEL-NEXT:    cset w0, lt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ugt_slt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hi
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ugt_slt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, hi
+; CHECK-SD-NEXT:    cset w0, lt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ugt_slt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hi
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ugt i32 %s0, %s1
   %c1 = icmp slt i32 %s2, %s3
@@ -1131,21 +1131,21 @@ entry:
 }
 
 define i32 @and_ugt_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ugt_sle:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, hi
-; SDISEL-NEXT:    cset w0, le
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ugt_sle:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hi
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, le
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ugt_sle:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, hi
+; CHECK-SD-NEXT:    cset w0, le
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ugt_sle:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hi
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, le
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ugt i32 %s0, %s1
   %c1 = icmp sle i32 %s2, %s3
@@ -1155,21 +1155,21 @@ entry:
 }
 
 define i32 @and_ugt_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ugt_sgt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, hi
-; SDISEL-NEXT:    cset w0, gt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ugt_sgt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hi
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, gt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ugt_sgt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, hi
+; CHECK-SD-NEXT:    cset w0, gt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ugt_sgt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hi
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, gt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ugt i32 %s0, %s1
   %c1 = icmp sgt i32 %s2, %s3
@@ -1179,21 +1179,21 @@ entry:
 }
 
 define i32 @and_ugt_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_ugt_sge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #8, hi
-; SDISEL-NEXT:    cset w0, ge
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_ugt_sge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hi
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ge
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ugt_sge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #8, hi
+; CHECK-SD-NEXT:    cset w0, ge
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_ugt_sge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hi
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ge
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ugt i32 %s0, %s1
   %c1 = icmp sge i32 %s2, %s3
@@ -1203,21 +1203,21 @@ entry:
 }
 
 define i32 @and_uge_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_uge_eq:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, hs
-; SDISEL-NEXT:    cset w0, eq
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_uge_eq:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hs
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, eq
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_uge_eq:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, hs
+; CHECK-SD-NEXT:    cset w0, eq
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_uge_eq:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hs
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, eq
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp uge i32 %s0, %s1
   %c1 = icmp eq i32 %s2, %s3
@@ -1227,21 +1227,21 @@ entry:
 }
 
 define i32 @and_uge_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_uge_ne:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, hs
-; SDISEL-NEXT:    cset w0, ne
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_uge_ne:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hs
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ne
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_uge_ne:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, hs
+; CHECK-SD-NEXT:    cset w0, ne
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_uge_ne:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hs
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ne
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp uge i32 %s0, %s1
   %c1 = icmp ne i32 %s2, %s3
@@ -1251,21 +1251,21 @@ entry:
 }
 
 define i32 @and_uge_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_uge_ult:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, hs
-; SDISEL-NEXT:    cset w0, lo
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_uge_ult:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hs
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lo
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_uge_ult:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, hs
+; CHECK-SD-NEXT:    cset w0, lo
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_uge_ult:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hs
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lo
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp uge i32 %s0, %s1
   %c1 = icmp ult i32 %s2, %s3
@@ -1275,21 +1275,21 @@ entry:
 }
 
 define i32 @and_uge_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_uge_ule:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, hs
-; SDISEL-NEXT:    cset w0, ls
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_uge_ule:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hs
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ls
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_uge_ule:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, hs
+; CHECK-SD-NEXT:    cset w0, ls
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_uge_ule:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hs
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ls
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp uge i32 %s0, %s1
   %c1 = icmp ule i32 %s2, %s3
@@ -1299,21 +1299,21 @@ entry:
 }
 
 define i32 @and_uge_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_uge_ugt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, hs
-; SDISEL-NEXT:    cset w0, hi
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_uge_ugt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hs
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hi
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_uge_ugt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, hs
+; CHECK-SD-NEXT:    cset w0, hi
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_uge_ugt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hs
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hi
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp uge i32 %s0, %s1
   %c1 = icmp ugt i32 %s2, %s3
@@ -1323,21 +1323,21 @@ entry:
 }
 
 define i32 @and_uge_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_uge_uge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, hs
-; SDISEL-NEXT:    cset w0, hs
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_uge_uge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hs
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hs
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_uge_uge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, hs
+; CHECK-SD-NEXT:    cset w0, hs
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_uge_uge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hs
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hs
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp uge i32 %s0, %s1
   %c1 = icmp uge i32 %s2, %s3
@@ -1347,21 +1347,21 @@ entry:
 }
 
 define i32 @and_uge_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_uge_slt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, hs
-; SDISEL-NEXT:    cset w0, lt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_uge_slt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hs
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_uge_slt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, hs
+; CHECK-SD-NEXT:    cset w0, lt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_uge_slt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hs
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp uge i32 %s0, %s1
   %c1 = icmp slt i32 %s2, %s3
@@ -1371,21 +1371,21 @@ entry:
 }
 
 define i32 @and_uge_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_uge_sle:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, hs
-; SDISEL-NEXT:    cset w0, le
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_uge_sle:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hs
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, le
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_uge_sle:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, hs
+; CHECK-SD-NEXT:    cset w0, le
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_uge_sle:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hs
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, le
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp uge i32 %s0, %s1
   %c1 = icmp sle i32 %s2, %s3
@@ -1395,21 +1395,21 @@ entry:
 }
 
 define i32 @and_uge_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_uge_sgt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, hs
-; SDISEL-NEXT:    cset w0, gt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_uge_sgt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hs
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, gt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_uge_sgt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, hs
+; CHECK-SD-NEXT:    cset w0, gt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_uge_sgt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hs
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, gt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp uge i32 %s0, %s1
   %c1 = icmp sgt i32 %s2, %s3
@@ -1419,21 +1419,21 @@ entry:
 }
 
 define i32 @and_uge_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_uge_sge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #8, hs
-; SDISEL-NEXT:    cset w0, ge
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_uge_sge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hs
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ge
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_uge_sge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #8, hs
+; CHECK-SD-NEXT:    cset w0, ge
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_uge_sge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hs
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ge
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp uge i32 %s0, %s1
   %c1 = icmp sge i32 %s2, %s3
@@ -1443,21 +1443,21 @@ entry:
 }
 
 define i32 @and_slt_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_slt_eq:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, lt
-; SDISEL-NEXT:    cset w0, eq
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_slt_eq:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, eq
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_slt_eq:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, lt
+; CHECK-SD-NEXT:    cset w0, eq
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_slt_eq:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, eq
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp slt i32 %s0, %s1
   %c1 = icmp eq i32 %s2, %s3
@@ -1467,21 +1467,21 @@ entry:
 }
 
 define i32 @and_slt_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_slt_ne:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, lt
-; SDISEL-NEXT:    cset w0, ne
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_slt_ne:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ne
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_slt_ne:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, lt
+; CHECK-SD-NEXT:    cset w0, ne
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_slt_ne:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ne
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp slt i32 %s0, %s1
   %c1 = icmp ne i32 %s2, %s3
@@ -1491,21 +1491,21 @@ entry:
 }
 
 define i32 @and_slt_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_slt_ult:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, lt
-; SDISEL-NEXT:    cset w0, lo
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_slt_ult:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lo
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_slt_ult:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, lt
+; CHECK-SD-NEXT:    cset w0, lo
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_slt_ult:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lo
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp slt i32 %s0, %s1
   %c1 = icmp ult i32 %s2, %s3
@@ -1515,21 +1515,21 @@ entry:
 }
 
 define i32 @and_slt_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_slt_ule:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, lt
-; SDISEL-NEXT:    cset w0, ls
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_slt_ule:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ls
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_slt_ule:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, lt
+; CHECK-SD-NEXT:    cset w0, ls
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_slt_ule:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ls
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp slt i32 %s0, %s1
   %c1 = icmp ule i32 %s2, %s3
@@ -1539,21 +1539,21 @@ entry:
 }
 
 define i32 @and_slt_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_slt_ugt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, lt
-; SDISEL-NEXT:    cset w0, hi
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_slt_ugt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hi
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_slt_ugt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, lt
+; CHECK-SD-NEXT:    cset w0, hi
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_slt_ugt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hi
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp slt i32 %s0, %s1
   %c1 = icmp ugt i32 %s2, %s3
@@ -1563,21 +1563,21 @@ entry:
 }
 
 define i32 @and_slt_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_slt_uge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, lt
-; SDISEL-NEXT:    cset w0, hs
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_slt_uge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hs
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_slt_uge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, lt
+; CHECK-SD-NEXT:    cset w0, hs
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_slt_uge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hs
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp slt i32 %s0, %s1
   %c1 = icmp uge i32 %s2, %s3
@@ -1587,21 +1587,21 @@ entry:
 }
 
 define i32 @and_slt_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_slt_slt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, lt
-; SDISEL-NEXT:    cset w0, lt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_slt_slt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_slt_slt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, lt
+; CHECK-SD-NEXT:    cset w0, lt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_slt_slt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp slt i32 %s0, %s1
   %c1 = icmp slt i32 %s2, %s3
@@ -1611,21 +1611,21 @@ entry:
 }
 
 define i32 @and_slt_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_slt_sle:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, lt
-; SDISEL-NEXT:    cset w0, le
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_slt_sle:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, le
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_slt_sle:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, lt
+; CHECK-SD-NEXT:    cset w0, le
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_slt_sle:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, le
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp slt i32 %s0, %s1
   %c1 = icmp sle i32 %s2, %s3
@@ -1635,21 +1635,21 @@ entry:
 }
 
 define i32 @and_slt_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_slt_sgt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, lt
-; SDISEL-NEXT:    cset w0, gt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_slt_sgt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, gt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_slt_sgt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, lt
+; CHECK-SD-NEXT:    cset w0, gt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_slt_sgt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, gt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp slt i32 %s0, %s1
   %c1 = icmp sgt i32 %s2, %s3
@@ -1659,21 +1659,21 @@ entry:
 }
 
 define i32 @and_slt_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_slt_sge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #8, lt
-; SDISEL-NEXT:    cset w0, ge
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_slt_sge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ge
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_slt_sge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #8, lt
+; CHECK-SD-NEXT:    cset w0, ge
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_slt_sge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ge
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp slt i32 %s0, %s1
   %c1 = icmp sge i32 %s2, %s3
@@ -1683,21 +1683,21 @@ entry:
 }
 
 define i32 @and_sle_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sle_eq:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, le
-; SDISEL-NEXT:    cset w0, eq
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sle_eq:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, le
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, eq
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sle_eq:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, le
+; CHECK-SD-NEXT:    cset w0, eq
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sle_eq:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, le
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, eq
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sle i32 %s0, %s1
   %c1 = icmp eq i32 %s2, %s3
@@ -1707,21 +1707,21 @@ entry:
 }
 
 define i32 @and_sle_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sle_ne:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, le
-; SDISEL-NEXT:    cset w0, ne
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sle_ne:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, le
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ne
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sle_ne:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, le
+; CHECK-SD-NEXT:    cset w0, ne
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sle_ne:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, le
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ne
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sle i32 %s0, %s1
   %c1 = icmp ne i32 %s2, %s3
@@ -1731,21 +1731,21 @@ entry:
 }
 
 define i32 @and_sle_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sle_ult:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, le
-; SDISEL-NEXT:    cset w0, lo
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sle_ult:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, le
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lo
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sle_ult:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, le
+; CHECK-SD-NEXT:    cset w0, lo
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sle_ult:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, le
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lo
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sle i32 %s0, %s1
   %c1 = icmp ult i32 %s2, %s3
@@ -1755,21 +1755,21 @@ entry:
 }
 
 define i32 @and_sle_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sle_ule:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, le
-; SDISEL-NEXT:    cset w0, ls
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sle_ule:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, le
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ls
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sle_ule:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, le
+; CHECK-SD-NEXT:    cset w0, ls
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sle_ule:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, le
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ls
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sle i32 %s0, %s1
   %c1 = icmp ule i32 %s2, %s3
@@ -1779,21 +1779,21 @@ entry:
 }
 
 define i32 @and_sle_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sle_ugt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, le
-; SDISEL-NEXT:    cset w0, hi
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sle_ugt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, le
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hi
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sle_ugt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, le
+; CHECK-SD-NEXT:    cset w0, hi
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sle_ugt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, le
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hi
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sle i32 %s0, %s1
   %c1 = icmp ugt i32 %s2, %s3
@@ -1803,21 +1803,21 @@ entry:
 }
 
 define i32 @and_sle_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sle_uge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, le
-; SDISEL-NEXT:    cset w0, hs
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sle_uge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, le
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hs
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sle_uge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, le
+; CHECK-SD-NEXT:    cset w0, hs
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sle_uge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, le
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hs
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sle i32 %s0, %s1
   %c1 = icmp uge i32 %s2, %s3
@@ -1827,21 +1827,21 @@ entry:
 }
 
 define i32 @and_sle_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sle_slt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, le
-; SDISEL-NEXT:    cset w0, lt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sle_slt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, le
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sle_slt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, le
+; CHECK-SD-NEXT:    cset w0, lt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sle_slt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, le
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sle i32 %s0, %s1
   %c1 = icmp slt i32 %s2, %s3
@@ -1851,21 +1851,21 @@ entry:
 }
 
 define i32 @and_sle_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sle_sle:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, le
-; SDISEL-NEXT:    cset w0, le
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sle_sle:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, le
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, le
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sle_sle:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, le
+; CHECK-SD-NEXT:    cset w0, le
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sle_sle:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, le
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, le
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sle i32 %s0, %s1
   %c1 = icmp sle i32 %s2, %s3
@@ -1875,21 +1875,21 @@ entry:
 }
 
 define i32 @and_sle_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sle_sgt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, le
-; SDISEL-NEXT:    cset w0, gt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sle_sgt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, le
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, gt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sle_sgt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, le
+; CHECK-SD-NEXT:    cset w0, gt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sle_sgt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, le
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, gt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sle i32 %s0, %s1
   %c1 = icmp sgt i32 %s2, %s3
@@ -1899,21 +1899,21 @@ entry:
 }
 
 define i32 @and_sle_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sle_sge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #8, le
-; SDISEL-NEXT:    cset w0, ge
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sle_sge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, le
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ge
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sle_sge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #8, le
+; CHECK-SD-NEXT:    cset w0, ge
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sle_sge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, le
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ge
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sle i32 %s0, %s1
   %c1 = icmp sge i32 %s2, %s3
@@ -1923,21 +1923,21 @@ entry:
 }
 
 define i32 @and_sgt_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sgt_eq:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, gt
-; SDISEL-NEXT:    cset w0, eq
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sgt_eq:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, gt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, eq
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sgt_eq:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, gt
+; CHECK-SD-NEXT:    cset w0, eq
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sgt_eq:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, gt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, eq
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sgt i32 %s0, %s1
   %c1 = icmp eq i32 %s2, %s3
@@ -1947,21 +1947,21 @@ entry:
 }
 
 define i32 @and_sgt_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sgt_ne:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, gt
-; SDISEL-NEXT:    cset w0, ne
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sgt_ne:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, gt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ne
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sgt_ne:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, gt
+; CHECK-SD-NEXT:    cset w0, ne
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sgt_ne:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, gt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ne
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sgt i32 %s0, %s1
   %c1 = icmp ne i32 %s2, %s3
@@ -1971,21 +1971,21 @@ entry:
 }
 
 define i32 @and_sgt_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sgt_ult:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, gt
-; SDISEL-NEXT:    cset w0, lo
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sgt_ult:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, gt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lo
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sgt_ult:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, gt
+; CHECK-SD-NEXT:    cset w0, lo
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sgt_ult:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, gt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lo
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sgt i32 %s0, %s1
   %c1 = icmp ult i32 %s2, %s3
@@ -1995,21 +1995,21 @@ entry:
 }
 
 define i32 @and_sgt_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sgt_ule:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, gt
-; SDISEL-NEXT:    cset w0, ls
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sgt_ule:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, gt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ls
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sgt_ule:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, gt
+; CHECK-SD-NEXT:    cset w0, ls
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sgt_ule:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, gt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ls
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sgt i32 %s0, %s1
   %c1 = icmp ule i32 %s2, %s3
@@ -2019,21 +2019,21 @@ entry:
 }
 
 define i32 @and_sgt_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sgt_ugt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, gt
-; SDISEL-NEXT:    cset w0, hi
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sgt_ugt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, gt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hi
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sgt_ugt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, gt
+; CHECK-SD-NEXT:    cset w0, hi
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sgt_ugt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, gt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hi
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sgt i32 %s0, %s1
   %c1 = icmp ugt i32 %s2, %s3
@@ -2043,21 +2043,21 @@ entry:
 }
 
 define i32 @and_sgt_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sgt_uge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, gt
-; SDISEL-NEXT:    cset w0, hs
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sgt_uge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, gt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hs
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sgt_uge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, gt
+; CHECK-SD-NEXT:    cset w0, hs
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sgt_uge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, gt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hs
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sgt i32 %s0, %s1
   %c1 = icmp uge i32 %s2, %s3
@@ -2067,21 +2067,21 @@ entry:
 }
 
 define i32 @and_sgt_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sgt_slt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, gt
-; SDISEL-NEXT:    cset w0, lt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sgt_slt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, gt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sgt_slt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, gt
+; CHECK-SD-NEXT:    cset w0, lt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sgt_slt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, gt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sgt i32 %s0, %s1
   %c1 = icmp slt i32 %s2, %s3
@@ -2091,21 +2091,21 @@ entry:
 }
 
 define i32 @and_sgt_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sgt_sle:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, gt
-; SDISEL-NEXT:    cset w0, le
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sgt_sle:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, gt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, le
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sgt_sle:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, gt
+; CHECK-SD-NEXT:    cset w0, le
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sgt_sle:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, gt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, le
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sgt i32 %s0, %s1
   %c1 = icmp sle i32 %s2, %s3
@@ -2115,21 +2115,21 @@ entry:
 }
 
 define i32 @and_sgt_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sgt_sgt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, gt
-; SDISEL-NEXT:    cset w0, gt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sgt_sgt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, gt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, gt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sgt_sgt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, gt
+; CHECK-SD-NEXT:    cset w0, gt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sgt_sgt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, gt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, gt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sgt i32 %s0, %s1
   %c1 = icmp sgt i32 %s2, %s3
@@ -2139,21 +2139,21 @@ entry:
 }
 
 define i32 @and_sgt_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sgt_sge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #8, gt
-; SDISEL-NEXT:    cset w0, ge
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sgt_sge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, gt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ge
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sgt_sge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #8, gt
+; CHECK-SD-NEXT:    cset w0, ge
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sgt_sge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, gt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ge
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sgt i32 %s0, %s1
   %c1 = icmp sge i32 %s2, %s3
@@ -2163,21 +2163,21 @@ entry:
 }
 
 define i32 @and_sge_eq(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sge_eq:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, ge
-; SDISEL-NEXT:    cset w0, eq
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sge_eq:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ge
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, eq
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sge_eq:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, ge
+; CHECK-SD-NEXT:    cset w0, eq
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sge_eq:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ge
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, eq
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sge i32 %s0, %s1
   %c1 = icmp eq i32 %s2, %s3
@@ -2187,21 +2187,21 @@ entry:
 }
 
 define i32 @and_sge_ne(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sge_ne:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, ge
-; SDISEL-NEXT:    cset w0, ne
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sge_ne:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ge
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ne
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sge_ne:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, ge
+; CHECK-SD-NEXT:    cset w0, ne
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sge_ne:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ge
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ne
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sge i32 %s0, %s1
   %c1 = icmp ne i32 %s2, %s3
@@ -2211,21 +2211,21 @@ entry:
 }
 
 define i32 @and_sge_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sge_ult:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, ge
-; SDISEL-NEXT:    cset w0, lo
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sge_ult:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ge
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lo
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sge_ult:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, ge
+; CHECK-SD-NEXT:    cset w0, lo
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sge_ult:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ge
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lo
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sge i32 %s0, %s1
   %c1 = icmp ult i32 %s2, %s3
@@ -2235,21 +2235,21 @@ entry:
 }
 
 define i32 @and_sge_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sge_ule:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, ge
-; SDISEL-NEXT:    cset w0, ls
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sge_ule:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ge
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ls
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sge_ule:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, ge
+; CHECK-SD-NEXT:    cset w0, ls
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sge_ule:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ge
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ls
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sge i32 %s0, %s1
   %c1 = icmp ule i32 %s2, %s3
@@ -2259,21 +2259,21 @@ entry:
 }
 
 define i32 @and_sge_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sge_ugt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, ge
-; SDISEL-NEXT:    cset w0, hi
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sge_ugt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ge
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hi
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sge_ugt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, ge
+; CHECK-SD-NEXT:    cset w0, hi
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sge_ugt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ge
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hi
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sge i32 %s0, %s1
   %c1 = icmp ugt i32 %s2, %s3
@@ -2283,21 +2283,21 @@ entry:
 }
 
 define i32 @and_sge_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sge_uge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, ge
-; SDISEL-NEXT:    cset w0, hs
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sge_uge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ge
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hs
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sge_uge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, ge
+; CHECK-SD-NEXT:    cset w0, hs
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sge_uge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ge
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hs
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sge i32 %s0, %s1
   %c1 = icmp uge i32 %s2, %s3
@@ -2307,21 +2307,21 @@ entry:
 }
 
 define i32 @and_sge_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sge_slt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, ge
-; SDISEL-NEXT:    cset w0, lt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sge_slt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ge
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sge_slt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, ge
+; CHECK-SD-NEXT:    cset w0, lt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sge_slt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ge
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sge i32 %s0, %s1
   %c1 = icmp slt i32 %s2, %s3
@@ -2331,21 +2331,21 @@ entry:
 }
 
 define i32 @and_sge_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sge_sle:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, ge
-; SDISEL-NEXT:    cset w0, le
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sge_sle:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ge
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, le
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sge_sle:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, ge
+; CHECK-SD-NEXT:    cset w0, le
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sge_sle:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ge
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, le
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sge i32 %s0, %s1
   %c1 = icmp sle i32 %s2, %s3
@@ -2355,21 +2355,21 @@ entry:
 }
 
 define i32 @and_sge_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sge_sgt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #4, ge
-; SDISEL-NEXT:    cset w0, gt
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sge_sgt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ge
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, gt
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sge_sgt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #4, ge
+; CHECK-SD-NEXT:    cset w0, gt
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sge_sgt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ge
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, gt
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sge i32 %s0, %s1
   %c1 = icmp sgt i32 %s2, %s3
@@ -2379,21 +2379,21 @@ entry:
 }
 
 define i32 @and_sge_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3) {
-; SDISEL-LABEL: and_sge_sge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #8, ge
-; SDISEL-NEXT:    cset w0, ge
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_sge_sge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ge
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ge
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sge_sge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #8, ge
+; CHECK-SD-NEXT:    cset w0, ge
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_sge_sge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ge
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ge
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sge i32 %s0, %s1
   %c1 = icmp sge i32 %s2, %s3
@@ -2403,19 +2403,19 @@ entry:
 }
 
 define i32 @cmp_to_ands1(i32 %num) {
-; SDISEL-LABEL: cmp_to_ands1:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    and w8, w0, #0xff
-; SDISEL-NEXT:    tst w0, #0xfe
-; SDISEL-NEXT:    csel w0, w8, wzr, ne
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: cmp_to_ands1:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    and w8, w0, #0xff
-; GISEL-NEXT:    cmp w8, #1
-; GISEL-NEXT:    csel w0, w8, wzr, hi
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: cmp_to_ands1:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    and w8, w0, #0xff
+; CHECK-SD-NEXT:    tst w0, #0xfe
+; CHECK-SD-NEXT:    csel w0, w8, wzr, ne
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: cmp_to_ands1:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    and w8, w0, #0xff
+; CHECK-GI-NEXT:    cmp w8, #1
+; CHECK-GI-NEXT:    csel w0, w8, wzr, hi
+; CHECK-GI-NEXT:    ret
   %and = and i32 %num, 255
   %cmp = icmp ugt i32 %and, 1
   %r = select i1 %cmp, i32 %and, i32 0
@@ -2423,19 +2423,19 @@ define i32 @cmp_to_ands1(i32 %num) {
 }
 
 define i32 @cmp_to_ands2(i32 %num) {
-; SDISEL-LABEL: cmp_to_ands2:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    and w8, w0, #0xfe
-; SDISEL-NEXT:    tst w0, #0xc0
-; SDISEL-NEXT:    csel w0, w8, wzr, ne
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: cmp_to_ands2:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    and w8, w0, #0xfe
-; GISEL-NEXT:    cmp w8, #63
-; GISEL-NEXT:    csel w0, w8, wzr, hi
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: cmp_to_ands2:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    and w8, w0, #0xfe
+; CHECK-SD-NEXT:    tst w0, #0xc0
+; CHECK-SD-NEXT:    csel w0, w8, wzr, ne
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: cmp_to_ands2:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    and w8, w0, #0xfe
+; CHECK-GI-NEXT:    cmp w8, #63
+; CHECK-GI-NEXT:    csel w0, w8, wzr, hi
+; CHECK-GI-NEXT:    ret
   %and = and i32 %num, 254
   %cmp = icmp ugt i32 %and, 63
   %r = select i1 %cmp, i32 %and, i32 0
@@ -2443,19 +2443,19 @@ define i32 @cmp_to_ands2(i32 %num) {
 }
 
 define i32 @cmp_to_ands3(i32 %num, i32 %a) {
-; SDISEL-LABEL: cmp_to_ands3:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    tst w0, #0x10
-; SDISEL-NEXT:    csel w0, w1, wzr, ne
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: cmp_to_ands3:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    mov w8, #23 // =0x17
-; GISEL-NEXT:    and w8, w0, w8
-; GISEL-NEXT:    cmp w8, #7
-; GISEL-NEXT:    csel w0, w1, wzr, hi
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: cmp_to_ands3:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    tst w0, #0x10
+; CHECK-SD-NEXT:    csel w0, w1, wzr, ne
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: cmp_to_ands3:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    mov w8, #23 // =0x17
+; CHECK-GI-NEXT:    and w8, w0, w8
+; CHECK-GI-NEXT:    cmp w8, #7
+; CHECK-GI-NEXT:    csel w0, w1, wzr, hi
+; CHECK-GI-NEXT:    ret
   %and = and i32 %num, 23
   %cmp = icmp ugt i32 %and, 7
   %r = select i1 %cmp, i32 %a, i32 0
@@ -2463,19 +2463,19 @@ define i32 @cmp_to_ands3(i32 %num, i32 %a) {
 }
 
 define i32 @cmp_to_ands4(i32 %num, i32 %a) {
-; SDISEL-LABEL: cmp_to_ands4:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    and w8, w0, #0x30
-; SDISEL-NEXT:    tst w0, #0x20
-; SDISEL-NEXT:    csel w0, w8, w1, eq
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: cmp_to_ands4:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    and w8, w0, #0x30
-; GISEL-NEXT:    cmp w8, #31
-; GISEL-NEXT:    csel w0, w8, w1, ls
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: cmp_to_ands4:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    and w8, w0, #0x30
+; CHECK-SD-NEXT:    tst w0, #0x20
+; CHECK-SD-NEXT:    csel w0, w8, w1, eq
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: cmp_to_ands4:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    and w8, w0, #0x30
+; CHECK-GI-NEXT:    cmp w8, #31
+; CHECK-GI-NEXT:    csel w0, w8, w1, ls
+; CHECK-GI-NEXT:    ret
   %and = and i32 %num, 48
   %cmp = icmp ule i32 %and, 31
   %r = select i1 %cmp, i32 %and, i32 %a
@@ -2483,19 +2483,19 @@ define i32 @cmp_to_ands4(i32 %num, i32 %a) {
 }
 
 define i32 @cmp_to_ands5(i32 %num, i32 %a) {
-; SDISEL-LABEL: cmp_to_ands5:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    and w8, w0, #0xf8
-; SDISEL-NEXT:    tst w0, #0xc0
-; SDISEL-NEXT:    csel w0, w8, w1, eq
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: cmp_to_ands5:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    and w8, w0, #0xf8
-; GISEL-NEXT:    cmp w8, #64
-; GISEL-NEXT:    csel w0, w8, w1, lo
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: cmp_to_ands5:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    and w8, w0, #0xf8
+; CHECK-SD-NEXT:    tst w0, #0xc0
+; CHECK-SD-NEXT:    csel w0, w8, w1, eq
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: cmp_to_ands5:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    and w8, w0, #0xf8
+; CHECK-GI-NEXT:    cmp w8, #64
+; CHECK-GI-NEXT:    csel w0, w8, w1, lo
+; CHECK-GI-NEXT:    ret
   %and = and i32 %num, 248
   %cmp = icmp ult i32 %and, 64
   %r = select i1 %cmp, i32 %and, i32 %a
@@ -2503,19 +2503,19 @@ define i32 @cmp_to_ands5(i32 %num, i32 %a) {
 }
 
 define i32 @cmp_to_ands6(i32 %num) {
-; SDISEL-LABEL: cmp_to_ands6:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    and w8, w0, #0xfe
-; SDISEL-NEXT:    tst w0, #0xf0
-; SDISEL-NEXT:    csel w0, w8, wzr, ne
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: cmp_to_ands6:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    and w8, w0, #0xfe
-; GISEL-NEXT:    cmp w8, #16
-; GISEL-NEXT:    csel w0, w8, wzr, hs
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: cmp_to_ands6:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    and w8, w0, #0xfe
+; CHECK-SD-NEXT:    tst w0, #0xf0
+; CHECK-SD-NEXT:    csel w0, w8, wzr, ne
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: cmp_to_ands6:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    and w8, w0, #0xfe
+; CHECK-GI-NEXT:    cmp w8, #16
+; CHECK-GI-NEXT:    csel w0, w8, wzr, hs
+; CHECK-GI-NEXT:    ret
   %and = and i32 %num, 254
   %cmp = icmp uge i32 %and, 16
   %r = select i1 %cmp, i32 %and, i32 0
@@ -2523,21 +2523,21 @@ define i32 @cmp_to_ands6(i32 %num) {
 }
 
 define i1 @and_fcmp(float %0, float %1) {
-; SDISEL-LABEL: and_fcmp:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    fcmp s1, s1
-; SDISEL-NEXT:    fccmp s0, s0, #0, vs
-; SDISEL-NEXT:    cset w0, vs
-; SDISEL-NEXT:    ret
-;
-; GISEL-LABEL: and_fcmp:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    fcmp s0, #0.0
-; GISEL-NEXT:    cset w8, vs
-; GISEL-NEXT:    fcmp s1, #0.0
-; GISEL-NEXT:    cset w9, vs
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_fcmp:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    fcmp s1, s1
+; CHECK-SD-NEXT:    fccmp s0, s0, #0, vs
+; CHECK-SD-NEXT:    cset w0, vs
+; CHECK-SD-NEXT:    ret
+;
+; CHECK-GI-LABEL: and_fcmp:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    fcmp s0, #0.0
+; CHECK-GI-NEXT:    cset w8, vs
+; CHECK-GI-NEXT:    fcmp s1, #0.0
+; CHECK-GI-NEXT:    cset w9, vs
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 
   %3 = fcmp uno float %0, 0.000000e+00
   %4 = fcmp uno float %1, 0.000000e+00
diff --git a/llvm/test/CodeGen/AArch64/andorbrcompare.ll b/llvm/test/CodeGen/AArch64/andorbrcompare.ll
index 951a5cd..5bc06ec 100644
--- a/llvm/test/CodeGen/AArch64/andorbrcompare.ll
+++ b/llvm/test/CodeGen/AArch64/andorbrcompare.ll
@@ -1,44 +1,44 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -mtriple=aarch64-none-elf -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,SDISEL
-; RUN: llc -mtriple=aarch64-none-elf -global-isel -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,GISEL
+; RUN: llc -mtriple=aarch64-none-elf -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-SD
+; RUN: llc -mtriple=aarch64-none-elf -global-isel -verify-machineinstrs %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-GI
 
 declare void @dummy()
 
 define i32 @and_eq_ne_ult(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %s4, i32 %s5, ptr %p) {
-; SDISEL-LABEL: and_eq_ne_ult:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w2, w3
-; SDISEL-NEXT:    ccmp w0, w1, #0, ne
-; SDISEL-NEXT:    b.eq .LBB0_3
-; SDISEL-NEXT:  // %bb.1: // %entry
-; SDISEL-NEXT:    cmp w4, w5
-; SDISEL-NEXT:    b.lo .LBB0_3
-; SDISEL-NEXT:  // %bb.2:
-; SDISEL-NEXT:    mov w0, wzr
-; SDISEL-NEXT:    ret
-; SDISEL-NEXT:  .LBB0_3: // %if
-; SDISEL-NEXT:    mov w0, #1 // =0x1
-; SDISEL-NEXT:    str w0, [x6]
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_eq_ne_ult:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w2, w3
+; CHECK-SD-NEXT:    ccmp w0, w1, #0, ne
+; CHECK-SD-NEXT:    b.eq .LBB0_3
+; CHECK-SD-NEXT:  // %bb.1: // %entry
+; CHECK-SD-NEXT:    cmp w4, w5
+; CHECK-SD-NEXT:    b.lo .LBB0_3
+; CHECK-SD-NEXT:  // %bb.2:
+; CHECK-SD-NEXT:    mov w0, wzr
+; CHECK-SD-NEXT:    ret
+; CHECK-SD-NEXT:  .LBB0_3: // %if
+; CHECK-SD-NEXT:    mov w0, #1 // =0x1
+; CHECK-SD-NEXT:    str w0, [x6]
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: and_eq_ne_ult:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, eq
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ne
-; GISEL-NEXT:    and w8, w8, w9
-; GISEL-NEXT:    tbnz w8, #0, .LBB0_3
-; GISEL-NEXT:  // %bb.1: // %entry
-; GISEL-NEXT:    cmp w4, w5
-; GISEL-NEXT:    mov w0, wzr
-; GISEL-NEXT:    b.lo .LBB0_3
-; GISEL-NEXT:  // %bb.2: // %common.ret
-; GISEL-NEXT:    ret
-; GISEL-NEXT:  .LBB0_3: // %if
-; GISEL-NEXT:    mov w0, #1 // =0x1
-; GISEL-NEXT:    str w0, [x6]
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: and_eq_ne_ult:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, eq
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ne
+; CHECK-GI-NEXT:    and w8, w8, w9
+; CHECK-GI-NEXT:    tbnz w8, #0, .LBB0_3
+; CHECK-GI-NEXT:  // %bb.1: // %entry
+; CHECK-GI-NEXT:    cmp w4, w5
+; CHECK-GI-NEXT:    mov w0, wzr
+; CHECK-GI-NEXT:    b.lo .LBB0_3
+; CHECK-GI-NEXT:  // %bb.2: // %common.ret
+; CHECK-GI-NEXT:    ret
+; CHECK-GI-NEXT:  .LBB0_3: // %if
+; CHECK-GI-NEXT:    mov w0, #1 // =0x1
+; CHECK-GI-NEXT:    str w0, [x6]
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp eq i32 %s0, %s1
   %c1 = icmp ne i32 %s2, %s3
@@ -56,40 +56,40 @@ else:
 }
 
 define i32 @and_ne_ult_ule(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %s4, i32 %s5, ptr %p) {
-; SDISEL-LABEL: and_ne_ult_ule:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w2, w3
-; SDISEL-NEXT:    ccmp w0, w1, #4, lo
-; SDISEL-NEXT:    b.ne .LBB1_3
-; SDISEL-NEXT:  // %bb.1: // %entry
-; SDISEL-NEXT:    cmp w4, w5
-; SDISEL-NEXT:    b.ls .LBB1_3
-; SDISEL-NEXT:  // %bb.2:
-; SDISEL-NEXT:    mov w0, wzr
-; SDISEL-NEXT:    ret
-; SDISEL-NEXT:  .LBB1_3: // %if
-; SDISEL-NEXT:    mov w0, #1 // =0x1
-; SDISEL-NEXT:    str w0, [x6]
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ne_ult_ule:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w2, w3
+; CHECK-SD-NEXT:    ccmp w0, w1, #4, lo
+; CHECK-SD-NEXT:    b.ne .LBB1_3
+; CHECK-SD-NEXT:  // %bb.1: // %entry
+; CHECK-SD-NEXT:    cmp w4, w5
+; CHECK-SD-NEXT:    b.ls .LBB1_3
+; CHECK-SD-NEXT:  // %bb.2:
+; CHECK-SD-NEXT:    mov w0, wzr
+; CHECK-SD-NEXT:    ret
+; CHECK-SD-NEXT:  .LBB1_3: // %if
+; CHECK-SD-NEXT:    mov w0, #1 // =0x1
+; CHECK-SD-NEXT:    str w0, [x6]
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: and_ne_ult_ule:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ne
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lo
-; GISEL-NEXT:    and w8, w8, w9
-; GISEL-NEXT:    tbnz w8, #0, .LBB1_3
-; GISEL-NEXT:  // %bb.1: // %entry
-; GISEL-NEXT:    cmp w4, w5
-; GISEL-NEXT:    mov w0, wzr
-; GISEL-NEXT:    b.ls .LBB1_3
-; GISEL-NEXT:  // %bb.2: // %common.ret
-; GISEL-NEXT:    ret
-; GISEL-NEXT:  .LBB1_3: // %if
-; GISEL-NEXT:    mov w0, #1 // =0x1
-; GISEL-NEXT:    str w0, [x6]
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: and_ne_ult_ule:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ne
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lo
+; CHECK-GI-NEXT:    and w8, w8, w9
+; CHECK-GI-NEXT:    tbnz w8, #0, .LBB1_3
+; CHECK-GI-NEXT:  // %bb.1: // %entry
+; CHECK-GI-NEXT:    cmp w4, w5
+; CHECK-GI-NEXT:    mov w0, wzr
+; CHECK-GI-NEXT:    b.ls .LBB1_3
+; CHECK-GI-NEXT:  // %bb.2: // %common.ret
+; CHECK-GI-NEXT:    ret
+; CHECK-GI-NEXT:  .LBB1_3: // %if
+; CHECK-GI-NEXT:    mov w0, #1 // =0x1
+; CHECK-GI-NEXT:    str w0, [x6]
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ne i32 %s0, %s1
   %c1 = icmp ult i32 %s2, %s3
@@ -107,40 +107,40 @@ else:
 }
 
 define i32 @and_ult_ule_ugt(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %s4, i32 %s5, ptr %p) {
-; SDISEL-LABEL: and_ult_ule_ugt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w2, w3
-; SDISEL-NEXT:    ccmp w0, w1, #2, ls
-; SDISEL-NEXT:    b.lo .LBB2_3
-; SDISEL-NEXT:  // %bb.1: // %entry
-; SDISEL-NEXT:    cmp w4, w5
-; SDISEL-NEXT:    b.hi .LBB2_3
-; SDISEL-NEXT:  // %bb.2:
-; SDISEL-NEXT:    mov w0, wzr
-; SDISEL-NEXT:    ret
-; SDISEL-NEXT:  .LBB2_3: // %if
-; SDISEL-NEXT:    mov w0, #1 // =0x1
-; SDISEL-NEXT:    str w0, [x6]
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ult_ule_ugt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w2, w3
+; CHECK-SD-NEXT:    ccmp w0, w1, #2, ls
+; CHECK-SD-NEXT:    b.lo .LBB2_3
+; CHECK-SD-NEXT:  // %bb.1: // %entry
+; CHECK-SD-NEXT:    cmp w4, w5
+; CHECK-SD-NEXT:    b.hi .LBB2_3
+; CHECK-SD-NEXT:  // %bb.2:
+; CHECK-SD-NEXT:    mov w0, wzr
+; CHECK-SD-NEXT:    ret
+; CHECK-SD-NEXT:  .LBB2_3: // %if
+; CHECK-SD-NEXT:    mov w0, #1 // =0x1
+; CHECK-SD-NEXT:    str w0, [x6]
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: and_ult_ule_ugt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lo
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ls
-; GISEL-NEXT:    and w8, w8, w9
-; GISEL-NEXT:    tbnz w8, #0, .LBB2_3
-; GISEL-NEXT:  // %bb.1: // %entry
-; GISEL-NEXT:    cmp w4, w5
-; GISEL-NEXT:    mov w0, wzr
-; GISEL-NEXT:    b.hi .LBB2_3
-; GISEL-NEXT:  // %bb.2: // %common.ret
-; GISEL-NEXT:    ret
-; GISEL-NEXT:  .LBB2_3: // %if
-; GISEL-NEXT:    mov w0, #1 // =0x1
-; GISEL-NEXT:    str w0, [x6]
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: and_ult_ule_ugt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lo
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ls
+; CHECK-GI-NEXT:    and w8, w8, w9
+; CHECK-GI-NEXT:    tbnz w8, #0, .LBB2_3
+; CHECK-GI-NEXT:  // %bb.1: // %entry
+; CHECK-GI-NEXT:    cmp w4, w5
+; CHECK-GI-NEXT:    mov w0, wzr
+; CHECK-GI-NEXT:    b.hi .LBB2_3
+; CHECK-GI-NEXT:  // %bb.2: // %common.ret
+; CHECK-GI-NEXT:    ret
+; CHECK-GI-NEXT:  .LBB2_3: // %if
+; CHECK-GI-NEXT:    mov w0, #1 // =0x1
+; CHECK-GI-NEXT:    str w0, [x6]
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ult i32 %s0, %s1
   %c1 = icmp ule i32 %s2, %s3
@@ -158,40 +158,40 @@ else:
 }
 
 define i32 @and_ule_ugt_uge(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %s4, i32 %s5, ptr %p) {
-; SDISEL-LABEL: and_ule_ugt_uge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w2, w3
-; SDISEL-NEXT:    ccmp w0, w1, #2, hi
-; SDISEL-NEXT:    b.ls .LBB3_3
-; SDISEL-NEXT:  // %bb.1: // %entry
-; SDISEL-NEXT:    cmp w4, w5
-; SDISEL-NEXT:    b.hs .LBB3_3
-; SDISEL-NEXT:  // %bb.2:
-; SDISEL-NEXT:    mov w0, wzr
-; SDISEL-NEXT:    ret
-; SDISEL-NEXT:  .LBB3_3: // %if
-; SDISEL-NEXT:    mov w0, #1 // =0x1
-; SDISEL-NEXT:    str w0, [x6]
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ule_ugt_uge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w2, w3
+; CHECK-SD-NEXT:    ccmp w0, w1, #2, hi
+; CHECK-SD-NEXT:    b.ls .LBB3_3
+; CHECK-SD-NEXT:  // %bb.1: // %entry
+; CHECK-SD-NEXT:    cmp w4, w5
+; CHECK-SD-NEXT:    b.hs .LBB3_3
+; CHECK-SD-NEXT:  // %bb.2:
+; CHECK-SD-NEXT:    mov w0, wzr
+; CHECK-SD-NEXT:    ret
+; CHECK-SD-NEXT:  .LBB3_3: // %if
+; CHECK-SD-NEXT:    mov w0, #1 // =0x1
+; CHECK-SD-NEXT:    str w0, [x6]
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: and_ule_ugt_uge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, ls
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hi
-; GISEL-NEXT:    and w8, w8, w9
-; GISEL-NEXT:    tbnz w8, #0, .LBB3_3
-; GISEL-NEXT:  // %bb.1: // %entry
-; GISEL-NEXT:    cmp w4, w5
-; GISEL-NEXT:    mov w0, wzr
-; GISEL-NEXT:    b.hs .LBB3_3
-; GISEL-NEXT:  // %bb.2: // %common.ret
-; GISEL-NEXT:    ret
-; GISEL-NEXT:  .LBB3_3: // %if
-; GISEL-NEXT:    mov w0, #1 // =0x1
-; GISEL-NEXT:    str w0, [x6]
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: and_ule_ugt_uge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, ls
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hi
+; CHECK-GI-NEXT:    and w8, w8, w9
+; CHECK-GI-NEXT:    tbnz w8, #0, .LBB3_3
+; CHECK-GI-NEXT:  // %bb.1: // %entry
+; CHECK-GI-NEXT:    cmp w4, w5
+; CHECK-GI-NEXT:    mov w0, wzr
+; CHECK-GI-NEXT:    b.hs .LBB3_3
+; CHECK-GI-NEXT:  // %bb.2: // %common.ret
+; CHECK-GI-NEXT:    ret
+; CHECK-GI-NEXT:  .LBB3_3: // %if
+; CHECK-GI-NEXT:    mov w0, #1 // =0x1
+; CHECK-GI-NEXT:    str w0, [x6]
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ule i32 %s0, %s1
   %c1 = icmp ugt i32 %s2, %s3
@@ -209,40 +209,40 @@ else:
 }
 
 define i32 @and_ugt_uge_slt(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %s4, i32 %s5, ptr %p) {
-; SDISEL-LABEL: and_ugt_uge_slt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w2, w3
-; SDISEL-NEXT:    ccmp w0, w1, #0, hs
-; SDISEL-NEXT:    b.hi .LBB4_3
-; SDISEL-NEXT:  // %bb.1: // %entry
-; SDISEL-NEXT:    cmp w4, w5
-; SDISEL-NEXT:    b.lt .LBB4_3
-; SDISEL-NEXT:  // %bb.2:
-; SDISEL-NEXT:    mov w0, wzr
-; SDISEL-NEXT:    ret
-; SDISEL-NEXT:  .LBB4_3: // %if
-; SDISEL-NEXT:    mov w0, #1 // =0x1
-; SDISEL-NEXT:    str w0, [x6]
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_ugt_uge_slt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w2, w3
+; CHECK-SD-NEXT:    ccmp w0, w1, #0, hs
+; CHECK-SD-NEXT:    b.hi .LBB4_3
+; CHECK-SD-NEXT:  // %bb.1: // %entry
+; CHECK-SD-NEXT:    cmp w4, w5
+; CHECK-SD-NEXT:    b.lt .LBB4_3
+; CHECK-SD-NEXT:  // %bb.2:
+; CHECK-SD-NEXT:    mov w0, wzr
+; CHECK-SD-NEXT:    ret
+; CHECK-SD-NEXT:  .LBB4_3: // %if
+; CHECK-SD-NEXT:    mov w0, #1 // =0x1
+; CHECK-SD-NEXT:    str w0, [x6]
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: and_ugt_uge_slt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hi
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hs
-; GISEL-NEXT:    and w8, w8, w9
-; GISEL-NEXT:    tbnz w8, #0, .LBB4_3
-; GISEL-NEXT:  // %bb.1: // %entry
-; GISEL-NEXT:    cmp w4, w5
-; GISEL-NEXT:    mov w0, wzr
-; GISEL-NEXT:    b.lt .LBB4_3
-; GISEL-NEXT:  // %bb.2: // %common.ret
-; GISEL-NEXT:    ret
-; GISEL-NEXT:  .LBB4_3: // %if
-; GISEL-NEXT:    mov w0, #1 // =0x1
-; GISEL-NEXT:    str w0, [x6]
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: and_ugt_uge_slt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hi
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hs
+; CHECK-GI-NEXT:    and w8, w8, w9
+; CHECK-GI-NEXT:    tbnz w8, #0, .LBB4_3
+; CHECK-GI-NEXT:  // %bb.1: // %entry
+; CHECK-GI-NEXT:    cmp w4, w5
+; CHECK-GI-NEXT:    mov w0, wzr
+; CHECK-GI-NEXT:    b.lt .LBB4_3
+; CHECK-GI-NEXT:  // %bb.2: // %common.ret
+; CHECK-GI-NEXT:    ret
+; CHECK-GI-NEXT:  .LBB4_3: // %if
+; CHECK-GI-NEXT:    mov w0, #1 // =0x1
+; CHECK-GI-NEXT:    str w0, [x6]
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp ugt i32 %s0, %s1
   %c1 = icmp uge i32 %s2, %s3
@@ -260,40 +260,40 @@ else:
 }
 
 define i32 @and_uge_slt_sle(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %s4, i32 %s5, ptr %p) {
-; SDISEL-LABEL: and_uge_slt_sle:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w2, w3
-; SDISEL-NEXT:    ccmp w0, w1, #0, lt
-; SDISEL-NEXT:    b.hs .LBB5_3
-; SDISEL-NEXT:  // %bb.1: // %entry
-; SDISEL-NEXT:    cmp w4, w5
-; SDISEL-NEXT:    b.le .LBB5_3
-; SDISEL-NEXT:  // %bb.2:
-; SDISEL-NEXT:    mov w0, wzr
-; SDISEL-NEXT:    ret
-; SDISEL-NEXT:  .LBB5_3: // %if
-; SDISEL-NEXT:    mov w0, #1 // =0x1
-; SDISEL-NEXT:    str w0, [x6]
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_uge_slt_sle:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w2, w3
+; CHECK-SD-NEXT:    ccmp w0, w1, #0, lt
+; CHECK-SD-NEXT:    b.hs .LBB5_3
+; CHECK-SD-NEXT:  // %bb.1: // %entry
+; CHECK-SD-NEXT:    cmp w4, w5
+; CHECK-SD-NEXT:    b.le .LBB5_3
+; CHECK-SD-NEXT:  // %bb.2:
+; CHECK-SD-NEXT:    mov w0, wzr
+; CHECK-SD-NEXT:    ret
+; CHECK-SD-NEXT:  .LBB5_3: // %if
+; CHECK-SD-NEXT:    mov w0, #1 // =0x1
+; CHECK-SD-NEXT:    str w0, [x6]
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: and_uge_slt_sle:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, hs
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, lt
-; GISEL-NEXT:    and w8, w8, w9
-; GISEL-NEXT:    tbnz w8, #0, .LBB5_3
-; GISEL-NEXT:  // %bb.1: // %entry
-; GISEL-NEXT:    cmp w4, w5
-; GISEL-NEXT:    mov w0, wzr
-; GISEL-NEXT:    b.le .LBB5_3
-; GISEL-NEXT:  // %bb.2: // %common.ret
-; GISEL-NEXT:    ret
-; GISEL-NEXT:  .LBB5_3: // %if
-; GISEL-NEXT:    mov w0, #1 // =0x1
-; GISEL-NEXT:    str w0, [x6]
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: and_uge_slt_sle:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, hs
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, lt
+; CHECK-GI-NEXT:    and w8, w8, w9
+; CHECK-GI-NEXT:    tbnz w8, #0, .LBB5_3
+; CHECK-GI-NEXT:  // %bb.1: // %entry
+; CHECK-GI-NEXT:    cmp w4, w5
+; CHECK-GI-NEXT:    mov w0, wzr
+; CHECK-GI-NEXT:    b.le .LBB5_3
+; CHECK-GI-NEXT:  // %bb.2: // %common.ret
+; CHECK-GI-NEXT:    ret
+; CHECK-GI-NEXT:  .LBB5_3: // %if
+; CHECK-GI-NEXT:    mov w0, #1 // =0x1
+; CHECK-GI-NEXT:    str w0, [x6]
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp uge i32 %s0, %s1
   %c1 = icmp slt i32 %s2, %s3
@@ -311,40 +311,40 @@ else:
 }
 
 define i32 @and_slt_sle_sgt(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %s4, i32 %s5, ptr %p) {
-; SDISEL-LABEL: and_slt_sle_sgt:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w2, w3
-; SDISEL-NEXT:    ccmp w0, w1, #0, le
-; SDISEL-NEXT:    b.lt .LBB6_3
-; SDISEL-NEXT:  // %bb.1: // %entry
-; SDISEL-NEXT:    cmp w4, w5
-; SDISEL-NEXT:    b.gt .LBB6_3
-; SDISEL-NEXT:  // %bb.2:
-; SDISEL-NEXT:    mov w0, wzr
-; SDISEL-NEXT:    ret
-; SDISEL-NEXT:  .LBB6_3: // %if
-; SDISEL-NEXT:    mov w0, #1 // =0x1
-; SDISEL-NEXT:    str w0, [x6]
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_slt_sle_sgt:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w2, w3
+; CHECK-SD-NEXT:    ccmp w0, w1, #0, le
+; CHECK-SD-NEXT:    b.lt .LBB6_3
+; CHECK-SD-NEXT:  // %bb.1: // %entry
+; CHECK-SD-NEXT:    cmp w4, w5
+; CHECK-SD-NEXT:    b.gt .LBB6_3
+; CHECK-SD-NEXT:  // %bb.2:
+; CHECK-SD-NEXT:    mov w0, wzr
+; CHECK-SD-NEXT:    ret
+; CHECK-SD-NEXT:  .LBB6_3: // %if
+; CHECK-SD-NEXT:    mov w0, #1 // =0x1
+; CHECK-SD-NEXT:    str w0, [x6]
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: and_slt_sle_sgt:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, le
-; GISEL-NEXT:    and w8, w8, w9
-; GISEL-NEXT:    tbnz w8, #0, .LBB6_3
-; GISEL-NEXT:  // %bb.1: // %entry
-; GISEL-NEXT:    cmp w4, w5
-; GISEL-NEXT:    mov w0, wzr
-; GISEL-NEXT:    b.gt .LBB6_3
-; GISEL-NEXT:  // %bb.2: // %common.ret
-; GISEL-NEXT:    ret
-; GISEL-NEXT:  .LBB6_3: // %if
-; GISEL-NEXT:    mov w0, #1 // =0x1
-; GISEL-NEXT:    str w0, [x6]
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: and_slt_sle_sgt:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, le
+; CHECK-GI-NEXT:    and w8, w8, w9
+; CHECK-GI-NEXT:    tbnz w8, #0, .LBB6_3
+; CHECK-GI-NEXT:  // %bb.1: // %entry
+; CHECK-GI-NEXT:    cmp w4, w5
+; CHECK-GI-NEXT:    mov w0, wzr
+; CHECK-GI-NEXT:    b.gt .LBB6_3
+; CHECK-GI-NEXT:  // %bb.2: // %common.ret
+; CHECK-GI-NEXT:    ret
+; CHECK-GI-NEXT:  .LBB6_3: // %if
+; CHECK-GI-NEXT:    mov w0, #1 // =0x1
+; CHECK-GI-NEXT:    str w0, [x6]
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp slt i32 %s0, %s1
   %c1 = icmp sle i32 %s2, %s3
@@ -362,40 +362,40 @@ else:
 }
 
 define i32 @and_sle_sgt_sge(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %s4, i32 %s5, ptr %p) {
-; SDISEL-LABEL: and_sle_sgt_sge:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    cmp w2, w3
-; SDISEL-NEXT:    ccmp w0, w1, #0, gt
-; SDISEL-NEXT:    b.le .LBB7_3
-; SDISEL-NEXT:  // %bb.1: // %entry
-; SDISEL-NEXT:    cmp w4, w5
-; SDISEL-NEXT:    b.ge .LBB7_3
-; SDISEL-NEXT:  // %bb.2:
-; SDISEL-NEXT:    mov w0, wzr
-; SDISEL-NEXT:    ret
-; SDISEL-NEXT:  .LBB7_3: // %if
-; SDISEL-NEXT:    mov w0, #1 // =0x1
-; SDISEL-NEXT:    str w0, [x6]
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: and_sle_sgt_sge:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    cmp w2, w3
+; CHECK-SD-NEXT:    ccmp w0, w1, #0, gt
+; CHECK-SD-NEXT:    b.le .LBB7_3
+; CHECK-SD-NEXT:  // %bb.1: // %entry
+; CHECK-SD-NEXT:    cmp w4, w5
+; CHECK-SD-NEXT:    b.ge .LBB7_3
+; CHECK-SD-NEXT:  // %bb.2:
+; CHECK-SD-NEXT:    mov w0, wzr
+; CHECK-SD-NEXT:    ret
+; CHECK-SD-NEXT:  .LBB7_3: // %if
+; CHECK-SD-NEXT:    mov w0, #1 // =0x1
+; CHECK-SD-NEXT:    str w0, [x6]
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: and_sle_sgt_sge:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, le
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, gt
-; GISEL-NEXT:    and w8, w8, w9
-; GISEL-NEXT:    tbnz w8, #0, .LBB7_3
-; GISEL-NEXT:  // %bb.1: // %entry
-; GISEL-NEXT:    cmp w4, w5
-; GISEL-NEXT:    mov w0, wzr
-; GISEL-NEXT:    b.ge .LBB7_3
-; GISEL-NEXT:  // %bb.2: // %common.ret
-; GISEL-NEXT:    ret
-; GISEL-NEXT:  .LBB7_3: // %if
-; GISEL-NEXT:    mov w0, #1 // =0x1
-; GISEL-NEXT:    str w0, [x6]
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: and_sle_sgt_sge:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, le
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, gt
+; CHECK-GI-NEXT:    and w8, w8, w9
+; CHECK-GI-NEXT:    tbnz w8, #0, .LBB7_3
+; CHECK-GI-NEXT:  // %bb.1: // %entry
+; CHECK-GI-NEXT:    cmp w4, w5
+; CHECK-GI-NEXT:    mov w0, wzr
+; CHECK-GI-NEXT:    b.ge .LBB7_3
+; CHECK-GI-NEXT:  // %bb.2: // %common.ret
+; CHECK-GI-NEXT:    ret
+; CHECK-GI-NEXT:  .LBB7_3: // %if
+; CHECK-GI-NEXT:    mov w0, #1 // =0x1
+; CHECK-GI-NEXT:    str w0, [x6]
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sle i32 %s0, %s1
   %c1 = icmp sgt i32 %s2, %s3
diff --git a/llvm/test/CodeGen/AArch64/arm64-ccmp.ll b/llvm/test/CodeGen/AArch64/arm64-ccmp.ll
index 06e957f..a546ffd 100644
--- a/llvm/test/CodeGen/AArch64/arm64-ccmp.ll
+++ b/llvm/test/CodeGen/AArch64/arm64-ccmp.ll
@@ -1,6 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -debugify-and-strip-all-safe -mcpu=cyclone -verify-machineinstrs -aarch64-enable-ccmp -aarch64-stress-ccmp | FileCheck %s --check-prefixes=CHECK,SDISEL
-; RUN: llc < %s -debugify-and-strip-all-safe -mcpu=cyclone -verify-machineinstrs -aarch64-enable-ccmp -aarch64-stress-ccmp -global-isel | FileCheck %s --check-prefixes=CHECK,GISEL
+; RUN: llc < %s -debugify-and-strip-all-safe -mcpu=cyclone -verify-machineinstrs -aarch64-enable-ccmp -aarch64-stress-ccmp | FileCheck %s --check-prefixes=CHECK,CHECK-SD
+; RUN: llc < %s -debugify-and-strip-all-safe -mcpu=cyclone -verify-machineinstrs -aarch64-enable-ccmp -aarch64-stress-ccmp -global-isel | FileCheck %s --check-prefixes=CHECK,CHECK-GI
 target triple = "arm64-apple-ios"
 
 define i32 @single_same(i32 %a, i32 %b) nounwind ssp {
@@ -32,31 +32,31 @@ if.end:
 
 ; Different condition codes for the two compares.
 define i32 @single_different(i32 %a, i32 %b) nounwind ssp {
-; SDISEL-LABEL: single_different:
-; SDISEL:       ; %bb.0: ; %entry
-; SDISEL-NEXT:    cmp w0, #6
-; SDISEL-NEXT:    ccmp w1, #17, #0, ge
-; SDISEL-NEXT:    b.eq LBB1_2
-; SDISEL-NEXT:  ; %bb.1: ; %if.then
-; SDISEL-NEXT:    stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill
-; SDISEL-NEXT:    bl _foo
-; SDISEL-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
-; SDISEL-NEXT:  LBB1_2: ; %if.end
-; SDISEL-NEXT:    mov w0, #7 ; =0x7
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: single_different:
+; CHECK-SD:       ; %bb.0: ; %entry
+; CHECK-SD-NEXT:    cmp w0, #6
+; CHECK-SD-NEXT:    ccmp w1, #17, #0, ge
+; CHECK-SD-NEXT:    b.eq LBB1_2
+; CHECK-SD-NEXT:  ; %bb.1: ; %if.then
+; CHECK-SD-NEXT:    stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill
+; CHECK-SD-NEXT:    bl _foo
+; CHECK-SD-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
+; CHECK-SD-NEXT:  LBB1_2: ; %if.end
+; CHECK-SD-NEXT:    mov w0, #7 ; =0x7
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: single_different:
-; GISEL:       ; %bb.0: ; %entry
-; GISEL-NEXT:    cmp w0, #5
-; GISEL-NEXT:    ccmp w1, #17, #0, gt
-; GISEL-NEXT:    b.eq LBB1_2
-; GISEL-NEXT:  ; %bb.1: ; %if.then
-; GISEL-NEXT:    stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill
-; GISEL-NEXT:    bl _foo
-; GISEL-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
-; GISEL-NEXT:  LBB1_2: ; %if.end
-; GISEL-NEXT:    mov w0, #7 ; =0x7
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: single_different:
+; CHECK-GI:       ; %bb.0: ; %entry
+; CHECK-GI-NEXT:    cmp w0, #5
+; CHECK-GI-NEXT:    ccmp w1, #17, #0, gt
+; CHECK-GI-NEXT:    b.eq LBB1_2
+; CHECK-GI-NEXT:  ; %bb.1: ; %if.then
+; CHECK-GI-NEXT:    stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill
+; CHECK-GI-NEXT:    bl _foo
+; CHECK-GI-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
+; CHECK-GI-NEXT:  LBB1_2: ; %if.end
+; CHECK-GI-NEXT:    mov w0, #7 ; =0x7
+; CHECK-GI-NEXT:    ret
 entry:
   %cmp = icmp sle i32 %a, 5
   %cmp1 = icmp ne i32 %b, 17
@@ -73,41 +73,41 @@ if.end:
 
 ; Second block clobbers the flags, can't convert (easily).
 define i32 @single_flagclobber(i32 %a, i32 %b) nounwind ssp {
-; SDISEL-LABEL: single_flagclobber:
-; SDISEL:       ; %bb.0: ; %entry
-; SDISEL-NEXT:    cmp w0, #5
-; SDISEL-NEXT:    b.eq LBB2_2
-; SDISEL-NEXT:  ; %bb.1: ; %lor.lhs.false
-; SDISEL-NEXT:    lsl w8, w1, #1
-; SDISEL-NEXT:    cmp w1, #7
-; SDISEL-NEXT:    csinc w8, w8, w1, lt
-; SDISEL-NEXT:    cmp w8, #16
-; SDISEL-NEXT:    b.gt LBB2_3
-; SDISEL-NEXT:  LBB2_2: ; %if.then
-; SDISEL-NEXT:    stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill
-; SDISEL-NEXT:    bl _foo
-; SDISEL-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
-; SDISEL-NEXT:  LBB2_3: ; %if.end
-; SDISEL-NEXT:    mov w0, #7 ; =0x7
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: single_flagclobber:
+; CHECK-SD:       ; %bb.0: ; %entry
+; CHECK-SD-NEXT:    cmp w0, #5
+; CHECK-SD-NEXT:    b.eq LBB2_2
+; CHECK-SD-NEXT:  ; %bb.1: ; %lor.lhs.false
+; CHECK-SD-NEXT:    lsl w8, w1, #1
+; CHECK-SD-NEXT:    cmp w1, #7
+; CHECK-SD-NEXT:    csinc w8, w8, w1, lt
+; CHECK-SD-NEXT:    cmp w8, #16
+; CHECK-SD-NEXT:    b.gt LBB2_3
+; CHECK-SD-NEXT:  LBB2_2: ; %if.then
+; CHECK-SD-NEXT:    stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill
+; CHECK-SD-NEXT:    bl _foo
+; CHECK-SD-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
+; CHECK-SD-NEXT:  LBB2_3: ; %if.end
+; CHECK-SD-NEXT:    mov w0, #7 ; =0x7
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: single_flagclobber:
-; GISEL:       ; %bb.0: ; %entry
-; GISEL-NEXT:    cmp w0, #5
-; GISEL-NEXT:    b.eq LBB2_2
-; GISEL-NEXT:  ; %bb.1: ; %lor.lhs.false
-; GISEL-NEXT:    lsl w8, w1, #1
-; GISEL-NEXT:    cmp w1, #7
-; GISEL-NEXT:    csinc w8, w8, w1, lt
-; GISEL-NEXT:    cmp w8, #17
-; GISEL-NEXT:    b.ge LBB2_3
-; GISEL-NEXT:  LBB2_2: ; %if.then
-; GISEL-NEXT:    stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill
-; GISEL-NEXT:    bl _foo
-; GISEL-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
-; GISEL-NEXT:  LBB2_3: ; %if.end
-; GISEL-NEXT:    mov w0, #7 ; =0x7
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: single_flagclobber:
+; CHECK-GI:       ; %bb.0: ; %entry
+; CHECK-GI-NEXT:    cmp w0, #5
+; CHECK-GI-NEXT:    b.eq LBB2_2
+; CHECK-GI-NEXT:  ; %bb.1: ; %lor.lhs.false
+; CHECK-GI-NEXT:    lsl w8, w1, #1
+; CHECK-GI-NEXT:    cmp w1, #7
+; CHECK-GI-NEXT:    csinc w8, w8, w1, lt
+; CHECK-GI-NEXT:    cmp w8, #17
+; CHECK-GI-NEXT:    b.ge LBB2_3
+; CHECK-GI-NEXT:  LBB2_2: ; %if.then
+; CHECK-GI-NEXT:    stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill
+; CHECK-GI-NEXT:    bl _foo
+; CHECK-GI-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
+; CHECK-GI-NEXT:  LBB2_3: ; %if.end
+; CHECK-GI-NEXT:    mov w0, #7 ; =0x7
+; CHECK-GI-NEXT:    ret
 entry:
   %cmp = icmp eq i32 %a, 5
   br i1 %cmp, label %if.then, label %lor.lhs.false
@@ -171,37 +171,37 @@ if.end:                                           ; preds = %if.then, %lor.lhs.f
 ; The sdiv/udiv instructions do not trap when the divisor is zero, so they are
 ; safe to speculate.
 define i32 @speculate_division(i32 %a, i32 %b) nounwind ssp {
-; SDISEL-LABEL: speculate_division:
-; SDISEL:       ; %bb.0: ; %entry
-; SDISEL-NEXT:    cmp w0, #1
-; SDISEL-NEXT:    sdiv w8, w1, w0
-; SDISEL-NEXT:    ccmp w8, #16, #0, ge
-; SDISEL-NEXT:    b.le LBB4_2
-; SDISEL-NEXT:  ; %bb.1: ; %if.end
-; SDISEL-NEXT:    mov w0, #7 ; =0x7
-; SDISEL-NEXT:    ret
-; SDISEL-NEXT:  LBB4_2: ; %if.then
-; SDISEL-NEXT:    stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill
-; SDISEL-NEXT:    bl _foo
-; SDISEL-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
-; SDISEL-NEXT:    mov w0, #7 ; =0x7
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: speculate_division:
+; CHECK-SD:       ; %bb.0: ; %entry
+; CHECK-SD-NEXT:    cmp w0, #1
+; CHECK-SD-NEXT:    sdiv w8, w1, w0
+; CHECK-SD-NEXT:    ccmp w8, #16, #0, ge
+; CHECK-SD-NEXT:    b.le LBB4_2
+; CHECK-SD-NEXT:  ; %bb.1: ; %if.end
+; CHECK-SD-NEXT:    mov w0, #7 ; =0x7
+; CHECK-SD-NEXT:    ret
+; CHECK-SD-NEXT:  LBB4_2: ; %if.then
+; CHECK-SD-NEXT:    stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill
+; CHECK-SD-NEXT:    bl _foo
+; CHECK-SD-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
+; CHECK-SD-NEXT:    mov w0, #7 ; =0x7
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: speculate_division:
-; GISEL:       ; %bb.0: ; %entry
-; GISEL-NEXT:    cmp w0, #0
-; GISEL-NEXT:    sdiv w8, w1, w0
-; GISEL-NEXT:    ccmp w8, #17, #0, gt
-; GISEL-NEXT:    b.lt LBB4_2
-; GISEL-NEXT:  ; %bb.1: ; %if.end
-; GISEL-NEXT:    mov w0, #7 ; =0x7
-; GISEL-NEXT:    ret
-; GISEL-NEXT:  LBB4_2: ; %if.then
-; GISEL-NEXT:    stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill
-; GISEL-NEXT:    bl _foo
-; GISEL-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
-; GISEL-NEXT:    mov w0, #7 ; =0x7
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: speculate_division:
+; CHECK-GI:       ; %bb.0: ; %entry
+; CHECK-GI-NEXT:    cmp w0, #0
+; CHECK-GI-NEXT:    sdiv w8, w1, w0
+; CHECK-GI-NEXT:    ccmp w8, #17, #0, gt
+; CHECK-GI-NEXT:    b.lt LBB4_2
+; CHECK-GI-NEXT:  ; %bb.1: ; %if.end
+; CHECK-GI-NEXT:    mov w0, #7 ; =0x7
+; CHECK-GI-NEXT:    ret
+; CHECK-GI-NEXT:  LBB4_2: ; %if.then
+; CHECK-GI-NEXT:    stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill
+; CHECK-GI-NEXT:    bl _foo
+; CHECK-GI-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
+; CHECK-GI-NEXT:    mov w0, #7 ; =0x7
+; CHECK-GI-NEXT:    ret
 entry:
   %cmp = icmp sgt i32 %a, 0
   br i1 %cmp, label %land.lhs.true, label %if.end
@@ -221,41 +221,41 @@ if.end:
 
 ; Floating point compare.
 define i32 @single_fcmp(i32 %a, float %b) nounwind ssp {
-; SDISEL-LABEL: single_fcmp:
-; SDISEL:       ; %bb.0: ; %entry
-; SDISEL-NEXT:    cmp w0, #1
-; SDISEL-NEXT:    scvtf s1, w0
-; SDISEL-NEXT:    fdiv s0, s0, s1
-; SDISEL-NEXT:    fmov s1, #17.00000000
-; SDISEL-NEXT:    fccmp s0, s1, #8, ge
-; SDISEL-NEXT:    b.ge LBB5_2
-; SDISEL-NEXT:  ; %bb.1: ; %if.end
-; SDISEL-NEXT:    mov w0, #7 ; =0x7
-; SDISEL-NEXT:    ret
-; SDISEL-NEXT:  LBB5_2: ; %if.then
-; SDISEL-NEXT:    stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill
-; SDISEL-NEXT:    bl _foo
-; SDISEL-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
-; SDISEL-NEXT:    mov w0, #7 ; =0x7
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: single_fcmp:
+; CHECK-SD:       ; %bb.0: ; %entry
+; CHECK-SD-NEXT:    cmp w0, #1
+; CHECK-SD-NEXT:    scvtf s1, w0
+; CHECK-SD-NEXT:    fdiv s0, s0, s1
+; CHECK-SD-NEXT:    fmov s1, #17.00000000
+; CHECK-SD-NEXT:    fccmp s0, s1, #8, ge
+; CHECK-SD-NEXT:    b.ge LBB5_2
+; CHECK-SD-NEXT:  ; %bb.1: ; %if.end
+; CHECK-SD-NEXT:    mov w0, #7 ; =0x7
+; CHECK-SD-NEXT:    ret
+; CHECK-SD-NEXT:  LBB5_2: ; %if.then
+; CHECK-SD-NEXT:    stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill
+; CHECK-SD-NEXT:    bl _foo
+; CHECK-SD-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
+; CHECK-SD-NEXT:    mov w0, #7 ; =0x7
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: single_fcmp:
-; GISEL:       ; %bb.0: ; %entry
-; GISEL-NEXT:    cmp w0, #0
-; GISEL-NEXT:    scvtf s1, w0
-; GISEL-NEXT:    fdiv s0, s0, s1
-; GISEL-NEXT:    fmov s1, #17.00000000
-; GISEL-NEXT:    fccmp s0, s1, #8, gt
-; GISEL-NEXT:    b.ge LBB5_2
-; GISEL-NEXT:  ; %bb.1: ; %if.end
-; GISEL-NEXT:    mov w0, #7 ; =0x7
-; GISEL-NEXT:    ret
-; GISEL-NEXT:  LBB5_2: ; %if.then
-; GISEL-NEXT:    stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill
-; GISEL-NEXT:    bl _foo
-; GISEL-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
-; GISEL-NEXT:    mov w0, #7 ; =0x7
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: single_fcmp:
+; CHECK-GI:       ; %bb.0: ; %entry
+; CHECK-GI-NEXT:    cmp w0, #0
+; CHECK-GI-NEXT:    scvtf s1, w0
+; CHECK-GI-NEXT:    fdiv s0, s0, s1
+; CHECK-GI-NEXT:    fmov s1, #17.00000000
+; CHECK-GI-NEXT:    fccmp s0, s1, #8, gt
+; CHECK-GI-NEXT:    b.ge LBB5_2
+; CHECK-GI-NEXT:  ; %bb.1: ; %if.end
+; CHECK-GI-NEXT:    mov w0, #7 ; =0x7
+; CHECK-GI-NEXT:    ret
+; CHECK-GI-NEXT:  LBB5_2: ; %if.then
+; CHECK-GI-NEXT:    stp x29, x30, [sp, #-16]! ; 16-byte Folded Spill
+; CHECK-GI-NEXT:    bl _foo
+; CHECK-GI-NEXT:    ldp x29, x30, [sp], #16 ; 16-byte Folded Reload
+; CHECK-GI-NEXT:    mov w0, #7 ; =0x7
+; CHECK-GI-NEXT:    ret
 entry:
   %cmp = icmp sgt i32 %a, 0
   br i1 %cmp, label %land.lhs.true, label %if.end
@@ -499,28 +499,28 @@ define float @select_or_float(i32 %w0, i32 %w1, float %x2, float %x3) {
 }
 
 define i64 @gccbug(i64 %x0, i64 %x1) {
-; SDISEL-LABEL: gccbug:
-; SDISEL:       ; %bb.0:
-; SDISEL-NEXT:    cmp x0, #2
-; SDISEL-NEXT:    ccmp x0, #4, #4, ne
-; SDISEL-NEXT:    ccmp x1, #0, #0, eq
-; SDISEL-NEXT:    mov w8, #1 ; =0x1
-; SDISEL-NEXT:    cinc x0, x8, eq
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: gccbug:
+; CHECK-SD:       ; %bb.0:
+; CHECK-SD-NEXT:    cmp x0, #2
+; CHECK-SD-NEXT:    ccmp x0, #4, #4, ne
+; CHECK-SD-NEXT:    ccmp x1, #0, #0, eq
+; CHECK-SD-NEXT:    mov w8, #1 ; =0x1
+; CHECK-SD-NEXT:    cinc x0, x8, eq
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: gccbug:
-; GISEL:       ; %bb.0:
-; GISEL-NEXT:    cmp x1, #0
-; GISEL-NEXT:    cset w8, eq
-; GISEL-NEXT:    cmp x0, #2
-; GISEL-NEXT:    cset w9, eq
-; GISEL-NEXT:    cmp x0, #4
-; GISEL-NEXT:    cset w10, eq
-; GISEL-NEXT:    orr w9, w10, w9
-; GISEL-NEXT:    and w8, w9, w8
-; GISEL-NEXT:    and x8, x8, #0x1
-; GISEL-NEXT:    add x0, x8, #1
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: gccbug:
+; CHECK-GI:       ; %bb.0:
+; CHECK-GI-NEXT:    cmp x1, #0
+; CHECK-GI-NEXT:    cset w8, eq
+; CHECK-GI-NEXT:    cmp x0, #2
+; CHECK-GI-NEXT:    cset w9, eq
+; CHECK-GI-NEXT:    cmp x0, #4
+; CHECK-GI-NEXT:    cset w10, eq
+; CHECK-GI-NEXT:    orr w9, w10, w9
+; CHECK-GI-NEXT:    and w8, w9, w8
+; CHECK-GI-NEXT:    and x8, x8, #0x1
+; CHECK-GI-NEXT:    add x0, x8, #1
+; CHECK-GI-NEXT:    ret
   %cmp0 = icmp eq i64 %x1, 0
   %cmp1 = icmp eq i64 %x0, 2
   %cmp2 = icmp eq i64 %x0, 4
@@ -570,23 +570,23 @@ define i32 @select_andor(i32 %v1, i32 %v2, i32 %v3) {
 }
 
 define i32 @select_andor32(i32 %v1, i32 %v2, i32 %v3) {
-; SDISEL-LABEL: select_andor32:
-; SDISEL:       ; %bb.0:
-; SDISEL-NEXT:    cmp w1, w2
-; SDISEL-NEXT:    mov w8, #32 ; =0x20
-; SDISEL-NEXT:    ccmp w0, w8, #4, lt
-; SDISEL-NEXT:    ccmp w0, w1, #0, eq
-; SDISEL-NEXT:    csel w0, w0, w1, eq
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: select_andor32:
+; CHECK-SD:       ; %bb.0:
+; CHECK-SD-NEXT:    cmp w1, w2
+; CHECK-SD-NEXT:    mov w8, #32 ; =0x20
+; CHECK-SD-NEXT:    ccmp w0, w8, #4, lt
+; CHECK-SD-NEXT:    ccmp w0, w1, #0, eq
+; CHECK-SD-NEXT:    csel w0, w0, w1, eq
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: select_andor32:
-; GISEL:       ; %bb.0:
-; GISEL-NEXT:    mov w8, #32 ; =0x20
-; GISEL-NEXT:    cmp w1, w2
-; GISEL-NEXT:    ccmp w0, w8, #4, lt
-; GISEL-NEXT:    ccmp w0, w1, #0, eq
-; GISEL-NEXT:    csel w0, w0, w1, eq
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: select_andor32:
+; CHECK-GI:       ; %bb.0:
+; CHECK-GI-NEXT:    mov w8, #32 ; =0x20
+; CHECK-GI-NEXT:    cmp w1, w2
+; CHECK-GI-NEXT:    ccmp w0, w8, #4, lt
+; CHECK-GI-NEXT:    ccmp w0, w1, #0, eq
+; CHECK-GI-NEXT:    csel w0, w0, w1, eq
+; CHECK-GI-NEXT:    ret
   %c0 = icmp eq i32 %v1, %v2
   %c1 = icmp sge i32 %v2, %v3
   %c2 = icmp eq i32 %v1, 32
@@ -597,22 +597,22 @@ define i32 @select_andor32(i32 %v1, i32 %v2, i32 %v3) {
 }
 
 define i64 @select_noccmp1(i64 %v1, i64 %v2, i64 %v3, i64 %r) {
-; SDISEL-LABEL: select_noccmp1:
-; SDISEL:       ; %bb.0:
-; SDISEL-NEXT:    cmp x0, #0
-; SDISEL-NEXT:    ccmp x0, #13, #4, lt
-; SDISEL-NEXT:    cset w8, gt
-; SDISEL-NEXT:    cmp x2, #2
-; SDISEL-NEXT:    ccmp x2, #4, #4, lt
-; SDISEL-NEXT:    csinc w8, w8, wzr, le
-; SDISEL-NEXT:    cmp w8, #0
-; SDISEL-NEXT:    csel x0, xzr, x3, ne
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: select_noccmp1:
+; CHECK-SD:       ; %bb.0:
+; CHECK-SD-NEXT:    cmp x0, #0
+; CHECK-SD-NEXT:    ccmp x0, #13, #4, lt
+; CHECK-SD-NEXT:    cset w8, gt
+; CHECK-SD-NEXT:    cmp x2, #2
+; CHECK-SD-NEXT:    ccmp x2, #4, #4, lt
+; CHECK-SD-NEXT:    csinc w8, w8, wzr, le
+; CHECK-SD-NEXT:    cmp w8, #0
+; CHECK-SD-NEXT:    csel x0, xzr, x3, ne
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: select_noccmp1:
-; GISEL:       ; %bb.0:
-; GISEL-NEXT:    mov x0, x3
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: select_noccmp1:
+; CHECK-GI:       ; %bb.0:
+; CHECK-GI-NEXT:    mov x0, x3
+; CHECK-GI-NEXT:    ret
   %c0 = icmp slt i64 %v1, 0
   %c1 = icmp sgt i64 %v1, 13
   %c2 = icmp slt i64 %v3, 2
@@ -627,28 +627,28 @@ define i64 @select_noccmp1(i64 %v1, i64 %v2, i64 %v3, i64 %r) {
 @g = global i32 0
 
 define i64 @select_noccmp2(i64 %v1, i64 %v2, i64 %v3, i64 %r) {
-; SDISEL-LABEL: select_noccmp2:
-; SDISEL:       ; %bb.0:
-; SDISEL-NEXT:    cmp x0, #0
-; SDISEL-NEXT:    ccmp x0, #13, #0, ge
-; SDISEL-NEXT:    cset w8, gt
-; SDISEL-NEXT:    cmp w8, #0
-; SDISEL-NEXT:    csel x0, xzr, x3, ne
-; SDISEL-NEXT:    sbfx w8, w8, #0, #1
-; SDISEL-NEXT:    adrp x9, _g@PAGE
-; SDISEL-NEXT:    str w8, [x9, _g@PAGEOFF]
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: select_noccmp2:
+; CHECK-SD:       ; %bb.0:
+; CHECK-SD-NEXT:    cmp x0, #0
+; CHECK-SD-NEXT:    ccmp x0, #13, #0, ge
+; CHECK-SD-NEXT:    cset w8, gt
+; CHECK-SD-NEXT:    cmp w8, #0
+; CHECK-SD-NEXT:    csel x0, xzr, x3, ne
+; CHECK-SD-NEXT:    sbfx w8, w8, #0, #1
+; CHECK-SD-NEXT:    adrp x9, _g@PAGE
+; CHECK-SD-NEXT:    str w8, [x9, _g@PAGEOFF]
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: select_noccmp2:
-; GISEL:       ; %bb.0:
-; GISEL-NEXT:    cmp x0, #14
-; GISEL-NEXT:    cset w8, hs
-; GISEL-NEXT:    tst w8, #0x1
-; GISEL-NEXT:    csel x0, xzr, x3, ne
-; GISEL-NEXT:    sbfx w8, w8, #0, #1
-; GISEL-NEXT:    adrp x9, _g@PAGE
-; GISEL-NEXT:    str w8, [x9, _g@PAGEOFF]
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: select_noccmp2:
+; CHECK-GI:       ; %bb.0:
+; CHECK-GI-NEXT:    cmp x0, #14
+; CHECK-GI-NEXT:    cset w8, hs
+; CHECK-GI-NEXT:    tst w8, #0x1
+; CHECK-GI-NEXT:    csel x0, xzr, x3, ne
+; CHECK-GI-NEXT:    sbfx w8, w8, #0, #1
+; CHECK-GI-NEXT:    adrp x9, _g@PAGE
+; CHECK-GI-NEXT:    str w8, [x9, _g@PAGEOFF]
+; CHECK-GI-NEXT:    ret
   %c0 = icmp slt i64 %v1, 0
   %c1 = icmp sgt i64 %v1, 13
   %or = or i1 %c0, %c1
@@ -661,33 +661,33 @@ define i64 @select_noccmp2(i64 %v1, i64 %v2, i64 %v3, i64 %r) {
 ; The following is not possible to implement with a single cmp;ccmp;csel
 ; sequence.
 define i32 @select_noccmp3(i32 %v0, i32 %v1, i32 %v2) {
-; SDISEL-LABEL: select_noccmp3:
-; SDISEL:       ; %bb.0:
-; SDISEL-NEXT:    cmp w0, #0
-; SDISEL-NEXT:    ccmp w0, #13, #0, ge
-; SDISEL-NEXT:    cset w8, gt
-; SDISEL-NEXT:    cmp w0, #22
-; SDISEL-NEXT:    mov w9, #44 ; =0x2c
-; SDISEL-NEXT:    ccmp w0, w9, #0, ge
-; SDISEL-NEXT:    csel w8, wzr, w8, le
-; SDISEL-NEXT:    cmp w0, #99
-; SDISEL-NEXT:    mov w9, #77 ; =0x4d
-; SDISEL-NEXT:    ccmp w0, w9, #4, ne
-; SDISEL-NEXT:    cset w9, eq
-; SDISEL-NEXT:    tst w8, w9
-; SDISEL-NEXT:    csel w0, w1, w2, ne
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: select_noccmp3:
+; CHECK-SD:       ; %bb.0:
+; CHECK-SD-NEXT:    cmp w0, #0
+; CHECK-SD-NEXT:    ccmp w0, #13, #0, ge
+; CHECK-SD-NEXT:    cset w8, gt
+; CHECK-SD-NEXT:    cmp w0, #22
+; CHECK-SD-NEXT:    mov w9, #44 ; =0x2c
+; CHECK-SD-NEXT:    ccmp w0, w9, #0, ge
+; CHECK-SD-NEXT:    csel w8, wzr, w8, le
+; CHECK-SD-NEXT:    cmp w0, #99
+; CHECK-SD-NEXT:    mov w9, #77 ; =0x4d
+; CHECK-SD-NEXT:    ccmp w0, w9, #4, ne
+; CHECK-SD-NEXT:    cset w9, eq
+; CHECK-SD-NEXT:    tst w8, w9
+; CHECK-SD-NEXT:    csel w0, w1, w2, ne
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: select_noccmp3:
-; GISEL:       ; %bb.0:
-; GISEL-NEXT:    mov w8, #99 ; =0x63
-; GISEL-NEXT:    sub w9, w0, #45
-; GISEL-NEXT:    cmp w0, #77
-; GISEL-NEXT:    ccmp w0, w8, #4, ne
-; GISEL-NEXT:    ccmn w9, #23, #2, eq
-; GISEL-NEXT:    ccmp w0, #14, #0, lo
-; GISEL-NEXT:    csel w0, w1, w2, hs
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: select_noccmp3:
+; CHECK-GI:       ; %bb.0:
+; CHECK-GI-NEXT:    mov w8, #99 ; =0x63
+; CHECK-GI-NEXT:    sub w9, w0, #45
+; CHECK-GI-NEXT:    cmp w0, #77
+; CHECK-GI-NEXT:    ccmp w0, w8, #4, ne
+; CHECK-GI-NEXT:    ccmn w9, #23, #2, eq
+; CHECK-GI-NEXT:    ccmp w0, #14, #0, lo
+; CHECK-GI-NEXT:    csel w0, w1, w2, hs
+; CHECK-GI-NEXT:    ret
   %c0 = icmp slt i32 %v0, 0
   %c1 = icmp sgt i32 %v0, 13
   %c2 = icmp slt i32 %v0, 22
@@ -864,27 +864,27 @@ define i32 @select_or_olt_ueq_ogt(double %v0, double %v1, double %v2, double %v3
 ; Verify that we correctly promote f16.
 
 define i32 @half_select_and_olt_oge(half %v0, half %v1, half %v2, half %v3, i32 %a, i32 %b) #0 {
-; SDISEL-LABEL: half_select_and_olt_oge:
-; SDISEL:       ; %bb.0:
-; SDISEL-NEXT:    fcvt s1, h1
-; SDISEL-NEXT:    fcvt s0, h0
-; SDISEL-NEXT:    fcmp s0, s1
-; SDISEL-NEXT:    fcvt s0, h3
-; SDISEL-NEXT:    fcvt s1, h2
-; SDISEL-NEXT:    fccmp s1, s0, #8, mi
-; SDISEL-NEXT:    csel w0, w0, w1, ge
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: half_select_and_olt_oge:
+; CHECK-SD:       ; %bb.0:
+; CHECK-SD-NEXT:    fcvt s1, h1
+; CHECK-SD-NEXT:    fcvt s0, h0
+; CHECK-SD-NEXT:    fcmp s0, s1
+; CHECK-SD-NEXT:    fcvt s0, h3
+; CHECK-SD-NEXT:    fcvt s1, h2
+; CHECK-SD-NEXT:    fccmp s1, s0, #8, mi
+; CHECK-SD-NEXT:    csel w0, w0, w1, ge
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: half_select_and_olt_oge:
-; GISEL:       ; %bb.0:
-; GISEL-NEXT:    fcvt s0, h0
-; GISEL-NEXT:    fcvt s1, h1
-; GISEL-NEXT:    fcvt s2, h2
-; GISEL-NEXT:    fcvt s3, h3
-; GISEL-NEXT:    fcmp s0, s1
-; GISEL-NEXT:    fccmp s2, s3, #8, mi
-; GISEL-NEXT:    csel w0, w0, w1, ge
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: half_select_and_olt_oge:
+; CHECK-GI:       ; %bb.0:
+; CHECK-GI-NEXT:    fcvt s0, h0
+; CHECK-GI-NEXT:    fcvt s1, h1
+; CHECK-GI-NEXT:    fcvt s2, h2
+; CHECK-GI-NEXT:    fcvt s3, h3
+; CHECK-GI-NEXT:    fcmp s0, s1
+; CHECK-GI-NEXT:    fccmp s2, s3, #8, mi
+; CHECK-GI-NEXT:    csel w0, w0, w1, ge
+; CHECK-GI-NEXT:    ret
   %c0 = fcmp olt half %v0, %v1
   %c1 = fcmp oge half %v2, %v3
   %cr = and i1 %c1, %c0
@@ -893,29 +893,29 @@ define i32 @half_select_and_olt_oge(half %v0, half %v1, half %v2, half %v3, i32
 }
 
 define i32 @half_select_and_olt_one(half %v0, half %v1, half %v2, half %v3, i32 %a, i32 %b) #0 {
-; SDISEL-LABEL: half_select_and_olt_one:
-; SDISEL:       ; %bb.0:
-; SDISEL-NEXT:    fcvt s1, h1
-; SDISEL-NEXT:    fcvt s0, h0
-; SDISEL-NEXT:    fcmp s0, s1
-; SDISEL-NEXT:    fcvt s0, h3
-; SDISEL-NEXT:    fcvt s1, h2
-; SDISEL-NEXT:    fccmp s1, s0, #4, mi
-; SDISEL-NEXT:    fccmp s1, s0, #1, ne
-; SDISEL-NEXT:    csel w0, w0, w1, vc
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: half_select_and_olt_one:
+; CHECK-SD:       ; %bb.0:
+; CHECK-SD-NEXT:    fcvt s1, h1
+; CHECK-SD-NEXT:    fcvt s0, h0
+; CHECK-SD-NEXT:    fcmp s0, s1
+; CHECK-SD-NEXT:    fcvt s0, h3
+; CHECK-SD-NEXT:    fcvt s1, h2
+; CHECK-SD-NEXT:    fccmp s1, s0, #4, mi
+; CHECK-SD-NEXT:    fccmp s1, s0, #1, ne
+; CHECK-SD-NEXT:    csel w0, w0, w1, vc
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: half_select_and_olt_one:
-; GISEL:       ; %bb.0:
-; GISEL-NEXT:    fcvt s0, h0
-; GISEL-NEXT:    fcvt s1, h1
-; GISEL-NEXT:    fcvt s2, h2
-; GISEL-NEXT:    fcvt s3, h3
-; GISEL-NEXT:    fcmp s0, s1
-; GISEL-NEXT:    fccmp s2, s3, #4, mi
-; GISEL-NEXT:    fccmp s2, s3, #1, ne
-; GISEL-NEXT:    csel w0, w0, w1, vc
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: half_select_and_olt_one:
+; CHECK-GI:       ; %bb.0:
+; CHECK-GI-NEXT:    fcvt s0, h0
+; CHECK-GI-NEXT:    fcvt s1, h1
+; CHECK-GI-NEXT:    fcvt s2, h2
+; CHECK-GI-NEXT:    fcvt s3, h3
+; CHECK-GI-NEXT:    fcmp s0, s1
+; CHECK-GI-NEXT:    fccmp s2, s3, #4, mi
+; CHECK-GI-NEXT:    fccmp s2, s3, #1, ne
+; CHECK-GI-NEXT:    csel w0, w0, w1, vc
+; CHECK-GI-NEXT:    ret
   %c0 = fcmp olt half %v0, %v1
   %c1 = fcmp one half %v2, %v3
   %cr = and i1 %c1, %c0
@@ -926,51 +926,51 @@ define i32 @half_select_and_olt_one(half %v0, half %v1, half %v2, half %v3, i32
 ; Also verify that we don't try to generate f128 FCCMPs, using RT calls instead.
 
 define i32 @f128_select_and_olt_oge(fp128 %v0, fp128 %v1, fp128 %v2, fp128 %v3, i32 %a, i32 %b) #0 {
-; SDISEL-LABEL: f128_select_and_olt_oge:
-; SDISEL:       ; %bb.0:
-; SDISEL-NEXT:    sub sp, sp, #80
-; SDISEL-NEXT:    stp x22, x21, [sp, #32] ; 16-byte Folded Spill
-; SDISEL-NEXT:    stp x20, x19, [sp, #48] ; 16-byte Folded Spill
-; SDISEL-NEXT:    stp x29, x30, [sp, #64] ; 16-byte Folded Spill
-; SDISEL-NEXT:    mov x19, x1
-; SDISEL-NEXT:    mov x20, x0
-; SDISEL-NEXT:    stp q2, q3, [sp] ; 32-byte Folded Spill
-; SDISEL-NEXT:    bl ___lttf2
-; SDISEL-NEXT:    cmp w0, #0
-; SDISEL-NEXT:    cset w21, lt
-; SDISEL-NEXT:    ldp q0, q1, [sp] ; 32-byte Folded Reload
-; SDISEL-NEXT:    bl ___getf2
-; SDISEL-NEXT:    cmp w0, #0
-; SDISEL-NEXT:    cset w8, ge
-; SDISEL-NEXT:    tst w8, w21
-; SDISEL-NEXT:    csel w0, w20, w19, ne
-; SDISEL-NEXT:    ldp x29, x30, [sp, #64] ; 16-byte Folded Reload
-; SDISEL-NEXT:    ldp x20, x19, [sp, #48] ; 16-byte Folded Reload
-; SDISEL-NEXT:    ldp x22, x21, [sp, #32] ; 16-byte Folded Reload
-; SDISEL-NEXT:    add sp, sp, #80
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: f128_select_and_olt_oge:
+; CHECK-SD:       ; %bb.0:
+; CHECK-SD-NEXT:    sub sp, sp, #80
+; CHECK-SD-NEXT:    stp x22, x21, [sp, #32] ; 16-byte Folded Spill
+; CHECK-SD-NEXT:    stp x20, x19, [sp, #48] ; 16-byte Folded Spill
+; CHECK-SD-NEXT:    stp x29, x30, [sp, #64] ; 16-byte Folded Spill
+; CHECK-SD-NEXT:    mov x19, x1
+; CHECK-SD-NEXT:    mov x20, x0
+; CHECK-SD-NEXT:    stp q2, q3, [sp] ; 32-byte Folded Spill
+; CHECK-SD-NEXT:    bl ___lttf2
+; CHECK-SD-NEXT:    cmp w0, #0
+; CHECK-SD-NEXT:    cset w21, lt
+; CHECK-SD-NEXT:    ldp q0, q1, [sp] ; 32-byte Folded Reload
+; CHECK-SD-NEXT:    bl ___getf2
+; CHECK-SD-NEXT:    cmp w0, #0
+; CHECK-SD-NEXT:    cset w8, ge
+; CHECK-SD-NEXT:    tst w8, w21
+; CHECK-SD-NEXT:    csel w0, w20, w19, ne
+; CHECK-SD-NEXT:    ldp x29, x30, [sp, #64] ; 16-byte Folded Reload
+; CHECK-SD-NEXT:    ldp x20, x19, [sp, #48] ; 16-byte Folded Reload
+; CHECK-SD-NEXT:    ldp x22, x21, [sp, #32] ; 16-byte Folded Reload
+; CHECK-SD-NEXT:    add sp, sp, #80
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: f128_select_and_olt_oge:
-; GISEL:       ; %bb.0:
-; GISEL-NEXT:    sub sp, sp, #80
-; GISEL-NEXT:    stp x22, x21, [sp, #32] ; 16-byte Folded Spill
-; GISEL-NEXT:    stp x20, x19, [sp, #48] ; 16-byte Folded Spill
-; GISEL-NEXT:    stp x29, x30, [sp, #64] ; 16-byte Folded Spill
-; GISEL-NEXT:    stp q3, q2, [sp] ; 32-byte Folded Spill
-; GISEL-NEXT:    mov x19, x0
-; GISEL-NEXT:    mov x20, x1
-; GISEL-NEXT:    bl ___lttf2
-; GISEL-NEXT:    mov x21, x0
-; GISEL-NEXT:    ldp q1, q0, [sp] ; 32-byte Folded Reload
-; GISEL-NEXT:    bl ___getf2
-; GISEL-NEXT:    cmp w21, #0
-; GISEL-NEXT:    ccmp w0, #0, #8, lt
-; GISEL-NEXT:    csel w0, w19, w20, ge
-; GISEL-NEXT:    ldp x29, x30, [sp, #64] ; 16-byte Folded Reload
-; GISEL-NEXT:    ldp x20, x19, [sp, #48] ; 16-byte Folded Reload
-; GISEL-NEXT:    ldp x22, x21, [sp, #32] ; 16-byte Folded Reload
-; GISEL-NEXT:    add sp, sp, #80
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: f128_select_and_olt_oge:
+; CHECK-GI:       ; %bb.0:
+; CHECK-GI-NEXT:    sub sp, sp, #80
+; CHECK-GI-NEXT:    stp x22, x21, [sp, #32] ; 16-byte Folded Spill
+; CHECK-GI-NEXT:    stp x20, x19, [sp, #48] ; 16-byte Folded Spill
+; CHECK-GI-NEXT:    stp x29, x30, [sp, #64] ; 16-byte Folded Spill
+; CHECK-GI-NEXT:    stp q3, q2, [sp] ; 32-byte Folded Spill
+; CHECK-GI-NEXT:    mov x19, x0
+; CHECK-GI-NEXT:    mov x20, x1
+; CHECK-GI-NEXT:    bl ___lttf2
+; CHECK-GI-NEXT:    mov x21, x0
+; CHECK-GI-NEXT:    ldp q1, q0, [sp] ; 32-byte Folded Reload
+; CHECK-GI-NEXT:    bl ___getf2
+; CHECK-GI-NEXT:    cmp w21, #0
+; CHECK-GI-NEXT:    ccmp w0, #0, #8, lt
+; CHECK-GI-NEXT:    csel w0, w19, w20, ge
+; CHECK-GI-NEXT:    ldp x29, x30, [sp, #64] ; 16-byte Folded Reload
+; CHECK-GI-NEXT:    ldp x20, x19, [sp, #48] ; 16-byte Folded Reload
+; CHECK-GI-NEXT:    ldp x22, x21, [sp, #32] ; 16-byte Folded Reload
+; CHECK-GI-NEXT:    add sp, sp, #80
+; CHECK-GI-NEXT:    ret
   %c0 = fcmp olt fp128 %v0, %v1
   %c1 = fcmp oge fp128 %v2, %v3
   %cr = and i1 %c1, %c0
@@ -1048,46 +1048,46 @@ define i32 @deep_or2(i32 %a0, i32 %a1, i32 %a2, i32 %a3, i32 %x, i32 %y) {
 ; This test is trying to test that multiple ccmp's don't get created in a way
 ; that they would have multiple uses. It doesn't seem to.
 define i32 @multiccmp(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %x, i32 %y) #0 {
-; SDISEL-LABEL: multiccmp:
-; SDISEL:       ; %bb.0: ; %entry
-; SDISEL-NEXT:    stp x22, x21, [sp, #-48]! ; 16-byte Folded Spill
-; SDISEL-NEXT:    stp x20, x19, [sp, #16] ; 16-byte Folded Spill
-; SDISEL-NEXT:    stp x29, x30, [sp, #32] ; 16-byte Folded Spill
-; SDISEL-NEXT:    mov x19, x5
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    cset w20, gt
-; SDISEL-NEXT:    cmp w2, w3
-; SDISEL-NEXT:    cset w21, ne
-; SDISEL-NEXT:    tst w20, w21
-; SDISEL-NEXT:    csel w0, w5, w4, ne
-; SDISEL-NEXT:    bl _callee
-; SDISEL-NEXT:    tst w20, w21
-; SDISEL-NEXT:    csel w0, w0, w19, ne
-; SDISEL-NEXT:    bl _callee
-; SDISEL-NEXT:    ldp x29, x30, [sp, #32] ; 16-byte Folded Reload
-; SDISEL-NEXT:    ldp x20, x19, [sp, #16] ; 16-byte Folded Reload
-; SDISEL-NEXT:    ldp x22, x21, [sp], #48 ; 16-byte Folded Reload
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: multiccmp:
+; CHECK-SD:       ; %bb.0: ; %entry
+; CHECK-SD-NEXT:    stp x22, x21, [sp, #-48]! ; 16-byte Folded Spill
+; CHECK-SD-NEXT:    stp x20, x19, [sp, #16] ; 16-byte Folded Spill
+; CHECK-SD-NEXT:    stp x29, x30, [sp, #32] ; 16-byte Folded Spill
+; CHECK-SD-NEXT:    mov x19, x5
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    cset w20, gt
+; CHECK-SD-NEXT:    cmp w2, w3
+; CHECK-SD-NEXT:    cset w21, ne
+; CHECK-SD-NEXT:    tst w20, w21
+; CHECK-SD-NEXT:    csel w0, w5, w4, ne
+; CHECK-SD-NEXT:    bl _callee
+; CHECK-SD-NEXT:    tst w20, w21
+; CHECK-SD-NEXT:    csel w0, w0, w19, ne
+; CHECK-SD-NEXT:    bl _callee
+; CHECK-SD-NEXT:    ldp x29, x30, [sp, #32] ; 16-byte Folded Reload
+; CHECK-SD-NEXT:    ldp x20, x19, [sp, #16] ; 16-byte Folded Reload
+; CHECK-SD-NEXT:    ldp x22, x21, [sp], #48 ; 16-byte Folded Reload
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: multiccmp:
-; GISEL:       ; %bb.0: ; %entry
-; GISEL-NEXT:    stp x20, x19, [sp, #-32]! ; 16-byte Folded Spill
-; GISEL-NEXT:    stp x29, x30, [sp, #16] ; 16-byte Folded Spill
-; GISEL-NEXT:    mov x19, x5
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, gt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ne
-; GISEL-NEXT:    and w20, w8, w9
-; GISEL-NEXT:    tst w20, #0x1
-; GISEL-NEXT:    csel w0, w5, w4, ne
-; GISEL-NEXT:    bl _callee
-; GISEL-NEXT:    tst w20, #0x1
-; GISEL-NEXT:    csel w0, w0, w19, ne
-; GISEL-NEXT:    bl _callee
-; GISEL-NEXT:    ldp x29, x30, [sp, #16] ; 16-byte Folded Reload
-; GISEL-NEXT:    ldp x20, x19, [sp], #32 ; 16-byte Folded Reload
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: multiccmp:
+; CHECK-GI:       ; %bb.0: ; %entry
+; CHECK-GI-NEXT:    stp x20, x19, [sp, #-32]! ; 16-byte Folded Spill
+; CHECK-GI-NEXT:    stp x29, x30, [sp, #16] ; 16-byte Folded Spill
+; CHECK-GI-NEXT:    mov x19, x5
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, gt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ne
+; CHECK-GI-NEXT:    and w20, w8, w9
+; CHECK-GI-NEXT:    tst w20, #0x1
+; CHECK-GI-NEXT:    csel w0, w5, w4, ne
+; CHECK-GI-NEXT:    bl _callee
+; CHECK-GI-NEXT:    tst w20, #0x1
+; CHECK-GI-NEXT:    csel w0, w0, w19, ne
+; CHECK-GI-NEXT:    bl _callee
+; CHECK-GI-NEXT:    ldp x29, x30, [sp, #16] ; 16-byte Folded Reload
+; CHECK-GI-NEXT:    ldp x20, x19, [sp], #32 ; 16-byte Folded Reload
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sgt i32 %s0, %s1
   %c1 = icmp ne i32 %s2, %s3
@@ -1100,57 +1100,57 @@ entry:
 }
 
 define i32 @multiccmp2(i32 %s0, i32 %s1, i32 %s2, i32 %s3, i32 %x, i32 %y) #0 {
-; SDISEL-LABEL: multiccmp2:
-; SDISEL:       ; %bb.0: ; %entry
-; SDISEL-NEXT:    stp x22, x21, [sp, #-48]! ; 16-byte Folded Spill
-; SDISEL-NEXT:    stp x20, x19, [sp, #16] ; 16-byte Folded Spill
-; SDISEL-NEXT:    stp x29, x30, [sp, #32] ; 16-byte Folded Spill
-; SDISEL-NEXT:    mov x19, x5
-; SDISEL-NEXT:    mov x20, x3
-; SDISEL-NEXT:    mov x21, x0
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    cset w8, gt
-; SDISEL-NEXT:    cmp w2, w3
-; SDISEL-NEXT:    cset w22, ne
-; SDISEL-NEXT:    tst w8, w22
-; SDISEL-NEXT:    csel w0, w5, w4, ne
-; SDISEL-NEXT:    bl _callee
-; SDISEL-NEXT:    cmp w21, w20
-; SDISEL-NEXT:    cset w8, eq
-; SDISEL-NEXT:    tst w22, w8
-; SDISEL-NEXT:    csel w0, w0, w19, ne
-; SDISEL-NEXT:    bl _callee
-; SDISEL-NEXT:    ldp x29, x30, [sp, #32] ; 16-byte Folded Reload
-; SDISEL-NEXT:    ldp x20, x19, [sp, #16] ; 16-byte Folded Reload
-; SDISEL-NEXT:    ldp x22, x21, [sp], #48 ; 16-byte Folded Reload
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: multiccmp2:
+; CHECK-SD:       ; %bb.0: ; %entry
+; CHECK-SD-NEXT:    stp x22, x21, [sp, #-48]! ; 16-byte Folded Spill
+; CHECK-SD-NEXT:    stp x20, x19, [sp, #16] ; 16-byte Folded Spill
+; CHECK-SD-NEXT:    stp x29, x30, [sp, #32] ; 16-byte Folded Spill
+; CHECK-SD-NEXT:    mov x19, x5
+; CHECK-SD-NEXT:    mov x20, x3
+; CHECK-SD-NEXT:    mov x21, x0
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    cset w8, gt
+; CHECK-SD-NEXT:    cmp w2, w3
+; CHECK-SD-NEXT:    cset w22, ne
+; CHECK-SD-NEXT:    tst w8, w22
+; CHECK-SD-NEXT:    csel w0, w5, w4, ne
+; CHECK-SD-NEXT:    bl _callee
+; CHECK-SD-NEXT:    cmp w21, w20
+; CHECK-SD-NEXT:    cset w8, eq
+; CHECK-SD-NEXT:    tst w22, w8
+; CHECK-SD-NEXT:    csel w0, w0, w19, ne
+; CHECK-SD-NEXT:    bl _callee
+; CHECK-SD-NEXT:    ldp x29, x30, [sp, #32] ; 16-byte Folded Reload
+; CHECK-SD-NEXT:    ldp x20, x19, [sp, #16] ; 16-byte Folded Reload
+; CHECK-SD-NEXT:    ldp x22, x21, [sp], #48 ; 16-byte Folded Reload
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: multiccmp2:
-; GISEL:       ; %bb.0: ; %entry
-; GISEL-NEXT:    stp x22, x21, [sp, #-48]! ; 16-byte Folded Spill
-; GISEL-NEXT:    stp x20, x19, [sp, #16] ; 16-byte Folded Spill
-; GISEL-NEXT:    stp x29, x30, [sp, #32] ; 16-byte Folded Spill
-; GISEL-NEXT:    mov x19, x0
-; GISEL-NEXT:    mov x20, x3
-; GISEL-NEXT:    mov x21, x5
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, gt
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w22, ne
-; GISEL-NEXT:    and w8, w8, w22
-; GISEL-NEXT:    tst w8, #0x1
-; GISEL-NEXT:    csel w0, w5, w4, ne
-; GISEL-NEXT:    bl _callee
-; GISEL-NEXT:    cmp w19, w20
-; GISEL-NEXT:    cset w8, eq
-; GISEL-NEXT:    and w8, w22, w8
-; GISEL-NEXT:    tst w8, #0x1
-; GISEL-NEXT:    csel w0, w0, w21, ne
-; GISEL-NEXT:    bl _callee
-; GISEL-NEXT:    ldp x29, x30, [sp, #32] ; 16-byte Folded Reload
-; GISEL-NEXT:    ldp x20, x19, [sp, #16] ; 16-byte Folded Reload
-; GISEL-NEXT:    ldp x22, x21, [sp], #48 ; 16-byte Folded Reload
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: multiccmp2:
+; CHECK-GI:       ; %bb.0: ; %entry
+; CHECK-GI-NEXT:    stp x22, x21, [sp, #-48]! ; 16-byte Folded Spill
+; CHECK-GI-NEXT:    stp x20, x19, [sp, #16] ; 16-byte Folded Spill
+; CHECK-GI-NEXT:    stp x29, x30, [sp, #32] ; 16-byte Folded Spill
+; CHECK-GI-NEXT:    mov x19, x0
+; CHECK-GI-NEXT:    mov x20, x3
+; CHECK-GI-NEXT:    mov x21, x5
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, gt
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w22, ne
+; CHECK-GI-NEXT:    and w8, w8, w22
+; CHECK-GI-NEXT:    tst w8, #0x1
+; CHECK-GI-NEXT:    csel w0, w5, w4, ne
+; CHECK-GI-NEXT:    bl _callee
+; CHECK-GI-NEXT:    cmp w19, w20
+; CHECK-GI-NEXT:    cset w8, eq
+; CHECK-GI-NEXT:    and w8, w22, w8
+; CHECK-GI-NEXT:    tst w8, #0x1
+; CHECK-GI-NEXT:    csel w0, w0, w21, ne
+; CHECK-GI-NEXT:    bl _callee
+; CHECK-GI-NEXT:    ldp x29, x30, [sp, #32] ; 16-byte Folded Reload
+; CHECK-GI-NEXT:    ldp x20, x19, [sp, #16] ; 16-byte Folded Reload
+; CHECK-GI-NEXT:    ldp x22, x21, [sp], #48 ; 16-byte Folded Reload
+; CHECK-GI-NEXT:    ret
 entry:
   %c0 = icmp sgt i32 %s0, %s1
   %c1 = icmp ne i32 %s2, %s3
@@ -1168,21 +1168,21 @@ entry:
 declare i32 @callee(i32)
 
 define i1 @cmp_and_negative_const(i32 %0, i32 %1) {
-; SDISEL-LABEL: cmp_and_negative_const:
-; SDISEL:       ; %bb.0:
-; SDISEL-NEXT:    cmn w0, #1
-; SDISEL-NEXT:    ccmn w1, #2, #0, eq
-; SDISEL-NEXT:    cset w0, eq
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: cmp_and_negative_const:
+; CHECK-SD:       ; %bb.0:
+; CHECK-SD-NEXT:    cmn w0, #1
+; CHECK-SD-NEXT:    ccmn w1, #2, #0, eq
+; CHECK-SD-NEXT:    cset w0, eq
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: cmp_and_negative_const:
-; GISEL:       ; %bb.0:
-; GISEL-NEXT:    cmn w0, #1
-; GISEL-NEXT:    cset w8, eq
-; GISEL-NEXT:    cmn w1, #2
-; GISEL-NEXT:    cset w9, eq
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: cmp_and_negative_const:
+; CHECK-GI:       ; %bb.0:
+; CHECK-GI-NEXT:    cmn w0, #1
+; CHECK-GI-NEXT:    cset w8, eq
+; CHECK-GI-NEXT:    cmn w1, #2
+; CHECK-GI-NEXT:    cset w9, eq
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
   %3 = icmp eq i32 %0, -1
   %4 = icmp eq i32 %1, -2
   %5 = and i1 %3, %4
@@ -1190,21 +1190,21 @@ define i1 @cmp_and_negative_const(i32 %0, i32 %1) {
 }
 
 define i1 @cmp_or_negative_const(i32 %a, i32 %b) {
-; SDISEL-LABEL: cmp_or_negative_const:
-; SDISEL:       ; %bb.0:
-; SDISEL-NEXT:    cmn w0, #1
-; SDISEL-NEXT:    ccmn w1, #2, #4, ne
-; SDISEL-NEXT:    cset w0, eq
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: cmp_or_negative_const:
+; CHECK-SD:       ; %bb.0:
+; CHECK-SD-NEXT:    cmn w0, #1
+; CHECK-SD-NEXT:    ccmn w1, #2, #4, ne
+; CHECK-SD-NEXT:    cset w0, eq
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: cmp_or_negative_const:
-; GISEL:       ; %bb.0:
-; GISEL-NEXT:    cmn w0, #1
-; GISEL-NEXT:    cset w8, eq
-; GISEL-NEXT:    cmn w1, #2
-; GISEL-NEXT:    cset w9, eq
-; GISEL-NEXT:    orr w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: cmp_or_negative_const:
+; CHECK-GI:       ; %bb.0:
+; CHECK-GI-NEXT:    cmn w0, #1
+; CHECK-GI-NEXT:    cset w8, eq
+; CHECK-GI-NEXT:    cmn w1, #2
+; CHECK-GI-NEXT:    cset w9, eq
+; CHECK-GI-NEXT:    orr w0, w8, w9
+; CHECK-GI-NEXT:    ret
   %cmp = icmp eq i32 %a, -1
   %cmp1 = icmp eq i32 %b, -2
   %or.cond = or i1 %cmp, %cmp1
diff --git a/llvm/test/CodeGen/AArch64/cmp-chains.ll b/llvm/test/CodeGen/AArch64/cmp-chains.ll
index 4b816df..3620444 100644
--- a/llvm/test/CodeGen/AArch64/cmp-chains.ll
+++ b/llvm/test/CodeGen/AArch64/cmp-chains.ll
@@ -1,26 +1,26 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=aarch64-- | FileCheck %s --check-prefixes=CHECK,SDISEL
-; RUN: llc < %s -mtriple=aarch64-- -global-isel | FileCheck %s --check-prefixes=CHECK,GISEL
+; RUN: llc < %s -mtriple=aarch64-- | FileCheck %s --check-prefixes=CHECK,CHECK-SD
+; RUN: llc < %s -mtriple=aarch64-- -global-isel | FileCheck %s --check-prefixes=CHECK,CHECK-GI
 
 ; Ensure chains of comparisons produce chains of `ccmp`
 
 ; (x0 < x1) && (x2 > x3)
 define i32 @cmp_and2(i32 %0, i32 %1, i32 %2, i32 %3) {
-; SDISEL-LABEL: cmp_and2:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, lo
-; SDISEL-NEXT:    cset w0, hi
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: cmp_and2:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, lo
+; CHECK-SD-NEXT:    cset w0, hi
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: cmp_and2:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lo
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hi
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: cmp_and2:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lo
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hi
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
   %5 = icmp ult i32 %0, %1
   %6 = icmp ugt i32 %2, %3
   %7 = select i1 %5, i1 %6, i1 false
@@ -30,25 +30,25 @@ define i32 @cmp_and2(i32 %0, i32 %1, i32 %2, i32 %3) {
 
 ; (x0 < x1) && (x2 > x3) && (x4 != x5)
 define i32 @cmp_and3(i32 %0, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5) {
-; SDISEL-LABEL: cmp_and3:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, lo
-; SDISEL-NEXT:    ccmp w4, w5, #4, hi
-; SDISEL-NEXT:    cset w0, ne
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: cmp_and3:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, lo
+; CHECK-SD-NEXT:    ccmp w4, w5, #4, hi
+; CHECK-SD-NEXT:    cset w0, ne
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: cmp_and3:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lo
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hi
-; GISEL-NEXT:    cmp w4, w5
-; GISEL-NEXT:    and w8, w8, w9
-; GISEL-NEXT:    cset w9, ne
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: cmp_and3:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lo
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hi
+; CHECK-GI-NEXT:    cmp w4, w5
+; CHECK-GI-NEXT:    and w8, w8, w9
+; CHECK-GI-NEXT:    cset w9, ne
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
   %7 = icmp ult i32 %0, %1
   %8 = icmp ugt i32 %2, %3
   %9 = select i1 %7, i1 %8, i1 false
@@ -60,29 +60,29 @@ define i32 @cmp_and3(i32 %0, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5) {
 
 ; (x0 < x1) && (x2 > x3) && (x4 != x5) && (x6 == x7)
 define i32 @cmp_and4(i32 %0, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5, i32 %6, i32 %7) {
-; SDISEL-LABEL: cmp_and4:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    cmp w2, w3
-; SDISEL-NEXT:    ccmp w0, w1, #2, hi
-; SDISEL-NEXT:    ccmp w4, w5, #4, lo
-; SDISEL-NEXT:    ccmp w6, w7, #0, ne
-; SDISEL-NEXT:    cset w0, eq
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: cmp_and4:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    cmp w2, w3
+; CHECK-SD-NEXT:    ccmp w0, w1, #2, hi
+; CHECK-SD-NEXT:    ccmp w4, w5, #4, lo
+; CHECK-SD-NEXT:    ccmp w6, w7, #0, ne
+; CHECK-SD-NEXT:    cset w0, eq
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: cmp_and4:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w8, hi
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w9, lo
-; GISEL-NEXT:    cmp w4, w5
-; GISEL-NEXT:    cset w10, ne
-; GISEL-NEXT:    cmp w6, w7
-; GISEL-NEXT:    and w8, w8, w9
-; GISEL-NEXT:    cset w11, eq
-; GISEL-NEXT:    and w9, w10, w11
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: cmp_and4:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w8, hi
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w9, lo
+; CHECK-GI-NEXT:    cmp w4, w5
+; CHECK-GI-NEXT:    cset w10, ne
+; CHECK-GI-NEXT:    cmp w6, w7
+; CHECK-GI-NEXT:    and w8, w8, w9
+; CHECK-GI-NEXT:    cset w11, eq
+; CHECK-GI-NEXT:    and w9, w10, w11
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
   %9 = icmp ugt i32 %2, %3
   %10 = icmp ult i32 %0, %1
   %11 = select i1 %9, i1 %10, i1 false
@@ -96,22 +96,22 @@ define i32 @cmp_and4(i32 %0, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5, i32 %6, i32
 
 ; (x0 < x1) || (x2 > x3)
 define i32 @cmp_or2(i32 %0, i32 %1, i32 %2, i32 %3) {
-; SDISEL-LABEL: cmp_or2:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #0, hs
-; SDISEL-NEXT:    cset w0, ne
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: cmp_or2:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #0, hs
+; CHECK-SD-NEXT:    cset w0, ne
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: cmp_or2:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lo
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, ne
-; GISEL-NEXT:    orr w8, w8, w9
-; GISEL-NEXT:    and w0, w8, #0x1
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: cmp_or2:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lo
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, ne
+; CHECK-GI-NEXT:    orr w8, w8, w9
+; CHECK-GI-NEXT:    and w0, w8, #0x1
+; CHECK-GI-NEXT:    ret
   %5 = icmp ult i32 %0, %1
   %6 = icmp ne i32 %2, %3
   %7 = select i1 %5, i1 true, i1 %6
@@ -121,26 +121,26 @@ define i32 @cmp_or2(i32 %0, i32 %1, i32 %2, i32 %3) {
 
 ; (x0 < x1) || (x2 > x3) || (x4 != x5)
 define i32 @cmp_or3(i32 %0, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5) {
-; SDISEL-LABEL: cmp_or3:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, hs
-; SDISEL-NEXT:    ccmp w4, w5, #0, ls
-; SDISEL-NEXT:    cset w0, ne
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: cmp_or3:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, hs
+; CHECK-SD-NEXT:    ccmp w4, w5, #0, ls
+; CHECK-SD-NEXT:    cset w0, ne
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: cmp_or3:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lo
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hi
-; GISEL-NEXT:    cmp w4, w5
-; GISEL-NEXT:    orr w8, w8, w9
-; GISEL-NEXT:    cset w9, ne
-; GISEL-NEXT:    orr w8, w8, w9
-; GISEL-NEXT:    and w0, w8, #0x1
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: cmp_or3:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lo
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hi
+; CHECK-GI-NEXT:    cmp w4, w5
+; CHECK-GI-NEXT:    orr w8, w8, w9
+; CHECK-GI-NEXT:    cset w9, ne
+; CHECK-GI-NEXT:    orr w8, w8, w9
+; CHECK-GI-NEXT:    and w0, w8, #0x1
+; CHECK-GI-NEXT:    ret
   %7 = icmp ult i32 %0, %1
   %8 = icmp ugt i32 %2, %3
   %9 = select i1 %7, i1 true, i1 %8
@@ -152,30 +152,30 @@ define i32 @cmp_or3(i32 %0, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5) {
 
 ; (x0 < x1) || (x2 > x3) || (x4 != x5) || (x6 == x7)
 define i32 @cmp_or4(i32 %0, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5, i32 %6, i32 %7) {
-; SDISEL-LABEL: cmp_or4:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    cmp w0, w1
-; SDISEL-NEXT:    ccmp w2, w3, #2, hs
-; SDISEL-NEXT:    ccmp w4, w5, #0, ls
-; SDISEL-NEXT:    ccmp w6, w7, #4, eq
-; SDISEL-NEXT:    cset w0, eq
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: cmp_or4:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    cmp w0, w1
+; CHECK-SD-NEXT:    ccmp w2, w3, #2, hs
+; CHECK-SD-NEXT:    ccmp w4, w5, #0, ls
+; CHECK-SD-NEXT:    ccmp w6, w7, #4, eq
+; CHECK-SD-NEXT:    cset w0, eq
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: cmp_or4:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    cmp w0, w1
-; GISEL-NEXT:    cset w8, lo
-; GISEL-NEXT:    cmp w2, w3
-; GISEL-NEXT:    cset w9, hi
-; GISEL-NEXT:    cmp w4, w5
-; GISEL-NEXT:    cset w10, ne
-; GISEL-NEXT:    cmp w6, w7
-; GISEL-NEXT:    orr w8, w8, w9
-; GISEL-NEXT:    cset w11, eq
-; GISEL-NEXT:    orr w9, w10, w11
-; GISEL-NEXT:    orr w8, w8, w9
-; GISEL-NEXT:    and w0, w8, #0x1
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: cmp_or4:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    cmp w0, w1
+; CHECK-GI-NEXT:    cset w8, lo
+; CHECK-GI-NEXT:    cmp w2, w3
+; CHECK-GI-NEXT:    cset w9, hi
+; CHECK-GI-NEXT:    cmp w4, w5
+; CHECK-GI-NEXT:    cset w10, ne
+; CHECK-GI-NEXT:    cmp w6, w7
+; CHECK-GI-NEXT:    orr w8, w8, w9
+; CHECK-GI-NEXT:    cset w11, eq
+; CHECK-GI-NEXT:    orr w9, w10, w11
+; CHECK-GI-NEXT:    orr w8, w8, w9
+; CHECK-GI-NEXT:    and w0, w8, #0x1
+; CHECK-GI-NEXT:    ret
   %9 = icmp ult i32 %0, %1
   %10 = icmp ugt i32 %2, %3
   %11 = select i1 %9, i1 true, i1 %10
@@ -189,22 +189,22 @@ define i32 @cmp_or4(i32 %0, i32 %1, i32 %2, i32 %3, i32 %4, i32 %5, i32 %6, i32
 
 ; (x0 != 0) || (x1 != 0)
 define i32 @true_or2(i32 %0, i32 %1) {
-; SDISEL-LABEL: true_or2:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    orr w8, w0, w1
-; SDISEL-NEXT:    cmp w8, #0
-; SDISEL-NEXT:    cset w0, ne
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: true_or2:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    orr w8, w0, w1
+; CHECK-SD-NEXT:    cmp w8, #0
+; CHECK-SD-NEXT:    cset w0, ne
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: true_or2:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    cmp w0, #0
-; GISEL-NEXT:    cset w8, ne
-; GISEL-NEXT:    cmp w1, #0
-; GISEL-NEXT:    cset w9, ne
-; GISEL-NEXT:    orr w8, w8, w9
-; GISEL-NEXT:    and w0, w8, #0x1
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: true_or2:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    cmp w0, #0
+; CHECK-GI-NEXT:    cset w8, ne
+; CHECK-GI-NEXT:    cmp w1, #0
+; CHECK-GI-NEXT:    cset w9, ne
+; CHECK-GI-NEXT:    orr w8, w8, w9
+; CHECK-GI-NEXT:    and w0, w8, #0x1
+; CHECK-GI-NEXT:    ret
   %3 = icmp ne i32 %0, 0
   %4 = icmp ne i32 %1, 0
   %5 = select i1 %3, i1 true, i1 %4
@@ -214,26 +214,26 @@ define i32 @true_or2(i32 %0, i32 %1) {
 
 ; (x0 != 0) || (x1 != 0) || (x2 != 0)
 define i32 @true_or3(i32 %0, i32 %1, i32 %2) {
-; SDISEL-LABEL: true_or3:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    orr w8, w0, w1
-; SDISEL-NEXT:    orr w8, w8, w2
-; SDISEL-NEXT:    cmp w8, #0
-; SDISEL-NEXT:    cset w0, ne
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: true_or3:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    orr w8, w0, w1
+; CHECK-SD-NEXT:    orr w8, w8, w2
+; CHECK-SD-NEXT:    cmp w8, #0
+; CHECK-SD-NEXT:    cset w0, ne
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: true_or3:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    cmp w0, #0
-; GISEL-NEXT:    cset w8, ne
-; GISEL-NEXT:    cmp w1, #0
-; GISEL-NEXT:    cset w9, ne
-; GISEL-NEXT:    cmp w2, #0
-; GISEL-NEXT:    orr w8, w8, w9
-; GISEL-NEXT:    cset w9, ne
-; GISEL-NEXT:    orr w8, w8, w9
-; GISEL-NEXT:    and w0, w8, #0x1
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: true_or3:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    cmp w0, #0
+; CHECK-GI-NEXT:    cset w8, ne
+; CHECK-GI-NEXT:    cmp w1, #0
+; CHECK-GI-NEXT:    cset w9, ne
+; CHECK-GI-NEXT:    cmp w2, #0
+; CHECK-GI-NEXT:    orr w8, w8, w9
+; CHECK-GI-NEXT:    cset w9, ne
+; CHECK-GI-NEXT:    orr w8, w8, w9
+; CHECK-GI-NEXT:    and w0, w8, #0x1
+; CHECK-GI-NEXT:    ret
   %4 = icmp ne i32 %0, 0
   %5 = icmp ne i32 %1, 0
   %6 = select i1 %4, i1 true, i1 %5
@@ -260,22 +260,22 @@ define i32 @neg_range_int(i32 %a, i32 %b, i32 %c) {
 
 ; (b > -(d | 1) && a < c)
 define i32 @neg_range_int_comp(i32 %a, i32 %b, i32 %c, i32 %d) {
-; SDISEL-LABEL: neg_range_int_comp:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    orr w8, w3, #0x1
-; SDISEL-NEXT:    cmp w0, w2
-; SDISEL-NEXT:    ccmn w1, w8, #4, lt
-; SDISEL-NEXT:    csel w0, w1, w0, gt
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: neg_range_int_comp:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    orr w8, w3, #0x1
+; CHECK-SD-NEXT:    cmp w0, w2
+; CHECK-SD-NEXT:    ccmn w1, w8, #4, lt
+; CHECK-SD-NEXT:    csel w0, w1, w0, gt
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: neg_range_int_comp:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    orr w8, w3, #0x1
-; GISEL-NEXT:    cmp w0, w2
-; GISEL-NEXT:    neg w8, w8
-; GISEL-NEXT:    ccmp w1, w8, #4, lt
-; GISEL-NEXT:    csel w0, w1, w0, gt
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: neg_range_int_comp:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    orr w8, w3, #0x1
+; CHECK-GI-NEXT:    cmp w0, w2
+; CHECK-GI-NEXT:    neg w8, w8
+; CHECK-GI-NEXT:    ccmp w1, w8, #4, lt
+; CHECK-GI-NEXT:    csel w0, w1, w0, gt
+; CHECK-GI-NEXT:    ret
   %dor = or i32 %d, 1
   %negd = sub i32 0, %dor
   %cmp = icmp sgt i32 %b, %negd
@@ -287,22 +287,22 @@ define i32 @neg_range_int_comp(i32 %a, i32 %b, i32 %c, i32 %d) {
 
 ; (b >u -(d | 1) && a < c)
 define i32 @neg_range_int_comp_u(i32 %a, i32 %b, i32 %c, i32 %d) {
-; SDISEL-LABEL: neg_range_int_comp_u:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    orr w8, w3, #0x1
-; SDISEL-NEXT:    cmp w0, w2
-; SDISEL-NEXT:    ccmn w1, w8, #0, lt
-; SDISEL-NEXT:    csel w0, w1, w0, hi
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: neg_range_int_comp_u:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    orr w8, w3, #0x1
+; CHECK-SD-NEXT:    cmp w0, w2
+; CHECK-SD-NEXT:    ccmn w1, w8, #0, lt
+; CHECK-SD-NEXT:    csel w0, w1, w0, hi
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: neg_range_int_comp_u:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    orr w8, w3, #0x1
-; GISEL-NEXT:    cmp w0, w2
-; GISEL-NEXT:    neg w8, w8
-; GISEL-NEXT:    ccmp w1, w8, #0, lt
-; GISEL-NEXT:    csel w0, w1, w0, hi
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: neg_range_int_comp_u:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    orr w8, w3, #0x1
+; CHECK-GI-NEXT:    cmp w0, w2
+; CHECK-GI-NEXT:    neg w8, w8
+; CHECK-GI-NEXT:    ccmp w1, w8, #0, lt
+; CHECK-GI-NEXT:    csel w0, w1, w0, hi
+; CHECK-GI-NEXT:    ret
   %dor = or i32 %d, 1
   %negd = sub i32 0, %dor
   %cmp = icmp ugt i32 %b, %negd
@@ -314,22 +314,22 @@ define i32 @neg_range_int_comp_u(i32 %a, i32 %b, i32 %c, i32 %d) {
 
 ; (b > -(d | 1) && a u < c)
 define i32 @neg_range_int_comp_ua(i32 %a, i32 %b, i32 %c, i32 %d) {
-; SDISEL-LABEL: neg_range_int_comp_ua:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    orr w8, w3, #0x1
-; SDISEL-NEXT:    cmp w0, w2
-; SDISEL-NEXT:    ccmn w1, w8, #4, lo
-; SDISEL-NEXT:    csel w0, w1, w0, gt
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: neg_range_int_comp_ua:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    orr w8, w3, #0x1
+; CHECK-SD-NEXT:    cmp w0, w2
+; CHECK-SD-NEXT:    ccmn w1, w8, #4, lo
+; CHECK-SD-NEXT:    csel w0, w1, w0, gt
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: neg_range_int_comp_ua:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    orr w8, w3, #0x1
-; GISEL-NEXT:    cmp w0, w2
-; GISEL-NEXT:    neg w8, w8
-; GISEL-NEXT:    ccmp w1, w8, #4, lo
-; GISEL-NEXT:    csel w0, w1, w0, gt
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: neg_range_int_comp_ua:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    orr w8, w3, #0x1
+; CHECK-GI-NEXT:    cmp w0, w2
+; CHECK-GI-NEXT:    neg w8, w8
+; CHECK-GI-NEXT:    ccmp w1, w8, #4, lo
+; CHECK-GI-NEXT:    csel w0, w1, w0, gt
+; CHECK-GI-NEXT:    ret
   %dor = or i32 %d, 1
   %negd = sub i32 0, %dor
   %cmp = icmp sgt i32 %b, %negd
@@ -341,19 +341,19 @@ define i32 @neg_range_int_comp_ua(i32 %a, i32 %b, i32 %c, i32 %d) {
 
 ; (b <= -3 && a > c)
 define i32 @neg_range_int_2(i32 %a, i32 %b, i32 %c) {
-; SDISEL-LABEL: neg_range_int_2:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    cmp w0, w2
-; SDISEL-NEXT:    ccmn w1, #4, #4, gt
-; SDISEL-NEXT:    csel w0, w1, w0, gt
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: neg_range_int_2:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    cmp w0, w2
+; CHECK-SD-NEXT:    ccmn w1, #4, #4, gt
+; CHECK-SD-NEXT:    csel w0, w1, w0, gt
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: neg_range_int_2:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    cmp w0, w2
-; GISEL-NEXT:    ccmn w1, #3, #8, gt
-; GISEL-NEXT:    csel w0, w1, w0, ge
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: neg_range_int_2:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    cmp w0, w2
+; CHECK-GI-NEXT:    ccmn w1, #3, #8, gt
+; CHECK-GI-NEXT:    csel w0, w1, w0, ge
+; CHECK-GI-NEXT:    ret
   %cmp = icmp sge i32 %b, -3
   %cmp1 = icmp sgt i32 %a, %c
   %or.cond = and i1 %cmp, %cmp1
@@ -363,22 +363,22 @@ define i32 @neg_range_int_2(i32 %a, i32 %b, i32 %c) {
 
 ; (b < -(d | 1) && a >= c)
 define i32 @neg_range_int_comp2(i32 %a, i32 %b, i32 %c, i32 %d) {
-; SDISEL-LABEL: neg_range_int_comp2:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    orr w8, w3, #0x1
-; SDISEL-NEXT:    cmp w0, w2
-; SDISEL-NEXT:    ccmn w1, w8, #0, ge
-; SDISEL-NEXT:    csel w0, w1, w0, lt
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: neg_range_int_comp2:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    orr w8, w3, #0x1
+; CHECK-SD-NEXT:    cmp w0, w2
+; CHECK-SD-NEXT:    ccmn w1, w8, #0, ge
+; CHECK-SD-NEXT:    csel w0, w1, w0, lt
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: neg_range_int_comp2:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    orr w8, w3, #0x1
-; GISEL-NEXT:    cmp w0, w2
-; GISEL-NEXT:    neg w8, w8
-; GISEL-NEXT:    ccmp w1, w8, #0, ge
-; GISEL-NEXT:    csel w0, w1, w0, lt
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: neg_range_int_comp2:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    orr w8, w3, #0x1
+; CHECK-GI-NEXT:    cmp w0, w2
+; CHECK-GI-NEXT:    neg w8, w8
+; CHECK-GI-NEXT:    ccmp w1, w8, #0, ge
+; CHECK-GI-NEXT:    csel w0, w1, w0, lt
+; CHECK-GI-NEXT:    ret
   %dor = or i32 %d, 1
   %negd = sub i32 0, %dor
   %cmp = icmp slt i32 %b, %negd
@@ -390,22 +390,22 @@ define i32 @neg_range_int_comp2(i32 %a, i32 %b, i32 %c, i32 %d) {
 
 ; (b <u -(d | 1) && a > c)
 define i32 @neg_range_int_comp_u2(i32 %a, i32 %b, i32 %c, i32 %d) {
-; SDISEL-LABEL: neg_range_int_comp_u2:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    orr w8, w3, #0x1
-; SDISEL-NEXT:    cmp w0, w2
-; SDISEL-NEXT:    ccmn w1, w8, #2, gt
-; SDISEL-NEXT:    csel w0, w1, w0, lo
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: neg_range_int_comp_u2:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    orr w8, w3, #0x1
+; CHECK-SD-NEXT:    cmp w0, w2
+; CHECK-SD-NEXT:    ccmn w1, w8, #2, gt
+; CHECK-SD-NEXT:    csel w0, w1, w0, lo
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: neg_range_int_comp_u2:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    orr w8, w3, #0x1
-; GISEL-NEXT:    cmp w0, w2
-; GISEL-NEXT:    neg w8, w8
-; GISEL-NEXT:    ccmp w1, w8, #2, gt
-; GISEL-NEXT:    csel w0, w1, w0, lo
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: neg_range_int_comp_u2:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    orr w8, w3, #0x1
+; CHECK-GI-NEXT:    cmp w0, w2
+; CHECK-GI-NEXT:    neg w8, w8
+; CHECK-GI-NEXT:    ccmp w1, w8, #2, gt
+; CHECK-GI-NEXT:    csel w0, w1, w0, lo
+; CHECK-GI-NEXT:    ret
   %dor = or i32 %d, 1
   %negd = sub i32 0, %dor
   %cmp = icmp ult i32 %b, %negd
@@ -417,22 +417,22 @@ define i32 @neg_range_int_comp_u2(i32 %a, i32 %b, i32 %c, i32 %d) {
 
 ; (b > -(d | 1) && a u > c)
 define i32 @neg_range_int_comp_ua2(i32 %a, i32 %b, i32 %c, i32 %d) {
-; SDISEL-LABEL: neg_range_int_comp_ua2:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    orr w8, w3, #0x1
-; SDISEL-NEXT:    cmp w0, w2
-; SDISEL-NEXT:    ccmn w1, w8, #4, hi
-; SDISEL-NEXT:    csel w0, w1, w0, gt
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: neg_range_int_comp_ua2:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    orr w8, w3, #0x1
+; CHECK-SD-NEXT:    cmp w0, w2
+; CHECK-SD-NEXT:    ccmn w1, w8, #4, hi
+; CHECK-SD-NEXT:    csel w0, w1, w0, gt
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: neg_range_int_comp_ua2:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    orr w8, w3, #0x1
-; GISEL-NEXT:    cmp w0, w2
-; GISEL-NEXT:    neg w8, w8
-; GISEL-NEXT:    ccmp w1, w8, #4, hi
-; GISEL-NEXT:    csel w0, w1, w0, gt
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: neg_range_int_comp_ua2:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    orr w8, w3, #0x1
+; CHECK-GI-NEXT:    cmp w0, w2
+; CHECK-GI-NEXT:    neg w8, w8
+; CHECK-GI-NEXT:    ccmp w1, w8, #4, hi
+; CHECK-GI-NEXT:    csel w0, w1, w0, gt
+; CHECK-GI-NEXT:    ret
   %dor = or i32 %d, 1
   %negd = sub i32 0, %dor
   %cmp = icmp sgt i32 %b, %negd
@@ -444,22 +444,22 @@ define i32 @neg_range_int_comp_ua2(i32 %a, i32 %b, i32 %c, i32 %d) {
 
 ; (b > -(d | 1) && a u == c)
 define i32 @neg_range_int_comp_ua3(i32 %a, i32 %b, i32 %c, i32 %d) {
-; SDISEL-LABEL: neg_range_int_comp_ua3:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    orr w8, w3, #0x1
-; SDISEL-NEXT:    cmp w0, w2
-; SDISEL-NEXT:    ccmn w1, w8, #4, eq
-; SDISEL-NEXT:    csel w0, w1, w0, gt
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: neg_range_int_comp_ua3:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    orr w8, w3, #0x1
+; CHECK-SD-NEXT:    cmp w0, w2
+; CHECK-SD-NEXT:    ccmn w1, w8, #4, eq
+; CHECK-SD-NEXT:    csel w0, w1, w0, gt
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: neg_range_int_comp_ua3:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    orr w8, w3, #0x1
-; GISEL-NEXT:    cmp w0, w2
-; GISEL-NEXT:    neg w8, w8
-; GISEL-NEXT:    ccmp w1, w8, #4, eq
-; GISEL-NEXT:    csel w0, w1, w0, gt
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: neg_range_int_comp_ua3:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    orr w8, w3, #0x1
+; CHECK-GI-NEXT:    cmp w0, w2
+; CHECK-GI-NEXT:    neg w8, w8
+; CHECK-GI-NEXT:    ccmp w1, w8, #4, eq
+; CHECK-GI-NEXT:    csel w0, w1, w0, gt
+; CHECK-GI-NEXT:    ret
   %dor = or i32 %d, 1
   %negd = sub i32 0, %dor
   %cmp = icmp sgt i32 %b, %negd
@@ -471,26 +471,26 @@ define i32 @neg_range_int_comp_ua3(i32 %a, i32 %b, i32 %c, i32 %d) {
 
 ; -(a | 1) > (b | 3) && a < c
 define i32 @neg_range_int_c(i32 %a, i32 %b, i32 %c) {
-; SDISEL-LABEL: neg_range_int_c:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    orr w8, w0, #0x1
-; SDISEL-NEXT:    orr w9, w1, #0x3
-; SDISEL-NEXT:    cmn w9, w8
-; SDISEL-NEXT:    ccmp w2, w0, #2, lo
-; SDISEL-NEXT:    cset w0, lo
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: neg_range_int_c:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    orr w8, w0, #0x1
+; CHECK-SD-NEXT:    orr w9, w1, #0x3
+; CHECK-SD-NEXT:    cmn w9, w8
+; CHECK-SD-NEXT:    ccmp w2, w0, #2, lo
+; CHECK-SD-NEXT:    cset w0, lo
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: neg_range_int_c:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    orr w8, w0, #0x1
-; GISEL-NEXT:    orr w9, w1, #0x3
-; GISEL-NEXT:    neg w8, w8
-; GISEL-NEXT:    cmp w9, w8
-; GISEL-NEXT:    cset w8, lo
-; GISEL-NEXT:    cmp w2, w0
-; GISEL-NEXT:    cset w9, lo
-; GISEL-NEXT:    and w0, w8, w9
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: neg_range_int_c:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    orr w8, w0, #0x1
+; CHECK-GI-NEXT:    orr w9, w1, #0x3
+; CHECK-GI-NEXT:    neg w8, w8
+; CHECK-GI-NEXT:    cmp w9, w8
+; CHECK-GI-NEXT:    cset w8, lo
+; CHECK-GI-NEXT:    cmp w2, w0
+; CHECK-GI-NEXT:    cset w9, lo
+; CHECK-GI-NEXT:    and w0, w8, w9
+; CHECK-GI-NEXT:    ret
 entry:
   %or = or i32 %a, 1
   %sub = sub i32 0, %or
diff --git a/llvm/test/CodeGen/AArch64/complex-deinterleaving-uniform-cases.ll b/llvm/test/CodeGen/AArch64/complex-deinterleaving-uniform-cases.ll
index 13434fa..7686740 100644
--- a/llvm/test/CodeGen/AArch64/complex-deinterleaving-uniform-cases.ll
+++ b/llvm/test/CodeGen/AArch64/complex-deinterleaving-uniform-cases.ll
@@ -203,93 +203,89 @@ define <12 x float> @abp90c12(<12 x float> %a, <12 x float> %b, <12 x float> %c)
 ; CHECK:       // %bb.0: // %entry
 ; CHECK-NEXT:    // kill: def $s1 killed $s1 def $q1
 ; CHECK-NEXT:    // kill: def $s3 killed $s3 def $q3
+; CHECK-NEXT:    ldr s17, [sp, #40]
+; CHECK-NEXT:    add x10, sp, #56
 ; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
-; CHECK-NEXT:    // kill: def $s2 killed $s2 def $q2
-; CHECK-NEXT:    ldr s17, [sp, #32]
-; CHECK-NEXT:    // kill: def $s5 killed $s5 def $q5
 ; CHECK-NEXT:    add x9, sp, #48
-; CHECK-NEXT:    add x10, sp, #64
 ; CHECK-NEXT:    mov v1.s[1], v3.s[0]
+; CHECK-NEXT:    ldr s3, [sp, #32]
+; CHECK-NEXT:    // kill: def $s2 killed $s2 def $q2
 ; CHECK-NEXT:    mov v0.s[1], v2.s[0]
+; CHECK-NEXT:    ld1 { v17.s }[1], [x10]
+; CHECK-NEXT:    // kill: def $s5 killed $s5 def $q5
+; CHECK-NEXT:    ldr s16, [sp, #8]
 ; CHECK-NEXT:    // kill: def $s4 killed $s4 def $q4
-; CHECK-NEXT:    add x11, sp, #72
-; CHECK-NEXT:    ld1 { v17.s }[1], [x9]
-; CHECK-NEXT:    ldr s18, [x10]
-; CHECK-NEXT:    add x9, sp, #80
-; CHECK-NEXT:    add x10, sp, #56
-; CHECK-NEXT:    // kill: def $s6 killed $s6 def $q6
+; CHECK-NEXT:    add x10, sp, #24
+; CHECK-NEXT:    ld1 { v3.s }[1], [x9]
+; CHECK-NEXT:    add x9, sp, #72
 ; CHECK-NEXT:    // kill: def $s7 killed $s7 def $q7
-; CHECK-NEXT:    ldr s16, [sp, #8]
-; CHECK-NEXT:    ldr s3, [sp, #96]
-; CHECK-NEXT:    ld1 { v18.s }[1], [x9]
-; CHECK-NEXT:    add x9, sp, #88
+; CHECK-NEXT:    // kill: def $s6 killed $s6 def $q6
 ; CHECK-NEXT:    ldr s2, [sp]
+; CHECK-NEXT:    ld1 { v16.s }[1], [x10]
+; CHECK-NEXT:    add x10, sp, #112
+; CHECK-NEXT:    ldr s20, [sp, #136]
 ; CHECK-NEXT:    mov v1.s[2], v5.s[0]
-; CHECK-NEXT:    ldr s5, [sp, #40]
+; CHECK-NEXT:    ld1 { v17.s }[2], [x9]
+; CHECK-NEXT:    add x9, sp, #64
+; CHECK-NEXT:    ldr s5, [sp, #96]
+; CHECK-NEXT:    ld1 { v3.s }[2], [x9]
 ; CHECK-NEXT:    mov v0.s[2], v4.s[0]
+; CHECK-NEXT:    add x9, sp, #88
+; CHECK-NEXT:    ldr s4, [sp, #104]
+; CHECK-NEXT:    ldr s19, [sp, #192]
 ; CHECK-NEXT:    ld1 { v5.s }[1], [x10]
-; CHECK-NEXT:    ldr s19, [x11]
+; CHECK-NEXT:    add x10, sp, #80
+; CHECK-NEXT:    ld1 { v17.s }[3], [x9]
+; CHECK-NEXT:    mov v1.s[3], v7.s[0]
+; CHECK-NEXT:    add x9, sp, #120
+; CHECK-NEXT:    ld1 { v3.s }[3], [x10]
+; CHECK-NEXT:    ld1 { v4.s }[1], [x9]
+; CHECK-NEXT:    ldr s7, [sp, #128]
 ; CHECK-NEXT:    add x10, sp, #144
-; CHECK-NEXT:    zip1 v4.2d, v17.2d, v18.2d
-; CHECK-NEXT:    add x11, sp, #160
-; CHECK-NEXT:    ldr s18, [sp, #136]
-; CHECK-NEXT:    ld1 { v19.s }[1], [x9]
 ; CHECK-NEXT:    mov v0.s[3], v6.s[0]
-; CHECK-NEXT:    ldr s6, [sp, #128]
-; CHECK-NEXT:    mov v1.s[3], v7.s[0]
-; CHECK-NEXT:    add x9, sp, #24
-; CHECK-NEXT:    ldr s7, [sp, #104]
-; CHECK-NEXT:    ld1 { v16.s }[1], [x9]
-; CHECK-NEXT:    add x9, sp, #112
-; CHECK-NEXT:    ld1 { v6.s }[1], [x10]
-; CHECK-NEXT:    zip1 v5.2d, v5.2d, v19.2d
-; CHECK-NEXT:    add x10, sp, #120
-; CHECK-NEXT:    ld1 { v3.s }[1], [x9]
+; CHECK-NEXT:    add x9, sp, #16
 ; CHECK-NEXT:    ld1 { v7.s }[1], [x10]
-; CHECK-NEXT:    ldr s17, [x11]
-; CHECK-NEXT:    add x9, sp, #176
-; CHECK-NEXT:    add x10, sp, #16
-; CHECK-NEXT:    add x11, sp, #168
-; CHECK-NEXT:    ld1 { v17.s }[1], [x9]
-; CHECK-NEXT:    ld1 { v2.s }[1], [x10]
-; CHECK-NEXT:    add x9, sp, #152
-; CHECK-NEXT:    fmul v19.4s, v5.4s, v1.4s
-; CHECK-NEXT:    fmul v20.4s, v7.4s, v16.4s
-; CHECK-NEXT:    fmul v16.4s, v3.4s, v16.4s
-; CHECK-NEXT:    fmul v1.4s, v4.4s, v1.4s
-; CHECK-NEXT:    ld1 { v18.s }[1], [x9]
-; CHECK-NEXT:    ldr s21, [x11]
-; CHECK-NEXT:    zip1 v6.2d, v6.2d, v17.2d
-; CHECK-NEXT:    ldr s17, [sp, #192]
-; CHECK-NEXT:    add x9, sp, #184
+; CHECK-NEXT:    ld1 { v2.s }[1], [x9]
+; CHECK-NEXT:    add x9, sp, #160
+; CHECK-NEXT:    fmul v6.4s, v17.4s, v1.4s
+; CHECK-NEXT:    fmul v18.4s, v4.4s, v16.4s
+; CHECK-NEXT:    fmul v16.4s, v5.4s, v16.4s
+; CHECK-NEXT:    fmul v1.4s, v3.4s, v1.4s
 ; CHECK-NEXT:    add x10, sp, #208
-; CHECK-NEXT:    ld1 { v21.s }[1], [x9]
+; CHECK-NEXT:    ld1 { v7.s }[2], [x9]
+; CHECK-NEXT:    add x9, sp, #152
+; CHECK-NEXT:    ld1 { v19.s }[1], [x10]
+; CHECK-NEXT:    ld1 { v20.s }[1], [x9]
+; CHECK-NEXT:    add x9, sp, #176
+; CHECK-NEXT:    add x10, sp, #184
+; CHECK-NEXT:    fneg v6.4s, v6.4s
+; CHECK-NEXT:    fneg v18.4s, v18.4s
+; CHECK-NEXT:    fmla v16.4s, v2.4s, v4.4s
+; CHECK-NEXT:    fmla v1.4s, v0.4s, v17.4s
+; CHECK-NEXT:    ld1 { v7.s }[3], [x9]
+; CHECK-NEXT:    add x9, sp, #168
+; CHECK-NEXT:    ld1 { v20.s }[2], [x9]
+; CHECK-NEXT:    ldr s4, [sp, #200]
 ; CHECK-NEXT:    add x9, sp, #216
-; CHECK-NEXT:    fneg v19.4s, v19.4s
-; CHECK-NEXT:    fneg v20.4s, v20.4s
-; CHECK-NEXT:    fmla v16.4s, v2.4s, v7.4s
-; CHECK-NEXT:    fmla v1.4s, v0.4s, v5.4s
-; CHECK-NEXT:    ld1 { v17.s }[1], [x10]
-; CHECK-NEXT:    ldr s5, [sp, #200]
-; CHECK-NEXT:    zip1 v7.2d, v18.2d, v21.2d
-; CHECK-NEXT:    ld1 { v5.s }[1], [x9]
-; CHECK-NEXT:    fmla v19.4s, v0.4s, v4.4s
-; CHECK-NEXT:    fmla v20.4s, v2.4s, v3.4s
-; CHECK-NEXT:    fsub v0.4s, v6.4s, v1.4s
-; CHECK-NEXT:    fsub v1.4s, v17.4s, v16.4s
-; CHECK-NEXT:    fadd v2.4s, v7.4s, v19.4s
-; CHECK-NEXT:    fadd v3.4s, v5.4s, v20.4s
+; CHECK-NEXT:    fmla v6.4s, v0.4s, v3.4s
+; CHECK-NEXT:    fmla v18.4s, v2.4s, v5.4s
+; CHECK-NEXT:    ld1 { v4.s }[1], [x9]
+; CHECK-NEXT:    fsub v0.4s, v7.4s, v1.4s
+; CHECK-NEXT:    fsub v1.4s, v19.4s, v16.4s
+; CHECK-NEXT:    ld1 { v20.s }[3], [x10]
+; CHECK-NEXT:    fadd v2.4s, v4.4s, v18.4s
+; CHECK-NEXT:    fadd v3.4s, v20.4s, v6.4s
 ; CHECK-NEXT:    ext v4.16b, v0.16b, v1.16b, #12
-; CHECK-NEXT:    ext v5.16b, v2.16b, v3.16b, #12
-; CHECK-NEXT:    trn2 v1.4s, v1.4s, v3.4s
+; CHECK-NEXT:    ext v5.16b, v3.16b, v2.16b, #12
+; CHECK-NEXT:    trn2 v1.4s, v1.4s, v2.4s
 ; CHECK-NEXT:    ext v4.16b, v0.16b, v4.16b, #12
-; CHECK-NEXT:    ext v5.16b, v2.16b, v5.16b, #8
+; CHECK-NEXT:    ext v5.16b, v3.16b, v5.16b, #8
 ; CHECK-NEXT:    rev64 v4.4s, v4.4s
-; CHECK-NEXT:    trn2 v3.4s, v4.4s, v5.4s
-; CHECK-NEXT:    zip2 v4.4s, v0.4s, v2.4s
-; CHECK-NEXT:    zip1 v0.4s, v0.4s, v2.4s
-; CHECK-NEXT:    ext v1.16b, v3.16b, v1.16b, #8
-; CHECK-NEXT:    mov v4.d[1], v3.d[0]
+; CHECK-NEXT:    trn2 v2.4s, v4.4s, v5.4s
+; CHECK-NEXT:    zip2 v4.4s, v0.4s, v3.4s
+; CHECK-NEXT:    zip1 v0.4s, v0.4s, v3.4s
+; CHECK-NEXT:    ext v1.16b, v2.16b, v1.16b, #8
+; CHECK-NEXT:    mov v4.d[1], v2.d[0]
 ; CHECK-NEXT:    str q0, [x8]
 ; CHECK-NEXT:    stp q4, q1, [x8, #16]
 ; CHECK-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/concat-vector.ll b/llvm/test/CodeGen/AArch64/concat-vector.ll
index e6f27b9..acf15f1 100644
--- a/llvm/test/CodeGen/AArch64/concat-vector.ll
+++ b/llvm/test/CodeGen/AArch64/concat-vector.ll
@@ -186,9 +186,8 @@ define <16 x i8> @concat_v16s8_v4s8_load(ptr %ptrA, ptr %ptrB, ptr %ptrC, ptr %p
 ; CHECK:       // %bb.0:
 ; CHECK-NEXT:    ldr s0, [x0]
 ; CHECK-NEXT:    ld1 { v0.s }[1], [x1]
-; CHECK-NEXT:    ldr s1, [x2]
-; CHECK-NEXT:    ld1 { v1.s }[1], [x3]
-; CHECK-NEXT:    zip1 v0.2d, v0.2d, v1.2d
+; CHECK-NEXT:    ld1 { v0.s }[2], [x2]
+; CHECK-NEXT:    ld1 { v0.s }[3], [x3]
 ; CHECK-NEXT:    ret
     %A = load <4 x i8>, ptr %ptrA
     %B = load <4 x i8>, ptr %ptrB
diff --git a/llvm/test/CodeGen/AArch64/dag-combine-select.ll b/llvm/test/CodeGen/AArch64/dag-combine-select.ll
index 56208f1..02b0077 100644
--- a/llvm/test/CodeGen/AArch64/dag-combine-select.ll
+++ b/llvm/test/CodeGen/AArch64/dag-combine-select.ll
@@ -1,26 +1,26 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc -mtriple arm64-none-eabi -o - %s | FileCheck %s --check-prefixes=CHECK,SDISEL
-; RUN: llc -mtriple arm64-none-eabi -global-isel -o - %s | FileCheck %s --check-prefixes=CHECK,GISEL
+; RUN: llc -mtriple arm64-none-eabi -o - %s | FileCheck %s --check-prefixes=CHECK,CHECK-SD
+; RUN: llc -mtriple arm64-none-eabi -global-isel -o - %s | FileCheck %s --check-prefixes=CHECK,CHECK-GI
 
 @out = internal global i32 0, align 4
 
 ; Ensure that we transform select(C0, x, select(C1, x, y)) towards
 ; select(C0 | C1, x, y) so we can use CMP;CCMP for the implementation.
 define i32 @test0(i32 %v0, i32 %v1, i32 %v2) {
-; SDISEL-LABEL: test0:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    cmp w0, #7
-; SDISEL-NEXT:    ccmp w1, #0, #0, ne
-; SDISEL-NEXT:    csel w0, w1, w2, gt
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: test0:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    cmp w0, #7
+; CHECK-SD-NEXT:    ccmp w1, #0, #0, ne
+; CHECK-SD-NEXT:    csel w0, w1, w2, gt
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: test0:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    cmp w0, #7
-; GISEL-NEXT:    csel w8, w1, w2, eq
-; GISEL-NEXT:    cmp w1, #0
-; GISEL-NEXT:    csel w0, w1, w8, gt
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: test0:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    cmp w0, #7
+; CHECK-GI-NEXT:    csel w8, w1, w2, eq
+; CHECK-GI-NEXT:    cmp w1, #0
+; CHECK-GI-NEXT:    csel w0, w1, w8, gt
+; CHECK-GI-NEXT:    ret
   %cmp1 = icmp eq i32 %v0, 7
   %cmp2 = icmp sgt i32 %v1, 0
   %sel0 = select i1 %cmp1, i32 %v1, i32 %v2
@@ -32,36 +32,36 @@ define i32 @test0(i32 %v0, i32 %v1, i32 %v2) {
 ; sequences. This case should be transformed to select(C0, select(C1, x, y), y)
 ; anyway to get CSE effects.
 define void @test1(i32 %bitset, i32 %val0, i32 %val1) {
-; SDISEL-LABEL: test1:
-; SDISEL:       // %bb.0:
-; SDISEL-NEXT:    cmp w0, #7
-; SDISEL-NEXT:    adrp x9, out
-; SDISEL-NEXT:    csel w8, w1, w2, eq
-; SDISEL-NEXT:    cmp w8, #13
-; SDISEL-NEXT:    csel w8, w1, w2, lo
-; SDISEL-NEXT:    cmp w0, #42
-; SDISEL-NEXT:    csel w10, w1, w8, eq
-; SDISEL-NEXT:    str w8, [x9, :lo12:out]
-; SDISEL-NEXT:    str w10, [x9, :lo12:out]
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: test1:
+; CHECK-SD:       // %bb.0:
+; CHECK-SD-NEXT:    cmp w0, #7
+; CHECK-SD-NEXT:    adrp x9, out
+; CHECK-SD-NEXT:    csel w8, w1, w2, eq
+; CHECK-SD-NEXT:    cmp w8, #13
+; CHECK-SD-NEXT:    csel w8, w1, w2, lo
+; CHECK-SD-NEXT:    cmp w0, #42
+; CHECK-SD-NEXT:    csel w10, w1, w8, eq
+; CHECK-SD-NEXT:    str w8, [x9, :lo12:out]
+; CHECK-SD-NEXT:    str w10, [x9, :lo12:out]
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: test1:
-; GISEL:       // %bb.0:
-; GISEL-NEXT:    cmp w0, #7
-; GISEL-NEXT:    csel w8, w1, w2, eq
-; GISEL-NEXT:    cmp w8, #13
-; GISEL-NEXT:    cset w8, lo
-; GISEL-NEXT:    tst w8, #0x1
-; GISEL-NEXT:    csel w9, w1, w2, ne
-; GISEL-NEXT:    cmp w0, #42
-; GISEL-NEXT:    cset w10, eq
-; GISEL-NEXT:    orr w8, w10, w8
-; GISEL-NEXT:    tst w8, #0x1
-; GISEL-NEXT:    adrp x8, out
-; GISEL-NEXT:    csel w10, w1, w2, ne
-; GISEL-NEXT:    str w9, [x8, :lo12:out]
-; GISEL-NEXT:    str w10, [x8, :lo12:out]
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: test1:
+; CHECK-GI:       // %bb.0:
+; CHECK-GI-NEXT:    cmp w0, #7
+; CHECK-GI-NEXT:    csel w8, w1, w2, eq
+; CHECK-GI-NEXT:    cmp w8, #13
+; CHECK-GI-NEXT:    cset w8, lo
+; CHECK-GI-NEXT:    tst w8, #0x1
+; CHECK-GI-NEXT:    csel w9, w1, w2, ne
+; CHECK-GI-NEXT:    cmp w0, #42
+; CHECK-GI-NEXT:    cset w10, eq
+; CHECK-GI-NEXT:    orr w8, w10, w8
+; CHECK-GI-NEXT:    tst w8, #0x1
+; CHECK-GI-NEXT:    adrp x8, out
+; CHECK-GI-NEXT:    csel w10, w1, w2, ne
+; CHECK-GI-NEXT:    str w9, [x8, :lo12:out]
+; CHECK-GI-NEXT:    str w10, [x8, :lo12:out]
+; CHECK-GI-NEXT:    ret
   %cmp1 = icmp eq i32 %bitset, 7
   %cond = select i1 %cmp1, i32 %val0, i32 %val1
   %cmp5 = icmp ult i32 %cond, 13
diff --git a/llvm/test/CodeGen/AArch64/fp-maximumnum-minimumnum.ll b/llvm/test/CodeGen/AArch64/fp-maximumnum-minimumnum.ll
index 4906e2e..c6b8e41 100644
--- a/llvm/test/CodeGen/AArch64/fp-maximumnum-minimumnum.ll
+++ b/llvm/test/CodeGen/AArch64/fp-maximumnum-minimumnum.ll
@@ -1431,7 +1431,6 @@ define <9 x half> @max_v9f16(<9 x half> %a, <9 x half> %b) {
 ; FULLFP16-NEXT:    add x9, sp, #16
 ; FULLFP16-NEXT:    // kill: def $h3 killed $h3 def $q3
 ; FULLFP16-NEXT:    // kill: def $h4 killed $h4 def $q4
-; FULLFP16-NEXT:    add x10, sp, #40
 ; FULLFP16-NEXT:    // kill: def $h5 killed $h5 def $q5
 ; FULLFP16-NEXT:    // kill: def $h6 killed $h6 def $q6
 ; FULLFP16-NEXT:    // kill: def $h7 killed $h7 def $q7
@@ -1440,30 +1439,30 @@ define <9 x half> @max_v9f16(<9 x half> %a, <9 x half> %b) {
 ; FULLFP16-NEXT:    ld1 { v1.h }[1], [x9]
 ; FULLFP16-NEXT:    add x9, sp, #24
 ; FULLFP16-NEXT:    mov v0.h[2], v2.h[0]
+; FULLFP16-NEXT:    ldr h2, [sp]
 ; FULLFP16-NEXT:    ld1 { v1.h }[2], [x9]
 ; FULLFP16-NEXT:    add x9, sp, #32
+; FULLFP16-NEXT:    fminnm v2.8h, v2.8h, v2.8h
 ; FULLFP16-NEXT:    mov v0.h[3], v3.h[0]
 ; FULLFP16-NEXT:    ld1 { v1.h }[3], [x9]
-; FULLFP16-NEXT:    ldr h2, [x10]
-; FULLFP16-NEXT:    add x9, sp, #48
+; FULLFP16-NEXT:    add x9, sp, #40
 ; FULLFP16-NEXT:    ldr h3, [sp, #72]
-; FULLFP16-NEXT:    ld1 { v2.h }[1], [x9]
-; FULLFP16-NEXT:    add x9, sp, #56
+; FULLFP16-NEXT:    ld1 { v1.h }[4], [x9]
+; FULLFP16-NEXT:    add x9, sp, #48
 ; FULLFP16-NEXT:    fminnm v3.8h, v3.8h, v3.8h
 ; FULLFP16-NEXT:    mov v0.h[4], v4.h[0]
-; FULLFP16-NEXT:    ld1 { v2.h }[2], [x9]
-; FULLFP16-NEXT:    add x9, sp, #64
+; FULLFP16-NEXT:    ld1 { v1.h }[5], [x9]
+; FULLFP16-NEXT:    add x9, sp, #56
+; FULLFP16-NEXT:    fmaxnm v2.8h, v2.8h, v3.8h
 ; FULLFP16-NEXT:    mov v0.h[5], v5.h[0]
-; FULLFP16-NEXT:    ld1 { v2.h }[3], [x9]
-; FULLFP16-NEXT:    zip1 v1.2d, v1.2d, v2.2d
-; FULLFP16-NEXT:    ldr h2, [sp]
+; FULLFP16-NEXT:    ld1 { v1.h }[6], [x9]
+; FULLFP16-NEXT:    add x9, sp, #64
+; FULLFP16-NEXT:    str h2, [x8, #16]
 ; FULLFP16-NEXT:    mov v0.h[6], v6.h[0]
-; FULLFP16-NEXT:    fminnm v2.8h, v2.8h, v2.8h
+; FULLFP16-NEXT:    ld1 { v1.h }[7], [x9]
 ; FULLFP16-NEXT:    fminnm v1.8h, v1.8h, v1.8h
 ; FULLFP16-NEXT:    mov v0.h[7], v7.h[0]
-; FULLFP16-NEXT:    fmaxnm v2.8h, v2.8h, v3.8h
 ; FULLFP16-NEXT:    fminnm v0.8h, v0.8h, v0.8h
-; FULLFP16-NEXT:    str h2, [x8, #16]
 ; FULLFP16-NEXT:    fmaxnm v0.8h, v0.8h, v1.8h
 ; FULLFP16-NEXT:    str q0, [x8]
 ; FULLFP16-NEXT:    ret
@@ -2013,7 +2012,6 @@ define <9 x half> @min_v9f16(<9 x half> %a, <9 x half> %b) {
 ; FULLFP16-NEXT:    add x9, sp, #16
 ; FULLFP16-NEXT:    // kill: def $h3 killed $h3 def $q3
 ; FULLFP16-NEXT:    // kill: def $h4 killed $h4 def $q4
-; FULLFP16-NEXT:    add x10, sp, #40
 ; FULLFP16-NEXT:    // kill: def $h5 killed $h5 def $q5
 ; FULLFP16-NEXT:    // kill: def $h6 killed $h6 def $q6
 ; FULLFP16-NEXT:    // kill: def $h7 killed $h7 def $q7
@@ -2022,30 +2020,30 @@ define <9 x half> @min_v9f16(<9 x half> %a, <9 x half> %b) {
 ; FULLFP16-NEXT:    ld1 { v1.h }[1], [x9]
 ; FULLFP16-NEXT:    add x9, sp, #24
 ; FULLFP16-NEXT:    mov v0.h[2], v2.h[0]
+; FULLFP16-NEXT:    ldr h2, [sp]
 ; FULLFP16-NEXT:    ld1 { v1.h }[2], [x9]
 ; FULLFP16-NEXT:    add x9, sp, #32
+; FULLFP16-NEXT:    fminnm v2.8h, v2.8h, v2.8h
 ; FULLFP16-NEXT:    mov v0.h[3], v3.h[0]
 ; FULLFP16-NEXT:    ld1 { v1.h }[3], [x9]
-; FULLFP16-NEXT:    ldr h2, [x10]
-; FULLFP16-NEXT:    add x9, sp, #48
+; FULLFP16-NEXT:    add x9, sp, #40
 ; FULLFP16-NEXT:    ldr h3, [sp, #72]
-; FULLFP16-NEXT:    ld1 { v2.h }[1], [x9]
-; FULLFP16-NEXT:    add x9, sp, #56
+; FULLFP16-NEXT:    ld1 { v1.h }[4], [x9]
+; FULLFP16-NEXT:    add x9, sp, #48
 ; FULLFP16-NEXT:    fminnm v3.8h, v3.8h, v3.8h
 ; FULLFP16-NEXT:    mov v0.h[4], v4.h[0]
-; FULLFP16-NEXT:    ld1 { v2.h }[2], [x9]
-; FULLFP16-NEXT:    add x9, sp, #64
+; FULLFP16-NEXT:    ld1 { v1.h }[5], [x9]
+; FULLFP16-NEXT:    add x9, sp, #56
+; FULLFP16-NEXT:    fminnm v2.8h, v2.8h, v3.8h
 ; FULLFP16-NEXT:    mov v0.h[5], v5.h[0]
-; FULLFP16-NEXT:    ld1 { v2.h }[3], [x9]
-; FULLFP16-NEXT:    zip1 v1.2d, v1.2d, v2.2d
-; FULLFP16-NEXT:    ldr h2, [sp]
+; FULLFP16-NEXT:    ld1 { v1.h }[6], [x9]
+; FULLFP16-NEXT:    add x9, sp, #64
+; FULLFP16-NEXT:    str h2, [x8, #16]
 ; FULLFP16-NEXT:    mov v0.h[6], v6.h[0]
-; FULLFP16-NEXT:    fminnm v2.8h, v2.8h, v2.8h
+; FULLFP16-NEXT:    ld1 { v1.h }[7], [x9]
 ; FULLFP16-NEXT:    fminnm v1.8h, v1.8h, v1.8h
 ; FULLFP16-NEXT:    mov v0.h[7], v7.h[0]
-; FULLFP16-NEXT:    fminnm v2.8h, v2.8h, v3.8h
 ; FULLFP16-NEXT:    fminnm v0.8h, v0.8h, v0.8h
-; FULLFP16-NEXT:    str h2, [x8, #16]
 ; FULLFP16-NEXT:    fminnm v0.8h, v0.8h, v1.8h
 ; FULLFP16-NEXT:    str q0, [x8]
 ; FULLFP16-NEXT:    ret
diff --git a/llvm/test/CodeGen/AArch64/fp16_intrinsic_scalar_1op.ll b/llvm/test/CodeGen/AArch64/fp16_intrinsic_scalar_1op.ll
index 1b98954..b056460 100644
--- a/llvm/test/CodeGen/AArch64/fp16_intrinsic_scalar_1op.ll
+++ b/llvm/test/CodeGen/AArch64/fp16_intrinsic_scalar_1op.ll
@@ -1,6 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=aarch64 -global-isel=0 -mattr=+v8.2a,+fullfp16  | FileCheck %s --check-prefixes=CHECK,SDISEL
-; RUN: llc < %s -mtriple=aarch64 -global-isel=1 -mattr=+v8.2a,+fullfp16  | FileCheck %s --check-prefixes=CHECK,GISEL
+; RUN: llc < %s -mtriple=aarch64 -global-isel=0 -mattr=+v8.2a,+fullfp16  | FileCheck %s --check-prefixes=CHECK,CHECK-SD
+; RUN: llc < %s -mtriple=aarch64 -global-isel=1 -mattr=+v8.2a,+fullfp16  | FileCheck %s --check-prefixes=CHECK,CHECK-GI
 
 declare i64 @llvm.aarch64.neon.fcvtpu.i64.f16(half)
 declare i32 @llvm.aarch64.neon.fcvtpu.i32.f16(half)
@@ -27,18 +27,18 @@ declare half @llvm.aarch64.neon.frecpx.f16(half)
 declare half @llvm.aarch64.neon.frecpe.f16(half)
 
 define dso_local i16 @t2(half %a) {
-; SDISEL-LABEL: t2:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    fcmp h0, #0.0
-; SDISEL-NEXT:    csetm w0, eq
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: t2:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    fcmp h0, #0.0
+; CHECK-SD-NEXT:    csetm w0, eq
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: t2:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    fcmp h0, #0.0
-; GISEL-NEXT:    cset w8, eq
-; GISEL-NEXT:    sbfx w0, w8, #0, #1
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: t2:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    fcmp h0, #0.0
+; CHECK-GI-NEXT:    cset w8, eq
+; CHECK-GI-NEXT:    sbfx w0, w8, #0, #1
+; CHECK-GI-NEXT:    ret
 entry:
   %0 = fcmp oeq half %a, 0xH0000
   %vceqz = sext i1 %0 to i16
@@ -46,18 +46,18 @@ entry:
 }
 
 define dso_local i16 @t3(half %a) {
-; SDISEL-LABEL: t3:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    fcmp h0, #0.0
-; SDISEL-NEXT:    csetm w0, ge
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: t3:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    fcmp h0, #0.0
+; CHECK-SD-NEXT:    csetm w0, ge
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: t3:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    fcmp h0, #0.0
-; GISEL-NEXT:    cset w8, ge
-; GISEL-NEXT:    sbfx w0, w8, #0, #1
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: t3:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    fcmp h0, #0.0
+; CHECK-GI-NEXT:    cset w8, ge
+; CHECK-GI-NEXT:    sbfx w0, w8, #0, #1
+; CHECK-GI-NEXT:    ret
 entry:
   %0 = fcmp oge half %a, 0xH0000
   %vcgez = sext i1 %0 to i16
@@ -65,18 +65,18 @@ entry:
 }
 
 define dso_local i16 @t4(half %a) {
-; SDISEL-LABEL: t4:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    fcmp h0, #0.0
-; SDISEL-NEXT:    csetm w0, gt
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: t4:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    fcmp h0, #0.0
+; CHECK-SD-NEXT:    csetm w0, gt
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: t4:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    fcmp h0, #0.0
-; GISEL-NEXT:    cset w8, gt
-; GISEL-NEXT:    sbfx w0, w8, #0, #1
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: t4:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    fcmp h0, #0.0
+; CHECK-GI-NEXT:    cset w8, gt
+; CHECK-GI-NEXT:    sbfx w0, w8, #0, #1
+; CHECK-GI-NEXT:    ret
 entry:
   %0 = fcmp ogt half %a, 0xH0000
   %vcgtz = sext i1 %0 to i16
@@ -84,18 +84,18 @@ entry:
 }
 
 define dso_local i16 @t5(half %a) {
-; SDISEL-LABEL: t5:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    fcmp h0, #0.0
-; SDISEL-NEXT:    csetm w0, ls
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: t5:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    fcmp h0, #0.0
+; CHECK-SD-NEXT:    csetm w0, ls
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: t5:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    fcmp h0, #0.0
-; GISEL-NEXT:    cset w8, ls
-; GISEL-NEXT:    sbfx w0, w8, #0, #1
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: t5:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    fcmp h0, #0.0
+; CHECK-GI-NEXT:    cset w8, ls
+; CHECK-GI-NEXT:    sbfx w0, w8, #0, #1
+; CHECK-GI-NEXT:    ret
 entry:
   %0 = fcmp ole half %a, 0xH0000
   %vclez = sext i1 %0 to i16
@@ -103,18 +103,18 @@ entry:
 }
 
 define dso_local i16 @t6(half %a) {
-; SDISEL-LABEL: t6:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    fcmp h0, #0.0
-; SDISEL-NEXT:    csetm w0, mi
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: t6:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    fcmp h0, #0.0
+; CHECK-SD-NEXT:    csetm w0, mi
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: t6:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    fcmp h0, #0.0
-; GISEL-NEXT:    cset w8, mi
-; GISEL-NEXT:    sbfx w0, w8, #0, #1
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: t6:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    fcmp h0, #0.0
+; CHECK-GI-NEXT:    cset w8, mi
+; CHECK-GI-NEXT:    sbfx w0, w8, #0, #1
+; CHECK-GI-NEXT:    ret
 entry:
   %0 = fcmp olt half %a, 0xH0000
   %vcltz = sext i1 %0 to i16
@@ -172,15 +172,15 @@ entry:
 }
 
 define dso_local i16 @t16(half %a) {
-; SDISEL-LABEL: t16:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    fcvtzs w0, h0
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: t16:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    fcvtzs w0, h0
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: t16:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    fcvtzu w0, h0
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: t16:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    fcvtzu w0, h0
+; CHECK-GI-NEXT:    ret
 entry:
   %0 = fptoui half %a to i16
   ret i16 %0
diff --git a/llvm/test/CodeGen/AArch64/fp16_intrinsic_scalar_2op.ll b/llvm/test/CodeGen/AArch64/fp16_intrinsic_scalar_2op.ll
index 5b08ef2..da70599 100644
--- a/llvm/test/CodeGen/AArch64/fp16_intrinsic_scalar_2op.ll
+++ b/llvm/test/CodeGen/AArch64/fp16_intrinsic_scalar_2op.ll
@@ -1,6 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
-; RUN: llc < %s -mtriple=aarch64 -global-isel=0 -mattr=+v8.2a,+fullfp16  | FileCheck %s --check-prefixes=CHECK,SDISEL
-; RUN: llc < %s -mtriple=aarch64 -global-isel=1 -mattr=+v8.2a,+fullfp16  | FileCheck %s --check-prefixes=CHECK,GISEL
+; RUN: llc < %s -mtriple=aarch64 -global-isel=0 -mattr=+v8.2a,+fullfp16  | FileCheck %s --check-prefixes=CHECK,CHECK-SD
+; RUN: llc < %s -mtriple=aarch64 -global-isel=1 -mattr=+v8.2a,+fullfp16  | FileCheck %s --check-prefixes=CHECK,CHECK-GI
 
 
 declare half @llvm.aarch64.sisd.fabd.f16(half, half)
@@ -35,18 +35,18 @@ entry:
 }
 
 define dso_local i16 @t_vceqh_f16(half %a, half %b) {
-; SDISEL-LABEL: t_vceqh_f16:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    fcmp h0, h1
-; SDISEL-NEXT:    csetm w0, eq
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: t_vceqh_f16:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    fcmp h0, h1
+; CHECK-SD-NEXT:    csetm w0, eq
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: t_vceqh_f16:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    fcmp h0, h1
-; GISEL-NEXT:    cset w8, eq
-; GISEL-NEXT:    sbfx w0, w8, #0, #1
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: t_vceqh_f16:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    fcmp h0, h1
+; CHECK-GI-NEXT:    cset w8, eq
+; CHECK-GI-NEXT:    sbfx w0, w8, #0, #1
+; CHECK-GI-NEXT:    ret
 entry:
   %0 = fcmp oeq half %a, %b
   %vcmpd = sext i1 %0 to i16
@@ -54,18 +54,18 @@ entry:
 }
 
 define dso_local i16 @t_vcgeh_f16(half %a, half %b) {
-; SDISEL-LABEL: t_vcgeh_f16:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    fcmp h0, h1
-; SDISEL-NEXT:    csetm w0, ge
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: t_vcgeh_f16:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    fcmp h0, h1
+; CHECK-SD-NEXT:    csetm w0, ge
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: t_vcgeh_f16:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    fcmp h0, h1
-; GISEL-NEXT:    cset w8, ge
-; GISEL-NEXT:    sbfx w0, w8, #0, #1
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: t_vcgeh_f16:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    fcmp h0, h1
+; CHECK-GI-NEXT:    cset w8, ge
+; CHECK-GI-NEXT:    sbfx w0, w8, #0, #1
+; CHECK-GI-NEXT:    ret
 entry:
   %0 = fcmp oge half %a, %b
   %vcmpd = sext i1 %0 to i16
@@ -73,18 +73,18 @@ entry:
 }
 
 define dso_local i16 @t_vcgth_f16(half %a, half %b) {
-; SDISEL-LABEL: t_vcgth_f16:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    fcmp h0, h1
-; SDISEL-NEXT:    csetm w0, gt
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: t_vcgth_f16:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    fcmp h0, h1
+; CHECK-SD-NEXT:    csetm w0, gt
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: t_vcgth_f16:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    fcmp h0, h1
-; GISEL-NEXT:    cset w8, gt
-; GISEL-NEXT:    sbfx w0, w8, #0, #1
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: t_vcgth_f16:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    fcmp h0, h1
+; CHECK-GI-NEXT:    cset w8, gt
+; CHECK-GI-NEXT:    sbfx w0, w8, #0, #1
+; CHECK-GI-NEXT:    ret
 entry:
   %0 = fcmp ogt half %a, %b
   %vcmpd = sext i1 %0 to i16
@@ -92,18 +92,18 @@ entry:
 }
 
 define dso_local i16 @t_vcleh_f16(half %a, half %b) {
-; SDISEL-LABEL: t_vcleh_f16:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    fcmp h0, h1
-; SDISEL-NEXT:    csetm w0, ls
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: t_vcleh_f16:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    fcmp h0, h1
+; CHECK-SD-NEXT:    csetm w0, ls
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: t_vcleh_f16:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    fcmp h0, h1
-; GISEL-NEXT:    cset w8, ls
-; GISEL-NEXT:    sbfx w0, w8, #0, #1
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: t_vcleh_f16:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    fcmp h0, h1
+; CHECK-GI-NEXT:    cset w8, ls
+; CHECK-GI-NEXT:    sbfx w0, w8, #0, #1
+; CHECK-GI-NEXT:    ret
 entry:
   %0 = fcmp ole half %a, %b
   %vcmpd = sext i1 %0 to i16
@@ -111,18 +111,18 @@ entry:
 }
 
 define dso_local i16 @t_vclth_f16(half %a, half %b) {
-; SDISEL-LABEL: t_vclth_f16:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    fcmp h0, h1
-; SDISEL-NEXT:    csetm w0, mi
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: t_vclth_f16:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    fcmp h0, h1
+; CHECK-SD-NEXT:    csetm w0, mi
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: t_vclth_f16:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    fcmp h0, h1
-; GISEL-NEXT:    cset w8, mi
-; GISEL-NEXT:    sbfx w0, w8, #0, #1
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: t_vclth_f16:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    fcmp h0, h1
+; CHECK-GI-NEXT:    cset w8, mi
+; CHECK-GI-NEXT:    sbfx w0, w8, #0, #1
+; CHECK-GI-NEXT:    ret
 entry:
   %0 = fcmp olt half %a, %b
   %vcmpd = sext i1 %0 to i16
@@ -187,18 +187,18 @@ declare half @llvm.aarch64.neon.vcvtfxu2fp.f16.i32(i32, i32) #1
 declare i32 @llvm.aarch64.neon.vcvtfp2fxu.i32.f16(half, i32) #1
 
 define dso_local half @test_vcvth_n_f16_s16_1(i16 %a) {
-; SDISEL-LABEL: test_vcvth_n_f16_s16_1:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    fmov s0, w0
-; SDISEL-NEXT:    scvtf h0, h0, #1
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: test_vcvth_n_f16_s16_1:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    fmov s0, w0
+; CHECK-SD-NEXT:    scvtf h0, h0, #1
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: test_vcvth_n_f16_s16_1:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    sxth w8, w0
-; GISEL-NEXT:    fmov s0, w8
-; GISEL-NEXT:    scvtf h0, h0, #1
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: test_vcvth_n_f16_s16_1:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    sxth w8, w0
+; CHECK-GI-NEXT:    fmov s0, w8
+; CHECK-GI-NEXT:    scvtf h0, h0, #1
+; CHECK-GI-NEXT:    ret
 entry:
   %sext = sext i16 %a to i32
   %fcvth_n = tail call half @llvm.aarch64.neon.vcvtfxs2fp.f16.i32(i32 %sext, i32 1)
@@ -206,18 +206,18 @@ entry:
 }
 
 define dso_local half @test_vcvth_n_f16_s16_16(i16 %a) {
-; SDISEL-LABEL: test_vcvth_n_f16_s16_16:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    fmov s0, w0
-; SDISEL-NEXT:    scvtf h0, h0, #16
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: test_vcvth_n_f16_s16_16:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    fmov s0, w0
+; CHECK-SD-NEXT:    scvtf h0, h0, #16
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: test_vcvth_n_f16_s16_16:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    sxth w8, w0
-; GISEL-NEXT:    fmov s0, w8
-; GISEL-NEXT:    scvtf h0, h0, #16
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: test_vcvth_n_f16_s16_16:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    sxth w8, w0
+; CHECK-GI-NEXT:    fmov s0, w8
+; CHECK-GI-NEXT:    scvtf h0, h0, #16
+; CHECK-GI-NEXT:    ret
 entry:
   %sext = sext i16 %a to i32
   %fcvth_n = tail call half @llvm.aarch64.neon.vcvtfxs2fp.f16.i32(i32 %sext, i32 16)
@@ -315,18 +315,18 @@ entry:
 }
 
 define dso_local half @test_vcvth_n_f16_u16_1(i16 %a) {
-; SDISEL-LABEL: test_vcvth_n_f16_u16_1:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    fmov s0, w0
-; SDISEL-NEXT:    ucvtf h0, h0, #1
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: test_vcvth_n_f16_u16_1:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    fmov s0, w0
+; CHECK-SD-NEXT:    ucvtf h0, h0, #1
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: test_vcvth_n_f16_u16_1:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    and w8, w0, #0xffff
-; GISEL-NEXT:    fmov s0, w8
-; GISEL-NEXT:    ucvtf h0, h0, #1
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: test_vcvth_n_f16_u16_1:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    and w8, w0, #0xffff
+; CHECK-GI-NEXT:    fmov s0, w8
+; CHECK-GI-NEXT:    ucvtf h0, h0, #1
+; CHECK-GI-NEXT:    ret
 entry:
   %0 = zext i16 %a to i32
   %fcvth_n = tail call half @llvm.aarch64.neon.vcvtfxu2fp.f16.i32(i32 %0, i32 1)
@@ -334,18 +334,18 @@ entry:
 }
 
 define dso_local half @test_vcvth_n_f16_u16_16(i16 %a) {
-; SDISEL-LABEL: test_vcvth_n_f16_u16_16:
-; SDISEL:       // %bb.0: // %entry
-; SDISEL-NEXT:    fmov s0, w0
-; SDISEL-NEXT:    ucvtf h0, h0, #16
-; SDISEL-NEXT:    ret
+; CHECK-SD-LABEL: test_vcvth_n_f16_u16_16:
+; CHECK-SD:       // %bb.0: // %entry
+; CHECK-SD-NEXT:    fmov s0, w0
+; CHECK-SD-NEXT:    ucvtf h0, h0, #16
+; CHECK-SD-NEXT:    ret
 ;
-; GISEL-LABEL: test_vcvth_n_f16_u16_16:
-; GISEL:       // %bb.0: // %entry
-; GISEL-NEXT:    and w8, w0, #0xffff
-; GISEL-NEXT:    fmov s0, w8
-; GISEL-NEXT:    ucvtf h0, h0, #16
-; GISEL-NEXT:    ret
+; CHECK-GI-LABEL: test_vcvth_n_f16_u16_16:
+; CHECK-GI:       // %bb.0: // %entry
+; CHECK-GI-NEXT:    and w8, w0, #0xffff
+; CHECK-GI-NEXT:    fmov s0, w8
+; CHECK-GI-NEXT:    ucvtf h0, h0, #16
+; CHECK-GI-NEXT:    ret
 entry:
   %0 = zext i16 %a to i32
   %fcvth_n = tail call half @llvm.aarch64.neon.vcvtfxu2fp.f16.i32(i32 %0, i32 16)
diff --git a/llvm/test/CodeGen/AArch64/fsh.ll b/llvm/test/CodeGen/AArch64/fsh.ll
index ae2ef26..4c28c90 100644
--- a/llvm/test/CodeGen/AArch64/fsh.ll
+++ b/llvm/test/CodeGen/AArch64/fsh.ll
@@ -2509,88 +2509,87 @@ define <7 x i32> @fshl_v7i32(<7 x i32> %a, <7 x i32> %b, <7 x i32> %c) {
 ;
 ; CHECK-GI-LABEL: fshl_v7i32:
 ; CHECK-GI:       // %bb.0: // %entry
-; CHECK-GI-NEXT:    ldr s17, [sp, #48]
-; CHECK-GI-NEXT:    add x8, sp, #56
-; CHECK-GI-NEXT:    add x9, sp, #64
+; CHECK-GI-NEXT:    ldr s3, [sp, #48]
+; CHECK-GI-NEXT:    ldr s20, [sp, #56]
+; CHECK-GI-NEXT:    add x9, sp, #56
 ; CHECK-GI-NEXT:    ldr s4, [sp, #48]
-; CHECK-GI-NEXT:    ldr s21, [sp, #56]
-; CHECK-GI-NEXT:    mov w10, #-1 // =0xffffffff
-; CHECK-GI-NEXT:    ld1 { v17.s }[1], [x8]
-; CHECK-GI-NEXT:    ldr s20, [x9]
-; CHECK-GI-NEXT:    add x8, sp, #72
-; CHECK-GI-NEXT:    mov v4.s[1], v21.s[0]
+; CHECK-GI-NEXT:    ldr s7, [sp, #80]
+; CHECK-GI-NEXT:    mov w12, #-1 // =0xffffffff
+; CHECK-GI-NEXT:    ldr s21, [sp, #88]
+; CHECK-GI-NEXT:    mov v3.s[1], v20.s[0]
+; CHECK-GI-NEXT:    fmov s20, w12
+; CHECK-GI-NEXT:    ld1 { v4.s }[1], [x9]
+; CHECK-GI-NEXT:    ldr s17, [sp]
+; CHECK-GI-NEXT:    add x13, sp, #64
+; CHECK-GI-NEXT:    mov v7.s[1], v21.s[0]
 ; CHECK-GI-NEXT:    fmov s21, w7
-; CHECK-GI-NEXT:    ldr s6, [sp]
-; CHECK-GI-NEXT:    ld1 { v20.s }[1], [x8]
 ; CHECK-GI-NEXT:    ldr s19, [sp, #64]
-; CHECK-GI-NEXT:    ldr s7, [sp, #80]
-; CHECK-GI-NEXT:    ldr s22, [sp, #88]
-; CHECK-GI-NEXT:    mov w9, #31 // =0x1f
-; CHECK-GI-NEXT:    mov w11, #1 // =0x1
-; CHECK-GI-NEXT:    mov v21.s[1], v6.s[0]
-; CHECK-GI-NEXT:    fmov s6, w9
+; CHECK-GI-NEXT:    mov w11, #31 // =0x1f
+; CHECK-GI-NEXT:    mov v20.s[1], w12
 ; CHECK-GI-NEXT:    ldr s18, [sp, #96]
-; CHECK-GI-NEXT:    zip1 v17.2d, v17.2d, v20.2d
-; CHECK-GI-NEXT:    fmov s20, w10
-; CHECK-GI-NEXT:    mov v7.s[1], v22.s[0]
-; CHECK-GI-NEXT:    mov v4.s[2], v19.s[0]
-; CHECK-GI-NEXT:    fmov s19, w11
+; CHECK-GI-NEXT:    ld1 { v4.s }[2], [x13]
+; CHECK-GI-NEXT:    mov w13, #1 // =0x1
+; CHECK-GI-NEXT:    mov v3.s[2], v19.s[0]
+; CHECK-GI-NEXT:    mov v21.s[1], v17.s[0]
+; CHECK-GI-NEXT:    fmov s17, w11
+; CHECK-GI-NEXT:    fmov s19, w13
 ; CHECK-GI-NEXT:    fmov s23, w0
-; CHECK-GI-NEXT:    mov v6.s[1], w9
-; CHECK-GI-NEXT:    fmov s24, w9
-; CHECK-GI-NEXT:    ldr s2, [sp, #8]
-; CHECK-GI-NEXT:    mov v20.s[1], w10
+; CHECK-GI-NEXT:    fmov s24, w11
+; CHECK-GI-NEXT:    ldr s6, [sp, #8]
 ; CHECK-GI-NEXT:    ldr s0, [sp, #24]
 ; CHECK-GI-NEXT:    ldr s5, [sp, #32]
-; CHECK-GI-NEXT:    mov v19.s[1], w11
 ; CHECK-GI-NEXT:    mov v7.s[2], v18.s[0]
+; CHECK-GI-NEXT:    mov v17.s[1], w11
+; CHECK-GI-NEXT:    mov v19.s[1], w13
+; CHECK-GI-NEXT:    mov v20.s[2], w12
 ; CHECK-GI-NEXT:    ldr s16, [sp, #72]
 ; CHECK-GI-NEXT:    mov v23.s[1], w1
 ; CHECK-GI-NEXT:    ldr s18, [sp, #80]
-; CHECK-GI-NEXT:    mov v21.s[2], v2.s[0]
-; CHECK-GI-NEXT:    mov v24.s[1], w9
+; CHECK-GI-NEXT:    mov v21.s[2], v6.s[0]
+; CHECK-GI-NEXT:    mov v24.s[1], w11
 ; CHECK-GI-NEXT:    mov v0.s[1], v5.s[0]
-; CHECK-GI-NEXT:    fmov s5, w4
-; CHECK-GI-NEXT:    mov v20.s[2], w10
-; CHECK-GI-NEXT:    add x8, sp, #88
+; CHECK-GI-NEXT:    fmov s6, w4
+; CHECK-GI-NEXT:    add x10, sp, #88
 ; CHECK-GI-NEXT:    movi v22.4s, #31
-; CHECK-GI-NEXT:    mov v4.s[3], v16.s[0]
-; CHECK-GI-NEXT:    mov v6.s[2], w9
-; CHECK-GI-NEXT:    mov v19.s[2], w11
-; CHECK-GI-NEXT:    ldr s1, [sp, #16]
-; CHECK-GI-NEXT:    ldr s3, [sp, #40]
-; CHECK-GI-NEXT:    ld1 { v18.s }[1], [x8]
+; CHECK-GI-NEXT:    mov v3.s[3], v16.s[0]
+; CHECK-GI-NEXT:    mov v17.s[2], w11
+; CHECK-GI-NEXT:    mov v19.s[2], w13
+; CHECK-GI-NEXT:    ldr s2, [sp, #16]
+; CHECK-GI-NEXT:    ldr s1, [sp, #40]
+; CHECK-GI-NEXT:    ld1 { v18.s }[1], [x10]
+; CHECK-GI-NEXT:    eor v5.16b, v7.16b, v20.16b
 ; CHECK-GI-NEXT:    mov v23.s[2], w2
-; CHECK-GI-NEXT:    mov v5.s[1], w5
-; CHECK-GI-NEXT:    add x8, sp, #96
-; CHECK-GI-NEXT:    eor v2.16b, v7.16b, v20.16b
-; CHECK-GI-NEXT:    mov v21.s[3], v1.s[0]
-; CHECK-GI-NEXT:    mov v24.s[2], w9
-; CHECK-GI-NEXT:    mov v0.s[2], v3.s[0]
-; CHECK-GI-NEXT:    bic v1.16b, v22.16b, v4.16b
-; CHECK-GI-NEXT:    ld1 { v18.s }[2], [x8]
+; CHECK-GI-NEXT:    mov v6.s[1], w5
+; CHECK-GI-NEXT:    add x8, sp, #72
+; CHECK-GI-NEXT:    add x9, sp, #96
+; CHECK-GI-NEXT:    mov v21.s[3], v2.s[0]
+; CHECK-GI-NEXT:    mov v24.s[2], w11
+; CHECK-GI-NEXT:    mov v0.s[2], v1.s[0]
+; CHECK-GI-NEXT:    ld1 { v4.s }[3], [x8]
+; CHECK-GI-NEXT:    bic v2.16b, v22.16b, v3.16b
+; CHECK-GI-NEXT:    ld1 { v18.s }[2], [x9]
+; CHECK-GI-NEXT:    and v1.16b, v5.16b, v17.16b
 ; CHECK-GI-NEXT:    neg v3.4s, v19.4s
-; CHECK-GI-NEXT:    and v4.16b, v17.16b, v22.16b
-; CHECK-GI-NEXT:    and v2.16b, v2.16b, v6.16b
 ; CHECK-GI-NEXT:    mov v23.s[3], w3
-; CHECK-GI-NEXT:    mov v5.s[2], w6
-; CHECK-GI-NEXT:    ushr v6.4s, v21.4s, #1
-; CHECK-GI-NEXT:    neg v1.4s, v1.4s
+; CHECK-GI-NEXT:    mov v6.s[2], w6
+; CHECK-GI-NEXT:    and v4.16b, v4.16b, v22.16b
+; CHECK-GI-NEXT:    ushr v5.4s, v21.4s, #1
+; CHECK-GI-NEXT:    neg v2.4s, v2.4s
 ; CHECK-GI-NEXT:    and v7.16b, v18.16b, v24.16b
+; CHECK-GI-NEXT:    neg v1.4s, v1.4s
 ; CHECK-GI-NEXT:    ushl v0.4s, v0.4s, v3.4s
-; CHECK-GI-NEXT:    neg v2.4s, v2.4s
 ; CHECK-GI-NEXT:    ushl v3.4s, v23.4s, v4.4s
-; CHECK-GI-NEXT:    ushl v1.4s, v6.4s, v1.4s
-; CHECK-GI-NEXT:    ushl v4.4s, v5.4s, v7.4s
-; CHECK-GI-NEXT:    ushl v0.4s, v0.4s, v2.4s
-; CHECK-GI-NEXT:    orr v1.16b, v3.16b, v1.16b
+; CHECK-GI-NEXT:    ushl v2.4s, v5.4s, v2.4s
+; CHECK-GI-NEXT:    ushl v4.4s, v6.4s, v7.4s
+; CHECK-GI-NEXT:    ushl v0.4s, v0.4s, v1.4s
+; CHECK-GI-NEXT:    orr v1.16b, v3.16b, v2.16b
 ; CHECK-GI-NEXT:    orr v0.16b, v4.16b, v0.16b
 ; CHECK-GI-NEXT:    mov s2, v1.s[1]
 ; CHECK-GI-NEXT:    mov s3, v1.s[2]
 ; CHECK-GI-NEXT:    mov s4, v1.s[3]
-; CHECK-GI-NEXT:    fmov w0, s1
 ; CHECK-GI-NEXT:    mov s5, v0.s[1]
 ; CHECK-GI-NEXT:    mov s6, v0.s[2]
+; CHECK-GI-NEXT:    fmov w0, s1
 ; CHECK-GI-NEXT:    fmov w4, s0
 ; CHECK-GI-NEXT:    fmov w1, s2
 ; CHECK-GI-NEXT:    fmov w2, s3
diff --git a/llvm/test/CodeGen/AArch64/llvm.frexp.ll b/llvm/test/CodeGen/AArch64/llvm.frexp.ll
index 4e1876db7..2213aa1 100644
--- a/llvm/test/CodeGen/AArch64/llvm.frexp.ll
+++ b/llvm/test/CodeGen/AArch64/llvm.frexp.ll
@@ -700,14 +700,13 @@ define { <4 x float>, <4 x i32> } @test_frexp_v4f32_v4i32(<4 x float> %a) nounwi
 ; CHECK-NEXT:    ldr s1, [sp, #44]
 ; CHECK-NEXT:    ldr q2, [sp] // 16-byte Folded Reload
 ; CHECK-NEXT:    // kill: def $s0 killed $s0 def $q0
-; CHECK-NEXT:    mov v2.s[3], v0.s[0]
 ; CHECK-NEXT:    ld1 { v1.s }[1], [x19]
-; CHECK-NEXT:    ldr s0, [x20]
-; CHECK-NEXT:    ld1 { v0.s }[1], [x21]
+; CHECK-NEXT:    mov v2.s[3], v0.s[0]
+; CHECK-NEXT:    ld1 { v1.s }[2], [x20]
 ; CHECK-NEXT:    ldp x20, x19, [sp, #64] // 16-byte Folded Reload
-; CHECK-NEXT:    ldp x30, x21, [sp, #48] // 16-byte Folded Reload
-; CHECK-NEXT:    zip1 v1.2d, v1.2d, v0.2d
 ; CHECK-NEXT:    mov v0.16b, v2.16b
+; CHECK-NEXT:    ld1 { v1.s }[3], [x21]
+; CHECK-NEXT:    ldp x30, x21, [sp, #48] // 16-byte Folded Reload
 ; CHECK-NEXT:    add sp, sp, #80
 ; CHECK-NEXT:    ret
 ;
@@ -873,11 +872,10 @@ define <4 x i32> @test_frexp_v4f32_v4i32_only_use_exp(<4 x float> %a) nounwind {
 ; CHECK-NEXT:    bl frexpf
 ; CHECK-NEXT:    ldr s0, [sp, #28]
 ; CHECK-NEXT:    ld1 { v0.s }[1], [x19]
-; CHECK-NEXT:    ldr s1, [x20]
-; CHECK-NEXT:    ld1 { v1.s }[1], [x21]
+; CHECK-NEXT:    ld1 { v0.s }[2], [x20]
 ; CHECK-NEXT:    ldp x20, x19, [sp, #48] // 16-byte Folded Reload
+; CHECK-NEXT:    ld1 { v0.s }[3], [x21]
 ; CHECK-NEXT:    ldp x30, x21, [sp, #32] // 16-byte Folded Reload
-; CHECK-NEXT:    zip1 v0.2d, v0.2d, v1.2d
 ; CHECK-NEXT:    add sp, sp, #64
 ; CHECK-NEXT:    ret
 ;
diff --git a/llvm/test/CodeGen/AArch64/neon-dotreduce.ll b/llvm/test/CodeGen/AArch64/neon-dotreduce.ll
index 9443004..4f0c408 100644
--- a/llvm/test/CodeGen/AArch64/neon-dotreduce.ll
+++ b/llvm/test/CodeGen/AArch64/neon-dotreduce.ll
@@ -6810,200 +6810,195 @@ define i32 @test_sdot_v48i8_double_nomla(<48 x i8> %a, <48 x i8> %b, <48 x i8> %
 ; CHECK-SD-NEXT:    str x29, [sp, #-16]! // 8-byte Folded Spill
 ; CHECK-SD-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-SD-NEXT:    .cfi_offset w29, -16
-; CHECK-SD-NEXT:    ldr b0, [sp, #208]
+; CHECK-SD-NEXT:    ldr b5, [sp, #208]
 ; CHECK-SD-NEXT:    add x8, sp, #216
-; CHECK-SD-NEXT:    add x9, sp, #272
-; CHECK-SD-NEXT:    ldr b2, [sp, #80]
+; CHECK-SD-NEXT:    fmov s0, w0
 ; CHECK-SD-NEXT:    ldr b4, [sp, #976]
-; CHECK-SD-NEXT:    ldr b6, [sp, #720]
-; CHECK-SD-NEXT:    ld1 { v0.b }[1], [x8]
+; CHECK-SD-NEXT:    add x9, sp, #984
+; CHECK-SD-NEXT:    add x12, sp, #328
+; CHECK-SD-NEXT:    ld1 { v5.b }[1], [x8]
 ; CHECK-SD-NEXT:    add x8, sp, #224
-; CHECK-SD-NEXT:    fmov s16, w0
-; CHECK-SD-NEXT:    ldr b17, [sp, #848]
-; CHECK-SD-NEXT:    add x10, sp, #24
-; CHECK-SD-NEXT:    movi v19.2d, #0000000000000000
-; CHECK-SD-NEXT:    ld1 { v0.b }[2], [x8]
+; CHECK-SD-NEXT:    movi v1.16b, #1
+; CHECK-SD-NEXT:    mov v0.b[1], w1
+; CHECK-SD-NEXT:    ld1 { v4.b }[1], [x9]
+; CHECK-SD-NEXT:    movi v3.2d, #0000000000000000
+; CHECK-SD-NEXT:    add x11, sp, #992
+; CHECK-SD-NEXT:    ldr b6, [sp, #720]
+; CHECK-SD-NEXT:    ldr b7, [sp, #80]
+; CHECK-SD-NEXT:    ld1 { v5.b }[2], [x8]
 ; CHECK-SD-NEXT:    add x8, sp, #232
-; CHECK-SD-NEXT:    mov v16.b[1], w1
-; CHECK-SD-NEXT:    ld1 { v0.b }[3], [x8]
+; CHECK-SD-NEXT:    add x13, sp, #88
+; CHECK-SD-NEXT:    ld1 { v4.b }[2], [x11]
+; CHECK-SD-NEXT:    ld1 { v7.b }[1], [x13]
+; CHECK-SD-NEXT:    add x13, sp, #856
+; CHECK-SD-NEXT:    mov v0.b[2], w2
+; CHECK-SD-NEXT:    add x14, sp, #1008
+; CHECK-SD-NEXT:    add x15, sp, #872
+; CHECK-SD-NEXT:    ld1 { v5.b }[3], [x8]
 ; CHECK-SD-NEXT:    add x8, sp, #240
-; CHECK-SD-NEXT:    mov v16.b[2], w2
-; CHECK-SD-NEXT:    ld1 { v0.b }[4], [x8]
+; CHECK-SD-NEXT:    add x16, sp, #888
+; CHECK-SD-NEXT:    add x10, sp, #16
+; CHECK-SD-NEXT:    add x9, sp, #24
+; CHECK-SD-NEXT:    add x11, sp, #40
+; CHECK-SD-NEXT:    movi v2.2d, #0000000000000000
+; CHECK-SD-NEXT:    ld1 { v5.b }[4], [x8]
 ; CHECK-SD-NEXT:    add x8, sp, #248
-; CHECK-SD-NEXT:    mov v16.b[3], w3
-; CHECK-SD-NEXT:    ld1 { v0.b }[5], [x8]
+; CHECK-SD-NEXT:    mov v0.b[3], w3
+; CHECK-SD-NEXT:    ld1 { v5.b }[5], [x8]
 ; CHECK-SD-NEXT:    add x8, sp, #256
-; CHECK-SD-NEXT:    ld1 { v0.b }[6], [x8]
+; CHECK-SD-NEXT:    mov v0.b[4], w4
+; CHECK-SD-NEXT:    ld1 { v5.b }[6], [x8]
 ; CHECK-SD-NEXT:    add x8, sp, #264
-; CHECK-SD-NEXT:    mov v16.b[4], w4
-; CHECK-SD-NEXT:    ld1 { v0.b }[7], [x8]
-; CHECK-SD-NEXT:    ldr b1, [x9]
+; CHECK-SD-NEXT:    mov v0.b[5], w5
+; CHECK-SD-NEXT:    ld1 { v5.b }[7], [x8]
+; CHECK-SD-NEXT:    add x8, sp, #272
+; CHECK-SD-NEXT:    ld1 { v5.b }[8], [x8]
 ; CHECK-SD-NEXT:    add x8, sp, #280
-; CHECK-SD-NEXT:    add x9, sp, #88
-; CHECK-SD-NEXT:    mov v16.b[5], w5
-; CHECK-SD-NEXT:    ld1 { v1.b }[1], [x8]
+; CHECK-SD-NEXT:    mov v0.b[6], w6
+; CHECK-SD-NEXT:    ld1 { v5.b }[9], [x8]
 ; CHECK-SD-NEXT:    add x8, sp, #288
-; CHECK-SD-NEXT:    ld1 { v1.b }[2], [x8]
+; CHECK-SD-NEXT:    mov v0.b[7], w7
+; CHECK-SD-NEXT:    ld1 { v5.b }[10], [x8]
 ; CHECK-SD-NEXT:    add x8, sp, #296
-; CHECK-SD-NEXT:    mov v16.b[6], w6
-; CHECK-SD-NEXT:    ld1 { v1.b }[3], [x8]
+; CHECK-SD-NEXT:    ld1 { v0.b }[8], [x10]
+; CHECK-SD-NEXT:    add x10, sp, #128
+; CHECK-SD-NEXT:    ld1 { v5.b }[11], [x8]
 ; CHECK-SD-NEXT:    add x8, sp, #304
-; CHECK-SD-NEXT:    mov v16.b[7], w7
-; CHECK-SD-NEXT:    ld1 { v1.b }[4], [x8]
+; CHECK-SD-NEXT:    ld1 { v0.b }[9], [x9]
+; CHECK-SD-NEXT:    add x9, sp, #136
+; CHECK-SD-NEXT:    ld1 { v5.b }[12], [x8]
 ; CHECK-SD-NEXT:    add x8, sp, #312
-; CHECK-SD-NEXT:    ld1 { v1.b }[5], [x8]
+; CHECK-SD-NEXT:    ld1 { v5.b }[13], [x8]
 ; CHECK-SD-NEXT:    add x8, sp, #320
-; CHECK-SD-NEXT:    ld1 { v1.b }[6], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #328
-; CHECK-SD-NEXT:    ld1 { v1.b }[7], [x8]
-; CHECK-SD-NEXT:    ld1 { v2.b }[1], [x9]
-; CHECK-SD-NEXT:    add x8, sp, #96
-; CHECK-SD-NEXT:    add x9, sp, #144
-; CHECK-SD-NEXT:    ld1 { v2.b }[2], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #104
-; CHECK-SD-NEXT:    zip1 v0.2d, v0.2d, v1.2d
-; CHECK-SD-NEXT:    movi v1.16b, #1
-; CHECK-SD-NEXT:    ld1 { v2.b }[3], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #112
-; CHECK-SD-NEXT:    ld1 { v2.b }[4], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #120
-; CHECK-SD-NEXT:    ld1 { v2.b }[5], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #128
-; CHECK-SD-NEXT:    ld1 { v2.b }[6], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #136
-; CHECK-SD-NEXT:    ld1 { v2.b }[7], [x8]
-; CHECK-SD-NEXT:    ldr b3, [x9]
+; CHECK-SD-NEXT:    ld1 { v5.b }[14], [x8]
+; CHECK-SD-NEXT:    add x8, sp, #32
+; CHECK-SD-NEXT:    ld1 { v0.b }[10], [x8]
+; CHECK-SD-NEXT:    add x8, sp, #144
+; CHECK-SD-NEXT:    ld1 { v5.b }[15], [x12]
+; CHECK-SD-NEXT:    add x12, sp, #728
+; CHECK-SD-NEXT:    ld1 { v6.b }[1], [x12]
+; CHECK-SD-NEXT:    add x12, sp, #1000
+; CHECK-SD-NEXT:    ld1 { v0.b }[11], [x11]
+; CHECK-SD-NEXT:    ld1 { v4.b }[3], [x12]
+; CHECK-SD-NEXT:    add x12, sp, #736
+; CHECK-SD-NEXT:    add x11, sp, #920
+; CHECK-SD-NEXT:    sdot v3.4s, v5.16b, v1.16b
+; CHECK-SD-NEXT:    ldr b5, [sp, #848]
+; CHECK-SD-NEXT:    ld1 { v6.b }[2], [x12]
+; CHECK-SD-NEXT:    add x12, sp, #48
+; CHECK-SD-NEXT:    ld1 { v5.b }[1], [x13]
+; CHECK-SD-NEXT:    add x13, sp, #744
+; CHECK-SD-NEXT:    ld1 { v4.b }[4], [x14]
+; CHECK-SD-NEXT:    add x14, sp, #96
+; CHECK-SD-NEXT:    ld1 { v0.b }[12], [x12]
+; CHECK-SD-NEXT:    ld1 { v6.b }[3], [x13]
+; CHECK-SD-NEXT:    add x13, sp, #864
+; CHECK-SD-NEXT:    ld1 { v7.b }[2], [x14]
+; CHECK-SD-NEXT:    add x14, sp, #1016
+; CHECK-SD-NEXT:    ld1 { v5.b }[2], [x13]
+; CHECK-SD-NEXT:    add x13, sp, #752
+; CHECK-SD-NEXT:    ld1 { v4.b }[5], [x14]
+; CHECK-SD-NEXT:    add x14, sp, #104
+; CHECK-SD-NEXT:    ld1 { v6.b }[4], [x13]
+; CHECK-SD-NEXT:    add x13, sp, #1024
+; CHECK-SD-NEXT:    ld1 { v7.b }[3], [x14]
+; CHECK-SD-NEXT:    ld1 { v5.b }[3], [x15]
+; CHECK-SD-NEXT:    add x15, sp, #760
+; CHECK-SD-NEXT:    add x14, sp, #112
+; CHECK-SD-NEXT:    ld1 { v4.b }[6], [x13]
+; CHECK-SD-NEXT:    add x13, sp, #880
+; CHECK-SD-NEXT:    ld1 { v6.b }[5], [x15]
+; CHECK-SD-NEXT:    add x15, sp, #1032
+; CHECK-SD-NEXT:    ld1 { v7.b }[4], [x14]
+; CHECK-SD-NEXT:    ld1 { v5.b }[4], [x13]
+; CHECK-SD-NEXT:    add x14, sp, #768
+; CHECK-SD-NEXT:    add x13, sp, #120
+; CHECK-SD-NEXT:    ld1 { v4.b }[7], [x15]
+; CHECK-SD-NEXT:    add x15, sp, #1040
+; CHECK-SD-NEXT:    ld1 { v6.b }[6], [x14]
+; CHECK-SD-NEXT:    ld1 { v7.b }[5], [x13]
+; CHECK-SD-NEXT:    add x13, sp, #776
+; CHECK-SD-NEXT:    ld1 { v5.b }[5], [x16]
+; CHECK-SD-NEXT:    add x14, sp, #1048
+; CHECK-SD-NEXT:    ld1 { v4.b }[8], [x15]
+; CHECK-SD-NEXT:    add x15, sp, #896
+; CHECK-SD-NEXT:    ld1 { v6.b }[7], [x13]
+; CHECK-SD-NEXT:    ld1 { v7.b }[6], [x10]
+; CHECK-SD-NEXT:    add x10, sp, #784
+; CHECK-SD-NEXT:    ld1 { v5.b }[6], [x15]
+; CHECK-SD-NEXT:    add x13, sp, #1056
+; CHECK-SD-NEXT:    ld1 { v4.b }[9], [x14]
+; CHECK-SD-NEXT:    add x14, sp, #904
+; CHECK-SD-NEXT:    ld1 { v6.b }[8], [x10]
+; CHECK-SD-NEXT:    ld1 { v7.b }[7], [x9]
+; CHECK-SD-NEXT:    add x9, sp, #792
+; CHECK-SD-NEXT:    ld1 { v5.b }[7], [x14]
+; CHECK-SD-NEXT:    add x10, sp, #1064
+; CHECK-SD-NEXT:    ld1 { v4.b }[10], [x13]
+; CHECK-SD-NEXT:    add x13, sp, #912
+; CHECK-SD-NEXT:    ld1 { v6.b }[9], [x9]
+; CHECK-SD-NEXT:    ld1 { v7.b }[8], [x8]
+; CHECK-SD-NEXT:    add x9, sp, #800
+; CHECK-SD-NEXT:    ld1 { v5.b }[8], [x13]
 ; CHECK-SD-NEXT:    add x8, sp, #152
-; CHECK-SD-NEXT:    add x9, sp, #984
-; CHECK-SD-NEXT:    ld1 { v3.b }[1], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #160
-; CHECK-SD-NEXT:    ld1 { v3.b }[2], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #168
-; CHECK-SD-NEXT:    ld1 { v3.b }[3], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #176
-; CHECK-SD-NEXT:    ld1 { v3.b }[4], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #184
-; CHECK-SD-NEXT:    ld1 { v3.b }[5], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #192
-; CHECK-SD-NEXT:    ld1 { v3.b }[6], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #200
-; CHECK-SD-NEXT:    ld1 { v3.b }[7], [x8]
-; CHECK-SD-NEXT:    ld1 { v4.b }[1], [x9]
-; CHECK-SD-NEXT:    add x8, sp, #992
-; CHECK-SD-NEXT:    add x9, sp, #1040
-; CHECK-SD-NEXT:    ld1 { v4.b }[2], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #1000
-; CHECK-SD-NEXT:    zip1 v2.2d, v2.2d, v3.2d
-; CHECK-SD-NEXT:    ld1 { v4.b }[3], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #1008
-; CHECK-SD-NEXT:    ld1 { v4.b }[4], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #1016
-; CHECK-SD-NEXT:    ld1 { v4.b }[5], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #1024
-; CHECK-SD-NEXT:    ld1 { v4.b }[6], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #1032
-; CHECK-SD-NEXT:    ld1 { v4.b }[7], [x8]
-; CHECK-SD-NEXT:    ldr b5, [x9]
-; CHECK-SD-NEXT:    add x8, sp, #1048
-; CHECK-SD-NEXT:    add x9, sp, #728
-; CHECK-SD-NEXT:    ld1 { v5.b }[1], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #1056
-; CHECK-SD-NEXT:    ld1 { v5.b }[2], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #1064
-; CHECK-SD-NEXT:    ld1 { v5.b }[3], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #1072
-; CHECK-SD-NEXT:    ld1 { v5.b }[4], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #1080
-; CHECK-SD-NEXT:    ld1 { v5.b }[5], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #1088
-; CHECK-SD-NEXT:    ld1 { v5.b }[6], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #1096
-; CHECK-SD-NEXT:    ld1 { v5.b }[7], [x8]
-; CHECK-SD-NEXT:    ld1 { v6.b }[1], [x9]
-; CHECK-SD-NEXT:    add x8, sp, #736
-; CHECK-SD-NEXT:    add x9, sp, #784
-; CHECK-SD-NEXT:    ld1 { v6.b }[2], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #744
-; CHECK-SD-NEXT:    zip1 v4.2d, v4.2d, v5.2d
-; CHECK-SD-NEXT:    movi v5.2d, #0000000000000000
-; CHECK-SD-NEXT:    ld1 { v6.b }[3], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #752
-; CHECK-SD-NEXT:    sdot v19.4s, v4.16b, v1.16b
-; CHECK-SD-NEXT:    sdot v5.4s, v0.16b, v1.16b
-; CHECK-SD-NEXT:    ld1 { v6.b }[4], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #760
-; CHECK-SD-NEXT:    ld1 { v6.b }[5], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #768
-; CHECK-SD-NEXT:    ld1 { v6.b }[6], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #776
-; CHECK-SD-NEXT:    ld1 { v6.b }[7], [x8]
-; CHECK-SD-NEXT:    ldr b7, [x9]
-; CHECK-SD-NEXT:    add x8, sp, #792
-; CHECK-SD-NEXT:    add x9, sp, #856
-; CHECK-SD-NEXT:    ld1 { v7.b }[1], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #800
-; CHECK-SD-NEXT:    ld1 { v7.b }[2], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #808
-; CHECK-SD-NEXT:    ld1 { v7.b }[3], [x8]
+; CHECK-SD-NEXT:    ld1 { v4.b }[11], [x10]
+; CHECK-SD-NEXT:    add x10, sp, #1072
+; CHECK-SD-NEXT:    ld1 { v6.b }[10], [x9]
+; CHECK-SD-NEXT:    ld1 { v7.b }[9], [x8]
+; CHECK-SD-NEXT:    add x9, sp, #808
+; CHECK-SD-NEXT:    ld1 { v5.b }[9], [x11]
+; CHECK-SD-NEXT:    add x8, sp, #56
+; CHECK-SD-NEXT:    ld1 { v4.b }[12], [x10]
+; CHECK-SD-NEXT:    add x10, sp, #160
+; CHECK-SD-NEXT:    ld1 { v0.b }[13], [x8]
+; CHECK-SD-NEXT:    ld1 { v6.b }[11], [x9]
+; CHECK-SD-NEXT:    add x9, sp, #928
+; CHECK-SD-NEXT:    ld1 { v7.b }[10], [x10]
+; CHECK-SD-NEXT:    add x10, sp, #1080
+; CHECK-SD-NEXT:    ld1 { v5.b }[10], [x9]
 ; CHECK-SD-NEXT:    add x8, sp, #816
-; CHECK-SD-NEXT:    ld1 { v7.b }[4], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #824
-; CHECK-SD-NEXT:    ld1 { v7.b }[5], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #832
-; CHECK-SD-NEXT:    ld1 { v7.b }[6], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #840
-; CHECK-SD-NEXT:    ld1 { v7.b }[7], [x8]
-; CHECK-SD-NEXT:    ld1 { v17.b }[1], [x9]
-; CHECK-SD-NEXT:    add x8, sp, #864
-; CHECK-SD-NEXT:    add x9, sp, #16
-; CHECK-SD-NEXT:    ld1 { v16.b }[8], [x9]
-; CHECK-SD-NEXT:    add x9, sp, #912
-; CHECK-SD-NEXT:    ld1 { v17.b }[2], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #872
-; CHECK-SD-NEXT:    zip1 v0.2d, v6.2d, v7.2d
-; CHECK-SD-NEXT:    ld1 { v16.b }[9], [x10]
-; CHECK-SD-NEXT:    ld1 { v17.b }[3], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #880
-; CHECK-SD-NEXT:    sdot v19.4s, v0.16b, v1.16b
-; CHECK-SD-NEXT:    ld1 { v17.b }[4], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #888
-; CHECK-SD-NEXT:    ld1 { v17.b }[5], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #896
-; CHECK-SD-NEXT:    ld1 { v17.b }[6], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #904
-; CHECK-SD-NEXT:    ld1 { v17.b }[7], [x8]
-; CHECK-SD-NEXT:    ldr b18, [x9]
-; CHECK-SD-NEXT:    add x8, sp, #920
-; CHECK-SD-NEXT:    ld1 { v18.b }[1], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #32
-; CHECK-SD-NEXT:    ld1 { v16.b }[10], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #928
-; CHECK-SD-NEXT:    ld1 { v18.b }[2], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #40
-; CHECK-SD-NEXT:    ld1 { v16.b }[11], [x8]
+; CHECK-SD-NEXT:    ld1 { v4.b }[13], [x10]
+; CHECK-SD-NEXT:    add x9, sp, #168
+; CHECK-SD-NEXT:    add x10, sp, #176
+; CHECK-SD-NEXT:    ld1 { v6.b }[12], [x8]
 ; CHECK-SD-NEXT:    add x8, sp, #936
-; CHECK-SD-NEXT:    ld1 { v18.b }[3], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #48
-; CHECK-SD-NEXT:    ld1 { v16.b }[12], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #944
-; CHECK-SD-NEXT:    ld1 { v18.b }[4], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #56
-; CHECK-SD-NEXT:    ld1 { v16.b }[13], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #952
-; CHECK-SD-NEXT:    ld1 { v18.b }[5], [x8]
+; CHECK-SD-NEXT:    ld1 { v7.b }[11], [x9]
+; CHECK-SD-NEXT:    add x9, sp, #1088
+; CHECK-SD-NEXT:    ld1 { v5.b }[11], [x8]
 ; CHECK-SD-NEXT:    add x8, sp, #64
-; CHECK-SD-NEXT:    ld1 { v16.b }[14], [x8]
+; CHECK-SD-NEXT:    ld1 { v4.b }[14], [x9]
+; CHECK-SD-NEXT:    add x9, sp, #824
+; CHECK-SD-NEXT:    ld1 { v0.b }[14], [x8]
+; CHECK-SD-NEXT:    ld1 { v6.b }[13], [x9]
+; CHECK-SD-NEXT:    add x9, sp, #944
+; CHECK-SD-NEXT:    ld1 { v7.b }[12], [x10]
+; CHECK-SD-NEXT:    add x10, sp, #1096
+; CHECK-SD-NEXT:    ld1 { v5.b }[12], [x9]
+; CHECK-SD-NEXT:    add x8, sp, #832
+; CHECK-SD-NEXT:    ld1 { v4.b }[15], [x10]
+; CHECK-SD-NEXT:    add x9, sp, #184
+; CHECK-SD-NEXT:    add x10, sp, #72
+; CHECK-SD-NEXT:    ld1 { v6.b }[14], [x8]
+; CHECK-SD-NEXT:    add x8, sp, #952
+; CHECK-SD-NEXT:    ld1 { v7.b }[13], [x9]
+; CHECK-SD-NEXT:    ld1 { v5.b }[13], [x8]
+; CHECK-SD-NEXT:    add x8, sp, #840
+; CHECK-SD-NEXT:    ld1 { v0.b }[15], [x10]
+; CHECK-SD-NEXT:    sdot v2.4s, v4.16b, v1.16b
+; CHECK-SD-NEXT:    add x9, sp, #192
+; CHECK-SD-NEXT:    ld1 { v6.b }[15], [x8]
 ; CHECK-SD-NEXT:    add x8, sp, #960
-; CHECK-SD-NEXT:    ld1 { v18.b }[6], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #72
-; CHECK-SD-NEXT:    ld1 { v16.b }[15], [x8]
-; CHECK-SD-NEXT:    add x8, sp, #968
-; CHECK-SD-NEXT:    ld1 { v18.b }[7], [x8]
-; CHECK-SD-NEXT:    sdot v5.4s, v16.16b, v1.16b
-; CHECK-SD-NEXT:    zip1 v0.2d, v17.2d, v18.2d
-; CHECK-SD-NEXT:    sdot v5.4s, v2.16b, v1.16b
-; CHECK-SD-NEXT:    sdot v19.4s, v0.16b, v1.16b
-; CHECK-SD-NEXT:    add v0.4s, v5.4s, v19.4s
+; CHECK-SD-NEXT:    ld1 { v7.b }[14], [x9]
+; CHECK-SD-NEXT:    ld1 { v5.b }[14], [x8]
+; CHECK-SD-NEXT:    sdot v3.4s, v0.16b, v1.16b
+; CHECK-SD-NEXT:    add x8, sp, #200
+; CHECK-SD-NEXT:    add x9, sp, #968
+; CHECK-SD-NEXT:    sdot v2.4s, v6.16b, v1.16b
+; CHECK-SD-NEXT:    ld1 { v7.b }[15], [x8]
+; CHECK-SD-NEXT:    ld1 { v5.b }[15], [x9]
+; CHECK-SD-NEXT:    sdot v3.4s, v7.16b, v1.16b
+; CHECK-SD-NEXT:    sdot v2.4s, v5.16b, v1.16b
+; CHECK-SD-NEXT:    add v0.4s, v3.4s, v2.4s
 ; CHECK-SD-NEXT:    addv s0, v0.4s
 ; CHECK-SD-NEXT:    fmov w0, s0
 ; CHECK-SD-NEXT:    ldr x29, [sp], #16 // 8-byte Folded Reload
diff --git a/llvm/test/CodeGen/AArch64/nontemporal.ll b/llvm/test/CodeGen/AArch64/nontemporal.ll
index f7a87ae..f8ba150 100644
--- a/llvm/test/CodeGen/AArch64/nontemporal.ll
+++ b/llvm/test/CodeGen/AArch64/nontemporal.ll
@@ -683,43 +683,41 @@ define void @test_stnp_v17f32(<17 x float> %v, ptr %ptr) {
 ;
 ; CHECK-BE-LABEL: test_stnp_v17f32:
 ; CHECK-BE:       // %bb.0: // %entry
-; CHECK-BE-NEXT:    // kill: def $s1 killed $s1 def $q1
-; CHECK-BE-NEXT:    // kill: def $s0 killed $s0 def $q0
 ; CHECK-BE-NEXT:    // kill: def $s4 killed $s4 def $q4
-; CHECK-BE-NEXT:    // kill: def $s5 killed $s5 def $q5
-; CHECK-BE-NEXT:    add x8, sp, #12
-; CHECK-BE-NEXT:    add x9, sp, #20
+; CHECK-BE-NEXT:    // kill: def $s0 killed $s0 def $q0
 ; CHECK-BE-NEXT:    ldr s16, [sp, #36]
-; CHECK-BE-NEXT:    mov v0.s[1], v1.s[0]
-; CHECK-BE-NEXT:    ldr s1, [sp, #4]
+; CHECK-BE-NEXT:    // kill: def $s5 killed $s5 def $q5
+; CHECK-BE-NEXT:    // kill: def $s1 killed $s1 def $q1
+; CHECK-BE-NEXT:    ldr s17, [sp, #4]
+; CHECK-BE-NEXT:    add x8, sp, #44
 ; CHECK-BE-NEXT:    mov v4.s[1], v5.s[0]
-; CHECK-BE-NEXT:    add x10, sp, #52
+; CHECK-BE-NEXT:    mov v0.s[1], v1.s[0]
 ; CHECK-BE-NEXT:    // kill: def $s6 killed $s6 def $q6
 ; CHECK-BE-NEXT:    // kill: def $s2 killed $s2 def $q2
 ; CHECK-BE-NEXT:    // kill: def $s7 killed $s7 def $q7
 ; CHECK-BE-NEXT:    // kill: def $s3 killed $s3 def $q3
-; CHECK-BE-NEXT:    ld1 { v1.s }[1], [x8]
-; CHECK-BE-NEXT:    ldr s5, [x9]
-; CHECK-BE-NEXT:    add x8, sp, #28
-; CHECK-BE-NEXT:    add x9, sp, #44
-; CHECK-BE-NEXT:    ld1 { v5.s }[1], [x8]
-; CHECK-BE-NEXT:    ld1 { v16.s }[1], [x9]
-; CHECK-BE-NEXT:    ldr s17, [x10]
-; CHECK-BE-NEXT:    add x8, sp, #60
+; CHECK-BE-NEXT:    ldr s1, [sp, #68]
+; CHECK-BE-NEXT:    ld1 { v16.s }[1], [x8]
+; CHECK-BE-NEXT:    add x8, sp, #12
+; CHECK-BE-NEXT:    ld1 { v17.s }[1], [x8]
+; CHECK-BE-NEXT:    add x8, sp, #52
+; CHECK-BE-NEXT:    str s1, [x0, #64]
+; CHECK-BE-NEXT:    ld1 { v16.s }[2], [x8]
+; CHECK-BE-NEXT:    add x8, sp, #20
 ; CHECK-BE-NEXT:    mov v4.s[2], v6.s[0]
 ; CHECK-BE-NEXT:    mov v0.s[2], v2.s[0]
-; CHECK-BE-NEXT:    ld1 { v17.s }[1], [x8]
-; CHECK-BE-NEXT:    ldr s2, [sp, #68]
-; CHECK-BE-NEXT:    add x8, x0, #32
-; CHECK-BE-NEXT:    zip1 v1.2d, v1.2d, v5.2d
-; CHECK-BE-NEXT:    add x9, x0, #48
-; CHECK-BE-NEXT:    str s2, [x0, #64]
-; CHECK-BE-NEXT:    zip1 v5.2d, v16.2d, v17.2d
+; CHECK-BE-NEXT:    ld1 { v17.s }[2], [x8]
+; CHECK-BE-NEXT:    add x8, sp, #60
+; CHECK-BE-NEXT:    ld1 { v16.s }[3], [x8]
+; CHECK-BE-NEXT:    add x8, sp, #28
+; CHECK-BE-NEXT:    ld1 { v17.s }[3], [x8]
 ; CHECK-BE-NEXT:    mov v4.s[3], v7.s[0]
+; CHECK-BE-NEXT:    add x8, x0, #48
 ; CHECK-BE-NEXT:    mov v0.s[3], v3.s[0]
-; CHECK-BE-NEXT:    st1 { v1.4s }, [x8]
+; CHECK-BE-NEXT:    st1 { v16.4s }, [x8]
+; CHECK-BE-NEXT:    add x8, x0, #32
+; CHECK-BE-NEXT:    st1 { v17.4s }, [x8]
 ; CHECK-BE-NEXT:    add x8, x0, #16
-; CHECK-BE-NEXT:    st1 { v5.4s }, [x9]
 ; CHECK-BE-NEXT:    st1 { v4.4s }, [x8]
 ; CHECK-BE-NEXT:    st1 { v0.4s }, [x0]
 ; CHECK-BE-NEXT:    ret
diff --git a/llvm/test/CodeGen/AMDGPU/GlobalISel/fp64-atomics-gfx90a.ll b/llvm/test/CodeGen/AMDGPU/GlobalISel/fp64-atomics-gfx90a.ll
index 8192d4a..0e132f1 100644
--- a/llvm/test/CodeGen/AMDGPU/GlobalISel/fp64-atomics-gfx90a.ll
+++ b/llvm/test/CodeGen/AMDGPU/GlobalISel/fp64-atomics-gfx90a.ll
@@ -70,12 +70,12 @@ define amdgpu_kernel void @raw_buffer_atomic_add_rtn_f64_off4_slc(<4 x i32> %rsr
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_add_f64 v[0:1], v2, s[0:3], 4 offen glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_add_f64 v[2:3], v0, s[0:3], 4 offen glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: raw_buffer_atomic_add_rtn_f64_off4_slc:
@@ -85,12 +85,12 @@ define amdgpu_kernel void @raw_buffer_atomic_add_rtn_f64_off4_slc(<4 x i32> %rsr
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_add_f64 v[0:1], v2, s[0:3], 4 offen sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_add_f64 v[2:3], v0, s[0:3], 4 offen sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.raw.buffer.atomic.fadd.f64(double %data, <4 x i32> %rsrc, i32 %vindex, i32 4, i32 2)
@@ -153,12 +153,12 @@ define amdgpu_kernel void @raw_ptr_buffer_atomic_add_rtn_f64_off4_slc(ptr addrsp
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_add_f64 v[0:1], v2, s[0:3], 4 offen glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_add_f64 v[2:3], v0, s[0:3], 4 offen glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: raw_ptr_buffer_atomic_add_rtn_f64_off4_slc:
@@ -168,12 +168,12 @@ define amdgpu_kernel void @raw_ptr_buffer_atomic_add_rtn_f64_off4_slc(ptr addrsp
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_add_f64 v[0:1], v2, s[0:3], 4 offen sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_add_f64 v[2:3], v0, s[0:3], 4 offen sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.raw.ptr.buffer.atomic.fadd.f64(double %data, ptr addrspace(8) %rsrc, i32 %vindex, i32 4, i32 2)
@@ -236,12 +236,12 @@ define amdgpu_kernel void @struct_buffer_atomic_add_rtn_f64_off4_slc(<4 x i32> %
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_add_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_add_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: struct_buffer_atomic_add_rtn_f64_off4_slc:
@@ -251,12 +251,12 @@ define amdgpu_kernel void @struct_buffer_atomic_add_rtn_f64_off4_slc(<4 x i32> %
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_add_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_add_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.struct.buffer.atomic.fadd.f64(double %data, <4 x i32> %rsrc, i32 %vindex, i32 4, i32 0, i32 2)
@@ -319,12 +319,12 @@ define amdgpu_kernel void @struct_ptr_buffer_atomic_add_rtn_f64_off4_slc(ptr add
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_add_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_add_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: struct_ptr_buffer_atomic_add_rtn_f64_off4_slc:
@@ -334,12 +334,12 @@ define amdgpu_kernel void @struct_ptr_buffer_atomic_add_rtn_f64_off4_slc(ptr add
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_add_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_add_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.struct.ptr.buffer.atomic.fadd.f64(double %data, ptr addrspace(8) %rsrc, i32 %vindex, i32 4, i32 0, i32 2)
@@ -402,12 +402,12 @@ define amdgpu_kernel void @raw_buffer_atomic_min_rtn_f64_off4_slc(<4 x i32> %rsr
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_min_f64 v[0:1], v2, s[0:3], 4 offen glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_min_f64 v[2:3], v0, s[0:3], 4 offen glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: raw_buffer_atomic_min_rtn_f64_off4_slc:
@@ -417,12 +417,12 @@ define amdgpu_kernel void @raw_buffer_atomic_min_rtn_f64_off4_slc(<4 x i32> %rsr
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_min_f64 v[0:1], v2, s[0:3], 4 offen sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_min_f64 v[2:3], v0, s[0:3], 4 offen sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.raw.buffer.atomic.fmin.f64(double %data, <4 x i32> %rsrc, i32 %vindex, i32 4, i32 2)
@@ -485,12 +485,12 @@ define amdgpu_kernel void @raw_ptr_buffer_atomic_min_rtn_f64_off4_slc(ptr addrsp
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_min_f64 v[0:1], v2, s[0:3], 4 offen glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_min_f64 v[2:3], v0, s[0:3], 4 offen glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: raw_ptr_buffer_atomic_min_rtn_f64_off4_slc:
@@ -500,12 +500,12 @@ define amdgpu_kernel void @raw_ptr_buffer_atomic_min_rtn_f64_off4_slc(ptr addrsp
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_min_f64 v[0:1], v2, s[0:3], 4 offen sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_min_f64 v[2:3], v0, s[0:3], 4 offen sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.raw.ptr.buffer.atomic.fmin.f64(double %data, ptr addrspace(8) %rsrc, i32 %vindex, i32 4, i32 2)
@@ -568,12 +568,12 @@ define amdgpu_kernel void @struct_buffer_atomic_min_rtn_f64_off4_slc(<4 x i32> %
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_min_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_min_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: struct_buffer_atomic_min_rtn_f64_off4_slc:
@@ -583,12 +583,12 @@ define amdgpu_kernel void @struct_buffer_atomic_min_rtn_f64_off4_slc(<4 x i32> %
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_min_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_min_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.struct.buffer.atomic.fmin.f64(double %data, <4 x i32> %rsrc, i32 %vindex, i32 4, i32 0, i32 2)
@@ -651,12 +651,12 @@ define amdgpu_kernel void @struct_ptr_buffer_atomic_min_rtn_f64_off4_slc(ptr add
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_min_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_min_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: struct_ptr_buffer_atomic_min_rtn_f64_off4_slc:
@@ -666,12 +666,12 @@ define amdgpu_kernel void @struct_ptr_buffer_atomic_min_rtn_f64_off4_slc(ptr add
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_min_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_min_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.struct.ptr.buffer.atomic.fmin.f64(double %data, ptr addrspace(8) %rsrc, i32 %vindex, i32 4, i32 0, i32 2)
@@ -734,12 +734,12 @@ define amdgpu_kernel void @raw_buffer_atomic_max_rtn_f64_off4_slc(<4 x i32> %rsr
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_max_f64 v[0:1], v2, s[0:3], 4 offen glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_max_f64 v[2:3], v0, s[0:3], 4 offen glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: raw_buffer_atomic_max_rtn_f64_off4_slc:
@@ -749,12 +749,12 @@ define amdgpu_kernel void @raw_buffer_atomic_max_rtn_f64_off4_slc(<4 x i32> %rsr
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_max_f64 v[0:1], v2, s[0:3], 4 offen sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_max_f64 v[2:3], v0, s[0:3], 4 offen sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.raw.buffer.atomic.fmax.f64(double %data, <4 x i32> %rsrc, i32 %vindex, i32 4, i32 2)
@@ -817,12 +817,12 @@ define amdgpu_kernel void @raw_ptr_buffer_atomic_max_rtn_f64_off4_slc(ptr addrsp
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_max_f64 v[0:1], v2, s[0:3], 4 offen glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_max_f64 v[2:3], v0, s[0:3], 4 offen glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: raw_ptr_buffer_atomic_max_rtn_f64_off4_slc:
@@ -832,12 +832,12 @@ define amdgpu_kernel void @raw_ptr_buffer_atomic_max_rtn_f64_off4_slc(ptr addrsp
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_max_f64 v[0:1], v2, s[0:3], 4 offen sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_max_f64 v[2:3], v0, s[0:3], 4 offen sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.raw.ptr.buffer.atomic.fmax.f64(double %data, ptr addrspace(8) %rsrc, i32 %vindex, i32 4, i32 2)
@@ -900,12 +900,12 @@ define amdgpu_kernel void @struct_buffer_atomic_max_rtn_f64_off4_slc(<4 x i32> %
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_max_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_max_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: struct_buffer_atomic_max_rtn_f64_off4_slc:
@@ -915,12 +915,12 @@ define amdgpu_kernel void @struct_buffer_atomic_max_rtn_f64_off4_slc(<4 x i32> %
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_max_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_max_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.struct.buffer.atomic.fmax.f64(double %data, <4 x i32> %rsrc, i32 %vindex, i32 4, i32 0, i32 2)
@@ -983,12 +983,12 @@ define amdgpu_kernel void @struct_ptr_buffer_atomic_max_rtn_f64_off4_slc(ptr add
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_max_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_max_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: struct_ptr_buffer_atomic_max_rtn_f64_off4_slc:
@@ -998,12 +998,12 @@ define amdgpu_kernel void @struct_ptr_buffer_atomic_max_rtn_f64_off4_slc(ptr add
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_max_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_max_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.struct.ptr.buffer.atomic.fmax.f64(double %data, ptr addrspace(8) %rsrc, i32 %vindex, i32 4, i32 0, i32 2)
diff --git a/llvm/test/CodeGen/AMDGPU/bad-agpr-vgpr-regalloc-priority.mir b/llvm/test/CodeGen/AMDGPU/bad-agpr-vgpr-regalloc-priority.mir
index 1a457c9..9241a23 100644
--- a/llvm/test/CodeGen/AMDGPU/bad-agpr-vgpr-regalloc-priority.mir
+++ b/llvm/test/CodeGen/AMDGPU/bad-agpr-vgpr-regalloc-priority.mir
@@ -38,20 +38,20 @@ body:             |
     ; CHECK-NEXT: {{  $}}
     ; CHECK-NEXT: early-clobber renamable $sgpr6_sgpr7 = S_LOAD_DWORDX2_IMM_ec renamable $sgpr4_sgpr5, 36, 0 :: (dereferenceable invariant load (s64), align 4, addrspace 4)
     ; CHECK-NEXT: renamable $sgpr0_sgpr1_sgpr2_sgpr3 = S_LOAD_DWORDX4_IMM renamable $sgpr6_sgpr7, 0, 0 :: ("amdgpu-noclobber" load (s128), addrspace 1)
-    ; CHECK-NEXT: renamable $vgpr4 = V_MOV_B32_e32 1065353216, implicit $exec
-    ; CHECK-NEXT: renamable $vgpr5 = V_MOV_B32_e32 0, implicit $exec
-    ; CHECK-NEXT: renamable $vgpr6 = V_MOV_B32_e32 1073741824, implicit $exec
+    ; CHECK-NEXT: renamable $vgpr0 = V_MOV_B32_e32 1065353216, implicit $exec
+    ; CHECK-NEXT: renamable $vgpr4 = V_MOV_B32_e32 0, implicit $exec
+    ; CHECK-NEXT: renamable $vgpr1 = V_MOV_B32_e32 1073741824, implicit $exec
     ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3 = COPY killed renamable $sgpr0_sgpr1_sgpr2_sgpr3
-    ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3 = V_MFMA_F32_4X4X1F32_e64 $vgpr4, $vgpr6, killed $agpr0_agpr1_agpr2_agpr3, 0, 0, 0, implicit $mode, implicit $exec
-    ; CHECK-NEXT: renamable $vgpr1 = COPY renamable $agpr1
-    ; CHECK-NEXT: renamable $vgpr0 = COPY renamable $agpr0
-    ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3 = V_MFMA_F32_4X4X1F32_e64 $vgpr4, $vgpr6, killed $agpr0_agpr1_agpr2_agpr3, 0, 0, 0, implicit $mode, implicit $exec
-    ; CHECK-NEXT: renamable $vgpr3 = COPY renamable $agpr1
-    ; CHECK-NEXT: renamable $vgpr2 = COPY killed renamable $agpr0
-    ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3 = COPY killed renamable $vgpr0_vgpr1_vgpr2_vgpr3
-    ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3 = V_MFMA_F32_4X4X1F32_e64 killed $vgpr4, killed $vgpr6, killed $agpr0_agpr1_agpr2_agpr3, 0, 0, 0, implicit $mode, implicit $exec
+    ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3 = V_MFMA_F32_4X4X1F32_e64 $vgpr0, $vgpr1, killed $agpr0_agpr1_agpr2_agpr3, 0, 0, 0, implicit $mode, implicit $exec
+    ; CHECK-NEXT: renamable $vgpr6 = COPY renamable $agpr1
+    ; CHECK-NEXT: renamable $vgpr5 = COPY renamable $agpr0
+    ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3 = V_MFMA_F32_4X4X1F32_e64 $vgpr0, $vgpr1, killed $agpr0_agpr1_agpr2_agpr3, 0, 0, 0, implicit $mode, implicit $exec
+    ; CHECK-NEXT: renamable $vgpr8 = COPY renamable $agpr1
+    ; CHECK-NEXT: renamable $vgpr7 = COPY killed renamable $agpr0
+    ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3 = COPY killed renamable $vgpr5_vgpr6_vgpr7_vgpr8
+    ; CHECK-NEXT: renamable $agpr0_agpr1_agpr2_agpr3 = V_MFMA_F32_4X4X1F32_e64 killed $vgpr0, killed $vgpr1, killed $agpr0_agpr1_agpr2_agpr3, 0, 0, 0, implicit $mode, implicit $exec
     ; CHECK-NEXT: renamable $vgpr0_vgpr1_vgpr2_vgpr3 = COPY killed renamable $agpr0_agpr1_agpr2_agpr3
-    ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr5, killed renamable $vgpr0_vgpr1_vgpr2_vgpr3, killed renamable $sgpr6_sgpr7, 0, 0, implicit $exec :: (store (s128), addrspace 1)
+    ; CHECK-NEXT: GLOBAL_STORE_DWORDX4_SADDR killed renamable $vgpr4, killed renamable $vgpr0_vgpr1_vgpr2_vgpr3, killed renamable $sgpr6_sgpr7, 0, 0, implicit $exec :: (store (s128), addrspace 1)
     ; CHECK-NEXT: S_ENDPGM 0
     early-clobber renamable $sgpr6_sgpr7 = S_LOAD_DWORDX2_IMM_ec killed renamable $sgpr4_sgpr5, 36, 0 :: (dereferenceable invariant load (s64), align 4, addrspace 4)
     renamable $sgpr0_sgpr1_sgpr2_sgpr3 = S_LOAD_DWORDX4_IMM renamable $sgpr6_sgpr7, 0, 0 :: ("amdgpu-noclobber" load (s128), addrspace 1)
diff --git a/llvm/test/CodeGen/AMDGPU/bf16-math.ll b/llvm/test/CodeGen/AMDGPU/bf16-math.ll
index 1adf542..9979e83 100644
--- a/llvm/test/CodeGen/AMDGPU/bf16-math.ll
+++ b/llvm/test/CodeGen/AMDGPU/bf16-math.ll
@@ -323,6 +323,146 @@ define amdgpu_ps void @v_test_max_v2bf16_vl(ptr addrspace(1) %out, <2 x bfloat>
   ret void
 }
 
+define amdgpu_ps bfloat @test_clamp_bf16(bfloat %src) {
+; GCN-LABEL: test_clamp_bf16:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    v_pk_max_num_bf16 v0, v0, v0 clamp
+; GCN-NEXT:    ; return to shader part epilog
+  %max = call bfloat @llvm.maxnum.bf16(bfloat %src, bfloat 0.0)
+  %clamp = call bfloat @llvm.minnum.bf16(bfloat %max, bfloat 1.0)
+  ret bfloat %clamp
+}
+
+define amdgpu_ps bfloat @test_clamp_bf16_s(bfloat inreg %src) {
+; GCN-LABEL: test_clamp_bf16_s:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    v_pk_max_num_bf16 v0, s0, s0 clamp
+; GCN-NEXT:    ; return to shader part epilog
+  %max = call bfloat @llvm.maxnum.bf16(bfloat %src, bfloat 0.0)
+  %clamp = call bfloat @llvm.minnum.bf16(bfloat %max, bfloat 1.0)
+  ret bfloat %clamp
+}
+
+define amdgpu_ps float @test_clamp_v2bf16(<2 x bfloat> %src) {
+; GCN-LABEL: test_clamp_v2bf16:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    v_pk_max_num_bf16 v0, v0, v0 clamp
+; GCN-NEXT:    ; return to shader part epilog
+  %max = call <2 x bfloat> @llvm.maxnum.v2bf16(<2 x bfloat> %src, <2 x bfloat> <bfloat 0.0, bfloat 0.0>)
+  %clamp = call <2 x bfloat> @llvm.minnum.v2bf16(<2 x bfloat> %max, <2 x bfloat> <bfloat 1.0, bfloat 1.0>)
+  %ret = bitcast <2 x bfloat> %clamp to float
+  ret float %ret
+}
+
+define amdgpu_ps float @test_clamp_v2bf16_s(<2 x bfloat> inreg %src) {
+; GCN-LABEL: test_clamp_v2bf16_s:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    v_pk_max_num_bf16 v0, s0, s0 clamp
+; GCN-NEXT:    ; return to shader part epilog
+  %max = call <2 x bfloat> @llvm.maxnum.v2bf16(<2 x bfloat> %src, <2 x bfloat> <bfloat 0.0, bfloat 0.0>)
+  %clamp = call <2 x bfloat> @llvm.minnum.v2bf16(<2 x bfloat> %max, <2 x bfloat> <bfloat 1.0, bfloat 1.0>)
+  %ret = bitcast <2 x bfloat> %clamp to float
+  ret float %ret
+}
+
+define amdgpu_ps bfloat @test_clamp_bf16_folding(bfloat %src) {
+; GCN-LABEL: test_clamp_bf16_folding:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    v_exp_bf16_e32 v0, v0
+; GCN-NEXT:    v_nop
+; GCN-NEXT:    s_delay_alu instid0(TRANS32_DEP_1)
+; GCN-NEXT:    v_pk_max_num_bf16 v0, v0, v0 clamp
+; GCN-NEXT:    ; return to shader part epilog
+  %exp = call bfloat @llvm.exp2.bf16(bfloat %src)
+  %max = call bfloat @llvm.maxnum.bf16(bfloat %exp, bfloat 0.0)
+  %clamp = call bfloat @llvm.minnum.bf16(bfloat %max, bfloat 1.0)
+  ret bfloat %clamp
+}
+
+define amdgpu_ps float @test_clamp_v2bf16_folding(<2 x bfloat> %src0, <2 x bfloat> %src1) {
+; GCN-LABEL: test_clamp_v2bf16_folding:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    v_pk_mul_bf16 v0, v0, v1
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GCN-NEXT:    v_pk_max_num_bf16 v0, v0, v0 clamp
+; GCN-NEXT:    ; return to shader part epilog
+  %mul = fmul <2 x bfloat> %src0, %src1
+  %max = call <2 x bfloat> @llvm.maxnum.v2bf16(<2 x bfloat> %mul, <2 x bfloat> <bfloat 0.0, bfloat 0.0>)
+  %clamp = call <2 x bfloat> @llvm.minnum.v2bf16(<2 x bfloat> %max, <2 x bfloat> <bfloat 1.0, bfloat 1.0>)
+  %ret = bitcast <2 x bfloat> %clamp to float
+  ret float %ret
+}
+
+define amdgpu_ps void @v_test_mul_add_v2bf16_vvv(ptr addrspace(1) %out, <2 x bfloat> %a, <2 x bfloat> %b, <2 x bfloat> %c) {
+; GCN-LABEL: v_test_mul_add_v2bf16_vvv:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    v_pk_mul_bf16 v2, v2, v3
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GCN-NEXT:    v_pk_add_bf16 v2, v2, v4
+; GCN-NEXT:    global_store_b32 v[0:1], v2, off
+; GCN-NEXT:    s_endpgm
+  %mul = fmul contract <2 x bfloat> %a, %b
+  %add = fadd contract <2 x bfloat> %mul, %c
+  store <2 x bfloat> %add, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_ps void @v_test_mul_add_v2bf16_vss(ptr addrspace(1) %out, <2 x bfloat> %a, <2 x bfloat> inreg %b, <2 x bfloat> inreg %c) {
+; GCN-LABEL: v_test_mul_add_v2bf16_vss:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    v_pk_mul_bf16 v2, v2, s0
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GCN-NEXT:    v_pk_add_bf16 v2, v2, s1
+; GCN-NEXT:    global_store_b32 v[0:1], v2, off
+; GCN-NEXT:    s_endpgm
+  %mul = fmul contract <2 x bfloat> %a, %b
+  %add = fadd contract <2 x bfloat> %mul, %c
+  store <2 x bfloat> %add, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_ps void @v_test_mul_add_v2bf16_sss(ptr addrspace(1) %out, <2 x bfloat> inreg %a, <2 x bfloat> inreg %b, <2 x bfloat> inreg %c) {
+; GCN-LABEL: v_test_mul_add_v2bf16_sss:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    v_pk_mul_bf16 v2, s0, s1
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GCN-NEXT:    v_pk_add_bf16 v2, v2, s2
+; GCN-NEXT:    global_store_b32 v[0:1], v2, off
+; GCN-NEXT:    s_endpgm
+  %mul = fmul contract <2 x bfloat> %a, %b
+  %add = fadd contract <2 x bfloat> %mul, %c
+  store <2 x bfloat> %add, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_ps void @v_test_mul_add_v2bf16_vsc(ptr addrspace(1) %out, <2 x bfloat> %a, <2 x bfloat> inreg %b) {
+; GCN-LABEL: v_test_mul_add_v2bf16_vsc:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    v_pk_mul_bf16 v2, v2, s0
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GCN-NEXT:    v_pk_add_bf16 v2, v2, 0.5 op_sel_hi:[1,0]
+; GCN-NEXT:    global_store_b32 v[0:1], v2, off
+; GCN-NEXT:    s_endpgm
+  %mul = fmul contract <2 x bfloat> %a, %b
+  %add = fadd contract <2 x bfloat> %mul, <bfloat 0.5, bfloat 0.5>
+  store <2 x bfloat> %add, ptr addrspace(1) %out
+  ret void
+}
+
+define amdgpu_ps void @v_test_mul_add_v2bf16_vll(ptr addrspace(1) %out, <2 x bfloat> %a) {
+; GCN-LABEL: v_test_mul_add_v2bf16_vll:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    v_pk_mul_bf16 v2, 0x42c83f80, v2
+; GCN-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GCN-NEXT:    v_pk_add_bf16 v2, 0x43484000, v2
+; GCN-NEXT:    global_store_b32 v[0:1], v2, off
+; GCN-NEXT:    s_endpgm
+  %mul = fmul contract <2 x bfloat> %a, <bfloat 1.0, bfloat 100.0>
+  %add = fadd contract <2 x bfloat> %mul, <bfloat 2.0, bfloat 200.0>
+  store <2 x bfloat> %add, ptr addrspace(1) %out
+  ret void
+}
+
 define amdgpu_ps void @v_test_fma_v2bf16_vvv(ptr addrspace(1) %out, <2 x bfloat> %a, <2 x bfloat> %b, <2 x bfloat> %c) {
 ; GCN-LABEL: v_test_fma_v2bf16_vvv:
 ; GCN:       ; %bb.0:
@@ -426,6 +566,8 @@ define amdgpu_ps void @llvm_exp2_bf16_s(ptr addrspace(1) %out, bfloat inreg %src
   ret void
 }
 
+declare bfloat @llvm.minnum.bf16(bfloat, bfloat)
+declare bfloat @llvm.maxnum.bf16(bfloat, bfloat)
 declare <2 x bfloat> @llvm.minnum.v2bf16(<2 x bfloat> %a, <2 x bfloat> %b)
 declare <2 x bfloat> @llvm.maxnum.v2bf16(<2 x bfloat> %a, <2 x bfloat> %b)
 declare <2 x bfloat> @llvm.fma.v2bf16(<2 x bfloat>, <2 x bfloat>, <2 x bfloat>)
diff --git a/llvm/test/CodeGen/AMDGPU/bf16.ll b/llvm/test/CodeGen/AMDGPU/bf16.ll
index 7859fcdf..52e697c 100644
--- a/llvm/test/CodeGen/AMDGPU/bf16.ll
+++ b/llvm/test/CodeGen/AMDGPU/bf16.ll
@@ -468,15 +468,28 @@ define <16 x bfloat> @v_load_global_v16bf16(ptr addrspace(1) %ptr) {
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
-; GFX9-LABEL: v_load_global_v16bf16:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v9, v1
-; GFX9-NEXT:    v_mov_b32_e32 v8, v0
-; GFX9-NEXT:    global_load_dwordx4 v[0:3], v[8:9], off
-; GFX9-NEXT:    global_load_dwordx4 v[4:7], v[8:9], off offset:16
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    s_setpc_b64 s[30:31]
+; GFX900-LABEL: v_load_global_v16bf16:
+; GFX900:       ; %bb.0:
+; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX900-NEXT:    v_mov_b32_e32 v9, v1
+; GFX900-NEXT:    v_mov_b32_e32 v8, v0
+; GFX900-NEXT:    global_load_dwordx4 v[0:3], v[8:9], off
+; GFX900-NEXT:    global_load_dwordx4 v[4:7], v[8:9], off offset:16
+; GFX900-NEXT:    s_waitcnt vmcnt(0)
+; GFX900-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX950-LABEL: v_load_global_v16bf16:
+; GFX950:       ; %bb.0:
+; GFX950-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX950-NEXT:    global_load_dwordx4 v[8:11], v[0:1], off
+; GFX950-NEXT:    global_load_dwordx4 v[4:7], v[0:1], off offset:16
+; GFX950-NEXT:    s_waitcnt vmcnt(1)
+; GFX950-NEXT:    v_mov_b32_e32 v0, v8
+; GFX950-NEXT:    v_mov_b32_e32 v1, v9
+; GFX950-NEXT:    v_mov_b32_e32 v2, v10
+; GFX950-NEXT:    v_mov_b32_e32 v3, v11
+; GFX950-NEXT:    s_waitcnt vmcnt(0)
+; GFX950-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-LABEL: v_load_global_v16bf16:
 ; GFX10:       ; %bb.0:
@@ -619,17 +632,32 @@ define <32 x bfloat> @v_load_global_v32bf16(ptr addrspace(1) %ptr) {
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
-; GFX9-LABEL: v_load_global_v32bf16:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v17, v1
-; GFX9-NEXT:    v_mov_b32_e32 v16, v0
-; GFX9-NEXT:    global_load_dwordx4 v[0:3], v[16:17], off
-; GFX9-NEXT:    global_load_dwordx4 v[4:7], v[16:17], off offset:16
-; GFX9-NEXT:    global_load_dwordx4 v[8:11], v[16:17], off offset:32
-; GFX9-NEXT:    global_load_dwordx4 v[12:15], v[16:17], off offset:48
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    s_setpc_b64 s[30:31]
+; GFX900-LABEL: v_load_global_v32bf16:
+; GFX900:       ; %bb.0:
+; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX900-NEXT:    v_mov_b32_e32 v17, v1
+; GFX900-NEXT:    v_mov_b32_e32 v16, v0
+; GFX900-NEXT:    global_load_dwordx4 v[0:3], v[16:17], off
+; GFX900-NEXT:    global_load_dwordx4 v[4:7], v[16:17], off offset:16
+; GFX900-NEXT:    global_load_dwordx4 v[8:11], v[16:17], off offset:32
+; GFX900-NEXT:    global_load_dwordx4 v[12:15], v[16:17], off offset:48
+; GFX900-NEXT:    s_waitcnt vmcnt(0)
+; GFX900-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX950-LABEL: v_load_global_v32bf16:
+; GFX950:       ; %bb.0:
+; GFX950-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX950-NEXT:    global_load_dwordx4 v[16:19], v[0:1], off
+; GFX950-NEXT:    global_load_dwordx4 v[4:7], v[0:1], off offset:16
+; GFX950-NEXT:    global_load_dwordx4 v[8:11], v[0:1], off offset:32
+; GFX950-NEXT:    global_load_dwordx4 v[12:15], v[0:1], off offset:48
+; GFX950-NEXT:    s_waitcnt vmcnt(3)
+; GFX950-NEXT:    v_mov_b32_e32 v0, v16
+; GFX950-NEXT:    v_mov_b32_e32 v1, v17
+; GFX950-NEXT:    v_mov_b32_e32 v2, v18
+; GFX950-NEXT:    v_mov_b32_e32 v3, v19
+; GFX950-NEXT:    s_waitcnt vmcnt(0)
+; GFX950-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-LABEL: v_load_global_v32bf16:
 ; GFX10:       ; %bb.0:
@@ -877,22 +905,41 @@ define <64 x bfloat> @v_load_global_v64bf16(ptr addrspace(1) %ptr) {
 ; GFX8-NEXT:    s_waitcnt vmcnt(0)
 ; GFX8-NEXT:    s_setpc_b64 s[30:31]
 ;
-; GFX9-LABEL: v_load_global_v64bf16:
-; GFX9:       ; %bb.0:
-; GFX9-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX9-NEXT:    v_mov_b32_e32 v29, v1
-; GFX9-NEXT:    v_mov_b32_e32 v28, v0
-; GFX9-NEXT:    global_load_dwordx4 v[0:3], v[28:29], off
-; GFX9-NEXT:    global_load_dwordx4 v[4:7], v[28:29], off offset:16
-; GFX9-NEXT:    global_load_dwordx4 v[8:11], v[28:29], off offset:32
-; GFX9-NEXT:    global_load_dwordx4 v[12:15], v[28:29], off offset:48
-; GFX9-NEXT:    global_load_dwordx4 v[16:19], v[28:29], off offset:64
-; GFX9-NEXT:    global_load_dwordx4 v[20:23], v[28:29], off offset:80
-; GFX9-NEXT:    global_load_dwordx4 v[24:27], v[28:29], off offset:96
-; GFX9-NEXT:    s_nop 0
-; GFX9-NEXT:    global_load_dwordx4 v[28:31], v[28:29], off offset:112
-; GFX9-NEXT:    s_waitcnt vmcnt(0)
-; GFX9-NEXT:    s_setpc_b64 s[30:31]
+; GFX900-LABEL: v_load_global_v64bf16:
+; GFX900:       ; %bb.0:
+; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX900-NEXT:    v_mov_b32_e32 v29, v1
+; GFX900-NEXT:    v_mov_b32_e32 v28, v0
+; GFX900-NEXT:    global_load_dwordx4 v[0:3], v[28:29], off
+; GFX900-NEXT:    global_load_dwordx4 v[4:7], v[28:29], off offset:16
+; GFX900-NEXT:    global_load_dwordx4 v[8:11], v[28:29], off offset:32
+; GFX900-NEXT:    global_load_dwordx4 v[12:15], v[28:29], off offset:48
+; GFX900-NEXT:    global_load_dwordx4 v[16:19], v[28:29], off offset:64
+; GFX900-NEXT:    global_load_dwordx4 v[20:23], v[28:29], off offset:80
+; GFX900-NEXT:    global_load_dwordx4 v[24:27], v[28:29], off offset:96
+; GFX900-NEXT:    s_nop 0
+; GFX900-NEXT:    global_load_dwordx4 v[28:31], v[28:29], off offset:112
+; GFX900-NEXT:    s_waitcnt vmcnt(0)
+; GFX900-NEXT:    s_setpc_b64 s[30:31]
+;
+; GFX950-LABEL: v_load_global_v64bf16:
+; GFX950:       ; %bb.0:
+; GFX950-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX950-NEXT:    global_load_dwordx4 v[32:35], v[0:1], off
+; GFX950-NEXT:    global_load_dwordx4 v[4:7], v[0:1], off offset:16
+; GFX950-NEXT:    global_load_dwordx4 v[8:11], v[0:1], off offset:32
+; GFX950-NEXT:    global_load_dwordx4 v[12:15], v[0:1], off offset:48
+; GFX950-NEXT:    global_load_dwordx4 v[16:19], v[0:1], off offset:64
+; GFX950-NEXT:    global_load_dwordx4 v[20:23], v[0:1], off offset:80
+; GFX950-NEXT:    global_load_dwordx4 v[24:27], v[0:1], off offset:96
+; GFX950-NEXT:    global_load_dwordx4 v[28:31], v[0:1], off offset:112
+; GFX950-NEXT:    s_waitcnt vmcnt(7)
+; GFX950-NEXT:    v_mov_b32_e32 v0, v32
+; GFX950-NEXT:    v_mov_b32_e32 v1, v33
+; GFX950-NEXT:    v_mov_b32_e32 v2, v34
+; GFX950-NEXT:    v_mov_b32_e32 v3, v35
+; GFX950-NEXT:    s_waitcnt vmcnt(0)
+; GFX950-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-LABEL: v_load_global_v64bf16:
 ; GFX10:       ; %bb.0:
diff --git a/llvm/test/CodeGen/AMDGPU/build_vector.ll b/llvm/test/CodeGen/AMDGPU/build_vector.ll
index 7208eae..763f436 100644
--- a/llvm/test/CodeGen/AMDGPU/build_vector.ll
+++ b/llvm/test/CodeGen/AMDGPU/build_vector.ll
@@ -51,11 +51,11 @@ define amdgpu_kernel void @build_vector2 (ptr addrspace(1) %out) {
 ; GFX942-LABEL: build_vector2:
 ; GFX942:       ; %bb.0: ; %entry
 ; GFX942-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
-; GFX942-NEXT:    v_mov_b32_e32 v0, 5
-; GFX942-NEXT:    v_mov_b32_e32 v1, 6
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-NEXT:    v_mov_b32_e32 v2, 5
+; GFX942-NEXT:    v_mov_b32_e32 v3, 6
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[0:1]
 ; GFX942-NEXT:    s_endpgm
 entry:
   store <2 x i32> <i32 5, i32 6>, ptr addrspace(1) %out
@@ -116,13 +116,13 @@ define amdgpu_kernel void @build_vector4 (ptr addrspace(1) %out) {
 ; GFX942-LABEL: build_vector4:
 ; GFX942:       ; %bb.0: ; %entry
 ; GFX942-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
-; GFX942-NEXT:    v_mov_b32_e32 v4, 0
-; GFX942-NEXT:    v_mov_b32_e32 v0, 5
-; GFX942-NEXT:    v_mov_b32_e32 v1, 6
-; GFX942-NEXT:    v_mov_b32_e32 v2, 7
-; GFX942-NEXT:    v_mov_b32_e32 v3, 8
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-NEXT:    v_mov_b32_e32 v2, 5
+; GFX942-NEXT:    v_mov_b32_e32 v3, 6
+; GFX942-NEXT:    v_mov_b32_e32 v4, 7
+; GFX942-NEXT:    v_mov_b32_e32 v5, 8
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX942-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1]
 ; GFX942-NEXT:    s_endpgm
 entry:
   store <4 x i32> <i32 5, i32 6, i32 7, i32 8>, ptr addrspace(1) %out
@@ -307,13 +307,13 @@ define amdgpu_kernel void @build_v2i32_from_v4i16_shuffle(ptr addrspace(1) %out,
 ; GFX942-LABEL: build_v2i32_from_v4i16_shuffle:
 ; GFX942:       ; %bb.0: ; %entry
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-NEXT:    s_lshl_b32 s3, s3, 16
 ; GFX942-NEXT:    s_lshl_b32 s2, s2, 16
-; GFX942-NEXT:    v_mov_b32_e32 v0, s2
-; GFX942-NEXT:    v_mov_b32_e32 v1, s3
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX942-NEXT:    v_mov_b32_e32 v2, s2
+; GFX942-NEXT:    v_mov_b32_e32 v3, s3
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[0:1]
 ; GFX942-NEXT:    s_endpgm
 entry:
   %shuf = shufflevector <4 x i16> %in, <4 x i16> zeroinitializer, <2 x i32> <i32 0, i32 2>
diff --git a/llvm/test/CodeGen/AMDGPU/fix-crash-valu-hazard.ll b/llvm/test/CodeGen/AMDGPU/fix-crash-valu-hazard.ll
index 8781196..4f752d1 100644
--- a/llvm/test/CodeGen/AMDGPU/fix-crash-valu-hazard.ll
+++ b/llvm/test/CodeGen/AMDGPU/fix-crash-valu-hazard.ll
@@ -8,10 +8,10 @@ define amdgpu_ps void @global_load_lds_dword_saddr(ptr addrspace(1) inreg nocapt
 ; GFX942-LABEL: global_load_lds_dword_saddr:
 ; GFX942:       ; %bb.0: ; %main_body
 ; GFX942-NEXT:    v_readfirstlane_b32 s2, v0
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b32_e32 v1, 0
 ; GFX942-NEXT:    s_mov_b32 m0, s2
 ; GFX942-NEXT:    s_nop 0
-; GFX942-NEXT:    global_load_lds_dword v2, s[0:1] offset:32 nt
+; GFX942-NEXT:    global_load_lds_dword v1, s[0:1] offset:32 nt
 ; GFX942-NEXT:    s_getpc_b64 s[0:1]
 ; GFX942-NEXT:    s_add_u32 s0, s0, G@gotpcrel32@lo+4
 ; GFX942-NEXT:    s_addc_u32 s1, s1, G@gotpcrel32@hi+12
@@ -21,9 +21,9 @@ define amdgpu_ps void @global_load_lds_dword_saddr(ptr addrspace(1) inreg nocapt
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-NEXT:    s_mul_i32 s3, s3, 10
 ; GFX942-NEXT:    s_mul_i32 s2, s2, 10
-; GFX942-NEXT:    v_mov_b32_e32 v0, s2
-; GFX942-NEXT:    v_mov_b32_e32 v1, s3
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX942-NEXT:    v_mov_b32_e32 v2, s2
+; GFX942-NEXT:    v_mov_b32_e32 v3, s3
+; GFX942-NEXT:    global_store_dwordx2 v1, v[2:3], s[0:1]
 ; GFX942-NEXT:    s_endpgm
 ;
 ; GFX90A-LABEL: global_load_lds_dword_saddr:
diff --git a/llvm/test/CodeGen/AMDGPU/fp64-atomics-gfx90a.ll b/llvm/test/CodeGen/AMDGPU/fp64-atomics-gfx90a.ll
index 873fcee..6067194 100644
--- a/llvm/test/CodeGen/AMDGPU/fp64-atomics-gfx90a.ll
+++ b/llvm/test/CodeGen/AMDGPU/fp64-atomics-gfx90a.ll
@@ -71,12 +71,12 @@ define amdgpu_kernel void @raw_buffer_atomic_add_rtn_f64_off4_slc(<4 x i32> %rsr
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_add_f64 v[0:1], v2, s[0:3], 4 offen glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_add_f64 v[2:3], v0, s[0:3], 4 offen glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: raw_buffer_atomic_add_rtn_f64_off4_slc:
@@ -86,12 +86,12 @@ define amdgpu_kernel void @raw_buffer_atomic_add_rtn_f64_off4_slc(<4 x i32> %rsr
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_add_f64 v[0:1], v2, s[0:3], 4 offen sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_add_f64 v[2:3], v0, s[0:3], 4 offen sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.raw.buffer.atomic.fadd.f64(double %data, <4 x i32> %rsrc, i32 %vindex, i32 4, i32 2)
@@ -154,12 +154,12 @@ define amdgpu_kernel void @raw_ptr_buffer_atomic_add_rtn_f64_off4_slc(ptr addrsp
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_add_f64 v[0:1], v2, s[0:3], 4 offen glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_add_f64 v[2:3], v0, s[0:3], 4 offen glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: raw_ptr_buffer_atomic_add_rtn_f64_off4_slc:
@@ -169,12 +169,12 @@ define amdgpu_kernel void @raw_ptr_buffer_atomic_add_rtn_f64_off4_slc(ptr addrsp
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_add_f64 v[0:1], v2, s[0:3], 4 offen sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_add_f64 v[2:3], v0, s[0:3], 4 offen sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.raw.ptr.buffer.atomic.fadd.f64(double %data, ptr addrspace(8) %rsrc, i32 %vindex, i32 4, i32 2)
@@ -237,12 +237,12 @@ define amdgpu_kernel void @struct_buffer_atomic_add_rtn_f64_off4_slc(<4 x i32> %
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_add_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_add_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: struct_buffer_atomic_add_rtn_f64_off4_slc:
@@ -252,12 +252,12 @@ define amdgpu_kernel void @struct_buffer_atomic_add_rtn_f64_off4_slc(<4 x i32> %
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_add_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_add_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.struct.buffer.atomic.fadd.f64(double %data, <4 x i32> %rsrc, i32 %vindex, i32 4, i32 0, i32 2)
@@ -320,12 +320,12 @@ define amdgpu_kernel void @struct_ptr_buffer_atomic_add_rtn_f64_off4_slc(ptr add
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_add_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_add_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: struct_ptr_buffer_atomic_add_rtn_f64_off4_slc:
@@ -335,12 +335,12 @@ define amdgpu_kernel void @struct_ptr_buffer_atomic_add_rtn_f64_off4_slc(ptr add
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_add_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_add_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.struct.ptr.buffer.atomic.fadd.f64(double %data, ptr addrspace(8) %rsrc, i32 %vindex, i32 4, i32 0, i32 2)
@@ -403,12 +403,12 @@ define amdgpu_kernel void @raw_buffer_atomic_min_rtn_f64_off4_slc(<4 x i32> %rsr
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_min_f64 v[0:1], v2, s[0:3], 4 offen glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_min_f64 v[2:3], v0, s[0:3], 4 offen glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: raw_buffer_atomic_min_rtn_f64_off4_slc:
@@ -418,12 +418,12 @@ define amdgpu_kernel void @raw_buffer_atomic_min_rtn_f64_off4_slc(<4 x i32> %rsr
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_min_f64 v[0:1], v2, s[0:3], 4 offen sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_min_f64 v[2:3], v0, s[0:3], 4 offen sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.raw.buffer.atomic.fmin.f64(double %data, <4 x i32> %rsrc, i32 %vindex, i32 4, i32 2)
@@ -486,12 +486,12 @@ define amdgpu_kernel void @raw_ptr_buffer_atomic_min_rtn_f64_off4_slc(ptr addrsp
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_min_f64 v[0:1], v2, s[0:3], 4 offen glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_min_f64 v[2:3], v0, s[0:3], 4 offen glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: raw_ptr_buffer_atomic_min_rtn_f64_off4_slc:
@@ -501,12 +501,12 @@ define amdgpu_kernel void @raw_ptr_buffer_atomic_min_rtn_f64_off4_slc(ptr addrsp
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_min_f64 v[0:1], v2, s[0:3], 4 offen sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_min_f64 v[2:3], v0, s[0:3], 4 offen sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.raw.ptr.buffer.atomic.fmin.f64(double %data, ptr addrspace(8) %rsrc, i32 %vindex, i32 4, i32 2)
@@ -569,12 +569,12 @@ define amdgpu_kernel void @struct_buffer_atomic_min_rtn_f64_off4_slc(<4 x i32> %
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_min_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_min_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: struct_buffer_atomic_min_rtn_f64_off4_slc:
@@ -584,12 +584,12 @@ define amdgpu_kernel void @struct_buffer_atomic_min_rtn_f64_off4_slc(<4 x i32> %
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_min_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_min_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.struct.buffer.atomic.fmin.f64(double %data, <4 x i32> %rsrc, i32 %vindex, i32 4, i32 0, i32 2)
@@ -652,12 +652,12 @@ define amdgpu_kernel void @struct_ptr_buffer_atomic_min_rtn_f64_off4_slc(ptr add
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_min_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_min_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: struct_ptr_buffer_atomic_min_rtn_f64_off4_slc:
@@ -667,12 +667,12 @@ define amdgpu_kernel void @struct_ptr_buffer_atomic_min_rtn_f64_off4_slc(ptr add
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_min_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_min_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.struct.ptr.buffer.atomic.fmin.f64(double %data, ptr addrspace(8) %rsrc, i32 %vindex, i32 4, i32 0, i32 2)
@@ -735,12 +735,12 @@ define amdgpu_kernel void @raw_buffer_atomic_max_rtn_f64_off4_slc(<4 x i32> %rsr
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_max_f64 v[0:1], v2, s[0:3], 4 offen glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_max_f64 v[2:3], v0, s[0:3], 4 offen glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: raw_buffer_atomic_max_rtn_f64_off4_slc:
@@ -750,12 +750,12 @@ define amdgpu_kernel void @raw_buffer_atomic_max_rtn_f64_off4_slc(<4 x i32> %rsr
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_max_f64 v[0:1], v2, s[0:3], 4 offen sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_max_f64 v[2:3], v0, s[0:3], 4 offen sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.raw.buffer.atomic.fmax.f64(double %data, <4 x i32> %rsrc, i32 %vindex, i32 4, i32 2)
@@ -818,12 +818,12 @@ define amdgpu_kernel void @raw_ptr_buffer_atomic_max_rtn_f64_off4_slc(ptr addrsp
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_max_f64 v[0:1], v2, s[0:3], 4 offen glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_max_f64 v[2:3], v0, s[0:3], 4 offen glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: raw_ptr_buffer_atomic_max_rtn_f64_off4_slc:
@@ -833,12 +833,12 @@ define amdgpu_kernel void @raw_ptr_buffer_atomic_max_rtn_f64_off4_slc(ptr addrsp
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_max_f64 v[0:1], v2, s[0:3], 4 offen sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_max_f64 v[2:3], v0, s[0:3], 4 offen sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.raw.ptr.buffer.atomic.fmax.f64(double %data, ptr addrspace(8) %rsrc, i32 %vindex, i32 4, i32 2)
@@ -901,12 +901,12 @@ define amdgpu_kernel void @struct_buffer_atomic_max_rtn_f64_off4_slc(<4 x i32> %
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_max_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_max_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: struct_buffer_atomic_max_rtn_f64_off4_slc:
@@ -916,12 +916,12 @@ define amdgpu_kernel void @struct_buffer_atomic_max_rtn_f64_off4_slc(<4 x i32> %
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_max_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_max_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.struct.buffer.atomic.fmax.f64(double %data, <4 x i32> %rsrc, i32 %vindex, i32 4, i32 0, i32 2)
@@ -984,12 +984,12 @@ define amdgpu_kernel void @struct_ptr_buffer_atomic_max_rtn_f64_off4_slc(ptr add
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX90A-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
-; GFX90A-NEXT:    buffer_atomic_max_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 glc slc
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, s10
+; GFX90A-NEXT:    buffer_atomic_max_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 glc slc
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: struct_ptr_buffer_atomic_max_rtn_f64_off4_slc:
@@ -999,12 +999,12 @@ define amdgpu_kernel void @struct_ptr_buffer_atomic_max_rtn_f64_off4_slc(ptr add
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[8:9], s[4:5], 0x44
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    buffer_atomic_max_f64 v[0:1], v2, s[0:3], 0 idxen offset:4 sc0 nt
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-NEXT:    buffer_atomic_max_f64 v[2:3], v0, s[0:3], 0 idxen offset:4 sc0 nt
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[8:9]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[8:9]
 ; GFX942-NEXT:    s_endpgm
 main_body:
   %ret = call double @llvm.amdgcn.struct.ptr.buffer.atomic.fmax.f64(double %data, ptr addrspace(8) %rsrc, i32 %vindex, i32 4, i32 0, i32 2)
diff --git a/llvm/test/CodeGen/AMDGPU/gfx1250-scratch-scope-se.ll b/llvm/test/CodeGen/AMDGPU/gfx1250-scratch-scope-se.ll
new file mode 100644
index 0000000..d1e82a0
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/gfx1250-scratch-scope-se.ll
@@ -0,0 +1,95 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1250 < %s | FileCheck -check-prefixes=GCN,GCN-SDAG %s
+; RUN: llc -global-isel -mtriple=amdgcn-amd-amdhsa -mcpu=gfx1250 < %s | FileCheck -check-prefixes=GCN,GCN-GISEL %s
+
+; Test that stores that may hit scratch are correctly promoted to SCOPE_SE.
+
+define void @test_scratch_store(ptr addrspace(5) %ptr, i32 %val) {
+; GCN-LABEL: test_scratch_store:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GCN-NEXT:    s_wait_kmcnt 0x0
+; GCN-NEXT:    scratch_store_b32 v0, v1, off scope:SCOPE_SE
+; GCN-NEXT:    s_set_pc_i64 s[30:31]
+    store i32 %val, ptr addrspace(5) %ptr
+    ret void
+}
+
+define void @test_unknown_flat_store(ptr %ptr, i32 %val) {
+; GCN-LABEL: test_unknown_flat_store:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GCN-NEXT:    s_wait_kmcnt 0x0
+; GCN-NEXT:    flat_store_b32 v[0:1], v2 scope:SCOPE_SE
+; GCN-NEXT:    s_wait_dscnt 0x0
+; GCN-NEXT:    s_set_pc_i64 s[30:31]
+    store i32 %val, ptr %ptr
+    ret void
+}
+
+define void @test_flat_store_no_scratch_alloc(ptr %ptr, i32 %val) #0 {
+; GCN-LABEL: test_flat_store_no_scratch_alloc:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GCN-NEXT:    s_wait_kmcnt 0x0
+; GCN-NEXT:    flat_store_b32 v[0:1], v2
+; GCN-NEXT:    s_wait_dscnt 0x0
+; GCN-NEXT:    s_set_pc_i64 s[30:31]
+    store i32 %val, ptr %ptr
+    ret void
+}
+
+; TODO: handle
+define void @test_flat_store_noalias_addrspace(ptr %ptr, i32 %val) {
+; GCN-LABEL: test_flat_store_noalias_addrspace:
+; GCN:       ; %bb.0:
+; GCN-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GCN-NEXT:    s_wait_kmcnt 0x0
+; GCN-NEXT:    flat_store_b32 v[0:1], v2 scope:SCOPE_SE
+; GCN-NEXT:    s_wait_dscnt 0x0
+; GCN-NEXT:    s_set_pc_i64 s[30:31]
+    store i32 %val, ptr %ptr, !noalias.addrspace !{i32 5, i32 6}
+    ret void
+}
+
+; TODO: would be nice to handle too
+define void @test_flat_store_select(ptr addrspace(1) %a, ptr addrspace(3) %b, i1 %cond, i32 %val) {
+; GCN-SDAG-LABEL: test_flat_store_select:
+; GCN-SDAG:       ; %bb.0:
+; GCN-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GCN-SDAG-NEXT:    s_wait_kmcnt 0x0
+; GCN-SDAG-NEXT:    v_cmp_ne_u32_e32 vcc_lo, -1, v2
+; GCN-SDAG-NEXT:    v_and_b32_e32 v3, 1, v3
+; GCN-SDAG-NEXT:    s_mov_b64 s[0:1], src_shared_base
+; GCN-SDAG-NEXT:    v_cndmask_b32_e32 v2, 0, v2, vcc_lo
+; GCN-SDAG-NEXT:    v_cndmask_b32_e64 v5, 0, s1, vcc_lo
+; GCN-SDAG-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GCN-SDAG-NEXT:    v_cmp_eq_u32_e32 vcc_lo, 1, v3
+; GCN-SDAG-NEXT:    v_dual_cndmask_b32 v1, v5, v1 :: v_dual_cndmask_b32 v0, v2, v0
+; GCN-SDAG-NEXT:    flat_store_b32 v[0:1], v4 scope:SCOPE_SE
+; GCN-SDAG-NEXT:    s_wait_dscnt 0x0
+; GCN-SDAG-NEXT:    s_set_pc_i64 s[30:31]
+;
+; GCN-GISEL-LABEL: test_flat_store_select:
+; GCN-GISEL:       ; %bb.0:
+; GCN-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
+; GCN-GISEL-NEXT:    s_wait_kmcnt 0x0
+; GCN-GISEL-NEXT:    v_cmp_ne_u32_e32 vcc_lo, -1, v2
+; GCN-GISEL-NEXT:    v_and_b32_e32 v3, 1, v3
+; GCN-GISEL-NEXT:    s_mov_b64 s[0:1], src_shared_base
+; GCN-GISEL-NEXT:    v_cndmask_b32_e32 v2, 0, v2, vcc_lo
+; GCN-GISEL-NEXT:    v_cndmask_b32_e64 v5, 0, s1, vcc_lo
+; GCN-GISEL-NEXT:    s_delay_alu instid0(VALU_DEP_3) | instskip(NEXT) | instid1(VALU_DEP_2)
+; GCN-GISEL-NEXT:    v_cmp_ne_u32_e32 vcc_lo, 0, v3
+; GCN-GISEL-NEXT:    v_dual_cndmask_b32 v0, v2, v0 :: v_dual_cndmask_b32 v1, v5, v1
+; GCN-GISEL-NEXT:    flat_store_b32 v[0:1], v4 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    s_wait_dscnt 0x0
+; GCN-GISEL-NEXT:    s_set_pc_i64 s[30:31]
+    %a.ascast = addrspacecast ptr addrspace(1) %a to ptr
+    %b.ascast = addrspacecast ptr addrspace(3) %b to ptr
+    %ptr = select i1 %cond, ptr %a.ascast, ptr %b.ascast
+    store i32 %val, ptr %ptr
+    ret void
+}
+
+attributes #0 = { "amdgpu-no-flat-scratch-init" }
diff --git a/llvm/test/CodeGen/AMDGPU/global-load-xcnt.ll b/llvm/test/CodeGen/AMDGPU/global-load-xcnt.ll
index fd644a3..3a898a9 100644
--- a/llvm/test/CodeGen/AMDGPU/global-load-xcnt.ll
+++ b/llvm/test/CodeGen/AMDGPU/global-load-xcnt.ll
@@ -124,27 +124,27 @@ define i32 @test_v64i32_load_store(ptr addrspace(1) %ptr, i32 %idx, ptr addrspac
 ; GCN-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GCN-SDAG-NEXT:    s_wait_kmcnt 0x0
 ; GCN-SDAG-NEXT:    s_clause 0xd
-; GCN-SDAG-NEXT:    scratch_store_b32 off, v40, s32 offset:52
-; GCN-SDAG-NEXT:    scratch_store_b32 off, v41, s32 offset:48
-; GCN-SDAG-NEXT:    scratch_store_b32 off, v42, s32 offset:44
-; GCN-SDAG-NEXT:    scratch_store_b32 off, v43, s32 offset:40
-; GCN-SDAG-NEXT:    scratch_store_b32 off, v44, s32 offset:36
-; GCN-SDAG-NEXT:    scratch_store_b32 off, v45, s32 offset:32
-; GCN-SDAG-NEXT:    scratch_store_b32 off, v56, s32 offset:28
-; GCN-SDAG-NEXT:    scratch_store_b32 off, v57, s32 offset:24
-; GCN-SDAG-NEXT:    scratch_store_b32 off, v58, s32 offset:20
-; GCN-SDAG-NEXT:    scratch_store_b32 off, v59, s32 offset:16
-; GCN-SDAG-NEXT:    scratch_store_b32 off, v60, s32 offset:12
-; GCN-SDAG-NEXT:    scratch_store_b32 off, v61, s32 offset:8
-; GCN-SDAG-NEXT:    scratch_store_b32 off, v62, s32 offset:4
-; GCN-SDAG-NEXT:    scratch_store_b32 off, v63, s32
+; GCN-SDAG-NEXT:    scratch_store_b32 off, v40, s32 offset:52 scope:SCOPE_SE
+; GCN-SDAG-NEXT:    scratch_store_b32 off, v41, s32 offset:48 scope:SCOPE_SE
+; GCN-SDAG-NEXT:    scratch_store_b32 off, v42, s32 offset:44 scope:SCOPE_SE
+; GCN-SDAG-NEXT:    scratch_store_b32 off, v43, s32 offset:40 scope:SCOPE_SE
+; GCN-SDAG-NEXT:    scratch_store_b32 off, v44, s32 offset:36 scope:SCOPE_SE
+; GCN-SDAG-NEXT:    scratch_store_b32 off, v45, s32 offset:32 scope:SCOPE_SE
+; GCN-SDAG-NEXT:    scratch_store_b32 off, v56, s32 offset:28 scope:SCOPE_SE
+; GCN-SDAG-NEXT:    scratch_store_b32 off, v57, s32 offset:24 scope:SCOPE_SE
+; GCN-SDAG-NEXT:    scratch_store_b32 off, v58, s32 offset:20 scope:SCOPE_SE
+; GCN-SDAG-NEXT:    scratch_store_b32 off, v59, s32 offset:16 scope:SCOPE_SE
+; GCN-SDAG-NEXT:    scratch_store_b32 off, v60, s32 offset:12 scope:SCOPE_SE
+; GCN-SDAG-NEXT:    scratch_store_b32 off, v61, s32 offset:8 scope:SCOPE_SE
+; GCN-SDAG-NEXT:    scratch_store_b32 off, v62, s32 offset:4 scope:SCOPE_SE
+; GCN-SDAG-NEXT:    scratch_store_b32 off, v63, s32 scope:SCOPE_SE
 ; GCN-SDAG-NEXT:    global_load_b128 v[6:9], v[0:1], off offset:224
 ; GCN-SDAG-NEXT:    v_dual_mov_b32 v5, v4 :: v_dual_mov_b32 v4, v3
 ; GCN-SDAG-NEXT:    s_wait_loadcnt 0x0
-; GCN-SDAG-NEXT:    scratch_store_b128 off, v[6:9], s32 offset:56 ; 16-byte Folded Spill
+; GCN-SDAG-NEXT:    scratch_store_b128 off, v[6:9], s32 offset:56 scope:SCOPE_SE ; 16-byte Folded Spill
 ; GCN-SDAG-NEXT:    global_load_b128 v[6:9], v[0:1], off offset:240
 ; GCN-SDAG-NEXT:    s_wait_loadcnt 0x0
-; GCN-SDAG-NEXT:    scratch_store_b128 off, v[6:9], s32 offset:72 ; 16-byte Folded Spill
+; GCN-SDAG-NEXT:    scratch_store_b128 off, v[6:9], s32 offset:72 scope:SCOPE_SE ; 16-byte Folded Spill
 ; GCN-SDAG-NEXT:    s_clause 0xd
 ; GCN-SDAG-NEXT:    global_load_b128 v[10:13], v[0:1], off offset:192
 ; GCN-SDAG-NEXT:    global_load_b128 v[14:17], v[0:1], off offset:208
@@ -206,27 +206,27 @@ define i32 @test_v64i32_load_store(ptr addrspace(1) %ptr, i32 %idx, ptr addrspac
 ; GCN-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GCN-GISEL-NEXT:    s_wait_kmcnt 0x0
 ; GCN-GISEL-NEXT:    s_clause 0xf
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v40, s32 offset:60
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v41, s32 offset:56
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v42, s32 offset:52
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v43, s32 offset:48
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v44, s32 offset:44
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v45, s32 offset:40
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v46, s32 offset:36
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v47, s32 offset:32
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v56, s32 offset:28
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v57, s32 offset:24
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v58, s32 offset:20
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v59, s32 offset:16
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v60, s32 offset:12
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v61, s32 offset:8
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v62, s32 offset:4
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v63, s32
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v40, s32 offset:60 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v41, s32 offset:56 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v42, s32 offset:52 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v43, s32 offset:48 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v44, s32 offset:44 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v45, s32 offset:40 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v46, s32 offset:36 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v47, s32 offset:32 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v56, s32 offset:28 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v57, s32 offset:24 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v58, s32 offset:20 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v59, s32 offset:16 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v60, s32 offset:12 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v61, s32 offset:8 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v62, s32 offset:4 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v63, s32 scope:SCOPE_SE
 ; GCN-GISEL-NEXT:    s_wait_xcnt 0x8
 ; GCN-GISEL-NEXT:    v_dual_mov_b32 v46, v3 :: v_dual_mov_b32 v47, v4
 ; GCN-GISEL-NEXT:    global_load_b128 v[2:5], v[0:1], off offset:32
 ; GCN-GISEL-NEXT:    s_wait_loadcnt 0x0
-; GCN-GISEL-NEXT:    scratch_store_b128 off, v[2:5], s32 offset:80 ; 16-byte Folded Spill
+; GCN-GISEL-NEXT:    scratch_store_b128 off, v[2:5], s32 offset:80 scope:SCOPE_SE ; 16-byte Folded Spill
 ; GCN-GISEL-NEXT:    s_clause 0xe
 ; GCN-GISEL-NEXT:    global_load_b128 v[6:9], v[0:1], off offset:48
 ; GCN-GISEL-NEXT:    global_load_b128 v[10:13], v[0:1], off offset:64
@@ -244,7 +244,7 @@ define i32 @test_v64i32_load_store(ptr addrspace(1) %ptr, i32 %idx, ptr addrspac
 ; GCN-GISEL-NEXT:    global_load_b128 v[60:63], v[0:1], off offset:16
 ; GCN-GISEL-NEXT:    global_load_b128 v[0:3], v[0:1], off offset:240
 ; GCN-GISEL-NEXT:    s_wait_loadcnt 0x0
-; GCN-GISEL-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:64 ; 16-byte Folded Spill
+; GCN-GISEL-NEXT:    scratch_store_b128 off, v[0:3], s32 offset:64 scope:SCOPE_SE ; 16-byte Folded Spill
 ; GCN-GISEL-NEXT:    scratch_load_b128 v[0:3], off, s32 offset:80 th:TH_LOAD_LU ; 16-byte Folded Reload
 ; GCN-GISEL-NEXT:    s_wait_loadcnt 0x0
 ; GCN-GISEL-NEXT:    s_clause 0xe
@@ -299,10 +299,10 @@ define i64 @test_v16i64_load_store(ptr addrspace(1) %ptr_a, ptr addrspace(1) %pt
 ; GCN-SDAG-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GCN-SDAG-NEXT:    s_wait_kmcnt 0x0
 ; GCN-SDAG-NEXT:    s_clause 0x3
-; GCN-SDAG-NEXT:    scratch_store_b32 off, v40, s32 offset:12
-; GCN-SDAG-NEXT:    scratch_store_b32 off, v41, s32 offset:8
-; GCN-SDAG-NEXT:    scratch_store_b32 off, v42, s32 offset:4
-; GCN-SDAG-NEXT:    scratch_store_b32 off, v43, s32
+; GCN-SDAG-NEXT:    scratch_store_b32 off, v40, s32 offset:12 scope:SCOPE_SE
+; GCN-SDAG-NEXT:    scratch_store_b32 off, v41, s32 offset:8 scope:SCOPE_SE
+; GCN-SDAG-NEXT:    scratch_store_b32 off, v42, s32 offset:4 scope:SCOPE_SE
+; GCN-SDAG-NEXT:    scratch_store_b32 off, v43, s32 scope:SCOPE_SE
 ; GCN-SDAG-NEXT:    s_clause 0x7
 ; GCN-SDAG-NEXT:    global_load_b128 v[10:13], v[0:1], off offset:112
 ; GCN-SDAG-NEXT:    global_load_b128 v[18:21], v[0:1], off offset:96
@@ -385,12 +385,12 @@ define i64 @test_v16i64_load_store(ptr addrspace(1) %ptr_a, ptr addrspace(1) %pt
 ; GCN-GISEL-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GCN-GISEL-NEXT:    s_wait_kmcnt 0x0
 ; GCN-GISEL-NEXT:    s_clause 0x5
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v40, s32 offset:20
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v41, s32 offset:16
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v42, s32 offset:12
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v43, s32 offset:8
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v44, s32 offset:4
-; GCN-GISEL-NEXT:    scratch_store_b32 off, v45, s32
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v40, s32 offset:20 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v41, s32 offset:16 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v42, s32 offset:12 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v43, s32 offset:8 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v44, s32 offset:4 scope:SCOPE_SE
+; GCN-GISEL-NEXT:    scratch_store_b32 off, v45, s32 scope:SCOPE_SE
 ; GCN-GISEL-NEXT:    s_clause 0x7
 ; GCN-GISEL-NEXT:    global_load_b128 v[6:9], v[0:1], off offset:80
 ; GCN-GISEL-NEXT:    global_load_b128 v[10:13], v[0:1], off
diff --git a/llvm/test/CodeGen/AMDGPU/infer-addrspace-flat-atomic.ll b/llvm/test/CodeGen/AMDGPU/infer-addrspace-flat-atomic.ll
index 258aa9e..0a493e51 100644
--- a/llvm/test/CodeGen/AMDGPU/infer-addrspace-flat-atomic.ll
+++ b/llvm/test/CodeGen/AMDGPU/infer-addrspace-flat-atomic.ll
@@ -8,15 +8,15 @@ define protected amdgpu_kernel void @InferNothing(i32 %a, ptr %b, double %c) {
 ; CHECK-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
 ; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
 ; CHECK-NEXT:    s_ashr_i32 s7, s6, 31
-; CHECK-NEXT:    v_mov_b32_e32 v0, s2
-; CHECK-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-NEXT:    v_mov_b32_e32 v2, s2
+; CHECK-NEXT:    v_mov_b32_e32 v3, s3
 ; CHECK-NEXT:    s_lshl_b64 s[2:3], s[6:7], 3
 ; CHECK-NEXT:    s_add_u32 s0, s2, s0
 ; CHECK-NEXT:    s_addc_u32 s1, s3, s1
-; CHECK-NEXT:    v_mov_b32_e32 v3, s1
-; CHECK-NEXT:    v_add_co_u32_e64 v2, vcc, -8, s0
-; CHECK-NEXT:    v_addc_co_u32_e32 v3, vcc, -1, v3, vcc
-; CHECK-NEXT:    flat_atomic_add_f64 v[2:3], v[0:1]
+; CHECK-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-NEXT:    v_add_co_u32_e64 v0, vcc, -8, s0
+; CHECK-NEXT:    v_addc_co_u32_e32 v1, vcc, -1, v1, vcc
+; CHECK-NEXT:    flat_atomic_add_f64 v[0:1], v[2:3]
 ; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; CHECK-NEXT:    buffer_wbinvl1_vol
 ; CHECK-NEXT:    s_endpgm
@@ -35,15 +35,15 @@ define protected amdgpu_kernel void @InferFadd(i32 %a, ptr addrspace(1) %b, doub
 ; CHECK-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
 ; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
 ; CHECK-NEXT:    s_ashr_i32 s7, s6, 31
-; CHECK-NEXT:    v_mov_b32_e32 v0, s2
-; CHECK-NEXT:    v_mov_b32_e32 v1, s3
+; CHECK-NEXT:    v_mov_b32_e32 v2, s2
+; CHECK-NEXT:    v_mov_b32_e32 v3, s3
 ; CHECK-NEXT:    s_lshl_b64 s[2:3], s[6:7], 3
 ; CHECK-NEXT:    s_add_u32 s0, s0, s2
 ; CHECK-NEXT:    s_addc_u32 s1, s1, s3
-; CHECK-NEXT:    v_mov_b32_e32 v3, s1
-; CHECK-NEXT:    v_add_co_u32_e64 v2, vcc, -8, s0
-; CHECK-NEXT:    v_addc_co_u32_e32 v3, vcc, -1, v3, vcc
-; CHECK-NEXT:    flat_atomic_add_f64 v[2:3], v[0:1]
+; CHECK-NEXT:    v_mov_b32_e32 v1, s1
+; CHECK-NEXT:    v_add_co_u32_e64 v0, vcc, -8, s0
+; CHECK-NEXT:    v_addc_co_u32_e32 v1, vcc, -1, v1, vcc
+; CHECK-NEXT:    flat_atomic_add_f64 v[0:1], v[2:3]
 ; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; CHECK-NEXT:    buffer_wbinvl1_vol
 ; CHECK-NEXT:    s_endpgm
diff --git a/llvm/test/CodeGen/AMDGPU/inflated-reg-class-snippet-copy-use-after-free.mir b/llvm/test/CodeGen/AMDGPU/inflated-reg-class-snippet-copy-use-after-free.mir
index 11de6c8..06c3da0 100644
--- a/llvm/test/CodeGen/AMDGPU/inflated-reg-class-snippet-copy-use-after-free.mir
+++ b/llvm/test/CodeGen/AMDGPU/inflated-reg-class-snippet-copy-use-after-free.mir
@@ -32,32 +32,14 @@
 # CHECK-NEXT: undef [[SPLIT0:%[0-9]+]].sub2_sub3:av_512_align2 = COPY undef $vgpr2_vgpr3 {
 # CHECK-NEXT: internal [[SPLIT0]].sub0:av_512_align2 = COPY undef $vgpr0
 # CHECK-NEXT: }
-# CHECK-NEXT: undef [[SPLIT1:%[0-9]+]].sub2_sub3:av_512_align2 = COPY [[SPLIT0]].sub2_sub3 {
-# CHECK-NEXT: internal [[SPLIT1]].sub0:av_512_align2 = COPY [[SPLIT0]].sub0
-# CHECK-NEXT: }
-# CHECK-NEXT: undef [[SPLIT2:%[0-9]+]].sub2_sub3:av_512_align2 = COPY [[SPLIT1]].sub2_sub3 {
-# CHECK-NEXT: internal [[SPLIT2]].sub0:av_512_align2 = COPY [[SPLIT1]].sub0
-# CHECK-NEXT: }
-# CHECK-NEXT: SI_SPILL_AV512_SAVE [[SPLIT2]], %stack.1, $sgpr32, 0, implicit $exec :: (store (s512) into %stack.1, align 4, addrspace 5)
-# CHECK-NEXT: [[RESTORE1:%[0-9]+]]:av_512_align2 = SI_SPILL_AV512_RESTORE %stack.0, $sgpr32, 0, implicit $exec :: (load (s512) from %stack.0, align 4, addrspace 5)
-# CHECK-NEXT: undef [[SPLIT3:%[0-9]+]].sub0_sub1:av_512_align2 = COPY [[RESTORE1]].sub0_sub1
-# CHECK-NEXT: [[RESTORE2:%[0-9]+]]:av_512_align2 = SI_SPILL_AV512_RESTORE %stack.1, $sgpr32, 0, implicit $exec :: (load (s512) from %stack.1, align 4, addrspace 5)
-# CHECK-NEXT: undef [[SPLIT3:%[0-9]+]].sub2_sub3:av_512_align2 = COPY [[RESTORE2]].sub2_sub3 {
-# CHECK-NEXT: internal [[SPLIT3]].sub0:av_512_align2 = COPY [[RESTORE2]].sub0
-# CHECK-NEXT: }
-# CHECK-NEXT: undef [[SPLIT4:%[0-9]+]].sub2_sub3:av_512_align2 = COPY [[SPLIT3]].sub2_sub3 {
-# CHECK-NEXT: internal [[SPLIT4]].sub0:av_512_align2 = COPY [[SPLIT3]].sub0
-# CHECK-NEXT: }
-# CHECK-NEXT: [[SPLIT5:%[0-9]+]].sub2:av_512_align2 = COPY [[SPLIT4]].sub3
-# CHECK-NEXT: undef [[SPLIT6:%[0-9]+]].sub0_sub1_sub2:av_512_align2 = COPY [[SPLIT5]].sub0_sub1_sub2
-# CHECK-NEXT: undef [[SPLIT7:%[0-9]+]].sub0_sub1_sub2:av_512_align2 = COPY [[SPLIT6]].sub0_sub1_sub2
-# CHECK-NEXT: undef [[SPLIT8:%[0-9]+]].sub0:av_512_align2 = COPY [[SPLIT4]].sub0 {
-# CHECK-NEXT: internal [[SPLIT8]].sub2:av_512_align2 = COPY [[SPLIT4]].sub2
+# CHECK-NEXT: undef [[SPLIT2:%[0-9]+]].sub2_sub3:av_512_align2 = COPY [[SPLIT0]].sub2_sub3 {
+# CHECK-NEXT: internal [[SPLIT2]].sub0:av_512_align2 = COPY [[SPLIT0]].sub0
 # CHECK-NEXT: }
-# CHECK-NEXT: [[SPLIT9:%[0-9]+]].sub3:av_512_align2 = COPY [[SPLIT8]].sub2
-# CHECK-NEXT: undef [[SPLIT10:%[0-9]+]].sub0_sub1_sub2_sub3:av_512_align2 = COPY [[SPLIT9]].sub0_sub1_sub2_sub3
-# CHECK-NEXT: undef [[SPLIT13:%[0-9]+]].sub0_sub1_sub2_sub3:vreg_512_align2 = COPY [[SPLIT10]].sub0_sub1_sub2_sub3
-# CHECK-NEXT: [[MFMA_USE1:%[0-9]+]].sub4:vreg_512_align2 = COPY [[SPLIT8]].sub0
+# CHECK-NEXT: [[RESTORE2:%[0-9]+]]:av_512_align2 = SI_SPILL_AV512_RESTORE %stack.0, $sgpr32, 0, implicit $exec :: (load (s512) from %stack.0, align 4, addrspace 5)
+# CHECK-NEXT: [[MFMA_USE1:%[0-9]+]].sub0_sub1:vreg_512_align2 = COPY [[RESTORE2]].sub0_sub1
+# CHECK-NEXT: [[MFMA_USE1]].sub2:vreg_512_align2 = COPY [[SPLIT2]].sub3 
+# CHECK-NEXT: [[MFMA_USE1]].sub3:vreg_512_align2 = COPY [[SPLIT2]].sub2
+# CHECK-NEXT: [[MFMA_USE1]].sub4:vreg_512_align2 = COPY [[SPLIT2]].sub0
 # CHECK-NEXT: [[MFMA_USE1]].sub5:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec
 # CHECK-NEXT: [[MFMA_USE1]].sub6:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec
 # CHECK-NEXT: [[MFMA_USE1]].sub7:vreg_512_align2 = V_MOV_B32_e32 0, implicit $exec
diff --git a/llvm/test/CodeGen/AMDGPU/insert_vector_elt.v2bf16.ll b/llvm/test/CodeGen/AMDGPU/insert_vector_elt.v2bf16.ll
index 1ac75d3..d8c983a 100644
--- a/llvm/test/CodeGen/AMDGPU/insert_vector_elt.v2bf16.ll
+++ b/llvm/test/CodeGen/AMDGPU/insert_vector_elt.v2bf16.ll
@@ -1331,16 +1331,16 @@ define amdgpu_kernel void @v_insertelement_v16bf16_3(ptr addrspace(1) %out, ptr
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x0
 ; GFX942-NEXT:    s_load_dword s6, s[4:5], 0x10
 ; GFX942-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
-; GFX942-NEXT:    v_lshlrev_b32_e32 v8, 5, v0
-; GFX942-NEXT:    v_mov_b32_e32 v9, 0x5040100
+; GFX942-NEXT:    v_lshlrev_b32_e32 v4, 5, v0
+; GFX942-NEXT:    v_mov_b32_e32 v5, 0x5040100
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    global_load_dwordx4 v[0:3], v8, s[2:3]
-; GFX942-NEXT:    global_load_dwordx4 v[4:7], v8, s[2:3] offset:16
+; GFX942-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3]
+; GFX942-NEXT:    global_load_dwordx4 v[6:9], v4, s[2:3] offset:16
 ; GFX942-NEXT:    s_waitcnt vmcnt(1)
-; GFX942-NEXT:    v_perm_b32 v1, s6, v1, v9
+; GFX942-NEXT:    v_perm_b32 v1, s6, v1, v5
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
-; GFX942-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-NEXT:    global_store_dwordx4 v4, v[6:9], s[0:1] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
 ; GFX942-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x() #1
   %tid.ext = sext i32 %tid to i64
diff --git a/llvm/test/CodeGen/AMDGPU/large-avgpr-assign-last.mir b/llvm/test/CodeGen/AMDGPU/large-avgpr-assign-last.mir
new file mode 100644
index 0000000..58e9b0a
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/large-avgpr-assign-last.mir
@@ -0,0 +1,94 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 5
+# RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx950 -verify-regalloc -greedy-regclass-priority-trumps-globalness=1 -start-after=machine-scheduler -stop-after=virtregrewriter,2 -o - %s | FileCheck %s
+
+--- |
+  define void @temp_vgpr_to_agpr_should_not_undo_split_with_remat() #0 {
+  entry:
+    unreachable
+  }
+
+  attributes #0 = { "amdgpu-agpr-alloc"="0,0" }
+...
+
+
+---
+name:            temp_vgpr_to_agpr_should_not_undo_split_with_remat
+tracksRegLiveness: true
+machineFunctionInfo:
+  isEntryFunction: true
+  scratchRSrcReg:  '$sgpr96_sgpr97_sgpr98_sgpr99'
+  stackPtrOffsetReg: '$sgpr32'
+  argumentInfo:
+    privateSegmentBuffer: { reg: '$sgpr0_sgpr1_sgpr2_sgpr3' }
+    kernargSegmentPtr: { reg: '$sgpr4_sgpr5' }
+    workGroupIDX:    { reg: '$sgpr6' }
+    privateSegmentWaveByteOffset: { reg: '$sgpr7' }
+    workItemIDX:     { reg: '$vgpr0' }
+  sgprForEXECCopy: '$sgpr100_sgpr101'
+body:             |
+  bb.0:
+   liveins: $vgpr0, $sgpr4_sgpr5
+    ; CHECK-LABEL: name: temp_vgpr_to_agpr_should_not_undo_split_with_remat
+    ; CHECK: liveins: $vgpr0, $sgpr4_sgpr5
+    ; CHECK-NEXT: {{  $}}
+    ; CHECK-NEXT: renamable $vgpr0 = IMPLICIT_DEF
+    ; CHECK-NEXT: dead renamable $vgpr1 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr1 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr2 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr3 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr4 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr5 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr6 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr7 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr8 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr9 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr10 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr11 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr12 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr13 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr14 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr15 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr16 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr17 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr18 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr19 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr20 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr21 = IMPLICIT_DEF
+    ; CHECK-NEXT: renamable $vgpr22 = IMPLICIT_DEF
+    ; CHECK-NEXT: KILL killed renamable $vgpr2, killed renamable $vgpr3, killed renamable $vgpr4, killed renamable $vgpr5, killed renamable $vgpr6, killed renamable $vgpr7, killed renamable $vgpr8, killed renamable $vgpr9, killed renamable $vgpr10, killed renamable $vgpr11, killed renamable $vgpr12, killed renamable $vgpr13, killed renamable $vgpr14, killed renamable $vgpr15, killed renamable $vgpr16
+    ; CHECK-NEXT: S_NOP 0, implicit-def renamable $vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31_vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38
+    ; CHECK-NEXT: S_NOP 0, implicit-def renamable $vgpr39_vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47_vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54
+    ; CHECK-NEXT: KILL killed renamable $vgpr0, killed renamable $vgpr1, killed renamable $vgpr17, killed renamable $vgpr18, killed renamable $vgpr19, killed renamable $vgpr20, killed renamable $vgpr21, killed renamable $vgpr22
+    ; CHECK-NEXT: S_NOP 0, implicit killed renamable $vgpr23_vgpr24_vgpr25_vgpr26_vgpr27_vgpr28_vgpr29_vgpr30_vgpr31_vgpr32_vgpr33_vgpr34_vgpr35_vgpr36_vgpr37_vgpr38, implicit killed renamable $vgpr39_vgpr40_vgpr41_vgpr42_vgpr43_vgpr44_vgpr45_vgpr46_vgpr47_vgpr48_vgpr49_vgpr50_vgpr51_vgpr52_vgpr53_vgpr54
+    ; CHECK-NEXT: S_ENDPGM 0
+    %1:vgpr_32 = IMPLICIT_DEF
+    %2:vgpr_32 = IMPLICIT_DEF
+    %2:vgpr_32 = IMPLICIT_DEF
+    %3:vgpr_32 = IMPLICIT_DEF
+    %4:vgpr_32 = IMPLICIT_DEF
+    %5:vgpr_32 = IMPLICIT_DEF
+    %6:vgpr_32 = IMPLICIT_DEF
+    %7:vgpr_32 = IMPLICIT_DEF
+    %8:vgpr_32 = IMPLICIT_DEF
+    %9:vgpr_32 = IMPLICIT_DEF
+    %10:vgpr_32 = IMPLICIT_DEF
+    %11:vgpr_32 = IMPLICIT_DEF
+    %12:vgpr_32 = IMPLICIT_DEF
+    %13:vgpr_32 = IMPLICIT_DEF
+    %14:vgpr_32 = IMPLICIT_DEF
+    %15:vgpr_32 = IMPLICIT_DEF
+    %16:vgpr_32 = IMPLICIT_DEF
+    %17:vgpr_32 = IMPLICIT_DEF
+    %18:vgpr_32 = IMPLICIT_DEF
+    %19:vgpr_32 = IMPLICIT_DEF
+    %20:vgpr_32 = IMPLICIT_DEF
+    %21:vgpr_32 = IMPLICIT_DEF
+    %22:vgpr_32 = IMPLICIT_DEF
+    %23:vgpr_32 = IMPLICIT_DEF
+    KILL %3, %4, %5, %6, %7, %8, %9, %10, %11, %12, %13, %14, %15, %16, %17
+    S_NOP 0, implicit-def %50:av_512
+    S_NOP 0, implicit-def %51:av_512
+    KILL %1, %2, %18, %19, %20, %21, %22, %23
+    S_NOP 0, implicit %50, implicit %51
+    S_ENDPGM 0
+...
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.iglp.AFLCustomIRMutator.opt.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.iglp.AFLCustomIRMutator.opt.ll
index 85dd275..fcdad53 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.iglp.AFLCustomIRMutator.opt.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.iglp.AFLCustomIRMutator.opt.ll
@@ -4,30 +4,30 @@
 define amdgpu_kernel void @test_iglp_opt_rev_mfma_gemm(<1 x i64> %L1) {
 ; GCN-LABEL: test_iglp_opt_rev_mfma_gemm:
 ; GCN:       ; %bb.0: ; %entry
-; GCN-NEXT:    v_mov_b32_e32 v32, 0
-; GCN-NEXT:    ds_read_b128 v[0:3], v32
+; GCN-NEXT:    v_mov_b32_e32 v0, 0
+; GCN-NEXT:    ds_read_b128 v[2:5], v0
 ; GCN-NEXT:    s_load_dwordx2 s[0:1], s[8:9], 0x0
-; GCN-NEXT:    ds_read_b128 v[28:31], v32 offset:112
-; GCN-NEXT:    ds_read_b128 v[24:27], v32 offset:96
-; GCN-NEXT:    ds_read_b128 v[20:23], v32 offset:80
-; GCN-NEXT:    ds_read_b128 v[16:19], v32 offset:64
-; GCN-NEXT:    ds_read_b128 v[4:7], v32 offset:16
-; GCN-NEXT:    ds_read_b128 v[8:11], v32 offset:32
-; GCN-NEXT:    ds_read_b128 v[12:15], v32 offset:48
+; GCN-NEXT:    ds_read_b128 v[30:33], v0 offset:112
+; GCN-NEXT:    ds_read_b128 v[26:29], v0 offset:96
+; GCN-NEXT:    ds_read_b128 v[22:25], v0 offset:80
+; GCN-NEXT:    ds_read_b128 v[18:21], v0 offset:64
+; GCN-NEXT:    ds_read_b128 v[6:9], v0 offset:16
+; GCN-NEXT:    ds_read_b128 v[10:13], v0 offset:32
+; GCN-NEXT:    ds_read_b128 v[14:17], v0 offset:48
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NEXT:    ds_write_b128 v32, v[0:3]
-; GCN-NEXT:    v_mov_b32_e32 v0, 0
-; GCN-NEXT:    v_mov_b32_e32 v1, v0
+; GCN-NEXT:    ds_write_b128 v0, v[2:5]
+; GCN-NEXT:    v_mov_b32_e32 v2, 0
+; GCN-NEXT:    v_mov_b32_e32 v3, v2
 ; GCN-NEXT:    s_cmp_lg_u64 s[0:1], 0
 ; GCN-NEXT:    ; iglp_opt mask(0x00000001)
-; GCN-NEXT:    ds_write_b128 v32, v[28:31] offset:112
-; GCN-NEXT:    ds_write_b128 v32, v[24:27] offset:96
-; GCN-NEXT:    ds_write_b128 v32, v[20:23] offset:80
-; GCN-NEXT:    ds_write_b128 v32, v[16:19] offset:64
-; GCN-NEXT:    ds_write_b128 v32, v[12:15] offset:48
-; GCN-NEXT:    ds_write_b128 v32, v[8:11] offset:32
-; GCN-NEXT:    ds_write_b128 v32, v[4:7] offset:16
-; GCN-NEXT:    ds_write_b64 v32, v[0:1]
+; GCN-NEXT:    ds_write_b128 v0, v[30:33] offset:112
+; GCN-NEXT:    ds_write_b128 v0, v[26:29] offset:96
+; GCN-NEXT:    ds_write_b128 v0, v[22:25] offset:80
+; GCN-NEXT:    ds_write_b128 v0, v[18:21] offset:64
+; GCN-NEXT:    ds_write_b128 v0, v[14:17] offset:48
+; GCN-NEXT:    ds_write_b128 v0, v[10:13] offset:32
+; GCN-NEXT:    ds_write_b128 v0, v[6:9] offset:16
+; GCN-NEXT:    ds_write_b64 v0, v[2:3]
 ; GCN-NEXT:    s_endpgm
 entry:
   call void @llvm.amdgcn.iglp.opt(i32 1)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.dim.gfx90a.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.dim.gfx90a.ll
index ed7d88b..dcac419 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.dim.gfx90a.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.image.dim.gfx90a.ll
@@ -18,19 +18,22 @@ define amdgpu_ps <4 x float> @load_1d_lwe(<8 x i32> inreg %rsrc, ptr addrspace(1
 ; GCN-LABEL: load_1d_lwe:
 ; GCN:       ; %bb.0: ; %main_body
 ; GCN-NEXT:    v_mov_b32_e32 v8, 0
-; GCN-NEXT:    v_mov_b32_e32 v6, v0
 ; GCN-NEXT:    v_mov_b32_e32 v9, v8
 ; GCN-NEXT:    v_mov_b32_e32 v10, v8
 ; GCN-NEXT:    v_mov_b32_e32 v11, v8
 ; GCN-NEXT:    v_mov_b32_e32 v12, v8
-; GCN-NEXT:    v_mov_b32_e32 v0, v8
-; GCN-NEXT:    v_mov_b32_e32 v1, v9
-; GCN-NEXT:    v_mov_b32_e32 v2, v10
-; GCN-NEXT:    v_mov_b32_e32 v3, v11
-; GCN-NEXT:    v_mov_b32_e32 v4, v12
-; GCN-NEXT:    image_load v[0:4], v6, s[0:7] dmask:0xf unorm lwe
+; GCN-NEXT:    v_mov_b32_e32 v2, v8
+; GCN-NEXT:    v_mov_b32_e32 v3, v9
+; GCN-NEXT:    v_mov_b32_e32 v4, v10
+; GCN-NEXT:    v_mov_b32_e32 v5, v11
+; GCN-NEXT:    v_mov_b32_e32 v6, v12
+; GCN-NEXT:    image_load v[2:6], v0, s[0:7] dmask:0xf unorm lwe
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dword v8, v4, s[8:9]
+; GCN-NEXT:    v_mov_b32_e32 v0, v2
+; GCN-NEXT:    v_mov_b32_e32 v1, v3
+; GCN-NEXT:    v_mov_b32_e32 v2, v4
+; GCN-NEXT:    v_mov_b32_e32 v3, v5
+; GCN-NEXT:    global_store_dword v8, v6, s[8:9]
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    ; return to shader part epilog
 main_body:
@@ -75,6 +78,27 @@ main_body:
 }
 
 define amdgpu_ps <4 x float> @load_cube_lwe(<8 x i32> inreg %rsrc, ptr addrspace(1) inreg %out, i32 %s, i32 %t, i32 %slice) {
+; GCN-LABEL: load_cube_lwe:
+; GCN:       ; %bb.0: ; %main_body
+; GCN-NEXT:    v_mov_b32_e32 v10, 0
+; GCN-NEXT:    v_mov_b32_e32 v11, v10
+; GCN-NEXT:    v_mov_b32_e32 v12, v10
+; GCN-NEXT:    v_mov_b32_e32 v13, v10
+; GCN-NEXT:    v_mov_b32_e32 v14, v10
+; GCN-NEXT:    v_mov_b32_e32 v4, v10
+; GCN-NEXT:    v_mov_b32_e32 v5, v11
+; GCN-NEXT:    v_mov_b32_e32 v6, v12
+; GCN-NEXT:    v_mov_b32_e32 v7, v13
+; GCN-NEXT:    v_mov_b32_e32 v8, v14
+; GCN-NEXT:    image_load v[4:8], v[0:2], s[0:7] dmask:0xf unorm lwe da
+; GCN-NEXT:    s_waitcnt vmcnt(0)
+; GCN-NEXT:    v_mov_b32_e32 v0, v4
+; GCN-NEXT:    v_mov_b32_e32 v1, v5
+; GCN-NEXT:    v_mov_b32_e32 v2, v6
+; GCN-NEXT:    v_mov_b32_e32 v3, v7
+; GCN-NEXT:    global_store_dword v10, v8, s[8:9]
+; GCN-NEXT:    s_waitcnt vmcnt(0)
+; GCN-NEXT:    ; return to shader part epilog
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.cube.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 2, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0
@@ -106,6 +130,27 @@ main_body:
 }
 
 define amdgpu_ps <4 x float> @load_2darray_lwe(<8 x i32> inreg %rsrc, ptr addrspace(1) inreg %out, i32 %s, i32 %t, i32 %slice) {
+; GCN-LABEL: load_2darray_lwe:
+; GCN:       ; %bb.0: ; %main_body
+; GCN-NEXT:    v_mov_b32_e32 v10, 0
+; GCN-NEXT:    v_mov_b32_e32 v11, v10
+; GCN-NEXT:    v_mov_b32_e32 v12, v10
+; GCN-NEXT:    v_mov_b32_e32 v13, v10
+; GCN-NEXT:    v_mov_b32_e32 v14, v10
+; GCN-NEXT:    v_mov_b32_e32 v4, v10
+; GCN-NEXT:    v_mov_b32_e32 v5, v11
+; GCN-NEXT:    v_mov_b32_e32 v6, v12
+; GCN-NEXT:    v_mov_b32_e32 v7, v13
+; GCN-NEXT:    v_mov_b32_e32 v8, v14
+; GCN-NEXT:    image_load v[4:8], v[0:2], s[0:7] dmask:0xf unorm lwe da
+; GCN-NEXT:    s_waitcnt vmcnt(0)
+; GCN-NEXT:    v_mov_b32_e32 v0, v4
+; GCN-NEXT:    v_mov_b32_e32 v1, v5
+; GCN-NEXT:    v_mov_b32_e32 v2, v6
+; GCN-NEXT:    v_mov_b32_e32 v3, v7
+; GCN-NEXT:    global_store_dword v10, v8, s[8:9]
+; GCN-NEXT:    s_waitcnt vmcnt(0)
+; GCN-NEXT:    ; return to shader part epilog
 main_body:
   %v = call {<4 x float>,i32} @llvm.amdgcn.image.load.2darray.v4f32i32.i32(i32 15, i32 %s, i32 %t, i32 %slice, <8 x i32> %rsrc, i32 2, i32 0)
   %v.vec = extractvalue {<4 x float>, i32} %v, 0
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.bf16.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.bf16.ll
index 1585a2c..303ea50 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.bf16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.bf16.ll
@@ -1,6 +1,7 @@
-; RUN: llc -mtriple=amdgcn -mcpu=gfx908 < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX908 %s
-; RUN: llc -mtriple=amdgcn -mcpu=gfx908 -mattr=-mfma-inline-literal-bug < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX908 %s
-; RUN: llc -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX90A %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -mtriple=amdgcn -mcpu=gfx908 < %s | FileCheck --check-prefixes=GCN,GFX908 %s
+; RUN: llc -mtriple=amdgcn -mcpu=gfx908 -mattr=-mfma-inline-literal-bug < %s | FileCheck --check-prefixes=GCN,GFX908 %s
+; RUN: llc -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck --check-prefixes=GCN,GFX90A %s
 
 declare <32 x float> @llvm.amdgcn.mfma.f32.32x32x2bf16(<2 x i16>, <2 x i16>, <32 x float>, i32, i32, i32)
 declare <16 x float> @llvm.amdgcn.mfma.f32.16x16x2bf16(<2 x i16>, <2 x i16>, <16 x float>, i32, i32, i32)
@@ -9,50 +10,199 @@ declare <16 x float> @llvm.amdgcn.mfma.f32.32x32x4bf16(<2 x i16>, <2 x i16>, <16
 declare <4 x float> @llvm.amdgcn.mfma.f32.16x16x8bf16(<2 x i16>, <2 x i16>, <4 x float>, i32, i32, i32)
 declare i32 @llvm.amdgcn.workitem.id.x()
 
-; GCN-LABEL: {{^}}test_mfma_f32_32x32x2bf16:
-; GCN-DAG:         v_mov_b32_e32 [[TWO:v[0-9]+]], 2
-; GCN-DAG:         v_mov_b32_e32 [[ONE:v[0-9]+]], 1
-; GCN-DAG:         s_load_dwordx16
-; GCN-DAG:         s_load_dwordx16
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX908-DAG:      v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX90A-COUNT-32: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GCN:             v_mfma_f32_32x32x2bf16 a[{{[0-9]+:[0-9]+}}], [[ONE]], [[TWO]], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GFX908-COUNT-32: v_accvgpr_read_b32
-; GFX908:          global_store_dwordx4
-; GFX90A-NOT:      v_accvgpr_read_b32
-; GFX90A-COUNT-8:  global_store_dwordx4 v{{[0-9]+}}, a[{{[0-9:]+}}],
 define amdgpu_kernel void @test_mfma_f32_32x32x2bf16(ptr addrspace(1) %arg) #0 {
+; GFX908-LABEL: test_mfma_f32_32x32x2bf16:
+; GFX908:       ; %bb.0: ; %bb
+; GFX908-NEXT:    s_load_dwordx2 s[34:35], s[4:5], 0x24
+; GFX908-NEXT:    v_mov_b32_e32 v4, 0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    s_load_dwordx16 s[16:31], s[34:35], 0x0
+; GFX908-NEXT:    s_load_dwordx16 s[0:15], s[34:35], 0x40
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    v_mov_b32_e32 v0, s16
+; GFX908-NEXT:    v_mov_b32_e32 v1, s17
+; GFX908-NEXT:    v_mov_b32_e32 v2, s18
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s21
+; GFX908-NEXT:    v_mov_b32_e32 v1, s22
+; GFX908-NEXT:    v_mov_b32_e32 v2, s23
+; GFX908-NEXT:    v_accvgpr_write_b32 a5, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a6, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a7, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s24
+; GFX908-NEXT:    v_mov_b32_e32 v1, s25
+; GFX908-NEXT:    v_mov_b32_e32 v2, s26
+; GFX908-NEXT:    v_accvgpr_write_b32 a8, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a9, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a10, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s27
+; GFX908-NEXT:    v_mov_b32_e32 v1, s28
+; GFX908-NEXT:    v_mov_b32_e32 v2, s29
+; GFX908-NEXT:    v_accvgpr_write_b32 a11, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a12, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a13, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s30
+; GFX908-NEXT:    v_mov_b32_e32 v1, s31
+; GFX908-NEXT:    v_mov_b32_e32 v2, s0
+; GFX908-NEXT:    v_accvgpr_write_b32 a14, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a15, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a16, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s1
+; GFX908-NEXT:    v_mov_b32_e32 v1, s2
+; GFX908-NEXT:    v_mov_b32_e32 v2, s3
+; GFX908-NEXT:    v_accvgpr_write_b32 a17, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a18, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a19, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s4
+; GFX908-NEXT:    v_mov_b32_e32 v1, s5
+; GFX908-NEXT:    v_mov_b32_e32 v2, s6
+; GFX908-NEXT:    v_accvgpr_write_b32 a20, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a21, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a22, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s7
+; GFX908-NEXT:    v_mov_b32_e32 v1, s8
+; GFX908-NEXT:    v_mov_b32_e32 v2, s9
+; GFX908-NEXT:    v_mov_b32_e32 v3, s19
+; GFX908-NEXT:    v_accvgpr_write_b32 a23, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a24, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a25, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s10
+; GFX908-NEXT:    v_mov_b32_e32 v1, s11
+; GFX908-NEXT:    v_mov_b32_e32 v2, s12
+; GFX908-NEXT:    v_mov_b32_e32 v5, s20
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a26, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a27, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a28, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s13
+; GFX908-NEXT:    v_mov_b32_e32 v1, s14
+; GFX908-NEXT:    v_mov_b32_e32 v2, s15
+; GFX908-NEXT:    v_mov_b32_e32 v3, 1
+; GFX908-NEXT:    v_accvgpr_write_b32 a4, v5
+; GFX908-NEXT:    v_accvgpr_write_b32 a29, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a30, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a31, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, 2
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_mfma_f32_32x32x2bf16 a[0:31], v3, v0, a[0:31] cbsz:1 abid:2 blgp:3
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a27
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a26
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a25
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a24
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:96
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a31
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a30
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a29
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a28
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:112
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a19
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a18
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a17
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a16
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:64
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a23
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a22
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a21
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a20
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:80
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a11
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a10
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a9
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a8
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:32
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a15
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a14
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a13
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a12
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:48
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a3
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a2
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a1
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a0
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35]
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a7
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a6
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a5
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a4
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:16
+; GFX908-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: test_mfma_f32_32x32x2bf16:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx2 s[34:35], s[4:5], 0x24
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 1
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 2
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_load_dwordx16 s[16:31], s[34:35], 0x0
+; GFX90A-NEXT:    s_load_dwordx16 s[0:15], s[34:35], 0x40
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, s16
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, s17
+; GFX90A-NEXT:    v_accvgpr_write_b32 a2, s18
+; GFX90A-NEXT:    v_accvgpr_write_b32 a3, s19
+; GFX90A-NEXT:    v_accvgpr_write_b32 a4, s20
+; GFX90A-NEXT:    v_accvgpr_write_b32 a5, s21
+; GFX90A-NEXT:    v_accvgpr_write_b32 a6, s22
+; GFX90A-NEXT:    v_accvgpr_write_b32 a7, s23
+; GFX90A-NEXT:    v_accvgpr_write_b32 a8, s24
+; GFX90A-NEXT:    v_accvgpr_write_b32 a9, s25
+; GFX90A-NEXT:    v_accvgpr_write_b32 a10, s26
+; GFX90A-NEXT:    v_accvgpr_write_b32 a11, s27
+; GFX90A-NEXT:    v_accvgpr_write_b32 a12, s28
+; GFX90A-NEXT:    v_accvgpr_write_b32 a13, s29
+; GFX90A-NEXT:    v_accvgpr_write_b32 a14, s30
+; GFX90A-NEXT:    v_accvgpr_write_b32 a15, s31
+; GFX90A-NEXT:    v_accvgpr_write_b32 a16, s0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a17, s1
+; GFX90A-NEXT:    v_accvgpr_write_b32 a18, s2
+; GFX90A-NEXT:    v_accvgpr_write_b32 a19, s3
+; GFX90A-NEXT:    v_accvgpr_write_b32 a20, s4
+; GFX90A-NEXT:    v_accvgpr_write_b32 a21, s5
+; GFX90A-NEXT:    v_accvgpr_write_b32 a22, s6
+; GFX90A-NEXT:    v_accvgpr_write_b32 a23, s7
+; GFX90A-NEXT:    v_accvgpr_write_b32 a24, s8
+; GFX90A-NEXT:    v_accvgpr_write_b32 a25, s9
+; GFX90A-NEXT:    v_accvgpr_write_b32 a26, s10
+; GFX90A-NEXT:    v_accvgpr_write_b32 a27, s11
+; GFX90A-NEXT:    v_accvgpr_write_b32 a28, s12
+; GFX90A-NEXT:    v_accvgpr_write_b32 a29, s13
+; GFX90A-NEXT:    v_accvgpr_write_b32 a30, s14
+; GFX90A-NEXT:    v_accvgpr_write_b32 a31, s15
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f32_32x32x2bf16 a[0:31], v1, v2, a[0:31] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 2
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[24:27], s[34:35] offset:96
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[28:31], s[34:35] offset:112
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[16:19], s[34:35] offset:64
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[20:23], s[34:35] offset:80
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[8:11], s[34:35] offset:32
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[12:15], s[34:35] offset:48
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[34:35]
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[4:7], s[34:35] offset:16
+; GFX90A-NEXT:    s_endpgm
 bb:
   %in.1 = load <32 x float>, ptr addrspace(1) %arg
   %a = bitcast i32 1 to <2 x i16>
@@ -62,18 +212,109 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_16x16x2bf16:
-; GCN-DAG:         v_mov_b32_e32 [[TWO:v[0-9]+]], 2
-; GCN-DAG:         v_mov_b32_e32 [[ONE:v[0-9]+]], 1
-; GCN-DAG:         s_load_dwordx16
-; GFX908-DAG-COUNT-16: v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX90A-COUNT-16: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GCN:             v_mfma_f32_16x16x2bf16 a[{{[0-9]+:[0-9]+}}], [[ONE]], [[TWO]], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GFX908-COUNT-16: v_accvgpr_read_b32
-; GFX908:          global_store_dwordx4
-; GFX90A-NOT:      v_accvgpr_read_b32
-; GFX90A-COUNT-4:  global_store_dwordx4 v{{[0-9]+}}, a[{{[0-9:]+}}],
 define amdgpu_kernel void @test_mfma_f32_16x16x2bf16(ptr addrspace(1) %arg) #0 {
+; GFX908-LABEL: test_mfma_f32_16x16x2bf16:
+; GFX908:       ; %bb.0: ; %bb
+; GFX908-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX908-NEXT:    v_mov_b32_e32 v0, 1
+; GFX908-NEXT:    v_mov_b32_e32 v12, 0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    v_mov_b32_e32 v13, s0
+; GFX908-NEXT:    v_mov_b32_e32 v1, s1
+; GFX908-NEXT:    v_mov_b32_e32 v2, s2
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, v13
+; GFX908-NEXT:    v_mov_b32_e32 v13, s3
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, v13
+; GFX908-NEXT:    v_mov_b32_e32 v1, s4
+; GFX908-NEXT:    v_mov_b32_e32 v2, s5
+; GFX908-NEXT:    v_mov_b32_e32 v13, s6
+; GFX908-NEXT:    v_accvgpr_write_b32 a4, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a5, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a6, v13
+; GFX908-NEXT:    v_mov_b32_e32 v1, s7
+; GFX908-NEXT:    v_mov_b32_e32 v2, s8
+; GFX908-NEXT:    v_mov_b32_e32 v13, s9
+; GFX908-NEXT:    v_accvgpr_write_b32 a7, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a8, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a9, v13
+; GFX908-NEXT:    v_mov_b32_e32 v1, s10
+; GFX908-NEXT:    v_mov_b32_e32 v2, s11
+; GFX908-NEXT:    v_mov_b32_e32 v13, s12
+; GFX908-NEXT:    v_accvgpr_write_b32 a10, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a11, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a12, v13
+; GFX908-NEXT:    v_mov_b32_e32 v1, s13
+; GFX908-NEXT:    v_mov_b32_e32 v2, s14
+; GFX908-NEXT:    v_mov_b32_e32 v13, s15
+; GFX908-NEXT:    v_accvgpr_write_b32 a13, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a14, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a15, v13
+; GFX908-NEXT:    v_mov_b32_e32 v1, 2
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, a[0:15] cbsz:1 abid:2 blgp:3
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a15
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a14
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a13
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a12
+; GFX908-NEXT:    v_accvgpr_read_b32 v7, a11
+; GFX908-NEXT:    v_accvgpr_read_b32 v6, a10
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a9
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a8
+; GFX908-NEXT:    v_accvgpr_read_b32 v11, a7
+; GFX908-NEXT:    v_accvgpr_read_b32 v10, a6
+; GFX908-NEXT:    v_accvgpr_read_b32 v9, a5
+; GFX908-NEXT:    v_accvgpr_read_b32 v8, a4
+; GFX908-NEXT:    global_store_dwordx4 v12, v[0:3], s[16:17] offset:48
+; GFX908-NEXT:    global_store_dwordx4 v12, v[4:7], s[16:17] offset:32
+; GFX908-NEXT:    global_store_dwordx4 v12, v[8:11], s[16:17] offset:16
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a3
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a2
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a1
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a0
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v12, v[0:3], s[16:17]
+; GFX908-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: test_mfma_f32_16x16x2bf16:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 1
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 2
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX90A-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX90A-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX90A-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX90A-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX90A-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX90A-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX90A-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX90A-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX90A-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX90A-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX90A-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX90A-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX90A-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX90A-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, a[0:15] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX90A-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %a = bitcast i32 1 to <2 x i16>
@@ -83,18 +324,53 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_4x4x2bf16:
-; GCN-DAG:        v_mov_b32_e32 [[TWO:v[0-9]+]], 2
-; GCN-DAG:        v_mov_b32_e32 [[ONE:v[0-9]+]], 1
-; GCN:            s_load_dwordx4
-; GFX908-COUNT-4: v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX90A-COUNT-4: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GCN:            v_mfma_f32_4x4x2bf16 [[RES:a\[[0-9]+:[0-9]+\]]], [[ONE]], [[TWO]], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GFX908-COUNT-4: v_accvgpr_read_b32
-; GFX908:         global_store_dwordx4
-; GFX90A-NOT:     v_accvgpr_read_b32
-; GFX90A:         global_store_dwordx4 v{{[0-9]+}}, [[RES]],
 define amdgpu_kernel void @test_mfma_f32_4x4x2bf16(ptr addrspace(1) %arg) #0 {
+; GFX908-LABEL: test_mfma_f32_4x4x2bf16:
+; GFX908:       ; %bb.0: ; %bb
+; GFX908-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX908-NEXT:    v_mov_b32_e32 v0, 1
+; GFX908-NEXT:    v_mov_b32_e32 v1, 2
+; GFX908-NEXT:    v_mov_b32_e32 v4, 0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    v_mov_b32_e32 v5, s0
+; GFX908-NEXT:    v_mov_b32_e32 v2, s1
+; GFX908-NEXT:    v_mov_b32_e32 v3, s2
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, v5
+; GFX908-NEXT:    v_mov_b32_e32 v5, s3
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, v5
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_mfma_f32_4x4x2bf16 a[0:3], v0, v1, a[0:3] cbsz:1 abid:2 blgp:3
+; GFX908-NEXT:    s_nop 3
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a0
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a1
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a2
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a3
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX908-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: test_mfma_f32_4x4x2bf16:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 1
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 2
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX90A-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX90A-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f32_4x4x2bf16 a[0:3], v0, v2, a[0:3] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    s_nop 4
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[0:3], s[6:7]
+; GFX90A-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %a = bitcast i32 1 to <2 x i16>
@@ -104,18 +380,110 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_32x32x4bf16:
-; GCN-DAG:         v_mov_b32_e32 [[TWO:v[0-9]+]], 2
-; GCN-DAG:         v_mov_b32_e32 [[ONE:v[0-9]+]], 1
-; GCN-DAG:         s_load_dwordx16
-; GFX908-DAG-COUNT-16: v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX90A-COUNT-4:  v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GCN:             v_mfma_f32_32x32x4bf16 a[{{[0-9]+:[0-9]+}}], [[ONE]], [[TWO]], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GFX908-COUNT-16: v_accvgpr_read_b32
-; GFX908:          global_store_dwordx4
-; GFX90A-NOT:      v_accvgpr_read_b32
-; GFX90A-COUNT-4:  global_store_dwordx4 v{{[0-9]+}}, a[{{[0-9:]+}}],
 define amdgpu_kernel void @test_mfma_f32_32x32x4bf16(ptr addrspace(1) %arg) #0 {
+; GFX908-LABEL: test_mfma_f32_32x32x4bf16:
+; GFX908:       ; %bb.0: ; %bb
+; GFX908-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX908-NEXT:    v_mov_b32_e32 v0, 1
+; GFX908-NEXT:    v_mov_b32_e32 v16, 0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    v_mov_b32_e32 v17, s0
+; GFX908-NEXT:    v_mov_b32_e32 v1, s1
+; GFX908-NEXT:    v_mov_b32_e32 v2, s2
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, v17
+; GFX908-NEXT:    v_mov_b32_e32 v17, s3
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, v17
+; GFX908-NEXT:    v_mov_b32_e32 v1, s4
+; GFX908-NEXT:    v_mov_b32_e32 v2, s5
+; GFX908-NEXT:    v_mov_b32_e32 v17, s6
+; GFX908-NEXT:    v_accvgpr_write_b32 a4, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a5, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a6, v17
+; GFX908-NEXT:    v_mov_b32_e32 v1, s7
+; GFX908-NEXT:    v_mov_b32_e32 v2, s8
+; GFX908-NEXT:    v_mov_b32_e32 v17, s9
+; GFX908-NEXT:    v_accvgpr_write_b32 a7, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a8, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a9, v17
+; GFX908-NEXT:    v_mov_b32_e32 v1, s10
+; GFX908-NEXT:    v_mov_b32_e32 v2, s11
+; GFX908-NEXT:    v_mov_b32_e32 v17, s12
+; GFX908-NEXT:    v_accvgpr_write_b32 a10, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a11, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a12, v17
+; GFX908-NEXT:    v_mov_b32_e32 v1, s13
+; GFX908-NEXT:    v_mov_b32_e32 v2, s14
+; GFX908-NEXT:    v_mov_b32_e32 v17, s15
+; GFX908-NEXT:    v_accvgpr_write_b32 a13, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a14, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a15, v17
+; GFX908-NEXT:    v_mov_b32_e32 v1, 2
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, a[0:15] cbsz:1 abid:2 blgp:3
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a15
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a14
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a13
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a12
+; GFX908-NEXT:    v_accvgpr_read_b32 v7, a11
+; GFX908-NEXT:    v_accvgpr_read_b32 v6, a10
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a9
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a8
+; GFX908-NEXT:    v_accvgpr_read_b32 v11, a7
+; GFX908-NEXT:    v_accvgpr_read_b32 v10, a6
+; GFX908-NEXT:    v_accvgpr_read_b32 v9, a5
+; GFX908-NEXT:    v_accvgpr_read_b32 v8, a4
+; GFX908-NEXT:    v_accvgpr_read_b32 v15, a3
+; GFX908-NEXT:    v_accvgpr_read_b32 v14, a2
+; GFX908-NEXT:    v_accvgpr_read_b32 v13, a1
+; GFX908-NEXT:    v_accvgpr_read_b32 v12, a0
+; GFX908-NEXT:    global_store_dwordx4 v16, v[0:3], s[16:17] offset:48
+; GFX908-NEXT:    global_store_dwordx4 v16, v[4:7], s[16:17] offset:32
+; GFX908-NEXT:    global_store_dwordx4 v16, v[8:11], s[16:17] offset:16
+; GFX908-NEXT:    global_store_dwordx4 v16, v[12:15], s[16:17]
+; GFX908-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: test_mfma_f32_32x32x4bf16:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 1
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 2
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX90A-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX90A-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX90A-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX90A-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX90A-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX90A-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX90A-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX90A-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX90A-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX90A-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX90A-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX90A-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX90A-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX90A-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, a[0:15] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX90A-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %a = bitcast i32 1 to <2 x i16>
@@ -125,18 +493,55 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_16x16x8bf16:
-; GCN-DAG:        v_mov_b32_e32 [[TWO:v[0-9]+]], 2
-; GCN-DAG:        v_mov_b32_e32 [[ONE:v[0-9]+]], 1
-; GCN:            s_load_dwordx4
-; GFX908-COUNT-4: v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX90A-COUNT-4: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GCN:            v_mfma_f32_16x16x8bf16 [[RES:a\[[0-9]+:[0-9]+\]]], [[ONE]], [[TWO]], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GFX908-COUNT-4: v_accvgpr_read_b32
-; GFX908:         global_store_dwordx4
-; GFX90A-NOT:     v_accvgpr_read_b32
-; GFX90A:         global_store_dwordx4 v{{[0-9]+}}, [[RES]],
 define amdgpu_kernel void @test_mfma_f32_16x16x8bf16(ptr addrspace(1) %arg) #0 {
+; GFX908-LABEL: test_mfma_f32_16x16x8bf16:
+; GFX908:       ; %bb.0: ; %bb
+; GFX908-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX908-NEXT:    v_mov_b32_e32 v0, 1
+; GFX908-NEXT:    v_mov_b32_e32 v1, 2
+; GFX908-NEXT:    v_mov_b32_e32 v4, 0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    v_mov_b32_e32 v5, s0
+; GFX908-NEXT:    v_mov_b32_e32 v2, s1
+; GFX908-NEXT:    v_mov_b32_e32 v3, s2
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, v5
+; GFX908-NEXT:    v_mov_b32_e32 v5, s3
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, v5
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_mfma_f32_16x16x8bf16 a[0:3], v0, v1, a[0:3] cbsz:1 abid:2 blgp:3
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a0
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a1
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a2
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a3
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX908-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: test_mfma_f32_16x16x8bf16:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 1
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 2
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX90A-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX90A-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f32_16x16x8bf16 a[0:3], v0, v2, a[0:3] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 2
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[0:3], s[6:7]
+; GFX90A-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %a = bitcast i32 1 to <2 x i16>
@@ -147,3 +552,5 @@ bb:
 }
 
 attributes #0 = { "amdgpu-flat-work-group-size"="1,256" }
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; GCN: {{.*}}
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx90a.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx90a.ll
index 4c26961..ff77d5cc 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx90a.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx90a.ll
@@ -1,5 +1,8 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
 ; RUN: llc -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX90A %s
 ; RUN: llc -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX942 %s
+; RUN: llc -mtriple=amdgcn -mcpu=gfx90a -amdgpu-mfma-vgpr-form < %s | FileCheck -enable-var-scope --check-prefixes=VGPR,GFX90A-VGPR %s
+; RUN: llc -mtriple=amdgcn -mcpu=gfx942 -amdgpu-mfma-vgpr-form < %s | FileCheck -enable-var-scope --check-prefixes=VGPR,GFX942-VGPR %s
 
 declare <32 x float> @llvm.amdgcn.mfma.f32.32x32x4bf16.1k(<4 x i16>, <4 x i16>, <32 x float>, i32, i32, i32)
 declare <16 x float> @llvm.amdgcn.mfma.f32.16x16x4bf16.1k(<4 x i16>, <4 x i16>, <16 x float>, i32, i32, i32)
@@ -10,17 +13,238 @@ declare <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double, double, <4 x doubl
 declare double @llvm.amdgcn.mfma.f64.4x4x4f64(double, double, double, i32, i32, i32)
 declare i32 @llvm.amdgcn.workitem.id.x()
 
-; GCN-LABEL: {{^}}test_mfma_f32_32x32x4bf16_1k:
-; GCN-DAG:     s_load_dwordx16
-; GCN-DAG:     s_load_dwordx16
-; GCN-DAG:     v_mov_b32_e32 v[[TWO:[0-9]+]], 2
-; GCN-DAG:     v_mov_b32_e32 v[[ONE:[0-9]+]], 1
-; GCN-COUNT-32: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GFX90A:      v_mfma_f32_32x32x4bf16_1k a[{{[0-9]+:[0-9]+}}], v[[[ONE]]:{{[0-9]+}}], v[[[TWO]]:{{[0-9]+}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GFX942:      v_mfma_f32_32x32x4_2b_bf16 a[{{[0-9]+:[0-9]+}}], v[[[ONE]]:{{[0-9+]}}], v[[[TWO]]:{{[0-9+]}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GCN-NOT:     v_accvgpr_read_b32
-; GCN-COUNT-8: global_store_dwordx4 v{{[0-9]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_32x32x4bf16_1k(ptr addrspace(1) %arg) #0 {
+; GFX90A-LABEL: test_mfma_f32_32x32x4bf16_1k:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx2 s[34:35], s[4:5], 0x24
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 0
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 1
+; GFX90A-NEXT:    v_mov_b32_e32 v3, v1
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 2
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_load_dwordx16 s[16:31], s[34:35], 0x0
+; GFX90A-NEXT:    s_load_dwordx16 s[0:15], s[34:35], 0x40
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, s16
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, s17
+; GFX90A-NEXT:    v_accvgpr_write_b32 a2, s18
+; GFX90A-NEXT:    v_accvgpr_write_b32 a3, s19
+; GFX90A-NEXT:    v_accvgpr_write_b32 a4, s20
+; GFX90A-NEXT:    v_accvgpr_write_b32 a5, s21
+; GFX90A-NEXT:    v_accvgpr_write_b32 a6, s22
+; GFX90A-NEXT:    v_accvgpr_write_b32 a7, s23
+; GFX90A-NEXT:    v_accvgpr_write_b32 a8, s24
+; GFX90A-NEXT:    v_accvgpr_write_b32 a9, s25
+; GFX90A-NEXT:    v_accvgpr_write_b32 a10, s26
+; GFX90A-NEXT:    v_accvgpr_write_b32 a11, s27
+; GFX90A-NEXT:    v_accvgpr_write_b32 a12, s28
+; GFX90A-NEXT:    v_accvgpr_write_b32 a13, s29
+; GFX90A-NEXT:    v_accvgpr_write_b32 a14, s30
+; GFX90A-NEXT:    v_accvgpr_write_b32 a15, s31
+; GFX90A-NEXT:    v_accvgpr_write_b32 a16, s0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a17, s1
+; GFX90A-NEXT:    v_accvgpr_write_b32 a18, s2
+; GFX90A-NEXT:    v_accvgpr_write_b32 a19, s3
+; GFX90A-NEXT:    v_accvgpr_write_b32 a20, s4
+; GFX90A-NEXT:    v_accvgpr_write_b32 a21, s5
+; GFX90A-NEXT:    v_accvgpr_write_b32 a22, s6
+; GFX90A-NEXT:    v_accvgpr_write_b32 a23, s7
+; GFX90A-NEXT:    v_accvgpr_write_b32 a24, s8
+; GFX90A-NEXT:    v_accvgpr_write_b32 a25, s9
+; GFX90A-NEXT:    v_accvgpr_write_b32 a26, s10
+; GFX90A-NEXT:    v_accvgpr_write_b32 a27, s11
+; GFX90A-NEXT:    v_accvgpr_write_b32 a28, s12
+; GFX90A-NEXT:    v_accvgpr_write_b32 a29, s13
+; GFX90A-NEXT:    v_accvgpr_write_b32 a30, s14
+; GFX90A-NEXT:    v_accvgpr_write_b32 a31, s15
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f32_32x32x4bf16_1k a[0:31], v[2:3], v[0:1], a[0:31] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 2
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[24:27], s[34:35] offset:96
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[28:31], s[34:35] offset:112
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[16:19], s[34:35] offset:64
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[20:23], s[34:35] offset:80
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[8:11], s[34:35] offset:32
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[12:15], s[34:35] offset:48
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[0:3], s[34:35]
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[4:7], s[34:35] offset:16
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_f32_32x32x4bf16_1k:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    s_load_dwordx2 s[34:35], s[4:5], 0x24
+; GFX942-NEXT:    v_mov_b32_e32 v1, 0
+; GFX942-NEXT:    v_mov_b32_e32 v2, 1
+; GFX942-NEXT:    v_mov_b32_e32 v3, v1
+; GFX942-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    s_load_dwordx16 s[16:31], s[34:35], 0x0
+; GFX942-NEXT:    s_load_dwordx16 s[0:15], s[34:35], 0x40
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    v_accvgpr_write_b32 a0, s16
+; GFX942-NEXT:    v_accvgpr_write_b32 a1, s17
+; GFX942-NEXT:    v_accvgpr_write_b32 a2, s18
+; GFX942-NEXT:    v_accvgpr_write_b32 a3, s19
+; GFX942-NEXT:    v_accvgpr_write_b32 a4, s20
+; GFX942-NEXT:    v_accvgpr_write_b32 a5, s21
+; GFX942-NEXT:    v_accvgpr_write_b32 a6, s22
+; GFX942-NEXT:    v_accvgpr_write_b32 a7, s23
+; GFX942-NEXT:    v_accvgpr_write_b32 a8, s24
+; GFX942-NEXT:    v_accvgpr_write_b32 a9, s25
+; GFX942-NEXT:    v_accvgpr_write_b32 a10, s26
+; GFX942-NEXT:    v_accvgpr_write_b32 a11, s27
+; GFX942-NEXT:    v_accvgpr_write_b32 a12, s28
+; GFX942-NEXT:    v_accvgpr_write_b32 a13, s29
+; GFX942-NEXT:    v_accvgpr_write_b32 a14, s30
+; GFX942-NEXT:    v_accvgpr_write_b32 a15, s31
+; GFX942-NEXT:    v_accvgpr_write_b32 a16, s0
+; GFX942-NEXT:    v_accvgpr_write_b32 a17, s1
+; GFX942-NEXT:    v_accvgpr_write_b32 a18, s2
+; GFX942-NEXT:    v_accvgpr_write_b32 a19, s3
+; GFX942-NEXT:    v_accvgpr_write_b32 a20, s4
+; GFX942-NEXT:    v_accvgpr_write_b32 a21, s5
+; GFX942-NEXT:    v_accvgpr_write_b32 a22, s6
+; GFX942-NEXT:    v_accvgpr_write_b32 a23, s7
+; GFX942-NEXT:    v_accvgpr_write_b32 a24, s8
+; GFX942-NEXT:    v_accvgpr_write_b32 a25, s9
+; GFX942-NEXT:    v_accvgpr_write_b32 a26, s10
+; GFX942-NEXT:    v_accvgpr_write_b32 a27, s11
+; GFX942-NEXT:    v_accvgpr_write_b32 a28, s12
+; GFX942-NEXT:    v_accvgpr_write_b32 a29, s13
+; GFX942-NEXT:    v_accvgpr_write_b32 a30, s14
+; GFX942-NEXT:    v_accvgpr_write_b32 a31, s15
+; GFX942-NEXT:    s_nop 1
+; GFX942-NEXT:    v_mfma_f32_32x32x4_2b_bf16 a[0:31], v[2:3], v[0:1], a[0:31] cbsz:1 abid:2 blgp:3
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 2
+; GFX942-NEXT:    global_store_dwordx4 v1, a[24:27], s[34:35] offset:96
+; GFX942-NEXT:    global_store_dwordx4 v1, a[28:31], s[34:35] offset:112
+; GFX942-NEXT:    global_store_dwordx4 v1, a[16:19], s[34:35] offset:64
+; GFX942-NEXT:    global_store_dwordx4 v1, a[20:23], s[34:35] offset:80
+; GFX942-NEXT:    global_store_dwordx4 v1, a[8:11], s[34:35] offset:32
+; GFX942-NEXT:    global_store_dwordx4 v1, a[12:15], s[34:35] offset:48
+; GFX942-NEXT:    global_store_dwordx4 v1, a[0:3], s[34:35]
+; GFX942-NEXT:    global_store_dwordx4 v1, a[4:7], s[34:35] offset:16
+; GFX942-NEXT:    s_endpgm
+;
+; GFX90A-VGPR-LABEL: test_mfma_f32_32x32x4bf16_1k:
+; GFX90A-VGPR:       ; %bb.0: ; %bb
+; GFX90A-VGPR-NEXT:    s_load_dwordx2 s[34:35], s[4:5], 0x24
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v33, 0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v34, 1
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v35, v33
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v32, 2
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    s_load_dwordx16 s[16:31], s[34:35], 0x0
+; GFX90A-VGPR-NEXT:    s_load_dwordx16 s[0:15], s[34:35], 0x40
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v0, s16
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v1, s17
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v2, s18
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v3, s19
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v4, s20
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v5, s21
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v6, s22
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v7, s23
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v8, s24
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v9, s25
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v10, s26
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v11, s27
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v12, s28
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v13, s29
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v14, s30
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v15, s31
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v16, s0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v17, s1
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v18, s2
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v19, s3
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v20, s4
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v21, s5
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v22, s6
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v23, s7
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v24, s8
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v25, s9
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v26, s10
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v27, s11
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v28, s12
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v29, s13
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v30, s14
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v31, s15
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    v_mfma_f32_32x32x4bf16_1k v[0:31], v[34:35], v[32:33], v[0:31] cbsz:1 abid:2 blgp:3
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 2
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v33, v[24:27], s[34:35] offset:96
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v33, v[28:31], s[34:35] offset:112
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v33, v[16:19], s[34:35] offset:64
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v33, v[20:23], s[34:35] offset:80
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v33, v[8:11], s[34:35] offset:32
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v33, v[12:15], s[34:35] offset:48
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v33, v[0:3], s[34:35]
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v33, v[4:7], s[34:35] offset:16
+; GFX90A-VGPR-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_32x32x4bf16_1k:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[34:35], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v33, 0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v34, 1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v35, v33
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v32, 2
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx16 s[16:31], s[34:35], 0x0
+; GFX942-VGPR-NEXT:    s_load_dwordx16 s[0:15], s[34:35], 0x40
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, s16
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, s17
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, s18
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, s19
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, s20
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, s21
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, s22
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v7, s23
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v8, s24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v9, s25
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v10, s26
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v11, s27
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v12, s28
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v13, s29
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v14, s30
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v15, s31
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, s0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v17, s1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v18, s2
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v19, s3
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v20, s4
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v21, s5
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v22, s6
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v23, s7
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v24, s8
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v25, s9
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v26, s10
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v27, s11
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v28, s12
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v29, s13
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v30, s14
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v31, s15
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_32x32x4_2b_bf16 v[0:31], v[34:35], v[32:33], v[0:31] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 2
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v33, v[24:27], s[34:35] offset:96
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v33, v[28:31], s[34:35] offset:112
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v33, v[16:19], s[34:35] offset:64
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v33, v[20:23], s[34:35] offset:80
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v33, v[8:11], s[34:35] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v33, v[12:15], s[34:35] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v33, v[0:3], s[34:35]
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v33, v[4:7], s[34:35] offset:16
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <32 x float>, ptr addrspace(1) %arg
   %a = bitcast i64 1 to <4 x i16>
@@ -30,16 +254,134 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_16x16x4bf16_1k:
-; GCN-DAG:      s_load_dwordx16
-; GCN-DAG:      v_mov_b32_e32 v[[TWO:[0-9]+]], 2
-; GCN-DAG:      v_mov_b32_e32 v[[ONE:[0-9]+]], 1
-; GCN-COUNT-16: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GFX90A:       v_mfma_f32_16x16x4bf16_1k a[{{[0-9]+:[0-9]+}}], v[[[ONE]]:{{[0-9]+}}], v[[[TWO]]:{{[0-9]+}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GFX942:       v_mfma_f32_16x16x4_4b_bf16 a[{{[0-9]+:[0-9]+}}], v[[[ONE]]:{{[0-9+]}}], v[[[TWO]]:{{[0-9+]}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GCN-NOT:      v_accvgpr_read_b32
-; GCN-COUNT-4:  global_store_dwordx4 v{{[0-9]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_16x16x4bf16_1k(ptr addrspace(1) %arg) #0 {
+; GFX90A-LABEL: test_mfma_f32_16x16x4bf16_1k:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 0
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 1
+; GFX90A-NEXT:    v_mov_b32_e32 v3, v1
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 2
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX90A-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX90A-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX90A-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX90A-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX90A-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX90A-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX90A-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX90A-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX90A-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX90A-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX90A-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX90A-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX90A-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX90A-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f32_16x16x4bf16_1k a[0:15], v[2:3], v[0:1], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 2
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[12:15], s[16:17] offset:48
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[8:11], s[16:17] offset:32
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[4:7], s[16:17] offset:16
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[0:3], s[16:17]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_f32_16x16x4bf16_1k:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-NEXT:    v_mov_b32_e32 v1, 0
+; GFX942-NEXT:    v_mov_b32_e32 v2, 1
+; GFX942-NEXT:    v_mov_b32_e32 v3, v1
+; GFX942-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX942-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX942-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX942-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX942-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX942-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX942-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX942-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX942-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX942-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX942-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX942-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX942-NEXT:    s_nop 1
+; GFX942-NEXT:    v_mfma_f32_16x16x4_4b_bf16 a[0:15], v[2:3], v[0:1], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 2
+; GFX942-NEXT:    global_store_dwordx4 v1, a[12:15], s[16:17] offset:48
+; GFX942-NEXT:    global_store_dwordx4 v1, a[8:11], s[16:17] offset:32
+; GFX942-NEXT:    global_store_dwordx4 v1, a[4:7], s[16:17] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v1, a[0:3], s[16:17]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX90A-VGPR-LABEL: test_mfma_f32_16x16x4bf16_1k:
+; GFX90A-VGPR:       ; %bb.0: ; %bb
+; GFX90A-VGPR-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v17, 0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v18, 1
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v19, v17
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v16, 2
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[8:9], s[8:9], s[8:9] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[10:11], s[10:11], s[10:11] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[12:13], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[14:15], s[14:15], s[14:15] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    v_mfma_f32_16x16x4bf16_1k v[0:15], v[18:19], v[16:17], v[0:15] cbsz:1 abid:2 blgp:3
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 2
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v17, v[12:15], s[16:17] offset:48
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v17, v[8:11], s[16:17] offset:32
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v17, v[4:7], s[16:17] offset:16
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v17, v[0:3], s[16:17]
+; GFX90A-VGPR-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_16x16x4bf16_1k:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v17, 0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v18, 1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v19, v17
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, 2
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_16x16x4_4b_bf16 v[0:15], v[18:19], v[16:17], v[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 2
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v17, v[12:15], s[16:17] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v17, v[8:11], s[16:17] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v17, v[4:7], s[16:17] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v17, v[0:3], s[16:17]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %a = bitcast i64 1 to <4 x i16>
@@ -49,16 +391,82 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_4x4x4bf16_1k:
-; GCN-DAG:     s_load_dwordx4
-; GCN-DAG:     v_mov_b32_e32 v[[TWO:[0-9]+]], 2
-; GCN-DAG:     v_mov_b32_e32 v[[ONE:[0-9]+]], 1
-; GCN-COUNT-4: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GFX90A:      v_mfma_f32_4x4x4bf16_1k [[RES:a\[[0-9]+:[0-9]+\]]], v[[[ONE]]:{{[0-9]+}}], v[[[TWO]]:{{[0-9]+}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GFX942:      v_mfma_f32_4x4x4_16b_bf16 [[RES:a\[[0-9]+:[0-9]+\]]], v[[[ONE]]:{{[0-9+]}}], v[[[TWO]]:{{[0-9+]}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GCN-NOT:     v_accvgpr_read_b32
-; GCN:         global_store_dwordx4 v{{[0-9]+}}, [[RES]],
 define amdgpu_kernel void @test_mfma_f32_4x4x4bf16_1k(ptr addrspace(1) %arg) #0 {
+; GFX90A-LABEL: test_mfma_f32_4x4x4bf16_1k:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 0
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 1
+; GFX90A-NEXT:    v_mov_b32_e32 v3, v1
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 2
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX90A-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX90A-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f32_4x4x4bf16_1k a[0:3], v[2:3], v[0:1], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    s_nop 4
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[0:3], s[6:7]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_f32_4x4x4bf16_1k:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-NEXT:    v_mov_b32_e32 v1, 0
+; GFX942-NEXT:    v_mov_b32_e32 v2, 1
+; GFX942-NEXT:    v_mov_b32_e32 v3, v1
+; GFX942-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-NEXT:    s_nop 1
+; GFX942-NEXT:    v_mfma_f32_4x4x4_16b_bf16 a[0:3], v[2:3], v[0:1], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-NEXT:    s_nop 4
+; GFX942-NEXT:    global_store_dwordx4 v1, a[0:3], s[6:7]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX90A-VGPR-LABEL: test_mfma_f32_4x4x4bf16_1k:
+; GFX90A-VGPR:       ; %bb.0: ; %bb
+; GFX90A-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v5, 0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v6, 1
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v7, v5
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v4, 2
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    v_mfma_f32_4x4x4bf16_1k v[0:3], v[6:7], v[4:5], v[0:3] cbsz:1 abid:2 blgp:3
+; GFX90A-VGPR-NEXT:    s_nop 4
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v5, v[0:3], s[6:7]
+; GFX90A-VGPR-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_4x4x4bf16_1k:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, 0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, 1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v7, v5
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, 2
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_4x4x4_16b_bf16 v[0:3], v[6:7], v[4:5], v[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_nop 4
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v5, v[0:3], s[6:7]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %a = bitcast i64 1 to <4 x i16>
@@ -68,16 +476,136 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_32x32x8bf16_1k:
-; GCN-DAG:      s_load_dwordx16
-; GCN-DAG:      v_mov_b32_e32 v[[TWO:[0-9]+]], 2
-; GCN-DAG:      v_mov_b32_e32 v[[ONE:[0-9]+]], 1
-; GCN-COUNT-16: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GFX90A:       v_mfma_f32_32x32x8bf16_1k a[{{[0-9]+:[0-9]+}}], v[[[ONE]]:{{[0-9]+}}], v[[[TWO]]:{{[0-9]+}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GFX942:       v_mfma_f32_32x32x8_bf16 a[{{[0-9]+:[0-9]+}}], v[[[ONE]]:{{[0-9+]}}], v[[[TWO]]:{{[0-9+]}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GCN-NOT:      v_accvgpr_read_b32
-; GCN-COUNT-4:  global_store_dwordx4 v{{[0-9]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_32x32x8bf16_1k(ptr addrspace(1) %arg) #0 {
+; GFX90A-LABEL: test_mfma_f32_32x32x8bf16_1k:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 0
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 1
+; GFX90A-NEXT:    v_mov_b32_e32 v3, v1
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 2
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX90A-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX90A-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX90A-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX90A-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX90A-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX90A-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX90A-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX90A-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX90A-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX90A-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX90A-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX90A-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX90A-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX90A-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f32_32x32x8bf16_1k a[0:15], v[2:3], v[0:1], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 2
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[12:15], s[16:17] offset:48
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[8:11], s[16:17] offset:32
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[4:7], s[16:17] offset:16
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[0:3], s[16:17]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_f32_32x32x8bf16_1k:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-NEXT:    v_mov_b32_e32 v1, 0
+; GFX942-NEXT:    v_mov_b32_e32 v2, 1
+; GFX942-NEXT:    v_mov_b32_e32 v3, v1
+; GFX942-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX942-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX942-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX942-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX942-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX942-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX942-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX942-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX942-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX942-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX942-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX942-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX942-NEXT:    s_nop 1
+; GFX942-NEXT:    v_mfma_f32_32x32x8_bf16 a[0:15], v[2:3], v[0:1], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 2
+; GFX942-NEXT:    global_store_dwordx4 v1, a[12:15], s[16:17] offset:48
+; GFX942-NEXT:    global_store_dwordx4 v1, a[8:11], s[16:17] offset:32
+; GFX942-NEXT:    global_store_dwordx4 v1, a[4:7], s[16:17] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v1, a[0:3], s[16:17]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX90A-VGPR-LABEL: test_mfma_f32_32x32x8bf16_1k:
+; GFX90A-VGPR:       ; %bb.0: ; %bb
+; GFX90A-VGPR-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v17, 0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v18, 1
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v19, v17
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v16, 2
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[8:9], s[8:9], s[8:9] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[10:11], s[10:11], s[10:11] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[12:13], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[14:15], s[14:15], s[14:15] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    v_mfma_f32_32x32x8bf16_1k v[0:15], v[18:19], v[16:17], v[0:15] cbsz:1 abid:2 blgp:3
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 2
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v17, v[12:15], s[16:17] offset:48
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v17, v[8:11], s[16:17] offset:32
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v17, v[4:7], s[16:17] offset:16
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v17, v[0:3], s[16:17]
+; GFX90A-VGPR-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_32x32x8bf16_1k:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v17, 0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v18, 1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v19, v17
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, 2
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_32x32x8_bf16 v[0:15], v[18:19], v[16:17], v[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 2
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v17, v[12:15], s[16:17] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v17, v[8:11], s[16:17] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v17, v[4:7], s[16:17] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v17, v[0:3], s[16:17]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %a = bitcast i64 1 to <4 x i16>
@@ -87,16 +615,84 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_16x16x16bf16_1k:
-; GCN-DAG:     s_load_dwordx4
-; GCN-DAG:     v_mov_b32_e32 v[[TWO:[0-9]+]], 2
-; GCN-DAG:     v_mov_b32_e32 v[[ONE:[0-9]+]], 1
-; GCN-COUNT-4: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GFX90A:      v_mfma_f32_16x16x16bf16_1k [[RES:a\[[0-9]+:[0-9]+\]]], v[[[ONE]]:{{[0-9]+}}], v[[[TWO]]:{{[0-9]+}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GFX942:      v_mfma_f32_16x16x16_bf16 [[RES:a\[[0-9]+:[0-9]+\]]], v[[[ONE]]:{{[0-9+]}}], v[[[TWO]]:{{[0-9+]}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GCN-NOT:     v_accvgpr_read_b32
-; GCN:         global_store_dwordx4 v{{[0-9]+}}, [[RES]],
 define amdgpu_kernel void @test_mfma_f32_16x16x16bf16_1k(ptr addrspace(1) %arg) #0 {
+; GFX90A-LABEL: test_mfma_f32_16x16x16bf16_1k:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 0
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 1
+; GFX90A-NEXT:    v_mov_b32_e32 v3, v1
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 2
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX90A-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX90A-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f32_16x16x16bf16_1k a[0:3], v[2:3], v[0:1], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 2
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[0:3], s[6:7]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_f32_16x16x16bf16_1k:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-NEXT:    v_mov_b32_e32 v1, 0
+; GFX942-NEXT:    v_mov_b32_e32 v2, 1
+; GFX942-NEXT:    v_mov_b32_e32 v3, v1
+; GFX942-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-NEXT:    s_nop 1
+; GFX942-NEXT:    v_mfma_f32_16x16x16_bf16 a[0:3], v[2:3], v[0:1], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-NEXT:    s_nop 6
+; GFX942-NEXT:    global_store_dwordx4 v1, a[0:3], s[6:7]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX90A-VGPR-LABEL: test_mfma_f32_16x16x16bf16_1k:
+; GFX90A-VGPR:       ; %bb.0: ; %bb
+; GFX90A-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v5, 0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v6, 1
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v7, v5
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v4, 2
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    v_mfma_f32_16x16x16bf16_1k v[0:3], v[6:7], v[4:5], v[0:3] cbsz:1 abid:2 blgp:3
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 2
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v5, v[0:3], s[6:7]
+; GFX90A-VGPR-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_16x16x16bf16_1k:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, 0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, 1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v7, v5
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, 2
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_16x16x16_bf16 v[0:3], v[6:7], v[4:5], v[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_nop 6
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v5, v[0:3], s[6:7]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %a = bitcast i64 1 to <4 x i16>
@@ -106,13 +702,70 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f64_4x4x4f64:
-; GFX90A: v_mfma_f64_4x4x4f64 [[M1:a\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], 0{{$}}
-; GFX90A: v_mfma_f64_4x4x4f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], [[M1]] cbsz:1 abid:2 blgp:3
-; GFX942: v_mfma_f64_4x4x4_4b_f64 [[M1:a\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], 0{{$}}
-; GFX942: v_mfma_f64_4x4x4_4b_f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], [[M1]] cbsz:1 abid:2 neg:[1,1,0]
-; GCN:    global_store_dwordx2
 define amdgpu_kernel void @test_mfma_f64_4x4x4f64(ptr addrspace(1) %arg, double %a, double %b) #0 {
+; GFX90A-LABEL: test_mfma_f64_4x4x4f64:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f64_4x4x4f64 a[0:1], v[0:1], v[2:3], 0
+; GFX90A-NEXT:    s_nop 3
+; GFX90A-NEXT:    v_mfma_f64_4x4x4f64 a[0:1], v[0:1], v[2:3], a[0:1] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    global_store_dwordx2 v0, a[0:1], s[0:1]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_f64_4x4x4f64:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    s_nop 1
+; GFX942-NEXT:    v_mfma_f64_4x4x4_4b_f64 a[0:1], v[0:1], v[2:3], 0
+; GFX942-NEXT:    s_nop 3
+; GFX942-NEXT:    v_mfma_f64_4x4x4_4b_f64 a[0:1], v[0:1], v[2:3], a[0:1] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    global_store_dwordx2 v0, a[0:1], s[0:1]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX90A-VGPR-LABEL: test_mfma_f64_4x4x4f64:
+; GFX90A-VGPR:       ; %bb.0: ; %bb
+; GFX90A-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[0:1], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    v_mfma_f64_4x4x4f64 v[4:5], v[0:1], v[2:3], 0
+; GFX90A-VGPR-NEXT:    s_nop 3
+; GFX90A-VGPR-NEXT:    v_mfma_f64_4x4x4f64 v[0:1], v[0:1], v[2:3], v[4:5] cbsz:1 abid:2 blgp:3
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX90A-VGPR-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f64_4x4x4f64:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f64_4x4x4_4b_f64 v[4:5], v[0:1], v[2:3], 0
+; GFX942-VGPR-NEXT:    s_nop 3
+; GFX942-VGPR-NEXT:    v_mfma_f64_4x4x4_4b_f64 v[0:1], v[0:1], v[2:3], v[4:5] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call double @llvm.amdgcn.mfma.f64.4x4x4f64(double %a, double %b, double 0.0, i32 0, i32 0, i32 0)
   %mai.2 = tail call double @llvm.amdgcn.mfma.f64.4x4x4f64(double %a, double %b, double %mai.1, i32 1, i32 2, i32 3)
@@ -120,13 +773,110 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f64_16x16x4f64:
-; GCN:    s_load_dwordx8
-; GFX90A: v_mfma_f64_16x16x4f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GFX942: v_mfma_f64_16x16x4_f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 neg:[1,1,0]
-; GCN:    global_store_dwordx4
-; GCN:    global_store_dwordx4
 define amdgpu_kernel void @test_mfma_f64_16x16x4f64(ptr addrspace(1) %arg, double %a, double %b) #0 {
+; GFX90A-LABEL: test_mfma_f64_16x16x4f64:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
+; GFX90A-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_mov_b32_e32 v2, s10
+; GFX90A-NEXT:    s_load_dwordx8 s[0:7], s[8:9], 0x0
+; GFX90A-NEXT:    v_mov_b32_e32 v3, s11
+; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX90A-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX90A-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX90A-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX90A-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX90A-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX90A-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f64_16x16x4f64 a[0:7], v[2:3], v[0:1], a[0:7] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[4:7], s[8:9] offset:16
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[8:9]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_f64_16x16x4f64:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
+; GFX942-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    v_mov_b32_e32 v2, s10
+; GFX942-NEXT:    s_load_dwordx8 s[0:7], s[8:9], 0x0
+; GFX942-NEXT:    v_mov_b32_e32 v3, s11
+; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[12:13]
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX942-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX942-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX942-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX942-NEXT:    s_nop 1
+; GFX942-NEXT:    v_mfma_f64_16x16x4_f64 a[0:7], v[2:3], v[0:1], a[0:7] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[8:9] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[8:9]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX90A-VGPR-LABEL: test_mfma_f64_16x16x4f64:
+; GFX90A-VGPR:       ; %bb.0: ; %bb
+; GFX90A-VGPR-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
+; GFX90A-VGPR-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v10, s10
+; GFX90A-VGPR-NEXT:    s_load_dwordx8 s[0:7], s[8:9], 0x0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v11, s11
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[8:9], s[12:13], s[12:13] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[0:1], s[0:1], s[0:1] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[2:3], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[4:5], s[4:5], s[4:5] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[6:7], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[10:11], v[8:9], v[0:7] cbsz:1 abid:2 blgp:3
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v8, 0
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 0
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
+; GFX90A-VGPR-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f64_16x16x4f64:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x34
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v10, s10
+; GFX942-VGPR-NEXT:    s_load_dwordx8 s[0:7], s[8:9], 0x0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v11, s11
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], s[12:13]
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[10:11], v[8:9], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v8, 0
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v8, v[4:7], s[8:9] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v8, v[0:3], s[8:9]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x double>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> %in.1, i32 1, i32 2, i32 3)
@@ -134,14 +884,78 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f64_16x16x4f64_splat_imm_0:
-; GFX90A: v_mfma_f64_16x16x4f64 [[M1:a\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], 0{{$}}
-; GFX90A: v_mfma_f64_16x16x4f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], [[M1]] cbsz:1 abid:2 blgp:3
-; GFX942: v_mfma_f64_16x16x4_f64 [[M1:a\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], 0{{$}}
-; GFX942: v_mfma_f64_16x16x4_f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], [[M1]] cbsz:1 abid:2 neg:[1,1,0]
-; GCN:    global_store_dwordx4
-; GCN:    global_store_dwordx4
 define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_0(ptr addrspace(1) %arg, double %a, double %b) #0 {
+; GFX90A-LABEL: test_mfma_f64_16x16x4f64_splat_imm_0:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], 0
+; GFX90A-NEXT:    v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], a[0:7] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_f64_16x16x4f64_splat_imm_0:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    s_nop 1
+; GFX942-NEXT:    v_mfma_f64_16x16x4_f64 a[0:7], v[0:1], v[2:3], 0
+; GFX942-NEXT:    v_mfma_f64_16x16x4_f64 a[0:7], v[0:1], v[2:3], a[0:7] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX90A-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_0:
+; GFX90A-VGPR:       ; %bb.0: ; %bb
+; GFX90A-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], 0
+; GFX90A-VGPR-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v8, 0
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 0
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-VGPR-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_0:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], 0
+; GFX942-VGPR-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v8, 0
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> zeroinitializer, i32 0, i32 0, i32 0)
   %mai.2 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> %mai.1, i32 1, i32 2, i32 3)
@@ -149,14 +963,78 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f64_16x16x4f64_splat_imm_int_neg1:
-; GFX90A: v_mfma_f64_16x16x4f64 [[M1:a\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], -1{{$}}
-; GFX90A: v_mfma_f64_16x16x4f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], [[M1]] cbsz:1 abid:2 blgp:3
-; GFX942: v_mfma_f64_16x16x4_f64 [[M1:a\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], -1{{$}}
-; GFX942: v_mfma_f64_16x16x4_f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], [[M1]] cbsz:1 abid:2 neg:[1,1,0]
-; GCN:    global_store_dwordx4
-; GCN:    global_store_dwordx4
 define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_int_neg1(ptr addrspace(1) %arg, double %a, double %b) #0 {
+; GFX90A-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_neg1:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], -1
+; GFX90A-NEXT:    v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], a[0:7] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_neg1:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    s_nop 1
+; GFX942-NEXT:    v_mfma_f64_16x16x4_f64 a[0:7], v[0:1], v[2:3], -1
+; GFX942-NEXT:    v_mfma_f64_16x16x4_f64 a[0:7], v[0:1], v[2:3], a[0:7] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX90A-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_neg1:
+; GFX90A-VGPR:       ; %bb.0: ; %bb
+; GFX90A-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], -1
+; GFX90A-VGPR-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v8, 0
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 0
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-VGPR-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_neg1:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], -1
+; GFX942-VGPR-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v8, 0
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> splat (double bitcast (i64 -1 to double)), i32 0, i32 0, i32 0)
   %mai.2 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> %mai.1, i32 1, i32 2, i32 3)
@@ -164,14 +1042,78 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f64_16x16x4f64_splat_imm_1:
-; GFX90A: v_mfma_f64_16x16x4f64 [[M1:a\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], 1.0{{$}}
-; GFX90A: v_mfma_f64_16x16x4f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], [[M1]] cbsz:1 abid:2 blgp:3
-; GFX942: v_mfma_f64_16x16x4_f64 [[M1:a\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], 1.0{{$}}
-; GFX942: v_mfma_f64_16x16x4_f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], [[M1]] cbsz:1 abid:2 neg:[1,1,0]
-; GCN:    global_store_dwordx4
-; GCN:    global_store_dwordx4
 define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_1(ptr addrspace(1) %arg, double %a, double %b) #0 {
+; GFX90A-LABEL: test_mfma_f64_16x16x4f64_splat_imm_1:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], 1.0
+; GFX90A-NEXT:    v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], a[0:7] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_f64_16x16x4f64_splat_imm_1:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    s_nop 1
+; GFX942-NEXT:    v_mfma_f64_16x16x4_f64 a[0:7], v[0:1], v[2:3], 1.0
+; GFX942-NEXT:    v_mfma_f64_16x16x4_f64 a[0:7], v[0:1], v[2:3], a[0:7] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX90A-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_1:
+; GFX90A-VGPR:       ; %bb.0: ; %bb
+; GFX90A-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], 1.0
+; GFX90A-VGPR-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v8, 0
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 0
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-VGPR-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_1:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], 1.0
+; GFX942-VGPR-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v8, 0
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> splat (double 1.0), i32 0, i32 0, i32 0)
   %mai.2 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> %mai.1, i32 1, i32 2, i32 3)
@@ -179,14 +1121,78 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f64_16x16x4f64_splat_imm_neg1:
-; GFX90A: v_mfma_f64_16x16x4f64 [[M1:a\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], -1.0{{$}}
-; GFX90A: v_mfma_f64_16x16x4f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], [[M1]] cbsz:1 abid:2 blgp:3
-; GFX942: v_mfma_f64_16x16x4_f64 [[M1:a\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], -1.0{{$}}
-; GFX942: v_mfma_f64_16x16x4_f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], [[M1]] cbsz:1 abid:2 neg:[1,1,0]
-; GCN:    global_store_dwordx4
-; GCN:    global_store_dwordx4
 define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_neg1(ptr addrspace(1) %arg, double %a, double %b) #0 {
+; GFX90A-LABEL: test_mfma_f64_16x16x4f64_splat_imm_neg1:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], -1.0
+; GFX90A-NEXT:    v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], a[0:7] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_f64_16x16x4f64_splat_imm_neg1:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    s_nop 1
+; GFX942-NEXT:    v_mfma_f64_16x16x4_f64 a[0:7], v[0:1], v[2:3], -1.0
+; GFX942-NEXT:    v_mfma_f64_16x16x4_f64 a[0:7], v[0:1], v[2:3], a[0:7] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX90A-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_neg1:
+; GFX90A-VGPR:       ; %bb.0: ; %bb
+; GFX90A-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], -1.0
+; GFX90A-VGPR-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v8, 0
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 0
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-VGPR-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_neg1:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], -1.0
+; GFX942-VGPR-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v8, 0
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> splat (double -1.0), i32 0, i32 0, i32 0)
   %mai.2 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> %mai.1, i32 1, i32 2, i32 3)
@@ -194,14 +1200,78 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f64_16x16x4f64_splat_imm_int_64:
-; GFX90A: v_mfma_f64_16x16x4f64 [[M1:a\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], 64{{$}}
-; GFX90A: v_mfma_f64_16x16x4f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], [[M1]] cbsz:1 abid:2 blgp:3
-; GFX942: v_mfma_f64_16x16x4_f64 [[M1:a\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], 64{{$}}
-; GFX942: v_mfma_f64_16x16x4_f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], [[M1]] cbsz:1 abid:2 neg:[1,1,0]
-; GCN:    global_store_dwordx4
-; GCN:    global_store_dwordx4
 define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_int_64(ptr addrspace(1) %arg, double %a, double %b) #0 {
+; GFX90A-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], 64
+; GFX90A-NEXT:    v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], a[0:7] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    s_nop 1
+; GFX942-NEXT:    v_mfma_f64_16x16x4_f64 a[0:7], v[0:1], v[2:3], 64
+; GFX942-NEXT:    v_mfma_f64_16x16x4_f64 a[0:7], v[0:1], v[2:3], a[0:7] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX90A-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64:
+; GFX90A-VGPR:       ; %bb.0: ; %bb
+; GFX90A-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], 64
+; GFX90A-VGPR-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v8, 0
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 0
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-VGPR-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], 64
+; GFX942-VGPR-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v8, 0
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> splat (double bitcast (i64 64 to double)), i32 0, i32 0, i32 0)
   %mai.2 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> %mai.1, i32 1, i32 2, i32 3)
@@ -209,23 +1279,116 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_bits:
-; GCN: v_accvgpr_write_b32 a[[A_LOW_BITS_0:[0-9]+]], 0{{$}}
-; GCN: v_accvgpr_write_b32 a[[A_HIGH_BITS_0:[0-9]+]], 64
-; GCN: v_accvgpr_mov_b32 a{{[0-9]+}}, a[[A_LOW_BITS_0]]
-; GCN: v_accvgpr_mov_b32 a{{[0-9]+}}, a[[A_HIGH_BITS_0]]
-; GCN: v_accvgpr_mov_b32 a{{[0-9]+}}, a[[A_LOW_BITS_0]]
-; GCN: v_accvgpr_mov_b32 a{{[0-9]+}}, a[[A_HIGH_BITS_0]]
-; GCN: v_accvgpr_mov_b32 a{{[0-9]+}}, a[[A_LOW_BITS_0]]
-; GCN: v_accvgpr_mov_b32 a[[LAST_CONST_REG:[0-9]+]], a[[A_HIGH_BITS_0]]
-
-; GFX90A: v_mfma_f64_16x16x4f64 [[M1:a\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a{{\[}}[[A_LOW_BITS_0]]:[[LAST_CONST_REG]]{{\]$}}
-; GFX90A: v_mfma_f64_16x16x4f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], [[M1]] cbsz:1 abid:2 blgp:3
-; GFX942: v_mfma_f64_16x16x4_f64 [[M1:a\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a{{\[}}[[A_LOW_BITS_0]]:[[LAST_CONST_REG]]{{\]$}}
-; GFX942: v_mfma_f64_16x16x4_f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], [[M1]] cbsz:1 abid:2 neg:[1,1,0]
-; GCN:    global_store_dwordx4
-; GCN:    global_store_dwordx4
 define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_bits(ptr addrspace(1) %arg, double %a, double %b) #0 {
+; GFX90A-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_bits:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, 0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, 64
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a2, a0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a3, a1
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a4, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a5, a1
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a6, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a7, a1
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], a[0:7]
+; GFX90A-NEXT:    v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], a[0:7] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_bits:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-NEXT:    v_accvgpr_write_b32 a0, 0
+; GFX942-NEXT:    v_accvgpr_write_b32 a1, 64
+; GFX942-NEXT:    v_accvgpr_mov_b32 a2, a0
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
+; GFX942-NEXT:    v_accvgpr_mov_b32 a3, a1
+; GFX942-NEXT:    v_accvgpr_mov_b32 a4, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a5, a1
+; GFX942-NEXT:    v_accvgpr_mov_b32 a6, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a7, a1
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    s_nop 1
+; GFX942-NEXT:    v_mfma_f64_16x16x4_f64 a[0:7], v[0:1], v[2:3], a[0:7]
+; GFX942-NEXT:    v_mfma_f64_16x16x4_f64 a[0:7], v[0:1], v[2:3], a[0:7] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX90A-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_bits:
+; GFX90A-VGPR:       ; %bb.0: ; %bb
+; GFX90A-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v1, 64
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v2, v0
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[10:11], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v3, v1
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v4, v0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v5, v1
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v6, v0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v7, v1
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[8:9], v[6:7], v[6:7] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[12:13], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[6:7], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[4:5], v[2:3], v[2:3] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    v_mfma_f64_16x16x4f64 v[2:9], v[10:11], v[12:13], v[2:9]
+; GFX90A-VGPR-NEXT:    v_mfma_f64_16x16x4f64 v[2:9], v[10:11], v[12:13], v[2:9] cbsz:1 abid:2 blgp:3
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v0, v[6:9], s[0:1] offset:16
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1]
+; GFX90A-VGPR-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_bits:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, 64
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[10:11], s[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v7, v1
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], v[6:7]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[12:13], s[6:7]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[6:7], v[4:5]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[4:5], v[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f64_16x16x4_f64 v[2:9], v[10:11], v[12:13], v[2:9]
+; GFX942-VGPR-NEXT:    v_mfma_f64_16x16x4_f64 v[2:9], v[10:11], v[12:13], v[2:9] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v0, v[6:9], s[0:1] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> splat (double bitcast (i64 274877906944 to double)), i32 0, i32 0, i32 0)
   %mai.2 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> %mai.1, i32 1, i32 2, i32 3)
@@ -233,23 +1396,110 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_and_low:
-; GCN: v_accvgpr_write_b32 a[[A_LOW_BITS_0:[0-9]+]], 64{{$}}
-; GCN: v_accvgpr_mov_b32 a{{[0-9]+}}, a[[A_LOW_BITS_0]]
-; GCN: v_accvgpr_mov_b32 a{{[0-9]+}}, a[[A_LOW_BITS_0]]
-; GCN: v_accvgpr_mov_b32 a{{[0-9]+}}, a[[A_LOW_BITS_0]]
-; GCN: v_accvgpr_mov_b32 a{{[0-9]+}}, a[[A_LOW_BITS_0]]
-; GCN: v_accvgpr_mov_b32 a{{[0-9]+}}, a[[A_LOW_BITS_0]]
-; GCN: v_accvgpr_mov_b32 a{{[0-9]+}}, a[[A_LOW_BITS_0]]
-; GCN: v_accvgpr_mov_b32 a[[LAST_CONST_REG:[0-9]+]], a[[A_LOW_BITS_0]]
-
-; GFX90A: v_mfma_f64_16x16x4f64 [[M1:a\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a{{\[}}[[A_LOW_BITS_0]]:[[LAST_CONST_REG]]{{\]$}}
-; GFX90A: v_mfma_f64_16x16x4f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], [[M1]] cbsz:1 abid:2 blgp:3
-; GFX942: v_mfma_f64_16x16x4_f64 [[M1:a\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a{{\[}}[[A_LOW_BITS_0]]:[[LAST_CONST_REG]]{{\]$}}
-; GFX942: v_mfma_f64_16x16x4_f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], [[M1]] cbsz:1 abid:2 neg:[1,1,0]
-; GCN:    global_store_dwordx4
-; GCN:    global_store_dwordx4
 define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_and_low(ptr addrspace(1) %arg, double %a, double %b) #0 {
+; GFX90A-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_and_low:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, 64
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a1, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a2, a0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a3, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a4, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a5, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a6, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a7, a0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], a[0:7]
+; GFX90A-NEXT:    v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], a[0:7] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_and_low:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-NEXT:    v_accvgpr_write_b32 a0, 64
+; GFX942-NEXT:    v_accvgpr_mov_b32 a1, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a2, a0
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
+; GFX942-NEXT:    v_accvgpr_mov_b32 a3, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a4, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a5, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a6, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a7, a0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    s_nop 1
+; GFX942-NEXT:    v_mfma_f64_16x16x4_f64 a[0:7], v[0:1], v[2:3], a[0:7]
+; GFX942-NEXT:    v_mfma_f64_16x16x4_f64 a[0:7], v[0:1], v[2:3], a[0:7] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX90A-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_and_low:
+; GFX90A-VGPR:       ; %bb.0: ; %bb
+; GFX90A-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v0, 64
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v1, v0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v2, v0
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v3, v0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v4, v0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v5, v0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v6, v0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v7, v0
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
+; GFX90A-VGPR-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v8, 0
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 0
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-VGPR-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_int_64_in_high_and_low:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 64
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v7, v0
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7]
+; GFX942-VGPR-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v8, 0
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> splat (double bitcast (i64 274877907008 to double)), i32 0, i32 0, i32 0)
   %mai.2 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> %mai.1, i32 1, i32 2, i32 3)
@@ -257,23 +1507,110 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f64_16x16x4f64_splat_imm_f32_1_in_high_and_low:
-; GCN: v_accvgpr_write_b32 a[[A_LOW_BITS_0:[0-9]+]], 1.0
-; GCN: v_accvgpr_mov_b32 a{{[0-9]+}}, a[[A_LOW_BITS_0]]
-; GCN: v_accvgpr_mov_b32 a{{[0-9]+}}, a[[A_LOW_BITS_0]]
-; GCN: v_accvgpr_mov_b32 a{{[0-9]+}}, a[[A_LOW_BITS_0]]
-; GCN: v_accvgpr_mov_b32 a{{[0-9]+}}, a[[A_LOW_BITS_0]]
-; GCN: v_accvgpr_mov_b32 a{{[0-9]+}}, a[[A_LOW_BITS_0]]
-; GCN: v_accvgpr_mov_b32 a{{[0-9]+}}, a[[A_LOW_BITS_0]]
-; GCN: v_accvgpr_mov_b32 a[[LAST_CONST_REG:[0-9]+]], a[[A_LOW_BITS_0]]
-
-; GFX90A: v_mfma_f64_16x16x4f64 [[M1:a\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a{{\[}}[[A_LOW_BITS_0]]:[[LAST_CONST_REG]]{{\]$}}
-; GFX90A: v_mfma_f64_16x16x4f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], [[M1]] cbsz:1 abid:2 blgp:3
-; GFX942: v_mfma_f64_16x16x4_f64 [[M1:a\[[0-9]+:[0-9]+\]]], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a{{\[}}[[A_LOW_BITS_0]]:[[LAST_CONST_REG]]{{\]$}}
-; GFX942: v_mfma_f64_16x16x4_f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], [[M1]] cbsz:1 abid:2 neg:[1,1,0]
-; GCN:    global_store_dwordx4
-; GCN:    global_store_dwordx4
 define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_imm_f32_1_in_high_and_low(ptr addrspace(1) %arg, double %a, double %b) #0 {
+; GFX90A-LABEL: test_mfma_f64_16x16x4f64_splat_imm_f32_1_in_high_and_low:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, 1.0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a1, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a2, a0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a3, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a4, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a5, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a6, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a7, a0
+; GFX90A-NEXT:    v_pk_mov_b32 v[2:3], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], a[0:7]
+; GFX90A-NEXT:    v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], a[0:7] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_f64_16x16x4f64_splat_imm_f32_1_in_high_and_low:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-NEXT:    v_accvgpr_write_b32 a0, 1.0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a1, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a2, a0
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[2:3]
+; GFX942-NEXT:    v_accvgpr_mov_b32 a3, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a4, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a5, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a6, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a7, a0
+; GFX942-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-NEXT:    s_nop 1
+; GFX942-NEXT:    v_mfma_f64_16x16x4_f64 a[0:7], v[0:1], v[2:3], a[0:7]
+; GFX942-NEXT:    v_mfma_f64_16x16x4_f64 a[0:7], v[0:1], v[2:3], a[0:7] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX90A-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_f32_1_in_high_and_low:
+; GFX90A-VGPR:       ; %bb.0: ; %bb
+; GFX90A-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v1, v0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v2, v0
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[8:9], s[2:3], s[2:3] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v3, v0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v4, v0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v5, v0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v6, v0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v7, v0
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7]
+; GFX90A-VGPR-NEXT:    v_mfma_f64_16x16x4f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 blgp:3
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v8, 0
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 0
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX90A-VGPR-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_imm_f32_1_in_high_and_low:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v7, v0
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7]
+; GFX942-VGPR-NEXT:    v_mfma_f64_16x16x4_f64 v[0:7], v[8:9], v[10:11], v[0:7] cbsz:1 abid:2 neg:[1,1,0]
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v8, 0
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v8, v[4:7], s[0:1] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> splat (double bitcast (<2 x float> splat (float 1.0) to double)), i32 0, i32 0, i32 0)
   %mai.2 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> %mai.1, i32 1, i32 2, i32 3)
@@ -281,26 +1618,236 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f64_16x16x4f64_imm:
-; GFX90A: v_mfma_f64_16x16x4f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a[{{[0-9]+:[0-9]+}}]{{$}}
-; GFX942: v_mfma_f64_16x16x4_f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a[{{[0-9]+:[0-9]+}}]{{$}}
-; GCN:    global_store_dwordx4
-; GCN:    global_store_dwordx4
 define amdgpu_kernel void @test_mfma_f64_16x16x4f64_imm(ptr addrspace(1) %arg, double %a, double %b) #0 {
+; GFX90A-LABEL: test_mfma_f64_16x16x4f64_imm:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, 0
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0x3ff00000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a7, v0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_mov_b32_e32 v2, s2
+; GFX90A-NEXT:    v_mov_b32_e32 v3, s3
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a1, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a2, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a3, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a4, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a5, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a6, a0
+; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f64_16x16x4f64 a[0:7], v[2:3], v[0:1], a[0:7]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_f64_16x16x4f64_imm:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-NEXT:    v_accvgpr_write_b32 a0, 0
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0x3ff00000
+; GFX942-NEXT:    v_accvgpr_write_b32 a7, v0
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    v_mov_b32_e32 v2, s2
+; GFX942-NEXT:    v_mov_b32_e32 v3, s3
+; GFX942-NEXT:    v_accvgpr_mov_b32 a1, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a2, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a3, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a4, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a5, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a6, a0
+; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
+; GFX942-NEXT:    s_nop 1
+; GFX942-NEXT:    v_mfma_f64_16x16x4_f64 a[0:7], v[2:3], v[0:1], a[0:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX90A-VGPR-LABEL: test_mfma_f64_16x16x4f64_imm:
+; GFX90A-VGPR:       ; %bb.0: ; %bb
+; GFX90A-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v7, 0x3ff00000
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v2, v0
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v12, s2
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v13, s3
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v3, v0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v4, v0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v5, v0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v6, v0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v1, v0
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[8:9], v[6:7], v[6:7] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[6:7], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[4:5], v[2:3], v[2:3] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    v_mfma_f64_16x16x4f64 v[2:9], v[12:13], v[10:11], v[2:9]
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v0, v[6:9], s[0:1] offset:16
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1]
+; GFX90A-VGPR-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f64_16x16x4f64_imm:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v7, 0x3ff00000
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v12, s2
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v13, s3
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], v[6:7]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[6:7], v[4:5]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[4:5], v[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f64_16x16x4_f64 v[2:9], v[12:13], v[10:11], v[2:9]
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v0, v[6:9], s[0:1] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> <double 0.0, double 0.0, double 0.0, double 1.0>, i32 0, i32 0, i32 0)
   store <4 x double> %mai.1, ptr addrspace(1) %arg
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f64_16x16x4f64_splat_lit:
-; GCN-DAG: v_accvgpr_write_b32 a{{[0-9]+}}, 0{{$}}
-; GCN-DAG: v_mov_b32_e32 v{{[0-9]+}}, 0x405ec000
-; GFX90A:  v_mfma_f64_16x16x4f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a[{{[0-9]+:[0-9]+}}]{{$}}
-; GFX942:  v_mfma_f64_16x16x4_f64 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a[{{[0-9]+:[0-9]+}}]{{$}}
-; GCN:     global_store_dwordx4
-; GCN:     global_store_dwordx4
 define amdgpu_kernel void @test_mfma_f64_16x16x4f64_splat_lit(ptr addrspace(1) %arg, double %a, double %b) #0 {
+; GFX90A-LABEL: test_mfma_f64_16x16x4f64_splat_lit:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0x405ec000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, 0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_mov_b32_e32 v2, s2
+; GFX90A-NEXT:    v_mov_b32_e32 v3, s3
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a2, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a3, a1
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a4, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a5, a1
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a6, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a7, a1
+; GFX90A-NEXT:    v_pk_mov_b32 v[0:1], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f64_16x16x4f64 a[0:7], v[2:3], v[0:1], a[0:7]
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_f64_16x16x4f64_splat_lit:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0x405ec000
+; GFX942-NEXT:    v_accvgpr_write_b32 a0, 0
+; GFX942-NEXT:    v_accvgpr_write_b32 a1, v0
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    v_mov_b32_e32 v2, s2
+; GFX942-NEXT:    v_mov_b32_e32 v3, s3
+; GFX942-NEXT:    v_accvgpr_mov_b32 a2, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a3, a1
+; GFX942-NEXT:    v_accvgpr_mov_b32 a4, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a5, a1
+; GFX942-NEXT:    v_accvgpr_mov_b32 a6, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a7, a1
+; GFX942-NEXT:    v_mov_b64_e32 v[0:1], s[6:7]
+; GFX942-NEXT:    s_nop 1
+; GFX942-NEXT:    v_mfma_f64_16x16x4_f64 a[0:7], v[2:3], v[0:1], a[0:7]
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX90A-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_lit:
+; GFX90A-VGPR:       ; %bb.0: ; %bb
+; GFX90A-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX90A-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v1, 0x405ec000
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v2, v0
+; GFX90A-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v12, s2
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v13, s3
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v3, v1
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v4, v0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v5, v1
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v6, v0
+; GFX90A-VGPR-NEXT:    v_mov_b32_e32 v7, v1
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[8:9], v[6:7], v[6:7] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[10:11], s[6:7], s[6:7] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[6:7], v[4:5], v[4:5] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[4:5], v[2:3], v[2:3] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    v_pk_mov_b32 v[2:3], v[0:1], v[0:1] op_sel:[0,1]
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    v_mfma_f64_16x16x4f64 v[2:9], v[12:13], v[10:11], v[2:9]
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 7
+; GFX90A-VGPR-NEXT:    s_nop 1
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v0, v[6:9], s[0:1] offset:16
+; GFX90A-VGPR-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1]
+; GFX90A-VGPR-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f64_16x16x4f64_splat_lit:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, 0x405ec000
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v12, s2
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v13, s3
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v7, v1
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], v[6:7]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[10:11], s[6:7]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[6:7], v[4:5]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[4:5], v[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f64_16x16x4_f64 v[2:9], v[12:13], v[10:11], v[2:9]
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v0, v[6:9], s[0:1] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x double> @llvm.amdgcn.mfma.f64.16x16x4f64(double %a, double %b, <4 x double> <double 123.0, double 123.0, double 123.0, double 123.0>, i32 0, i32 0, i32 0)
   store <4 x double> %mai.1, ptr addrspace(1) %arg
@@ -308,3 +1855,6 @@ bb:
 }
 
 attributes #0 = { "amdgpu-flat-work-group-size"="1,256" }
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; GCN: {{.*}}
+; VGPR: {{.*}}
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx942.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx942.ll
index b792a12..7d85d34 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx942.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx942.ll
@@ -1,12 +1,13 @@
-; RUN: llc -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX942,VGPRCD %s
-; RUN: llc -mtriple=amdgcn -mcpu=gfx942 -global-isel < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GISEL,VGPRCD %s
-; RUN: llc -mtriple=amdgcn -mcpu=gfx942 -stress-regalloc=10 < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX942,AGPRCD %s
-; RUN: llc -mtriple=amdgcn -mcpu=gfx942 -stress-regalloc=10 -global-isel < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GISEL,AGPRCD %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck --check-prefixes=GFX942,GFX942-VGPRCD,GFX942-SDAG,GFX942-VGPRCD-SDAG %s
+; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck --check-prefixes=GFX942,GFX942-VGPRCD,GFX942-GISEL,GFX942-VGPRCD-GISEL %s
+; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx942 -stress-regalloc=10 < %s | FileCheck --check-prefixes=GFX942,GFX942-AGPRCD,GFX942-SDAG,GFX942-AGPRCD-SDAG %s
+; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx942 -stress-regalloc=10 < %s | FileCheck --check-prefixes=GFX942,GFX942-AGPRCD,GFX942-GISEL,GFX942-AGPRCD-GISEL %s
 
-; RUN: llc -mtriple=amdgcn -mcpu=gfx950 < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX942,VGPRCD %s
-; RUN: llc -mtriple=amdgcn -mcpu=gfx950 -global-isel < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GISEL,VGPRCD %s
-; RUN: llc -mtriple=amdgcn -mcpu=gfx950 -stress-regalloc=10 < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX942,AGPRCD %s
-; RUN: llc -mtriple=amdgcn -mcpu=gfx950 -stress-regalloc=10 -global-isel < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GISEL,AGPRCD %s
+; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx950 < %s | FileCheck --check-prefixes=GFX950,GFX950-VGPRCD,GFX950-SDAG,GFX950-VGPRCD-SDAG %s
+; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx950 < %s | FileCheck --check-prefixes=GFX950,GFX950-VGPRCD,GFX950-GISEL,GFX950-VGPRCD-GISEL %s
+; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx950 -stress-regalloc=10 < %s | FileCheck --check-prefixes=GFX950,GFX950-AGPRCD,GFX950-SDAG,GFX950-AGPRCD-SDAG %s
+; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx950 -stress-regalloc=10 < %s | FileCheck --check-prefixes=GFX950,GFX950-AGPRCD,GFX950-GISEL,GFX950-AGPRCD-GISEL %s
 
 declare <4 x i32> @llvm.amdgcn.mfma.i32.16x16x32.i8(i64, i64, <4 x i32>, i32, i32, i32)
 declare <16 x i32> @llvm.amdgcn.mfma.i32.32x32x16.i8(i64, i64, <16 x i32>, i32, i32, i32)
@@ -33,17 +34,132 @@ declare <16 x float> @llvm.amdgcn.smfmac.f32.32x32x32.bf8.fp8(<2 x i32>, <4 x i3
 declare <16 x float> @llvm.amdgcn.smfmac.f32.32x32x32.fp8.bf8(<2 x i32>, <4 x i32>, <16 x float>, i32, i32, i32)
 declare <16 x float> @llvm.amdgcn.smfmac.f32.32x32x32.fp8.fp8(<2 x i32>, <4 x i32>, <16 x float>, i32, i32, i32)
 
-; GCN-LABEL: {{^}}test_mfma_i32_16x16x32i8:
-; GFX942-DAG:  v_mov_b32_e32 v[[ONE:[0-9]+]], 1
-; GFX942-DAG:  v_mov_b32_e32 v[[TWO:[0-9]+]], 2
-; GFX942-DAG:  v_mov_b32_e32 v[[THREE:[0-9]+]], 3
-; GFX942-DAG:  v_mov_b32_e32 v[[FOUR:[0-9]+]], 4
-; GCN-COUNT-4: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GFX942:      v_mfma_i32_16x16x32_i8 a[{{[0-9]+:[0-9]+}}], v[[[TWO]]:[[ONE]]], v[[[FOUR]]:[[THREE]]], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GISEL:       v_mfma_i32_16x16x32_i8 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GCN-NOT:     v_accvgpr_read_b32
-; GCN:         global_store_dwordx4 v{{[0-9]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_i32_16x16x32i8(ptr addrspace(1) %arg) #0 {
+; GFX942-VGPRCD-SDAG-LABEL: test_mfma_i32_16x16x32i8:
+; GFX942-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 2
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 4
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, 3
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_mfma_i32_16x16x32_i8 a[0:3], v[2:3], v[4:5], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 6
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX942-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-GISEL-LABEL: test_mfma_i32_16x16x32i8:
+; GFX942-GISEL:       ; %bb.0: ; %bb
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-GISEL-NEXT:    s_nop 1
+; GFX942-GISEL-NEXT:    v_mfma_i32_16x16x32_i8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-GISEL-NEXT:    s_nop 5
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX942-GISEL-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-SDAG-LABEL: test_mfma_i32_16x16x32i8:
+; GFX942-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_mfma_i32_16x16x32_i8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 6
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v4, a[0:3], s[6:7]
+; GFX942-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-SDAG-LABEL: test_mfma_i32_16x16x32i8:
+; GFX950-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 2
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 4
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, 3
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_mfma_i32_16x16x32_i8 a[0:3], v[2:3], v[4:5], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX950-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-GISEL-LABEL: test_mfma_i32_16x16x32i8:
+; GFX950-GISEL:       ; %bb.0: ; %bb
+; GFX950-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-GISEL-NEXT:    s_nop 1
+; GFX950-GISEL-NEXT:    v_mfma_i32_16x16x32_i8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-GISEL-NEXT:    s_nop 6
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX950-GISEL-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-SDAG-LABEL: test_mfma_i32_16x16x32i8:
+; GFX950-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 0
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_mfma_i32_16x16x32_i8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v4, a[0:3], s[6:7]
+; GFX950-AGPRCD-SDAG-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x i32>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x i32> @llvm.amdgcn.mfma.i32.16x16x32.i8(i64 4294967298, i64 12884901892, <4 x i32> %in.1, i32 1, i32 2, i32 3)
@@ -51,17 +167,154 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_i32_32x32x16i8:
-; GFX942-DAG:   v_mov_b32_e32 v[[ONE:[0-9]+]], 1
-; GFX942-DAG:   v_mov_b32_e32 v[[TWO:[0-9]+]], 2
-; GFX942-DAG:   v_mov_b32_e32 v[[THREE:[0-9]+]], 3
-; GFX942-DAG:   v_mov_b32_e32 v[[FOUR:[0-9]+]], 4
-; GCN-COUNT-16: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GFX942:       v_mfma_i32_32x32x16_i8 a[{{[0-9]+:[0-9]+}}], v[[[TWO]]:[[ONE]]], v[[[FOUR]]:[[THREE]]], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GISEL:        v_mfma_i32_32x32x16_i8 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GCN-NOT:      v_accvgpr_read_b32
-; GCN-COUNT-4:  global_store_dwordx4 v{{[0-9]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_i32_32x32x16i8(ptr addrspace(1) %arg) #0 {
+; GFX942-SDAG-LABEL: test_mfma_i32_32x32x16i8:
+; GFX942-SDAG:       ; %bb.0: ; %bb
+; GFX942-SDAG-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX942-SDAG-NEXT:    s_nop 1
+; GFX942-SDAG-NEXT:    v_mfma_i32_32x32x16_i8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-SDAG-NEXT:    s_nop 7
+; GFX942-SDAG-NEXT:    s_nop 1
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX942-SDAG-NEXT:    s_endpgm
+;
+; GFX942-GISEL-LABEL: test_mfma_i32_32x32x16i8:
+; GFX942-GISEL:       ; %bb.0: ; %bb
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX942-GISEL-NEXT:    s_nop 1
+; GFX942-GISEL-NEXT:    v_mfma_i32_32x32x16_i8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-GISEL-NEXT:    s_nop 7
+; GFX942-GISEL-NEXT:    s_nop 1
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX942-GISEL-NEXT:    s_endpgm
+;
+; GFX950-SDAG-LABEL: test_mfma_i32_32x32x16i8:
+; GFX950-SDAG:       ; %bb.0: ; %bb
+; GFX950-SDAG-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX950-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX950-SDAG-NEXT:    s_nop 1
+; GFX950-SDAG-NEXT:    v_mfma_i32_32x32x16_i8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-SDAG-NEXT:    s_nop 7
+; GFX950-SDAG-NEXT:    s_nop 2
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX950-SDAG-NEXT:    s_endpgm
+;
+; GFX950-GISEL-LABEL: test_mfma_i32_32x32x16i8:
+; GFX950-GISEL:       ; %bb.0: ; %bb
+; GFX950-GISEL-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX950-GISEL-NEXT:    s_nop 1
+; GFX950-GISEL-NEXT:    v_mfma_i32_32x32x16_i8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-GISEL-NEXT:    s_nop 7
+; GFX950-GISEL-NEXT:    s_nop 2
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX950-GISEL-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x i32>, ptr addrspace(1) %arg
   %mai.1 = tail call <16 x i32> @llvm.amdgcn.mfma.i32.32x32x16.i8(i64 4294967298, i64 12884901892, <16 x i32> %in.1, i32 1, i32 2, i32 3)
@@ -69,17 +322,132 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_16x16x32_bf8_bf8:
-; GFX942-DAG:  v_mov_b32_e32 v[[ONE:[0-9]+]], 1
-; GFX942-DAG:  v_mov_b32_e32 v[[TWO:[0-9]+]], 2
-; GFX942-DAG:  v_mov_b32_e32 v[[THREE:[0-9]+]], 3
-; GFX942-DAG:  v_mov_b32_e32 v[[FOUR:[0-9]+]], 4
-; GCN-COUNT-4: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GFX942:      v_mfma_f32_16x16x32_bf8_bf8 a[{{[0-9]+:[0-9]+}}], v{{\[}}[[TWO]]:[[ONE]]], v{{\[}}[[FOUR]]:[[THREE]]], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GISEL:       v_mfma_f32_16x16x32_bf8_bf8 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GCN-NOT:     v_accvgpr_read_b32
-; GCN:         global_store_dwordx4 v{{[0-9]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_16x16x32_bf8_bf8(ptr addrspace(1) %arg) #0 {
+; GFX942-VGPRCD-SDAG-LABEL: test_mfma_f32_16x16x32_bf8_bf8:
+; GFX942-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 2
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 4
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, 3
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_mfma_f32_16x16x32_bf8_bf8 a[0:3], v[2:3], v[4:5], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 6
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX942-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-GISEL-LABEL: test_mfma_f32_16x16x32_bf8_bf8:
+; GFX942-GISEL:       ; %bb.0: ; %bb
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-GISEL-NEXT:    s_nop 1
+; GFX942-GISEL-NEXT:    v_mfma_f32_16x16x32_bf8_bf8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-GISEL-NEXT:    s_nop 5
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX942-GISEL-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-SDAG-LABEL: test_mfma_f32_16x16x32_bf8_bf8:
+; GFX942-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_mfma_f32_16x16x32_bf8_bf8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 6
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v4, a[0:3], s[6:7]
+; GFX942-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-SDAG-LABEL: test_mfma_f32_16x16x32_bf8_bf8:
+; GFX950-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 2
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 4
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, 3
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_mfma_f32_16x16x32_bf8_bf8 a[0:3], v[2:3], v[4:5], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX950-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-GISEL-LABEL: test_mfma_f32_16x16x32_bf8_bf8:
+; GFX950-GISEL:       ; %bb.0: ; %bb
+; GFX950-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-GISEL-NEXT:    s_nop 1
+; GFX950-GISEL-NEXT:    v_mfma_f32_16x16x32_bf8_bf8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-GISEL-NEXT:    s_nop 6
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX950-GISEL-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-SDAG-LABEL: test_mfma_f32_16x16x32_bf8_bf8:
+; GFX950-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 0
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_mfma_f32_16x16x32_bf8_bf8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v4, a[0:3], s[6:7]
+; GFX950-AGPRCD-SDAG-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x float> @llvm.amdgcn.mfma.f32.16x16x32.bf8.bf8(i64 4294967298, i64 12884901892, <4 x float> %in.1, i32 1, i32 2, i32 3)
@@ -87,17 +455,132 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_16x16x32_bf8_fp8:
-; GFX942-DAG:  v_mov_b32_e32 v[[ONE:[0-9]+]], 1
-; GFX942-DAG:  v_mov_b32_e32 v[[TWO:[0-9]+]], 2
-; GFX942-DAG:  v_mov_b32_e32 v[[THREE:[0-9]+]], 3
-; GFX942-DAG:  v_mov_b32_e32 v[[FOUR:[0-9]+]], 4
-; GCN-COUNT-4: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GFX942:      v_mfma_f32_16x16x32_bf8_fp8 a[{{[0-9]+:[0-9]+}}], v{{\[}}[[TWO]]:[[ONE]]], v{{\[}}[[FOUR]]:[[THREE]]], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GISEL:       v_mfma_f32_16x16x32_bf8_fp8 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GCN-NOT:     v_accvgpr_read_b32
-; GCN:         global_store_dwordx4 v{{[0-9]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_16x16x32_bf8_fp8(ptr addrspace(1) %arg) #0 {
+; GFX942-VGPRCD-SDAG-LABEL: test_mfma_f32_16x16x32_bf8_fp8:
+; GFX942-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 2
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 4
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, 3
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_mfma_f32_16x16x32_bf8_fp8 a[0:3], v[2:3], v[4:5], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 6
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX942-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-GISEL-LABEL: test_mfma_f32_16x16x32_bf8_fp8:
+; GFX942-GISEL:       ; %bb.0: ; %bb
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-GISEL-NEXT:    s_nop 1
+; GFX942-GISEL-NEXT:    v_mfma_f32_16x16x32_bf8_fp8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-GISEL-NEXT:    s_nop 5
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX942-GISEL-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-SDAG-LABEL: test_mfma_f32_16x16x32_bf8_fp8:
+; GFX942-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_mfma_f32_16x16x32_bf8_fp8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 6
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v4, a[0:3], s[6:7]
+; GFX942-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-SDAG-LABEL: test_mfma_f32_16x16x32_bf8_fp8:
+; GFX950-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 2
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 4
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, 3
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_mfma_f32_16x16x32_bf8_fp8 a[0:3], v[2:3], v[4:5], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX950-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-GISEL-LABEL: test_mfma_f32_16x16x32_bf8_fp8:
+; GFX950-GISEL:       ; %bb.0: ; %bb
+; GFX950-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-GISEL-NEXT:    s_nop 1
+; GFX950-GISEL-NEXT:    v_mfma_f32_16x16x32_bf8_fp8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-GISEL-NEXT:    s_nop 6
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX950-GISEL-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-SDAG-LABEL: test_mfma_f32_16x16x32_bf8_fp8:
+; GFX950-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 0
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_mfma_f32_16x16x32_bf8_fp8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v4, a[0:3], s[6:7]
+; GFX950-AGPRCD-SDAG-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x float> @llvm.amdgcn.mfma.f32.16x16x32.bf8.fp8(i64 4294967298, i64 12884901892, <4 x float> %in.1, i32 1, i32 2, i32 3)
@@ -105,17 +588,132 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_16x16x32_fp8_bf8:
-; GFX942-DAG:  v_mov_b32_e32 v[[ONE:[0-9]+]], 1
-; GFX942-DAG:  v_mov_b32_e32 v[[TWO:[0-9]+]], 2
-; GFX942-DAG:  v_mov_b32_e32 v[[THREE:[0-9]+]], 3
-; GFX942-DAG:  v_mov_b32_e32 v[[FOUR:[0-9]+]], 4
-; GCN-COUNT-4: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GFX942:      v_mfma_f32_16x16x32_fp8_bf8 a[{{[0-9]+:[0-9]+}}], v{{\[}}[[TWO]]:[[ONE]]], v{{\[}}[[FOUR]]:[[THREE]]], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GISEL:       v_mfma_f32_16x16x32_fp8_bf8 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GCN-NOT:     v_accvgpr_read_b32
-; GCN:         global_store_dwordx4 v{{[0-9]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_16x16x32_fp8_bf8(ptr addrspace(1) %arg) #0 {
+; GFX942-VGPRCD-SDAG-LABEL: test_mfma_f32_16x16x32_fp8_bf8:
+; GFX942-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 2
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 4
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, 3
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_mfma_f32_16x16x32_fp8_bf8 a[0:3], v[2:3], v[4:5], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 6
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX942-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-GISEL-LABEL: test_mfma_f32_16x16x32_fp8_bf8:
+; GFX942-GISEL:       ; %bb.0: ; %bb
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-GISEL-NEXT:    s_nop 1
+; GFX942-GISEL-NEXT:    v_mfma_f32_16x16x32_fp8_bf8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-GISEL-NEXT:    s_nop 5
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX942-GISEL-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-SDAG-LABEL: test_mfma_f32_16x16x32_fp8_bf8:
+; GFX942-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_mfma_f32_16x16x32_fp8_bf8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 6
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v4, a[0:3], s[6:7]
+; GFX942-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-SDAG-LABEL: test_mfma_f32_16x16x32_fp8_bf8:
+; GFX950-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 2
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 4
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, 3
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_mfma_f32_16x16x32_fp8_bf8 a[0:3], v[2:3], v[4:5], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX950-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-GISEL-LABEL: test_mfma_f32_16x16x32_fp8_bf8:
+; GFX950-GISEL:       ; %bb.0: ; %bb
+; GFX950-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-GISEL-NEXT:    s_nop 1
+; GFX950-GISEL-NEXT:    v_mfma_f32_16x16x32_fp8_bf8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-GISEL-NEXT:    s_nop 6
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX950-GISEL-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-SDAG-LABEL: test_mfma_f32_16x16x32_fp8_bf8:
+; GFX950-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 0
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_mfma_f32_16x16x32_fp8_bf8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v4, a[0:3], s[6:7]
+; GFX950-AGPRCD-SDAG-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x float> @llvm.amdgcn.mfma.f32.16x16x32.fp8.bf8(i64 4294967298, i64 12884901892, <4 x float> %in.1, i32 1, i32 2, i32 3)
@@ -123,17 +721,132 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_16x16x32_fp8_fp8:
-; GFX942-DAG:  v_mov_b32_e32 v[[ONE:[0-9]+]], 1
-; GFX942-DAG:  v_mov_b32_e32 v[[TWO:[0-9]+]], 2
-; GFX942-DAG:  v_mov_b32_e32 v[[THREE:[0-9]+]], 3
-; GFX942-DAG:  v_mov_b32_e32 v[[FOUR:[0-9]+]], 4
-; GCN-COUNT-4: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GFX942:      v_mfma_f32_16x16x32_fp8_fp8 a[{{[0-9]+:[0-9]+}}], v{{\[}}[[TWO]]:[[ONE]]], v{{\[}}[[FOUR]]:[[THREE]]], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GISEL:       v_mfma_f32_16x16x32_fp8_fp8 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GCN-NOT:     v_accvgpr_read_b32
-; GCN:         global_store_dwordx4 v{{[0-9]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_16x16x32_fp8_fp8(ptr addrspace(1) %arg) #0 {
+; GFX942-VGPRCD-SDAG-LABEL: test_mfma_f32_16x16x32_fp8_fp8:
+; GFX942-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 2
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 4
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, 3
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_mfma_f32_16x16x32_fp8_fp8 a[0:3], v[2:3], v[4:5], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 6
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX942-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-GISEL-LABEL: test_mfma_f32_16x16x32_fp8_fp8:
+; GFX942-GISEL:       ; %bb.0: ; %bb
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-GISEL-NEXT:    s_nop 1
+; GFX942-GISEL-NEXT:    v_mfma_f32_16x16x32_fp8_fp8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-GISEL-NEXT:    s_nop 5
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX942-GISEL-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-SDAG-LABEL: test_mfma_f32_16x16x32_fp8_fp8:
+; GFX942-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_mfma_f32_16x16x32_fp8_fp8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 6
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v4, a[0:3], s[6:7]
+; GFX942-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-SDAG-LABEL: test_mfma_f32_16x16x32_fp8_fp8:
+; GFX950-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 2
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 4
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, 3
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-VGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_mfma_f32_16x16x32_fp8_fp8 a[0:3], v[2:3], v[4:5], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX950-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-GISEL-LABEL: test_mfma_f32_16x16x32_fp8_fp8:
+; GFX950-GISEL:       ; %bb.0: ; %bb
+; GFX950-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-GISEL-NEXT:    s_nop 1
+; GFX950-GISEL-NEXT:    v_mfma_f32_16x16x32_fp8_fp8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-GISEL-NEXT:    s_nop 6
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX950-GISEL-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-SDAG-LABEL: test_mfma_f32_16x16x32_fp8_fp8:
+; GFX950-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, 0
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_mfma_f32_16x16x32_fp8_fp8 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v4, a[0:3], s[6:7]
+; GFX950-AGPRCD-SDAG-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x float> @llvm.amdgcn.mfma.f32.16x16x32.fp8.fp8(i64 4294967298, i64 12884901892, <4 x float> %in.1, i32 1, i32 2, i32 3)
@@ -141,17 +854,154 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_32x32x16_bf8_bf8:
-; GFX942-DAG:  v_mov_b32_e32 v[[ONE:[0-9]+]], 1
-; GFX942-DAG:  v_mov_b32_e32 v[[TWO:[0-9]+]], 2
-; GFX942-DAG:  v_mov_b32_e32 v[[THREE:[0-9]+]], 3
-; GFX942-DAG:  v_mov_b32_e32 v[[FOUR:[0-9]+]], 4
-; GCN-COUNT-4: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GFX942:      v_mfma_f32_32x32x16_bf8_bf8 a[{{[0-9]+:[0-9]+}}], v{{\[}}[[TWO]]:[[ONE]]], v{{\[}}[[FOUR]]:[[THREE]]], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GISEL:       v_mfma_f32_32x32x16_bf8_bf8 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GCN-NOT:     v_accvgpr_read_b32
-; GCN:         global_store_dwordx4 v{{[0-9]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_32x32x16_bf8_bf8(ptr addrspace(1) %arg) #0 {
+; GFX942-SDAG-LABEL: test_mfma_f32_32x32x16_bf8_bf8:
+; GFX942-SDAG:       ; %bb.0: ; %bb
+; GFX942-SDAG-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX942-SDAG-NEXT:    s_nop 1
+; GFX942-SDAG-NEXT:    v_mfma_f32_32x32x16_bf8_bf8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-SDAG-NEXT:    s_nop 7
+; GFX942-SDAG-NEXT:    s_nop 1
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX942-SDAG-NEXT:    s_endpgm
+;
+; GFX942-GISEL-LABEL: test_mfma_f32_32x32x16_bf8_bf8:
+; GFX942-GISEL:       ; %bb.0: ; %bb
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX942-GISEL-NEXT:    s_nop 1
+; GFX942-GISEL-NEXT:    v_mfma_f32_32x32x16_bf8_bf8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-GISEL-NEXT:    s_nop 7
+; GFX942-GISEL-NEXT:    s_nop 1
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX942-GISEL-NEXT:    s_endpgm
+;
+; GFX950-SDAG-LABEL: test_mfma_f32_32x32x16_bf8_bf8:
+; GFX950-SDAG:       ; %bb.0: ; %bb
+; GFX950-SDAG-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX950-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX950-SDAG-NEXT:    s_nop 1
+; GFX950-SDAG-NEXT:    v_mfma_f32_32x32x16_bf8_bf8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-SDAG-NEXT:    s_nop 7
+; GFX950-SDAG-NEXT:    s_nop 2
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX950-SDAG-NEXT:    s_endpgm
+;
+; GFX950-GISEL-LABEL: test_mfma_f32_32x32x16_bf8_bf8:
+; GFX950-GISEL:       ; %bb.0: ; %bb
+; GFX950-GISEL-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX950-GISEL-NEXT:    s_nop 1
+; GFX950-GISEL-NEXT:    v_mfma_f32_32x32x16_bf8_bf8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-GISEL-NEXT:    s_nop 7
+; GFX950-GISEL-NEXT:    s_nop 2
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX950-GISEL-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.bf8.bf8(i64 4294967298, i64 12884901892, <16 x float> %in.1, i32 1, i32 2, i32 3)
@@ -159,17 +1009,154 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_32x32x16_bf8_fp8:
-; GFX942-DAG:  v_mov_b32_e32 v[[ONE:[0-9]+]], 1
-; GFX942-DAG:  v_mov_b32_e32 v[[TWO:[0-9]+]], 2
-; GFX942-DAG:  v_mov_b32_e32 v[[THREE:[0-9]+]], 3
-; GFX942-DAG:  v_mov_b32_e32 v[[FOUR:[0-9]+]], 4
-; GCN-COUNT-4: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GFX942:      v_mfma_f32_32x32x16_bf8_fp8 a[{{[0-9]+:[0-9]+}}], v{{\[}}[[TWO]]:[[ONE]]], v{{\[}}[[FOUR]]:[[THREE]]], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GISEL:       v_mfma_f32_32x32x16_bf8_fp8 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GCN-NOT:     v_accvgpr_read_b32
-; GCN:         global_store_dwordx4 v{{[0-9]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_32x32x16_bf8_fp8(ptr addrspace(1) %arg) #0 {
+; GFX942-SDAG-LABEL: test_mfma_f32_32x32x16_bf8_fp8:
+; GFX942-SDAG:       ; %bb.0: ; %bb
+; GFX942-SDAG-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX942-SDAG-NEXT:    s_nop 1
+; GFX942-SDAG-NEXT:    v_mfma_f32_32x32x16_bf8_fp8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-SDAG-NEXT:    s_nop 7
+; GFX942-SDAG-NEXT:    s_nop 1
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX942-SDAG-NEXT:    s_endpgm
+;
+; GFX942-GISEL-LABEL: test_mfma_f32_32x32x16_bf8_fp8:
+; GFX942-GISEL:       ; %bb.0: ; %bb
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX942-GISEL-NEXT:    s_nop 1
+; GFX942-GISEL-NEXT:    v_mfma_f32_32x32x16_bf8_fp8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-GISEL-NEXT:    s_nop 7
+; GFX942-GISEL-NEXT:    s_nop 1
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX942-GISEL-NEXT:    s_endpgm
+;
+; GFX950-SDAG-LABEL: test_mfma_f32_32x32x16_bf8_fp8:
+; GFX950-SDAG:       ; %bb.0: ; %bb
+; GFX950-SDAG-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX950-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX950-SDAG-NEXT:    s_nop 1
+; GFX950-SDAG-NEXT:    v_mfma_f32_32x32x16_bf8_fp8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-SDAG-NEXT:    s_nop 7
+; GFX950-SDAG-NEXT:    s_nop 2
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX950-SDAG-NEXT:    s_endpgm
+;
+; GFX950-GISEL-LABEL: test_mfma_f32_32x32x16_bf8_fp8:
+; GFX950-GISEL:       ; %bb.0: ; %bb
+; GFX950-GISEL-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX950-GISEL-NEXT:    s_nop 1
+; GFX950-GISEL-NEXT:    v_mfma_f32_32x32x16_bf8_fp8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-GISEL-NEXT:    s_nop 7
+; GFX950-GISEL-NEXT:    s_nop 2
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX950-GISEL-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.bf8.fp8(i64 4294967298, i64 12884901892, <16 x float> %in.1, i32 1, i32 2, i32 3)
@@ -177,17 +1164,154 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_32x32x16_fp8_bf8:
-; GFX942-DAG:  v_mov_b32_e32 v[[ONE:[0-9]+]], 1
-; GFX942-DAG:  v_mov_b32_e32 v[[TWO:[0-9]+]], 2
-; GFX942-DAG:  v_mov_b32_e32 v[[THREE:[0-9]+]], 3
-; GFX942-DAG:  v_mov_b32_e32 v[[FOUR:[0-9]+]], 4
-; GCN-COUNT-4: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GFX942:      v_mfma_f32_32x32x16_fp8_bf8 a[{{[0-9]+:[0-9]+}}], v{{\[}}[[TWO]]:[[ONE]]], v{{\[}}[[FOUR]]:[[THREE]]], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GISEL:       v_mfma_f32_32x32x16_fp8_bf8 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GCN-NOT:     v_accvgpr_read_b32
-; GCN:         global_store_dwordx4 v{{[0-9]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_32x32x16_fp8_bf8(ptr addrspace(1) %arg) #0 {
+; GFX942-SDAG-LABEL: test_mfma_f32_32x32x16_fp8_bf8:
+; GFX942-SDAG:       ; %bb.0: ; %bb
+; GFX942-SDAG-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX942-SDAG-NEXT:    s_nop 1
+; GFX942-SDAG-NEXT:    v_mfma_f32_32x32x16_fp8_bf8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-SDAG-NEXT:    s_nop 7
+; GFX942-SDAG-NEXT:    s_nop 1
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX942-SDAG-NEXT:    s_endpgm
+;
+; GFX942-GISEL-LABEL: test_mfma_f32_32x32x16_fp8_bf8:
+; GFX942-GISEL:       ; %bb.0: ; %bb
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX942-GISEL-NEXT:    s_nop 1
+; GFX942-GISEL-NEXT:    v_mfma_f32_32x32x16_fp8_bf8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-GISEL-NEXT:    s_nop 7
+; GFX942-GISEL-NEXT:    s_nop 1
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX942-GISEL-NEXT:    s_endpgm
+;
+; GFX950-SDAG-LABEL: test_mfma_f32_32x32x16_fp8_bf8:
+; GFX950-SDAG:       ; %bb.0: ; %bb
+; GFX950-SDAG-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX950-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX950-SDAG-NEXT:    s_nop 1
+; GFX950-SDAG-NEXT:    v_mfma_f32_32x32x16_fp8_bf8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-SDAG-NEXT:    s_nop 7
+; GFX950-SDAG-NEXT:    s_nop 2
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX950-SDAG-NEXT:    s_endpgm
+;
+; GFX950-GISEL-LABEL: test_mfma_f32_32x32x16_fp8_bf8:
+; GFX950-GISEL:       ; %bb.0: ; %bb
+; GFX950-GISEL-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX950-GISEL-NEXT:    s_nop 1
+; GFX950-GISEL-NEXT:    v_mfma_f32_32x32x16_fp8_bf8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-GISEL-NEXT:    s_nop 7
+; GFX950-GISEL-NEXT:    s_nop 2
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX950-GISEL-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.fp8.bf8(i64 4294967298, i64 12884901892, <16 x float> %in.1, i32 1, i32 2, i32 3)
@@ -195,17 +1319,154 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_32x32x16_fp8_fp8:
-; GFX942-DAG:  v_mov_b32_e32 v[[ONE:[0-9]+]], 1
-; GFX942-DAG:  v_mov_b32_e32 v[[TWO:[0-9]+]], 2
-; GFX942-DAG:  v_mov_b32_e32 v[[THREE:[0-9]+]], 3
-; GFX942-DAG:  v_mov_b32_e32 v[[FOUR:[0-9]+]], 4
-; GCN-COUNT-4: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GFX942:      v_mfma_f32_32x32x16_fp8_fp8 a[{{[0-9]+:[0-9]+}}], v{{\[}}[[TWO]]:[[ONE]]], v{{\[}}[[FOUR]]:[[THREE]]], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GISEL:       v_mfma_f32_32x32x16_fp8_fp8 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GCN-NOT:     v_accvgpr_read_b32
-; GCN:         global_store_dwordx4 v{{[0-9]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_32x32x16_fp8_fp8(ptr addrspace(1) %arg) #0 {
+; GFX942-SDAG-LABEL: test_mfma_f32_32x32x16_fp8_fp8:
+; GFX942-SDAG:       ; %bb.0: ; %bb
+; GFX942-SDAG-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX942-SDAG-NEXT:    s_nop 1
+; GFX942-SDAG-NEXT:    v_mfma_f32_32x32x16_fp8_fp8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-SDAG-NEXT:    s_nop 7
+; GFX942-SDAG-NEXT:    s_nop 1
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX942-SDAG-NEXT:    s_endpgm
+;
+; GFX942-GISEL-LABEL: test_mfma_f32_32x32x16_fp8_fp8:
+; GFX942-GISEL:       ; %bb.0: ; %bb
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX942-GISEL-NEXT:    s_nop 1
+; GFX942-GISEL-NEXT:    v_mfma_f32_32x32x16_fp8_fp8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-GISEL-NEXT:    s_nop 7
+; GFX942-GISEL-NEXT:    s_nop 1
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX942-GISEL-NEXT:    s_endpgm
+;
+; GFX950-SDAG-LABEL: test_mfma_f32_32x32x16_fp8_fp8:
+; GFX950-SDAG:       ; %bb.0: ; %bb
+; GFX950-SDAG-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX950-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX950-SDAG-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX950-SDAG-NEXT:    s_nop 1
+; GFX950-SDAG-NEXT:    v_mfma_f32_32x32x16_fp8_fp8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX950-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-SDAG-NEXT:    s_nop 7
+; GFX950-SDAG-NEXT:    s_nop 2
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX950-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX950-SDAG-NEXT:    s_endpgm
+;
+; GFX950-GISEL-LABEL: test_mfma_f32_32x32x16_fp8_fp8:
+; GFX950-GISEL:       ; %bb.0: ; %bb
+; GFX950-GISEL-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 2
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v1, 1
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v2, 4
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v3, 3
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX950-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX950-GISEL-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX950-GISEL-NEXT:    s_nop 1
+; GFX950-GISEL-NEXT:    v_mfma_f32_32x32x16_fp8_fp8 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX950-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-GISEL-NEXT:    s_nop 7
+; GFX950-GISEL-NEXT:    s_nop 2
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX950-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX950-GISEL-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.fp8.fp8(i64 4294967298, i64 12884901892, <16 x float> %in.1, i32 1, i32 2, i32 3)
@@ -213,15 +1474,132 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_smfmac_f32_16x16x32_f16:
-; GCN:        s_load_dwordx4 s[[[SLO:[0-9]+]]:[[SHI:[0-9]+]]], s[{{[0-9:]+}}], 0x0{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 [[CD:v]][[[RLO:[0-9]+]]:{{[0-9]+}}], s[[[SLO]]:{{[0-9]+}}]{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 v[{{[0-9]+}}:[[RHI:[0-9]+]]], s[{{[0-9]+}}:[[SHI]]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 [[CD:a]][[RLO:[0-9]+]], s[[SLO]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 a[[RHI:[0-9]+]], s[[SHI]]{{$}}
-; GCN:        v_smfmac_f32_16x16x32_f16 [[CD]][[[RLO]]:[[RHI]]], {{[av]}}[{{[0-9:]+}}], {{[av]}}[{{[0-9:]+}}], v{{[0-9]+}} cbsz:1 abid:2
-; GCN:        global_store_dwordx4 v{{[0-9]+}}, [[CD]][[[RLO]]:[[RHI]]]
 define amdgpu_kernel void @test_smfmac_f32_16x16x32_f16(ptr addrspace(1) %arg, <4 x half> %a, <8 x half> %b, i32 %idx) #0 {
+; GFX942-VGPRCD-SDAG-LABEL: test_smfmac_f32_16x16x32_f16:
+; GFX942-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x24
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dword s6, s[4:5], 0x44
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, 0
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[10:11]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[12:13]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[14:15]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v7, s6
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[10:11], s[2:3]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[0:1]
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x32_f16 v[8:11], v[4:5], v[0:3], v7 cbsz:1 abid:2
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 6
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v6, v[8:11], s[8:9]
+; GFX942-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-VGPRCD-GISEL-LABEL: test_smfmac_f32_16x16x32_f16:
+; GFX942-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x24
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dword s6, s[4:5], 0x44
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[10:11]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[12:13]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[14:15]
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[2:3]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[0:1]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s6
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x32_f16 v[8:11], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 5
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[8:9]
+; GFX942-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-LABEL: test_smfmac_f32_16x16x32_f16:
+; GFX942-AGPRCD:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x24
+; GFX942-AGPRCD-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-NEXT:    v_mov_b64_e32 v[4:5], s[10:11]
+; GFX942-AGPRCD-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
+; GFX942-AGPRCD-NEXT:    v_mov_b64_e32 v[0:1], s[12:13]
+; GFX942-AGPRCD-NEXT:    v_mov_b64_e32 v[2:3], s[14:15]
+; GFX942-AGPRCD-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-AGPRCD-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-AGPRCD-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-AGPRCD-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-AGPRCD-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX942-AGPRCD-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-NEXT:    v_mov_b32_e32 v6, s0
+; GFX942-AGPRCD-NEXT:    s_nop 1
+; GFX942-AGPRCD-NEXT:    v_smfmac_f32_16x16x32_f16 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-NEXT:    s_nop 5
+; GFX942-AGPRCD-NEXT:    global_store_dwordx4 v0, a[0:3], s[8:9]
+; GFX942-AGPRCD-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-SDAG-LABEL: test_smfmac_f32_16x16x32_f16:
+; GFX950-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x24
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dword s6, s[4:5], 0x44
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, 0
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[10:11]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[12:13]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[14:15]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v7, s6
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[10:11], s[2:3]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[0:1]
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x32_f16 v[8:11], v[4:5], v[0:3], v7 cbsz:1 abid:2
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v6, v[8:11], s[8:9]
+; GFX950-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-GISEL-LABEL: test_smfmac_f32_16x16x32_f16:
+; GFX950-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x24
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dword s6, s[4:5], 0x44
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[10:11]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[12:13]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[14:15]
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[2:3]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[0:1]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s6
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-VGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x32_f16 v[8:11], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 6
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[8:9]
+; GFX950-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-LABEL: test_smfmac_f32_16x16x32_f16:
+; GFX950-AGPRCD:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x24
+; GFX950-AGPRCD-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-NEXT:    v_mov_b64_e32 v[4:5], s[10:11]
+; GFX950-AGPRCD-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
+; GFX950-AGPRCD-NEXT:    v_mov_b64_e32 v[0:1], s[12:13]
+; GFX950-AGPRCD-NEXT:    v_mov_b64_e32 v[2:3], s[14:15]
+; GFX950-AGPRCD-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-AGPRCD-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-AGPRCD-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-AGPRCD-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-AGPRCD-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX950-AGPRCD-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-NEXT:    v_mov_b32_e32 v6, s0
+; GFX950-AGPRCD-NEXT:    s_nop 1
+; GFX950-AGPRCD-NEXT:    v_smfmac_f32_16x16x32_f16 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-NEXT:    s_nop 6
+; GFX950-AGPRCD-NEXT:    global_store_dwordx4 v0, a[0:3], s[8:9]
+; GFX950-AGPRCD-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x32.f16(<4 x half> %a, <8 x half> %b, <4 x float> %in.1, i32 %idx, i32 1, i32 2)
@@ -229,18 +1607,278 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_smfmac_f32_32x32x16_f16:
-; GCN:        s_load_dwordx16 s[[[SLO:[0-9]+]]:[[SHI:[0-9]+]]], s[{{[0-9:]+}}], 0x0{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 [[CD:v]][[[RLO:[0-9]+]]:{{[0-9]+}}], s[[[SLO]]:{{[0-9]+}}]{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 v[{{[0-9]+}}:[[RHI:[0-9]+]]], s[{{[0-9]+}}:[[SHI]]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 [[CD:a]][[RLO:[0-9]+]], s[[SLO]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 a[[RHI:[0-9]+]], s[[SHI]]{{$}}
-; GCN:        v_smfmac_f32_32x32x16_f16 [[CD]][[[RLO]]:[[RHI]]], {{[av]}}[{{[0-9:]+}}], {{[av]}}[{{[0-9:]+}}], v{{[0-9]+}} cbsz:1 abid:2
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][[[RLO]]:{{[0-9]+}}], s[{{[0-9:]+}}]{{$}}
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9:]+}}], s[{{[0-9:]+}}] offset:16
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9:]+}}], s[{{[0-9:]+}}] offset:32
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9]+}}:[[RHI]]], s[{{[0-9:]+}}] offset:48
 define amdgpu_kernel void @test_smfmac_f32_32x32x16_f16(ptr addrspace(1) %arg, <4 x half> %a, <8 x half> %b, i32 %idx) #0 {
+; GFX942-VGPRCD-SDAG-LABEL: test_smfmac_f32_32x32x16_f16:
+; GFX942-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[16:23], s[4:5], 0x24
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dword s24, s[4:5], 0x44
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[20:21], s[18:19]
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[16:17], s[20:21]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[18:19], s[22:23]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v22, s24
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x16_f16 v[0:15], v[20:21], v[16:19], v22 cbsz:1 abid:2
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[12:15], s[16:17] offset:48
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[16:17] offset:32
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[4:7], s[16:17] offset:16
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[0:3], s[16:17]
+; GFX942-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-VGPRCD-GISEL-LABEL: test_smfmac_f32_32x32x16_f16:
+; GFX942-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx8 s[16:23], s[4:5], 0x24
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dword s24, s[4:5], 0x44
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[18:19]
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[20:21]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[22:23]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v22, s24
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x16_f16 v[0:15], v[20:21], v[16:19], v22 cbsz:1 abid:2
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 7
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[0:3], s[16:17]
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[4:7], s[16:17] offset:16
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[8:11], s[16:17] offset:32
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[12:15], s[16:17] offset:48
+; GFX942-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-SDAG-LABEL: test_smfmac_f32_32x32x16_f16:
+; GFX942-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[26:27]
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx16 s[8:23], s[24:25], 0x0
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[28:29]
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s0
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[30:31]
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x16_f16 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 7
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[24:25] offset:48
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[24:25] offset:32
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[24:25] offset:16
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[24:25]
+; GFX942-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-GISEL-LABEL: test_smfmac_f32_32x32x16_f16:
+; GFX942-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[26:27]
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx16 s[8:23], s[24:25], 0x0
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[28:29]
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s0
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[30:31]
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x16_f16 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 7
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[24:25]
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[24:25] offset:16
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[24:25] offset:32
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[24:25] offset:48
+; GFX942-AGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-SDAG-LABEL: test_smfmac_f32_32x32x16_f16:
+; GFX950-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[16:23], s[4:5], 0x24
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dword s24, s[4:5], 0x44
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[20:21], s[18:19]
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[16:17], s[20:21]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[18:19], s[22:23]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v22, s24
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x16_f16 v[0:15], v[20:21], v[16:19], v22 cbsz:1 abid:2
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 2
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[12:15], s[16:17] offset:48
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[16:17] offset:32
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[4:7], s[16:17] offset:16
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[0:3], s[16:17]
+; GFX950-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-GISEL-LABEL: test_smfmac_f32_32x32x16_f16:
+; GFX950-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx8 s[16:23], s[4:5], 0x24
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dword s24, s[4:5], 0x44
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[18:19]
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[20:21]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[22:23]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v22, s24
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-VGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x16_f16 v[0:15], v[20:21], v[16:19], v22 cbsz:1 abid:2
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v16, 0
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 7
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 2
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[0:3], s[16:17]
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[4:7], s[16:17] offset:16
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[8:11], s[16:17] offset:32
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[12:15], s[16:17] offset:48
+; GFX950-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-SDAG-LABEL: test_smfmac_f32_32x32x16_f16:
+; GFX950-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[26:27]
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx16 s[8:23], s[24:25], 0x0
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[28:29]
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s0
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[30:31]
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x16_f16 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 2
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[24:25] offset:48
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[24:25] offset:32
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[24:25] offset:16
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[24:25]
+; GFX950-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-GISEL-LABEL: test_smfmac_f32_32x32x16_f16:
+; GFX950-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[26:27]
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx16 s[8:23], s[24:25], 0x0
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[28:29]
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s0
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[30:31]
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-AGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x16_f16 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 7
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 2
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[24:25]
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[24:25] offset:16
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[24:25] offset:32
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[24:25] offset:48
+; GFX950-AGPRCD-GISEL-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <16 x float> @llvm.amdgcn.smfmac.f32.32x32x16.f16(<4 x half> %a, <8 x half> %b, <16 x float> %in.1, i32 %idx, i32 1, i32 2)
@@ -248,15 +1886,132 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_smfmac_f32_16x16x32_bf16:
-; GCN:        s_load_dwordx4 s[[[SLO:[0-9]+]]:[[SHI:[0-9]+]]], s[{{[0-9:]+}}], 0x0{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 [[CD:v]][[[RLO:[0-9]+]]:{{[0-9]+}}], s[[[SLO]]:{{[0-9]+}}]{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 v[{{[0-9]+}}:[[RHI:[0-9]+]]], s[{{[0-9]+}}:[[SHI]]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 [[CD:a]][[RLO:[0-9]+]], s[[SLO]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 a[[RHI:[0-9]+]], s[[SHI]]{{$}}
-; GCN:        v_smfmac_f32_16x16x32_bf16 [[CD]][[[RLO]]:[[RHI]]], {{[av]}}[{{[0-9:]+}}], {{[av]}}[{{[0-9:]+}}], v{{[0-9]+}} cbsz:1 abid:2
-; GCN:        global_store_dwordx4 v{{[0-9]+}}, [[CD]][[[RLO]]:[[RHI]]]
 define amdgpu_kernel void @test_smfmac_f32_16x16x32_bf16(ptr addrspace(1) %arg, <4 x i16> %a, <8 x i16> %b, i32 %idx) #0 {
+; GFX942-VGPRCD-SDAG-LABEL: test_smfmac_f32_16x16x32_bf16:
+; GFX942-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x24
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dword s6, s[4:5], 0x44
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, 0
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[10:11]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[12:13]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[14:15]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v7, s6
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[10:11], s[2:3]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[0:1]
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x32_bf16 v[8:11], v[4:5], v[0:3], v7 cbsz:1 abid:2
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 6
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v6, v[8:11], s[8:9]
+; GFX942-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-VGPRCD-GISEL-LABEL: test_smfmac_f32_16x16x32_bf16:
+; GFX942-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x24
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dword s6, s[4:5], 0x44
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[10:11]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[12:13]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[14:15]
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[2:3]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[0:1]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s6
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x32_bf16 v[8:11], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 5
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[8:9]
+; GFX942-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-LABEL: test_smfmac_f32_16x16x32_bf16:
+; GFX942-AGPRCD:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x24
+; GFX942-AGPRCD-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-NEXT:    v_mov_b64_e32 v[4:5], s[10:11]
+; GFX942-AGPRCD-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
+; GFX942-AGPRCD-NEXT:    v_mov_b64_e32 v[0:1], s[12:13]
+; GFX942-AGPRCD-NEXT:    v_mov_b64_e32 v[2:3], s[14:15]
+; GFX942-AGPRCD-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-AGPRCD-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-AGPRCD-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-AGPRCD-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-AGPRCD-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX942-AGPRCD-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-NEXT:    v_mov_b32_e32 v6, s0
+; GFX942-AGPRCD-NEXT:    s_nop 1
+; GFX942-AGPRCD-NEXT:    v_smfmac_f32_16x16x32_bf16 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-NEXT:    s_nop 5
+; GFX942-AGPRCD-NEXT:    global_store_dwordx4 v0, a[0:3], s[8:9]
+; GFX942-AGPRCD-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-SDAG-LABEL: test_smfmac_f32_16x16x32_bf16:
+; GFX950-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x24
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dword s6, s[4:5], 0x44
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, 0
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[10:11]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[12:13]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[14:15]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v7, s6
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[10:11], s[2:3]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[0:1]
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x32_bf16 v[8:11], v[4:5], v[0:3], v7 cbsz:1 abid:2
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v6, v[8:11], s[8:9]
+; GFX950-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-GISEL-LABEL: test_smfmac_f32_16x16x32_bf16:
+; GFX950-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x24
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dword s6, s[4:5], 0x44
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[10:11]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[12:13]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[14:15]
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[2:3]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[0:1]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s6
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-VGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x32_bf16 v[8:11], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 6
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[8:9]
+; GFX950-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-LABEL: test_smfmac_f32_16x16x32_bf16:
+; GFX950-AGPRCD:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x24
+; GFX950-AGPRCD-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-NEXT:    v_mov_b64_e32 v[4:5], s[10:11]
+; GFX950-AGPRCD-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
+; GFX950-AGPRCD-NEXT:    v_mov_b64_e32 v[0:1], s[12:13]
+; GFX950-AGPRCD-NEXT:    v_mov_b64_e32 v[2:3], s[14:15]
+; GFX950-AGPRCD-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-AGPRCD-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-AGPRCD-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-AGPRCD-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-AGPRCD-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX950-AGPRCD-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-NEXT:    v_mov_b32_e32 v6, s0
+; GFX950-AGPRCD-NEXT:    s_nop 1
+; GFX950-AGPRCD-NEXT:    v_smfmac_f32_16x16x32_bf16 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-NEXT:    s_nop 6
+; GFX950-AGPRCD-NEXT:    global_store_dwordx4 v0, a[0:3], s[8:9]
+; GFX950-AGPRCD-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x32.bf16(<4 x i16> %a, <8 x i16> %b, <4 x float> %in.1, i32 %idx, i32 1, i32 2)
@@ -264,18 +2019,278 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_smfmac_f32_32x32x16_bf16:
-; GCN:        s_load_dwordx16 s[[[SLO:[0-9]+]]:[[SHI:[0-9]+]]], s[{{[0-9:]+}}], 0x0{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 [[CD:v]][[[RLO:[0-9]+]]:{{[0-9]+}}], s[[[SLO]]:{{[0-9]+}}]{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 v[{{[0-9]+}}:[[RHI:[0-9]+]]], s[{{[0-9]+}}:[[SHI]]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 [[CD:a]][[RLO:[0-9]+]], s[[SLO]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 a[[RHI:[0-9]+]], s[[SHI]]{{$}}
-; GCN:        v_smfmac_f32_32x32x16_bf16 [[CD]][[[RLO]]:[[RHI]]], {{[av]}}[{{[0-9:]+}}], {{[av]}}[{{[0-9:]+}}], v{{[0-9]+}} cbsz:1 abid:2
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][[[RLO]]:{{[0-9]+}}], s[{{[0-9:]+}}]{{$}}
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9:]+}}], s[{{[0-9:]+}}] offset:16
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9:]+}}], s[{{[0-9:]+}}] offset:32
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9]+}}:[[RHI]]], s[{{[0-9:]+}}] offset:48
 define amdgpu_kernel void @test_smfmac_f32_32x32x16_bf16(ptr addrspace(1) %arg, <4 x i16> %a, <8 x i16> %b, i32 %idx) #0 {
+; GFX942-VGPRCD-SDAG-LABEL: test_smfmac_f32_32x32x16_bf16:
+; GFX942-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[16:23], s[4:5], 0x24
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dword s24, s[4:5], 0x44
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[20:21], s[18:19]
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[16:17], s[20:21]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[18:19], s[22:23]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v22, s24
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x16_bf16 v[0:15], v[20:21], v[16:19], v22 cbsz:1 abid:2
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[12:15], s[16:17] offset:48
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[16:17] offset:32
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[4:7], s[16:17] offset:16
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[0:3], s[16:17]
+; GFX942-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-VGPRCD-GISEL-LABEL: test_smfmac_f32_32x32x16_bf16:
+; GFX942-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx8 s[16:23], s[4:5], 0x24
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dword s24, s[4:5], 0x44
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[18:19]
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[20:21]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[22:23]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v22, s24
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x16_bf16 v[0:15], v[20:21], v[16:19], v22 cbsz:1 abid:2
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 7
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[0:3], s[16:17]
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[4:7], s[16:17] offset:16
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[8:11], s[16:17] offset:32
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[12:15], s[16:17] offset:48
+; GFX942-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-SDAG-LABEL: test_smfmac_f32_32x32x16_bf16:
+; GFX942-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[26:27]
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx16 s[8:23], s[24:25], 0x0
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[28:29]
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s0
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[30:31]
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x16_bf16 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 7
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[24:25] offset:48
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[24:25] offset:32
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[24:25] offset:16
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[24:25]
+; GFX942-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-GISEL-LABEL: test_smfmac_f32_32x32x16_bf16:
+; GFX942-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[26:27]
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx16 s[8:23], s[24:25], 0x0
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[28:29]
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s0
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[30:31]
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x16_bf16 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 7
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[24:25]
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[24:25] offset:16
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[24:25] offset:32
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[24:25] offset:48
+; GFX942-AGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-SDAG-LABEL: test_smfmac_f32_32x32x16_bf16:
+; GFX950-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[16:23], s[4:5], 0x24
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dword s24, s[4:5], 0x44
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[20:21], s[18:19]
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[16:17], s[20:21]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[18:19], s[22:23]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v22, s24
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x16_bf16 v[0:15], v[20:21], v[16:19], v22 cbsz:1 abid:2
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 2
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[12:15], s[16:17] offset:48
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[16:17] offset:32
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[4:7], s[16:17] offset:16
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[0:3], s[16:17]
+; GFX950-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-GISEL-LABEL: test_smfmac_f32_32x32x16_bf16:
+; GFX950-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx8 s[16:23], s[4:5], 0x24
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dword s24, s[4:5], 0x44
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[18:19]
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[20:21]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[22:23]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v22, s24
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-VGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x16_bf16 v[0:15], v[20:21], v[16:19], v22 cbsz:1 abid:2
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v16, 0
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 7
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 2
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[0:3], s[16:17]
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[4:7], s[16:17] offset:16
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[8:11], s[16:17] offset:32
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[12:15], s[16:17] offset:48
+; GFX950-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-SDAG-LABEL: test_smfmac_f32_32x32x16_bf16:
+; GFX950-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[26:27]
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx16 s[8:23], s[24:25], 0x0
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[28:29]
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s0
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[30:31]
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x16_bf16 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 2
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[24:25] offset:48
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[24:25] offset:32
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[24:25] offset:16
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[24:25]
+; GFX950-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-GISEL-LABEL: test_smfmac_f32_32x32x16_bf16:
+; GFX950-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[26:27]
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx16 s[8:23], s[24:25], 0x0
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[28:29]
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s0
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[30:31]
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-AGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x16_bf16 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 7
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 2
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[24:25]
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[24:25] offset:16
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[24:25] offset:32
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[24:25] offset:48
+; GFX950-AGPRCD-GISEL-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <16 x float> @llvm.amdgcn.smfmac.f32.32x32x16.bf16(<4 x i16> %a, <8 x i16> %b, <16 x float> %in.1, i32 %idx, i32 1, i32 2)
@@ -283,15 +2298,214 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_smfmac_i32_16x16x64_i8:
-; GCN:        s_load_dwordx4 s[[[SLO:[0-9]+]]:[[SHI:[0-9]+]]], s[{{[0-9:]+}}], 0x0{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 [[CD:v]][[[RLO:[0-9]+]]:{{[0-9]+}}], s[[[SLO]]:{{[0-9]+}}]{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 v[{{[0-9]+}}:[[RHI:[0-9]+]]], s[{{[0-9]+}}:[[SHI]]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 [[CD:a]][[RLO:[0-9]+]], s[[SLO]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 a[[RHI:[0-9]+]], s[[SHI]]{{$}}
-; GCN:        v_smfmac_i32_16x16x64_i8 [[CD]][[[RLO]]:[[RHI]]], {{[av]}}[{{[0-9:]+}}], {{[av]}}[{{[0-9:]+}}], v{{[0-9]+}} cbsz:1 abid:2
-; GCN:        global_store_dwordx4 v{{[0-9]+}}, [[CD]][[[RLO]]:[[RHI]]]
 define amdgpu_kernel void @test_smfmac_i32_16x16x64_i8(ptr addrspace(1) %arg, <2 x i32> %a, <4 x i32> %b, i32 %idx) #0 {
+; GFX942-VGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_i8:
+; GFX942-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v10, s8
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v11, s9
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s10
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s11
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s12
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s13
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s14
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[0:1]
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_smfmac_i32_16x16x64_i8 v[6:9], v[10:11], v[2:5], v1 cbsz:1 abid:2
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 6
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, v[6:9], s[6:7]
+; GFX942-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-VGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_i8:
+; GFX942-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x24
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x3c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dword s14, s[4:5], 0x44
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[12:13], 0x0
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s4, s2
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s5, s3
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[4:5]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s14
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    v_smfmac_i32_16x16x64_i8 v[8:11], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 5
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[12:13]
+; GFX942-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_i8:
+; GFX942-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x0
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_smfmac_i32_16x16x64_i8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 5
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_i8:
+; GFX942-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-AGPRCD-GISEL-NEXT:    ; implicit-def: $vgpr7 : SGPR spill to VGPR lane
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s2
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s3
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX942-AGPRCD-GISEL-NEXT:    v_writelane_b32 v7, s0, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-AGPRCD-GISEL-NEXT:    v_readlane_b32 s0, v7, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    v_smfmac_i32_16x16x64_i8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 5
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX942-AGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_i8:
+; GFX950-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v10, s8
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v11, s9
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s10
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s11
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s12
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s13
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s14
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[0:1]
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_smfmac_i32_16x16x64_i8 v[6:9], v[10:11], v[2:5], v1 cbsz:1 abid:2
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, v[6:9], s[6:7]
+; GFX950-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_i8:
+; GFX950-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x24
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x3c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dword s14, s[4:5], 0x44
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[12:13], 0x0
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s4, s2
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s5, s3
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[4:5]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s14
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-VGPRCD-GISEL-NEXT:    v_smfmac_i32_16x16x64_i8 v[8:11], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 6
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[12:13]
+; GFX950-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_i8:
+; GFX950-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x0
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_smfmac_i32_16x16x64_i8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 6
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX950-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_i8:
+; GFX950-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-AGPRCD-GISEL-NEXT:    ; implicit-def: $vgpr7 : SGPR spill to VGPR lane
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s2
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s3
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX950-AGPRCD-GISEL-NEXT:    v_writelane_b32 v7, s0, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-AGPRCD-GISEL-NEXT:    v_readlane_b32 s0, v7, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-AGPRCD-GISEL-NEXT:    v_smfmac_i32_16x16x64_i8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 6
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX950-AGPRCD-GISEL-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x i32>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x i32> @llvm.amdgcn.smfmac.i32.16x16x64.i8(<2 x i32> %a, <4 x i32> %b, <4 x i32> %in.1, i32 %idx, i32 1, i32 2)
@@ -299,18 +2513,310 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_smfmac_i32_32x32x32_i8:
-; GCN:        s_load_dwordx16 s[[[SLO:[0-9]+]]:[[SHI:[0-9]+]]], s[{{[0-9:]+}}], 0x0{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 [[CD:v]][[[RLO:[0-9]+]]:{{[0-9]+}}], s[[[SLO]]:{{[0-9]+}}]{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 v[{{[0-9]+}}:[[RHI:[0-9]+]]], s[{{[0-9]+}}:[[SHI]]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 [[CD:a]][[RLO:[0-9]+]], s[[SLO]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 a[[RHI:[0-9]+]], s[[SHI]]{{$}}
-; GCN:        v_smfmac_i32_32x32x32_i8 [[CD]][[[RLO]]:[[RHI]]], {{[av]}}[{{[0-9:]+}}], {{[av]}}[{{[0-9:]+}}], v{{[0-9]+}} cbsz:1 abid:2
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][[[RLO]]:{{[0-9]+}}], s[{{[0-9:]+}}]{{$}}
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9:]+}}], s[{{[0-9:]+}}] offset:16
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9:]+}}], s[{{[0-9:]+}}] offset:32
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9]+}}:[[RHI]]], s[{{[0-9:]+}}] offset:48
 define amdgpu_kernel void @test_smfmac_i32_32x32x32_i8(ptr addrspace(1) %arg, <2 x i32> %a, <4 x i32> %b, i32 %idx) #0 {
+; GFX942-VGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_i8:
+; GFX942-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[16:23], s[4:5], 0x2c
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v22, s16
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v23, s17
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v18, s18
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v19, s19
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v20, s20
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v21, s21
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, s22
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_smfmac_i32_32x32x32_i8 v[0:15], v[22:23], v[18:21], v16 cbsz:1 abid:2
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX942-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-VGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_i8:
+; GFX942-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[16:19], s[4:5], 0x2c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[22:23], s[4:5], 0x3c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dword s26, s[4:5], 0x44
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[16:17]
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s20, s18
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s21, s19
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[20:21]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v22, s26
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[22:23]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    v_smfmac_i32_32x32x32_i8 v[0:15], v[20:21], v[16:19], v22 cbsz:1 abid:2
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 7
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX942-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_i8:
+; GFX942-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_smfmac_i32_32x32x32_i8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 7
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_i8:
+; GFX942-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[24:27], s[4:5], 0x2c
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[24:25]
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dword s2, s[4:5], 0x44
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s26
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s27
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s2
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    v_smfmac_i32_32x32x32_i8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 7
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX942-AGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_i8:
+; GFX950-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[16:23], s[4:5], 0x2c
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v22, s16
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v23, s17
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v18, s18
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v19, s19
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v20, s20
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v21, s21
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, s22
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_smfmac_i32_32x32x32_i8 v[0:15], v[22:23], v[18:21], v16 cbsz:1 abid:2
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 2
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX950-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_i8:
+; GFX950-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[16:19], s[4:5], 0x2c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[22:23], s[4:5], 0x3c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dword s26, s[4:5], 0x44
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[16:17]
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s20, s18
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s21, s19
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[20:21]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v22, s26
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[22:23]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-VGPRCD-GISEL-NEXT:    v_smfmac_i32_32x32x32_i8 v[0:15], v[20:21], v[16:19], v22 cbsz:1 abid:2
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v16, 0
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 7
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 2
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX950-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_i8:
+; GFX950-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_smfmac_i32_32x32x32_i8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 2
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX950-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_i8:
+; GFX950-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[24:27], s[4:5], 0x2c
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[24:25]
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dword s2, s[4:5], 0x44
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s26
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s27
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s2
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-AGPRCD-GISEL-NEXT:    v_smfmac_i32_32x32x32_i8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 7
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 2
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX950-AGPRCD-GISEL-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x i32>, ptr addrspace(1) %arg
   %mai.1 = tail call <16 x i32> @llvm.amdgcn.smfmac.i32.32x32x32.i8(<2 x i32> %a, <4 x i32> %b, <16 x i32> %in.1, i32 %idx, i32 1, i32 2)
@@ -318,15 +2824,214 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_smfmac_i32_16x16x64_bf8_bf8:
-; GCN:        s_load_dwordx4 s{{\[}}[[SLO:[0-9]+]]:[[SHI:[0-9]+]]], s[{{[0-9:]+}}], 0x0{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 [[CD:v]]{{\[}}[[RLO:[0-9]+]]:{{[0-9]+}}], s{{\[}}[[SLO]]:{{[0-9]+}}]{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 v[{{[0-9]+}}:[[RHI:[0-9]+]]], s[{{[0-9]+}}:[[SHI]]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 [[CD:a]][[RLO:[0-9]+]], s[[SLO]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 a[[RHI:[0-9]+]], s[[SHI]]{{$}}
-; GCN:        v_smfmac_f32_16x16x64_bf8_bf8 [[CD]]{{\[}}[[RLO]]:[[RHI]]], {{[av]}}[{{[0-9:]+}}], {{[av]}}[{{[0-9:]+}}], v{{[0-9]+}} cbsz:1 abid:2
-; GCN:        global_store_dwordx4 v{{[0-9]+}}, [[CD]]{{\[}}[[RLO]]:[[RHI]]]
 define amdgpu_kernel void @test_smfmac_i32_16x16x64_bf8_bf8(ptr addrspace(1) %arg, <2 x i32> %a, <4 x i32> %b, i32 %idx) #0 {
+; GFX942-VGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_bf8_bf8:
+; GFX942-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v10, s8
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v11, s9
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s10
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s11
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s12
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s13
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s14
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[0:1]
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x64_bf8_bf8 v[6:9], v[10:11], v[2:5], v1 cbsz:1 abid:2
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 6
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, v[6:9], s[6:7]
+; GFX942-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-VGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_bf8_bf8:
+; GFX942-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x24
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x3c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dword s14, s[4:5], 0x44
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[12:13], 0x0
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s4, s2
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s5, s3
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[4:5]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s14
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x64_bf8_bf8 v[8:11], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 5
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[12:13]
+; GFX942-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_bf8_bf8:
+; GFX942-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x0
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x64_bf8_bf8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 5
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_bf8_bf8:
+; GFX942-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-AGPRCD-GISEL-NEXT:    ; implicit-def: $vgpr7 : SGPR spill to VGPR lane
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s2
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s3
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX942-AGPRCD-GISEL-NEXT:    v_writelane_b32 v7, s0, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-AGPRCD-GISEL-NEXT:    v_readlane_b32 s0, v7, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x64_bf8_bf8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 5
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX942-AGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_bf8_bf8:
+; GFX950-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v10, s8
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v11, s9
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s10
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s11
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s12
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s13
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s14
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[0:1]
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x64_bf8_bf8 v[6:9], v[10:11], v[2:5], v1 cbsz:1 abid:2
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, v[6:9], s[6:7]
+; GFX950-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_bf8_bf8:
+; GFX950-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x24
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x3c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dword s14, s[4:5], 0x44
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[12:13], 0x0
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s4, s2
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s5, s3
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[4:5]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s14
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-VGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x64_bf8_bf8 v[8:11], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 6
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[12:13]
+; GFX950-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_bf8_bf8:
+; GFX950-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x0
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x64_bf8_bf8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 6
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX950-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_bf8_bf8:
+; GFX950-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-AGPRCD-GISEL-NEXT:    ; implicit-def: $vgpr7 : SGPR spill to VGPR lane
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s2
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s3
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX950-AGPRCD-GISEL-NEXT:    v_writelane_b32 v7, s0, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-AGPRCD-GISEL-NEXT:    v_readlane_b32 s0, v7, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-AGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x64_bf8_bf8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 6
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX950-AGPRCD-GISEL-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x64.bf8.bf8(<2 x i32> %a, <4 x i32> %b, <4 x float> %in.1, i32 %idx, i32 1, i32 2)
@@ -334,15 +3039,214 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_smfmac_i32_16x16x64_bf8_fp8:
-; GCN:        s_load_dwordx4 s{{\[}}[[SLO:[0-9]+]]:[[SHI:[0-9]+]]], s[{{[0-9:]+}}], 0x0{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 [[CD:v]]{{\[}}[[RLO:[0-9]+]]:{{[0-9]+}}], s{{\[}}[[SLO]]:{{[0-9]+}}]{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 v[{{[0-9]+}}:[[RHI:[0-9]+]]], s[{{[0-9]+}}:[[SHI]]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 [[CD:a]][[RLO:[0-9]+]], s[[SLO]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 a[[RHI:[0-9]+]], s[[SHI]]{{$}}
-; GCN:        v_smfmac_f32_16x16x64_bf8_fp8 [[CD]]{{\[}}[[RLO]]:[[RHI]]], {{[av]}}[{{[0-9:]+}}], {{[av]}}[{{[0-9:]+}}], v{{[0-9]+}} cbsz:1 abid:2
-; GCN:        global_store_dwordx4 v{{[0-9]+}}, [[CD]]{{\[}}[[RLO]]:[[RHI]]]
 define amdgpu_kernel void @test_smfmac_i32_16x16x64_bf8_fp8(ptr addrspace(1) %arg, <2 x i32> %a, <4 x i32> %b, i32 %idx) #0 {
+; GFX942-VGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_bf8_fp8:
+; GFX942-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v10, s8
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v11, s9
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s10
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s11
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s12
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s13
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s14
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[0:1]
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x64_bf8_fp8 v[6:9], v[10:11], v[2:5], v1 cbsz:1 abid:2
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 6
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, v[6:9], s[6:7]
+; GFX942-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-VGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_bf8_fp8:
+; GFX942-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x24
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x3c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dword s14, s[4:5], 0x44
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[12:13], 0x0
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s4, s2
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s5, s3
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[4:5]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s14
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x64_bf8_fp8 v[8:11], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 5
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[12:13]
+; GFX942-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_bf8_fp8:
+; GFX942-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x0
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x64_bf8_fp8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 5
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_bf8_fp8:
+; GFX942-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-AGPRCD-GISEL-NEXT:    ; implicit-def: $vgpr7 : SGPR spill to VGPR lane
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s2
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s3
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX942-AGPRCD-GISEL-NEXT:    v_writelane_b32 v7, s0, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-AGPRCD-GISEL-NEXT:    v_readlane_b32 s0, v7, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x64_bf8_fp8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 5
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX942-AGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_bf8_fp8:
+; GFX950-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v10, s8
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v11, s9
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s10
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s11
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s12
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s13
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s14
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[0:1]
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x64_bf8_fp8 v[6:9], v[10:11], v[2:5], v1 cbsz:1 abid:2
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, v[6:9], s[6:7]
+; GFX950-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_bf8_fp8:
+; GFX950-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x24
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x3c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dword s14, s[4:5], 0x44
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[12:13], 0x0
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s4, s2
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s5, s3
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[4:5]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s14
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-VGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x64_bf8_fp8 v[8:11], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 6
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[12:13]
+; GFX950-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_bf8_fp8:
+; GFX950-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x0
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x64_bf8_fp8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 6
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX950-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_bf8_fp8:
+; GFX950-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-AGPRCD-GISEL-NEXT:    ; implicit-def: $vgpr7 : SGPR spill to VGPR lane
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s2
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s3
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX950-AGPRCD-GISEL-NEXT:    v_writelane_b32 v7, s0, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-AGPRCD-GISEL-NEXT:    v_readlane_b32 s0, v7, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-AGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x64_bf8_fp8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 6
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX950-AGPRCD-GISEL-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x64.bf8.fp8(<2 x i32> %a, <4 x i32> %b, <4 x float> %in.1, i32 %idx, i32 1, i32 2)
@@ -350,15 +3254,214 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_smfmac_i32_16x16x64_fp8_bf8:
-; GCN:        s_load_dwordx4 s{{\[}}[[SLO:[0-9]+]]:[[SHI:[0-9]+]]], s[{{[0-9:]+}}], 0x0{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 [[CD:v]]{{\[}}[[RLO:[0-9]+]]:{{[0-9]+}}], s{{\[}}[[SLO]]:{{[0-9]+}}]{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 v[{{[0-9]+}}:[[RHI:[0-9]+]]], s[{{[0-9]+}}:[[SHI]]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 [[CD:a]][[RLO:[0-9]+]], s[[SLO]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 a[[RHI:[0-9]+]], s[[SHI]]{{$}}
-; GCN:        v_smfmac_f32_16x16x64_fp8_bf8 [[CD]]{{\[}}[[RLO]]:[[RHI]]], {{[av]}}[{{[0-9:]+}}], {{[av]}}[{{[0-9:]+}}], v{{[0-9]+}} cbsz:1 abid:2
-; GCN:        global_store_dwordx4 v{{[0-9]+}}, [[CD]]{{\[}}[[RLO]]:[[RHI]]]
 define amdgpu_kernel void @test_smfmac_i32_16x16x64_fp8_bf8(ptr addrspace(1) %arg, <2 x i32> %a, <4 x i32> %b, i32 %idx) #0 {
+; GFX942-VGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_fp8_bf8:
+; GFX942-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v10, s8
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v11, s9
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s10
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s11
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s12
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s13
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s14
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[0:1]
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x64_fp8_bf8 v[6:9], v[10:11], v[2:5], v1 cbsz:1 abid:2
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 6
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, v[6:9], s[6:7]
+; GFX942-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-VGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_fp8_bf8:
+; GFX942-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x24
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x3c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dword s14, s[4:5], 0x44
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[12:13], 0x0
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s4, s2
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s5, s3
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[4:5]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s14
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x64_fp8_bf8 v[8:11], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 5
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[12:13]
+; GFX942-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_fp8_bf8:
+; GFX942-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x0
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x64_fp8_bf8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 5
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_fp8_bf8:
+; GFX942-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-AGPRCD-GISEL-NEXT:    ; implicit-def: $vgpr7 : SGPR spill to VGPR lane
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s2
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s3
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX942-AGPRCD-GISEL-NEXT:    v_writelane_b32 v7, s0, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-AGPRCD-GISEL-NEXT:    v_readlane_b32 s0, v7, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x64_fp8_bf8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 5
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX942-AGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_fp8_bf8:
+; GFX950-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v10, s8
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v11, s9
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s10
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s11
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s12
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s13
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s14
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[0:1]
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x64_fp8_bf8 v[6:9], v[10:11], v[2:5], v1 cbsz:1 abid:2
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, v[6:9], s[6:7]
+; GFX950-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_fp8_bf8:
+; GFX950-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x24
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x3c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dword s14, s[4:5], 0x44
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[12:13], 0x0
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s4, s2
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s5, s3
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[4:5]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s14
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-VGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x64_fp8_bf8 v[8:11], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 6
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[12:13]
+; GFX950-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_fp8_bf8:
+; GFX950-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x0
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x64_fp8_bf8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 6
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX950-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_fp8_bf8:
+; GFX950-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-AGPRCD-GISEL-NEXT:    ; implicit-def: $vgpr7 : SGPR spill to VGPR lane
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s2
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s3
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX950-AGPRCD-GISEL-NEXT:    v_writelane_b32 v7, s0, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-AGPRCD-GISEL-NEXT:    v_readlane_b32 s0, v7, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-AGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x64_fp8_bf8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 6
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX950-AGPRCD-GISEL-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x64.fp8.bf8(<2 x i32> %a, <4 x i32> %b, <4 x float> %in.1, i32 %idx, i32 1, i32 2)
@@ -366,15 +3469,214 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_smfmac_i32_16x16x64_fp8_fp8:
-; GCN:        s_load_dwordx4 s{{\[}}[[SLO:[0-9]+]]:[[SHI:[0-9]+]]], s[{{[0-9:]+}}], 0x0{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 [[CD:v]]{{\[}}[[RLO:[0-9]+]]:{{[0-9]+}}], s{{\[}}[[SLO]]:{{[0-9]+}}]{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 v[{{[0-9]+}}:[[RHI:[0-9]+]]], s[{{[0-9]+}}:[[SHI]]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 [[CD:a]][[RLO:[0-9]+]], s[[SLO]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 a[[RHI:[0-9]+]], s[[SHI]]{{$}}
-; GCN:        v_smfmac_f32_16x16x64_fp8_fp8 [[CD]]{{\[}}[[RLO]]:[[RHI]]], {{[av]}}[{{[0-9:]+}}], {{[av]}}[{{[0-9:]+}}], v{{[0-9]+}} cbsz:1 abid:2
-; GCN:        global_store_dwordx4 v{{[0-9]+}}, [[CD]]{{\[}}[[RLO]]:[[RHI]]]
 define amdgpu_kernel void @test_smfmac_i32_16x16x64_fp8_fp8(ptr addrspace(1) %arg, <2 x i32> %a, <4 x i32> %b, i32 %idx) #0 {
+; GFX942-VGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_fp8_fp8:
+; GFX942-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v10, s8
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v11, s9
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s10
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s11
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s12
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s13
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s14
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[0:1]
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x64_fp8_fp8 v[6:9], v[10:11], v[2:5], v1 cbsz:1 abid:2
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 6
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, v[6:9], s[6:7]
+; GFX942-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-VGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_fp8_fp8:
+; GFX942-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x24
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x3c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dword s14, s[4:5], 0x44
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[12:13], 0x0
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s4, s2
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s5, s3
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[4:5]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s14
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x64_fp8_fp8 v[8:11], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 5
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[12:13]
+; GFX942-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_fp8_fp8:
+; GFX942-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x0
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x64_fp8_fp8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 5
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_fp8_fp8:
+; GFX942-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-AGPRCD-GISEL-NEXT:    ; implicit-def: $vgpr7 : SGPR spill to VGPR lane
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s2
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s3
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX942-AGPRCD-GISEL-NEXT:    v_writelane_b32 v7, s0, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-AGPRCD-GISEL-NEXT:    v_readlane_b32 s0, v7, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x64_fp8_fp8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 5
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX942-AGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_fp8_fp8:
+; GFX950-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v10, s8
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v11, s9
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s10
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s11
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s12
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[2:3]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s13
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s14
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[0:1]
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x64_fp8_fp8 v[6:9], v[10:11], v[2:5], v1 cbsz:1 abid:2
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, v[6:9], s[6:7]
+; GFX950-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_fp8_fp8:
+; GFX950-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[12:13], s[4:5], 0x24
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x3c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dword s14, s[4:5], 0x44
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[8:11], s[12:13], 0x0
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s4, s2
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s5, s3
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[4:5]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[6:7]
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s14
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-VGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x64_fp8_fp8 v[8:11], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 6
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[12:13]
+; GFX950-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-SDAG-LABEL: test_smfmac_i32_16x16x64_fp8_fp8:
+; GFX950-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x0
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_smfmac_f32_16x16x64_fp8_fp8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 6
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX950-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-GISEL-LABEL: test_smfmac_i32_16x16x64_fp8_fp8:
+; GFX950-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x2c
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX950-AGPRCD-GISEL-NEXT:    ; implicit-def: $vgpr7 : SGPR spill to VGPR lane
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[0:1]
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dword s0, s[4:5], 0x44
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s2
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s3
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX950-AGPRCD-GISEL-NEXT:    v_writelane_b32 v7, s0, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX950-AGPRCD-GISEL-NEXT:    v_readlane_b32 s0, v7, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-AGPRCD-GISEL-NEXT:    v_smfmac_f32_16x16x64_fp8_fp8 a[0:3], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 6
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX950-AGPRCD-GISEL-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x float> @llvm.amdgcn.smfmac.f32.16x16x64.fp8.fp8(<2 x i32> %a, <4 x i32> %b, <4 x float> %in.1, i32 %idx, i32 1, i32 2)
@@ -382,18 +3684,310 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_smfmac_i32_32x32x32_bf8_bf8:
-; GCN:        s_load_dwordx16 s{{\[}}[[SLO:[0-9]+]]:[[SHI:[0-9]+]]], s[{{[0-9:]+}}], 0x0{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 [[CD:v]]{{\[}}[[RLO:[0-9]+]]:{{[0-9]+}}], s{{\[}}[[SLO]]:{{[0-9]+}}]{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 v[{{[0-9]+}}:[[RHI:[0-9]+]]], s[{{[0-9]+}}:[[SHI]]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 [[CD:a]][[RLO:[0-9]+]], s[[SLO]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 a[[RHI:[0-9]+]], s[[SHI]]{{$}}
-; GCN:        v_smfmac_f32_32x32x32_bf8_bf8 [[CD]]{{\[}}[[RLO]]:[[RHI]]], {{[av]}}[{{[0-9:]+}}], {{[av]}}[{{[0-9:]+}}], v{{[0-9]+}} cbsz:1 abid:2
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]]{{\[}}[[RLO]]:{{[0-9]+}}], s[{{[0-9:]+}}]{{$}}
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9:]+}}], s[{{[0-9:]+}}] offset:16
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9:]+}}], s[{{[0-9:]+}}] offset:32
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9]+}}:[[RHI]]], s[{{[0-9:]+}}] offset:48
 define amdgpu_kernel void @test_smfmac_i32_32x32x32_bf8_bf8(ptr addrspace(1) %arg, <2 x i32> %a, <4 x i32> %b, i32 %idx) #0 {
+; GFX942-VGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_bf8_bf8:
+; GFX942-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[16:23], s[4:5], 0x2c
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v22, s16
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v23, s17
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v18, s18
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v19, s19
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v20, s20
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v21, s21
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, s22
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x32_bf8_bf8 v[0:15], v[22:23], v[18:21], v16 cbsz:1 abid:2
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX942-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-VGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_bf8_bf8:
+; GFX942-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[16:19], s[4:5], 0x2c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[22:23], s[4:5], 0x3c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dword s26, s[4:5], 0x44
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[16:17]
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s20, s18
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s21, s19
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[20:21]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v22, s26
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[22:23]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x32_bf8_bf8 v[0:15], v[20:21], v[16:19], v22 cbsz:1 abid:2
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 7
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX942-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_bf8_bf8:
+; GFX942-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x32_bf8_bf8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 7
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_bf8_bf8:
+; GFX942-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[24:27], s[4:5], 0x2c
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[24:25]
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dword s2, s[4:5], 0x44
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s26
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s27
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s2
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x32_bf8_bf8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 7
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX942-AGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_bf8_bf8:
+; GFX950-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[16:23], s[4:5], 0x2c
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v22, s16
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v23, s17
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v18, s18
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v19, s19
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v20, s20
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v21, s21
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, s22
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x32_bf8_bf8 v[0:15], v[22:23], v[18:21], v16 cbsz:1 abid:2
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 2
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX950-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_bf8_bf8:
+; GFX950-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[16:19], s[4:5], 0x2c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[22:23], s[4:5], 0x3c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dword s26, s[4:5], 0x44
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[16:17]
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s20, s18
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s21, s19
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[20:21]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v22, s26
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[22:23]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-VGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x32_bf8_bf8 v[0:15], v[20:21], v[16:19], v22 cbsz:1 abid:2
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v16, 0
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 7
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 2
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX950-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_bf8_bf8:
+; GFX950-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x32_bf8_bf8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 2
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX950-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_bf8_bf8:
+; GFX950-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[24:27], s[4:5], 0x2c
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[24:25]
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dword s2, s[4:5], 0x44
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s26
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s27
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s2
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-AGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x32_bf8_bf8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 7
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 2
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX950-AGPRCD-GISEL-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <16 x float> @llvm.amdgcn.smfmac.f32.32x32x32.bf8.bf8(<2 x i32> %a, <4 x i32> %b, <16 x float> %in.1, i32 %idx, i32 1, i32 2)
@@ -401,18 +3995,310 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_smfmac_i32_32x32x32_bf8_fp8:
-; GCN:        s_load_dwordx16 s{{\[}}[[SLO:[0-9]+]]:[[SHI:[0-9]+]]], s[{{[0-9:]+}}], 0x0{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 [[CD:v]]{{\[}}[[RLO:[0-9]+]]:{{[0-9]+}}], s{{\[}}[[SLO]]:{{[0-9]+}}]{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 v[{{[0-9]+}}:[[RHI:[0-9]+]]], s[{{[0-9]+}}:[[SHI]]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 [[CD:a]][[RLO:[0-9]+]], s[[SLO]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 a[[RHI:[0-9]+]], s[[SHI]]{{$}}
-; GCN:        v_smfmac_f32_32x32x32_bf8_fp8 [[CD]]{{\[}}[[RLO]]:[[RHI]]], {{[av]}}[{{[0-9:]+}}], {{[av]}}[{{[0-9:]+}}], v{{[0-9]+}} cbsz:1 abid:2
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]]{{\[}}[[RLO]]:{{[0-9]+}}], s[{{[0-9:]+}}]{{$}}
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9:]+}}], s[{{[0-9:]+}}] offset:16
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9:]+}}], s[{{[0-9:]+}}] offset:32
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9]+}}:[[RHI]]], s[{{[0-9:]+}}] offset:48
 define amdgpu_kernel void @test_smfmac_i32_32x32x32_bf8_fp8(ptr addrspace(1) %arg, <2 x i32> %a, <4 x i32> %b, i32 %idx) #0 {
+; GFX942-VGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_bf8_fp8:
+; GFX942-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[16:23], s[4:5], 0x2c
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v22, s16
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v23, s17
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v18, s18
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v19, s19
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v20, s20
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v21, s21
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, s22
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x32_bf8_fp8 v[0:15], v[22:23], v[18:21], v16 cbsz:1 abid:2
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX942-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-VGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_bf8_fp8:
+; GFX942-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[16:19], s[4:5], 0x2c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[22:23], s[4:5], 0x3c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dword s26, s[4:5], 0x44
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[16:17]
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s20, s18
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s21, s19
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[20:21]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v22, s26
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[22:23]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x32_bf8_fp8 v[0:15], v[20:21], v[16:19], v22 cbsz:1 abid:2
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 7
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX942-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_bf8_fp8:
+; GFX942-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x32_bf8_fp8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 7
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_bf8_fp8:
+; GFX942-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[24:27], s[4:5], 0x2c
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[24:25]
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dword s2, s[4:5], 0x44
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s26
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s27
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s2
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x32_bf8_fp8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 7
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX942-AGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_bf8_fp8:
+; GFX950-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[16:23], s[4:5], 0x2c
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v22, s16
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v23, s17
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v18, s18
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v19, s19
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v20, s20
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v21, s21
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, s22
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x32_bf8_fp8 v[0:15], v[22:23], v[18:21], v16 cbsz:1 abid:2
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 2
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX950-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_bf8_fp8:
+; GFX950-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[16:19], s[4:5], 0x2c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[22:23], s[4:5], 0x3c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dword s26, s[4:5], 0x44
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[16:17]
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s20, s18
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s21, s19
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[20:21]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v22, s26
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[22:23]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-VGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x32_bf8_fp8 v[0:15], v[20:21], v[16:19], v22 cbsz:1 abid:2
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v16, 0
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 7
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 2
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX950-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_bf8_fp8:
+; GFX950-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x32_bf8_fp8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 2
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX950-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_bf8_fp8:
+; GFX950-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[24:27], s[4:5], 0x2c
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[24:25]
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dword s2, s[4:5], 0x44
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s26
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s27
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s2
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-AGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x32_bf8_fp8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 7
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 2
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX950-AGPRCD-GISEL-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <16 x float> @llvm.amdgcn.smfmac.f32.32x32x32.bf8.fp8(<2 x i32> %a, <4 x i32> %b, <16 x float> %in.1, i32 %idx, i32 1, i32 2)
@@ -420,18 +4306,310 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_smfmac_i32_32x32x32_fp8_bf8:
-; GCN:        s_load_dwordx16 s{{\[}}[[SLO:[0-9]+]]:[[SHI:[0-9]+]]], s[{{[0-9:]+}}], 0x0{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 [[CD:v]]{{\[}}[[RLO:[0-9]+]]:{{[0-9]+}}], s{{\[}}[[SLO]]:{{[0-9]+}}]{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 v[{{[0-9]+}}:[[RHI:[0-9]+]]], s[{{[0-9]+}}:[[SHI]]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 [[CD:a]][[RLO:[0-9]+]], s[[SLO]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 a[[RHI:[0-9]+]], s[[SHI]]{{$}}
-; GCN:        v_smfmac_f32_32x32x32_fp8_bf8 [[CD]]{{\[}}[[RLO]]:[[RHI]]], {{[av]}}[{{[0-9:]+}}], {{[av]}}[{{[0-9:]+}}], v{{[0-9]+}} cbsz:1 abid:2
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]]{{\[}}[[RLO]]:{{[0-9]+}}], s[{{[0-9:]+}}]{{$}}
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9:]+}}], s[{{[0-9:]+}}] offset:16
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9:]+}}], s[{{[0-9:]+}}] offset:32
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9]+}}:[[RHI]]], s[{{[0-9:]+}}] offset:48
 define amdgpu_kernel void @test_smfmac_i32_32x32x32_fp8_bf8(ptr addrspace(1) %arg, <2 x i32> %a, <4 x i32> %b, i32 %idx) #0 {
+; GFX942-VGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_fp8_bf8:
+; GFX942-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[16:23], s[4:5], 0x2c
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v22, s16
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v23, s17
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v18, s18
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v19, s19
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v20, s20
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v21, s21
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, s22
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x32_fp8_bf8 v[0:15], v[22:23], v[18:21], v16 cbsz:1 abid:2
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX942-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-VGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_fp8_bf8:
+; GFX942-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[16:19], s[4:5], 0x2c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[22:23], s[4:5], 0x3c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dword s26, s[4:5], 0x44
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[16:17]
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s20, s18
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s21, s19
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[20:21]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v22, s26
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[22:23]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x32_fp8_bf8 v[0:15], v[20:21], v[16:19], v22 cbsz:1 abid:2
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 7
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX942-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_fp8_bf8:
+; GFX942-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x32_fp8_bf8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 7
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_fp8_bf8:
+; GFX942-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[24:27], s[4:5], 0x2c
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[24:25]
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dword s2, s[4:5], 0x44
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s26
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s27
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s2
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x32_fp8_bf8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 7
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX942-AGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_fp8_bf8:
+; GFX950-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[16:23], s[4:5], 0x2c
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v22, s16
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v23, s17
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v18, s18
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v19, s19
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v20, s20
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v21, s21
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, s22
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x32_fp8_bf8 v[0:15], v[22:23], v[18:21], v16 cbsz:1 abid:2
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 2
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX950-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_fp8_bf8:
+; GFX950-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[16:19], s[4:5], 0x2c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[22:23], s[4:5], 0x3c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dword s26, s[4:5], 0x44
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[16:17]
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s20, s18
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s21, s19
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[20:21]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v22, s26
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[22:23]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-VGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x32_fp8_bf8 v[0:15], v[20:21], v[16:19], v22 cbsz:1 abid:2
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v16, 0
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 7
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 2
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX950-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_fp8_bf8:
+; GFX950-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x32_fp8_bf8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 2
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX950-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_fp8_bf8:
+; GFX950-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[24:27], s[4:5], 0x2c
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[24:25]
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dword s2, s[4:5], 0x44
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s26
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s27
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s2
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-AGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x32_fp8_bf8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 7
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 2
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX950-AGPRCD-GISEL-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <16 x float> @llvm.amdgcn.smfmac.f32.32x32x32.fp8.bf8(<2 x i32> %a, <4 x i32> %b, <16 x float> %in.1, i32 %idx, i32 1, i32 2)
@@ -439,18 +4617,310 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_smfmac_i32_32x32x32_fp8_fp8:
-; GCN:        s_load_dwordx16 s{{\[}}[[SLO:[0-9]+]]:[[SHI:[0-9]+]]], s[{{[0-9:]+}}], 0x0{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 [[CD:v]]{{\[}}[[RLO:[0-9]+]]:{{[0-9]+}}], s{{\[}}[[SLO]]:{{[0-9]+}}]{{$}}
-; VGPRCD-DAG: v_mov_b64_e32 v[{{[0-9]+}}:[[RHI:[0-9]+]]], s[{{[0-9]+}}:[[SHI]]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 [[CD:a]][[RLO:[0-9]+]], s[[SLO]]{{$}}
-; AGPRCD-DAG: v_accvgpr_write_b32 a[[RHI:[0-9]+]], s[[SHI]]{{$}}
-; GCN:        v_smfmac_f32_32x32x32_fp8_fp8 [[CD]]{{\[}}[[RLO]]:[[RHI]]], {{[av]}}[{{[0-9:]+}}], {{[av]}}[{{[0-9:]+}}], v{{[0-9]+}} cbsz:1 abid:2
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]]{{\[}}[[RLO]]:{{[0-9]+}}], s[{{[0-9:]+}}]{{$}}
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9:]+}}], s[{{[0-9:]+}}] offset:16
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9:]+}}], s[{{[0-9:]+}}] offset:32
-; GCN-DAG:    global_store_dwordx4 v{{[0-9]+}}, [[CD]][{{[0-9]+}}:[[RHI]]], s[{{[0-9:]+}}] offset:48
 define amdgpu_kernel void @test_smfmac_i32_32x32x32_fp8_fp8(ptr addrspace(1) %arg, <2 x i32> %a, <4 x i32> %b, i32 %idx) #0 {
+; GFX942-VGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_fp8_fp8:
+; GFX942-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[16:23], s[4:5], 0x2c
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v22, s16
+; GFX942-VGPRCD-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v23, s17
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v18, s18
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v19, s19
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v20, s20
+; GFX942-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v21, s21
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, s22
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x32_fp8_fp8 v[0:15], v[22:23], v[18:21], v16 cbsz:1 abid:2
+; GFX942-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX942-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX942-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX942-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-VGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_fp8_fp8:
+; GFX942-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[16:19], s[4:5], 0x2c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[22:23], s[4:5], 0x3c
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dword s26, s[4:5], 0x44
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[16:17]
+; GFX942-VGPRCD-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s20, s18
+; GFX942-VGPRCD-GISEL-NEXT:    s_mov_b32 s21, s19
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[20:21]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v22, s26
+; GFX942-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[22:23]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x32_fp8_fp8 v[0:15], v[20:21], v[16:19], v22 cbsz:1 abid:2
+; GFX942-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 7
+; GFX942-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX942-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX942-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_fp8_fp8:
+; GFX942-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX942-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX942-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX942-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x32_fp8_fp8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 7
+; GFX942-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX942-AGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_fp8_fp8:
+; GFX942-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[24:27], s[4:5], 0x2c
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[24:25]
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX942-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX942-AGPRCD-GISEL-NEXT:    s_load_dword s2, s[4:5], 0x44
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s26
+; GFX942-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s27
+; GFX942-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s2
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x32_fp8_fp8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX942-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 7
+; GFX942-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX942-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX942-AGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_fp8_fp8:
+; GFX950-VGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx8 s[16:23], s[4:5], 0x2c
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v22, s16
+; GFX950-VGPRCD-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v23, s17
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v18, s18
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v19, s19
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v20, s20
+; GFX950-VGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v21, s21
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, s22
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-VGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x32_fp8_fp8 v[0:15], v[22:23], v[18:21], v16 cbsz:1 abid:2
+; GFX950-VGPRCD-SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-VGPRCD-SDAG-NEXT:    s_nop 2
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX950-VGPRCD-SDAG-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX950-VGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-VGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_fp8_fp8:
+; GFX950-VGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx4 s[16:19], s[4:5], 0x2c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[24:25], s[4:5], 0x24
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx2 s[22:23], s[4:5], 0x3c
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dword s26, s[4:5], 0x44
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[20:21], s[16:17]
+; GFX950-VGPRCD-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[24:25], 0x0
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s20, s18
+; GFX950-VGPRCD-GISEL-NEXT:    s_mov_b32 s21, s19
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[16:17], s[20:21]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v22, s26
+; GFX950-VGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[18:19], s[22:23]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-VGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x32_fp8_fp8 v[0:15], v[20:21], v[16:19], v22 cbsz:1 abid:2
+; GFX950-VGPRCD-GISEL-NEXT:    v_mov_b32_e32 v16, 0
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 7
+; GFX950-VGPRCD-GISEL-NEXT:    s_nop 2
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[0:3], s[24:25]
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[4:7], s[24:25] offset:16
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[8:11], s[24:25] offset:32
+; GFX950-VGPRCD-GISEL-NEXT:    global_store_dwordx4 v16, v[12:15], s[24:25] offset:48
+; GFX950-VGPRCD-GISEL-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-SDAG-LABEL: test_smfmac_i32_32x32x32_fp8_fp8:
+; GFX950-AGPRCD-SDAG:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX950-AGPRCD-SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX950-AGPRCD-SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x2c
+; GFX950-AGPRCD-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v4, s8
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v5, s9
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, s10
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v1, s11
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v6, s14
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 1
+; GFX950-AGPRCD-SDAG-NEXT:    v_smfmac_f32_32x32x32_fp8_fp8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 7
+; GFX950-AGPRCD-SDAG-NEXT:    s_nop 2
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX950-AGPRCD-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX950-AGPRCD-SDAG-NEXT:    s_endpgm
+;
+; GFX950-AGPRCD-GISEL-LABEL: test_smfmac_i32_32x32x32_fp8_fp8:
+; GFX950-AGPRCD-GISEL:       ; %bb.0: ; %bb
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx4 s[24:27], s[4:5], 0x2c
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[24:25]
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx16 s[8:23], s[0:1], 0x0
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a4, s12
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a5, s13
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a6, s14
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a7, s15
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a8, s16
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a9, s17
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a10, s18
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a11, s19
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a12, s20
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a13, s21
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a14, s22
+; GFX950-AGPRCD-GISEL-NEXT:    v_accvgpr_write_b32 a15, s23
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dwordx2 s[10:11], s[4:5], 0x3c
+; GFX950-AGPRCD-GISEL-NEXT:    s_load_dword s2, s[4:5], 0x44
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s8, s26
+; GFX950-AGPRCD-GISEL-NEXT:    s_mov_b32 s9, s27
+; GFX950-AGPRCD-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v6, s2
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 1
+; GFX950-AGPRCD-GISEL-NEXT:    v_smfmac_f32_32x32x32_fp8_fp8 a[0:15], v[4:5], v[0:3], v6 cbsz:1 abid:2
+; GFX950-AGPRCD-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 7
+; GFX950-AGPRCD-GISEL-NEXT:    s_nop 2
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX950-AGPRCD-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX950-AGPRCD-GISEL-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <16 x float> @llvm.amdgcn.smfmac.f32.32x32x32.fp8.fp8(<2 x i32> %a, <4 x i32> %b, <16 x float> %in.1, i32 %idx, i32 1, i32 2)
@@ -459,3 +4929,8 @@ bb:
 }
 
 attributes #0 = { "amdgpu-flat-work-group-size"="1,256" }
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; GFX942: {{.*}}
+; GFX942-VGPRCD: {{.*}}
+; GFX950: {{.*}}
+; GFX950-VGPRCD: {{.*}}
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.bf16.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.bf16.ll
index 452033f..d358837 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.bf16.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.bf16.ll
@@ -15,9 +15,9 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_bf16(<8 x bfloat> %arg0, <8 x
 ; GCN:       ; %bb.0:
 ; GCN-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; GCN-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; GCN-NEXT:    v_mov_b64_e32 v[12:13], 48
-; GCN-NEXT:    v_mov_b64_e32 v[14:15], 32
-; GCN-NEXT:    v_mov_b64_e32 v[16:17], 16
+; GCN-NEXT:    v_mov_b64_e32 v[8:9], 48
+; GCN-NEXT:    v_mov_b64_e32 v[10:11], 32
+; GCN-NEXT:    v_mov_b64_e32 v[12:13], 16
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    v_mov_b64_e32 v[0:1], s[24:25]
 ; GCN-NEXT:    v_mov_b64_e32 v[2:3], s[26:27]
@@ -39,42 +39,42 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_bf16(<8 x bfloat> %arg0, <8 x
 ; GCN-NEXT:    v_accvgpr_write_b32 a13, s21
 ; GCN-NEXT:    v_accvgpr_write_b32 a14, s22
 ; GCN-NEXT:    v_accvgpr_write_b32 a15, s23
-; GCN-NEXT:    v_mov_b64_e32 v[18:19], 0
-; GCN-NEXT:    v_mov_b32_e32 v8, s16
+; GCN-NEXT:    v_mov_b64_e32 v[14:15], 0
+; GCN-NEXT:    v_mov_b32_e32 v16, s16
 ; GCN-NEXT:    v_mfma_f32_32x32x16_bf16 a[16:31], v[0:3], v[4:7], a[0:15]
 ; GCN-NEXT:    v_mov_b32_e32 v0, s20
 ; GCN-NEXT:    v_mov_b32_e32 v1, s21
 ; GCN-NEXT:    v_mov_b32_e32 v2, s22
 ; GCN-NEXT:    v_mov_b32_e32 v3, s23
-; GCN-NEXT:    v_mov_b32_e32 v9, s17
-; GCN-NEXT:    v_mov_b32_e32 v10, s18
-; GCN-NEXT:    v_mov_b32_e32 v11, s19
+; GCN-NEXT:    v_mov_b32_e32 v17, s17
+; GCN-NEXT:    v_mov_b32_e32 v18, s18
+; GCN-NEXT:    v_mov_b32_e32 v19, s19
 ; GCN-NEXT:    s_nop 4
-; GCN-NEXT:    global_store_dwordx4 v[12:13], a[28:31], off sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v[8:9], a[28:31], off sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v[14:15], a[24:27], off sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v[10:11], a[24:27], off sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v[16:17], a[20:23], off sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v[12:13], a[20:23], off sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v[18:19], a[16:19], off sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v[14:15], a[16:19], off sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v[14:15], v[8:11], off sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v[10:11], v[16:19], off sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v[12:13], v[0:3], off sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v[8:9], v[0:3], off sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_nop 0
 ; GCN-NEXT:    v_mov_b32_e32 v0, s8
 ; GCN-NEXT:    v_mov_b32_e32 v1, s9
 ; GCN-NEXT:    v_mov_b32_e32 v2, s10
 ; GCN-NEXT:    v_mov_b32_e32 v3, s11
-; GCN-NEXT:    global_store_dwordx4 v[18:19], v[0:3], off sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v[14:15], v[0:3], off sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_nop 0
 ; GCN-NEXT:    v_mov_b32_e32 v0, s12
 ; GCN-NEXT:    v_mov_b32_e32 v1, s13
 ; GCN-NEXT:    v_mov_b32_e32 v2, s14
 ; GCN-NEXT:    v_mov_b32_e32 v3, s15
-; GCN-NEXT:    global_store_dwordx4 v[16:17], v[0:3], off sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v[12:13], v[0:3], off sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_endpgm
   %result = call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.bf16(<8 x bfloat> %arg0, <8 x bfloat> %arg1, <16 x float> %arg2, i32 0, i32 0, i32 0)
@@ -88,9 +88,9 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_bf16__flags(<8 x bfloat> %arg0
 ; GCN:       ; %bb.0:
 ; GCN-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; GCN-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; GCN-NEXT:    v_mov_b64_e32 v[12:13], 48
-; GCN-NEXT:    v_mov_b64_e32 v[14:15], 32
-; GCN-NEXT:    v_mov_b64_e32 v[16:17], 16
+; GCN-NEXT:    v_mov_b64_e32 v[8:9], 48
+; GCN-NEXT:    v_mov_b64_e32 v[10:11], 32
+; GCN-NEXT:    v_mov_b64_e32 v[12:13], 16
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    v_mov_b64_e32 v[0:1], s[24:25]
 ; GCN-NEXT:    v_mov_b64_e32 v[2:3], s[26:27]
@@ -112,42 +112,42 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_bf16__flags(<8 x bfloat> %arg0
 ; GCN-NEXT:    v_accvgpr_write_b32 a13, s21
 ; GCN-NEXT:    v_accvgpr_write_b32 a14, s22
 ; GCN-NEXT:    v_accvgpr_write_b32 a15, s23
-; GCN-NEXT:    v_mov_b64_e32 v[18:19], 0
-; GCN-NEXT:    v_mov_b32_e32 v8, s16
+; GCN-NEXT:    v_mov_b64_e32 v[14:15], 0
+; GCN-NEXT:    v_mov_b32_e32 v16, s16
 ; GCN-NEXT:    v_mfma_f32_32x32x16_bf16 a[16:31], v[0:3], v[4:7], a[0:15] cbsz:2 abid:3 blgp:1
 ; GCN-NEXT:    v_mov_b32_e32 v0, s20
 ; GCN-NEXT:    v_mov_b32_e32 v1, s21
 ; GCN-NEXT:    v_mov_b32_e32 v2, s22
 ; GCN-NEXT:    v_mov_b32_e32 v3, s23
-; GCN-NEXT:    v_mov_b32_e32 v9, s17
-; GCN-NEXT:    v_mov_b32_e32 v10, s18
-; GCN-NEXT:    v_mov_b32_e32 v11, s19
+; GCN-NEXT:    v_mov_b32_e32 v17, s17
+; GCN-NEXT:    v_mov_b32_e32 v18, s18
+; GCN-NEXT:    v_mov_b32_e32 v19, s19
 ; GCN-NEXT:    s_nop 4
-; GCN-NEXT:    global_store_dwordx4 v[12:13], a[28:31], off sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v[8:9], a[28:31], off sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v[14:15], a[24:27], off sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v[10:11], a[24:27], off sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v[16:17], a[20:23], off sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v[12:13], a[20:23], off sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v[18:19], a[16:19], off sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v[14:15], a[16:19], off sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v[14:15], v[8:11], off sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v[10:11], v[16:19], off sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v[12:13], v[0:3], off sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v[8:9], v[0:3], off sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_nop 0
 ; GCN-NEXT:    v_mov_b32_e32 v0, s8
 ; GCN-NEXT:    v_mov_b32_e32 v1, s9
 ; GCN-NEXT:    v_mov_b32_e32 v2, s10
 ; GCN-NEXT:    v_mov_b32_e32 v3, s11
-; GCN-NEXT:    global_store_dwordx4 v[18:19], v[0:3], off sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v[14:15], v[0:3], off sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_nop 0
 ; GCN-NEXT:    v_mov_b32_e32 v0, s12
 ; GCN-NEXT:    v_mov_b32_e32 v1, s13
 ; GCN-NEXT:    v_mov_b32_e32 v2, s14
 ; GCN-NEXT:    v_mov_b32_e32 v3, s15
-; GCN-NEXT:    global_store_dwordx4 v[16:17], v[0:3], off sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v[12:13], v[0:3], off sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_endpgm
   %result = call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.bf16(<8 x bfloat> %arg0, <8 x bfloat> %arg1, <16 x float> %arg2, i32 2, i32 3, i32 1)
@@ -252,7 +252,7 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_bf16__vgprcd(<8 x bfloat> %arg
 ; GCN-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; GCN-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
 ; GCN-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0xa4
-; GCN-NEXT:    v_mov_b32_e32 v12, 0
+; GCN-NEXT:    v_mov_b32_e32 v8, 0
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    v_mov_b64_e32 v[0:1], s[24:25]
 ; GCN-NEXT:    v_mov_b64_e32 v[2:3], s[26:27]
@@ -274,40 +274,40 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_bf16__vgprcd(<8 x bfloat> %arg
 ; GCN-NEXT:    v_accvgpr_write_b32 a18, s10
 ; GCN-NEXT:    v_accvgpr_write_b32 a17, s9
 ; GCN-NEXT:    v_accvgpr_write_b32 a16, s8
-; GCN-NEXT:    v_mov_b32_e32 v8, s20
-; GCN-NEXT:    v_mov_b32_e32 v9, s21
+; GCN-NEXT:    v_mov_b32_e32 v10, s20
+; GCN-NEXT:    v_mov_b32_e32 v11, s21
 ; GCN-NEXT:    v_mfma_f32_32x32x16_bf16 a[0:15], v[0:3], v[4:7], a[16:31]
-; GCN-NEXT:    v_mov_b32_e32 v10, s22
-; GCN-NEXT:    v_mov_b32_e32 v11, s23
+; GCN-NEXT:    v_mov_b32_e32 v12, s22
+; GCN-NEXT:    v_mov_b32_e32 v13, s23
 ; GCN-NEXT:    v_mov_b32_e32 v0, s16
 ; GCN-NEXT:    v_mov_b32_e32 v1, s17
 ; GCN-NEXT:    v_mov_b32_e32 v2, s18
 ; GCN-NEXT:    v_mov_b32_e32 v3, s19
-; GCN-NEXT:    global_store_dwordx4 v12, v[8:11], s[0:1] offset:48 sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v8, v[10:13], s[0:1] offset:48 sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] offset:32 sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:32 sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_nop 0
 ; GCN-NEXT:    v_mov_b32_e32 v0, s12
 ; GCN-NEXT:    v_mov_b32_e32 v1, s13
 ; GCN-NEXT:    v_mov_b32_e32 v2, s14
 ; GCN-NEXT:    v_mov_b32_e32 v3, s15
-; GCN-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] offset:16 sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:16 sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_nop 0
 ; GCN-NEXT:    v_mov_b32_e32 v0, s8
 ; GCN-NEXT:    v_mov_b32_e32 v1, s9
 ; GCN-NEXT:    v_mov_b32_e32 v2, s10
 ; GCN-NEXT:    v_mov_b32_e32 v3, s11
-; GCN-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v12, a[8:11], s[0:1] offset:32 sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v8, a[8:11], s[0:1] offset:32 sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v12, a[12:15], s[0:1] offset:48 sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v8, a[12:15], s[0:1] offset:48 sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v12, a[0:3], s[0:1] sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v8, a[0:3], s[0:1] sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v12, a[4:7], s[0:1] offset:16 sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v8, a[4:7], s[0:1] offset:16 sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_endpgm
   %result = call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.bf16(<8 x bfloat> %arg0, <8 x bfloat> %arg1, <16 x float> %arg2, i32 0, i32 0, i32 0)
@@ -322,7 +322,7 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_bf16__vgprcd__flags(<8 x bfloa
 ; GCN-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; GCN-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
 ; GCN-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0xa4
-; GCN-NEXT:    v_mov_b32_e32 v12, 0
+; GCN-NEXT:    v_mov_b32_e32 v8, 0
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    v_mov_b64_e32 v[0:1], s[24:25]
 ; GCN-NEXT:    v_mov_b64_e32 v[2:3], s[26:27]
@@ -344,40 +344,40 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_bf16__vgprcd__flags(<8 x bfloa
 ; GCN-NEXT:    v_accvgpr_write_b32 a18, s10
 ; GCN-NEXT:    v_accvgpr_write_b32 a17, s9
 ; GCN-NEXT:    v_accvgpr_write_b32 a16, s8
-; GCN-NEXT:    v_mov_b32_e32 v8, s20
-; GCN-NEXT:    v_mov_b32_e32 v9, s21
+; GCN-NEXT:    v_mov_b32_e32 v10, s20
+; GCN-NEXT:    v_mov_b32_e32 v11, s21
 ; GCN-NEXT:    v_mfma_f32_32x32x16_bf16 a[0:15], v[0:3], v[4:7], a[16:31] cbsz:1 abid:2 blgp:3
-; GCN-NEXT:    v_mov_b32_e32 v10, s22
-; GCN-NEXT:    v_mov_b32_e32 v11, s23
+; GCN-NEXT:    v_mov_b32_e32 v12, s22
+; GCN-NEXT:    v_mov_b32_e32 v13, s23
 ; GCN-NEXT:    v_mov_b32_e32 v0, s16
 ; GCN-NEXT:    v_mov_b32_e32 v1, s17
 ; GCN-NEXT:    v_mov_b32_e32 v2, s18
 ; GCN-NEXT:    v_mov_b32_e32 v3, s19
-; GCN-NEXT:    global_store_dwordx4 v12, v[8:11], s[0:1] offset:48 sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v8, v[10:13], s[0:1] offset:48 sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] offset:32 sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:32 sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_nop 0
 ; GCN-NEXT:    v_mov_b32_e32 v0, s12
 ; GCN-NEXT:    v_mov_b32_e32 v1, s13
 ; GCN-NEXT:    v_mov_b32_e32 v2, s14
 ; GCN-NEXT:    v_mov_b32_e32 v3, s15
-; GCN-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] offset:16 sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:16 sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_nop 0
 ; GCN-NEXT:    v_mov_b32_e32 v0, s8
 ; GCN-NEXT:    v_mov_b32_e32 v1, s9
 ; GCN-NEXT:    v_mov_b32_e32 v2, s10
 ; GCN-NEXT:    v_mov_b32_e32 v3, s11
-; GCN-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v12, a[8:11], s[0:1] offset:32 sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v8, a[8:11], s[0:1] offset:32 sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v12, a[12:15], s[0:1] offset:48 sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v8, a[12:15], s[0:1] offset:48 sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v12, a[0:3], s[0:1] sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v8, a[0:3], s[0:1] sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
-; GCN-NEXT:    global_store_dwordx4 v12, a[4:7], s[0:1] offset:16 sc0 sc1
+; GCN-NEXT:    global_store_dwordx4 v8, a[4:7], s[0:1] offset:16 sc0 sc1
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_endpgm
   %result = call <16 x float> @llvm.amdgcn.mfma.f32.32x32x16.bf16(<8 x bfloat> %arg0, <8 x bfloat> %arg1, <16 x float> %arg2, i32 1, i32 2, i32 3)
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.ll
index 9bdae28f..21465be 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.gfx950.ll
@@ -394,9 +394,9 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16(<8 x half> %arg0, <8 x hal
 ; SDAG:       ; %bb.0:
 ; SDAG-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; SDAG-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; SDAG-NEXT:    v_mov_b64_e32 v[12:13], 48
-; SDAG-NEXT:    v_mov_b64_e32 v[14:15], 32
-; SDAG-NEXT:    v_mov_b64_e32 v[16:17], 16
+; SDAG-NEXT:    v_mov_b64_e32 v[8:9], 48
+; SDAG-NEXT:    v_mov_b64_e32 v[10:11], 32
+; SDAG-NEXT:    v_mov_b64_e32 v[12:13], 16
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
 ; SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[24:25]
 ; SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[26:27]
@@ -418,42 +418,42 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16(<8 x half> %arg0, <8 x hal
 ; SDAG-NEXT:    v_accvgpr_write_b32 a13, s21
 ; SDAG-NEXT:    v_accvgpr_write_b32 a14, s22
 ; SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
-; SDAG-NEXT:    v_mov_b64_e32 v[18:19], 0
-; SDAG-NEXT:    v_mov_b32_e32 v8, s16
+; SDAG-NEXT:    v_mov_b64_e32 v[14:15], 0
+; SDAG-NEXT:    v_mov_b32_e32 v16, s16
 ; SDAG-NEXT:    v_mfma_f32_32x32x16_f16 a[16:31], v[0:3], v[4:7], a[0:15]
 ; SDAG-NEXT:    v_mov_b32_e32 v0, s20
 ; SDAG-NEXT:    v_mov_b32_e32 v1, s21
 ; SDAG-NEXT:    v_mov_b32_e32 v2, s22
 ; SDAG-NEXT:    v_mov_b32_e32 v3, s23
-; SDAG-NEXT:    v_mov_b32_e32 v9, s17
-; SDAG-NEXT:    v_mov_b32_e32 v10, s18
-; SDAG-NEXT:    v_mov_b32_e32 v11, s19
+; SDAG-NEXT:    v_mov_b32_e32 v17, s17
+; SDAG-NEXT:    v_mov_b32_e32 v18, s18
+; SDAG-NEXT:    v_mov_b32_e32 v19, s19
 ; SDAG-NEXT:    s_nop 4
-; SDAG-NEXT:    global_store_dwordx4 v[12:13], a[28:31], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[8:9], a[28:31], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[14:15], a[24:27], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[10:11], a[24:27], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[16:17], a[20:23], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[12:13], a[20:23], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[18:19], a[16:19], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[14:15], a[16:19], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[14:15], v[8:11], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[10:11], v[16:19], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[12:13], v[0:3], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[8:9], v[0:3], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
 ; SDAG-NEXT:    v_mov_b32_e32 v0, s8
 ; SDAG-NEXT:    v_mov_b32_e32 v1, s9
 ; SDAG-NEXT:    v_mov_b32_e32 v2, s10
 ; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    global_store_dwordx4 v[18:19], v[0:3], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[14:15], v[0:3], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
 ; SDAG-NEXT:    v_mov_b32_e32 v0, s12
 ; SDAG-NEXT:    v_mov_b32_e32 v1, s13
 ; SDAG-NEXT:    v_mov_b32_e32 v2, s14
 ; SDAG-NEXT:    v_mov_b32_e32 v3, s15
-; SDAG-NEXT:    global_store_dwordx4 v[16:17], v[0:3], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[12:13], v[0:3], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_endpgm
 ;
@@ -518,9 +518,9 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16(<8 x half> %arg0, <8 x hal
 ; HEURRC:       ; %bb.0:
 ; HEURRC-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; HEURRC-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; HEURRC-NEXT:    v_mov_b64_e32 v[12:13], 48
-; HEURRC-NEXT:    v_mov_b64_e32 v[14:15], 32
-; HEURRC-NEXT:    v_mov_b64_e32 v[16:17], 16
+; HEURRC-NEXT:    v_mov_b64_e32 v[8:9], 48
+; HEURRC-NEXT:    v_mov_b64_e32 v[10:11], 32
+; HEURRC-NEXT:    v_mov_b64_e32 v[12:13], 16
 ; HEURRC-NEXT:    s_waitcnt lgkmcnt(0)
 ; HEURRC-NEXT:    v_mov_b64_e32 v[0:1], s[24:25]
 ; HEURRC-NEXT:    v_mov_b64_e32 v[2:3], s[26:27]
@@ -542,42 +542,42 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16(<8 x half> %arg0, <8 x hal
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a13, s21
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a14, s22
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a15, s23
-; HEURRC-NEXT:    v_mov_b64_e32 v[18:19], 0
-; HEURRC-NEXT:    v_mov_b32_e32 v8, s16
+; HEURRC-NEXT:    v_mov_b64_e32 v[14:15], 0
+; HEURRC-NEXT:    v_mov_b32_e32 v16, s16
 ; HEURRC-NEXT:    v_mfma_f32_32x32x16_f16 a[16:31], v[0:3], v[4:7], a[0:15]
 ; HEURRC-NEXT:    v_mov_b32_e32 v0, s20
 ; HEURRC-NEXT:    v_mov_b32_e32 v1, s21
 ; HEURRC-NEXT:    v_mov_b32_e32 v2, s22
 ; HEURRC-NEXT:    v_mov_b32_e32 v3, s23
-; HEURRC-NEXT:    v_mov_b32_e32 v9, s17
-; HEURRC-NEXT:    v_mov_b32_e32 v10, s18
-; HEURRC-NEXT:    v_mov_b32_e32 v11, s19
+; HEURRC-NEXT:    v_mov_b32_e32 v17, s17
+; HEURRC-NEXT:    v_mov_b32_e32 v18, s18
+; HEURRC-NEXT:    v_mov_b32_e32 v19, s19
 ; HEURRC-NEXT:    s_nop 4
-; HEURRC-NEXT:    global_store_dwordx4 v[12:13], a[28:31], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[8:9], a[28:31], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v[14:15], a[24:27], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[10:11], a[24:27], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v[16:17], a[20:23], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[12:13], a[20:23], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v[18:19], a[16:19], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[14:15], a[16:19], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v[14:15], v[8:11], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[10:11], v[16:19], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v[12:13], v[0:3], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[8:9], v[0:3], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_nop 0
 ; HEURRC-NEXT:    v_mov_b32_e32 v0, s8
 ; HEURRC-NEXT:    v_mov_b32_e32 v1, s9
 ; HEURRC-NEXT:    v_mov_b32_e32 v2, s10
 ; HEURRC-NEXT:    v_mov_b32_e32 v3, s11
-; HEURRC-NEXT:    global_store_dwordx4 v[18:19], v[0:3], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[14:15], v[0:3], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_nop 0
 ; HEURRC-NEXT:    v_mov_b32_e32 v0, s12
 ; HEURRC-NEXT:    v_mov_b32_e32 v1, s13
 ; HEURRC-NEXT:    v_mov_b32_e32 v2, s14
 ; HEURRC-NEXT:    v_mov_b32_e32 v3, s15
-; HEURRC-NEXT:    global_store_dwordx4 v[16:17], v[0:3], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[12:13], v[0:3], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_endpgm
 ;
@@ -585,9 +585,9 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16(<8 x half> %arg0, <8 x hal
 ; VGPRRC:       ; %bb.0:
 ; VGPRRC-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; VGPRRC-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; VGPRRC-NEXT:    v_mov_b64_e32 v[44:45], 48
-; VGPRRC-NEXT:    v_mov_b64_e32 v[46:47], 32
-; VGPRRC-NEXT:    v_mov_b64_e32 v[48:49], 16
+; VGPRRC-NEXT:    v_mov_b64_e32 v[40:41], 48
+; VGPRRC-NEXT:    v_mov_b64_e32 v[42:43], 32
+; VGPRRC-NEXT:    v_mov_b64_e32 v[44:45], 16
 ; VGPRRC-NEXT:    s_waitcnt lgkmcnt(0)
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[34:35], s[26:27]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[32:33], s[24:25]
@@ -601,43 +601,43 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16(<8 x half> %arg0, <8 x hal
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[10:11], s[18:19]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[12:13], s[20:21]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[14:15], s[22:23]
-; VGPRRC-NEXT:    v_mov_b64_e32 v[50:51], 0
-; VGPRRC-NEXT:    v_mov_b32_e32 v40, s16
+; VGPRRC-NEXT:    v_mov_b64_e32 v[46:47], 0
+; VGPRRC-NEXT:    v_mov_b32_e32 v48, s16
 ; VGPRRC-NEXT:    v_mfma_f32_32x32x16_f16 v[16:31], v[32:35], v[36:39], v[0:15]
-; VGPRRC-NEXT:    v_mov_b32_e32 v41, s17
-; VGPRRC-NEXT:    v_mov_b32_e32 v42, s18
-; VGPRRC-NEXT:    v_mov_b32_e32 v43, s19
+; VGPRRC-NEXT:    v_mov_b32_e32 v49, s17
+; VGPRRC-NEXT:    v_mov_b32_e32 v50, s18
+; VGPRRC-NEXT:    v_mov_b32_e32 v51, s19
 ; VGPRRC-NEXT:    s_nop 7
 ; VGPRRC-NEXT:    s_nop 0
-; VGPRRC-NEXT:    global_store_dwordx4 v[44:45], v[28:31], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[40:41], v[28:31], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v[46:47], v[24:27], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[42:43], v[24:27], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v[48:49], v[20:23], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[44:45], v[20:23], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v[50:51], v[16:19], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[46:47], v[16:19], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v[46:47], v[40:43], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[42:43], v[48:51], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    v_mov_b32_e32 v0, s20
 ; VGPRRC-NEXT:    v_mov_b32_e32 v1, s21
 ; VGPRRC-NEXT:    v_mov_b32_e32 v2, s22
 ; VGPRRC-NEXT:    v_mov_b32_e32 v3, s23
-; VGPRRC-NEXT:    global_store_dwordx4 v[44:45], v[0:3], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[40:41], v[0:3], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v0, s8
 ; VGPRRC-NEXT:    v_mov_b32_e32 v1, s9
 ; VGPRRC-NEXT:    v_mov_b32_e32 v2, s10
 ; VGPRRC-NEXT:    v_mov_b32_e32 v3, s11
-; VGPRRC-NEXT:    global_store_dwordx4 v[50:51], v[0:3], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[46:47], v[0:3], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v0, s12
 ; VGPRRC-NEXT:    v_mov_b32_e32 v1, s13
 ; VGPRRC-NEXT:    v_mov_b32_e32 v2, s14
 ; VGPRRC-NEXT:    v_mov_b32_e32 v3, s15
-; VGPRRC-NEXT:    global_store_dwordx4 v[48:49], v[0:3], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[44:45], v[0:3], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_endpgm
 ; AGPR-LABEL: test_mfma_f32_32x32x16_f16:
@@ -776,9 +776,9 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__flags(<8 x half> %arg0, <
 ; SDAG:       ; %bb.0:
 ; SDAG-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; SDAG-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; SDAG-NEXT:    v_mov_b64_e32 v[12:13], 48
-; SDAG-NEXT:    v_mov_b64_e32 v[14:15], 32
-; SDAG-NEXT:    v_mov_b64_e32 v[16:17], 16
+; SDAG-NEXT:    v_mov_b64_e32 v[8:9], 48
+; SDAG-NEXT:    v_mov_b64_e32 v[10:11], 32
+; SDAG-NEXT:    v_mov_b64_e32 v[12:13], 16
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
 ; SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[24:25]
 ; SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[26:27]
@@ -800,42 +800,42 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__flags(<8 x half> %arg0, <
 ; SDAG-NEXT:    v_accvgpr_write_b32 a13, s21
 ; SDAG-NEXT:    v_accvgpr_write_b32 a14, s22
 ; SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
-; SDAG-NEXT:    v_mov_b64_e32 v[18:19], 0
-; SDAG-NEXT:    v_mov_b32_e32 v8, s16
+; SDAG-NEXT:    v_mov_b64_e32 v[14:15], 0
+; SDAG-NEXT:    v_mov_b32_e32 v16, s16
 ; SDAG-NEXT:    v_mfma_f32_32x32x16_f16 a[16:31], v[0:3], v[4:7], a[0:15] cbsz:2 abid:3 blgp:1
 ; SDAG-NEXT:    v_mov_b32_e32 v0, s20
 ; SDAG-NEXT:    v_mov_b32_e32 v1, s21
 ; SDAG-NEXT:    v_mov_b32_e32 v2, s22
 ; SDAG-NEXT:    v_mov_b32_e32 v3, s23
-; SDAG-NEXT:    v_mov_b32_e32 v9, s17
-; SDAG-NEXT:    v_mov_b32_e32 v10, s18
-; SDAG-NEXT:    v_mov_b32_e32 v11, s19
+; SDAG-NEXT:    v_mov_b32_e32 v17, s17
+; SDAG-NEXT:    v_mov_b32_e32 v18, s18
+; SDAG-NEXT:    v_mov_b32_e32 v19, s19
 ; SDAG-NEXT:    s_nop 4
-; SDAG-NEXT:    global_store_dwordx4 v[12:13], a[28:31], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[8:9], a[28:31], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[14:15], a[24:27], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[10:11], a[24:27], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[16:17], a[20:23], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[12:13], a[20:23], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[18:19], a[16:19], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[14:15], a[16:19], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[14:15], v[8:11], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[10:11], v[16:19], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[12:13], v[0:3], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[8:9], v[0:3], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
 ; SDAG-NEXT:    v_mov_b32_e32 v0, s8
 ; SDAG-NEXT:    v_mov_b32_e32 v1, s9
 ; SDAG-NEXT:    v_mov_b32_e32 v2, s10
 ; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    global_store_dwordx4 v[18:19], v[0:3], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[14:15], v[0:3], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
 ; SDAG-NEXT:    v_mov_b32_e32 v0, s12
 ; SDAG-NEXT:    v_mov_b32_e32 v1, s13
 ; SDAG-NEXT:    v_mov_b32_e32 v2, s14
 ; SDAG-NEXT:    v_mov_b32_e32 v3, s15
-; SDAG-NEXT:    global_store_dwordx4 v[16:17], v[0:3], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[12:13], v[0:3], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_endpgm
 ;
@@ -900,9 +900,9 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__flags(<8 x half> %arg0, <
 ; HEURRC:       ; %bb.0:
 ; HEURRC-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; HEURRC-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; HEURRC-NEXT:    v_mov_b64_e32 v[12:13], 48
-; HEURRC-NEXT:    v_mov_b64_e32 v[14:15], 32
-; HEURRC-NEXT:    v_mov_b64_e32 v[16:17], 16
+; HEURRC-NEXT:    v_mov_b64_e32 v[8:9], 48
+; HEURRC-NEXT:    v_mov_b64_e32 v[10:11], 32
+; HEURRC-NEXT:    v_mov_b64_e32 v[12:13], 16
 ; HEURRC-NEXT:    s_waitcnt lgkmcnt(0)
 ; HEURRC-NEXT:    v_mov_b64_e32 v[0:1], s[24:25]
 ; HEURRC-NEXT:    v_mov_b64_e32 v[2:3], s[26:27]
@@ -924,42 +924,42 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__flags(<8 x half> %arg0, <
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a13, s21
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a14, s22
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a15, s23
-; HEURRC-NEXT:    v_mov_b64_e32 v[18:19], 0
-; HEURRC-NEXT:    v_mov_b32_e32 v8, s16
+; HEURRC-NEXT:    v_mov_b64_e32 v[14:15], 0
+; HEURRC-NEXT:    v_mov_b32_e32 v16, s16
 ; HEURRC-NEXT:    v_mfma_f32_32x32x16_f16 a[16:31], v[0:3], v[4:7], a[0:15] cbsz:2 abid:3 blgp:1
 ; HEURRC-NEXT:    v_mov_b32_e32 v0, s20
 ; HEURRC-NEXT:    v_mov_b32_e32 v1, s21
 ; HEURRC-NEXT:    v_mov_b32_e32 v2, s22
 ; HEURRC-NEXT:    v_mov_b32_e32 v3, s23
-; HEURRC-NEXT:    v_mov_b32_e32 v9, s17
-; HEURRC-NEXT:    v_mov_b32_e32 v10, s18
-; HEURRC-NEXT:    v_mov_b32_e32 v11, s19
+; HEURRC-NEXT:    v_mov_b32_e32 v17, s17
+; HEURRC-NEXT:    v_mov_b32_e32 v18, s18
+; HEURRC-NEXT:    v_mov_b32_e32 v19, s19
 ; HEURRC-NEXT:    s_nop 4
-; HEURRC-NEXT:    global_store_dwordx4 v[12:13], a[28:31], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[8:9], a[28:31], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v[14:15], a[24:27], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[10:11], a[24:27], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v[16:17], a[20:23], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[12:13], a[20:23], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v[18:19], a[16:19], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[14:15], a[16:19], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v[14:15], v[8:11], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[10:11], v[16:19], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v[12:13], v[0:3], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[8:9], v[0:3], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_nop 0
 ; HEURRC-NEXT:    v_mov_b32_e32 v0, s8
 ; HEURRC-NEXT:    v_mov_b32_e32 v1, s9
 ; HEURRC-NEXT:    v_mov_b32_e32 v2, s10
 ; HEURRC-NEXT:    v_mov_b32_e32 v3, s11
-; HEURRC-NEXT:    global_store_dwordx4 v[18:19], v[0:3], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[14:15], v[0:3], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_nop 0
 ; HEURRC-NEXT:    v_mov_b32_e32 v0, s12
 ; HEURRC-NEXT:    v_mov_b32_e32 v1, s13
 ; HEURRC-NEXT:    v_mov_b32_e32 v2, s14
 ; HEURRC-NEXT:    v_mov_b32_e32 v3, s15
-; HEURRC-NEXT:    global_store_dwordx4 v[16:17], v[0:3], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[12:13], v[0:3], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_endpgm
 ;
@@ -967,9 +967,9 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__flags(<8 x half> %arg0, <
 ; VGPRRC:       ; %bb.0:
 ; VGPRRC-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; VGPRRC-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; VGPRRC-NEXT:    v_mov_b64_e32 v[44:45], 48
-; VGPRRC-NEXT:    v_mov_b64_e32 v[46:47], 32
-; VGPRRC-NEXT:    v_mov_b64_e32 v[48:49], 16
+; VGPRRC-NEXT:    v_mov_b64_e32 v[40:41], 48
+; VGPRRC-NEXT:    v_mov_b64_e32 v[42:43], 32
+; VGPRRC-NEXT:    v_mov_b64_e32 v[44:45], 16
 ; VGPRRC-NEXT:    s_waitcnt lgkmcnt(0)
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[34:35], s[26:27]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[32:33], s[24:25]
@@ -983,43 +983,43 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__flags(<8 x half> %arg0, <
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[10:11], s[18:19]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[12:13], s[20:21]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[14:15], s[22:23]
-; VGPRRC-NEXT:    v_mov_b64_e32 v[50:51], 0
-; VGPRRC-NEXT:    v_mov_b32_e32 v40, s16
+; VGPRRC-NEXT:    v_mov_b64_e32 v[46:47], 0
+; VGPRRC-NEXT:    v_mov_b32_e32 v48, s16
 ; VGPRRC-NEXT:    v_mfma_f32_32x32x16_f16 v[16:31], v[32:35], v[36:39], v[0:15] cbsz:2 abid:3 blgp:1
-; VGPRRC-NEXT:    v_mov_b32_e32 v41, s17
-; VGPRRC-NEXT:    v_mov_b32_e32 v42, s18
-; VGPRRC-NEXT:    v_mov_b32_e32 v43, s19
+; VGPRRC-NEXT:    v_mov_b32_e32 v49, s17
+; VGPRRC-NEXT:    v_mov_b32_e32 v50, s18
+; VGPRRC-NEXT:    v_mov_b32_e32 v51, s19
 ; VGPRRC-NEXT:    s_nop 7
 ; VGPRRC-NEXT:    s_nop 0
-; VGPRRC-NEXT:    global_store_dwordx4 v[44:45], v[28:31], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[40:41], v[28:31], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v[46:47], v[24:27], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[42:43], v[24:27], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v[48:49], v[20:23], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[44:45], v[20:23], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v[50:51], v[16:19], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[46:47], v[16:19], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v[46:47], v[40:43], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[42:43], v[48:51], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    v_mov_b32_e32 v0, s20
 ; VGPRRC-NEXT:    v_mov_b32_e32 v1, s21
 ; VGPRRC-NEXT:    v_mov_b32_e32 v2, s22
 ; VGPRRC-NEXT:    v_mov_b32_e32 v3, s23
-; VGPRRC-NEXT:    global_store_dwordx4 v[44:45], v[0:3], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[40:41], v[0:3], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v0, s8
 ; VGPRRC-NEXT:    v_mov_b32_e32 v1, s9
 ; VGPRRC-NEXT:    v_mov_b32_e32 v2, s10
 ; VGPRRC-NEXT:    v_mov_b32_e32 v3, s11
-; VGPRRC-NEXT:    global_store_dwordx4 v[50:51], v[0:3], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[46:47], v[0:3], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v0, s12
 ; VGPRRC-NEXT:    v_mov_b32_e32 v1, s13
 ; VGPRRC-NEXT:    v_mov_b32_e32 v2, s14
 ; VGPRRC-NEXT:    v_mov_b32_e32 v3, s15
-; VGPRRC-NEXT:    global_store_dwordx4 v[48:49], v[0:3], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[44:45], v[0:3], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_endpgm
 ; AGPR-LABEL: test_mfma_f32_32x32x16_f16__flags:
@@ -1505,7 +1505,7 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd(<8 x half> %arg0,
 ; SDAG-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; SDAG-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
 ; SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0xa4
-; SDAG-NEXT:    v_mov_b32_e32 v12, 0
+; SDAG-NEXT:    v_mov_b32_e32 v8, 0
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
 ; SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[24:25]
 ; SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[26:27]
@@ -1527,40 +1527,40 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd(<8 x half> %arg0,
 ; SDAG-NEXT:    v_accvgpr_write_b32 a18, s10
 ; SDAG-NEXT:    v_accvgpr_write_b32 a17, s9
 ; SDAG-NEXT:    v_accvgpr_write_b32 a16, s8
-; SDAG-NEXT:    v_mov_b32_e32 v8, s20
-; SDAG-NEXT:    v_mov_b32_e32 v9, s21
+; SDAG-NEXT:    v_mov_b32_e32 v10, s20
+; SDAG-NEXT:    v_mov_b32_e32 v11, s21
 ; SDAG-NEXT:    v_mfma_f32_32x32x16_f16 a[0:15], v[0:3], v[4:7], a[16:31]
-; SDAG-NEXT:    v_mov_b32_e32 v10, s22
-; SDAG-NEXT:    v_mov_b32_e32 v11, s23
+; SDAG-NEXT:    v_mov_b32_e32 v12, s22
+; SDAG-NEXT:    v_mov_b32_e32 v13, s23
 ; SDAG-NEXT:    v_mov_b32_e32 v0, s16
 ; SDAG-NEXT:    v_mov_b32_e32 v1, s17
 ; SDAG-NEXT:    v_mov_b32_e32 v2, s18
 ; SDAG-NEXT:    v_mov_b32_e32 v3, s19
-; SDAG-NEXT:    global_store_dwordx4 v12, v[8:11], s[0:1] offset:48 sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v8, v[10:13], s[0:1] offset:48 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] offset:32 sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:32 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
 ; SDAG-NEXT:    v_mov_b32_e32 v0, s12
 ; SDAG-NEXT:    v_mov_b32_e32 v1, s13
 ; SDAG-NEXT:    v_mov_b32_e32 v2, s14
 ; SDAG-NEXT:    v_mov_b32_e32 v3, s15
-; SDAG-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] offset:16 sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:16 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
 ; SDAG-NEXT:    v_mov_b32_e32 v0, s8
 ; SDAG-NEXT:    v_mov_b32_e32 v1, s9
 ; SDAG-NEXT:    v_mov_b32_e32 v2, s10
 ; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v12, a[8:11], s[0:1] offset:32 sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v8, a[8:11], s[0:1] offset:32 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v12, a[12:15], s[0:1] offset:48 sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v8, a[12:15], s[0:1] offset:48 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v12, a[0:3], s[0:1] sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v8, a[0:3], s[0:1] sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v12, a[4:7], s[0:1] offset:16 sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v8, a[4:7], s[0:1] offset:16 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_endpgm
 ;
@@ -1623,7 +1623,7 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd(<8 x half> %arg0,
 ; HEURRC-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; HEURRC-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
 ; HEURRC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0xa4
-; HEURRC-NEXT:    v_mov_b32_e32 v12, 0
+; HEURRC-NEXT:    v_mov_b32_e32 v8, 0
 ; HEURRC-NEXT:    s_waitcnt lgkmcnt(0)
 ; HEURRC-NEXT:    v_mov_b64_e32 v[0:1], s[24:25]
 ; HEURRC-NEXT:    v_mov_b64_e32 v[2:3], s[26:27]
@@ -1645,40 +1645,40 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd(<8 x half> %arg0,
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a18, s10
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a17, s9
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a16, s8
-; HEURRC-NEXT:    v_mov_b32_e32 v8, s20
-; HEURRC-NEXT:    v_mov_b32_e32 v9, s21
+; HEURRC-NEXT:    v_mov_b32_e32 v10, s20
+; HEURRC-NEXT:    v_mov_b32_e32 v11, s21
 ; HEURRC-NEXT:    v_mfma_f32_32x32x16_f16 a[0:15], v[0:3], v[4:7], a[16:31]
-; HEURRC-NEXT:    v_mov_b32_e32 v10, s22
-; HEURRC-NEXT:    v_mov_b32_e32 v11, s23
+; HEURRC-NEXT:    v_mov_b32_e32 v12, s22
+; HEURRC-NEXT:    v_mov_b32_e32 v13, s23
 ; HEURRC-NEXT:    v_mov_b32_e32 v0, s16
 ; HEURRC-NEXT:    v_mov_b32_e32 v1, s17
 ; HEURRC-NEXT:    v_mov_b32_e32 v2, s18
 ; HEURRC-NEXT:    v_mov_b32_e32 v3, s19
-; HEURRC-NEXT:    global_store_dwordx4 v12, v[8:11], s[0:1] offset:48 sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v8, v[10:13], s[0:1] offset:48 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] offset:32 sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:32 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_nop 0
 ; HEURRC-NEXT:    v_mov_b32_e32 v0, s12
 ; HEURRC-NEXT:    v_mov_b32_e32 v1, s13
 ; HEURRC-NEXT:    v_mov_b32_e32 v2, s14
 ; HEURRC-NEXT:    v_mov_b32_e32 v3, s15
-; HEURRC-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] offset:16 sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:16 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_nop 0
 ; HEURRC-NEXT:    v_mov_b32_e32 v0, s8
 ; HEURRC-NEXT:    v_mov_b32_e32 v1, s9
 ; HEURRC-NEXT:    v_mov_b32_e32 v2, s10
 ; HEURRC-NEXT:    v_mov_b32_e32 v3, s11
-; HEURRC-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v12, a[8:11], s[0:1] offset:32 sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v8, a[8:11], s[0:1] offset:32 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v12, a[12:15], s[0:1] offset:48 sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v8, a[12:15], s[0:1] offset:48 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v12, a[0:3], s[0:1] sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v8, a[0:3], s[0:1] sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v12, a[4:7], s[0:1] offset:16 sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v8, a[4:7], s[0:1] offset:16 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_endpgm
 ;
@@ -1687,7 +1687,7 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd(<8 x half> %arg0,
 ; VGPRRC-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; VGPRRC-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
 ; VGPRRC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0xa4
-; VGPRRC-NEXT:    v_mov_b32_e32 v44, 0
+; VGPRRC-NEXT:    v_mov_b32_e32 v40, 0
 ; VGPRRC-NEXT:    s_waitcnt lgkmcnt(0)
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[34:35], s[26:27]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[32:33], s[24:25]
@@ -1701,41 +1701,41 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd(<8 x half> %arg0,
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[20:21], s[12:13]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[18:19], s[10:11]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[16:17], s[8:9]
-; VGPRRC-NEXT:    v_mov_b32_e32 v40, s20
-; VGPRRC-NEXT:    v_mov_b32_e32 v41, s21
+; VGPRRC-NEXT:    v_mov_b32_e32 v42, s20
+; VGPRRC-NEXT:    v_mov_b32_e32 v43, s21
 ; VGPRRC-NEXT:    v_mfma_f32_32x32x16_f16 v[0:15], v[32:35], v[36:39], v[16:31]
-; VGPRRC-NEXT:    v_mov_b32_e32 v42, s22
-; VGPRRC-NEXT:    v_mov_b32_e32 v43, s23
-; VGPRRC-NEXT:    global_store_dwordx4 v44, v[40:43], s[0:1] offset:48 sc0 sc1
+; VGPRRC-NEXT:    v_mov_b32_e32 v44, s22
+; VGPRRC-NEXT:    v_mov_b32_e32 v45, s23
+; VGPRRC-NEXT:    global_store_dwordx4 v40, v[42:45], s[0:1] offset:48 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 2
 ; VGPRRC-NEXT:    v_mov_b32_e32 v16, s16
 ; VGPRRC-NEXT:    v_mov_b32_e32 v17, s17
 ; VGPRRC-NEXT:    v_mov_b32_e32 v18, s18
 ; VGPRRC-NEXT:    v_mov_b32_e32 v19, s19
-; VGPRRC-NEXT:    global_store_dwordx4 v44, v[16:19], s[0:1] offset:32 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v40, v[16:19], s[0:1] offset:32 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v16, s12
 ; VGPRRC-NEXT:    v_mov_b32_e32 v17, s13
 ; VGPRRC-NEXT:    v_mov_b32_e32 v18, s14
 ; VGPRRC-NEXT:    v_mov_b32_e32 v19, s15
-; VGPRRC-NEXT:    global_store_dwordx4 v44, v[16:19], s[0:1] offset:16 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v40, v[16:19], s[0:1] offset:16 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v16, s8
 ; VGPRRC-NEXT:    v_mov_b32_e32 v17, s9
 ; VGPRRC-NEXT:    v_mov_b32_e32 v18, s10
 ; VGPRRC-NEXT:    v_mov_b32_e32 v19, s11
-; VGPRRC-NEXT:    global_store_dwordx4 v44, v[16:19], s[0:1] sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v40, v[16:19], s[0:1] sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v44, v[8:11], s[0:1] offset:32 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v40, v[8:11], s[0:1] offset:32 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v44, v[12:15], s[0:1] offset:48 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v40, v[12:15], s[0:1] offset:48 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v44, v[0:3], s[0:1] sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v40, v[0:3], s[0:1] sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v44, v[4:7], s[0:1] offset:16 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v40, v[4:7], s[0:1] offset:16 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_endpgm
 ; AGPR-LABEL: test_mfma_f32_32x32x16_f16__vgprcd:
@@ -1869,7 +1869,7 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd__flags(<8 x half>
 ; SDAG-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; SDAG-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
 ; SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0xa4
-; SDAG-NEXT:    v_mov_b32_e32 v12, 0
+; SDAG-NEXT:    v_mov_b32_e32 v8, 0
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
 ; SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[24:25]
 ; SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[26:27]
@@ -1891,40 +1891,40 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd__flags(<8 x half>
 ; SDAG-NEXT:    v_accvgpr_write_b32 a18, s10
 ; SDAG-NEXT:    v_accvgpr_write_b32 a17, s9
 ; SDAG-NEXT:    v_accvgpr_write_b32 a16, s8
-; SDAG-NEXT:    v_mov_b32_e32 v8, s20
-; SDAG-NEXT:    v_mov_b32_e32 v9, s21
+; SDAG-NEXT:    v_mov_b32_e32 v10, s20
+; SDAG-NEXT:    v_mov_b32_e32 v11, s21
 ; SDAG-NEXT:    v_mfma_f32_32x32x16_f16 a[0:15], v[0:3], v[4:7], a[16:31] cbsz:1 abid:2 blgp:3
-; SDAG-NEXT:    v_mov_b32_e32 v10, s22
-; SDAG-NEXT:    v_mov_b32_e32 v11, s23
+; SDAG-NEXT:    v_mov_b32_e32 v12, s22
+; SDAG-NEXT:    v_mov_b32_e32 v13, s23
 ; SDAG-NEXT:    v_mov_b32_e32 v0, s16
 ; SDAG-NEXT:    v_mov_b32_e32 v1, s17
 ; SDAG-NEXT:    v_mov_b32_e32 v2, s18
 ; SDAG-NEXT:    v_mov_b32_e32 v3, s19
-; SDAG-NEXT:    global_store_dwordx4 v12, v[8:11], s[0:1] offset:48 sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v8, v[10:13], s[0:1] offset:48 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] offset:32 sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:32 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
 ; SDAG-NEXT:    v_mov_b32_e32 v0, s12
 ; SDAG-NEXT:    v_mov_b32_e32 v1, s13
 ; SDAG-NEXT:    v_mov_b32_e32 v2, s14
 ; SDAG-NEXT:    v_mov_b32_e32 v3, s15
-; SDAG-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] offset:16 sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:16 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
 ; SDAG-NEXT:    v_mov_b32_e32 v0, s8
 ; SDAG-NEXT:    v_mov_b32_e32 v1, s9
 ; SDAG-NEXT:    v_mov_b32_e32 v2, s10
 ; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v12, a[8:11], s[0:1] offset:32 sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v8, a[8:11], s[0:1] offset:32 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v12, a[12:15], s[0:1] offset:48 sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v8, a[12:15], s[0:1] offset:48 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v12, a[0:3], s[0:1] sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v8, a[0:3], s[0:1] sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v12, a[4:7], s[0:1] offset:16 sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v8, a[4:7], s[0:1] offset:16 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_endpgm
 ;
@@ -1987,7 +1987,7 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd__flags(<8 x half>
 ; HEURRC-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; HEURRC-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
 ; HEURRC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0xa4
-; HEURRC-NEXT:    v_mov_b32_e32 v12, 0
+; HEURRC-NEXT:    v_mov_b32_e32 v8, 0
 ; HEURRC-NEXT:    s_waitcnt lgkmcnt(0)
 ; HEURRC-NEXT:    v_mov_b64_e32 v[0:1], s[24:25]
 ; HEURRC-NEXT:    v_mov_b64_e32 v[2:3], s[26:27]
@@ -2009,40 +2009,40 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd__flags(<8 x half>
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a18, s10
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a17, s9
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a16, s8
-; HEURRC-NEXT:    v_mov_b32_e32 v8, s20
-; HEURRC-NEXT:    v_mov_b32_e32 v9, s21
+; HEURRC-NEXT:    v_mov_b32_e32 v10, s20
+; HEURRC-NEXT:    v_mov_b32_e32 v11, s21
 ; HEURRC-NEXT:    v_mfma_f32_32x32x16_f16 a[0:15], v[0:3], v[4:7], a[16:31] cbsz:1 abid:2 blgp:3
-; HEURRC-NEXT:    v_mov_b32_e32 v10, s22
-; HEURRC-NEXT:    v_mov_b32_e32 v11, s23
+; HEURRC-NEXT:    v_mov_b32_e32 v12, s22
+; HEURRC-NEXT:    v_mov_b32_e32 v13, s23
 ; HEURRC-NEXT:    v_mov_b32_e32 v0, s16
 ; HEURRC-NEXT:    v_mov_b32_e32 v1, s17
 ; HEURRC-NEXT:    v_mov_b32_e32 v2, s18
 ; HEURRC-NEXT:    v_mov_b32_e32 v3, s19
-; HEURRC-NEXT:    global_store_dwordx4 v12, v[8:11], s[0:1] offset:48 sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v8, v[10:13], s[0:1] offset:48 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] offset:32 sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:32 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_nop 0
 ; HEURRC-NEXT:    v_mov_b32_e32 v0, s12
 ; HEURRC-NEXT:    v_mov_b32_e32 v1, s13
 ; HEURRC-NEXT:    v_mov_b32_e32 v2, s14
 ; HEURRC-NEXT:    v_mov_b32_e32 v3, s15
-; HEURRC-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] offset:16 sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:16 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_nop 0
 ; HEURRC-NEXT:    v_mov_b32_e32 v0, s8
 ; HEURRC-NEXT:    v_mov_b32_e32 v1, s9
 ; HEURRC-NEXT:    v_mov_b32_e32 v2, s10
 ; HEURRC-NEXT:    v_mov_b32_e32 v3, s11
-; HEURRC-NEXT:    global_store_dwordx4 v12, v[0:3], s[0:1] sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v12, a[8:11], s[0:1] offset:32 sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v8, a[8:11], s[0:1] offset:32 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v12, a[12:15], s[0:1] offset:48 sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v8, a[12:15], s[0:1] offset:48 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v12, a[0:3], s[0:1] sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v8, a[0:3], s[0:1] sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v12, a[4:7], s[0:1] offset:16 sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v8, a[4:7], s[0:1] offset:16 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_endpgm
 ;
@@ -2051,7 +2051,7 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd__flags(<8 x half>
 ; VGPRRC-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; VGPRRC-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
 ; VGPRRC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0xa4
-; VGPRRC-NEXT:    v_mov_b32_e32 v44, 0
+; VGPRRC-NEXT:    v_mov_b32_e32 v40, 0
 ; VGPRRC-NEXT:    s_waitcnt lgkmcnt(0)
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[34:35], s[26:27]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[32:33], s[24:25]
@@ -2065,41 +2065,41 @@ define amdgpu_kernel void @test_mfma_f32_32x32x16_f16__vgprcd__flags(<8 x half>
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[20:21], s[12:13]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[18:19], s[10:11]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[16:17], s[8:9]
-; VGPRRC-NEXT:    v_mov_b32_e32 v40, s20
-; VGPRRC-NEXT:    v_mov_b32_e32 v41, s21
+; VGPRRC-NEXT:    v_mov_b32_e32 v42, s20
+; VGPRRC-NEXT:    v_mov_b32_e32 v43, s21
 ; VGPRRC-NEXT:    v_mfma_f32_32x32x16_f16 v[0:15], v[32:35], v[36:39], v[16:31] cbsz:1 abid:2 blgp:3
-; VGPRRC-NEXT:    v_mov_b32_e32 v42, s22
-; VGPRRC-NEXT:    v_mov_b32_e32 v43, s23
-; VGPRRC-NEXT:    global_store_dwordx4 v44, v[40:43], s[0:1] offset:48 sc0 sc1
+; VGPRRC-NEXT:    v_mov_b32_e32 v44, s22
+; VGPRRC-NEXT:    v_mov_b32_e32 v45, s23
+; VGPRRC-NEXT:    global_store_dwordx4 v40, v[42:45], s[0:1] offset:48 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 2
 ; VGPRRC-NEXT:    v_mov_b32_e32 v16, s16
 ; VGPRRC-NEXT:    v_mov_b32_e32 v17, s17
 ; VGPRRC-NEXT:    v_mov_b32_e32 v18, s18
 ; VGPRRC-NEXT:    v_mov_b32_e32 v19, s19
-; VGPRRC-NEXT:    global_store_dwordx4 v44, v[16:19], s[0:1] offset:32 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v40, v[16:19], s[0:1] offset:32 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v16, s12
 ; VGPRRC-NEXT:    v_mov_b32_e32 v17, s13
 ; VGPRRC-NEXT:    v_mov_b32_e32 v18, s14
 ; VGPRRC-NEXT:    v_mov_b32_e32 v19, s15
-; VGPRRC-NEXT:    global_store_dwordx4 v44, v[16:19], s[0:1] offset:16 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v40, v[16:19], s[0:1] offset:16 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v16, s8
 ; VGPRRC-NEXT:    v_mov_b32_e32 v17, s9
 ; VGPRRC-NEXT:    v_mov_b32_e32 v18, s10
 ; VGPRRC-NEXT:    v_mov_b32_e32 v19, s11
-; VGPRRC-NEXT:    global_store_dwordx4 v44, v[16:19], s[0:1] sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v40, v[16:19], s[0:1] sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v44, v[8:11], s[0:1] offset:32 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v40, v[8:11], s[0:1] offset:32 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v44, v[12:15], s[0:1] offset:48 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v40, v[12:15], s[0:1] offset:48 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v44, v[0:3], s[0:1] sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v40, v[0:3], s[0:1] sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v44, v[4:7], s[0:1] offset:16 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v40, v[4:7], s[0:1] offset:16 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_endpgm
 ; AGPR-LABEL: test_mfma_f32_32x32x16_f16__vgprcd__flags:
@@ -2781,24 +2781,24 @@ define amdgpu_kernel void @test_mfma_i32_16x16x64_i8_no_agpr__vgprcd(ptr addrspa
 ; SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x34
 ; SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
 ; SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x54
-; SDAG-NEXT:    v_mov_b32_e32 v8, 0
+; SDAG-NEXT:    v_mov_b32_e32 v0, 0
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v0, s8
-; SDAG-NEXT:    v_mov_b32_e32 v1, s9
-; SDAG-NEXT:    v_mov_b32_e32 v2, s10
-; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    v_mov_b32_e32 v4, s12
-; SDAG-NEXT:    v_mov_b32_e32 v5, s13
-; SDAG-NEXT:    v_mov_b32_e32 v6, s14
-; SDAG-NEXT:    v_mov_b32_e32 v7, s15
+; SDAG-NEXT:    v_mov_b32_e32 v2, s8
+; SDAG-NEXT:    v_mov_b32_e32 v3, s9
+; SDAG-NEXT:    v_mov_b32_e32 v4, s10
+; SDAG-NEXT:    v_mov_b32_e32 v5, s11
+; SDAG-NEXT:    v_mov_b32_e32 v6, s12
+; SDAG-NEXT:    v_mov_b32_e32 v7, s13
+; SDAG-NEXT:    v_mov_b32_e32 v8, s14
+; SDAG-NEXT:    v_mov_b32_e32 v9, s15
 ; SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
 ; SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
 ; SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
 ; SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_mfma_i32_16x16x64_i8 a[0:3], v[0:3], v[4:7], a[0:3]
+; SDAG-NEXT:    v_mfma_i32_16x16x64_i8 a[0:3], v[2:5], v[6:9], a[0:3]
 ; SDAG-NEXT:    s_nop 7
-; SDAG-NEXT:    global_store_dwordx4 v8, a[0:3], s[6:7]
+; SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
 ; SDAG-NEXT:    s_endpgm
 ;
 ; GISEL-LABEL: test_mfma_i32_16x16x64_i8_no_agpr__vgprcd:
@@ -2827,24 +2827,24 @@ define amdgpu_kernel void @test_mfma_i32_16x16x64_i8_no_agpr__vgprcd(ptr addrspa
 ; HEURRC-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x34
 ; HEURRC-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
 ; HEURRC-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x54
-; HEURRC-NEXT:    v_mov_b32_e32 v8, 0
+; HEURRC-NEXT:    v_mov_b32_e32 v0, 0
 ; HEURRC-NEXT:    s_waitcnt lgkmcnt(0)
-; HEURRC-NEXT:    v_mov_b32_e32 v0, s8
-; HEURRC-NEXT:    v_mov_b32_e32 v1, s9
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s10
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s11
-; HEURRC-NEXT:    v_mov_b32_e32 v4, s12
-; HEURRC-NEXT:    v_mov_b32_e32 v5, s13
-; HEURRC-NEXT:    v_mov_b32_e32 v6, s14
-; HEURRC-NEXT:    v_mov_b32_e32 v7, s15
+; HEURRC-NEXT:    v_mov_b32_e32 v2, s8
+; HEURRC-NEXT:    v_mov_b32_e32 v3, s9
+; HEURRC-NEXT:    v_mov_b32_e32 v4, s10
+; HEURRC-NEXT:    v_mov_b32_e32 v5, s11
+; HEURRC-NEXT:    v_mov_b32_e32 v6, s12
+; HEURRC-NEXT:    v_mov_b32_e32 v7, s13
+; HEURRC-NEXT:    v_mov_b32_e32 v8, s14
+; HEURRC-NEXT:    v_mov_b32_e32 v9, s15
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a0, s0
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a1, s1
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a2, s2
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a3, s3
 ; HEURRC-NEXT:    s_nop 1
-; HEURRC-NEXT:    v_mfma_i32_16x16x64_i8 a[0:3], v[0:3], v[4:7], a[0:3]
+; HEURRC-NEXT:    v_mfma_i32_16x16x64_i8 a[0:3], v[2:5], v[6:9], a[0:3]
 ; HEURRC-NEXT:    s_nop 7
-; HEURRC-NEXT:    global_store_dwordx4 v8, a[0:3], s[6:7]
+; HEURRC-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
 ; HEURRC-NEXT:    s_endpgm
 ;
 ; VGPRRC-LABEL: test_mfma_i32_16x16x64_i8_no_agpr__vgprcd:
@@ -2852,24 +2852,24 @@ define amdgpu_kernel void @test_mfma_i32_16x16x64_i8_no_agpr__vgprcd(ptr addrspa
 ; VGPRRC-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x34
 ; VGPRRC-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
 ; VGPRRC-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x54
-; VGPRRC-NEXT:    v_mov_b32_e32 v12, 0
+; VGPRRC-NEXT:    v_mov_b32_e32 v4, 0
 ; VGPRRC-NEXT:    s_waitcnt lgkmcnt(0)
-; VGPRRC-NEXT:    v_mov_b32_e32 v0, s8
-; VGPRRC-NEXT:    v_mov_b32_e32 v1, s9
-; VGPRRC-NEXT:    v_mov_b32_e32 v2, s10
-; VGPRRC-NEXT:    v_mov_b32_e32 v3, s11
-; VGPRRC-NEXT:    v_mov_b32_e32 v4, s12
-; VGPRRC-NEXT:    v_mov_b32_e32 v5, s13
-; VGPRRC-NEXT:    v_mov_b32_e32 v6, s14
-; VGPRRC-NEXT:    v_mov_b32_e32 v7, s15
-; VGPRRC-NEXT:    v_mov_b32_e32 v8, s0
-; VGPRRC-NEXT:    v_mov_b32_e32 v9, s1
-; VGPRRC-NEXT:    v_mov_b32_e32 v10, s2
-; VGPRRC-NEXT:    v_mov_b32_e32 v11, s3
+; VGPRRC-NEXT:    v_mov_b32_e32 v6, s8
+; VGPRRC-NEXT:    v_mov_b32_e32 v7, s9
+; VGPRRC-NEXT:    v_mov_b32_e32 v8, s10
+; VGPRRC-NEXT:    v_mov_b32_e32 v9, s11
+; VGPRRC-NEXT:    v_mov_b32_e32 v10, s12
+; VGPRRC-NEXT:    v_mov_b32_e32 v11, s13
+; VGPRRC-NEXT:    v_mov_b32_e32 v12, s14
+; VGPRRC-NEXT:    v_mov_b32_e32 v13, s15
+; VGPRRC-NEXT:    v_mov_b32_e32 v0, s0
+; VGPRRC-NEXT:    v_mov_b32_e32 v1, s1
+; VGPRRC-NEXT:    v_mov_b32_e32 v2, s2
+; VGPRRC-NEXT:    v_mov_b32_e32 v3, s3
 ; VGPRRC-NEXT:    s_nop 1
-; VGPRRC-NEXT:    v_mfma_i32_16x16x64_i8 v[0:3], v[0:3], v[4:7], v[8:11]
+; VGPRRC-NEXT:    v_mfma_i32_16x16x64_i8 v[0:3], v[6:9], v[10:13], v[0:3]
 ; VGPRRC-NEXT:    s_nop 7
-; VGPRRC-NEXT:    global_store_dwordx4 v12, v[0:3], s[6:7]
+; VGPRRC-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
 ; VGPRRC-NEXT:    s_endpgm
 ; AGPR-LABEL: test_mfma_i32_16x16x64_i8_no_agpr__vgprcd:
 ; AGPR:       ; %bb.0:
@@ -2930,24 +2930,24 @@ define amdgpu_kernel void @test_mfma_i32_16x16x64_i8_no_agpr__vgprcd__flags(ptr
 ; SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x34
 ; SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
 ; SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x54
-; SDAG-NEXT:    v_mov_b32_e32 v8, 0
+; SDAG-NEXT:    v_mov_b32_e32 v0, 0
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v0, s8
-; SDAG-NEXT:    v_mov_b32_e32 v1, s9
-; SDAG-NEXT:    v_mov_b32_e32 v2, s10
-; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    v_mov_b32_e32 v4, s12
-; SDAG-NEXT:    v_mov_b32_e32 v5, s13
-; SDAG-NEXT:    v_mov_b32_e32 v6, s14
-; SDAG-NEXT:    v_mov_b32_e32 v7, s15
+; SDAG-NEXT:    v_mov_b32_e32 v2, s8
+; SDAG-NEXT:    v_mov_b32_e32 v3, s9
+; SDAG-NEXT:    v_mov_b32_e32 v4, s10
+; SDAG-NEXT:    v_mov_b32_e32 v5, s11
+; SDAG-NEXT:    v_mov_b32_e32 v6, s12
+; SDAG-NEXT:    v_mov_b32_e32 v7, s13
+; SDAG-NEXT:    v_mov_b32_e32 v8, s14
+; SDAG-NEXT:    v_mov_b32_e32 v9, s15
 ; SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
 ; SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
 ; SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
 ; SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_mfma_i32_16x16x64_i8 a[0:3], v[0:3], v[4:7], a[0:3] cbsz:3 abid:2 blgp:1
+; SDAG-NEXT:    v_mfma_i32_16x16x64_i8 a[0:3], v[2:5], v[6:9], a[0:3] cbsz:3 abid:2 blgp:1
 ; SDAG-NEXT:    s_nop 7
-; SDAG-NEXT:    global_store_dwordx4 v8, a[0:3], s[6:7]
+; SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
 ; SDAG-NEXT:    s_endpgm
 ;
 ; GISEL-LABEL: test_mfma_i32_16x16x64_i8_no_agpr__vgprcd__flags:
@@ -2976,24 +2976,24 @@ define amdgpu_kernel void @test_mfma_i32_16x16x64_i8_no_agpr__vgprcd__flags(ptr
 ; HEURRC-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x34
 ; HEURRC-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
 ; HEURRC-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x54
-; HEURRC-NEXT:    v_mov_b32_e32 v8, 0
+; HEURRC-NEXT:    v_mov_b32_e32 v0, 0
 ; HEURRC-NEXT:    s_waitcnt lgkmcnt(0)
-; HEURRC-NEXT:    v_mov_b32_e32 v0, s8
-; HEURRC-NEXT:    v_mov_b32_e32 v1, s9
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s10
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s11
-; HEURRC-NEXT:    v_mov_b32_e32 v4, s12
-; HEURRC-NEXT:    v_mov_b32_e32 v5, s13
-; HEURRC-NEXT:    v_mov_b32_e32 v6, s14
-; HEURRC-NEXT:    v_mov_b32_e32 v7, s15
+; HEURRC-NEXT:    v_mov_b32_e32 v2, s8
+; HEURRC-NEXT:    v_mov_b32_e32 v3, s9
+; HEURRC-NEXT:    v_mov_b32_e32 v4, s10
+; HEURRC-NEXT:    v_mov_b32_e32 v5, s11
+; HEURRC-NEXT:    v_mov_b32_e32 v6, s12
+; HEURRC-NEXT:    v_mov_b32_e32 v7, s13
+; HEURRC-NEXT:    v_mov_b32_e32 v8, s14
+; HEURRC-NEXT:    v_mov_b32_e32 v9, s15
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a0, s0
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a1, s1
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a2, s2
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a3, s3
 ; HEURRC-NEXT:    s_nop 1
-; HEURRC-NEXT:    v_mfma_i32_16x16x64_i8 a[0:3], v[0:3], v[4:7], a[0:3] cbsz:3 abid:2 blgp:1
+; HEURRC-NEXT:    v_mfma_i32_16x16x64_i8 a[0:3], v[2:5], v[6:9], a[0:3] cbsz:3 abid:2 blgp:1
 ; HEURRC-NEXT:    s_nop 7
-; HEURRC-NEXT:    global_store_dwordx4 v8, a[0:3], s[6:7]
+; HEURRC-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
 ; HEURRC-NEXT:    s_endpgm
 ;
 ; VGPRRC-LABEL: test_mfma_i32_16x16x64_i8_no_agpr__vgprcd__flags:
@@ -3001,24 +3001,24 @@ define amdgpu_kernel void @test_mfma_i32_16x16x64_i8_no_agpr__vgprcd__flags(ptr
 ; VGPRRC-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x34
 ; VGPRRC-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
 ; VGPRRC-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x54
-; VGPRRC-NEXT:    v_mov_b32_e32 v12, 0
+; VGPRRC-NEXT:    v_mov_b32_e32 v4, 0
 ; VGPRRC-NEXT:    s_waitcnt lgkmcnt(0)
-; VGPRRC-NEXT:    v_mov_b32_e32 v0, s8
-; VGPRRC-NEXT:    v_mov_b32_e32 v1, s9
-; VGPRRC-NEXT:    v_mov_b32_e32 v2, s10
-; VGPRRC-NEXT:    v_mov_b32_e32 v3, s11
-; VGPRRC-NEXT:    v_mov_b32_e32 v4, s12
-; VGPRRC-NEXT:    v_mov_b32_e32 v5, s13
-; VGPRRC-NEXT:    v_mov_b32_e32 v6, s14
-; VGPRRC-NEXT:    v_mov_b32_e32 v7, s15
-; VGPRRC-NEXT:    v_mov_b32_e32 v8, s0
-; VGPRRC-NEXT:    v_mov_b32_e32 v9, s1
-; VGPRRC-NEXT:    v_mov_b32_e32 v10, s2
-; VGPRRC-NEXT:    v_mov_b32_e32 v11, s3
+; VGPRRC-NEXT:    v_mov_b32_e32 v6, s8
+; VGPRRC-NEXT:    v_mov_b32_e32 v7, s9
+; VGPRRC-NEXT:    v_mov_b32_e32 v8, s10
+; VGPRRC-NEXT:    v_mov_b32_e32 v9, s11
+; VGPRRC-NEXT:    v_mov_b32_e32 v10, s12
+; VGPRRC-NEXT:    v_mov_b32_e32 v11, s13
+; VGPRRC-NEXT:    v_mov_b32_e32 v12, s14
+; VGPRRC-NEXT:    v_mov_b32_e32 v13, s15
+; VGPRRC-NEXT:    v_mov_b32_e32 v0, s0
+; VGPRRC-NEXT:    v_mov_b32_e32 v1, s1
+; VGPRRC-NEXT:    v_mov_b32_e32 v2, s2
+; VGPRRC-NEXT:    v_mov_b32_e32 v3, s3
 ; VGPRRC-NEXT:    s_nop 1
-; VGPRRC-NEXT:    v_mfma_i32_16x16x64_i8 v[0:3], v[0:3], v[4:7], v[8:11] cbsz:3 abid:2 blgp:1
+; VGPRRC-NEXT:    v_mfma_i32_16x16x64_i8 v[0:3], v[6:9], v[10:13], v[0:3] cbsz:3 abid:2 blgp:1
 ; VGPRRC-NEXT:    s_nop 7
-; VGPRRC-NEXT:    global_store_dwordx4 v12, v[0:3], s[6:7]
+; VGPRRC-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
 ; VGPRRC-NEXT:    s_endpgm
 ; AGPR-LABEL: test_mfma_i32_16x16x64_i8_no_agpr__vgprcd__flags:
 ; AGPR:       ; %bb.0:
@@ -3084,19 +3084,19 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8(<4 x i32> %arg0, <4 x i32>
 ; SDAG:       ; %bb.0:
 ; SDAG-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; SDAG-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; SDAG-NEXT:    v_mov_b64_e32 v[8:9], 48
-; SDAG-NEXT:    v_mov_b64_e32 v[10:11], 32
-; SDAG-NEXT:    v_mov_b64_e32 v[12:13], 16
+; SDAG-NEXT:    v_mov_b64_e32 v[0:1], 48
+; SDAG-NEXT:    v_mov_b64_e32 v[2:3], 32
+; SDAG-NEXT:    v_mov_b64_e32 v[4:5], 16
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v0, s24
-; SDAG-NEXT:    v_mov_b32_e32 v1, s25
-; SDAG-NEXT:    v_mov_b32_e32 v2, s26
-; SDAG-NEXT:    v_mov_b32_e32 v3, s27
+; SDAG-NEXT:    v_mov_b32_e32 v8, s24
+; SDAG-NEXT:    v_mov_b32_e32 v9, s25
+; SDAG-NEXT:    v_mov_b32_e32 v10, s26
+; SDAG-NEXT:    v_mov_b32_e32 v11, s27
 ; SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
-; SDAG-NEXT:    v_mov_b32_e32 v4, s28
-; SDAG-NEXT:    v_mov_b32_e32 v5, s29
-; SDAG-NEXT:    v_mov_b32_e32 v6, s30
-; SDAG-NEXT:    v_mov_b32_e32 v7, s31
+; SDAG-NEXT:    v_mov_b32_e32 v12, s28
+; SDAG-NEXT:    v_mov_b32_e32 v13, s29
+; SDAG-NEXT:    v_mov_b32_e32 v14, s30
+; SDAG-NEXT:    v_mov_b32_e32 v15, s31
 ; SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
 ; SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
 ; SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
@@ -3112,44 +3112,42 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8(<4 x i32> %arg0, <4 x i32>
 ; SDAG-NEXT:    v_accvgpr_write_b32 a13, s21
 ; SDAG-NEXT:    v_accvgpr_write_b32 a14, s22
 ; SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
-; SDAG-NEXT:    v_mov_b64_e32 v[14:15], 0
+; SDAG-NEXT:    v_mov_b64_e32 v[6:7], 0
 ; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_mfma_i32_32x32x32_i8 a[16:31], v[0:3], v[4:7], a[0:15]
-; SDAG-NEXT:    v_mov_b32_e32 v0, s16
-; SDAG-NEXT:    v_mov_b32_e32 v1, s17
-; SDAG-NEXT:    v_mov_b32_e32 v2, s18
-; SDAG-NEXT:    v_mov_b32_e32 v3, s19
+; SDAG-NEXT:    v_mfma_i32_32x32x32_i8 a[16:31], v[8:11], v[12:15], a[0:15]
+; SDAG-NEXT:    v_mov_b32_e32 v8, s16
+; SDAG-NEXT:    v_mov_b32_e32 v9, s17
+; SDAG-NEXT:    v_mov_b32_e32 v10, s18
+; SDAG-NEXT:    v_mov_b32_e32 v11, s19
 ; SDAG-NEXT:    s_nop 7
-; SDAG-NEXT:    global_store_dwordx4 v[8:9], a[28:31], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[0:1], a[28:31], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[10:11], a[24:27], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[2:3], a[24:27], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[12:13], a[20:23], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[4:5], a[20:23], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[14:15], a[16:19], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[6:7], a[16:19], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[10:11], v[0:3], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[2:3], v[8:11], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_mov_b32_e32 v0, s20
-; SDAG-NEXT:    v_mov_b32_e32 v1, s21
-; SDAG-NEXT:    v_mov_b32_e32 v2, s22
-; SDAG-NEXT:    v_mov_b32_e32 v3, s23
-; SDAG-NEXT:    global_store_dwordx4 v[8:9], v[0:3], off sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v2, s10
+; SDAG-NEXT:    v_mov_b32_e32 v3, s11
+; SDAG-NEXT:    v_mov_b32_e32 v8, s20
+; SDAG-NEXT:    v_mov_b32_e32 v9, s21
+; SDAG-NEXT:    v_mov_b32_e32 v10, s22
+; SDAG-NEXT:    v_mov_b32_e32 v11, s23
+; SDAG-NEXT:    global_store_dwordx4 v[0:1], v[8:11], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    s_nop 0
 ; SDAG-NEXT:    v_mov_b32_e32 v0, s8
 ; SDAG-NEXT:    v_mov_b32_e32 v1, s9
-; SDAG-NEXT:    v_mov_b32_e32 v2, s10
-; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    global_store_dwordx4 v[14:15], v[0:3], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[6:7], v[0:3], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
 ; SDAG-NEXT:    v_mov_b32_e32 v0, s12
 ; SDAG-NEXT:    v_mov_b32_e32 v1, s13
 ; SDAG-NEXT:    v_mov_b32_e32 v2, s14
 ; SDAG-NEXT:    v_mov_b32_e32 v3, s15
-; SDAG-NEXT:    global_store_dwordx4 v[12:13], v[0:3], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[4:5], v[0:3], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_endpgm
 ;
@@ -3214,19 +3212,19 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8(<4 x i32> %arg0, <4 x i32>
 ; HEURRC:       ; %bb.0:
 ; HEURRC-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; HEURRC-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; HEURRC-NEXT:    v_mov_b64_e32 v[8:9], 48
-; HEURRC-NEXT:    v_mov_b64_e32 v[10:11], 32
-; HEURRC-NEXT:    v_mov_b64_e32 v[12:13], 16
+; HEURRC-NEXT:    v_mov_b64_e32 v[0:1], 48
+; HEURRC-NEXT:    v_mov_b64_e32 v[2:3], 32
+; HEURRC-NEXT:    v_mov_b64_e32 v[4:5], 16
 ; HEURRC-NEXT:    s_waitcnt lgkmcnt(0)
-; HEURRC-NEXT:    v_mov_b32_e32 v0, s24
-; HEURRC-NEXT:    v_mov_b32_e32 v1, s25
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s26
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s27
+; HEURRC-NEXT:    v_mov_b32_e32 v8, s24
+; HEURRC-NEXT:    v_mov_b32_e32 v9, s25
+; HEURRC-NEXT:    v_mov_b32_e32 v10, s26
+; HEURRC-NEXT:    v_mov_b32_e32 v11, s27
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a0, s8
-; HEURRC-NEXT:    v_mov_b32_e32 v4, s28
-; HEURRC-NEXT:    v_mov_b32_e32 v5, s29
-; HEURRC-NEXT:    v_mov_b32_e32 v6, s30
-; HEURRC-NEXT:    v_mov_b32_e32 v7, s31
+; HEURRC-NEXT:    v_mov_b32_e32 v12, s28
+; HEURRC-NEXT:    v_mov_b32_e32 v13, s29
+; HEURRC-NEXT:    v_mov_b32_e32 v14, s30
+; HEURRC-NEXT:    v_mov_b32_e32 v15, s31
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a1, s9
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a2, s10
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a3, s11
@@ -3242,44 +3240,42 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8(<4 x i32> %arg0, <4 x i32>
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a13, s21
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a14, s22
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a15, s23
-; HEURRC-NEXT:    v_mov_b64_e32 v[14:15], 0
+; HEURRC-NEXT:    v_mov_b64_e32 v[6:7], 0
 ; HEURRC-NEXT:    s_nop 0
-; HEURRC-NEXT:    v_mfma_i32_32x32x32_i8 a[16:31], v[0:3], v[4:7], a[0:15]
-; HEURRC-NEXT:    v_mov_b32_e32 v0, s16
-; HEURRC-NEXT:    v_mov_b32_e32 v1, s17
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s18
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s19
+; HEURRC-NEXT:    v_mfma_i32_32x32x32_i8 a[16:31], v[8:11], v[12:15], a[0:15]
+; HEURRC-NEXT:    v_mov_b32_e32 v8, s16
+; HEURRC-NEXT:    v_mov_b32_e32 v9, s17
+; HEURRC-NEXT:    v_mov_b32_e32 v10, s18
+; HEURRC-NEXT:    v_mov_b32_e32 v11, s19
 ; HEURRC-NEXT:    s_nop 7
-; HEURRC-NEXT:    global_store_dwordx4 v[8:9], a[28:31], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[0:1], a[28:31], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v[10:11], a[24:27], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[2:3], a[24:27], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v[12:13], a[20:23], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[4:5], a[20:23], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v[14:15], a[16:19], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[6:7], a[16:19], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v[10:11], v[0:3], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[2:3], v[8:11], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    s_nop 0
-; HEURRC-NEXT:    v_mov_b32_e32 v0, s20
-; HEURRC-NEXT:    v_mov_b32_e32 v1, s21
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s22
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s23
-; HEURRC-NEXT:    global_store_dwordx4 v[8:9], v[0:3], off sc0 sc1
+; HEURRC-NEXT:    v_mov_b32_e32 v2, s10
+; HEURRC-NEXT:    v_mov_b32_e32 v3, s11
+; HEURRC-NEXT:    v_mov_b32_e32 v8, s20
+; HEURRC-NEXT:    v_mov_b32_e32 v9, s21
+; HEURRC-NEXT:    v_mov_b32_e32 v10, s22
+; HEURRC-NEXT:    v_mov_b32_e32 v11, s23
+; HEURRC-NEXT:    global_store_dwordx4 v[0:1], v[8:11], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    s_nop 0
 ; HEURRC-NEXT:    v_mov_b32_e32 v0, s8
 ; HEURRC-NEXT:    v_mov_b32_e32 v1, s9
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s10
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s11
-; HEURRC-NEXT:    global_store_dwordx4 v[14:15], v[0:3], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[6:7], v[0:3], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_nop 0
 ; HEURRC-NEXT:    v_mov_b32_e32 v0, s12
 ; HEURRC-NEXT:    v_mov_b32_e32 v1, s13
 ; HEURRC-NEXT:    v_mov_b32_e32 v2, s14
 ; HEURRC-NEXT:    v_mov_b32_e32 v3, s15
-; HEURRC-NEXT:    global_store_dwordx4 v[12:13], v[0:3], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[4:5], v[0:3], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_endpgm
 ;
@@ -3287,19 +3283,19 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8(<4 x i32> %arg0, <4 x i32>
 ; VGPRRC:       ; %bb.0:
 ; VGPRRC-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; VGPRRC-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; VGPRRC-NEXT:    v_mov_b64_e32 v[40:41], 48
-; VGPRRC-NEXT:    v_mov_b64_e32 v[42:43], 32
-; VGPRRC-NEXT:    v_mov_b64_e32 v[44:45], 16
+; VGPRRC-NEXT:    v_mov_b64_e32 v[32:33], 48
+; VGPRRC-NEXT:    v_mov_b64_e32 v[34:35], 32
+; VGPRRC-NEXT:    v_mov_b64_e32 v[36:37], 16
 ; VGPRRC-NEXT:    s_waitcnt lgkmcnt(0)
-; VGPRRC-NEXT:    v_mov_b32_e32 v32, s24
-; VGPRRC-NEXT:    v_mov_b32_e32 v33, s25
-; VGPRRC-NEXT:    v_mov_b32_e32 v34, s26
-; VGPRRC-NEXT:    v_mov_b32_e32 v35, s27
+; VGPRRC-NEXT:    v_mov_b32_e32 v40, s24
+; VGPRRC-NEXT:    v_mov_b32_e32 v41, s25
+; VGPRRC-NEXT:    v_mov_b32_e32 v42, s26
+; VGPRRC-NEXT:    v_mov_b32_e32 v43, s27
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
-; VGPRRC-NEXT:    v_mov_b32_e32 v36, s28
-; VGPRRC-NEXT:    v_mov_b32_e32 v37, s29
-; VGPRRC-NEXT:    v_mov_b32_e32 v38, s30
-; VGPRRC-NEXT:    v_mov_b32_e32 v39, s31
+; VGPRRC-NEXT:    v_mov_b32_e32 v44, s28
+; VGPRRC-NEXT:    v_mov_b32_e32 v45, s29
+; VGPRRC-NEXT:    v_mov_b32_e32 v46, s30
+; VGPRRC-NEXT:    v_mov_b32_e32 v47, s31
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[4:5], s[12:13]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[6:7], s[14:15]
@@ -3307,45 +3303,45 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8(<4 x i32> %arg0, <4 x i32>
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[10:11], s[18:19]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[12:13], s[20:21]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[14:15], s[22:23]
-; VGPRRC-NEXT:    v_mov_b64_e32 v[46:47], 0
+; VGPRRC-NEXT:    v_mov_b64_e32 v[38:39], 0
 ; VGPRRC-NEXT:    s_nop 0
-; VGPRRC-NEXT:    v_mfma_i32_32x32x32_i8 v[16:31], v[32:35], v[36:39], v[0:15]
+; VGPRRC-NEXT:    v_mfma_i32_32x32x32_i8 v[16:31], v[40:43], v[44:47], v[0:15]
 ; VGPRRC-NEXT:    s_nop 7
 ; VGPRRC-NEXT:    s_nop 3
-; VGPRRC-NEXT:    global_store_dwordx4 v[40:41], v[28:31], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[32:33], v[28:31], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v[42:43], v[24:27], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[34:35], v[24:27], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v[44:45], v[20:23], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[36:37], v[20:23], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v[46:47], v[16:19], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[38:39], v[16:19], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    v_mov_b32_e32 v0, s16
 ; VGPRRC-NEXT:    v_mov_b32_e32 v1, s17
 ; VGPRRC-NEXT:    v_mov_b32_e32 v2, s18
 ; VGPRRC-NEXT:    v_mov_b32_e32 v3, s19
-; VGPRRC-NEXT:    global_store_dwordx4 v[42:43], v[0:3], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[34:35], v[0:3], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v0, s20
 ; VGPRRC-NEXT:    v_mov_b32_e32 v1, s21
 ; VGPRRC-NEXT:    v_mov_b32_e32 v2, s22
 ; VGPRRC-NEXT:    v_mov_b32_e32 v3, s23
-; VGPRRC-NEXT:    global_store_dwordx4 v[40:41], v[0:3], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[32:33], v[0:3], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v0, s8
 ; VGPRRC-NEXT:    v_mov_b32_e32 v1, s9
 ; VGPRRC-NEXT:    v_mov_b32_e32 v2, s10
 ; VGPRRC-NEXT:    v_mov_b32_e32 v3, s11
-; VGPRRC-NEXT:    global_store_dwordx4 v[46:47], v[0:3], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[38:39], v[0:3], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v0, s12
 ; VGPRRC-NEXT:    v_mov_b32_e32 v1, s13
 ; VGPRRC-NEXT:    v_mov_b32_e32 v2, s14
 ; VGPRRC-NEXT:    v_mov_b32_e32 v3, s15
-; VGPRRC-NEXT:    global_store_dwordx4 v[44:45], v[0:3], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[36:37], v[0:3], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_endpgm
 ; AGPR-LABEL: test_mfma_i32_32x32x32_i8:
@@ -3496,19 +3492,19 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__flags(<4 x i32> %arg0, <4
 ; SDAG:       ; %bb.0:
 ; SDAG-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; SDAG-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; SDAG-NEXT:    v_mov_b64_e32 v[8:9], 48
-; SDAG-NEXT:    v_mov_b64_e32 v[10:11], 32
-; SDAG-NEXT:    v_mov_b64_e32 v[12:13], 16
+; SDAG-NEXT:    v_mov_b64_e32 v[0:1], 48
+; SDAG-NEXT:    v_mov_b64_e32 v[2:3], 32
+; SDAG-NEXT:    v_mov_b64_e32 v[4:5], 16
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v0, s24
-; SDAG-NEXT:    v_mov_b32_e32 v1, s25
-; SDAG-NEXT:    v_mov_b32_e32 v2, s26
-; SDAG-NEXT:    v_mov_b32_e32 v3, s27
+; SDAG-NEXT:    v_mov_b32_e32 v8, s24
+; SDAG-NEXT:    v_mov_b32_e32 v9, s25
+; SDAG-NEXT:    v_mov_b32_e32 v10, s26
+; SDAG-NEXT:    v_mov_b32_e32 v11, s27
 ; SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
-; SDAG-NEXT:    v_mov_b32_e32 v4, s28
-; SDAG-NEXT:    v_mov_b32_e32 v5, s29
-; SDAG-NEXT:    v_mov_b32_e32 v6, s30
-; SDAG-NEXT:    v_mov_b32_e32 v7, s31
+; SDAG-NEXT:    v_mov_b32_e32 v12, s28
+; SDAG-NEXT:    v_mov_b32_e32 v13, s29
+; SDAG-NEXT:    v_mov_b32_e32 v14, s30
+; SDAG-NEXT:    v_mov_b32_e32 v15, s31
 ; SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
 ; SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
 ; SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
@@ -3524,44 +3520,42 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__flags(<4 x i32> %arg0, <4
 ; SDAG-NEXT:    v_accvgpr_write_b32 a13, s21
 ; SDAG-NEXT:    v_accvgpr_write_b32 a14, s22
 ; SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
-; SDAG-NEXT:    v_mov_b64_e32 v[14:15], 0
+; SDAG-NEXT:    v_mov_b64_e32 v[6:7], 0
 ; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_mfma_i32_32x32x32_i8 a[16:31], v[0:3], v[4:7], a[0:15] cbsz:2 abid:3 blgp:1
-; SDAG-NEXT:    v_mov_b32_e32 v0, s16
-; SDAG-NEXT:    v_mov_b32_e32 v1, s17
-; SDAG-NEXT:    v_mov_b32_e32 v2, s18
-; SDAG-NEXT:    v_mov_b32_e32 v3, s19
+; SDAG-NEXT:    v_mfma_i32_32x32x32_i8 a[16:31], v[8:11], v[12:15], a[0:15] cbsz:2 abid:3 blgp:1
+; SDAG-NEXT:    v_mov_b32_e32 v8, s16
+; SDAG-NEXT:    v_mov_b32_e32 v9, s17
+; SDAG-NEXT:    v_mov_b32_e32 v10, s18
+; SDAG-NEXT:    v_mov_b32_e32 v11, s19
 ; SDAG-NEXT:    s_nop 7
-; SDAG-NEXT:    global_store_dwordx4 v[8:9], a[28:31], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[0:1], a[28:31], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[10:11], a[24:27], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[2:3], a[24:27], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[12:13], a[20:23], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[4:5], a[20:23], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[14:15], a[16:19], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[6:7], a[16:19], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[10:11], v[0:3], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[2:3], v[8:11], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_mov_b32_e32 v0, s20
-; SDAG-NEXT:    v_mov_b32_e32 v1, s21
-; SDAG-NEXT:    v_mov_b32_e32 v2, s22
-; SDAG-NEXT:    v_mov_b32_e32 v3, s23
-; SDAG-NEXT:    global_store_dwordx4 v[8:9], v[0:3], off sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v2, s10
+; SDAG-NEXT:    v_mov_b32_e32 v3, s11
+; SDAG-NEXT:    v_mov_b32_e32 v8, s20
+; SDAG-NEXT:    v_mov_b32_e32 v9, s21
+; SDAG-NEXT:    v_mov_b32_e32 v10, s22
+; SDAG-NEXT:    v_mov_b32_e32 v11, s23
+; SDAG-NEXT:    global_store_dwordx4 v[0:1], v[8:11], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    s_nop 0
 ; SDAG-NEXT:    v_mov_b32_e32 v0, s8
 ; SDAG-NEXT:    v_mov_b32_e32 v1, s9
-; SDAG-NEXT:    v_mov_b32_e32 v2, s10
-; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    global_store_dwordx4 v[14:15], v[0:3], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[6:7], v[0:3], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
 ; SDAG-NEXT:    v_mov_b32_e32 v0, s12
 ; SDAG-NEXT:    v_mov_b32_e32 v1, s13
 ; SDAG-NEXT:    v_mov_b32_e32 v2, s14
 ; SDAG-NEXT:    v_mov_b32_e32 v3, s15
-; SDAG-NEXT:    global_store_dwordx4 v[12:13], v[0:3], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[4:5], v[0:3], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_endpgm
 ;
@@ -3626,19 +3620,19 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__flags(<4 x i32> %arg0, <4
 ; HEURRC:       ; %bb.0:
 ; HEURRC-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; HEURRC-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; HEURRC-NEXT:    v_mov_b64_e32 v[8:9], 48
-; HEURRC-NEXT:    v_mov_b64_e32 v[10:11], 32
-; HEURRC-NEXT:    v_mov_b64_e32 v[12:13], 16
+; HEURRC-NEXT:    v_mov_b64_e32 v[0:1], 48
+; HEURRC-NEXT:    v_mov_b64_e32 v[2:3], 32
+; HEURRC-NEXT:    v_mov_b64_e32 v[4:5], 16
 ; HEURRC-NEXT:    s_waitcnt lgkmcnt(0)
-; HEURRC-NEXT:    v_mov_b32_e32 v0, s24
-; HEURRC-NEXT:    v_mov_b32_e32 v1, s25
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s26
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s27
+; HEURRC-NEXT:    v_mov_b32_e32 v8, s24
+; HEURRC-NEXT:    v_mov_b32_e32 v9, s25
+; HEURRC-NEXT:    v_mov_b32_e32 v10, s26
+; HEURRC-NEXT:    v_mov_b32_e32 v11, s27
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a0, s8
-; HEURRC-NEXT:    v_mov_b32_e32 v4, s28
-; HEURRC-NEXT:    v_mov_b32_e32 v5, s29
-; HEURRC-NEXT:    v_mov_b32_e32 v6, s30
-; HEURRC-NEXT:    v_mov_b32_e32 v7, s31
+; HEURRC-NEXT:    v_mov_b32_e32 v12, s28
+; HEURRC-NEXT:    v_mov_b32_e32 v13, s29
+; HEURRC-NEXT:    v_mov_b32_e32 v14, s30
+; HEURRC-NEXT:    v_mov_b32_e32 v15, s31
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a1, s9
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a2, s10
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a3, s11
@@ -3654,44 +3648,42 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__flags(<4 x i32> %arg0, <4
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a13, s21
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a14, s22
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a15, s23
-; HEURRC-NEXT:    v_mov_b64_e32 v[14:15], 0
+; HEURRC-NEXT:    v_mov_b64_e32 v[6:7], 0
 ; HEURRC-NEXT:    s_nop 0
-; HEURRC-NEXT:    v_mfma_i32_32x32x32_i8 a[16:31], v[0:3], v[4:7], a[0:15] cbsz:2 abid:3 blgp:1
-; HEURRC-NEXT:    v_mov_b32_e32 v0, s16
-; HEURRC-NEXT:    v_mov_b32_e32 v1, s17
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s18
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s19
+; HEURRC-NEXT:    v_mfma_i32_32x32x32_i8 a[16:31], v[8:11], v[12:15], a[0:15] cbsz:2 abid:3 blgp:1
+; HEURRC-NEXT:    v_mov_b32_e32 v8, s16
+; HEURRC-NEXT:    v_mov_b32_e32 v9, s17
+; HEURRC-NEXT:    v_mov_b32_e32 v10, s18
+; HEURRC-NEXT:    v_mov_b32_e32 v11, s19
 ; HEURRC-NEXT:    s_nop 7
-; HEURRC-NEXT:    global_store_dwordx4 v[8:9], a[28:31], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[0:1], a[28:31], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v[10:11], a[24:27], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[2:3], a[24:27], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v[12:13], a[20:23], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[4:5], a[20:23], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v[14:15], a[16:19], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[6:7], a[16:19], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v[10:11], v[0:3], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[2:3], v[8:11], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    s_nop 0
-; HEURRC-NEXT:    v_mov_b32_e32 v0, s20
-; HEURRC-NEXT:    v_mov_b32_e32 v1, s21
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s22
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s23
-; HEURRC-NEXT:    global_store_dwordx4 v[8:9], v[0:3], off sc0 sc1
+; HEURRC-NEXT:    v_mov_b32_e32 v2, s10
+; HEURRC-NEXT:    v_mov_b32_e32 v3, s11
+; HEURRC-NEXT:    v_mov_b32_e32 v8, s20
+; HEURRC-NEXT:    v_mov_b32_e32 v9, s21
+; HEURRC-NEXT:    v_mov_b32_e32 v10, s22
+; HEURRC-NEXT:    v_mov_b32_e32 v11, s23
+; HEURRC-NEXT:    global_store_dwordx4 v[0:1], v[8:11], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    s_nop 0
 ; HEURRC-NEXT:    v_mov_b32_e32 v0, s8
 ; HEURRC-NEXT:    v_mov_b32_e32 v1, s9
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s10
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s11
-; HEURRC-NEXT:    global_store_dwordx4 v[14:15], v[0:3], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[6:7], v[0:3], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_nop 0
 ; HEURRC-NEXT:    v_mov_b32_e32 v0, s12
 ; HEURRC-NEXT:    v_mov_b32_e32 v1, s13
 ; HEURRC-NEXT:    v_mov_b32_e32 v2, s14
 ; HEURRC-NEXT:    v_mov_b32_e32 v3, s15
-; HEURRC-NEXT:    global_store_dwordx4 v[12:13], v[0:3], off sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v[4:5], v[0:3], off sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_endpgm
 ;
@@ -3699,19 +3691,19 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__flags(<4 x i32> %arg0, <4
 ; VGPRRC:       ; %bb.0:
 ; VGPRRC-NEXT:    s_load_dwordx8 s[24:31], s[4:5], 0x24
 ; VGPRRC-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; VGPRRC-NEXT:    v_mov_b64_e32 v[40:41], 48
-; VGPRRC-NEXT:    v_mov_b64_e32 v[42:43], 32
-; VGPRRC-NEXT:    v_mov_b64_e32 v[44:45], 16
+; VGPRRC-NEXT:    v_mov_b64_e32 v[32:33], 48
+; VGPRRC-NEXT:    v_mov_b64_e32 v[34:35], 32
+; VGPRRC-NEXT:    v_mov_b64_e32 v[36:37], 16
 ; VGPRRC-NEXT:    s_waitcnt lgkmcnt(0)
-; VGPRRC-NEXT:    v_mov_b32_e32 v32, s24
-; VGPRRC-NEXT:    v_mov_b32_e32 v33, s25
-; VGPRRC-NEXT:    v_mov_b32_e32 v34, s26
-; VGPRRC-NEXT:    v_mov_b32_e32 v35, s27
+; VGPRRC-NEXT:    v_mov_b32_e32 v40, s24
+; VGPRRC-NEXT:    v_mov_b32_e32 v41, s25
+; VGPRRC-NEXT:    v_mov_b32_e32 v42, s26
+; VGPRRC-NEXT:    v_mov_b32_e32 v43, s27
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
-; VGPRRC-NEXT:    v_mov_b32_e32 v36, s28
-; VGPRRC-NEXT:    v_mov_b32_e32 v37, s29
-; VGPRRC-NEXT:    v_mov_b32_e32 v38, s30
-; VGPRRC-NEXT:    v_mov_b32_e32 v39, s31
+; VGPRRC-NEXT:    v_mov_b32_e32 v44, s28
+; VGPRRC-NEXT:    v_mov_b32_e32 v45, s29
+; VGPRRC-NEXT:    v_mov_b32_e32 v46, s30
+; VGPRRC-NEXT:    v_mov_b32_e32 v47, s31
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[4:5], s[12:13]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[6:7], s[14:15]
@@ -3719,45 +3711,45 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__flags(<4 x i32> %arg0, <4
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[10:11], s[18:19]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[12:13], s[20:21]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[14:15], s[22:23]
-; VGPRRC-NEXT:    v_mov_b64_e32 v[46:47], 0
+; VGPRRC-NEXT:    v_mov_b64_e32 v[38:39], 0
 ; VGPRRC-NEXT:    s_nop 0
-; VGPRRC-NEXT:    v_mfma_i32_32x32x32_i8 v[16:31], v[32:35], v[36:39], v[0:15] cbsz:2 abid:3 blgp:1
+; VGPRRC-NEXT:    v_mfma_i32_32x32x32_i8 v[16:31], v[40:43], v[44:47], v[0:15] cbsz:2 abid:3 blgp:1
 ; VGPRRC-NEXT:    s_nop 7
 ; VGPRRC-NEXT:    s_nop 3
-; VGPRRC-NEXT:    global_store_dwordx4 v[40:41], v[28:31], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[32:33], v[28:31], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v[42:43], v[24:27], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[34:35], v[24:27], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v[44:45], v[20:23], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[36:37], v[20:23], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v[46:47], v[16:19], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[38:39], v[16:19], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    v_mov_b32_e32 v0, s16
 ; VGPRRC-NEXT:    v_mov_b32_e32 v1, s17
 ; VGPRRC-NEXT:    v_mov_b32_e32 v2, s18
 ; VGPRRC-NEXT:    v_mov_b32_e32 v3, s19
-; VGPRRC-NEXT:    global_store_dwordx4 v[42:43], v[0:3], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[34:35], v[0:3], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v0, s20
 ; VGPRRC-NEXT:    v_mov_b32_e32 v1, s21
 ; VGPRRC-NEXT:    v_mov_b32_e32 v2, s22
 ; VGPRRC-NEXT:    v_mov_b32_e32 v3, s23
-; VGPRRC-NEXT:    global_store_dwordx4 v[40:41], v[0:3], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[32:33], v[0:3], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v0, s8
 ; VGPRRC-NEXT:    v_mov_b32_e32 v1, s9
 ; VGPRRC-NEXT:    v_mov_b32_e32 v2, s10
 ; VGPRRC-NEXT:    v_mov_b32_e32 v3, s11
-; VGPRRC-NEXT:    global_store_dwordx4 v[46:47], v[0:3], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[38:39], v[0:3], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v0, s12
 ; VGPRRC-NEXT:    v_mov_b32_e32 v1, s13
 ; VGPRRC-NEXT:    v_mov_b32_e32 v2, s14
 ; VGPRRC-NEXT:    v_mov_b32_e32 v3, s15
-; VGPRRC-NEXT:    global_store_dwordx4 v[44:45], v[0:3], off sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v[36:37], v[0:3], off sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_endpgm
 ; AGPR-LABEL: test_mfma_i32_32x32x32_i8__flags:
@@ -4254,17 +4246,17 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd(<4 x i32> %arg0, <4
 ; SDAG:       ; %bb.0:
 ; SDAG-NEXT:    s_load_dwordx8 s[20:27], s[4:5], 0x24
 ; SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0xa4
-; SDAG-NEXT:    v_mov_b32_e32 v8, 0
+; SDAG-NEXT:    v_mov_b32_e32 v0, 0
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v0, s20
-; SDAG-NEXT:    v_mov_b32_e32 v1, s21
-; SDAG-NEXT:    v_mov_b32_e32 v2, s22
-; SDAG-NEXT:    v_mov_b32_e32 v3, s23
+; SDAG-NEXT:    v_mov_b32_e32 v2, s20
+; SDAG-NEXT:    v_mov_b32_e32 v3, s21
+; SDAG-NEXT:    v_mov_b32_e32 v4, s22
+; SDAG-NEXT:    v_mov_b32_e32 v5, s23
 ; SDAG-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; SDAG-NEXT:    v_mov_b32_e32 v4, s24
-; SDAG-NEXT:    v_mov_b32_e32 v5, s25
-; SDAG-NEXT:    v_mov_b32_e32 v6, s26
-; SDAG-NEXT:    v_mov_b32_e32 v7, s27
+; SDAG-NEXT:    v_mov_b32_e32 v6, s24
+; SDAG-NEXT:    v_mov_b32_e32 v7, s25
+; SDAG-NEXT:    v_mov_b32_e32 v8, s26
+; SDAG-NEXT:    v_mov_b32_e32 v9, s27
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
 ; SDAG-NEXT:    v_accvgpr_write_b32 a31, s23
 ; SDAG-NEXT:    v_accvgpr_write_b32 a30, s22
@@ -4283,41 +4275,41 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd(<4 x i32> %arg0, <4
 ; SDAG-NEXT:    v_accvgpr_write_b32 a17, s9
 ; SDAG-NEXT:    v_accvgpr_write_b32 a16, s8
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_mfma_i32_32x32x32_i8 a[0:15], v[0:3], v[4:7], a[16:31]
-; SDAG-NEXT:    v_mov_b32_e32 v0, s20
-; SDAG-NEXT:    v_mov_b32_e32 v1, s21
-; SDAG-NEXT:    v_mov_b32_e32 v2, s22
-; SDAG-NEXT:    v_mov_b32_e32 v3, s23
-; SDAG-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:48 sc0 sc1
+; SDAG-NEXT:    v_mfma_i32_32x32x32_i8 a[0:15], v[2:5], v[6:9], a[16:31]
+; SDAG-NEXT:    v_mov_b32_e32 v2, s20
+; SDAG-NEXT:    v_mov_b32_e32 v3, s21
+; SDAG-NEXT:    v_mov_b32_e32 v4, s22
+; SDAG-NEXT:    v_mov_b32_e32 v5, s23
+; SDAG-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1] offset:48 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_mov_b32_e32 v0, s16
-; SDAG-NEXT:    v_mov_b32_e32 v1, s17
-; SDAG-NEXT:    v_mov_b32_e32 v2, s18
-; SDAG-NEXT:    v_mov_b32_e32 v3, s19
-; SDAG-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:32 sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v2, s16
+; SDAG-NEXT:    v_mov_b32_e32 v3, s17
+; SDAG-NEXT:    v_mov_b32_e32 v4, s18
+; SDAG-NEXT:    v_mov_b32_e32 v5, s19
+; SDAG-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1] offset:32 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_mov_b32_e32 v0, s12
-; SDAG-NEXT:    v_mov_b32_e32 v1, s13
-; SDAG-NEXT:    v_mov_b32_e32 v2, s14
-; SDAG-NEXT:    v_mov_b32_e32 v3, s15
-; SDAG-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:16 sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; SDAG-NEXT:    v_mov_b32_e32 v4, s14
+; SDAG-NEXT:    v_mov_b32_e32 v5, s15
+; SDAG-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1] offset:16 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_mov_b32_e32 v0, s8
-; SDAG-NEXT:    v_mov_b32_e32 v1, s9
-; SDAG-NEXT:    v_mov_b32_e32 v2, s10
-; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v2, s8
+; SDAG-NEXT:    v_mov_b32_e32 v3, s9
+; SDAG-NEXT:    v_mov_b32_e32 v4, s10
+; SDAG-NEXT:    v_mov_b32_e32 v5, s11
+; SDAG-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1] sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v8, a[8:11], s[0:1] offset:32 sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v8, a[12:15], s[0:1] offset:48 sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v8, a[0:3], s[0:1] sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1] sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v8, a[4:7], s[0:1] offset:16 sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_endpgm
 ;
@@ -4379,17 +4371,17 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd(<4 x i32> %arg0, <4
 ; HEURRC:       ; %bb.0:
 ; HEURRC-NEXT:    s_load_dwordx8 s[20:27], s[4:5], 0x24
 ; HEURRC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0xa4
-; HEURRC-NEXT:    v_mov_b32_e32 v8, 0
+; HEURRC-NEXT:    v_mov_b32_e32 v0, 0
 ; HEURRC-NEXT:    s_waitcnt lgkmcnt(0)
-; HEURRC-NEXT:    v_mov_b32_e32 v0, s20
-; HEURRC-NEXT:    v_mov_b32_e32 v1, s21
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s22
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s23
+; HEURRC-NEXT:    v_mov_b32_e32 v2, s20
+; HEURRC-NEXT:    v_mov_b32_e32 v3, s21
+; HEURRC-NEXT:    v_mov_b32_e32 v4, s22
+; HEURRC-NEXT:    v_mov_b32_e32 v5, s23
 ; HEURRC-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; HEURRC-NEXT:    v_mov_b32_e32 v4, s24
-; HEURRC-NEXT:    v_mov_b32_e32 v5, s25
-; HEURRC-NEXT:    v_mov_b32_e32 v6, s26
-; HEURRC-NEXT:    v_mov_b32_e32 v7, s27
+; HEURRC-NEXT:    v_mov_b32_e32 v6, s24
+; HEURRC-NEXT:    v_mov_b32_e32 v7, s25
+; HEURRC-NEXT:    v_mov_b32_e32 v8, s26
+; HEURRC-NEXT:    v_mov_b32_e32 v9, s27
 ; HEURRC-NEXT:    s_waitcnt lgkmcnt(0)
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a31, s23
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a30, s22
@@ -4408,41 +4400,41 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd(<4 x i32> %arg0, <4
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a17, s9
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a16, s8
 ; HEURRC-NEXT:    s_nop 1
-; HEURRC-NEXT:    v_mfma_i32_32x32x32_i8 a[0:15], v[0:3], v[4:7], a[16:31]
-; HEURRC-NEXT:    v_mov_b32_e32 v0, s20
-; HEURRC-NEXT:    v_mov_b32_e32 v1, s21
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s22
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s23
-; HEURRC-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:48 sc0 sc1
+; HEURRC-NEXT:    v_mfma_i32_32x32x32_i8 a[0:15], v[2:5], v[6:9], a[16:31]
+; HEURRC-NEXT:    v_mov_b32_e32 v2, s20
+; HEURRC-NEXT:    v_mov_b32_e32 v3, s21
+; HEURRC-NEXT:    v_mov_b32_e32 v4, s22
+; HEURRC-NEXT:    v_mov_b32_e32 v5, s23
+; HEURRC-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1] offset:48 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_nop 0
-; HEURRC-NEXT:    v_mov_b32_e32 v0, s16
-; HEURRC-NEXT:    v_mov_b32_e32 v1, s17
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s18
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s19
-; HEURRC-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:32 sc0 sc1
+; HEURRC-NEXT:    v_mov_b32_e32 v2, s16
+; HEURRC-NEXT:    v_mov_b32_e32 v3, s17
+; HEURRC-NEXT:    v_mov_b32_e32 v4, s18
+; HEURRC-NEXT:    v_mov_b32_e32 v5, s19
+; HEURRC-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1] offset:32 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_nop 0
-; HEURRC-NEXT:    v_mov_b32_e32 v0, s12
-; HEURRC-NEXT:    v_mov_b32_e32 v1, s13
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s14
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s15
-; HEURRC-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:16 sc0 sc1
+; HEURRC-NEXT:    v_mov_b32_e32 v2, s12
+; HEURRC-NEXT:    v_mov_b32_e32 v3, s13
+; HEURRC-NEXT:    v_mov_b32_e32 v4, s14
+; HEURRC-NEXT:    v_mov_b32_e32 v5, s15
+; HEURRC-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1] offset:16 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_nop 0
-; HEURRC-NEXT:    v_mov_b32_e32 v0, s8
-; HEURRC-NEXT:    v_mov_b32_e32 v1, s9
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s10
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s11
-; HEURRC-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1
+; HEURRC-NEXT:    v_mov_b32_e32 v2, s8
+; HEURRC-NEXT:    v_mov_b32_e32 v3, s9
+; HEURRC-NEXT:    v_mov_b32_e32 v4, s10
+; HEURRC-NEXT:    v_mov_b32_e32 v5, s11
+; HEURRC-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1] sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v8, a[8:11], s[0:1] offset:32 sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v8, a[12:15], s[0:1] offset:48 sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v8, a[0:3], s[0:1] sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1] sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v8, a[4:7], s[0:1] offset:16 sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_endpgm
 ;
@@ -4450,17 +4442,17 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd(<4 x i32> %arg0, <4
 ; VGPRRC:       ; %bb.0:
 ; VGPRRC-NEXT:    s_load_dwordx8 s[20:27], s[4:5], 0x24
 ; VGPRRC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0xa4
-; VGPRRC-NEXT:    v_mov_b32_e32 v40, 0
+; VGPRRC-NEXT:    v_mov_b32_e32 v32, 0
 ; VGPRRC-NEXT:    s_waitcnt lgkmcnt(0)
-; VGPRRC-NEXT:    v_mov_b32_e32 v32, s20
-; VGPRRC-NEXT:    v_mov_b32_e32 v33, s21
-; VGPRRC-NEXT:    v_mov_b32_e32 v34, s22
-; VGPRRC-NEXT:    v_mov_b32_e32 v35, s23
+; VGPRRC-NEXT:    v_mov_b32_e32 v34, s20
+; VGPRRC-NEXT:    v_mov_b32_e32 v35, s21
+; VGPRRC-NEXT:    v_mov_b32_e32 v36, s22
+; VGPRRC-NEXT:    v_mov_b32_e32 v37, s23
 ; VGPRRC-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; VGPRRC-NEXT:    v_mov_b32_e32 v36, s24
-; VGPRRC-NEXT:    v_mov_b32_e32 v37, s25
-; VGPRRC-NEXT:    v_mov_b32_e32 v38, s26
-; VGPRRC-NEXT:    v_mov_b32_e32 v39, s27
+; VGPRRC-NEXT:    v_mov_b32_e32 v38, s24
+; VGPRRC-NEXT:    v_mov_b32_e32 v39, s25
+; VGPRRC-NEXT:    v_mov_b32_e32 v40, s26
+; VGPRRC-NEXT:    v_mov_b32_e32 v41, s27
 ; VGPRRC-NEXT:    s_waitcnt lgkmcnt(0)
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[30:31], s[22:23]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[28:29], s[20:21]
@@ -4471,42 +4463,42 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd(<4 x i32> %arg0, <4
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[18:19], s[10:11]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[16:17], s[8:9]
 ; VGPRRC-NEXT:    s_nop 1
-; VGPRRC-NEXT:    v_mfma_i32_32x32x32_i8 v[0:15], v[32:35], v[36:39], v[16:31]
+; VGPRRC-NEXT:    v_mfma_i32_32x32x32_i8 v[0:15], v[34:37], v[38:41], v[16:31]
 ; VGPRRC-NEXT:    s_nop 6
 ; VGPRRC-NEXT:    v_mov_b32_e32 v16, s20
 ; VGPRRC-NEXT:    v_mov_b32_e32 v17, s21
 ; VGPRRC-NEXT:    v_mov_b32_e32 v18, s22
 ; VGPRRC-NEXT:    v_mov_b32_e32 v19, s23
-; VGPRRC-NEXT:    global_store_dwordx4 v40, v[16:19], s[0:1] offset:48 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v32, v[16:19], s[0:1] offset:48 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v16, s16
 ; VGPRRC-NEXT:    v_mov_b32_e32 v17, s17
 ; VGPRRC-NEXT:    v_mov_b32_e32 v18, s18
 ; VGPRRC-NEXT:    v_mov_b32_e32 v19, s19
-; VGPRRC-NEXT:    global_store_dwordx4 v40, v[16:19], s[0:1] offset:32 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v32, v[16:19], s[0:1] offset:32 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v16, s12
 ; VGPRRC-NEXT:    v_mov_b32_e32 v17, s13
 ; VGPRRC-NEXT:    v_mov_b32_e32 v18, s14
 ; VGPRRC-NEXT:    v_mov_b32_e32 v19, s15
-; VGPRRC-NEXT:    global_store_dwordx4 v40, v[16:19], s[0:1] offset:16 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v32, v[16:19], s[0:1] offset:16 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v16, s8
 ; VGPRRC-NEXT:    v_mov_b32_e32 v17, s9
 ; VGPRRC-NEXT:    v_mov_b32_e32 v18, s10
 ; VGPRRC-NEXT:    v_mov_b32_e32 v19, s11
-; VGPRRC-NEXT:    global_store_dwordx4 v40, v[16:19], s[0:1] sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v32, v[16:19], s[0:1] sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v40, v[8:11], s[0:1] offset:32 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v32, v[8:11], s[0:1] offset:32 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v40, v[12:15], s[0:1] offset:48 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v32, v[12:15], s[0:1] offset:48 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v40, v[0:3], s[0:1] sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v32, v[0:3], s[0:1] sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v40, v[4:7], s[0:1] offset:16 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v32, v[4:7], s[0:1] offset:16 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_endpgm
 ; AGPR-LABEL: test_mfma_i32_32x32x32_i8__vgprcd:
@@ -4653,17 +4645,17 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd__flags(<4 x i32> %a
 ; SDAG:       ; %bb.0:
 ; SDAG-NEXT:    s_load_dwordx8 s[20:27], s[4:5], 0x24
 ; SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0xa4
-; SDAG-NEXT:    v_mov_b32_e32 v8, 0
+; SDAG-NEXT:    v_mov_b32_e32 v0, 0
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v0, s20
-; SDAG-NEXT:    v_mov_b32_e32 v1, s21
-; SDAG-NEXT:    v_mov_b32_e32 v2, s22
-; SDAG-NEXT:    v_mov_b32_e32 v3, s23
+; SDAG-NEXT:    v_mov_b32_e32 v2, s20
+; SDAG-NEXT:    v_mov_b32_e32 v3, s21
+; SDAG-NEXT:    v_mov_b32_e32 v4, s22
+; SDAG-NEXT:    v_mov_b32_e32 v5, s23
 ; SDAG-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; SDAG-NEXT:    v_mov_b32_e32 v4, s24
-; SDAG-NEXT:    v_mov_b32_e32 v5, s25
-; SDAG-NEXT:    v_mov_b32_e32 v6, s26
-; SDAG-NEXT:    v_mov_b32_e32 v7, s27
+; SDAG-NEXT:    v_mov_b32_e32 v6, s24
+; SDAG-NEXT:    v_mov_b32_e32 v7, s25
+; SDAG-NEXT:    v_mov_b32_e32 v8, s26
+; SDAG-NEXT:    v_mov_b32_e32 v9, s27
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
 ; SDAG-NEXT:    v_accvgpr_write_b32 a31, s23
 ; SDAG-NEXT:    v_accvgpr_write_b32 a30, s22
@@ -4682,41 +4674,41 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd__flags(<4 x i32> %a
 ; SDAG-NEXT:    v_accvgpr_write_b32 a17, s9
 ; SDAG-NEXT:    v_accvgpr_write_b32 a16, s8
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_mfma_i32_32x32x32_i8 a[0:15], v[0:3], v[4:7], a[16:31] cbsz:1 abid:2 blgp:3
-; SDAG-NEXT:    v_mov_b32_e32 v0, s20
-; SDAG-NEXT:    v_mov_b32_e32 v1, s21
-; SDAG-NEXT:    v_mov_b32_e32 v2, s22
-; SDAG-NEXT:    v_mov_b32_e32 v3, s23
-; SDAG-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:48 sc0 sc1
+; SDAG-NEXT:    v_mfma_i32_32x32x32_i8 a[0:15], v[2:5], v[6:9], a[16:31] cbsz:1 abid:2 blgp:3
+; SDAG-NEXT:    v_mov_b32_e32 v2, s20
+; SDAG-NEXT:    v_mov_b32_e32 v3, s21
+; SDAG-NEXT:    v_mov_b32_e32 v4, s22
+; SDAG-NEXT:    v_mov_b32_e32 v5, s23
+; SDAG-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1] offset:48 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_mov_b32_e32 v0, s16
-; SDAG-NEXT:    v_mov_b32_e32 v1, s17
-; SDAG-NEXT:    v_mov_b32_e32 v2, s18
-; SDAG-NEXT:    v_mov_b32_e32 v3, s19
-; SDAG-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:32 sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v2, s16
+; SDAG-NEXT:    v_mov_b32_e32 v3, s17
+; SDAG-NEXT:    v_mov_b32_e32 v4, s18
+; SDAG-NEXT:    v_mov_b32_e32 v5, s19
+; SDAG-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1] offset:32 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_mov_b32_e32 v0, s12
-; SDAG-NEXT:    v_mov_b32_e32 v1, s13
-; SDAG-NEXT:    v_mov_b32_e32 v2, s14
-; SDAG-NEXT:    v_mov_b32_e32 v3, s15
-; SDAG-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:16 sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; SDAG-NEXT:    v_mov_b32_e32 v4, s14
+; SDAG-NEXT:    v_mov_b32_e32 v5, s15
+; SDAG-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1] offset:16 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_mov_b32_e32 v0, s8
-; SDAG-NEXT:    v_mov_b32_e32 v1, s9
-; SDAG-NEXT:    v_mov_b32_e32 v2, s10
-; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v2, s8
+; SDAG-NEXT:    v_mov_b32_e32 v3, s9
+; SDAG-NEXT:    v_mov_b32_e32 v4, s10
+; SDAG-NEXT:    v_mov_b32_e32 v5, s11
+; SDAG-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1] sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v8, a[8:11], s[0:1] offset:32 sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v8, a[12:15], s[0:1] offset:48 sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v8, a[0:3], s[0:1] sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1] sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v8, a[4:7], s[0:1] offset:16 sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_endpgm
 ;
@@ -4778,17 +4770,17 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd__flags(<4 x i32> %a
 ; HEURRC:       ; %bb.0:
 ; HEURRC-NEXT:    s_load_dwordx8 s[20:27], s[4:5], 0x24
 ; HEURRC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0xa4
-; HEURRC-NEXT:    v_mov_b32_e32 v8, 0
+; HEURRC-NEXT:    v_mov_b32_e32 v0, 0
 ; HEURRC-NEXT:    s_waitcnt lgkmcnt(0)
-; HEURRC-NEXT:    v_mov_b32_e32 v0, s20
-; HEURRC-NEXT:    v_mov_b32_e32 v1, s21
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s22
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s23
+; HEURRC-NEXT:    v_mov_b32_e32 v2, s20
+; HEURRC-NEXT:    v_mov_b32_e32 v3, s21
+; HEURRC-NEXT:    v_mov_b32_e32 v4, s22
+; HEURRC-NEXT:    v_mov_b32_e32 v5, s23
 ; HEURRC-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; HEURRC-NEXT:    v_mov_b32_e32 v4, s24
-; HEURRC-NEXT:    v_mov_b32_e32 v5, s25
-; HEURRC-NEXT:    v_mov_b32_e32 v6, s26
-; HEURRC-NEXT:    v_mov_b32_e32 v7, s27
+; HEURRC-NEXT:    v_mov_b32_e32 v6, s24
+; HEURRC-NEXT:    v_mov_b32_e32 v7, s25
+; HEURRC-NEXT:    v_mov_b32_e32 v8, s26
+; HEURRC-NEXT:    v_mov_b32_e32 v9, s27
 ; HEURRC-NEXT:    s_waitcnt lgkmcnt(0)
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a31, s23
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a30, s22
@@ -4807,41 +4799,41 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd__flags(<4 x i32> %a
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a17, s9
 ; HEURRC-NEXT:    v_accvgpr_write_b32 a16, s8
 ; HEURRC-NEXT:    s_nop 1
-; HEURRC-NEXT:    v_mfma_i32_32x32x32_i8 a[0:15], v[0:3], v[4:7], a[16:31] cbsz:1 abid:2 blgp:3
-; HEURRC-NEXT:    v_mov_b32_e32 v0, s20
-; HEURRC-NEXT:    v_mov_b32_e32 v1, s21
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s22
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s23
-; HEURRC-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:48 sc0 sc1
+; HEURRC-NEXT:    v_mfma_i32_32x32x32_i8 a[0:15], v[2:5], v[6:9], a[16:31] cbsz:1 abid:2 blgp:3
+; HEURRC-NEXT:    v_mov_b32_e32 v2, s20
+; HEURRC-NEXT:    v_mov_b32_e32 v3, s21
+; HEURRC-NEXT:    v_mov_b32_e32 v4, s22
+; HEURRC-NEXT:    v_mov_b32_e32 v5, s23
+; HEURRC-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1] offset:48 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_nop 0
-; HEURRC-NEXT:    v_mov_b32_e32 v0, s16
-; HEURRC-NEXT:    v_mov_b32_e32 v1, s17
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s18
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s19
-; HEURRC-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:32 sc0 sc1
+; HEURRC-NEXT:    v_mov_b32_e32 v2, s16
+; HEURRC-NEXT:    v_mov_b32_e32 v3, s17
+; HEURRC-NEXT:    v_mov_b32_e32 v4, s18
+; HEURRC-NEXT:    v_mov_b32_e32 v5, s19
+; HEURRC-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1] offset:32 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_nop 0
-; HEURRC-NEXT:    v_mov_b32_e32 v0, s12
-; HEURRC-NEXT:    v_mov_b32_e32 v1, s13
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s14
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s15
-; HEURRC-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] offset:16 sc0 sc1
+; HEURRC-NEXT:    v_mov_b32_e32 v2, s12
+; HEURRC-NEXT:    v_mov_b32_e32 v3, s13
+; HEURRC-NEXT:    v_mov_b32_e32 v4, s14
+; HEURRC-NEXT:    v_mov_b32_e32 v5, s15
+; HEURRC-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1] offset:16 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_nop 0
-; HEURRC-NEXT:    v_mov_b32_e32 v0, s8
-; HEURRC-NEXT:    v_mov_b32_e32 v1, s9
-; HEURRC-NEXT:    v_mov_b32_e32 v2, s10
-; HEURRC-NEXT:    v_mov_b32_e32 v3, s11
-; HEURRC-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1] sc0 sc1
+; HEURRC-NEXT:    v_mov_b32_e32 v2, s8
+; HEURRC-NEXT:    v_mov_b32_e32 v3, s9
+; HEURRC-NEXT:    v_mov_b32_e32 v4, s10
+; HEURRC-NEXT:    v_mov_b32_e32 v5, s11
+; HEURRC-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1] sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v8, a[8:11], s[0:1] offset:32 sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v8, a[12:15], s[0:1] offset:48 sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v8, a[0:3], s[0:1] sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1] sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
-; HEURRC-NEXT:    global_store_dwordx4 v8, a[4:7], s[0:1] offset:16 sc0 sc1
+; HEURRC-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16 sc0 sc1
 ; HEURRC-NEXT:    s_waitcnt vmcnt(0)
 ; HEURRC-NEXT:    s_endpgm
 ;
@@ -4849,17 +4841,17 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd__flags(<4 x i32> %a
 ; VGPRRC:       ; %bb.0:
 ; VGPRRC-NEXT:    s_load_dwordx8 s[20:27], s[4:5], 0x24
 ; VGPRRC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0xa4
-; VGPRRC-NEXT:    v_mov_b32_e32 v40, 0
+; VGPRRC-NEXT:    v_mov_b32_e32 v32, 0
 ; VGPRRC-NEXT:    s_waitcnt lgkmcnt(0)
-; VGPRRC-NEXT:    v_mov_b32_e32 v32, s20
-; VGPRRC-NEXT:    v_mov_b32_e32 v33, s21
-; VGPRRC-NEXT:    v_mov_b32_e32 v34, s22
-; VGPRRC-NEXT:    v_mov_b32_e32 v35, s23
+; VGPRRC-NEXT:    v_mov_b32_e32 v34, s20
+; VGPRRC-NEXT:    v_mov_b32_e32 v35, s21
+; VGPRRC-NEXT:    v_mov_b32_e32 v36, s22
+; VGPRRC-NEXT:    v_mov_b32_e32 v37, s23
 ; VGPRRC-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x64
-; VGPRRC-NEXT:    v_mov_b32_e32 v36, s24
-; VGPRRC-NEXT:    v_mov_b32_e32 v37, s25
-; VGPRRC-NEXT:    v_mov_b32_e32 v38, s26
-; VGPRRC-NEXT:    v_mov_b32_e32 v39, s27
+; VGPRRC-NEXT:    v_mov_b32_e32 v38, s24
+; VGPRRC-NEXT:    v_mov_b32_e32 v39, s25
+; VGPRRC-NEXT:    v_mov_b32_e32 v40, s26
+; VGPRRC-NEXT:    v_mov_b32_e32 v41, s27
 ; VGPRRC-NEXT:    s_waitcnt lgkmcnt(0)
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[30:31], s[22:23]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[28:29], s[20:21]
@@ -4870,42 +4862,42 @@ define amdgpu_kernel void @test_mfma_i32_32x32x32_i8__vgprcd__flags(<4 x i32> %a
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[18:19], s[10:11]
 ; VGPRRC-NEXT:    v_mov_b64_e32 v[16:17], s[8:9]
 ; VGPRRC-NEXT:    s_nop 1
-; VGPRRC-NEXT:    v_mfma_i32_32x32x32_i8 v[0:15], v[32:35], v[36:39], v[16:31] cbsz:1 abid:2 blgp:3
+; VGPRRC-NEXT:    v_mfma_i32_32x32x32_i8 v[0:15], v[34:37], v[38:41], v[16:31] cbsz:1 abid:2 blgp:3
 ; VGPRRC-NEXT:    s_nop 6
 ; VGPRRC-NEXT:    v_mov_b32_e32 v16, s20
 ; VGPRRC-NEXT:    v_mov_b32_e32 v17, s21
 ; VGPRRC-NEXT:    v_mov_b32_e32 v18, s22
 ; VGPRRC-NEXT:    v_mov_b32_e32 v19, s23
-; VGPRRC-NEXT:    global_store_dwordx4 v40, v[16:19], s[0:1] offset:48 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v32, v[16:19], s[0:1] offset:48 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v16, s16
 ; VGPRRC-NEXT:    v_mov_b32_e32 v17, s17
 ; VGPRRC-NEXT:    v_mov_b32_e32 v18, s18
 ; VGPRRC-NEXT:    v_mov_b32_e32 v19, s19
-; VGPRRC-NEXT:    global_store_dwordx4 v40, v[16:19], s[0:1] offset:32 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v32, v[16:19], s[0:1] offset:32 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v16, s12
 ; VGPRRC-NEXT:    v_mov_b32_e32 v17, s13
 ; VGPRRC-NEXT:    v_mov_b32_e32 v18, s14
 ; VGPRRC-NEXT:    v_mov_b32_e32 v19, s15
-; VGPRRC-NEXT:    global_store_dwordx4 v40, v[16:19], s[0:1] offset:16 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v32, v[16:19], s[0:1] offset:16 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_nop 0
 ; VGPRRC-NEXT:    v_mov_b32_e32 v16, s8
 ; VGPRRC-NEXT:    v_mov_b32_e32 v17, s9
 ; VGPRRC-NEXT:    v_mov_b32_e32 v18, s10
 ; VGPRRC-NEXT:    v_mov_b32_e32 v19, s11
-; VGPRRC-NEXT:    global_store_dwordx4 v40, v[16:19], s[0:1] sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v32, v[16:19], s[0:1] sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v40, v[8:11], s[0:1] offset:32 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v32, v[8:11], s[0:1] offset:32 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v40, v[12:15], s[0:1] offset:48 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v32, v[12:15], s[0:1] offset:48 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v40, v[0:3], s[0:1] sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v32, v[0:3], s[0:1] sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
-; VGPRRC-NEXT:    global_store_dwordx4 v40, v[4:7], s[0:1] offset:16 sc0 sc1
+; VGPRRC-NEXT:    global_store_dwordx4 v32, v[4:7], s[0:1] offset:16 sc0 sc1
 ; VGPRRC-NEXT:    s_waitcnt vmcnt(0)
 ; VGPRRC-NEXT:    s_endpgm
 ; AGPR-LABEL: test_mfma_i32_32x32x32_i8__vgprcd__flags:
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.i8.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.i8.ll
index ccee113..856185b 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.i8.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.i8.ll
@@ -1,22 +1,116 @@
-; RUN: llc -mtriple=amdgcn -mcpu=gfx908 < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX908 %s
-; RUN: llc -mtriple=amdgcn -mcpu=gfx908 -mattr=-mfma-inline-literal-bug < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX908 %s
-; RUN: llc -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX90A %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -mtriple=amdgcn -mcpu=gfx908 < %s | FileCheck --check-prefixes=GCN,GFX908 %s
+; RUN: llc -mtriple=amdgcn -mcpu=gfx908 -mattr=-mfma-inline-literal-bug < %s | FileCheck --check-prefixes=GCN,GFX908 %s
+; RUN: llc -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck --check-prefixes=GCN,GFX90A %s
 
 declare <16 x i32> @llvm.amdgcn.mfma.i32.32x32x8i8(i32, i32, <16 x i32>, i32, i32, i32)
 declare <4 x i32> @llvm.amdgcn.mfma.i32.16x16x16i8(i32, i32, <4 x i32>, i32, i32, i32)
 
-; GCN-LABEL: {{^}}test_mfma_i32_32x32x8i8:
-; GCN-DAG:         v_mov_b32_e32 [[TWO:v[0-9]+]], 2
-; GCN-DAG:         v_mov_b32_e32 [[ONE:v[0-9]+]], 1
-; GCN-DAG:         s_load_dwordx16
-; GFX908-DAG-COUNT-16: v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX90A-COUNT-16: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GCN:             v_mfma_i32_32x32x8i8 a[{{[0-9]+:[0-9]+}}], [[ONE]], [[TWO]], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GFX908-COUNT-16: v_accvgpr_read_b32
-; GFX908:          global_store_dwordx4
-; GFX90A-NOT:      v_accvgpr_read_b32
-; GFX90A-COUNT-4:  global_store_dwordx4 v{{[0-9]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_i32_32x32x8i8(ptr addrspace(1) %arg) #0 {
+; GFX908-LABEL: test_mfma_i32_32x32x8i8:
+; GFX908:       ; %bb.0: ; %bb
+; GFX908-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX908-NEXT:    v_mov_b32_e32 v0, 1
+; GFX908-NEXT:    v_mov_b32_e32 v16, 0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    v_mov_b32_e32 v17, s0
+; GFX908-NEXT:    v_mov_b32_e32 v1, s1
+; GFX908-NEXT:    v_mov_b32_e32 v2, s2
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, v17
+; GFX908-NEXT:    v_mov_b32_e32 v17, s3
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, v17
+; GFX908-NEXT:    v_mov_b32_e32 v1, s4
+; GFX908-NEXT:    v_mov_b32_e32 v2, s5
+; GFX908-NEXT:    v_mov_b32_e32 v17, s6
+; GFX908-NEXT:    v_accvgpr_write_b32 a4, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a5, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a6, v17
+; GFX908-NEXT:    v_mov_b32_e32 v1, s7
+; GFX908-NEXT:    v_mov_b32_e32 v2, s8
+; GFX908-NEXT:    v_mov_b32_e32 v17, s9
+; GFX908-NEXT:    v_accvgpr_write_b32 a7, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a8, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a9, v17
+; GFX908-NEXT:    v_mov_b32_e32 v1, s10
+; GFX908-NEXT:    v_mov_b32_e32 v2, s11
+; GFX908-NEXT:    v_mov_b32_e32 v17, s12
+; GFX908-NEXT:    v_accvgpr_write_b32 a10, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a11, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a12, v17
+; GFX908-NEXT:    v_mov_b32_e32 v1, s13
+; GFX908-NEXT:    v_mov_b32_e32 v2, s14
+; GFX908-NEXT:    v_mov_b32_e32 v17, s15
+; GFX908-NEXT:    v_accvgpr_write_b32 a13, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a14, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a15, v17
+; GFX908-NEXT:    v_mov_b32_e32 v1, 2
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_mfma_i32_32x32x8i8 a[0:15], v0, v1, a[0:15] cbsz:1 abid:2 blgp:3
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_accvgpr_read_b32 v15, a15
+; GFX908-NEXT:    v_accvgpr_read_b32 v14, a14
+; GFX908-NEXT:    v_accvgpr_read_b32 v13, a13
+; GFX908-NEXT:    v_accvgpr_read_b32 v12, a12
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a3
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a2
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a1
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a0
+; GFX908-NEXT:    v_accvgpr_read_b32 v7, a7
+; GFX908-NEXT:    v_accvgpr_read_b32 v6, a6
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a5
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a4
+; GFX908-NEXT:    v_accvgpr_read_b32 v11, a11
+; GFX908-NEXT:    v_accvgpr_read_b32 v10, a10
+; GFX908-NEXT:    v_accvgpr_read_b32 v9, a9
+; GFX908-NEXT:    v_accvgpr_read_b32 v8, a8
+; GFX908-NEXT:    global_store_dwordx4 v16, v[12:15], s[16:17] offset:48
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    global_store_dwordx4 v16, v[8:11], s[16:17] offset:32
+; GFX908-NEXT:    global_store_dwordx4 v16, v[4:7], s[16:17] offset:16
+; GFX908-NEXT:    global_store_dwordx4 v16, v[0:3], s[16:17]
+; GFX908-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: test_mfma_i32_32x32x8i8:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 1
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 2
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX90A-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX90A-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX90A-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX90A-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX90A-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX90A-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX90A-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX90A-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX90A-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX90A-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX90A-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX90A-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX90A-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX90A-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_i32_32x32x8i8 a[0:15], v0, v1, a[0:15] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX90A-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x i32>, ptr addrspace(1) %arg
   %mai.1 = tail call <16 x i32> @llvm.amdgcn.mfma.i32.32x32x8i8(i32 1, i32 2, <16 x i32> %in.1, i32 1, i32 2, i32 3)
@@ -24,18 +118,55 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_i32_16x16x16i8:
-; GCN-DAG:        v_mov_b32_e32 [[TWO:v[0-9]+]], 2
-; GCN-DAG:        v_mov_b32_e32 [[ONE:v[0-9]+]], 1
-; GCN:            s_load_dwordx4
-; GFX908-COUNT-4: v_accvgpr_write_b32 a{{[0-9]+}}, v{{[0-9]+}}
-; GFX90A-COUNT-4: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GCN:            v_mfma_i32_16x16x16i8 [[RES:a\[[0-9]+:[0-9]+\]]], [[ONE]], [[TWO]], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GFX908-COUNT-4: v_accvgpr_read_b32
-; GFX908:         global_store_dwordx4
-; GFX90A-NOT:     v_accvgpr_read_b32
-; GFX90A:         global_store_dwordx4 v{{[0-9]+}}, [[RES]]
 define amdgpu_kernel void @test_mfma_i32_16x16x16i8(ptr addrspace(1) %arg) #0 {
+; GFX908-LABEL: test_mfma_i32_16x16x16i8:
+; GFX908:       ; %bb.0: ; %bb
+; GFX908-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX908-NEXT:    v_mov_b32_e32 v0, 1
+; GFX908-NEXT:    v_mov_b32_e32 v1, 2
+; GFX908-NEXT:    v_mov_b32_e32 v4, 0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    v_mov_b32_e32 v5, s0
+; GFX908-NEXT:    v_mov_b32_e32 v2, s1
+; GFX908-NEXT:    v_mov_b32_e32 v3, s2
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, v5
+; GFX908-NEXT:    v_mov_b32_e32 v5, s3
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, v5
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_mfma_i32_16x16x16i8 a[0:3], v0, v1, a[0:3] cbsz:1 abid:2 blgp:3
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a0
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a1
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a2
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a3
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX908-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: test_mfma_i32_16x16x16i8:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 1
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 2
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX90A-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX90A-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_i32_16x16x16i8 a[0:3], v0, v2, a[0:3] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 2
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[0:3], s[6:7]
+; GFX90A-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x i32>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x i32> @llvm.amdgcn.mfma.i32.16x16x16i8(i32 1, i32 2, <4 x i32> %in.1, i32 1, i32 2, i32 3)
@@ -44,3 +175,5 @@ bb:
 }
 
 attributes #0 = { "amdgpu-flat-work-group-size"="1,256" }
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; GCN: {{.*}}
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll
index ff305da..78be949 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.ll
@@ -3,6 +3,7 @@
 ; RUN: llc -mtriple=amdgcn -mcpu=gfx908 -mattr=-mfma-inline-literal-bug < %s | FileCheck -enable-var-scope --check-prefixes=GCN,LIT-SRCC,GFX908,GFX908_A %s
 ; RUN: llc -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX90A,GFX908_A,GFX90A_42 %s
 ; RUN: llc -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX942,GFX90A_42 %s
+; RUN: llc -mtriple=amdgcn -mcpu=gfx942 -amdgpu-mfma-vgpr-form < %s | FileCheck -enable-var-scope --check-prefix=GFX942-VGPR %s
 
 declare <32 x float> @llvm.amdgcn.mfma.f32.32x32x1f32(float, float, <32 x float>, i32, i32, i32)
 declare <16 x float> @llvm.amdgcn.mfma.f32.16x16x1f32(float, float, <16 x float>, i32, i32, i32)
@@ -405,6 +406,63 @@ define amdgpu_kernel void @test_mfma_f32_32x32x1f32(ptr addrspace(1) %arg) #0 {
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[34:35]
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[34:35] offset:16
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_32x32x1f32:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[34:35], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v33, 1.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v34, 2.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v32, 0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx16 s[16:31], s[34:35], 0x0
+; GFX942-VGPR-NEXT:    s_load_dwordx16 s[0:15], s[34:35], 0x40
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, s16
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, s17
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, s18
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, s19
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, s20
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, s21
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, s22
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v7, s23
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v8, s24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v9, s25
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v10, s26
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v11, s27
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v12, s28
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v13, s29
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v14, s30
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v15, s31
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, s0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v17, s1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v18, s2
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v19, s3
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v20, s4
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v21, s5
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v22, s6
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v23, s7
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v24, s8
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v25, s9
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v26, s10
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v27, s11
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v28, s12
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v29, s13
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v30, s14
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v31, s15
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_32x32x1_2b_f32 v[0:31], v33, v34, v[0:31] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[24:27], s[34:35] offset:96
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[28:31], s[34:35] offset:112
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[16:19], s[34:35] offset:64
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[20:23], s[34:35] offset:80
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[8:11], s[34:35] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[12:15], s[34:35] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[0:3], s[34:35]
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[4:7], s[34:35] offset:16
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <32 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <32 x float> @llvm.amdgcn.mfma.f32.32x32x1f32(float 1.0, float 2.0, <32 x float> %in.1, i32 1, i32 2, i32 3)
@@ -618,6 +676,33 @@ define amdgpu_kernel void @test_mfma_f32_16x16x1f32(ptr addrspace(1) %arg) #0 {
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_16x16x1f32:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, 1.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v17, 2.0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_16x16x1_4b_f32 v[0:15], v16, v17, v[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[12:15], s[16:17] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[8:11], s[16:17] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[4:7], s[16:17] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[0:3], s[16:17]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <16 x float> @llvm.amdgcn.mfma.f32.16x16x1f32(float 1.0, float 2.0, <16 x float> %in.1, i32 1, i32 2, i32 3)
@@ -719,6 +804,23 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32(ptr addrspace(1) %arg) #0 {
 ; GFX942-NEXT:    s_nop 3
 ; GFX942-NEXT:    global_store_dwordx4 v1, a[0:3], s[6:7]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_4x4x1f32:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, 1.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, 2.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, 0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_4x4x1_16b_f32 v[0:3], v4, v6, v[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_nop 3
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v5, v[0:3], s[6:7]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x float> @llvm.amdgcn.mfma.f32.4x4x1f32(float 1.0, float 2.0, <4 x float> %in.1, i32 1, i32 2, i32 3)
@@ -934,6 +1036,34 @@ define amdgpu_kernel void @test_mfma_f32_32x32x2f32(ptr addrspace(1) %arg) #0 {
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_32x32x2f32:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, 1.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v17, 2.0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_32x32x2_f32 v[0:15], v16, v17, v[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[12:15], s[16:17] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[8:11], s[16:17] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[4:7], s[16:17] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[0:3], s[16:17]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <16 x float> @llvm.amdgcn.mfma.f32.32x32x2f32(float 1.0, float 2.0, <16 x float> %in.1, i32 1, i32 2, i32 3)
@@ -1039,6 +1169,24 @@ define amdgpu_kernel void @test_mfma_f32_16x16x4f32(ptr addrspace(1) %arg) #0 {
 ; GFX942-NEXT:    s_nop 1
 ; GFX942-NEXT:    global_store_dwordx4 v1, a[0:3], s[6:7]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_16x16x4f32:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, 1.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, 2.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, 0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_16x16x4_f32 v[0:3], v4, v6, v[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v5, v[0:3], s[6:7]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x float> @llvm.amdgcn.mfma.f32.16x16x4f32(float 1.0, float 2.0, <4 x float> %in.1, i32 1, i32 2, i32 3)
@@ -1114,19 +1262,19 @@ define amdgpu_kernel void @test_mfma_f32_32x32x4f16(ptr addrspace(1) %arg, ptr a
 ; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a27, v2
 ; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a28, v0
 ; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a29, v1
-; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v2, s14
-; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v3, s15
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v0, s14
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v1, s15
 ; NOLIT-SRCC-NEXT:    s_waitcnt lgkmcnt(0)
-; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v0, s0
-; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v1, s1
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v2, s0
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v3, s1
 ; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a4, v5
 ; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a5, v6
-; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a30, v2
-; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a31, v3
-; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v2, s2
-; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v3, s3
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a30, v0
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a31, v1
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v0, s2
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v1, s3
 ; NOLIT-SRCC-NEXT:    s_nop 1
-; NOLIT-SRCC-NEXT:    v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[2:3], a[0:31] cbsz:1 abid:2 blgp:3
+; NOLIT-SRCC-NEXT:    v_mfma_f32_32x32x4f16 a[0:31], v[2:3], v[0:1], a[0:31] cbsz:1 abid:2 blgp:3
 ; NOLIT-SRCC-NEXT:    s_nop 7
 ; NOLIT-SRCC-NEXT:    s_nop 7
 ; NOLIT-SRCC-NEXT:    s_nop 1
@@ -1254,19 +1402,19 @@ define amdgpu_kernel void @test_mfma_f32_32x32x4f16(ptr addrspace(1) %arg, ptr a
 ; LIT-SRCC-NEXT:    v_accvgpr_write_b32 a27, v2
 ; LIT-SRCC-NEXT:    v_accvgpr_write_b32 a28, v0
 ; LIT-SRCC-NEXT:    v_accvgpr_write_b32 a29, v1
-; LIT-SRCC-NEXT:    v_mov_b32_e32 v2, s14
-; LIT-SRCC-NEXT:    v_mov_b32_e32 v3, s15
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v0, s14
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v1, s15
 ; LIT-SRCC-NEXT:    s_waitcnt lgkmcnt(0)
-; LIT-SRCC-NEXT:    v_mov_b32_e32 v0, s0
-; LIT-SRCC-NEXT:    v_mov_b32_e32 v1, s1
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v2, s0
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v3, s1
 ; LIT-SRCC-NEXT:    v_accvgpr_write_b32 a4, v5
 ; LIT-SRCC-NEXT:    v_accvgpr_write_b32 a5, v6
-; LIT-SRCC-NEXT:    v_accvgpr_write_b32 a30, v2
-; LIT-SRCC-NEXT:    v_accvgpr_write_b32 a31, v3
-; LIT-SRCC-NEXT:    v_mov_b32_e32 v2, s2
-; LIT-SRCC-NEXT:    v_mov_b32_e32 v3, s3
+; LIT-SRCC-NEXT:    v_accvgpr_write_b32 a30, v0
+; LIT-SRCC-NEXT:    v_accvgpr_write_b32 a31, v1
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v0, s2
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v1, s3
 ; LIT-SRCC-NEXT:    s_nop 1
-; LIT-SRCC-NEXT:    v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[2:3], a[0:31] cbsz:1 abid:2 blgp:3
+; LIT-SRCC-NEXT:    v_mfma_f32_32x32x4f16 a[0:31], v[2:3], v[0:1], a[0:31] cbsz:1 abid:2 blgp:3
 ; LIT-SRCC-NEXT:    s_nop 7
 ; LIT-SRCC-NEXT:    s_nop 7
 ; LIT-SRCC-NEXT:    s_nop 1
@@ -1330,7 +1478,7 @@ define amdgpu_kernel void @test_mfma_f32_32x32x4f16(ptr addrspace(1) %arg, ptr a
 ; GFX90A-LABEL: test_mfma_f32_32x32x4f16:
 ; GFX90A:       ; %bb.0: ; %bb
 ; GFX90A-NEXT:    s_load_dwordx4 s[36:39], s[4:5], 0x24
-; GFX90A-NEXT:    v_mov_b32_e32 v4, 0
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX90A-NEXT:    s_load_dwordx16 s[0:15], s[36:37], 0x40
 ; GFX90A-NEXT:    s_load_dwordx16 s[16:31], s[36:37], 0x0
@@ -1345,8 +1493,8 @@ define amdgpu_kernel void @test_mfma_f32_32x32x4f16(ptr addrspace(1) %arg, ptr a
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a2, s18
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a3, s19
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_mov_b32_e32 v0, s0
-; GFX90A-NEXT:    v_mov_b32_e32 v1, s1
+; GFX90A-NEXT:    v_mov_b32_e32 v2, s0
+; GFX90A-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a4, s20
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a5, s21
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a6, s22
@@ -1371,27 +1519,27 @@ define amdgpu_kernel void @test_mfma_f32_32x32x4f16(ptr addrspace(1) %arg, ptr a
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a29, s13
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a30, s14
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a31, s15
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s2
-; GFX90A-NEXT:    v_mov_b32_e32 v3, s3
+; GFX90A-NEXT:    v_mov_b32_e32 v4, s2
+; GFX90A-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX90A-NEXT:    s_nop 1
-; GFX90A-NEXT:    v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[2:3], a[0:31] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    v_mfma_f32_32x32x4f16 a[0:31], v[2:3], v[4:5], a[0:31] cbsz:1 abid:2 blgp:3
 ; GFX90A-NEXT:    s_nop 7
 ; GFX90A-NEXT:    s_nop 7
 ; GFX90A-NEXT:    s_nop 2
-; GFX90A-NEXT:    global_store_dwordx4 v4, a[24:27], s[36:37] offset:96
-; GFX90A-NEXT:    global_store_dwordx4 v4, a[28:31], s[36:37] offset:112
-; GFX90A-NEXT:    global_store_dwordx4 v4, a[16:19], s[36:37] offset:64
-; GFX90A-NEXT:    global_store_dwordx4 v4, a[20:23], s[36:37] offset:80
-; GFX90A-NEXT:    global_store_dwordx4 v4, a[8:11], s[36:37] offset:32
-; GFX90A-NEXT:    global_store_dwordx4 v4, a[12:15], s[36:37] offset:48
-; GFX90A-NEXT:    global_store_dwordx4 v4, a[0:3], s[36:37]
-; GFX90A-NEXT:    global_store_dwordx4 v4, a[4:7], s[36:37] offset:16
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[24:27], s[36:37] offset:96
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[28:31], s[36:37] offset:112
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[16:19], s[36:37] offset:64
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[20:23], s[36:37] offset:80
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[8:11], s[36:37] offset:32
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[12:15], s[36:37] offset:48
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[36:37]
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[4:7], s[36:37] offset:16
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: test_mfma_f32_32x32x4f16:
 ; GFX942:       ; %bb.0: ; %bb
 ; GFX942-NEXT:    s_load_dwordx4 s[36:39], s[4:5], 0x24
-; GFX942-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-NEXT:    s_load_dwordx16 s[0:15], s[36:37], 0x40
 ; GFX942-NEXT:    s_load_dwordx16 s[16:31], s[36:37], 0x0
@@ -1406,8 +1554,8 @@ define amdgpu_kernel void @test_mfma_f32_32x32x4f16(ptr addrspace(1) %arg, ptr a
 ; GFX942-NEXT:    v_accvgpr_write_b32 a2, s18
 ; GFX942-NEXT:    v_accvgpr_write_b32 a3, s19
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b32_e32 v0, s0
-; GFX942-NEXT:    v_mov_b32_e32 v1, s1
+; GFX942-NEXT:    v_mov_b32_e32 v2, s0
+; GFX942-NEXT:    v_mov_b32_e32 v3, s1
 ; GFX942-NEXT:    v_accvgpr_write_b32 a4, s20
 ; GFX942-NEXT:    v_accvgpr_write_b32 a5, s21
 ; GFX942-NEXT:    v_accvgpr_write_b32 a6, s22
@@ -1432,22 +1580,83 @@ define amdgpu_kernel void @test_mfma_f32_32x32x4f16(ptr addrspace(1) %arg, ptr a
 ; GFX942-NEXT:    v_accvgpr_write_b32 a29, s13
 ; GFX942-NEXT:    v_accvgpr_write_b32 a30, s14
 ; GFX942-NEXT:    v_accvgpr_write_b32 a31, s15
-; GFX942-NEXT:    v_mov_b32_e32 v2, s2
-; GFX942-NEXT:    v_mov_b32_e32 v3, s3
+; GFX942-NEXT:    v_mov_b32_e32 v4, s2
+; GFX942-NEXT:    v_mov_b32_e32 v5, s3
 ; GFX942-NEXT:    s_nop 1
-; GFX942-NEXT:    v_mfma_f32_32x32x4_2b_f16 a[0:31], v[0:1], v[2:3], a[0:31] cbsz:1 abid:2 blgp:3
+; GFX942-NEXT:    v_mfma_f32_32x32x4_2b_f16 a[0:31], v[2:3], v[4:5], a[0:31] cbsz:1 abid:2 blgp:3
 ; GFX942-NEXT:    s_nop 7
 ; GFX942-NEXT:    s_nop 7
 ; GFX942-NEXT:    s_nop 2
-; GFX942-NEXT:    global_store_dwordx4 v4, a[24:27], s[36:37] offset:96
-; GFX942-NEXT:    global_store_dwordx4 v4, a[28:31], s[36:37] offset:112
-; GFX942-NEXT:    global_store_dwordx4 v4, a[16:19], s[36:37] offset:64
-; GFX942-NEXT:    global_store_dwordx4 v4, a[20:23], s[36:37] offset:80
-; GFX942-NEXT:    global_store_dwordx4 v4, a[8:11], s[36:37] offset:32
-; GFX942-NEXT:    global_store_dwordx4 v4, a[12:15], s[36:37] offset:48
-; GFX942-NEXT:    global_store_dwordx4 v4, a[0:3], s[36:37]
-; GFX942-NEXT:    global_store_dwordx4 v4, a[4:7], s[36:37] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v0, a[24:27], s[36:37] offset:96
+; GFX942-NEXT:    global_store_dwordx4 v0, a[28:31], s[36:37] offset:112
+; GFX942-NEXT:    global_store_dwordx4 v0, a[16:19], s[36:37] offset:64
+; GFX942-NEXT:    global_store_dwordx4 v0, a[20:23], s[36:37] offset:80
+; GFX942-NEXT:    global_store_dwordx4 v0, a[8:11], s[36:37] offset:32
+; GFX942-NEXT:    global_store_dwordx4 v0, a[12:15], s[36:37] offset:48
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[36:37]
+; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[36:37] offset:16
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_32x32x4f16:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[36:39], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v32, 0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx16 s[0:15], s[36:37], 0x40
+; GFX942-VGPR-NEXT:    s_load_dwordx16 s[16:31], s[36:37], 0x0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, s0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v17, s1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v18, s2
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v19, s3
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[38:39], 0x0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, s16
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, s17
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, s18
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, s19
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v34, s0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v35, s1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, s20
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, s21
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, s22
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v7, s23
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v8, s24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v9, s25
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v10, s26
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v11, s27
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v12, s28
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v13, s29
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v14, s30
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v15, s31
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v20, s4
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v21, s5
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v22, s6
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v23, s7
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v24, s8
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v25, s9
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v26, s10
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v27, s11
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v28, s12
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v29, s13
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v30, s14
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v31, s15
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v36, s2
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v37, s3
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_32x32x4_2b_f16 v[0:31], v[34:35], v[36:37], v[0:31] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 2
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[24:27], s[36:37] offset:96
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[28:31], s[36:37] offset:112
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[16:19], s[36:37] offset:64
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[20:23], s[36:37] offset:80
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[8:11], s[36:37] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[12:15], s[36:37] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[0:3], s[36:37]
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[4:7], s[36:37] offset:16
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <32 x float>, ptr addrspace(1) %arg
   %c.1 = load <4 x half>, ptr addrspace(1) %c
@@ -1676,6 +1885,36 @@ define amdgpu_kernel void @test_mfma_f32_16x16x4f16(ptr addrspace(1) %arg, ptr a
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_16x16x4f16:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[16:19], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[20:23], s[18:19], 0x0
+; GFX942-VGPR-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, s20
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v17, s21
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v18, s22
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v19, s23
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_16x16x4_4b_f16 v[0:15], v[16:17], v[18:19], v[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[12:15], s[16:17] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[8:11], s[16:17] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[4:7], s[16:17] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[0:3], s[16:17]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %c.1 = load <4 x half>, ptr addrspace(1) %c
@@ -1752,46 +1991,66 @@ define amdgpu_kernel void @test_mfma_f32_4x4x4f16(ptr addrspace(1) %arg, ptr add
 ; GFX90A-LABEL: test_mfma_f32_4x4x4f16:
 ; GFX90A:       ; %bb.0: ; %bb
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
-; GFX90A-NEXT:    v_mov_b32_e32 v4, 0
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX90A-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GFX90A-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x0
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_mov_b32_e32 v0, s4
-; GFX90A-NEXT:    v_mov_b32_e32 v1, s5
+; GFX90A-NEXT:    v_mov_b32_e32 v2, s4
+; GFX90A-NEXT:    v_mov_b32_e32 v3, s5
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a0, s8
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s6
-; GFX90A-NEXT:    v_mov_b32_e32 v3, s7
+; GFX90A-NEXT:    v_mov_b32_e32 v4, s6
+; GFX90A-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a1, s9
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a2, s10
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a3, s11
 ; GFX90A-NEXT:    s_nop 1
-; GFX90A-NEXT:    v_mfma_f32_4x4x4f16 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    v_mfma_f32_4x4x4f16 a[0:3], v[2:3], v[4:5], a[0:3] cbsz:1 abid:2 blgp:3
 ; GFX90A-NEXT:    s_nop 4
-; GFX90A-NEXT:    global_store_dwordx4 v4, a[0:3], s[0:1]
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: test_mfma_f32_4x4x4f16:
 ; GFX942:       ; %bb.0: ; %bb
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
-; GFX942-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GFX942-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x0
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b32_e32 v0, s4
-; GFX942-NEXT:    v_mov_b32_e32 v1, s5
+; GFX942-NEXT:    v_mov_b32_e32 v2, s4
+; GFX942-NEXT:    v_mov_b32_e32 v3, s5
 ; GFX942-NEXT:    v_accvgpr_write_b32 a0, s8
-; GFX942-NEXT:    v_mov_b32_e32 v2, s6
-; GFX942-NEXT:    v_mov_b32_e32 v3, s7
+; GFX942-NEXT:    v_mov_b32_e32 v4, s6
+; GFX942-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX942-NEXT:    v_accvgpr_write_b32 a1, s9
 ; GFX942-NEXT:    v_accvgpr_write_b32 a2, s10
 ; GFX942-NEXT:    v_accvgpr_write_b32 a3, s11
 ; GFX942-NEXT:    s_nop 1
-; GFX942-NEXT:    v_mfma_f32_4x4x4_16b_f16 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-NEXT:    v_mfma_f32_4x4x4_16b_f16 a[0:3], v[2:3], v[4:5], a[0:3] cbsz:1 abid:2 blgp:3
 ; GFX942-NEXT:    s_nop 4
-; GFX942-NEXT:    global_store_dwordx4 v4, a[0:3], s[0:1]
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_4x4x4f16:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, s4
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v7, s5
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v8, s6
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v9, s7
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_4x4x4_16b_f16 v[0:3], v[6:7], v[8:9], v[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_nop 4
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %c.1 = load <4 x half>, ptr addrspace(1) %c
@@ -2021,6 +2280,36 @@ define amdgpu_kernel void @test_mfma_f32_32x32x8f16(ptr addrspace(1) %arg, ptr a
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_32x32x8f16:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[16:19], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[20:23], s[18:19], 0x0
+; GFX942-VGPR-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, s20
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v17, s21
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v18, s22
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v19, s23
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_32x32x8_f16 v[0:15], v[16:17], v[18:19], v[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[12:15], s[16:17] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[8:11], s[16:17] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[4:7], s[16:17] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[0:3], s[16:17]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %c.1 = load <4 x half>, ptr addrspace(1) %c
@@ -2099,47 +2388,67 @@ define amdgpu_kernel void @test_mfma_f32_16x16x16f16(ptr addrspace(1) %arg, ptr
 ; GFX90A-LABEL: test_mfma_f32_16x16x16f16:
 ; GFX90A:       ; %bb.0: ; %bb
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
-; GFX90A-NEXT:    v_mov_b32_e32 v4, 0
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX90A-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GFX90A-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x0
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_mov_b32_e32 v0, s4
-; GFX90A-NEXT:    v_mov_b32_e32 v1, s5
+; GFX90A-NEXT:    v_mov_b32_e32 v2, s4
+; GFX90A-NEXT:    v_mov_b32_e32 v3, s5
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a0, s8
-; GFX90A-NEXT:    v_mov_b32_e32 v2, s6
-; GFX90A-NEXT:    v_mov_b32_e32 v3, s7
+; GFX90A-NEXT:    v_mov_b32_e32 v4, s6
+; GFX90A-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a1, s9
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a2, s10
 ; GFX90A-NEXT:    v_accvgpr_write_b32 a3, s11
 ; GFX90A-NEXT:    s_nop 1
-; GFX90A-NEXT:    v_mfma_f32_16x16x16f16 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    v_mfma_f32_16x16x16f16 a[0:3], v[2:3], v[4:5], a[0:3] cbsz:1 abid:2 blgp:3
 ; GFX90A-NEXT:    s_nop 7
 ; GFX90A-NEXT:    s_nop 2
-; GFX90A-NEXT:    global_store_dwordx4 v4, a[0:3], s[0:1]
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
 ; GFX90A-NEXT:    s_endpgm
 ;
 ; GFX942-LABEL: test_mfma_f32_16x16x16f16:
 ; GFX942:       ; %bb.0: ; %bb
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
-; GFX942-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX942-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
 ; GFX942-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x0
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b32_e32 v0, s4
-; GFX942-NEXT:    v_mov_b32_e32 v1, s5
+; GFX942-NEXT:    v_mov_b32_e32 v2, s4
+; GFX942-NEXT:    v_mov_b32_e32 v3, s5
 ; GFX942-NEXT:    v_accvgpr_write_b32 a0, s8
-; GFX942-NEXT:    v_mov_b32_e32 v2, s6
-; GFX942-NEXT:    v_mov_b32_e32 v3, s7
+; GFX942-NEXT:    v_mov_b32_e32 v4, s6
+; GFX942-NEXT:    v_mov_b32_e32 v5, s7
 ; GFX942-NEXT:    v_accvgpr_write_b32 a1, s9
 ; GFX942-NEXT:    v_accvgpr_write_b32 a2, s10
 ; GFX942-NEXT:    v_accvgpr_write_b32 a3, s11
 ; GFX942-NEXT:    s_nop 1
-; GFX942-NEXT:    v_mfma_f32_16x16x16_f16 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-NEXT:    v_mfma_f32_16x16x16_f16 a[0:3], v[2:3], v[4:5], a[0:3] cbsz:1 abid:2 blgp:3
 ; GFX942-NEXT:    s_nop 6
-; GFX942-NEXT:    global_store_dwordx4 v4, a[0:3], s[0:1]
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_16x16x16f16:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[4:7], s[2:3], 0x0
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[8:11], s[0:1], 0x0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, s4
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v7, s5
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v8, s6
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v9, s7
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_16x16x16_f16 v[0:3], v[6:7], v[8:9], v[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_nop 6
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %c.1 = load <4 x half>, ptr addrspace(1) %c
@@ -2508,6 +2817,63 @@ define amdgpu_kernel void @test_mfma_i32_32x32x4i8(ptr addrspace(1) %arg) #0 {
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[34:35]
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[34:35] offset:16
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_i32_32x32x4i8:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[34:35], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v33, 1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v34, 2
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v32, 0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx16 s[16:31], s[34:35], 0x0
+; GFX942-VGPR-NEXT:    s_load_dwordx16 s[0:15], s[34:35], 0x40
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, s16
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, s17
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, s18
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, s19
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, s20
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, s21
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, s22
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v7, s23
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v8, s24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v9, s25
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v10, s26
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v11, s27
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v12, s28
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v13, s29
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v14, s30
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v15, s31
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, s0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v17, s1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v18, s2
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v19, s3
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v20, s4
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v21, s5
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v22, s6
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v23, s7
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v24, s8
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v25, s9
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v26, s10
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v27, s11
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v28, s12
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v29, s13
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v30, s14
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v31, s15
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_i32_32x32x4_2b_i8 v[0:31], v33, v34, v[0:31] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 2
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[24:27], s[34:35] offset:96
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[28:31], s[34:35] offset:112
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[16:19], s[34:35] offset:64
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[20:23], s[34:35] offset:80
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[8:11], s[34:35] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[12:15], s[34:35] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[0:3], s[34:35]
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[4:7], s[34:35] offset:16
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <32 x i32>, ptr addrspace(1) %arg
   %mai.1 = tail call <32 x i32> @llvm.amdgcn.mfma.i32.32x32x4i8(i32 1, i32 2, <32 x i32> %in.1, i32 1, i32 2, i32 3)
@@ -2721,6 +3087,33 @@ define amdgpu_kernel void @test_mfma_i32_16x16x4i8(ptr addrspace(1) %arg) #0 {
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_i32_16x16x4i8:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, 1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v17, 2
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_i32_16x16x4_4b_i8 v[0:15], v16, v17, v[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[12:15], s[16:17] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[8:11], s[16:17] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[4:7], s[16:17] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[0:3], s[16:17]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x i32>, ptr addrspace(1) %arg
   %mai.1 = tail call <16 x i32> @llvm.amdgcn.mfma.i32.16x16x4i8(i32 1, i32 2, <16 x i32> %in.1, i32 1, i32 2, i32 3)
@@ -2728,6 +3121,143 @@ bb:
   ret void
 }
 
+define amdgpu_kernel void @test_mfma_i32_16x16x4i8_splatimm_src2_64(ptr addrspace(1) %arg) #0 {
+; NOLIT-SRCC-LABEL: test_mfma_i32_16x16x4i8_splatimm_src2_64:
+; NOLIT-SRCC:       ; %bb.0: ; %bb
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v0, 1
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a0, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a1, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a2, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a3, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a4, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a5, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a6, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a7, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a8, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a9, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a10, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a11, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a12, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a13, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a14, 64
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a15, 64
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v1, 2
+; NOLIT-SRCC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v16, 0
+; NOLIT-SRCC-NEXT:    v_mfma_i32_16x16x4i8 a[0:15], v0, v1, a[0:15] cbsz:1 abid:2 blgp:3
+; NOLIT-SRCC-NEXT:    s_nop 7
+; NOLIT-SRCC-NEXT:    s_nop 1
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v15, a15
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v14, a14
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v13, a13
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v12, a12
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v3, a3
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v2, a2
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v1, a1
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v0, a0
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v7, a7
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v6, a6
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v5, a5
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v4, a4
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v11, a11
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v10, a10
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v9, a9
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v8, a8
+; NOLIT-SRCC-NEXT:    s_waitcnt lgkmcnt(0)
+; NOLIT-SRCC-NEXT:    global_store_dwordx4 v16, v[12:15], s[0:1] offset:48
+; NOLIT-SRCC-NEXT:    global_store_dwordx4 v16, v[8:11], s[0:1] offset:32
+; NOLIT-SRCC-NEXT:    global_store_dwordx4 v16, v[4:7], s[0:1] offset:16
+; NOLIT-SRCC-NEXT:    global_store_dwordx4 v16, v[0:3], s[0:1]
+; NOLIT-SRCC-NEXT:    s_endpgm
+;
+; LIT-SRCC-LABEL: test_mfma_i32_16x16x4i8_splatimm_src2_64:
+; LIT-SRCC:       ; %bb.0: ; %bb
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v0, 1
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v1, 2
+; LIT-SRCC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v16, 0
+; LIT-SRCC-NEXT:    v_mfma_i32_16x16x4i8 a[0:15], v0, v1, 64 cbsz:1 abid:2 blgp:3
+; LIT-SRCC-NEXT:    s_nop 7
+; LIT-SRCC-NEXT:    s_nop 1
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v15, a15
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v14, a14
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v13, a13
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v12, a12
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v3, a3
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v2, a2
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v1, a1
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v0, a0
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v7, a7
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v6, a6
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v5, a5
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v4, a4
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v11, a11
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v10, a10
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v9, a9
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v8, a8
+; LIT-SRCC-NEXT:    s_waitcnt lgkmcnt(0)
+; LIT-SRCC-NEXT:    global_store_dwordx4 v16, v[12:15], s[0:1] offset:48
+; LIT-SRCC-NEXT:    global_store_dwordx4 v16, v[8:11], s[0:1] offset:32
+; LIT-SRCC-NEXT:    global_store_dwordx4 v16, v[4:7], s[0:1] offset:16
+; LIT-SRCC-NEXT:    global_store_dwordx4 v16, v[0:3], s[0:1]
+; LIT-SRCC-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: test_mfma_i32_16x16x4i8_splatimm_src2_64:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 1
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 2
+; GFX90A-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    v_mfma_i32_16x16x4i8 a[0:15], v0, v1, 64 cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_i32_16x16x4i8_splatimm_src2_64:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    v_mov_b32_e32 v0, 1
+; GFX942-NEXT:    v_mov_b32_e32 v1, 2
+; GFX942-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    v_mfma_i32_16x16x4_4b_i8 a[0:15], v0, v1, 64 cbsz:1 abid:2 blgp:3
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    s_nop 7
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    global_store_dwordx4 v0, a[12:15], s[0:1] offset:48
+; GFX942-NEXT:    global_store_dwordx4 v0, a[8:11], s[0:1] offset:32
+; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_i32_16x16x4i8_splatimm_src2_64:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, 2
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPR-NEXT:    v_mfma_i32_16x16x4_4b_i8 v[0:15], v0, v1, 64 cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[12:15], s[0:1] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[8:11], s[0:1] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[4:7], s[0:1] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
+bb:
+  %in.1 = load <16 x i32>, ptr addrspace(1) %arg
+  %mai.1 = tail call <16 x i32> @llvm.amdgcn.mfma.i32.16x16x4i8(i32 1, i32 2, <16 x i32> splat (i32 64), i32 1, i32 2, i32 3)
+  store <16 x i32> %mai.1, ptr addrspace(1) %arg
+  ret void
+}
+
 define amdgpu_kernel void @test_mfma_i32_4x4x4i8(ptr addrspace(1) %arg) #0 {
 ; NOLIT-SRCC-LABEL: test_mfma_i32_4x4x4i8:
 ; NOLIT-SRCC:       ; %bb.0: ; %bb
@@ -2822,6 +3352,23 @@ define amdgpu_kernel void @test_mfma_i32_4x4x4i8(ptr addrspace(1) %arg) #0 {
 ; GFX942-NEXT:    s_nop 4
 ; GFX942-NEXT:    global_store_dwordx4 v1, a[0:3], s[6:7]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_i32_4x4x4i8:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, 1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, 2
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, 0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_i32_4x4x4_16b_i8 v[0:3], v4, v6, v[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_nop 4
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v5, v[0:3], s[6:7]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x i32>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x i32> @llvm.amdgcn.mfma.i32.4x4x4i8(i32 1, i32 2, <4 x i32> %in.1, i32 1, i32 2, i32 3)
@@ -2829,6 +3376,197 @@ bb:
   ret void
 }
 
+define amdgpu_kernel void @test_mfma_i32_4x4x4i8_splat_imm_src2_1(ptr addrspace(1) %arg) #0 {
+; NOLIT-SRCC-LABEL: test_mfma_i32_4x4x4i8_splat_imm_src2_1:
+; NOLIT-SRCC:       ; %bb.0: ; %bb
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v0, 1
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a0, 1
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a1, 1
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a2, 1
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a3, 1
+; NOLIT-SRCC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v1, 2
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v4, 0
+; NOLIT-SRCC-NEXT:    s_nop 0
+; NOLIT-SRCC-NEXT:    v_mfma_i32_4x4x4i8 a[0:3], v0, v1, a[0:3] cbsz:1 abid:2 blgp:3
+; NOLIT-SRCC-NEXT:    s_nop 3
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v0, a0
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v1, a1
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v2, a2
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v3, a3
+; NOLIT-SRCC-NEXT:    s_waitcnt lgkmcnt(0)
+; NOLIT-SRCC-NEXT:    s_nop 0
+; NOLIT-SRCC-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; NOLIT-SRCC-NEXT:    s_endpgm
+;
+; LIT-SRCC-LABEL: test_mfma_i32_4x4x4i8_splat_imm_src2_1:
+; LIT-SRCC:       ; %bb.0: ; %bb
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v0, 1
+; LIT-SRCC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v1, 2
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v4, 0
+; LIT-SRCC-NEXT:    s_nop 0
+; LIT-SRCC-NEXT:    v_mfma_i32_4x4x4i8 a[0:3], v0, v1, 1 cbsz:1 abid:2 blgp:3
+; LIT-SRCC-NEXT:    s_nop 3
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v0, a0
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v1, a1
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v2, a2
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v3, a3
+; LIT-SRCC-NEXT:    s_waitcnt lgkmcnt(0)
+; LIT-SRCC-NEXT:    s_nop 0
+; LIT-SRCC-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; LIT-SRCC-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: test_mfma_i32_4x4x4i8_splat_imm_src2_1:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 1
+; GFX90A-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 2
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 0
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    v_mfma_i32_4x4x4i8 a[0:3], v0, v2, 1 cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_nop 3
+; GFX90A-NEXT:    global_store_dwordx4 v1, a[0:3], s[0:1]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_i32_4x4x4i8_splat_imm_src2_1:
+; GFX942:       ; %bb.0: ; %bb
+; GFX942-NEXT:    v_mov_b32_e32 v0, 1
+; GFX942-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-NEXT:    v_mov_b32_e32 v2, 2
+; GFX942-NEXT:    v_mov_b32_e32 v1, 0
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    v_mfma_i32_4x4x4_16b_i8 a[0:3], v0, v2, 1 cbsz:1 abid:2 blgp:3
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    s_nop 3
+; GFX942-NEXT:    global_store_dwordx4 v1, a[0:3], s[0:1]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_i32_4x4x4i8_splat_imm_src2_1:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 1
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, 2
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    v_mfma_i32_4x4x4_16b_i8 v[0:3], v0, v1, 1 cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_nop 3
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
+bb:
+  %in.1 = load <4 x i32>, ptr addrspace(1) %arg
+  %mai.1 = tail call <4 x i32> @llvm.amdgcn.mfma.i32.4x4x4i8(i32 1, i32 2, <4 x i32> splat (i32 1), i32 1, i32 2, i32 3)
+  store <4 x i32> %mai.1, ptr addrspace(1) %arg
+  ret void
+}
+
+define amdgpu_kernel void @test_mfma_i32_4x4x4i8_splat_k_src2_1(ptr addrspace(1) %arg) #0 {
+; NOLIT-SRCC-LABEL: test_mfma_i32_4x4x4i8_splat_k_src2_1:
+; NOLIT-SRCC:       ; %bb.0:
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v0, 0x41
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v1, 1
+; NOLIT-SRCC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a0, v0
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a1, v0
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a2, v0
+; NOLIT-SRCC-NEXT:    v_accvgpr_write_b32 a3, v0
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v0, 2
+; NOLIT-SRCC-NEXT:    v_mov_b32_e32 v4, 0
+; NOLIT-SRCC-NEXT:    s_nop 0
+; NOLIT-SRCC-NEXT:    v_mfma_i32_4x4x4i8 a[0:3], v1, v0, a[0:3] cbsz:1 abid:2 blgp:3
+; NOLIT-SRCC-NEXT:    s_nop 3
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v0, a0
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v1, a1
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v2, a2
+; NOLIT-SRCC-NEXT:    v_accvgpr_read_b32 v3, a3
+; NOLIT-SRCC-NEXT:    s_waitcnt lgkmcnt(0)
+; NOLIT-SRCC-NEXT:    s_nop 0
+; NOLIT-SRCC-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; NOLIT-SRCC-NEXT:    s_endpgm
+;
+; LIT-SRCC-LABEL: test_mfma_i32_4x4x4i8_splat_k_src2_1:
+; LIT-SRCC:       ; %bb.0:
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v0, 0x41
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v1, 1
+; LIT-SRCC-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; LIT-SRCC-NEXT:    v_accvgpr_write_b32 a0, v0
+; LIT-SRCC-NEXT:    v_accvgpr_write_b32 a1, v0
+; LIT-SRCC-NEXT:    v_accvgpr_write_b32 a2, v0
+; LIT-SRCC-NEXT:    v_accvgpr_write_b32 a3, v0
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v0, 2
+; LIT-SRCC-NEXT:    v_mov_b32_e32 v4, 0
+; LIT-SRCC-NEXT:    s_nop 0
+; LIT-SRCC-NEXT:    v_mfma_i32_4x4x4i8 a[0:3], v1, v0, a[0:3] cbsz:1 abid:2 blgp:3
+; LIT-SRCC-NEXT:    s_nop 3
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v0, a0
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v1, a1
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v2, a2
+; LIT-SRCC-NEXT:    v_accvgpr_read_b32 v3, a3
+; LIT-SRCC-NEXT:    s_waitcnt lgkmcnt(0)
+; LIT-SRCC-NEXT:    s_nop 0
+; LIT-SRCC-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; LIT-SRCC-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: test_mfma_i32_4x4x4i8_splat_k_src2_1:
+; GFX90A:       ; %bb.0:
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 0x41
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, v1
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 1
+; GFX90A-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a1, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a2, a0
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a3, a0
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 2
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    v_mfma_i32_4x4x4i8 a[0:3], v1, v2, a[0:3] cbsz:1 abid:2 blgp:3
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_nop 3
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX90A-NEXT:    s_endpgm
+;
+; GFX942-LABEL: test_mfma_i32_4x4x4i8_splat_k_src2_1:
+; GFX942:       ; %bb.0:
+; GFX942-NEXT:    v_mov_b32_e32 v1, 0x41
+; GFX942-NEXT:    v_accvgpr_write_b32 a0, v1
+; GFX942-NEXT:    v_mov_b32_e32 v1, 1
+; GFX942-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-NEXT:    v_accvgpr_mov_b32 a1, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a2, a0
+; GFX942-NEXT:    v_accvgpr_mov_b32 a3, a0
+; GFX942-NEXT:    v_mov_b32_e32 v2, 2
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-NEXT:    s_nop 0
+; GFX942-NEXT:    v_mfma_i32_4x4x4_16b_i8 a[0:3], v1, v2, a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-NEXT:    s_nop 3
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
+; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_i32_4x4x4i8_splat_k_src2_1:
+; GFX942-VGPR:       ; %bb.0:
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, 1
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 0x41
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, 2
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    v_mfma_i32_4x4x4_16b_i8 v[0:3], v5, v6, v[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_nop 3
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
+  %in.1 = load <4 x i32>, ptr addrspace(1) %arg
+  %mai.1 = tail call <4 x i32> @llvm.amdgcn.mfma.i32.4x4x4i8(i32 1, i32 2, <4 x i32> splat (i32 65), i32 1, i32 2, i32 3)
+  store <4 x i32> %mai.1, ptr addrspace(1) %arg
+  ret void
+}
+
 define amdgpu_kernel void @test_mfma_f32_32x32x1f32_forward_acc(ptr addrspace(1) %arg) #0 {
 ; NOLIT-SRCC-LABEL: test_mfma_f32_32x32x1f32_forward_acc:
 ; NOLIT-SRCC:       ; %bb.0: ; %bb
@@ -3219,6 +3957,64 @@ define amdgpu_kernel void @test_mfma_f32_32x32x1f32_forward_acc(ptr addrspace(1)
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[34:35]
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[34:35] offset:16
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_32x32x1f32_forward_acc:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[34:35], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v32, 1.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v33, 2.0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx16 s[16:31], s[34:35], 0x0
+; GFX942-VGPR-NEXT:    s_load_dwordx16 s[0:15], s[34:35], 0x40
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, s16
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, s17
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, s18
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, s19
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, s20
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, s21
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, s22
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v7, s23
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v8, s24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v9, s25
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v10, s26
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v11, s27
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v12, s28
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v13, s29
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v14, s30
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v15, s31
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, s0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v17, s1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v18, s2
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v19, s3
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v20, s4
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v21, s5
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v22, s6
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v23, s7
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v24, s8
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v25, s9
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v26, s10
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v27, s11
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v28, s12
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v29, s13
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v30, s14
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v31, s15
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_32x32x1_2b_f32 v[0:31], v32, v33, v[0:31]
+; GFX942-VGPR-NEXT:    v_mfma_f32_32x32x1_2b_f32 v[0:31], v32, v33, v[0:31]
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v32, 0
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[24:27], s[34:35] offset:96
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[28:31], s[34:35] offset:112
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[16:19], s[34:35] offset:64
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[20:23], s[34:35] offset:80
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[8:11], s[34:35] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[12:15], s[34:35] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[0:3], s[34:35]
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[4:7], s[34:35] offset:16
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <32 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <32 x float> @llvm.amdgcn.mfma.f32.32x32x1f32(float 1.0, float 2.0, <32 x float> %in.1, i32 0, i32 0, i32 0)
@@ -3435,6 +4231,34 @@ define amdgpu_kernel void @test_mfma_f32_16x16x1f32_forward_acc(ptr addrspace(1)
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_16x16x1f32_forward_acc:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, 1.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v17, 2.0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[4:5], s[4:5]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[6:7], s[6:7]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[12:13], s[12:13]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[14:15], s[14:15]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_16x16x1_4b_f32 v[0:15], v16, v17, v[0:15]
+; GFX942-VGPR-NEXT:    v_mfma_f32_16x16x1_4b_f32 v[0:15], v16, v17, v[0:15]
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[12:15], s[16:17] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[8:11], s[16:17] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[4:7], s[16:17] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[0:3], s[16:17]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <16 x float> @llvm.amdgcn.mfma.f32.16x16x1f32(float 1.0, float 2.0, <16 x float> %in.1, i32 0, i32 0, i32 0)
@@ -3542,6 +4366,25 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32_forward_acc(ptr addrspace(1) %
 ; GFX942-NEXT:    s_nop 3
 ; GFX942-NEXT:    global_store_dwordx4 v2, a[0:3], s[6:7]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_4x4x1f32_forward_acc:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, 1.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, 2.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, 0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_4x4x1_16b_f32 v[0:3], v4, v5, v[0:3]
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_4x4x1_16b_f32 v[0:3], v4, v5, v[0:3]
+; GFX942-VGPR-NEXT:    s_nop 3
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v6, v[0:3], s[6:7]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x float> @llvm.amdgcn.mfma.f32.4x4x1f32(float 1.0, float 2.0, <4 x float> %in.1, i32 0, i32 0, i32 0)
@@ -3616,6 +4459,19 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32_imm_splat(ptr addrspace(1) %ar
 ; GFX942-NEXT:    s_nop 2
 ; GFX942-NEXT:    global_store_dwordx4 v1, a[0:3], s[0:1]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_4x4x1f32_imm_splat:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    v_mfma_f32_4x4x1_16b_f32 v[0:3], v0, v1, 1.0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_nop 2
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x float> @llvm.amdgcn.mfma.f32.4x4x1f32(float 1.0, float 2.0, <4 x float> <float 1.0, float 1.0, float 1.0, float 1.0>, i32 0, i32 0, i32 0)
   store <4 x float> %mai.1, ptr addrspace(1) %arg
@@ -3745,6 +4601,22 @@ define amdgpu_kernel void @test_mfma_f32_16x16x1f32_imm_splat(ptr addrspace(1) %
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_16x16x1f32_imm_splat:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPR-NEXT:    v_mfma_f32_16x16x1_4b_f32 v[0:15], v0, v1, 1.0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[12:15], s[0:1] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[8:11], s[0:1] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[4:7], s[0:1] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <16 x float> @llvm.amdgcn.mfma.f32.16x16x1f32(float 1.0, float 2.0, <16 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, i32 0, i32 0, i32 0)
   store <16 x float> %mai.1, ptr addrspace(1) %arg
@@ -3885,6 +4757,24 @@ define amdgpu_kernel void @test_mfma_f32_32x32x8f16_imm_splat(ptr addrspace(1) %
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_32x32x8f16_imm_splat:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 0x3c003c00
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, 0x40004000
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, v2
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPR-NEXT:    v_mfma_f32_32x32x8_f16 v[0:15], v[0:1], v[2:3], 1.0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[12:15], s[0:1] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[8:11], s[0:1] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[4:7], s[0:1] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <16 x float> @llvm.amdgcn.mfma.f32.32x32x8f16(<4 x half> <half 1.0, half 1.0, half 1.0, half 1.0>, <4 x half> <half 2.0, half 2.0, half 2.0, half 2.0>, <16 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0>, i32 0, i32 0, i32 0)
   store <16 x float> %mai.1, ptr addrspace(1) %arg
@@ -4091,6 +4981,27 @@ define amdgpu_kernel void @test_mfma_f32_32x32x1f32_imm_splat(ptr addrspace(1) %
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_32x32x1f32_imm_splat:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v32, 0
+; GFX942-VGPR-NEXT:    v_mfma_f32_32x32x1_2b_f32 v[0:31], v0, v1, 0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[28:31], s[0:1] offset:112
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[24:27], s[0:1] offset:96
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[20:23], s[0:1] offset:80
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[16:19], s[0:1] offset:64
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[12:15], s[0:1] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[8:11], s[0:1] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[4:7], s[0:1] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <32 x float> @llvm.amdgcn.mfma.f32.32x32x1f32(float 1.0, float 2.0, <32 x float> <float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0>, i32 0, i32 0, i32 0)
   store <32 x float> %mai.1, ptr addrspace(1) %arg
@@ -4175,6 +5086,21 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32_imm(ptr addrspace(1) %arg) #0
 ; GFX942-NEXT:    s_nop 2
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_4x4x1f32_imm:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    v_mfma_f32_4x4x1_16b_f32 v[0:3], v0, v1, v[0:3]
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_nop 2
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <4 x float> @llvm.amdgcn.mfma.f32.4x4x1f32(float 1.0, float 2.0, <4 x float> <float 1.0, float 2.0, float 1.0, float 1.0>, i32 0, i32 0, i32 0)
   store <4 x float> %mai.1, ptr addrspace(1) %arg
@@ -4355,6 +5281,36 @@ define amdgpu_kernel void @test_mfma_f32_16x16x1f32_imm(ptr addrspace(1) %arg) #
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_16x16x1f32_imm:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v15, 2.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v7, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v8, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v9, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v10, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v11, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v12, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v13, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v14, v0
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, 0
+; GFX942-VGPR-NEXT:    v_mfma_f32_16x16x1_4b_f32 v[0:15], v0, v15, v[0:15]
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[12:15], s[0:1] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[8:11], s[0:1] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[4:7], s[0:1] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v16, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <16 x float> @llvm.amdgcn.mfma.f32.16x16x1f32(float 1.0, float 2.0, <16 x float> <float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 1.0, float 2.0>, i32 0, i32 0, i32 0)
   store <16 x float> %mai.1, ptr addrspace(1) %arg
@@ -4667,6 +5623,74 @@ define amdgpu_kernel void @test_mfma_f32_32x32x1f32_imm(ptr addrspace(1) %arg) #
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_32x32x1f32_imm:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, 0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v7, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v8, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v9, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v10, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v11, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v12, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v13, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v14, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v15, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v16, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v17, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v18, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v19, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v20, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v21, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v22, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v23, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v24, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v25, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v26, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v27, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v28, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v29, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v30, v1
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v31, v1
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[32:33], v[30:31]
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v34, 2.0
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[30:31], v[28:29]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[28:29], v[26:27]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[26:27], v[24:25]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[24:25], v[22:23]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[22:23], v[20:21]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[20:21], v[18:19]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[18:19], v[16:17]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[16:17], v[14:15]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[14:15], v[12:13]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[12:13], v[10:11]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[10:11], v[8:9]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[8:9], v[6:7]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[6:7], v[4:5]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[4:5], v[2:3]
+; GFX942-VGPR-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    v_mfma_f32_32x32x1_2b_f32 v[2:33], v0, v34, v[2:33]
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v1, v[30:33], s[0:1] offset:112
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v1, v[26:29], s[0:1] offset:96
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v1, v[22:25], s[0:1] offset:80
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v1, v[18:21], s[0:1] offset:64
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v1, v[14:17], s[0:1] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v1, v[10:13], s[0:1] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v1, v[6:9], s[0:1] offset:16
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v1, v[2:5], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %mai.1 = tail call <32 x float> @llvm.amdgcn.mfma.f32.32x32x1f32(float 1.0, float 2.0, <32 x float> <float 1.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0, float 0.0>, i32 0, i32 0, i32 0)
   store <32 x float> %mai.1, ptr addrspace(1) %arg
@@ -4755,6 +5779,24 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32_lit_splat(ptr addrspace(1) %ar
 ; GFX942-NEXT:    s_nop 2
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_4x4x1f32_lit_splat:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, 1.0
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
+; GFX942-VGPR-NEXT:    v_lshlrev_b32_e32 v4, 4, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 0x42f60000
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, 2.0
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    v_mfma_f32_4x4x1_16b_f32 v[0:3], v5, v6, v[0:3]
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_nop 2
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds <4 x float>, ptr addrspace(1) %arg, i32 %tid
@@ -4846,6 +5888,23 @@ define amdgpu_kernel void @test_mfma_f32_4x4x1f32_lit_splat_bad_code(ptr addrspa
 ; GFX942-NEXT:    s_nop 2
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_4x4x1f32_lit_splat_bad_code:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v5, 1.0
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v0, 0x42f60000
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v2, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v3, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v6, 2.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-VGPR-NEXT:    s_nop 0
+; GFX942-VGPR-NEXT:    v_mfma_f32_4x4x1_16b_f32 v[0:3], v5, v6, v[0:3]
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    s_nop 2
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds <4 x float>, ptr addrspace(1) %arg, i32 %tid
@@ -5109,6 +6168,37 @@ define amdgpu_kernel void @test_mfma_f32_32x32x1f32_vecarg(ptr addrspace(1) %arg
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[0:1]
 ; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[0:1] offset:16
 ; GFX942-NEXT:    s_endpgm
+;
+; GFX942-VGPR-LABEL: test_mfma_f32_32x32x1f32_vecarg:
+; GFX942-VGPR:       ; %bb.0: ; %bb
+; GFX942-VGPR-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX942-VGPR-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
+; GFX942-VGPR-NEXT:    v_lshlrev_b32_e32 v32, 7, v0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v33, 1.0
+; GFX942-VGPR-NEXT:    v_mov_b32_e32 v34, 2.0
+; GFX942-VGPR-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-VGPR-NEXT:    global_load_dwordx4 v[28:31], v32, s[0:1] offset:112
+; GFX942-VGPR-NEXT:    global_load_dwordx4 v[24:27], v32, s[0:1] offset:96
+; GFX942-VGPR-NEXT:    global_load_dwordx4 v[20:23], v32, s[0:1] offset:80
+; GFX942-VGPR-NEXT:    global_load_dwordx4 v[16:19], v32, s[0:1] offset:64
+; GFX942-VGPR-NEXT:    global_load_dwordx4 v[12:15], v32, s[0:1] offset:48
+; GFX942-VGPR-NEXT:    global_load_dwordx4 v[8:11], v32, s[0:1] offset:32
+; GFX942-VGPR-NEXT:    global_load_dwordx4 v[4:7], v32, s[0:1] offset:16
+; GFX942-VGPR-NEXT:    global_load_dwordx4 v[0:3], v32, s[0:1]
+; GFX942-VGPR-NEXT:    s_waitcnt vmcnt(0)
+; GFX942-VGPR-NEXT:    v_mfma_f32_32x32x1_2b_f32 v[0:31], v33, v34, v[0:31] cbsz:1 abid:2 blgp:3
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 7
+; GFX942-VGPR-NEXT:    s_nop 1
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[24:27], s[0:1] offset:96
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[28:31], s[0:1] offset:112
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[16:19], s[0:1] offset:64
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[20:23], s[0:1] offset:80
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[8:11], s[0:1] offset:32
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[12:15], s[0:1] offset:48
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[0:3], s[0:1]
+; GFX942-VGPR-NEXT:    global_store_dwordx4 v32, v[4:7], s[0:1] offset:16
+; GFX942-VGPR-NEXT:    s_endpgm
 bb:
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %gep = getelementptr inbounds <32 x float>, ptr addrspace(1) %arg, i32 %tid
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.16x16x128.f8f6f4.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.16x16x128.f8f6f4.ll
index 04ee0bb..37809da 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.16x16x128.f8f6f4.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.16x16x128.f8f6f4.ll
@@ -1485,30 +1485,30 @@ define <4 x float> @test_mfma_scale_f32_16x16x128_f8f6f4_0_0_sgprs(<8 x i32> inr
 ; SDAG-LABEL: test_mfma_scale_f32_16x16x128_f8f6f4_0_0_sgprs:
 ; SDAG:       ; %bb.0:
 ; SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v12, s0
-; SDAG-NEXT:    v_mov_b32_e32 v13, s1
-; SDAG-NEXT:    v_mov_b32_e32 v14, s2
-; SDAG-NEXT:    v_mov_b32_e32 v15, s3
-; SDAG-NEXT:    v_mov_b32_e32 v16, s16
-; SDAG-NEXT:    v_mov_b32_e32 v17, s17
-; SDAG-NEXT:    v_mov_b32_e32 v18, s18
-; SDAG-NEXT:    v_mov_b32_e32 v19, s19
-; SDAG-NEXT:    v_mov_b32_e32 v20, s28
-; SDAG-NEXT:    v_mov_b32_e32 v21, s29
-; SDAG-NEXT:    v_mov_b32_e32 v4, s20
-; SDAG-NEXT:    v_mov_b32_e32 v5, s21
-; SDAG-NEXT:    v_mov_b32_e32 v6, s22
-; SDAG-NEXT:    v_mov_b32_e32 v7, s23
-; SDAG-NEXT:    v_mov_b32_e32 v8, s24
-; SDAG-NEXT:    v_mov_b32_e32 v9, s25
-; SDAG-NEXT:    v_mov_b32_e32 v10, s26
-; SDAG-NEXT:    v_mov_b32_e32 v11, s27
-; SDAG-NEXT:    v_accvgpr_write_b32 a0, v20
-; SDAG-NEXT:    v_accvgpr_write_b32 a1, v21
+; SDAG-NEXT:    v_mov_b32_e32 v14, s0
+; SDAG-NEXT:    v_mov_b32_e32 v15, s1
+; SDAG-NEXT:    v_mov_b32_e32 v16, s2
+; SDAG-NEXT:    v_mov_b32_e32 v17, s3
+; SDAG-NEXT:    v_mov_b32_e32 v18, s16
+; SDAG-NEXT:    v_mov_b32_e32 v19, s17
+; SDAG-NEXT:    v_mov_b32_e32 v20, s18
+; SDAG-NEXT:    v_mov_b32_e32 v21, s19
+; SDAG-NEXT:    v_mov_b32_e32 v4, s28
+; SDAG-NEXT:    v_mov_b32_e32 v5, s29
+; SDAG-NEXT:    v_mov_b32_e32 v6, s20
+; SDAG-NEXT:    v_mov_b32_e32 v7, s21
+; SDAG-NEXT:    v_mov_b32_e32 v8, s22
+; SDAG-NEXT:    v_mov_b32_e32 v9, s23
+; SDAG-NEXT:    v_mov_b32_e32 v10, s24
+; SDAG-NEXT:    v_mov_b32_e32 v11, s25
+; SDAG-NEXT:    v_mov_b32_e32 v12, s26
+; SDAG-NEXT:    v_mov_b32_e32 v13, s27
+; SDAG-NEXT:    v_accvgpr_write_b32 a0, v4
+; SDAG-NEXT:    v_accvgpr_write_b32 a1, v5
 ; SDAG-NEXT:    v_accvgpr_write_b32 a2, v0
 ; SDAG-NEXT:    v_accvgpr_write_b32 a3, v1
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[12:19], v[4:11], a[0:3], v2, v3 op_sel_hi:[0,0,0]
+; SDAG-NEXT:    v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[14:21], v[6:13], a[0:3], v2, v3 op_sel_hi:[0,0,0]
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    s_nop 3
 ; SDAG-NEXT:    v_accvgpr_read_b32 v0, a0
@@ -1895,36 +1895,36 @@ define amdgpu_kernel void @test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd(<8 x i32
 ; SDAG-LABEL: test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd:
 ; SDAG:       ; %bb.0:
 ; SDAG-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x0
-; SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; SDAG-NEXT:    v_mov_b32_e32 v0, 0
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v0, s8
-; SDAG-NEXT:    v_mov_b32_e32 v1, s9
-; SDAG-NEXT:    v_mov_b32_e32 v2, s10
-; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    v_mov_b32_e32 v4, s12
-; SDAG-NEXT:    v_mov_b32_e32 v5, s13
-; SDAG-NEXT:    v_mov_b32_e32 v6, s14
-; SDAG-NEXT:    v_mov_b32_e32 v7, s15
+; SDAG-NEXT:    v_mov_b32_e32 v2, s8
+; SDAG-NEXT:    v_mov_b32_e32 v3, s9
+; SDAG-NEXT:    v_mov_b32_e32 v4, s10
+; SDAG-NEXT:    v_mov_b32_e32 v5, s11
+; SDAG-NEXT:    v_mov_b32_e32 v6, s12
+; SDAG-NEXT:    v_mov_b32_e32 v7, s13
+; SDAG-NEXT:    v_mov_b32_e32 v8, s14
+; SDAG-NEXT:    v_mov_b32_e32 v9, s15
 ; SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x40
-; SDAG-NEXT:    v_mov_b32_e32 v8, s16
-; SDAG-NEXT:    v_mov_b32_e32 v9, s17
-; SDAG-NEXT:    v_mov_b32_e32 v10, s18
-; SDAG-NEXT:    v_mov_b32_e32 v11, s19
-; SDAG-NEXT:    v_mov_b32_e32 v12, s20
-; SDAG-NEXT:    v_mov_b32_e32 v13, s21
-; SDAG-NEXT:    v_mov_b32_e32 v14, s22
-; SDAG-NEXT:    v_mov_b32_e32 v15, s23
+; SDAG-NEXT:    v_mov_b32_e32 v10, s16
+; SDAG-NEXT:    v_mov_b32_e32 v11, s17
+; SDAG-NEXT:    v_mov_b32_e32 v12, s18
+; SDAG-NEXT:    v_mov_b32_e32 v13, s19
+; SDAG-NEXT:    v_mov_b32_e32 v14, s20
+; SDAG-NEXT:    v_mov_b32_e32 v15, s21
+; SDAG-NEXT:    v_mov_b32_e32 v16, s22
+; SDAG-NEXT:    v_mov_b32_e32 v17, s23
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
 ; SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
 ; SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
 ; SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
 ; SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
-; SDAG-NEXT:    v_mov_b32_e32 v17, s13
+; SDAG-NEXT:    v_mov_b32_e32 v1, s13
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[0:7], v[8:15], a[0:3], s12, v17 op_sel:[1,1,0] op_sel_hi:[1,0,0] blgp:2
+; SDAG-NEXT:    v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[2:9], v[10:17], a[0:3], s12, v1 op_sel:[1,1,0] op_sel_hi:[1,0,0] blgp:2
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    s_nop 3
-; SDAG-NEXT:    global_store_dwordx4 v16, a[0:3], s[14:15]
+; SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[14:15]
 ; SDAG-NEXT:    s_endpgm
 ;
 ; GISEL-LABEL: test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd:
@@ -1964,33 +1964,33 @@ define amdgpu_kernel void @test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA
 ; SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x40
 ; SDAG-NEXT:    s_movk_i32 s6, 0x41
 ; SDAG-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x50
-; SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; SDAG-NEXT:    v_mov_b32_e32 v0, 0
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v0, s8
-; SDAG-NEXT:    v_mov_b32_e32 v1, s9
-; SDAG-NEXT:    v_mov_b32_e32 v2, s10
-; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    v_mov_b32_e32 v4, s12
-; SDAG-NEXT:    v_mov_b32_e32 v5, s13
-; SDAG-NEXT:    v_mov_b32_e32 v6, s14
-; SDAG-NEXT:    v_mov_b32_e32 v7, s15
+; SDAG-NEXT:    v_mov_b32_e32 v2, s8
+; SDAG-NEXT:    v_mov_b32_e32 v3, s9
+; SDAG-NEXT:    v_mov_b32_e32 v4, s10
+; SDAG-NEXT:    v_mov_b32_e32 v5, s11
+; SDAG-NEXT:    v_mov_b32_e32 v6, s12
+; SDAG-NEXT:    v_mov_b32_e32 v7, s13
+; SDAG-NEXT:    v_mov_b32_e32 v8, s14
+; SDAG-NEXT:    v_mov_b32_e32 v9, s15
 ; SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
-; SDAG-NEXT:    v_mov_b32_e32 v8, s16
-; SDAG-NEXT:    v_mov_b32_e32 v9, s17
-; SDAG-NEXT:    v_mov_b32_e32 v10, s18
-; SDAG-NEXT:    v_mov_b32_e32 v11, s19
-; SDAG-NEXT:    v_mov_b32_e32 v12, s20
-; SDAG-NEXT:    v_mov_b32_e32 v13, s21
-; SDAG-NEXT:    v_mov_b32_e32 v14, s22
-; SDAG-NEXT:    v_mov_b32_e32 v15, s23
+; SDAG-NEXT:    v_mov_b32_e32 v10, s16
+; SDAG-NEXT:    v_mov_b32_e32 v11, s17
+; SDAG-NEXT:    v_mov_b32_e32 v12, s18
+; SDAG-NEXT:    v_mov_b32_e32 v13, s19
+; SDAG-NEXT:    v_mov_b32_e32 v14, s20
+; SDAG-NEXT:    v_mov_b32_e32 v15, s21
+; SDAG-NEXT:    v_mov_b32_e32 v16, s22
+; SDAG-NEXT:    v_mov_b32_e32 v17, s23
 ; SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
 ; SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
 ; SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[0:7], v[8:15], a[0:3], s6, -2 op_sel:[1,1,0] op_sel_hi:[1,0,0]
+; SDAG-NEXT:    v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[2:9], v[10:17], a[0:3], s6, -2 op_sel:[1,1,0] op_sel_hi:[1,0,0]
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    s_nop 3
-; SDAG-NEXT:    global_store_dwordx4 v16, a[0:3], s[4:5]
+; SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[4:5]
 ; SDAG-NEXT:    s_endpgm
 ;
 ; GISEL-LABEL: test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA_kimm__scaleB__inlineimm:
@@ -2031,33 +2031,33 @@ define amdgpu_kernel void @test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA
 ; SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x40
 ; SDAG-NEXT:    s_movk_i32 s6, 0x41
 ; SDAG-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x50
-; SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; SDAG-NEXT:    v_mov_b32_e32 v0, 0
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v0, s8
-; SDAG-NEXT:    v_mov_b32_e32 v1, s9
-; SDAG-NEXT:    v_mov_b32_e32 v2, s10
-; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    v_mov_b32_e32 v4, s12
-; SDAG-NEXT:    v_mov_b32_e32 v5, s13
-; SDAG-NEXT:    v_mov_b32_e32 v6, s14
-; SDAG-NEXT:    v_mov_b32_e32 v7, s15
+; SDAG-NEXT:    v_mov_b32_e32 v2, s8
+; SDAG-NEXT:    v_mov_b32_e32 v3, s9
+; SDAG-NEXT:    v_mov_b32_e32 v4, s10
+; SDAG-NEXT:    v_mov_b32_e32 v5, s11
+; SDAG-NEXT:    v_mov_b32_e32 v6, s12
+; SDAG-NEXT:    v_mov_b32_e32 v7, s13
+; SDAG-NEXT:    v_mov_b32_e32 v8, s14
+; SDAG-NEXT:    v_mov_b32_e32 v9, s15
 ; SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
-; SDAG-NEXT:    v_mov_b32_e32 v8, s16
-; SDAG-NEXT:    v_mov_b32_e32 v9, s17
-; SDAG-NEXT:    v_mov_b32_e32 v10, s18
-; SDAG-NEXT:    v_mov_b32_e32 v11, s19
-; SDAG-NEXT:    v_mov_b32_e32 v12, s20
-; SDAG-NEXT:    v_mov_b32_e32 v13, s21
-; SDAG-NEXT:    v_mov_b32_e32 v14, s22
-; SDAG-NEXT:    v_mov_b32_e32 v15, s23
+; SDAG-NEXT:    v_mov_b32_e32 v10, s16
+; SDAG-NEXT:    v_mov_b32_e32 v11, s17
+; SDAG-NEXT:    v_mov_b32_e32 v12, s18
+; SDAG-NEXT:    v_mov_b32_e32 v13, s19
+; SDAG-NEXT:    v_mov_b32_e32 v14, s20
+; SDAG-NEXT:    v_mov_b32_e32 v15, s21
+; SDAG-NEXT:    v_mov_b32_e32 v16, s22
+; SDAG-NEXT:    v_mov_b32_e32 v17, s23
 ; SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
 ; SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
 ; SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[0:7], v[8:15], a[0:3], s6, 1.0 op_sel:[1,1,0] op_sel_hi:[1,0,0]
+; SDAG-NEXT:    v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[2:9], v[10:17], a[0:3], s6, 1.0 op_sel:[1,1,0] op_sel_hi:[1,0,0]
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    s_nop 3
-; SDAG-NEXT:    global_store_dwordx4 v16, a[0:3], s[4:5]
+; SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[4:5]
 ; SDAG-NEXT:    s_endpgm
 ;
 ; GISEL-LABEL: test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA_kimm__scaleB__FP_literal:
@@ -2096,34 +2096,34 @@ define amdgpu_kernel void @test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA
 ; SDAG:       ; %bb.0:
 ; SDAG-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x0
 ; SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x40
-; SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; SDAG-NEXT:    v_mov_b32_e32 v0, 0
 ; SDAG-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x50
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v0, s8
-; SDAG-NEXT:    v_mov_b32_e32 v1, s9
-; SDAG-NEXT:    v_mov_b32_e32 v2, s10
-; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    v_mov_b32_e32 v4, s12
-; SDAG-NEXT:    v_mov_b32_e32 v5, s13
-; SDAG-NEXT:    v_mov_b32_e32 v6, s14
-; SDAG-NEXT:    v_mov_b32_e32 v7, s15
+; SDAG-NEXT:    v_mov_b32_e32 v2, s8
+; SDAG-NEXT:    v_mov_b32_e32 v3, s9
+; SDAG-NEXT:    v_mov_b32_e32 v4, s10
+; SDAG-NEXT:    v_mov_b32_e32 v5, s11
+; SDAG-NEXT:    v_mov_b32_e32 v6, s12
+; SDAG-NEXT:    v_mov_b32_e32 v7, s13
+; SDAG-NEXT:    v_mov_b32_e32 v8, s14
+; SDAG-NEXT:    v_mov_b32_e32 v9, s15
 ; SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
-; SDAG-NEXT:    v_mov_b32_e32 v8, s16
-; SDAG-NEXT:    v_mov_b32_e32 v9, s17
-; SDAG-NEXT:    v_mov_b32_e32 v10, s18
-; SDAG-NEXT:    v_mov_b32_e32 v11, s19
-; SDAG-NEXT:    v_mov_b32_e32 v12, s20
-; SDAG-NEXT:    v_mov_b32_e32 v13, s21
-; SDAG-NEXT:    v_mov_b32_e32 v14, s22
-; SDAG-NEXT:    v_mov_b32_e32 v15, s23
+; SDAG-NEXT:    v_mov_b32_e32 v10, s16
+; SDAG-NEXT:    v_mov_b32_e32 v11, s17
+; SDAG-NEXT:    v_mov_b32_e32 v12, s18
+; SDAG-NEXT:    v_mov_b32_e32 v13, s19
+; SDAG-NEXT:    v_mov_b32_e32 v14, s20
+; SDAG-NEXT:    v_mov_b32_e32 v15, s21
+; SDAG-NEXT:    v_mov_b32_e32 v16, s22
+; SDAG-NEXT:    v_mov_b32_e32 v17, s23
 ; SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
 ; SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
 ; SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[0:7], v[8:15], a[0:3], 1.0, -2 op_sel:[1,1,0] op_sel_hi:[1,0,0]
+; SDAG-NEXT:    v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[2:9], v[10:17], a[0:3], 1.0, -2 op_sel:[1,1,0] op_sel_hi:[1,0,0]
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    s_nop 3
-; SDAG-NEXT:    global_store_dwordx4 v16, a[0:3], s[4:5]
+; SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[4:5]
 ; SDAG-NEXT:    s_endpgm
 ;
 ; GISEL-LABEL: test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA_FP_literal__scaleB__inline_imm:
@@ -2162,34 +2162,34 @@ define amdgpu_kernel void @test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA
 ; SDAG:       ; %bb.0:
 ; SDAG-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x0
 ; SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x40
-; SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; SDAG-NEXT:    v_mov_b32_e32 v0, 0
 ; SDAG-NEXT:    s_load_dwordx2 s[4:5], s[4:5], 0x50
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v0, s8
-; SDAG-NEXT:    v_mov_b32_e32 v1, s9
-; SDAG-NEXT:    v_mov_b32_e32 v2, s10
-; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    v_mov_b32_e32 v4, s12
-; SDAG-NEXT:    v_mov_b32_e32 v5, s13
-; SDAG-NEXT:    v_mov_b32_e32 v6, s14
-; SDAG-NEXT:    v_mov_b32_e32 v7, s15
+; SDAG-NEXT:    v_mov_b32_e32 v2, s8
+; SDAG-NEXT:    v_mov_b32_e32 v3, s9
+; SDAG-NEXT:    v_mov_b32_e32 v4, s10
+; SDAG-NEXT:    v_mov_b32_e32 v5, s11
+; SDAG-NEXT:    v_mov_b32_e32 v6, s12
+; SDAG-NEXT:    v_mov_b32_e32 v7, s13
+; SDAG-NEXT:    v_mov_b32_e32 v8, s14
+; SDAG-NEXT:    v_mov_b32_e32 v9, s15
 ; SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
-; SDAG-NEXT:    v_mov_b32_e32 v8, s16
-; SDAG-NEXT:    v_mov_b32_e32 v9, s17
-; SDAG-NEXT:    v_mov_b32_e32 v10, s18
-; SDAG-NEXT:    v_mov_b32_e32 v11, s19
-; SDAG-NEXT:    v_mov_b32_e32 v12, s20
-; SDAG-NEXT:    v_mov_b32_e32 v13, s21
-; SDAG-NEXT:    v_mov_b32_e32 v14, s22
-; SDAG-NEXT:    v_mov_b32_e32 v15, s23
+; SDAG-NEXT:    v_mov_b32_e32 v10, s16
+; SDAG-NEXT:    v_mov_b32_e32 v11, s17
+; SDAG-NEXT:    v_mov_b32_e32 v12, s18
+; SDAG-NEXT:    v_mov_b32_e32 v13, s19
+; SDAG-NEXT:    v_mov_b32_e32 v14, s20
+; SDAG-NEXT:    v_mov_b32_e32 v15, s21
+; SDAG-NEXT:    v_mov_b32_e32 v16, s22
+; SDAG-NEXT:    v_mov_b32_e32 v17, s23
 ; SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
 ; SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
 ; SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[0:7], v[8:15], a[0:3], 1.0, 0.15915494 op_sel:[1,1,0] op_sel_hi:[1,0,0]
+; SDAG-NEXT:    v_mfma_scale_f32_16x16x128_f8f6f4 a[0:3], v[2:9], v[10:17], a[0:3], 1.0, 0.15915494 op_sel:[1,1,0] op_sel_hi:[1,0,0]
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    s_nop 3
-; SDAG-NEXT:    global_store_dwordx4 v16, a[0:3], s[4:5]
+; SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[4:5]
 ; SDAG-NEXT:    s_endpgm
 ;
 ; GISEL-LABEL: test_mfma_scale_f32_16x16x128_f8f6f4__vgprcd___scaleA_FP_literal__scaleB__FP_literal:
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.ll
index 91197f9..bc50058 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.scale.f32.32x32x64.f8f6f4.ll
@@ -3515,26 +3515,26 @@ define <16 x float> @test_mfma_scale_f32_32x32x64_f8f6f4_0_0_sgprs(<8 x i32> inr
 ; SDAG-LABEL: test_mfma_scale_f32_32x32x64_f8f6f4_0_0_sgprs:
 ; SDAG:       ; %bb.0:
 ; SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v24, s0
-; SDAG-NEXT:    v_mov_b32_e32 v25, s1
-; SDAG-NEXT:    v_mov_b32_e32 v26, s2
-; SDAG-NEXT:    v_mov_b32_e32 v27, s3
-; SDAG-NEXT:    v_mov_b32_e32 v28, s16
-; SDAG-NEXT:    v_mov_b32_e32 v29, s17
-; SDAG-NEXT:    v_mov_b32_e32 v30, s18
-; SDAG-NEXT:    v_mov_b32_e32 v31, s19
-; SDAG-NEXT:    v_mov_b32_e32 v32, s28
-; SDAG-NEXT:    v_mov_b32_e32 v33, s29
-; SDAG-NEXT:    v_mov_b32_e32 v16, s20
-; SDAG-NEXT:    v_mov_b32_e32 v17, s21
-; SDAG-NEXT:    v_mov_b32_e32 v18, s22
-; SDAG-NEXT:    v_mov_b32_e32 v19, s23
-; SDAG-NEXT:    v_mov_b32_e32 v20, s24
-; SDAG-NEXT:    v_mov_b32_e32 v21, s25
-; SDAG-NEXT:    v_mov_b32_e32 v22, s26
-; SDAG-NEXT:    v_mov_b32_e32 v23, s27
-; SDAG-NEXT:    v_accvgpr_write_b32 a0, v32
-; SDAG-NEXT:    v_accvgpr_write_b32 a1, v33
+; SDAG-NEXT:    v_mov_b32_e32 v26, s0
+; SDAG-NEXT:    v_mov_b32_e32 v27, s1
+; SDAG-NEXT:    v_mov_b32_e32 v28, s2
+; SDAG-NEXT:    v_mov_b32_e32 v29, s3
+; SDAG-NEXT:    v_mov_b32_e32 v30, s16
+; SDAG-NEXT:    v_mov_b32_e32 v31, s17
+; SDAG-NEXT:    v_mov_b32_e32 v32, s18
+; SDAG-NEXT:    v_mov_b32_e32 v33, s19
+; SDAG-NEXT:    v_mov_b32_e32 v16, s28
+; SDAG-NEXT:    v_mov_b32_e32 v17, s29
+; SDAG-NEXT:    v_mov_b32_e32 v18, s20
+; SDAG-NEXT:    v_mov_b32_e32 v19, s21
+; SDAG-NEXT:    v_mov_b32_e32 v20, s22
+; SDAG-NEXT:    v_mov_b32_e32 v21, s23
+; SDAG-NEXT:    v_mov_b32_e32 v22, s24
+; SDAG-NEXT:    v_mov_b32_e32 v23, s25
+; SDAG-NEXT:    v_mov_b32_e32 v24, s26
+; SDAG-NEXT:    v_mov_b32_e32 v25, s27
+; SDAG-NEXT:    v_accvgpr_write_b32 a0, v16
+; SDAG-NEXT:    v_accvgpr_write_b32 a1, v17
 ; SDAG-NEXT:    v_accvgpr_write_b32 a2, v0
 ; SDAG-NEXT:    v_accvgpr_write_b32 a3, v1
 ; SDAG-NEXT:    v_accvgpr_write_b32 a4, v2
@@ -3550,7 +3550,7 @@ define <16 x float> @test_mfma_scale_f32_32x32x64_f8f6f4_0_0_sgprs(<8 x i32> inr
 ; SDAG-NEXT:    v_accvgpr_write_b32 a14, v12
 ; SDAG-NEXT:    v_accvgpr_write_b32 a15, v13
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_mfma_scale_f32_32x32x64_f8f6f4 a[0:15], v[24:31], v[16:23], a[0:15], v14, v15 op_sel_hi:[0,0,0]
+; SDAG-NEXT:    v_mfma_scale_f32_32x32x64_f8f6f4 a[0:15], v[26:33], v[18:25], a[0:15], v14, v15 op_sel_hi:[0,0,0]
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    s_nop 3
@@ -3993,34 +3993,34 @@ define <16 x float> @test_mfma_scale_f32_32x32x64_f8f6f4_0_0_sgpr_vgpr_sgpr__vgp
 ; SDAG-LABEL: test_mfma_scale_f32_32x32x64_f8f6f4_0_0_sgpr_vgpr_sgpr__vgpr_sgpr:
 ; SDAG:       ; %bb.0:
 ; SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v16, s0
-; SDAG-NEXT:    v_mov_b32_e32 v17, s1
-; SDAG-NEXT:    v_mov_b32_e32 v18, s2
-; SDAG-NEXT:    v_mov_b32_e32 v19, s3
-; SDAG-NEXT:    v_mov_b32_e32 v20, s16
-; SDAG-NEXT:    v_mov_b32_e32 v21, s17
-; SDAG-NEXT:    v_mov_b32_e32 v22, s18
-; SDAG-NEXT:    v_mov_b32_e32 v23, s19
-; SDAG-NEXT:    v_mov_b32_e32 v24, s20
-; SDAG-NEXT:    v_mov_b32_e32 v25, s21
-; SDAG-NEXT:    v_mov_b32_e32 v26, s22
-; SDAG-NEXT:    v_mov_b32_e32 v27, s23
-; SDAG-NEXT:    v_mov_b32_e32 v28, s24
-; SDAG-NEXT:    v_mov_b32_e32 v29, s25
-; SDAG-NEXT:    v_mov_b32_e32 v30, s26
-; SDAG-NEXT:    v_mov_b32_e32 v31, s27
-; SDAG-NEXT:    v_mov_b32_e32 v32, s28
-; SDAG-NEXT:    v_mov_b32_e32 v33, s29
-; SDAG-NEXT:    v_accvgpr_write_b32 a0, v24
-; SDAG-NEXT:    v_accvgpr_write_b32 a1, v25
-; SDAG-NEXT:    v_accvgpr_write_b32 a2, v26
-; SDAG-NEXT:    v_accvgpr_write_b32 a3, v27
-; SDAG-NEXT:    v_accvgpr_write_b32 a4, v28
-; SDAG-NEXT:    v_accvgpr_write_b32 a5, v29
-; SDAG-NEXT:    v_accvgpr_write_b32 a6, v30
-; SDAG-NEXT:    v_accvgpr_write_b32 a7, v31
-; SDAG-NEXT:    v_accvgpr_write_b32 a8, v32
-; SDAG-NEXT:    v_accvgpr_write_b32 a9, v33
+; SDAG-NEXT:    v_mov_b32_e32 v26, s0
+; SDAG-NEXT:    v_mov_b32_e32 v27, s1
+; SDAG-NEXT:    v_mov_b32_e32 v28, s2
+; SDAG-NEXT:    v_mov_b32_e32 v29, s3
+; SDAG-NEXT:    v_mov_b32_e32 v30, s16
+; SDAG-NEXT:    v_mov_b32_e32 v31, s17
+; SDAG-NEXT:    v_mov_b32_e32 v32, s18
+; SDAG-NEXT:    v_mov_b32_e32 v33, s19
+; SDAG-NEXT:    v_mov_b32_e32 v16, s20
+; SDAG-NEXT:    v_mov_b32_e32 v17, s21
+; SDAG-NEXT:    v_mov_b32_e32 v18, s22
+; SDAG-NEXT:    v_mov_b32_e32 v19, s23
+; SDAG-NEXT:    v_mov_b32_e32 v20, s24
+; SDAG-NEXT:    v_mov_b32_e32 v21, s25
+; SDAG-NEXT:    v_mov_b32_e32 v22, s26
+; SDAG-NEXT:    v_mov_b32_e32 v23, s27
+; SDAG-NEXT:    v_mov_b32_e32 v24, s28
+; SDAG-NEXT:    v_mov_b32_e32 v25, s29
+; SDAG-NEXT:    v_accvgpr_write_b32 a0, v16
+; SDAG-NEXT:    v_accvgpr_write_b32 a1, v17
+; SDAG-NEXT:    v_accvgpr_write_b32 a2, v18
+; SDAG-NEXT:    v_accvgpr_write_b32 a3, v19
+; SDAG-NEXT:    v_accvgpr_write_b32 a4, v20
+; SDAG-NEXT:    v_accvgpr_write_b32 a5, v21
+; SDAG-NEXT:    v_accvgpr_write_b32 a6, v22
+; SDAG-NEXT:    v_accvgpr_write_b32 a7, v23
+; SDAG-NEXT:    v_accvgpr_write_b32 a8, v24
+; SDAG-NEXT:    v_accvgpr_write_b32 a9, v25
 ; SDAG-NEXT:    v_accvgpr_write_b32 a10, v8
 ; SDAG-NEXT:    v_accvgpr_write_b32 a11, v9
 ; SDAG-NEXT:    v_accvgpr_write_b32 a12, v10
@@ -4028,7 +4028,7 @@ define <16 x float> @test_mfma_scale_f32_32x32x64_f8f6f4_0_0_sgpr_vgpr_sgpr__vgp
 ; SDAG-NEXT:    v_accvgpr_write_b32 a14, v12
 ; SDAG-NEXT:    v_accvgpr_write_b32 a15, v13
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_mfma_scale_f32_32x32x64_f8f6f4 a[0:15], v[16:23], v[0:7], a[0:15], v14, v15 op_sel_hi:[0,0,0]
+; SDAG-NEXT:    v_mfma_scale_f32_32x32x64_f8f6f4 a[0:15], v[26:33], v[0:7], a[0:15], v14, v15 op_sel_hi:[0,0,0]
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    s_nop 3
@@ -4540,22 +4540,22 @@ define amdgpu_kernel void @test_mfma_scale_f32_32x32x64_f8f6f4__vgprcd(<8 x i32>
 ; SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x80
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
 ; SDAG-NEXT:    v_accvgpr_write_b32 a0, s36
-; SDAG-NEXT:    v_mov_b32_e32 v0, s8
-; SDAG-NEXT:    v_mov_b32_e32 v1, s9
-; SDAG-NEXT:    v_mov_b32_e32 v2, s10
-; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    v_mov_b32_e32 v4, s12
-; SDAG-NEXT:    v_mov_b32_e32 v5, s13
-; SDAG-NEXT:    v_mov_b32_e32 v6, s14
-; SDAG-NEXT:    v_mov_b32_e32 v7, s15
-; SDAG-NEXT:    v_mov_b32_e32 v8, s16
-; SDAG-NEXT:    v_mov_b32_e32 v9, s17
-; SDAG-NEXT:    v_mov_b32_e32 v10, s18
-; SDAG-NEXT:    v_mov_b32_e32 v11, s19
-; SDAG-NEXT:    v_mov_b32_e32 v12, s20
-; SDAG-NEXT:    v_mov_b32_e32 v13, s21
-; SDAG-NEXT:    v_mov_b32_e32 v14, s22
-; SDAG-NEXT:    v_mov_b32_e32 v15, s23
+; SDAG-NEXT:    v_mov_b32_e32 v2, s8
+; SDAG-NEXT:    v_mov_b32_e32 v3, s9
+; SDAG-NEXT:    v_mov_b32_e32 v4, s10
+; SDAG-NEXT:    v_mov_b32_e32 v5, s11
+; SDAG-NEXT:    v_mov_b32_e32 v6, s12
+; SDAG-NEXT:    v_mov_b32_e32 v7, s13
+; SDAG-NEXT:    v_mov_b32_e32 v8, s14
+; SDAG-NEXT:    v_mov_b32_e32 v9, s15
+; SDAG-NEXT:    v_mov_b32_e32 v10, s16
+; SDAG-NEXT:    v_mov_b32_e32 v11, s17
+; SDAG-NEXT:    v_mov_b32_e32 v12, s18
+; SDAG-NEXT:    v_mov_b32_e32 v13, s19
+; SDAG-NEXT:    v_mov_b32_e32 v14, s20
+; SDAG-NEXT:    v_mov_b32_e32 v15, s21
+; SDAG-NEXT:    v_mov_b32_e32 v16, s22
+; SDAG-NEXT:    v_mov_b32_e32 v17, s23
 ; SDAG-NEXT:    v_accvgpr_write_b32 a1, s37
 ; SDAG-NEXT:    v_accvgpr_write_b32 a2, s38
 ; SDAG-NEXT:    v_accvgpr_write_b32 a3, s39
@@ -4571,9 +4571,9 @@ define amdgpu_kernel void @test_mfma_scale_f32_32x32x64_f8f6f4__vgprcd(<8 x i32>
 ; SDAG-NEXT:    v_accvgpr_write_b32 a13, s49
 ; SDAG-NEXT:    v_accvgpr_write_b32 a14, s50
 ; SDAG-NEXT:    v_accvgpr_write_b32 a15, s51
-; SDAG-NEXT:    v_mov_b32_e32 v16, s1
+; SDAG-NEXT:    v_mov_b32_e32 v0, s1
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_mfma_scale_f32_32x32x64_f8f6f4 a[0:15], v[0:7], v[8:15], a[0:15], s0, v16 op_sel:[1,1,0] op_sel_hi:[1,0,0] blgp:2
+; SDAG-NEXT:    v_mfma_scale_f32_32x32x64_f8f6f4 a[0:15], v[2:9], v[10:17], a[0:15], s0, v0 op_sel:[1,1,0] op_sel_hi:[1,0,0] blgp:2
 ; SDAG-NEXT:    v_mov_b32_e32 v0, 0
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    s_nop 7
@@ -4735,26 +4735,26 @@ define amdgpu_kernel void @test_mfma_scale_f32_32x32x64_f8f6f4_0_0__nonmac(<8 x
 ; SDAG:       ; %bb.0:
 ; SDAG-NEXT:    s_load_dwordx16 s[12:27], s[4:5], 0x0
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v0, s12
-; SDAG-NEXT:    v_mov_b32_e32 v1, s13
-; SDAG-NEXT:    v_mov_b32_e32 v2, s14
-; SDAG-NEXT:    v_mov_b32_e32 v3, s15
-; SDAG-NEXT:    v_mov_b32_e32 v4, s16
-; SDAG-NEXT:    v_mov_b32_e32 v5, s17
-; SDAG-NEXT:    v_mov_b32_e32 v6, s18
-; SDAG-NEXT:    v_mov_b32_e32 v7, s19
-; SDAG-NEXT:    v_mov_b32_e32 v8, s20
-; SDAG-NEXT:    v_mov_b32_e32 v9, s21
-; SDAG-NEXT:    v_mov_b32_e32 v10, s22
-; SDAG-NEXT:    v_mov_b32_e32 v11, s23
+; SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; SDAG-NEXT:    v_mov_b32_e32 v4, s14
+; SDAG-NEXT:    v_mov_b32_e32 v5, s15
+; SDAG-NEXT:    v_mov_b32_e32 v6, s16
+; SDAG-NEXT:    v_mov_b32_e32 v7, s17
+; SDAG-NEXT:    v_mov_b32_e32 v8, s18
+; SDAG-NEXT:    v_mov_b32_e32 v9, s19
+; SDAG-NEXT:    v_mov_b32_e32 v10, s20
+; SDAG-NEXT:    v_mov_b32_e32 v11, s21
+; SDAG-NEXT:    v_mov_b32_e32 v12, s22
+; SDAG-NEXT:    v_mov_b32_e32 v13, s23
 ; SDAG-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x40
 ; SDAG-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x80
-; SDAG-NEXT:    v_mov_b32_e32 v12, s24
-; SDAG-NEXT:    v_mov_b32_e32 v13, s25
-; SDAG-NEXT:    v_mov_b32_e32 v14, s26
+; SDAG-NEXT:    v_mov_b32_e32 v14, s24
+; SDAG-NEXT:    v_mov_b32_e32 v15, s25
+; SDAG-NEXT:    v_mov_b32_e32 v16, s26
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
 ; SDAG-NEXT:    v_accvgpr_write_b32 a0, s8
-; SDAG-NEXT:    v_mov_b32_e32 v15, s27
+; SDAG-NEXT:    v_mov_b32_e32 v17, s27
 ; SDAG-NEXT:    v_accvgpr_write_b32 a1, s9
 ; SDAG-NEXT:    v_accvgpr_write_b32 a2, s10
 ; SDAG-NEXT:    v_accvgpr_write_b32 a3, s11
@@ -4770,45 +4770,44 @@ define amdgpu_kernel void @test_mfma_scale_f32_32x32x64_f8f6f4_0_0__nonmac(<8 x
 ; SDAG-NEXT:    v_accvgpr_write_b32 a13, s21
 ; SDAG-NEXT:    v_accvgpr_write_b32 a14, s22
 ; SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
-; SDAG-NEXT:    v_mov_b32_e32 v16, s1
+; SDAG-NEXT:    v_mov_b32_e32 v0, s1
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_mfma_scale_f32_32x32x64_f8f6f4 a[0:15], v[0:7], v[8:15], a[0:15], s0, v16 op_sel_hi:[0,0,0]
-; SDAG-NEXT:    v_mov_b32_e32 v0, s20
-; SDAG-NEXT:    v_mov_b32_e32 v1, s21
-; SDAG-NEXT:    v_mov_b32_e32 v2, s22
-; SDAG-NEXT:    v_mov_b32_e32 v3, s23
-; SDAG-NEXT:    v_mov_b64_e32 v[4:5], 48
-; SDAG-NEXT:    global_store_dwordx4 v[4:5], v[0:3], off sc0 sc1
+; SDAG-NEXT:    v_mfma_scale_f32_32x32x64_f8f6f4 a[0:15], v[2:9], v[10:17], a[0:15], s0, v0 op_sel_hi:[0,0,0]
+; SDAG-NEXT:    v_mov_b32_e32 v2, s20
+; SDAG-NEXT:    v_mov_b32_e32 v3, s21
+; SDAG-NEXT:    v_mov_b32_e32 v4, s22
+; SDAG-NEXT:    v_mov_b32_e32 v5, s23
+; SDAG-NEXT:    v_mov_b64_e32 v[0:1], 48
+; SDAG-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    v_mov_b64_e32 v[6:7], 32
-; SDAG-NEXT:    v_mov_b64_e32 v[8:9], 16
-; SDAG-NEXT:    v_mov_b32_e32 v0, s16
-; SDAG-NEXT:    v_mov_b32_e32 v1, s17
-; SDAG-NEXT:    v_mov_b32_e32 v2, s18
-; SDAG-NEXT:    v_mov_b32_e32 v3, s19
-; SDAG-NEXT:    global_store_dwordx4 v[6:7], v[0:3], off sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v6, s18
+; SDAG-NEXT:    v_mov_b32_e32 v7, s19
+; SDAG-NEXT:    v_mov_b32_e32 v4, s16
+; SDAG-NEXT:    v_mov_b32_e32 v5, s17
+; SDAG-NEXT:    v_mov_b64_e32 v[2:3], 32
+; SDAG-NEXT:    global_store_dwordx4 v[2:3], v[4:7], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    v_mov_b64_e32 v[10:11], 0
-; SDAG-NEXT:    v_mov_b32_e32 v0, s12
-; SDAG-NEXT:    v_mov_b32_e32 v1, s13
-; SDAG-NEXT:    v_mov_b32_e32 v2, s14
-; SDAG-NEXT:    v_mov_b32_e32 v3, s15
-; SDAG-NEXT:    global_store_dwordx4 v[8:9], v[0:3], off sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v8, s14
+; SDAG-NEXT:    v_mov_b32_e32 v9, s15
+; SDAG-NEXT:    v_mov_b32_e32 v6, s12
+; SDAG-NEXT:    v_mov_b32_e32 v7, s13
+; SDAG-NEXT:    v_mov_b64_e32 v[4:5], 16
+; SDAG-NEXT:    global_store_dwordx4 v[4:5], v[6:9], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_mov_b32_e32 v0, s8
-; SDAG-NEXT:    v_mov_b32_e32 v1, s9
-; SDAG-NEXT:    v_mov_b32_e32 v2, s10
-; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    global_store_dwordx4 v[10:11], v[0:3], off sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v10, s10
+; SDAG-NEXT:    v_mov_b32_e32 v11, s11
+; SDAG-NEXT:    v_mov_b32_e32 v8, s8
+; SDAG-NEXT:    v_mov_b32_e32 v9, s9
+; SDAG-NEXT:    v_mov_b64_e32 v[6:7], 0
+; SDAG-NEXT:    global_store_dwordx4 v[6:7], v[8:11], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[6:7], a[8:11], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[2:3], a[8:11], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[4:5], a[12:15], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[0:1], a[12:15], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[10:11], a[0:3], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[6:7], a[0:3], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[8:9], a[4:7], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[4:5], a[4:7], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_endpgm
 ;
@@ -4922,42 +4921,41 @@ define amdgpu_kernel void @test_mfma_scale_f32_32x32x64_f8f6f4_25_42__nonmac(<8
 ; SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
 ; SDAG-NEXT:    s_nop 1
 ; SDAG-NEXT:    v_mfma_scale_f32_32x32x64_f8f6f4 a[0:15], v[0:7], v[8:15], a[0:15], 25, 42 op_sel_hi:[0,0,0] blgp:2
-; SDAG-NEXT:    v_mov_b32_e32 v0, s20
-; SDAG-NEXT:    v_mov_b32_e32 v1, s21
-; SDAG-NEXT:    v_mov_b32_e32 v2, s22
-; SDAG-NEXT:    v_mov_b32_e32 v3, s23
-; SDAG-NEXT:    v_mov_b64_e32 v[4:5], 48
-; SDAG-NEXT:    global_store_dwordx4 v[4:5], v[0:3], off sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v2, s20
+; SDAG-NEXT:    v_mov_b32_e32 v3, s21
+; SDAG-NEXT:    v_mov_b32_e32 v4, s22
+; SDAG-NEXT:    v_mov_b32_e32 v5, s23
+; SDAG-NEXT:    v_mov_b64_e32 v[0:1], 48
+; SDAG-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    v_mov_b64_e32 v[6:7], 32
-; SDAG-NEXT:    v_mov_b64_e32 v[8:9], 16
-; SDAG-NEXT:    v_mov_b32_e32 v0, s16
-; SDAG-NEXT:    v_mov_b32_e32 v1, s17
-; SDAG-NEXT:    v_mov_b32_e32 v2, s18
-; SDAG-NEXT:    v_mov_b32_e32 v3, s19
-; SDAG-NEXT:    global_store_dwordx4 v[6:7], v[0:3], off sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v6, s18
+; SDAG-NEXT:    v_mov_b32_e32 v7, s19
+; SDAG-NEXT:    v_mov_b32_e32 v4, s16
+; SDAG-NEXT:    v_mov_b32_e32 v5, s17
+; SDAG-NEXT:    v_mov_b64_e32 v[2:3], 32
+; SDAG-NEXT:    global_store_dwordx4 v[2:3], v[4:7], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    v_mov_b64_e32 v[10:11], 0
-; SDAG-NEXT:    v_mov_b32_e32 v0, s12
-; SDAG-NEXT:    v_mov_b32_e32 v1, s13
-; SDAG-NEXT:    v_mov_b32_e32 v2, s14
-; SDAG-NEXT:    v_mov_b32_e32 v3, s15
-; SDAG-NEXT:    global_store_dwordx4 v[8:9], v[0:3], off sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v8, s14
+; SDAG-NEXT:    v_mov_b32_e32 v9, s15
+; SDAG-NEXT:    v_mov_b32_e32 v6, s12
+; SDAG-NEXT:    v_mov_b32_e32 v7, s13
+; SDAG-NEXT:    v_mov_b64_e32 v[4:5], 16
+; SDAG-NEXT:    global_store_dwordx4 v[4:5], v[6:9], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_mov_b32_e32 v0, s8
-; SDAG-NEXT:    v_mov_b32_e32 v1, s9
-; SDAG-NEXT:    v_mov_b32_e32 v2, s10
-; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    global_store_dwordx4 v[10:11], v[0:3], off sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v10, s10
+; SDAG-NEXT:    v_mov_b32_e32 v11, s11
+; SDAG-NEXT:    v_mov_b32_e32 v8, s8
+; SDAG-NEXT:    v_mov_b32_e32 v9, s9
+; SDAG-NEXT:    v_mov_b64_e32 v[6:7], 0
+; SDAG-NEXT:    global_store_dwordx4 v[6:7], v[8:11], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[6:7], a[8:11], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[2:3], a[8:11], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[4:5], a[12:15], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[0:1], a[12:15], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[10:11], a[0:3], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[6:7], a[0:3], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[8:9], a[4:7], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[4:5], a[4:7], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_endpgm
 ;
@@ -5069,42 +5067,41 @@ define amdgpu_kernel void @test_mfma_scale_f32_32x32x64_f8f6f4_0_0__vgprcd_nonma
 ; SDAG-NEXT:    v_accvgpr_write_b32 a16, s8
 ; SDAG-NEXT:    s_nop 1
 ; SDAG-NEXT:    v_mfma_f32_32x32x64_f8f6f4 a[0:15], v[0:7], v[8:15], a[16:31] blgp:2
-; SDAG-NEXT:    v_mov_b32_e32 v0, s20
-; SDAG-NEXT:    v_mov_b32_e32 v1, s21
-; SDAG-NEXT:    v_mov_b32_e32 v2, s22
-; SDAG-NEXT:    v_mov_b32_e32 v3, s23
-; SDAG-NEXT:    v_mov_b64_e32 v[4:5], 48
-; SDAG-NEXT:    global_store_dwordx4 v[4:5], v[0:3], off sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v2, s20
+; SDAG-NEXT:    v_mov_b32_e32 v3, s21
+; SDAG-NEXT:    v_mov_b32_e32 v4, s22
+; SDAG-NEXT:    v_mov_b32_e32 v5, s23
+; SDAG-NEXT:    v_mov_b64_e32 v[0:1], 48
+; SDAG-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    v_mov_b64_e32 v[6:7], 32
-; SDAG-NEXT:    v_mov_b64_e32 v[8:9], 16
-; SDAG-NEXT:    v_mov_b32_e32 v0, s16
-; SDAG-NEXT:    v_mov_b32_e32 v1, s17
-; SDAG-NEXT:    v_mov_b32_e32 v2, s18
-; SDAG-NEXT:    v_mov_b32_e32 v3, s19
-; SDAG-NEXT:    global_store_dwordx4 v[6:7], v[0:3], off sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v6, s18
+; SDAG-NEXT:    v_mov_b32_e32 v7, s19
+; SDAG-NEXT:    v_mov_b32_e32 v4, s16
+; SDAG-NEXT:    v_mov_b32_e32 v5, s17
+; SDAG-NEXT:    v_mov_b64_e32 v[2:3], 32
+; SDAG-NEXT:    global_store_dwordx4 v[2:3], v[4:7], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    v_mov_b64_e32 v[10:11], 0
-; SDAG-NEXT:    v_mov_b32_e32 v0, s12
-; SDAG-NEXT:    v_mov_b32_e32 v1, s13
-; SDAG-NEXT:    v_mov_b32_e32 v2, s14
-; SDAG-NEXT:    v_mov_b32_e32 v3, s15
-; SDAG-NEXT:    global_store_dwordx4 v[8:9], v[0:3], off sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v8, s14
+; SDAG-NEXT:    v_mov_b32_e32 v9, s15
+; SDAG-NEXT:    v_mov_b32_e32 v6, s12
+; SDAG-NEXT:    v_mov_b32_e32 v7, s13
+; SDAG-NEXT:    v_mov_b64_e32 v[4:5], 16
+; SDAG-NEXT:    global_store_dwordx4 v[4:5], v[6:9], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_mov_b32_e32 v0, s8
-; SDAG-NEXT:    v_mov_b32_e32 v1, s9
-; SDAG-NEXT:    v_mov_b32_e32 v2, s10
-; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    global_store_dwordx4 v[10:11], v[0:3], off sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v10, s10
+; SDAG-NEXT:    v_mov_b32_e32 v11, s11
+; SDAG-NEXT:    v_mov_b32_e32 v8, s8
+; SDAG-NEXT:    v_mov_b32_e32 v9, s9
+; SDAG-NEXT:    v_mov_b64_e32 v[6:7], 0
+; SDAG-NEXT:    global_store_dwordx4 v[6:7], v[8:11], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[6:7], a[8:11], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[2:3], a[8:11], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[4:5], a[12:15], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[0:1], a[12:15], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[10:11], a[0:3], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[6:7], a[0:3], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[8:9], a[4:7], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[4:5], a[4:7], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_endpgm
 ;
@@ -5216,42 +5213,41 @@ define amdgpu_kernel void @test_mfma_scale_f32_32x32x64_f8f6f4_25_42__vgprcd_non
 ; SDAG-NEXT:    v_accvgpr_write_b32 a15, s23
 ; SDAG-NEXT:    s_nop 1
 ; SDAG-NEXT:    v_mfma_scale_f32_32x32x64_f8f6f4 a[0:15], v[0:7], v[8:15], a[0:15], 25, 42 op_sel_hi:[0,0,0] blgp:2
-; SDAG-NEXT:    v_mov_b32_e32 v0, s20
-; SDAG-NEXT:    v_mov_b32_e32 v1, s21
-; SDAG-NEXT:    v_mov_b32_e32 v2, s22
-; SDAG-NEXT:    v_mov_b32_e32 v3, s23
-; SDAG-NEXT:    v_mov_b64_e32 v[4:5], 48
-; SDAG-NEXT:    global_store_dwordx4 v[4:5], v[0:3], off sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v2, s20
+; SDAG-NEXT:    v_mov_b32_e32 v3, s21
+; SDAG-NEXT:    v_mov_b32_e32 v4, s22
+; SDAG-NEXT:    v_mov_b32_e32 v5, s23
+; SDAG-NEXT:    v_mov_b64_e32 v[0:1], 48
+; SDAG-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    v_mov_b64_e32 v[6:7], 32
-; SDAG-NEXT:    v_mov_b64_e32 v[8:9], 16
-; SDAG-NEXT:    v_mov_b32_e32 v0, s16
-; SDAG-NEXT:    v_mov_b32_e32 v1, s17
-; SDAG-NEXT:    v_mov_b32_e32 v2, s18
-; SDAG-NEXT:    v_mov_b32_e32 v3, s19
-; SDAG-NEXT:    global_store_dwordx4 v[6:7], v[0:3], off sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v6, s18
+; SDAG-NEXT:    v_mov_b32_e32 v7, s19
+; SDAG-NEXT:    v_mov_b32_e32 v4, s16
+; SDAG-NEXT:    v_mov_b32_e32 v5, s17
+; SDAG-NEXT:    v_mov_b64_e32 v[2:3], 32
+; SDAG-NEXT:    global_store_dwordx4 v[2:3], v[4:7], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    v_mov_b64_e32 v[10:11], 0
-; SDAG-NEXT:    v_mov_b32_e32 v0, s12
-; SDAG-NEXT:    v_mov_b32_e32 v1, s13
-; SDAG-NEXT:    v_mov_b32_e32 v2, s14
-; SDAG-NEXT:    v_mov_b32_e32 v3, s15
-; SDAG-NEXT:    global_store_dwordx4 v[8:9], v[0:3], off sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v8, s14
+; SDAG-NEXT:    v_mov_b32_e32 v9, s15
+; SDAG-NEXT:    v_mov_b32_e32 v6, s12
+; SDAG-NEXT:    v_mov_b32_e32 v7, s13
+; SDAG-NEXT:    v_mov_b64_e32 v[4:5], 16
+; SDAG-NEXT:    global_store_dwordx4 v[4:5], v[6:9], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_mov_b32_e32 v0, s8
-; SDAG-NEXT:    v_mov_b32_e32 v1, s9
-; SDAG-NEXT:    v_mov_b32_e32 v2, s10
-; SDAG-NEXT:    v_mov_b32_e32 v3, s11
-; SDAG-NEXT:    global_store_dwordx4 v[10:11], v[0:3], off sc0 sc1
+; SDAG-NEXT:    v_mov_b32_e32 v10, s10
+; SDAG-NEXT:    v_mov_b32_e32 v11, s11
+; SDAG-NEXT:    v_mov_b32_e32 v8, s8
+; SDAG-NEXT:    v_mov_b32_e32 v9, s9
+; SDAG-NEXT:    v_mov_b64_e32 v[6:7], 0
+; SDAG-NEXT:    global_store_dwordx4 v[6:7], v[8:11], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[6:7], a[8:11], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[2:3], a[8:11], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[4:5], a[12:15], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[0:1], a[12:15], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[10:11], a[0:3], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[6:7], a[0:3], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
-; SDAG-NEXT:    global_store_dwordx4 v[8:9], a[4:7], off sc0 sc1
+; SDAG-NEXT:    global_store_dwordx4 v[4:5], a[4:7], off sc0 sc1
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_endpgm
 ;
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.xf32.gfx942.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.xf32.gfx942.ll
index 7193fee..ea9334a 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.xf32.gfx942.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.mfma.xf32.gfx942.ll
@@ -1,22 +1,100 @@
-; RUN: llc -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX942 %s
-; RUN: llc -mtriple=amdgcn -mcpu=gfx942 -global-isel < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GISEL %s
-; RUN: llc -mtriple=amdgcn -mcpu=gfx942 -stress-regalloc=10 < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GFX942 %s
-; RUN: llc -mtriple=amdgcn -mcpu=gfx942 -stress-regalloc=10 -global-isel < %s | FileCheck -enable-var-scope --check-prefixes=GCN,GISEL %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck --check-prefixes=GFX942,GFX942-SDAG %s
+; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck --check-prefixes=GFX942,GFX942-GISEL %s
+; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx942 -stress-regalloc=10 < %s | FileCheck --check-prefixes=GFX942-STRESS,GFX942-SDAG-STRESS %s
+; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx942 -stress-regalloc=10 < %s | FileCheck --check-prefixes=GFX942-STRESS,GFX942-GISEL-STRESS %s
 
 declare <4 x float> @llvm.amdgcn.mfma.f32.16x16x8.xf32(<2 x float>, <2 x float>, <4 x float>, i32, i32, i32)
 declare <16 x float> @llvm.amdgcn.mfma.f32.32x32x4.xf32(<2 x float>, <2 x float>, <16 x float>, i32, i32, i32)
 
-; GCN-LABEL: {{^}}test_mfma_f32_16x16x8xf32:
-; GFX942-DAG:  v_mov_b32_e32 v[[ONE:[0-9]+]], 1.0
-; GFX942-DAG:  v_mov_b32_e32 v[[TWO:[0-9]+]], 2.0
-; GFX942-DAG:  v_mov_b32_e32 v[[THREE:[0-9]+]], 0x40400000
-; GFX942-DAG:  v_mov_b32_e32 v[[FOUR:[0-9]+]], 4.0
-; GCN-COUNT-4: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GFX942:      v_mfma_f32_16x16x8_xf32 a[{{[0-9]+:[0-9]+}}], v[[[ONE]]:[[TWO]]], v[[[THREE]]:[[FOUR]]], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GISEL:       v_mfma_f32_16x16x8_xf32 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GCN-NOT:     v_accvgpr_read_b32
-; GCN:         global_store_dwordx4 v{{[0-9]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_16x16x8xf32(ptr addrspace(1) %arg) #0 {
+; GFX942-SDAG-LABEL: test_mfma_f32_16x16x8xf32:
+; GFX942-SDAG:       ; %bb.0: ; %bb
+; GFX942-SDAG-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v4, 1.0
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v5, 2.0
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v0, 0x40400000
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v1, 4.0
+; GFX942-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-SDAG-NEXT:    s_nop 1
+; GFX942-SDAG-NEXT:    v_mfma_f32_16x16x8_xf32 a[0:3], v[4:5], v[0:1], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-SDAG-NEXT:    s_nop 6
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v2, a[0:3], s[6:7]
+; GFX942-SDAG-NEXT:    s_endpgm
+;
+; GFX942-GISEL-LABEL: test_mfma_f32_16x16x8xf32:
+; GFX942-GISEL:       ; %bb.0: ; %bb
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    s_mov_b32 s4, 1.0
+; GFX942-GISEL-NEXT:    s_mov_b32 s5, 2.0
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[4:5]
+; GFX942-GISEL-NEXT:    s_mov_b32 s4, 0x40400000
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-GISEL-NEXT:    s_mov_b32 s5, 4.0
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[4:5]
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-GISEL-NEXT:    s_nop 1
+; GFX942-GISEL-NEXT:    v_mfma_f32_16x16x8_xf32 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-GISEL-NEXT:    s_nop 5
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX942-GISEL-NEXT:    s_endpgm
+;
+; GFX942-SDAG-STRESS-LABEL: test_mfma_f32_16x16x8xf32:
+; GFX942-SDAG-STRESS:       ; %bb.0: ; %bb
+; GFX942-SDAG-STRESS-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-SDAG-STRESS-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX942-SDAG-STRESS-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX942-SDAG-STRESS-NEXT:    v_mov_b32_e32 v2, 0x40400000
+; GFX942-SDAG-STRESS-NEXT:    v_mov_b32_e32 v3, 4.0
+; GFX942-SDAG-STRESS-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-SDAG-STRESS-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-SDAG-STRESS-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-SDAG-STRESS-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-SDAG-STRESS-NEXT:    s_nop 1
+; GFX942-SDAG-STRESS-NEXT:    v_mfma_f32_16x16x8_xf32 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-SDAG-STRESS-NEXT:    s_nop 6
+; GFX942-SDAG-STRESS-NEXT:    global_store_dwordx4 v4, a[0:3], s[6:7]
+; GFX942-SDAG-STRESS-NEXT:    s_endpgm
+;
+; GFX942-GISEL-STRESS-LABEL: test_mfma_f32_16x16x8xf32:
+; GFX942-GISEL-STRESS:       ; %bb.0: ; %bb
+; GFX942-GISEL-STRESS-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX942-GISEL-STRESS-NEXT:    s_mov_b32 s0, 1.0
+; GFX942-GISEL-STRESS-NEXT:    s_mov_b32 s2, 0x40400000
+; GFX942-GISEL-STRESS-NEXT:    s_mov_b32 s1, 2.0
+; GFX942-GISEL-STRESS-NEXT:    s_mov_b32 s3, 4.0
+; GFX942-GISEL-STRESS-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-GISEL-STRESS-NEXT:    v_mov_b64_e32 v[2:3], s[2:3]
+; GFX942-GISEL-STRESS-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-STRESS-NEXT:    s_load_dwordx4 s[0:3], s[6:7], 0x0
+; GFX942-GISEL-STRESS-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-GISEL-STRESS-NEXT:    s_nop 1
+; GFX942-GISEL-STRESS-NEXT:    v_mfma_f32_16x16x8_xf32 a[0:3], v[0:1], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-GISEL-STRESS-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-GISEL-STRESS-NEXT:    s_nop 5
+; GFX942-GISEL-STRESS-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7]
+; GFX942-GISEL-STRESS-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x float> @llvm.amdgcn.mfma.f32.16x16x8.xf32(<2 x float> <float 1.0, float 2.0>, <2 x float> <float 3.0, float 4.0>, <4 x float> %in.1, i32 1, i32 2, i32 3)
@@ -24,17 +102,158 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_32x32x4xf32:
-; GFX942-DAG:  v_mov_b32_e32 v[[ONE:[0-9]+]], 1.0
-; GFX942-DAG:  v_mov_b32_e32 v[[TWO:[0-9]+]], 2.0
-; GFX942-DAG:  v_mov_b32_e32 v[[THREE:[0-9]+]], 0x40400000
-; GFX942-DAG:  v_mov_b32_e32 v[[FOUR:[0-9]+]], 4.0
-; GCN-COUNT-4: v_accvgpr_write_b32 a{{[0-9]+}}, s{{[0-9]+}}
-; GFX942:      v_mfma_f32_32x32x4_xf32 a[{{[0-9]+:[0-9]+}}], v[[[ONE]]:[[TWO]]], v[[[THREE]]:[[FOUR]]], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GISEL:       v_mfma_f32_32x32x4_xf32 a[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], v[{{[0-9]+:[0-9]+}}], a[{{[0-9]+:[0-9]+}}] cbsz:1 abid:2 blgp:3
-; GCN-NOT:     v_accvgpr_read_b32
-; GCN:         global_store_dwordx4 v{{[0-9]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_32x32x4xf32(ptr addrspace(1) %arg) #0 {
+; GFX942-SDAG-LABEL: test_mfma_f32_32x32x4xf32:
+; GFX942-SDAG:       ; %bb.0: ; %bb
+; GFX942-SDAG-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v2, 1.0
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v3, 2.0
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v0, 0x40400000
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v1, 4.0
+; GFX942-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-SDAG-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX942-SDAG-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX942-SDAG-NEXT:    s_nop 1
+; GFX942-SDAG-NEXT:    v_mfma_f32_32x32x4_xf32 a[0:15], v[2:3], v[0:1], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-SDAG-NEXT:    s_nop 7
+; GFX942-SDAG-NEXT:    s_nop 1
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX942-SDAG-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX942-SDAG-NEXT:    s_endpgm
+;
+; GFX942-GISEL-LABEL: test_mfma_f32_32x32x4xf32:
+; GFX942-GISEL:       ; %bb.0: ; %bb
+; GFX942-GISEL-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-GISEL-NEXT:    s_mov_b32 s18, 1.0
+; GFX942-GISEL-NEXT:    s_mov_b32 s19, 2.0
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[18:19]
+; GFX942-GISEL-NEXT:    s_mov_b32 s18, 0x40400000
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-GISEL-NEXT:    s_mov_b32 s19, 4.0
+; GFX942-GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[18:19]
+; GFX942-GISEL-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX942-GISEL-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX942-GISEL-NEXT:    s_nop 1
+; GFX942-GISEL-NEXT:    v_mfma_f32_32x32x4_xf32 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-GISEL-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-GISEL-NEXT:    s_nop 7
+; GFX942-GISEL-NEXT:    s_nop 1
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX942-GISEL-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX942-GISEL-NEXT:    s_endpgm
+;
+; GFX942-SDAG-STRESS-LABEL: test_mfma_f32_32x32x4xf32:
+; GFX942-SDAG-STRESS:       ; %bb.0: ; %bb
+; GFX942-SDAG-STRESS-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-SDAG-STRESS-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX942-SDAG-STRESS-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX942-SDAG-STRESS-NEXT:    v_mov_b32_e32 v2, 0x40400000
+; GFX942-SDAG-STRESS-NEXT:    v_mov_b32_e32 v3, 4.0
+; GFX942-SDAG-STRESS-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-SDAG-STRESS-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-SDAG-STRESS-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX942-SDAG-STRESS-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX942-SDAG-STRESS-NEXT:    s_nop 1
+; GFX942-SDAG-STRESS-NEXT:    v_mfma_f32_32x32x4_xf32 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-SDAG-STRESS-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-SDAG-STRESS-NEXT:    s_nop 7
+; GFX942-SDAG-STRESS-NEXT:    s_nop 1
+; GFX942-SDAG-STRESS-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX942-SDAG-STRESS-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX942-SDAG-STRESS-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX942-SDAG-STRESS-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX942-SDAG-STRESS-NEXT:    s_endpgm
+;
+; GFX942-GISEL-STRESS-LABEL: test_mfma_f32_32x32x4xf32:
+; GFX942-GISEL-STRESS:       ; %bb.0: ; %bb
+; GFX942-GISEL-STRESS-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x24
+; GFX942-GISEL-STRESS-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-STRESS-NEXT:    s_load_dwordx16 s[0:15], s[16:17], 0x0
+; GFX942-GISEL-STRESS-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a0, s0
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a1, s1
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a2, s2
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a3, s3
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a4, s4
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a5, s5
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a6, s6
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a7, s7
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a8, s8
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a9, s9
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a10, s10
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a11, s11
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a12, s12
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a13, s13
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a14, s14
+; GFX942-GISEL-STRESS-NEXT:    v_accvgpr_write_b32 a15, s15
+; GFX942-GISEL-STRESS-NEXT:    s_mov_b32 s0, 1.0
+; GFX942-GISEL-STRESS-NEXT:    s_mov_b32 s1, 2.0
+; GFX942-GISEL-STRESS-NEXT:    v_mov_b64_e32 v[0:1], s[0:1]
+; GFX942-GISEL-STRESS-NEXT:    s_mov_b32 s0, 0x40400000
+; GFX942-GISEL-STRESS-NEXT:    s_mov_b32 s1, 4.0
+; GFX942-GISEL-STRESS-NEXT:    v_mov_b64_e32 v[2:3], s[0:1]
+; GFX942-GISEL-STRESS-NEXT:    s_nop 1
+; GFX942-GISEL-STRESS-NEXT:    v_mfma_f32_32x32x4_xf32 a[0:15], v[0:1], v[2:3], a[0:15] cbsz:1 abid:2 blgp:3
+; GFX942-GISEL-STRESS-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-GISEL-STRESS-NEXT:    s_nop 7
+; GFX942-GISEL-STRESS-NEXT:    s_nop 1
+; GFX942-GISEL-STRESS-NEXT:    global_store_dwordx4 v0, a[0:3], s[16:17]
+; GFX942-GISEL-STRESS-NEXT:    global_store_dwordx4 v0, a[4:7], s[16:17] offset:16
+; GFX942-GISEL-STRESS-NEXT:    global_store_dwordx4 v0, a[8:11], s[16:17] offset:32
+; GFX942-GISEL-STRESS-NEXT:    global_store_dwordx4 v0, a[12:15], s[16:17] offset:48
+; GFX942-GISEL-STRESS-NEXT:    s_endpgm
 bb:
   %in.1 = load <16 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <16 x float> @llvm.amdgcn.mfma.f32.32x32x4.xf32(<2 x float> <float 1.0, float 2.0>, <2 x float> <float 3.0, float 4.0>, <16 x float> %in.1, i32 1, i32 2, i32 3)
@@ -43,3 +262,6 @@ bb:
 }
 
 attributes #0 = { "amdgpu-flat-work-group-size"="1,256" }
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; GFX942: {{.*}}
+; GFX942-STRESS: {{.*}}
diff --git a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.smfmac.gfx950.ll b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.smfmac.gfx950.ll
index 77d4aad..8056881 100644
--- a/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.smfmac.gfx950.ll
+++ b/llvm/test/CodeGen/AMDGPU/llvm.amdgcn.smfmac.gfx950.ll
@@ -17,24 +17,24 @@ define amdgpu_kernel void @test_smfmac_f32_16x16x64_f16__vgpr(ptr addrspace(1) %
 ; SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x34
 ; SDAG-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
 ; SDAG-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
-; SDAG-NEXT:    v_mov_b32_e32 v16, 0
+; SDAG-NEXT:    v_mov_b32_e32 v12, 0
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    global_load_dwordx4 v[8:11], v0, s[6:7]
+; SDAG-NEXT:    global_load_dwordx4 v[14:17], v0, s[6:7]
 ; SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x44
 ; SDAG-NEXT:    s_load_dword s16, s[4:5], 0x64
-; SDAG-NEXT:    v_mov_b64_e32 v[14:15], s[2:3]
-; SDAG-NEXT:    v_mov_b64_e32 v[12:13], s[0:1]
+; SDAG-NEXT:    v_mov_b64_e32 v[10:11], s[2:3]
+; SDAG-NEXT:    v_mov_b64_e32 v[8:9], s[0:1]
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
 ; SDAG-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
 ; SDAG-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
 ; SDAG-NEXT:    v_mov_b64_e32 v[4:5], s[12:13]
 ; SDAG-NEXT:    v_mov_b64_e32 v[6:7], s[14:15]
-; SDAG-NEXT:    v_mov_b32_e32 v17, s16
+; SDAG-NEXT:    v_mov_b32_e32 v13, s16
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_smfmac_f32_16x16x64_f16 v[8:11], v[12:15], v[0:7], v17 cbsz:1 abid:2
+; SDAG-NEXT:    v_smfmac_f32_16x16x64_f16 v[14:17], v[8:11], v[0:7], v13 cbsz:1 abid:2
 ; SDAG-NEXT:    s_nop 7
-; SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[6:7]
+; SDAG-NEXT:    global_store_dwordx4 v12, v[14:17], s[6:7]
 ; SDAG-NEXT:    s_endpgm
 ;
 ; GISEL-LABEL: test_smfmac_f32_16x16x64_f16__vgpr:
@@ -44,23 +44,23 @@ define amdgpu_kernel void @test_smfmac_f32_16x16x64_f16__vgpr(ptr addrspace(1) %
 ; GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
 ; GISEL-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
 ; GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GISEL-NEXT:    global_load_dwordx4 v[8:11], v0, s[6:7]
+; GISEL-NEXT:    global_load_dwordx4 v[14:17], v0, s[6:7]
 ; GISEL-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x44
 ; GISEL-NEXT:    s_load_dword s16, s[4:5], 0x64
-; GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[2:3]
-; GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[0:1]
+; GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[2:3]
+; GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[0:1]
 ; GISEL-NEXT:    s_waitcnt lgkmcnt(0)
 ; GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
 ; GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
 ; GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[12:13]
 ; GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[14:15]
-; GISEL-NEXT:    v_mov_b32_e32 v16, s16
+; GISEL-NEXT:    v_mov_b32_e32 v12, s16
 ; GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GISEL-NEXT:    s_nop 0
-; GISEL-NEXT:    v_smfmac_f32_16x16x64_f16 v[8:11], v[12:15], v[0:7], v16 cbsz:1 abid:2
+; GISEL-NEXT:    v_smfmac_f32_16x16x64_f16 v[14:17], v[8:11], v[0:7], v12 cbsz:1 abid:2
 ; GISEL-NEXT:    v_mov_b32_e32 v0, 0
 ; GISEL-NEXT:    s_nop 6
-; GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[6:7]
+; GISEL-NEXT:    global_store_dwordx4 v0, v[14:17], s[6:7]
 ; GISEL-NEXT:    s_endpgm
 bb:
   %id = call i32 @llvm.amdgcn.workitem.id.x()
@@ -120,25 +120,25 @@ define <4 x float> @test_smfmac_f32_16x16x64_f16__sgpr(<8 x half> inreg %arg0, <
 ; SDAG-LABEL: test_smfmac_f32_16x16x64_f16__sgpr:
 ; SDAG:       ; %bb.0:
 ; SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v8, s0
-; SDAG-NEXT:    v_mov_b32_e32 v9, s1
-; SDAG-NEXT:    v_mov_b32_e32 v10, s2
-; SDAG-NEXT:    v_mov_b32_e32 v11, s3
-; SDAG-NEXT:    v_mov_b32_e32 v0, s16
-; SDAG-NEXT:    v_mov_b32_e32 v1, s17
-; SDAG-NEXT:    v_mov_b32_e32 v2, s18
-; SDAG-NEXT:    v_mov_b32_e32 v3, s19
-; SDAG-NEXT:    v_mov_b32_e32 v4, s20
-; SDAG-NEXT:    v_mov_b32_e32 v5, s21
-; SDAG-NEXT:    v_mov_b32_e32 v6, s22
-; SDAG-NEXT:    v_mov_b32_e32 v7, s23
+; SDAG-NEXT:    v_mov_b32_e32 v10, s0
+; SDAG-NEXT:    v_mov_b32_e32 v11, s1
+; SDAG-NEXT:    v_mov_b32_e32 v12, s2
+; SDAG-NEXT:    v_mov_b32_e32 v13, s3
+; SDAG-NEXT:    v_mov_b32_e32 v2, s16
+; SDAG-NEXT:    v_mov_b32_e32 v3, s17
+; SDAG-NEXT:    v_mov_b32_e32 v4, s18
+; SDAG-NEXT:    v_mov_b32_e32 v5, s19
+; SDAG-NEXT:    v_mov_b32_e32 v6, s20
+; SDAG-NEXT:    v_mov_b32_e32 v7, s21
+; SDAG-NEXT:    v_mov_b32_e32 v8, s22
+; SDAG-NEXT:    v_mov_b32_e32 v9, s23
 ; SDAG-NEXT:    v_accvgpr_write_b32 a0, s24
 ; SDAG-NEXT:    v_accvgpr_write_b32 a1, s25
 ; SDAG-NEXT:    v_accvgpr_write_b32 a2, s26
 ; SDAG-NEXT:    v_accvgpr_write_b32 a3, s27
-; SDAG-NEXT:    v_mov_b32_e32 v12, s28
+; SDAG-NEXT:    v_mov_b32_e32 v0, s28
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_smfmac_f32_16x16x64_f16 a[0:3], v[8:11], v[0:7], v12
+; SDAG-NEXT:    v_smfmac_f32_16x16x64_f16 a[0:3], v[10:13], v[2:9], v0
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    v_accvgpr_read_b32 v0, a0
 ; SDAG-NEXT:    v_accvgpr_read_b32 v1, a1
@@ -547,24 +547,24 @@ define amdgpu_kernel void @test_smfmac_f32_16x16x64_bf16__vgpr(ptr addrspace(1)
 ; GCN-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x34
 ; GCN-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
 ; GCN-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
-; GCN-NEXT:    v_mov_b32_e32 v16, 0
+; GCN-NEXT:    v_mov_b32_e32 v12, 0
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
-; GCN-NEXT:    global_load_dwordx4 v[8:11], v0, s[6:7]
+; GCN-NEXT:    global_load_dwordx4 v[14:17], v0, s[6:7]
 ; GCN-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x44
 ; GCN-NEXT:    s_load_dword s16, s[4:5], 0x64
-; GCN-NEXT:    v_mov_b64_e32 v[14:15], s[2:3]
-; GCN-NEXT:    v_mov_b64_e32 v[12:13], s[0:1]
+; GCN-NEXT:    v_mov_b64_e32 v[10:11], s[2:3]
+; GCN-NEXT:    v_mov_b64_e32 v[8:9], s[0:1]
 ; GCN-NEXT:    s_waitcnt lgkmcnt(0)
 ; GCN-NEXT:    v_mov_b64_e32 v[0:1], s[8:9]
 ; GCN-NEXT:    v_mov_b64_e32 v[2:3], s[10:11]
 ; GCN-NEXT:    v_mov_b64_e32 v[4:5], s[12:13]
 ; GCN-NEXT:    v_mov_b64_e32 v[6:7], s[14:15]
-; GCN-NEXT:    v_mov_b32_e32 v17, s16
+; GCN-NEXT:    v_mov_b32_e32 v13, s16
 ; GCN-NEXT:    s_waitcnt vmcnt(0)
 ; GCN-NEXT:    s_nop 0
-; GCN-NEXT:    v_smfmac_f32_16x16x64_bf16 v[8:11], v[12:15], v[0:7], v17 cbsz:1 abid:2
+; GCN-NEXT:    v_smfmac_f32_16x16x64_bf16 v[14:17], v[8:11], v[0:7], v13 cbsz:1 abid:2
 ; GCN-NEXT:    s_nop 7
-; GCN-NEXT:    global_store_dwordx4 v16, v[8:11], s[6:7]
+; GCN-NEXT:    global_store_dwordx4 v12, v[14:17], s[6:7]
 ; GCN-NEXT:    s_endpgm
 bb:
   %id = call i32 @llvm.amdgcn.workitem.id.x()
@@ -624,25 +624,25 @@ define <4 x float> @test_smfmac_f32_16x16x64_bf16__sgpr(<8 x bfloat> inreg %arg0
 ; GCN-LABEL: test_smfmac_f32_16x16x64_bf16__sgpr:
 ; GCN:       ; %bb.0:
 ; GCN-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GCN-NEXT:    v_mov_b32_e32 v8, s0
-; GCN-NEXT:    v_mov_b32_e32 v9, s1
-; GCN-NEXT:    v_mov_b32_e32 v10, s2
-; GCN-NEXT:    v_mov_b32_e32 v11, s3
-; GCN-NEXT:    v_mov_b32_e32 v0, s16
-; GCN-NEXT:    v_mov_b32_e32 v1, s17
-; GCN-NEXT:    v_mov_b32_e32 v2, s18
-; GCN-NEXT:    v_mov_b32_e32 v3, s19
-; GCN-NEXT:    v_mov_b32_e32 v4, s20
-; GCN-NEXT:    v_mov_b32_e32 v5, s21
-; GCN-NEXT:    v_mov_b32_e32 v6, s22
-; GCN-NEXT:    v_mov_b32_e32 v7, s23
+; GCN-NEXT:    v_mov_b32_e32 v10, s0
+; GCN-NEXT:    v_mov_b32_e32 v11, s1
+; GCN-NEXT:    v_mov_b32_e32 v12, s2
+; GCN-NEXT:    v_mov_b32_e32 v13, s3
+; GCN-NEXT:    v_mov_b32_e32 v2, s16
+; GCN-NEXT:    v_mov_b32_e32 v3, s17
+; GCN-NEXT:    v_mov_b32_e32 v4, s18
+; GCN-NEXT:    v_mov_b32_e32 v5, s19
+; GCN-NEXT:    v_mov_b32_e32 v6, s20
+; GCN-NEXT:    v_mov_b32_e32 v7, s21
+; GCN-NEXT:    v_mov_b32_e32 v8, s22
+; GCN-NEXT:    v_mov_b32_e32 v9, s23
 ; GCN-NEXT:    v_accvgpr_write_b32 a0, s24
 ; GCN-NEXT:    v_accvgpr_write_b32 a1, s25
 ; GCN-NEXT:    v_accvgpr_write_b32 a2, s26
 ; GCN-NEXT:    v_accvgpr_write_b32 a3, s27
-; GCN-NEXT:    v_mov_b32_e32 v12, s28
+; GCN-NEXT:    v_mov_b32_e32 v0, s28
 ; GCN-NEXT:    s_nop 1
-; GCN-NEXT:    v_smfmac_f32_16x16x64_bf16 a[0:3], v[8:11], v[0:7], v12
+; GCN-NEXT:    v_smfmac_f32_16x16x64_bf16 a[0:3], v[10:13], v[2:9], v0
 ; GCN-NEXT:    s_nop 7
 ; GCN-NEXT:    v_accvgpr_read_b32 v0, a0
 ; GCN-NEXT:    v_accvgpr_read_b32 v1, a1
@@ -855,30 +855,30 @@ define amdgpu_kernel void @test_smfmac_i32_16x16x128_i8__vgpr(ptr addrspace(1) %
 ; SDAG-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
 ; SDAG-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
 ; SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x34
-; SDAG-NEXT:    v_mov_b32_e32 v16, 0
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    global_load_dwordx4 v[8:11], v0, s[6:7]
+; SDAG-NEXT:    global_load_dwordx4 v[10:13], v0, s[6:7]
 ; SDAG-NEXT:    s_load_dword s16, s[4:5], 0x64
 ; SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x54
-; SDAG-NEXT:    v_mov_b32_e32 v12, s8
-; SDAG-NEXT:    v_mov_b32_e32 v13, s9
-; SDAG-NEXT:    v_mov_b32_e32 v14, s10
-; SDAG-NEXT:    v_mov_b32_e32 v15, s11
-; SDAG-NEXT:    v_mov_b32_e32 v0, s12
-; SDAG-NEXT:    v_mov_b32_e32 v1, s13
-; SDAG-NEXT:    v_mov_b32_e32 v2, s14
-; SDAG-NEXT:    v_mov_b32_e32 v3, s15
+; SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; SDAG-NEXT:    v_mov_b32_e32 v14, s8
+; SDAG-NEXT:    v_mov_b32_e32 v15, s9
+; SDAG-NEXT:    v_mov_b32_e32 v16, s10
+; SDAG-NEXT:    v_mov_b32_e32 v17, s11
+; SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; SDAG-NEXT:    v_mov_b32_e32 v4, s14
+; SDAG-NEXT:    v_mov_b32_e32 v5, s15
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v4, s0
-; SDAG-NEXT:    v_mov_b32_e32 v5, s1
-; SDAG-NEXT:    v_mov_b32_e32 v6, s2
-; SDAG-NEXT:    v_mov_b32_e32 v7, s3
-; SDAG-NEXT:    v_mov_b32_e32 v17, s16
+; SDAG-NEXT:    v_mov_b32_e32 v6, s0
+; SDAG-NEXT:    v_mov_b32_e32 v7, s1
+; SDAG-NEXT:    v_mov_b32_e32 v8, s2
+; SDAG-NEXT:    v_mov_b32_e32 v9, s3
+; SDAG-NEXT:    v_mov_b32_e32 v1, s16
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_smfmac_i32_16x16x128_i8 v[8:11], v[12:15], v[0:7], v17 cbsz:1 abid:2
+; SDAG-NEXT:    v_smfmac_i32_16x16x128_i8 v[10:13], v[14:17], v[2:9], v1 cbsz:1 abid:2
 ; SDAG-NEXT:    s_nop 7
-; SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[6:7]
+; SDAG-NEXT:    global_store_dwordx4 v0, v[10:13], s[6:7]
 ; SDAG-NEXT:    s_endpgm
 ;
 ; GISEL-LABEL: test_smfmac_i32_16x16x128_i8__vgpr:
@@ -887,24 +887,24 @@ define amdgpu_kernel void @test_smfmac_i32_16x16x128_i8__vgpr(ptr addrspace(1) %
 ; GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
 ; GISEL-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
 ; GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GISEL-NEXT:    global_load_dwordx4 v[8:11], v0, s[0:1]
+; GISEL-NEXT:    global_load_dwordx4 v[14:17], v0, s[0:1]
 ; GISEL-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x34
 ; GISEL-NEXT:    s_load_dwordx4 s[16:19], s[4:5], 0x54
 ; GISEL-NEXT:    s_load_dword s2, s[4:5], 0x64
 ; GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[10:11]
-; GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[8:9]
+; GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
 ; GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[12:13]
 ; GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[14:15]
 ; GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[16:17]
 ; GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[18:19]
-; GISEL-NEXT:    v_mov_b32_e32 v16, s2
+; GISEL-NEXT:    v_mov_b32_e32 v12, s2
 ; GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GISEL-NEXT:    s_nop 0
-; GISEL-NEXT:    v_smfmac_i32_16x16x128_i8 v[8:11], v[12:15], v[0:7], v16 cbsz:1 abid:2
+; GISEL-NEXT:    v_smfmac_i32_16x16x128_i8 v[14:17], v[8:11], v[0:7], v12 cbsz:1 abid:2
 ; GISEL-NEXT:    v_mov_b32_e32 v0, 0
 ; GISEL-NEXT:    s_nop 6
-; GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[0:1]
+; GISEL-NEXT:    global_store_dwordx4 v0, v[14:17], s[0:1]
 ; GISEL-NEXT:    s_endpgm
 bb:
   %id = call i32 @llvm.amdgcn.workitem.id.x()
@@ -964,25 +964,25 @@ define <4 x i32> @test_smfmac_i32_16x16x128_i8__sgpr(<4 x i32> inreg %arg0, <8 x
 ; SDAG-LABEL: test_smfmac_i32_16x16x128_i8__sgpr:
 ; SDAG:       ; %bb.0:
 ; SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v8, s0
-; SDAG-NEXT:    v_mov_b32_e32 v9, s1
-; SDAG-NEXT:    v_mov_b32_e32 v10, s2
-; SDAG-NEXT:    v_mov_b32_e32 v11, s3
-; SDAG-NEXT:    v_mov_b32_e32 v0, s16
-; SDAG-NEXT:    v_mov_b32_e32 v1, s17
-; SDAG-NEXT:    v_mov_b32_e32 v2, s18
-; SDAG-NEXT:    v_mov_b32_e32 v3, s19
-; SDAG-NEXT:    v_mov_b32_e32 v4, s20
-; SDAG-NEXT:    v_mov_b32_e32 v5, s21
-; SDAG-NEXT:    v_mov_b32_e32 v6, s22
-; SDAG-NEXT:    v_mov_b32_e32 v7, s23
+; SDAG-NEXT:    v_mov_b32_e32 v10, s0
+; SDAG-NEXT:    v_mov_b32_e32 v11, s1
+; SDAG-NEXT:    v_mov_b32_e32 v12, s2
+; SDAG-NEXT:    v_mov_b32_e32 v13, s3
+; SDAG-NEXT:    v_mov_b32_e32 v2, s16
+; SDAG-NEXT:    v_mov_b32_e32 v3, s17
+; SDAG-NEXT:    v_mov_b32_e32 v4, s18
+; SDAG-NEXT:    v_mov_b32_e32 v5, s19
+; SDAG-NEXT:    v_mov_b32_e32 v6, s20
+; SDAG-NEXT:    v_mov_b32_e32 v7, s21
+; SDAG-NEXT:    v_mov_b32_e32 v8, s22
+; SDAG-NEXT:    v_mov_b32_e32 v9, s23
 ; SDAG-NEXT:    v_accvgpr_write_b32 a0, s24
 ; SDAG-NEXT:    v_accvgpr_write_b32 a1, s25
 ; SDAG-NEXT:    v_accvgpr_write_b32 a2, s26
 ; SDAG-NEXT:    v_accvgpr_write_b32 a3, s27
-; SDAG-NEXT:    v_mov_b32_e32 v12, s28
+; SDAG-NEXT:    v_mov_b32_e32 v0, s28
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_smfmac_i32_16x16x128_i8 a[0:3], v[8:11], v[0:7], v12
+; SDAG-NEXT:    v_smfmac_i32_16x16x128_i8 a[0:3], v[10:13], v[2:9], v0
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    v_accvgpr_read_b32 v0, a0
 ; SDAG-NEXT:    v_accvgpr_read_b32 v1, a1
@@ -1032,22 +1032,22 @@ define amdgpu_kernel void @test_smfmac_i32_32x32x64_i8__vgpr(ptr addrspace(1) %a
 ; SDAG-NEXT:    s_load_dword s16, s[4:5], 0x64
 ; SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x54
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v24, s8
-; SDAG-NEXT:    v_mov_b32_e32 v25, s9
-; SDAG-NEXT:    v_mov_b32_e32 v26, s10
-; SDAG-NEXT:    v_mov_b32_e32 v27, s11
-; SDAG-NEXT:    v_mov_b32_e32 v16, s12
-; SDAG-NEXT:    v_mov_b32_e32 v17, s13
-; SDAG-NEXT:    v_mov_b32_e32 v18, s14
-; SDAG-NEXT:    v_mov_b32_e32 v19, s15
-; SDAG-NEXT:    v_mov_b32_e32 v20, s0
-; SDAG-NEXT:    v_mov_b32_e32 v21, s1
-; SDAG-NEXT:    v_mov_b32_e32 v22, s2
-; SDAG-NEXT:    v_mov_b32_e32 v23, s3
-; SDAG-NEXT:    v_mov_b32_e32 v28, s16
+; SDAG-NEXT:    v_mov_b32_e32 v26, s8
+; SDAG-NEXT:    v_mov_b32_e32 v27, s9
+; SDAG-NEXT:    v_mov_b32_e32 v28, s10
+; SDAG-NEXT:    v_mov_b32_e32 v29, s11
+; SDAG-NEXT:    v_mov_b32_e32 v18, s12
+; SDAG-NEXT:    v_mov_b32_e32 v19, s13
+; SDAG-NEXT:    v_mov_b32_e32 v20, s14
+; SDAG-NEXT:    v_mov_b32_e32 v21, s15
+; SDAG-NEXT:    v_mov_b32_e32 v22, s0
+; SDAG-NEXT:    v_mov_b32_e32 v23, s1
+; SDAG-NEXT:    v_mov_b32_e32 v24, s2
+; SDAG-NEXT:    v_mov_b32_e32 v25, s3
+; SDAG-NEXT:    v_mov_b32_e32 v16, s16
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_smfmac_i32_32x32x64_i8 v[0:15], v[24:27], v[16:23], v28 cbsz:1 abid:2
+; SDAG-NEXT:    v_smfmac_i32_32x32x64_i8 v[0:15], v[26:29], v[18:25], v16 cbsz:1 abid:2
 ; SDAG-NEXT:    v_mov_b32_e32 v16, 0
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    s_nop 2
@@ -1397,30 +1397,30 @@ define amdgpu_kernel void @test_smfmac_f32_16x16x128_bf8_bf8__vgpr(ptr addrspace
 ; SDAG-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
 ; SDAG-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
 ; SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x34
-; SDAG-NEXT:    v_mov_b32_e32 v16, 0
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    global_load_dwordx4 v[8:11], v0, s[6:7]
+; SDAG-NEXT:    global_load_dwordx4 v[10:13], v0, s[6:7]
 ; SDAG-NEXT:    s_load_dword s16, s[4:5], 0x64
 ; SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x54
-; SDAG-NEXT:    v_mov_b32_e32 v12, s8
-; SDAG-NEXT:    v_mov_b32_e32 v13, s9
-; SDAG-NEXT:    v_mov_b32_e32 v14, s10
-; SDAG-NEXT:    v_mov_b32_e32 v15, s11
-; SDAG-NEXT:    v_mov_b32_e32 v0, s12
-; SDAG-NEXT:    v_mov_b32_e32 v1, s13
-; SDAG-NEXT:    v_mov_b32_e32 v2, s14
-; SDAG-NEXT:    v_mov_b32_e32 v3, s15
+; SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; SDAG-NEXT:    v_mov_b32_e32 v14, s8
+; SDAG-NEXT:    v_mov_b32_e32 v15, s9
+; SDAG-NEXT:    v_mov_b32_e32 v16, s10
+; SDAG-NEXT:    v_mov_b32_e32 v17, s11
+; SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; SDAG-NEXT:    v_mov_b32_e32 v4, s14
+; SDAG-NEXT:    v_mov_b32_e32 v5, s15
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v4, s0
-; SDAG-NEXT:    v_mov_b32_e32 v5, s1
-; SDAG-NEXT:    v_mov_b32_e32 v6, s2
-; SDAG-NEXT:    v_mov_b32_e32 v7, s3
-; SDAG-NEXT:    v_mov_b32_e32 v17, s16
+; SDAG-NEXT:    v_mov_b32_e32 v6, s0
+; SDAG-NEXT:    v_mov_b32_e32 v7, s1
+; SDAG-NEXT:    v_mov_b32_e32 v8, s2
+; SDAG-NEXT:    v_mov_b32_e32 v9, s3
+; SDAG-NEXT:    v_mov_b32_e32 v1, s16
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_smfmac_f32_16x16x128_bf8_bf8 v[8:11], v[12:15], v[0:7], v17 cbsz:1 abid:2
+; SDAG-NEXT:    v_smfmac_f32_16x16x128_bf8_bf8 v[10:13], v[14:17], v[2:9], v1 cbsz:1 abid:2
 ; SDAG-NEXT:    s_nop 7
-; SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[6:7]
+; SDAG-NEXT:    global_store_dwordx4 v0, v[10:13], s[6:7]
 ; SDAG-NEXT:    s_endpgm
 ;
 ; GISEL-LABEL: test_smfmac_f32_16x16x128_bf8_bf8__vgpr:
@@ -1429,24 +1429,24 @@ define amdgpu_kernel void @test_smfmac_f32_16x16x128_bf8_bf8__vgpr(ptr addrspace
 ; GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
 ; GISEL-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
 ; GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GISEL-NEXT:    global_load_dwordx4 v[8:11], v0, s[0:1]
+; GISEL-NEXT:    global_load_dwordx4 v[14:17], v0, s[0:1]
 ; GISEL-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x34
 ; GISEL-NEXT:    s_load_dwordx4 s[16:19], s[4:5], 0x54
 ; GISEL-NEXT:    s_load_dword s2, s[4:5], 0x64
 ; GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[10:11]
-; GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[8:9]
+; GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
 ; GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[12:13]
 ; GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[14:15]
 ; GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[16:17]
 ; GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[18:19]
-; GISEL-NEXT:    v_mov_b32_e32 v16, s2
+; GISEL-NEXT:    v_mov_b32_e32 v12, s2
 ; GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GISEL-NEXT:    s_nop 0
-; GISEL-NEXT:    v_smfmac_f32_16x16x128_bf8_bf8 v[8:11], v[12:15], v[0:7], v16 cbsz:1 abid:2
+; GISEL-NEXT:    v_smfmac_f32_16x16x128_bf8_bf8 v[14:17], v[8:11], v[0:7], v12 cbsz:1 abid:2
 ; GISEL-NEXT:    v_mov_b32_e32 v0, 0
 ; GISEL-NEXT:    s_nop 6
-; GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[0:1]
+; GISEL-NEXT:    global_store_dwordx4 v0, v[14:17], s[0:1]
 ; GISEL-NEXT:    s_endpgm
 bb:
   %id = call i32 @llvm.amdgcn.workitem.id.x()
@@ -1506,25 +1506,25 @@ define <4 x float> @test_smfmac_f32_16x16x128_bf8_bf8__sgpr(<4 x i32> inreg %arg
 ; SDAG-LABEL: test_smfmac_f32_16x16x128_bf8_bf8__sgpr:
 ; SDAG:       ; %bb.0:
 ; SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v8, s0
-; SDAG-NEXT:    v_mov_b32_e32 v9, s1
-; SDAG-NEXT:    v_mov_b32_e32 v10, s2
-; SDAG-NEXT:    v_mov_b32_e32 v11, s3
-; SDAG-NEXT:    v_mov_b32_e32 v0, s16
-; SDAG-NEXT:    v_mov_b32_e32 v1, s17
-; SDAG-NEXT:    v_mov_b32_e32 v2, s18
-; SDAG-NEXT:    v_mov_b32_e32 v3, s19
-; SDAG-NEXT:    v_mov_b32_e32 v4, s20
-; SDAG-NEXT:    v_mov_b32_e32 v5, s21
-; SDAG-NEXT:    v_mov_b32_e32 v6, s22
-; SDAG-NEXT:    v_mov_b32_e32 v7, s23
+; SDAG-NEXT:    v_mov_b32_e32 v10, s0
+; SDAG-NEXT:    v_mov_b32_e32 v11, s1
+; SDAG-NEXT:    v_mov_b32_e32 v12, s2
+; SDAG-NEXT:    v_mov_b32_e32 v13, s3
+; SDAG-NEXT:    v_mov_b32_e32 v2, s16
+; SDAG-NEXT:    v_mov_b32_e32 v3, s17
+; SDAG-NEXT:    v_mov_b32_e32 v4, s18
+; SDAG-NEXT:    v_mov_b32_e32 v5, s19
+; SDAG-NEXT:    v_mov_b32_e32 v6, s20
+; SDAG-NEXT:    v_mov_b32_e32 v7, s21
+; SDAG-NEXT:    v_mov_b32_e32 v8, s22
+; SDAG-NEXT:    v_mov_b32_e32 v9, s23
 ; SDAG-NEXT:    v_accvgpr_write_b32 a0, s24
 ; SDAG-NEXT:    v_accvgpr_write_b32 a1, s25
 ; SDAG-NEXT:    v_accvgpr_write_b32 a2, s26
 ; SDAG-NEXT:    v_accvgpr_write_b32 a3, s27
-; SDAG-NEXT:    v_mov_b32_e32 v12, s28
+; SDAG-NEXT:    v_mov_b32_e32 v0, s28
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_smfmac_f32_16x16x128_bf8_bf8 a[0:3], v[8:11], v[0:7], v12
+; SDAG-NEXT:    v_smfmac_f32_16x16x128_bf8_bf8 a[0:3], v[10:13], v[2:9], v0
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    v_accvgpr_read_b32 v0, a0
 ; SDAG-NEXT:    v_accvgpr_read_b32 v1, a1
@@ -1566,30 +1566,30 @@ define amdgpu_kernel void @test_smfmac_f32_16x16x128_bf8_fp8__vgpr(ptr addrspace
 ; SDAG-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
 ; SDAG-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
 ; SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x34
-; SDAG-NEXT:    v_mov_b32_e32 v16, 0
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    global_load_dwordx4 v[8:11], v0, s[6:7]
+; SDAG-NEXT:    global_load_dwordx4 v[10:13], v0, s[6:7]
 ; SDAG-NEXT:    s_load_dword s16, s[4:5], 0x64
 ; SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x54
-; SDAG-NEXT:    v_mov_b32_e32 v12, s8
-; SDAG-NEXT:    v_mov_b32_e32 v13, s9
-; SDAG-NEXT:    v_mov_b32_e32 v14, s10
-; SDAG-NEXT:    v_mov_b32_e32 v15, s11
-; SDAG-NEXT:    v_mov_b32_e32 v0, s12
-; SDAG-NEXT:    v_mov_b32_e32 v1, s13
-; SDAG-NEXT:    v_mov_b32_e32 v2, s14
-; SDAG-NEXT:    v_mov_b32_e32 v3, s15
+; SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; SDAG-NEXT:    v_mov_b32_e32 v14, s8
+; SDAG-NEXT:    v_mov_b32_e32 v15, s9
+; SDAG-NEXT:    v_mov_b32_e32 v16, s10
+; SDAG-NEXT:    v_mov_b32_e32 v17, s11
+; SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; SDAG-NEXT:    v_mov_b32_e32 v4, s14
+; SDAG-NEXT:    v_mov_b32_e32 v5, s15
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v4, s0
-; SDAG-NEXT:    v_mov_b32_e32 v5, s1
-; SDAG-NEXT:    v_mov_b32_e32 v6, s2
-; SDAG-NEXT:    v_mov_b32_e32 v7, s3
-; SDAG-NEXT:    v_mov_b32_e32 v17, s16
+; SDAG-NEXT:    v_mov_b32_e32 v6, s0
+; SDAG-NEXT:    v_mov_b32_e32 v7, s1
+; SDAG-NEXT:    v_mov_b32_e32 v8, s2
+; SDAG-NEXT:    v_mov_b32_e32 v9, s3
+; SDAG-NEXT:    v_mov_b32_e32 v1, s16
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_smfmac_f32_16x16x128_bf8_fp8 v[8:11], v[12:15], v[0:7], v17 cbsz:1 abid:2
+; SDAG-NEXT:    v_smfmac_f32_16x16x128_bf8_fp8 v[10:13], v[14:17], v[2:9], v1 cbsz:1 abid:2
 ; SDAG-NEXT:    s_nop 7
-; SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[6:7]
+; SDAG-NEXT:    global_store_dwordx4 v0, v[10:13], s[6:7]
 ; SDAG-NEXT:    s_endpgm
 ;
 ; GISEL-LABEL: test_smfmac_f32_16x16x128_bf8_fp8__vgpr:
@@ -1598,24 +1598,24 @@ define amdgpu_kernel void @test_smfmac_f32_16x16x128_bf8_fp8__vgpr(ptr addrspace
 ; GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
 ; GISEL-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
 ; GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GISEL-NEXT:    global_load_dwordx4 v[8:11], v0, s[0:1]
+; GISEL-NEXT:    global_load_dwordx4 v[14:17], v0, s[0:1]
 ; GISEL-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x34
 ; GISEL-NEXT:    s_load_dwordx4 s[16:19], s[4:5], 0x54
 ; GISEL-NEXT:    s_load_dword s2, s[4:5], 0x64
 ; GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[10:11]
-; GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[8:9]
+; GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
 ; GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[12:13]
 ; GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[14:15]
 ; GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[16:17]
 ; GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[18:19]
-; GISEL-NEXT:    v_mov_b32_e32 v16, s2
+; GISEL-NEXT:    v_mov_b32_e32 v12, s2
 ; GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GISEL-NEXT:    s_nop 0
-; GISEL-NEXT:    v_smfmac_f32_16x16x128_bf8_fp8 v[8:11], v[12:15], v[0:7], v16 cbsz:1 abid:2
+; GISEL-NEXT:    v_smfmac_f32_16x16x128_bf8_fp8 v[14:17], v[8:11], v[0:7], v12 cbsz:1 abid:2
 ; GISEL-NEXT:    v_mov_b32_e32 v0, 0
 ; GISEL-NEXT:    s_nop 6
-; GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[0:1]
+; GISEL-NEXT:    global_store_dwordx4 v0, v[14:17], s[0:1]
 ; GISEL-NEXT:    s_endpgm
 bb:
   %id = call i32 @llvm.amdgcn.workitem.id.x()
@@ -1675,25 +1675,25 @@ define <4 x float> @test_smfmac_f32_16x16x128_bf8_fp8__sgpr(<4 x i32> inreg %arg
 ; SDAG-LABEL: test_smfmac_f32_16x16x128_bf8_fp8__sgpr:
 ; SDAG:       ; %bb.0:
 ; SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v8, s0
-; SDAG-NEXT:    v_mov_b32_e32 v9, s1
-; SDAG-NEXT:    v_mov_b32_e32 v10, s2
-; SDAG-NEXT:    v_mov_b32_e32 v11, s3
-; SDAG-NEXT:    v_mov_b32_e32 v0, s16
-; SDAG-NEXT:    v_mov_b32_e32 v1, s17
-; SDAG-NEXT:    v_mov_b32_e32 v2, s18
-; SDAG-NEXT:    v_mov_b32_e32 v3, s19
-; SDAG-NEXT:    v_mov_b32_e32 v4, s20
-; SDAG-NEXT:    v_mov_b32_e32 v5, s21
-; SDAG-NEXT:    v_mov_b32_e32 v6, s22
-; SDAG-NEXT:    v_mov_b32_e32 v7, s23
+; SDAG-NEXT:    v_mov_b32_e32 v10, s0
+; SDAG-NEXT:    v_mov_b32_e32 v11, s1
+; SDAG-NEXT:    v_mov_b32_e32 v12, s2
+; SDAG-NEXT:    v_mov_b32_e32 v13, s3
+; SDAG-NEXT:    v_mov_b32_e32 v2, s16
+; SDAG-NEXT:    v_mov_b32_e32 v3, s17
+; SDAG-NEXT:    v_mov_b32_e32 v4, s18
+; SDAG-NEXT:    v_mov_b32_e32 v5, s19
+; SDAG-NEXT:    v_mov_b32_e32 v6, s20
+; SDAG-NEXT:    v_mov_b32_e32 v7, s21
+; SDAG-NEXT:    v_mov_b32_e32 v8, s22
+; SDAG-NEXT:    v_mov_b32_e32 v9, s23
 ; SDAG-NEXT:    v_accvgpr_write_b32 a0, s24
 ; SDAG-NEXT:    v_accvgpr_write_b32 a1, s25
 ; SDAG-NEXT:    v_accvgpr_write_b32 a2, s26
 ; SDAG-NEXT:    v_accvgpr_write_b32 a3, s27
-; SDAG-NEXT:    v_mov_b32_e32 v12, s28
+; SDAG-NEXT:    v_mov_b32_e32 v0, s28
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_smfmac_f32_16x16x128_bf8_fp8 a[0:3], v[8:11], v[0:7], v12
+; SDAG-NEXT:    v_smfmac_f32_16x16x128_bf8_fp8 a[0:3], v[10:13], v[2:9], v0
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    v_accvgpr_read_b32 v0, a0
 ; SDAG-NEXT:    v_accvgpr_read_b32 v1, a1
@@ -1735,30 +1735,30 @@ define amdgpu_kernel void @test_smfmac_f32_16x16x128_fp8_bf8__vgpr(ptr addrspace
 ; SDAG-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
 ; SDAG-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
 ; SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x34
-; SDAG-NEXT:    v_mov_b32_e32 v16, 0
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    global_load_dwordx4 v[8:11], v0, s[6:7]
+; SDAG-NEXT:    global_load_dwordx4 v[10:13], v0, s[6:7]
 ; SDAG-NEXT:    s_load_dword s16, s[4:5], 0x64
 ; SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x54
-; SDAG-NEXT:    v_mov_b32_e32 v12, s8
-; SDAG-NEXT:    v_mov_b32_e32 v13, s9
-; SDAG-NEXT:    v_mov_b32_e32 v14, s10
-; SDAG-NEXT:    v_mov_b32_e32 v15, s11
-; SDAG-NEXT:    v_mov_b32_e32 v0, s12
-; SDAG-NEXT:    v_mov_b32_e32 v1, s13
-; SDAG-NEXT:    v_mov_b32_e32 v2, s14
-; SDAG-NEXT:    v_mov_b32_e32 v3, s15
+; SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; SDAG-NEXT:    v_mov_b32_e32 v14, s8
+; SDAG-NEXT:    v_mov_b32_e32 v15, s9
+; SDAG-NEXT:    v_mov_b32_e32 v16, s10
+; SDAG-NEXT:    v_mov_b32_e32 v17, s11
+; SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; SDAG-NEXT:    v_mov_b32_e32 v4, s14
+; SDAG-NEXT:    v_mov_b32_e32 v5, s15
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v4, s0
-; SDAG-NEXT:    v_mov_b32_e32 v5, s1
-; SDAG-NEXT:    v_mov_b32_e32 v6, s2
-; SDAG-NEXT:    v_mov_b32_e32 v7, s3
-; SDAG-NEXT:    v_mov_b32_e32 v17, s16
+; SDAG-NEXT:    v_mov_b32_e32 v6, s0
+; SDAG-NEXT:    v_mov_b32_e32 v7, s1
+; SDAG-NEXT:    v_mov_b32_e32 v8, s2
+; SDAG-NEXT:    v_mov_b32_e32 v9, s3
+; SDAG-NEXT:    v_mov_b32_e32 v1, s16
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_smfmac_f32_16x16x128_fp8_bf8 v[8:11], v[12:15], v[0:7], v17 cbsz:1 abid:2
+; SDAG-NEXT:    v_smfmac_f32_16x16x128_fp8_bf8 v[10:13], v[14:17], v[2:9], v1 cbsz:1 abid:2
 ; SDAG-NEXT:    s_nop 7
-; SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[6:7]
+; SDAG-NEXT:    global_store_dwordx4 v0, v[10:13], s[6:7]
 ; SDAG-NEXT:    s_endpgm
 ;
 ; GISEL-LABEL: test_smfmac_f32_16x16x128_fp8_bf8__vgpr:
@@ -1767,24 +1767,24 @@ define amdgpu_kernel void @test_smfmac_f32_16x16x128_fp8_bf8__vgpr(ptr addrspace
 ; GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
 ; GISEL-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
 ; GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GISEL-NEXT:    global_load_dwordx4 v[8:11], v0, s[0:1]
+; GISEL-NEXT:    global_load_dwordx4 v[14:17], v0, s[0:1]
 ; GISEL-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x34
 ; GISEL-NEXT:    s_load_dwordx4 s[16:19], s[4:5], 0x54
 ; GISEL-NEXT:    s_load_dword s2, s[4:5], 0x64
 ; GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[10:11]
-; GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[8:9]
+; GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
 ; GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[12:13]
 ; GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[14:15]
 ; GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[16:17]
 ; GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[18:19]
-; GISEL-NEXT:    v_mov_b32_e32 v16, s2
+; GISEL-NEXT:    v_mov_b32_e32 v12, s2
 ; GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GISEL-NEXT:    s_nop 0
-; GISEL-NEXT:    v_smfmac_f32_16x16x128_fp8_bf8 v[8:11], v[12:15], v[0:7], v16 cbsz:1 abid:2
+; GISEL-NEXT:    v_smfmac_f32_16x16x128_fp8_bf8 v[14:17], v[8:11], v[0:7], v12 cbsz:1 abid:2
 ; GISEL-NEXT:    v_mov_b32_e32 v0, 0
 ; GISEL-NEXT:    s_nop 6
-; GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[0:1]
+; GISEL-NEXT:    global_store_dwordx4 v0, v[14:17], s[0:1]
 ; GISEL-NEXT:    s_endpgm
 bb:
   %id = call i32 @llvm.amdgcn.workitem.id.x()
@@ -1844,25 +1844,25 @@ define <4 x float> @test_smfmac_f32_16x16x128_fp8_bf8__sgpr(<4 x i32> inreg %arg
 ; SDAG-LABEL: test_smfmac_f32_16x16x128_fp8_bf8__sgpr:
 ; SDAG:       ; %bb.0:
 ; SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v8, s0
-; SDAG-NEXT:    v_mov_b32_e32 v9, s1
-; SDAG-NEXT:    v_mov_b32_e32 v10, s2
-; SDAG-NEXT:    v_mov_b32_e32 v11, s3
-; SDAG-NEXT:    v_mov_b32_e32 v0, s16
-; SDAG-NEXT:    v_mov_b32_e32 v1, s17
-; SDAG-NEXT:    v_mov_b32_e32 v2, s18
-; SDAG-NEXT:    v_mov_b32_e32 v3, s19
-; SDAG-NEXT:    v_mov_b32_e32 v4, s20
-; SDAG-NEXT:    v_mov_b32_e32 v5, s21
-; SDAG-NEXT:    v_mov_b32_e32 v6, s22
-; SDAG-NEXT:    v_mov_b32_e32 v7, s23
+; SDAG-NEXT:    v_mov_b32_e32 v10, s0
+; SDAG-NEXT:    v_mov_b32_e32 v11, s1
+; SDAG-NEXT:    v_mov_b32_e32 v12, s2
+; SDAG-NEXT:    v_mov_b32_e32 v13, s3
+; SDAG-NEXT:    v_mov_b32_e32 v2, s16
+; SDAG-NEXT:    v_mov_b32_e32 v3, s17
+; SDAG-NEXT:    v_mov_b32_e32 v4, s18
+; SDAG-NEXT:    v_mov_b32_e32 v5, s19
+; SDAG-NEXT:    v_mov_b32_e32 v6, s20
+; SDAG-NEXT:    v_mov_b32_e32 v7, s21
+; SDAG-NEXT:    v_mov_b32_e32 v8, s22
+; SDAG-NEXT:    v_mov_b32_e32 v9, s23
 ; SDAG-NEXT:    v_accvgpr_write_b32 a0, s24
 ; SDAG-NEXT:    v_accvgpr_write_b32 a1, s25
 ; SDAG-NEXT:    v_accvgpr_write_b32 a2, s26
 ; SDAG-NEXT:    v_accvgpr_write_b32 a3, s27
-; SDAG-NEXT:    v_mov_b32_e32 v12, s28
+; SDAG-NEXT:    v_mov_b32_e32 v0, s28
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_smfmac_f32_16x16x128_fp8_bf8 a[0:3], v[8:11], v[0:7], v12
+; SDAG-NEXT:    v_smfmac_f32_16x16x128_fp8_bf8 a[0:3], v[10:13], v[2:9], v0
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    v_accvgpr_read_b32 v0, a0
 ; SDAG-NEXT:    v_accvgpr_read_b32 v1, a1
@@ -1904,30 +1904,30 @@ define amdgpu_kernel void @test_smfmac_f32_16x16x128_fp8_fp8__vgpr(ptr addrspace
 ; SDAG-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
 ; SDAG-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
 ; SDAG-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x34
-; SDAG-NEXT:    v_mov_b32_e32 v16, 0
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    global_load_dwordx4 v[8:11], v0, s[6:7]
+; SDAG-NEXT:    global_load_dwordx4 v[10:13], v0, s[6:7]
 ; SDAG-NEXT:    s_load_dword s16, s[4:5], 0x64
 ; SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x54
-; SDAG-NEXT:    v_mov_b32_e32 v12, s8
-; SDAG-NEXT:    v_mov_b32_e32 v13, s9
-; SDAG-NEXT:    v_mov_b32_e32 v14, s10
-; SDAG-NEXT:    v_mov_b32_e32 v15, s11
-; SDAG-NEXT:    v_mov_b32_e32 v0, s12
-; SDAG-NEXT:    v_mov_b32_e32 v1, s13
-; SDAG-NEXT:    v_mov_b32_e32 v2, s14
-; SDAG-NEXT:    v_mov_b32_e32 v3, s15
+; SDAG-NEXT:    v_mov_b32_e32 v0, 0
+; SDAG-NEXT:    v_mov_b32_e32 v14, s8
+; SDAG-NEXT:    v_mov_b32_e32 v15, s9
+; SDAG-NEXT:    v_mov_b32_e32 v16, s10
+; SDAG-NEXT:    v_mov_b32_e32 v17, s11
+; SDAG-NEXT:    v_mov_b32_e32 v2, s12
+; SDAG-NEXT:    v_mov_b32_e32 v3, s13
+; SDAG-NEXT:    v_mov_b32_e32 v4, s14
+; SDAG-NEXT:    v_mov_b32_e32 v5, s15
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v4, s0
-; SDAG-NEXT:    v_mov_b32_e32 v5, s1
-; SDAG-NEXT:    v_mov_b32_e32 v6, s2
-; SDAG-NEXT:    v_mov_b32_e32 v7, s3
-; SDAG-NEXT:    v_mov_b32_e32 v17, s16
+; SDAG-NEXT:    v_mov_b32_e32 v6, s0
+; SDAG-NEXT:    v_mov_b32_e32 v7, s1
+; SDAG-NEXT:    v_mov_b32_e32 v8, s2
+; SDAG-NEXT:    v_mov_b32_e32 v9, s3
+; SDAG-NEXT:    v_mov_b32_e32 v1, s16
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_smfmac_f32_16x16x128_fp8_fp8 v[8:11], v[12:15], v[0:7], v17 cbsz:1 abid:2
+; SDAG-NEXT:    v_smfmac_f32_16x16x128_fp8_fp8 v[10:13], v[14:17], v[2:9], v1 cbsz:1 abid:2
 ; SDAG-NEXT:    s_nop 7
-; SDAG-NEXT:    global_store_dwordx4 v16, v[8:11], s[6:7]
+; SDAG-NEXT:    global_store_dwordx4 v0, v[10:13], s[6:7]
 ; SDAG-NEXT:    s_endpgm
 ;
 ; GISEL-LABEL: test_smfmac_f32_16x16x128_fp8_fp8__vgpr:
@@ -1936,24 +1936,24 @@ define amdgpu_kernel void @test_smfmac_f32_16x16x128_fp8_fp8__vgpr(ptr addrspace
 ; GISEL-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
 ; GISEL-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
 ; GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GISEL-NEXT:    global_load_dwordx4 v[8:11], v0, s[0:1]
+; GISEL-NEXT:    global_load_dwordx4 v[14:17], v0, s[0:1]
 ; GISEL-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x34
 ; GISEL-NEXT:    s_load_dwordx4 s[16:19], s[4:5], 0x54
 ; GISEL-NEXT:    s_load_dword s2, s[4:5], 0x64
 ; GISEL-NEXT:    s_waitcnt lgkmcnt(0)
-; GISEL-NEXT:    v_mov_b64_e32 v[14:15], s[10:11]
-; GISEL-NEXT:    v_mov_b64_e32 v[12:13], s[8:9]
+; GISEL-NEXT:    v_mov_b64_e32 v[8:9], s[8:9]
+; GISEL-NEXT:    v_mov_b64_e32 v[10:11], s[10:11]
 ; GISEL-NEXT:    v_mov_b64_e32 v[0:1], s[12:13]
 ; GISEL-NEXT:    v_mov_b64_e32 v[2:3], s[14:15]
 ; GISEL-NEXT:    v_mov_b64_e32 v[4:5], s[16:17]
 ; GISEL-NEXT:    v_mov_b64_e32 v[6:7], s[18:19]
-; GISEL-NEXT:    v_mov_b32_e32 v16, s2
+; GISEL-NEXT:    v_mov_b32_e32 v12, s2
 ; GISEL-NEXT:    s_waitcnt vmcnt(0)
 ; GISEL-NEXT:    s_nop 0
-; GISEL-NEXT:    v_smfmac_f32_16x16x128_fp8_fp8 v[8:11], v[12:15], v[0:7], v16 cbsz:1 abid:2
+; GISEL-NEXT:    v_smfmac_f32_16x16x128_fp8_fp8 v[14:17], v[8:11], v[0:7], v12 cbsz:1 abid:2
 ; GISEL-NEXT:    v_mov_b32_e32 v0, 0
 ; GISEL-NEXT:    s_nop 6
-; GISEL-NEXT:    global_store_dwordx4 v0, v[8:11], s[0:1]
+; GISEL-NEXT:    global_store_dwordx4 v0, v[14:17], s[0:1]
 ; GISEL-NEXT:    s_endpgm
 bb:
   %id = call i32 @llvm.amdgcn.workitem.id.x()
@@ -2013,25 +2013,25 @@ define <4 x float> @test_smfmac_f32_16x16x128_fp8_fp8__sgpr(<4 x i32> inreg %arg
 ; SDAG-LABEL: test_smfmac_f32_16x16x128_fp8_fp8__sgpr:
 ; SDAG:       ; %bb.0:
 ; SDAG-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v8, s0
-; SDAG-NEXT:    v_mov_b32_e32 v9, s1
-; SDAG-NEXT:    v_mov_b32_e32 v10, s2
-; SDAG-NEXT:    v_mov_b32_e32 v11, s3
-; SDAG-NEXT:    v_mov_b32_e32 v0, s16
-; SDAG-NEXT:    v_mov_b32_e32 v1, s17
-; SDAG-NEXT:    v_mov_b32_e32 v2, s18
-; SDAG-NEXT:    v_mov_b32_e32 v3, s19
-; SDAG-NEXT:    v_mov_b32_e32 v4, s20
-; SDAG-NEXT:    v_mov_b32_e32 v5, s21
-; SDAG-NEXT:    v_mov_b32_e32 v6, s22
-; SDAG-NEXT:    v_mov_b32_e32 v7, s23
+; SDAG-NEXT:    v_mov_b32_e32 v10, s0
+; SDAG-NEXT:    v_mov_b32_e32 v11, s1
+; SDAG-NEXT:    v_mov_b32_e32 v12, s2
+; SDAG-NEXT:    v_mov_b32_e32 v13, s3
+; SDAG-NEXT:    v_mov_b32_e32 v2, s16
+; SDAG-NEXT:    v_mov_b32_e32 v3, s17
+; SDAG-NEXT:    v_mov_b32_e32 v4, s18
+; SDAG-NEXT:    v_mov_b32_e32 v5, s19
+; SDAG-NEXT:    v_mov_b32_e32 v6, s20
+; SDAG-NEXT:    v_mov_b32_e32 v7, s21
+; SDAG-NEXT:    v_mov_b32_e32 v8, s22
+; SDAG-NEXT:    v_mov_b32_e32 v9, s23
 ; SDAG-NEXT:    v_accvgpr_write_b32 a0, s24
 ; SDAG-NEXT:    v_accvgpr_write_b32 a1, s25
 ; SDAG-NEXT:    v_accvgpr_write_b32 a2, s26
 ; SDAG-NEXT:    v_accvgpr_write_b32 a3, s27
-; SDAG-NEXT:    v_mov_b32_e32 v12, s28
+; SDAG-NEXT:    v_mov_b32_e32 v0, s28
 ; SDAG-NEXT:    s_nop 1
-; SDAG-NEXT:    v_smfmac_f32_16x16x128_fp8_fp8 a[0:3], v[8:11], v[0:7], v12
+; SDAG-NEXT:    v_smfmac_f32_16x16x128_fp8_fp8 a[0:3], v[10:13], v[2:9], v0
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    v_accvgpr_read_b32 v0, a0
 ; SDAG-NEXT:    v_accvgpr_read_b32 v1, a1
@@ -2081,22 +2081,22 @@ define amdgpu_kernel void @test_smfmac_f32_32x32x64_bf8_bf8__vgpr(ptr addrspace(
 ; SDAG-NEXT:    s_load_dword s16, s[4:5], 0x64
 ; SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x54
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v24, s8
-; SDAG-NEXT:    v_mov_b32_e32 v25, s9
-; SDAG-NEXT:    v_mov_b32_e32 v26, s10
-; SDAG-NEXT:    v_mov_b32_e32 v27, s11
-; SDAG-NEXT:    v_mov_b32_e32 v16, s12
-; SDAG-NEXT:    v_mov_b32_e32 v17, s13
-; SDAG-NEXT:    v_mov_b32_e32 v18, s14
-; SDAG-NEXT:    v_mov_b32_e32 v19, s15
-; SDAG-NEXT:    v_mov_b32_e32 v20, s0
-; SDAG-NEXT:    v_mov_b32_e32 v21, s1
-; SDAG-NEXT:    v_mov_b32_e32 v22, s2
-; SDAG-NEXT:    v_mov_b32_e32 v23, s3
-; SDAG-NEXT:    v_mov_b32_e32 v28, s16
+; SDAG-NEXT:    v_mov_b32_e32 v26, s8
+; SDAG-NEXT:    v_mov_b32_e32 v27, s9
+; SDAG-NEXT:    v_mov_b32_e32 v28, s10
+; SDAG-NEXT:    v_mov_b32_e32 v29, s11
+; SDAG-NEXT:    v_mov_b32_e32 v18, s12
+; SDAG-NEXT:    v_mov_b32_e32 v19, s13
+; SDAG-NEXT:    v_mov_b32_e32 v20, s14
+; SDAG-NEXT:    v_mov_b32_e32 v21, s15
+; SDAG-NEXT:    v_mov_b32_e32 v22, s0
+; SDAG-NEXT:    v_mov_b32_e32 v23, s1
+; SDAG-NEXT:    v_mov_b32_e32 v24, s2
+; SDAG-NEXT:    v_mov_b32_e32 v25, s3
+; SDAG-NEXT:    v_mov_b32_e32 v16, s16
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_smfmac_f32_32x32x64_bf8_bf8 v[0:15], v[24:27], v[16:23], v28 cbsz:1 abid:2
+; SDAG-NEXT:    v_smfmac_f32_32x32x64_bf8_bf8 v[0:15], v[26:29], v[18:25], v16 cbsz:1 abid:2
 ; SDAG-NEXT:    v_mov_b32_e32 v16, 0
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    s_nop 2
@@ -2454,22 +2454,22 @@ define amdgpu_kernel void @test_smfmac_f32_32x32x64_bf8_fp8__vgpr(ptr addrspace(
 ; SDAG-NEXT:    s_load_dword s16, s[4:5], 0x64
 ; SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x54
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v24, s8
-; SDAG-NEXT:    v_mov_b32_e32 v25, s9
-; SDAG-NEXT:    v_mov_b32_e32 v26, s10
-; SDAG-NEXT:    v_mov_b32_e32 v27, s11
-; SDAG-NEXT:    v_mov_b32_e32 v16, s12
-; SDAG-NEXT:    v_mov_b32_e32 v17, s13
-; SDAG-NEXT:    v_mov_b32_e32 v18, s14
-; SDAG-NEXT:    v_mov_b32_e32 v19, s15
-; SDAG-NEXT:    v_mov_b32_e32 v20, s0
-; SDAG-NEXT:    v_mov_b32_e32 v21, s1
-; SDAG-NEXT:    v_mov_b32_e32 v22, s2
-; SDAG-NEXT:    v_mov_b32_e32 v23, s3
-; SDAG-NEXT:    v_mov_b32_e32 v28, s16
+; SDAG-NEXT:    v_mov_b32_e32 v26, s8
+; SDAG-NEXT:    v_mov_b32_e32 v27, s9
+; SDAG-NEXT:    v_mov_b32_e32 v28, s10
+; SDAG-NEXT:    v_mov_b32_e32 v29, s11
+; SDAG-NEXT:    v_mov_b32_e32 v18, s12
+; SDAG-NEXT:    v_mov_b32_e32 v19, s13
+; SDAG-NEXT:    v_mov_b32_e32 v20, s14
+; SDAG-NEXT:    v_mov_b32_e32 v21, s15
+; SDAG-NEXT:    v_mov_b32_e32 v22, s0
+; SDAG-NEXT:    v_mov_b32_e32 v23, s1
+; SDAG-NEXT:    v_mov_b32_e32 v24, s2
+; SDAG-NEXT:    v_mov_b32_e32 v25, s3
+; SDAG-NEXT:    v_mov_b32_e32 v16, s16
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_smfmac_f32_32x32x64_bf8_fp8 v[0:15], v[24:27], v[16:23], v28 cbsz:1 abid:2
+; SDAG-NEXT:    v_smfmac_f32_32x32x64_bf8_fp8 v[0:15], v[26:29], v[18:25], v16 cbsz:1 abid:2
 ; SDAG-NEXT:    v_mov_b32_e32 v16, 0
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    s_nop 2
@@ -2827,22 +2827,22 @@ define amdgpu_kernel void @test_smfmac_f32_32x32x64_fp8_bf8__vgpr(ptr addrspace(
 ; SDAG-NEXT:    s_load_dword s16, s[4:5], 0x64
 ; SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x54
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v24, s8
-; SDAG-NEXT:    v_mov_b32_e32 v25, s9
-; SDAG-NEXT:    v_mov_b32_e32 v26, s10
-; SDAG-NEXT:    v_mov_b32_e32 v27, s11
-; SDAG-NEXT:    v_mov_b32_e32 v16, s12
-; SDAG-NEXT:    v_mov_b32_e32 v17, s13
-; SDAG-NEXT:    v_mov_b32_e32 v18, s14
-; SDAG-NEXT:    v_mov_b32_e32 v19, s15
-; SDAG-NEXT:    v_mov_b32_e32 v20, s0
-; SDAG-NEXT:    v_mov_b32_e32 v21, s1
-; SDAG-NEXT:    v_mov_b32_e32 v22, s2
-; SDAG-NEXT:    v_mov_b32_e32 v23, s3
-; SDAG-NEXT:    v_mov_b32_e32 v28, s16
+; SDAG-NEXT:    v_mov_b32_e32 v26, s8
+; SDAG-NEXT:    v_mov_b32_e32 v27, s9
+; SDAG-NEXT:    v_mov_b32_e32 v28, s10
+; SDAG-NEXT:    v_mov_b32_e32 v29, s11
+; SDAG-NEXT:    v_mov_b32_e32 v18, s12
+; SDAG-NEXT:    v_mov_b32_e32 v19, s13
+; SDAG-NEXT:    v_mov_b32_e32 v20, s14
+; SDAG-NEXT:    v_mov_b32_e32 v21, s15
+; SDAG-NEXT:    v_mov_b32_e32 v22, s0
+; SDAG-NEXT:    v_mov_b32_e32 v23, s1
+; SDAG-NEXT:    v_mov_b32_e32 v24, s2
+; SDAG-NEXT:    v_mov_b32_e32 v25, s3
+; SDAG-NEXT:    v_mov_b32_e32 v16, s16
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_smfmac_f32_32x32x64_fp8_bf8 v[0:15], v[24:27], v[16:23], v28 cbsz:1 abid:2
+; SDAG-NEXT:    v_smfmac_f32_32x32x64_fp8_bf8 v[0:15], v[26:29], v[18:25], v16 cbsz:1 abid:2
 ; SDAG-NEXT:    v_mov_b32_e32 v16, 0
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    s_nop 2
@@ -3200,22 +3200,22 @@ define amdgpu_kernel void @test_smfmac_f32_32x32x64_fp8_fp8__vgpr(ptr addrspace(
 ; SDAG-NEXT:    s_load_dword s16, s[4:5], 0x64
 ; SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x54
 ; SDAG-NEXT:    s_waitcnt lgkmcnt(0)
-; SDAG-NEXT:    v_mov_b32_e32 v24, s8
-; SDAG-NEXT:    v_mov_b32_e32 v25, s9
-; SDAG-NEXT:    v_mov_b32_e32 v26, s10
-; SDAG-NEXT:    v_mov_b32_e32 v27, s11
-; SDAG-NEXT:    v_mov_b32_e32 v16, s12
-; SDAG-NEXT:    v_mov_b32_e32 v17, s13
-; SDAG-NEXT:    v_mov_b32_e32 v18, s14
-; SDAG-NEXT:    v_mov_b32_e32 v19, s15
-; SDAG-NEXT:    v_mov_b32_e32 v20, s0
-; SDAG-NEXT:    v_mov_b32_e32 v21, s1
-; SDAG-NEXT:    v_mov_b32_e32 v22, s2
-; SDAG-NEXT:    v_mov_b32_e32 v23, s3
-; SDAG-NEXT:    v_mov_b32_e32 v28, s16
+; SDAG-NEXT:    v_mov_b32_e32 v26, s8
+; SDAG-NEXT:    v_mov_b32_e32 v27, s9
+; SDAG-NEXT:    v_mov_b32_e32 v28, s10
+; SDAG-NEXT:    v_mov_b32_e32 v29, s11
+; SDAG-NEXT:    v_mov_b32_e32 v18, s12
+; SDAG-NEXT:    v_mov_b32_e32 v19, s13
+; SDAG-NEXT:    v_mov_b32_e32 v20, s14
+; SDAG-NEXT:    v_mov_b32_e32 v21, s15
+; SDAG-NEXT:    v_mov_b32_e32 v22, s0
+; SDAG-NEXT:    v_mov_b32_e32 v23, s1
+; SDAG-NEXT:    v_mov_b32_e32 v24, s2
+; SDAG-NEXT:    v_mov_b32_e32 v25, s3
+; SDAG-NEXT:    v_mov_b32_e32 v16, s16
 ; SDAG-NEXT:    s_waitcnt vmcnt(0)
 ; SDAG-NEXT:    s_nop 0
-; SDAG-NEXT:    v_smfmac_f32_32x32x64_fp8_fp8 v[0:15], v[24:27], v[16:23], v28 cbsz:1 abid:2
+; SDAG-NEXT:    v_smfmac_f32_32x32x64_fp8_fp8 v[0:15], v[26:29], v[18:25], v16 cbsz:1 abid:2
 ; SDAG-NEXT:    v_mov_b32_e32 v16, 0
 ; SDAG-NEXT:    s_nop 7
 ; SDAG-NEXT:    s_nop 2
diff --git a/llvm/test/CodeGen/AMDGPU/mad-mix-hi-bf16.ll b/llvm/test/CodeGen/AMDGPU/mad-mix-hi-bf16.ll
index 5b2de59..84123e6 100644
--- a/llvm/test/CodeGen/AMDGPU/mad-mix-hi-bf16.ll
+++ b/llvm/test/CodeGen/AMDGPU/mad-mix-hi-bf16.ll
@@ -121,18 +121,7 @@ define <2 x bfloat> @v_mad_mixhi_bf16_bf16lo_bf16lo_bf16lo_undeflo_clamp_postcvt
 ; GFX1250:       ; %bb.0:
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
-; GFX1250-NEXT:    v_fma_mixlo_bf16 v0, v0, v1, v2 op_sel_hi:[1,1,1]
-; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
-; GFX1250-NEXT:    v_max_num_f32_e32 v0, 0, v0
-; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-NEXT:    v_cvt_pk_bf16_f32 v0, v0, s0
-; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
-; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-NEXT:    v_min_num_f32_e32 v0, 1.0, v0
-; GFX1250-NEXT:    v_cvt_pk_bf16_f32 v0, v0, s0
-; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
+; GFX1250-NEXT:    v_fma_mixhi_bf16 v0, v0, v1, v2 op_sel_hi:[1,1,1] clamp
 ; GFX1250-NEXT:    s_set_pc_i64 s[30:31]
   %src0.ext = fpext bfloat %src0 to float
   %src1.ext = fpext bfloat %src1 to float
@@ -150,20 +139,10 @@ define <2 x bfloat> @v_mad_mixhi_bf16_bf16lo_bf16lo_bf16lo_undeflo_clamp_postcvt
 ; GFX1250:       ; %bb.0:
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
-; GFX1250-NEXT:    v_fma_mixlo_bf16 v1, v0, v1, v2 op_sel_hi:[1,1,1]
-; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 16, v1
-; GFX1250-NEXT:    v_max_num_f32_e32 v0, 0, v0
-; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-NEXT:    v_cvt_pk_bf16_f32 v0, v0, s0
-; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
-; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-NEXT:    v_min_num_f32_e32 v0, 1.0, v0
-; GFX1250-NEXT:    v_cvt_pk_bf16_f32 v0, v0, s0
-; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
-; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
+; GFX1250-NEXT:    v_fma_mixlo_bf16 v3, v0, v1, v2 op_sel_hi:[1,1,1]
+; GFX1250-NEXT:    v_fma_mixhi_bf16 v0, v0, v1, v2 op_sel_hi:[1,1,1] clamp
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
-; GFX1250-NEXT:    global_store_b16 v[0:1], v1, off scope:SCOPE_SYS
+; GFX1250-NEXT:    global_store_b16 v[0:1], v3, off scope:SCOPE_SYS
 ; GFX1250-NEXT:    s_wait_storecnt 0x0
 ; GFX1250-NEXT:    s_set_pc_i64 s[30:31]
   %src0.ext = fpext bfloat %src0 to float
diff --git a/llvm/test/CodeGen/AMDGPU/mad-mix-lo-bf16.ll b/llvm/test/CodeGen/AMDGPU/mad-mix-lo-bf16.ll
index 557080a..1b2eb83 100644
--- a/llvm/test/CodeGen/AMDGPU/mad-mix-lo-bf16.ll
+++ b/llvm/test/CodeGen/AMDGPU/mad-mix-lo-bf16.ll
@@ -75,15 +75,8 @@ define bfloat @v_mad_mixlo_bf16_bf16lo_bf16lo_f32_clamp_post_cvt(bfloat %src0, b
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
 ; GFX1250-NEXT:    v_fma_mixlo_bf16 v0, v0, v1, v2 op_sel_hi:[1,1,0]
-; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
-; GFX1250-NEXT:    v_max_num_f32_e32 v0, 0, v0
-; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-NEXT:    v_cvt_pk_bf16_f32 v0, v0, s0
-; GFX1250-NEXT:    v_lshlrev_b32_e32 v0, 16, v0
-; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-NEXT:    v_min_num_f32_e32 v0, 1.0, v0
-; GFX1250-NEXT:    v_cvt_pk_bf16_f32 v0, v0, s0
+; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1250-NEXT:    v_pk_max_num_bf16 v0, v0, v0 clamp
 ; GFX1250-NEXT:    s_set_pc_i64 s[30:31]
   %src0.ext = fpext bfloat %src0 to float
   %src1.ext = fpext bfloat %src1 to float
@@ -199,9 +192,8 @@ define <2 x bfloat> @v_mad_mix_v2f32_clamp_postcvt(<2 x bfloat> %src0, <2 x bflo
 ; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX1250-NEXT:    v_pk_fma_f32 v[0:1], v[4:5], v[6:7], v[0:1]
 ; GFX1250-NEXT:    v_cvt_pk_bf16_f32 v0, v0, v1
-; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-NEXT:    v_pk_max_num_bf16 v0, v0, 0
-; GFX1250-NEXT:    v_pk_min_num_bf16 v0, v0, 1.0 op_sel_hi:[1,0]
+; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1250-NEXT:    v_pk_max_num_bf16 v0, v0, v0 clamp
 ; GFX1250-NEXT:    s_set_pc_i64 s[30:31]
   %src0.ext = fpext <2 x bfloat> %src0 to <2 x float>
   %src1.ext = fpext <2 x bfloat> %src1 to <2 x float>
@@ -219,16 +211,13 @@ define <3 x bfloat> @v_mad_mix_v3f32_clamp_postcvt(<3 x bfloat> %src0, <3 x bflo
 ; GFX1250:       ; %bb.0:
 ; GFX1250-NEXT:    s_wait_loadcnt_dscnt 0x0
 ; GFX1250-NEXT:    s_wait_kmcnt 0x0
-; GFX1250-NEXT:    v_fma_mixlo_bf16 v6, v0, v2, v4 op_sel_hi:[1,1,1]
-; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(SKIP_1) | instid1(VALU_DEP_2)
-; GFX1250-NEXT:    v_fma_mixhi_bf16 v6, v0, v2, v4 op_sel:[1,1,1] op_sel_hi:[1,1,1]
-; GFX1250-NEXT:    v_fma_mixlo_bf16 v0, v1, v3, v5 op_sel_hi:[1,1,1]
-; GFX1250-NEXT:    v_pk_max_num_bf16 v1, v6, 0
+; GFX1250-NEXT:    v_fma_mixlo_bf16 v6, v0, v2, v4 op_sel_hi:[1,1,1] clamp
+; GFX1250-NEXT:    v_fma_mixlo_bf16 v1, v1, v3, v5 op_sel_hi:[1,1,1]
 ; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1250-NEXT:    v_pk_max_num_bf16 v2, v0, 0
-; GFX1250-NEXT:    v_pk_min_num_bf16 v0, v1, 1.0 op_sel_hi:[1,0]
+; GFX1250-NEXT:    v_fma_mixhi_bf16 v6, v0, v2, v4 op_sel:[1,1,1] op_sel_hi:[1,1,1] clamp
+; GFX1250-NEXT:    v_pk_max_num_bf16 v1, v1, v1 clamp
 ; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_2)
-; GFX1250-NEXT:    v_pk_min_num_bf16 v1, v2, 1.0
+; GFX1250-NEXT:    v_mov_b32_e32 v0, v6
 ; GFX1250-NEXT:    s_set_pc_i64 s[30:31]
   %src0.ext = fpext <3 x bfloat> %src0 to <3 x float>
   %src1.ext = fpext <3 x bfloat> %src1 to <3 x float>
@@ -261,11 +250,8 @@ define <4 x bfloat> @v_mad_mix_v4f32_clamp_postcvt(<4 x bfloat> %src0, <4 x bflo
 ; GFX1250-NEXT:    v_cvt_pk_bf16_f32 v0, v0, v1
 ; GFX1250-NEXT:    v_cvt_pk_bf16_f32 v1, v2, v3
 ; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1250-NEXT:    v_pk_max_num_bf16 v0, v0, 0
-; GFX1250-NEXT:    v_pk_max_num_bf16 v1, v1, 0
-; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_2) | instskip(NEXT) | instid1(VALU_DEP_2)
-; GFX1250-NEXT:    v_pk_min_num_bf16 v0, v0, 1.0 op_sel_hi:[1,0]
-; GFX1250-NEXT:    v_pk_min_num_bf16 v1, v1, 1.0 op_sel_hi:[1,0]
+; GFX1250-NEXT:    v_pk_max_num_bf16 v0, v0, v0 clamp
+; GFX1250-NEXT:    v_pk_max_num_bf16 v1, v1, v1 clamp
 ; GFX1250-NEXT:    s_set_pc_i64 s[30:31]
   %src0.ext = fpext <4 x bfloat> %src0 to <4 x float>
   %src1.ext = fpext <4 x bfloat> %src1 to <4 x float>
@@ -291,15 +277,7 @@ define <2 x bfloat> @v_mad_mix_v2f32_clamp_postcvt_lo(<2 x bfloat> %src0, <2 x b
 ; GFX1250-NEXT:    v_pk_fma_f32 v[0:1], v[4:5], v[6:7], v[0:1]
 ; GFX1250-NEXT:    v_cvt_pk_bf16_f32 v0, v0, v1
 ; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-NEXT:    v_lshlrev_b32_e32 v1, 16, v0
-; GFX1250-NEXT:    v_max_num_f32_e32 v1, 0, v1
-; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-NEXT:    v_cvt_pk_bf16_f32 v1, v1, s0
-; GFX1250-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
-; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-NEXT:    v_min_num_f32_e32 v1, 1.0, v1
-; GFX1250-NEXT:    v_cvt_pk_bf16_f32 v1, v1, s0
-; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
+; GFX1250-NEXT:    v_pk_max_num_bf16 v1, v0, v0 clamp
 ; GFX1250-NEXT:    v_bfi_b32 v0, 0xffff, v1, v0
 ; GFX1250-NEXT:    s_set_pc_i64 s[30:31]
   %src0.ext = fpext <2 x bfloat> %src0 to <2 x float>
@@ -328,14 +306,8 @@ define <2 x bfloat> @v_mad_mix_v2f32_clamp_postcvt_hi(<2 x bfloat> %src0, <2 x b
 ; GFX1250-NEXT:    v_pk_fma_f32 v[0:1], v[4:5], v[6:7], v[0:1]
 ; GFX1250-NEXT:    v_cvt_pk_bf16_f32 v0, v0, v1
 ; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-NEXT:    v_and_b32_e32 v1, 0xffff0000, v0
-; GFX1250-NEXT:    v_max_num_f32_e32 v1, 0, v1
-; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-NEXT:    v_cvt_pk_bf16_f32 v1, v1, s0
-; GFX1250-NEXT:    v_lshlrev_b32_e32 v1, 16, v1
-; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
-; GFX1250-NEXT:    v_min_num_f32_e32 v1, 1.0, v1
-; GFX1250-NEXT:    v_cvt_pk_bf16_f32 v1, v1, s0
+; GFX1250-NEXT:    v_lshrrev_b32_e32 v1, 16, v0
+; GFX1250-NEXT:    v_pk_max_num_bf16 v1, v1, v1 clamp
 ; GFX1250-NEXT:    s_delay_alu instid0(VALU_DEP_1)
 ; GFX1250-NEXT:    v_perm_b32 v0, v1, v0, 0x5040100
 ; GFX1250-NEXT:    s_set_pc_i64 s[30:31]
diff --git a/llvm/test/CodeGen/AMDGPU/memintrinsic-unroll.ll b/llvm/test/CodeGen/AMDGPU/memintrinsic-unroll.ll
index 9cc42ac..be02045 100644
--- a/llvm/test/CodeGen/AMDGPU/memintrinsic-unroll.ll
+++ b/llvm/test/CodeGen/AMDGPU/memintrinsic-unroll.ll
@@ -9851,8 +9851,8 @@ define void @memmove_p5_p5_sz2048(ptr addrspace(5) align 1 %dst, ptr addrspace(5
 ; CHECK-NEXT:    s_andn2_saveexec_b32 s6, s6
 ; CHECK-NEXT:    s_cbranch_execz .LBB8_6
 ; CHECK-NEXT:  ; %bb.4: ; %memmove_bwd_loop.preheader
-; CHECK-NEXT:    v_add_nc_u32_e32 v0, 0x700, v0
 ; CHECK-NEXT:    v_add_nc_u32_e32 v1, 0x700, v1
+; CHECK-NEXT:    v_add_nc_u32_e32 v0, 0x700, v0
 ; CHECK-NEXT:    s_movk_i32 s4, 0xf800
 ; CHECK-NEXT:    s_mov_b32 s5, -1
 ; CHECK-NEXT:  .LBB8_5: ; %memmove_bwd_loop
@@ -11167,8 +11167,8 @@ define void @memmove_p5_p5_sz2048(ptr addrspace(5) align 1 %dst, ptr addrspace(5
 ; ALIGNED-NEXT:    s_andn2_saveexec_b32 s6, s6
 ; ALIGNED-NEXT:    s_cbranch_execz .LBB8_6
 ; ALIGNED-NEXT:  ; %bb.4: ; %memmove_bwd_loop.preheader
-; ALIGNED-NEXT:    v_add_nc_u32_e32 v0, 0x700, v0
 ; ALIGNED-NEXT:    v_add_nc_u32_e32 v1, 0x700, v1
+; ALIGNED-NEXT:    v_add_nc_u32_e32 v0, 0x700, v0
 ; ALIGNED-NEXT:    s_movk_i32 s4, 0xf800
 ; ALIGNED-NEXT:    s_mov_b32 s5, -1
 ; ALIGNED-NEXT:  .LBB8_5: ; %memmove_bwd_loop
@@ -12381,8 +12381,8 @@ define void @memmove_p5_p5_sz2048(ptr addrspace(5) align 1 %dst, ptr addrspace(5
 ; UNROLL3-NEXT:    buffer_load_dword v4, v1, s[0:3], 0 offen offset:2024
 ; UNROLL3-NEXT:    buffer_load_dword v5, v1, s[0:3], 0 offen offset:2020
 ; UNROLL3-NEXT:    buffer_load_dword v6, v1, s[0:3], 0 offen offset:2016
-; UNROLL3-NEXT:    v_add_nc_u32_e32 v2, 0x7b0, v0
 ; UNROLL3-NEXT:    v_add_nc_u32_e32 v1, 0x7b0, v1
+; UNROLL3-NEXT:    v_add_nc_u32_e32 v2, 0x7b0, v0
 ; UNROLL3-NEXT:    s_waitcnt vmcnt(3)
 ; UNROLL3-NEXT:    buffer_store_dword v3, v0, s[0:3], 0 offen offset:2028
 ; UNROLL3-NEXT:    s_waitcnt vmcnt(2)
diff --git a/llvm/test/CodeGen/AMDGPU/mfma-cd-select.ll b/llvm/test/CodeGen/AMDGPU/mfma-cd-select.ll
index 6763957..f7aaa3e 100644
--- a/llvm/test/CodeGen/AMDGPU/mfma-cd-select.ll
+++ b/llvm/test/CodeGen/AMDGPU/mfma-cd-select.ll
@@ -1,15 +1,148 @@
-; RUN: llc -mtriple=amdgcn -mcpu=gfx908 < %s | FileCheck --enable-var-scope --check-prefixes=GCN,GFX908 %s
-; RUN: llc -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck --enable-var-scope --check-prefixes=GCN,GFX90A %s
-; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck --enable-var-scope --check-prefixes=GCN,GFX90A %s
-; RUN: llc -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck --enable-var-scope --check-prefixes=GCN,GFX90A %s
-; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck --enable-var-scope --check-prefixes=GCN,GFX90A %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx908 < %s | FileCheck --check-prefixes=GCN,GFX908 %s
+; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck --check-prefixes=GCN,GFX90A %s
+; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck --check-prefixes=GCN,GFX90A %s
+; RUN: llc -global-isel=0 -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck --check-prefixes=GCN,GFX90A %s
+; RUN: llc -global-isel -mtriple=amdgcn -mcpu=gfx942 < %s | FileCheck --check-prefixes=GCN,GFX90A %s
 
 declare <32 x float> @llvm.amdgcn.mfma.f32.32x32x1f32(float, float, <32 x float>, i32, i32, i32)
 
-; GCN-LABEL: {{^}}test_mfma_f32_32x32x1f32_vgpr:
-; GFX908: v_mfma_f32_32x32x1{{.*}} a[{{[0-9:]+}}], v{{[0-9]+}}, v{{[0-9:]+}}, a[{{[0-9:]+}}]
-; GFX90A: v_mfma_f32_32x32x1{{.*}} v[{{[0-9:]+}}], v{{[0-9]+}}, v{{[0-9:]+}}, v[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_32x32x1f32_vgpr(ptr addrspace(1) %arg) #0 {
+; GFX908-LABEL: test_mfma_f32_32x32x1f32_vgpr:
+; GFX908:       ; %bb.0: ; %bb
+; GFX908-NEXT:    s_load_dwordx2 s[34:35], s[4:5], 0x24
+; GFX908-NEXT:    v_mov_b32_e32 v4, 0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    s_load_dwordx16 s[16:31], s[34:35], 0x0
+; GFX908-NEXT:    s_load_dwordx16 s[0:15], s[34:35], 0x40
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    v_mov_b32_e32 v0, s16
+; GFX908-NEXT:    v_mov_b32_e32 v1, s17
+; GFX908-NEXT:    v_mov_b32_e32 v2, s18
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s21
+; GFX908-NEXT:    v_mov_b32_e32 v1, s22
+; GFX908-NEXT:    v_mov_b32_e32 v2, s23
+; GFX908-NEXT:    v_accvgpr_write_b32 a5, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a6, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a7, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s24
+; GFX908-NEXT:    v_mov_b32_e32 v1, s25
+; GFX908-NEXT:    v_mov_b32_e32 v2, s26
+; GFX908-NEXT:    v_accvgpr_write_b32 a8, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a9, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a10, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s27
+; GFX908-NEXT:    v_mov_b32_e32 v1, s28
+; GFX908-NEXT:    v_mov_b32_e32 v2, s29
+; GFX908-NEXT:    v_accvgpr_write_b32 a11, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a12, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a13, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s30
+; GFX908-NEXT:    v_mov_b32_e32 v1, s31
+; GFX908-NEXT:    v_mov_b32_e32 v2, s0
+; GFX908-NEXT:    v_accvgpr_write_b32 a14, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a15, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a16, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s1
+; GFX908-NEXT:    v_mov_b32_e32 v1, s2
+; GFX908-NEXT:    v_mov_b32_e32 v2, s3
+; GFX908-NEXT:    v_accvgpr_write_b32 a17, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a18, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a19, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s4
+; GFX908-NEXT:    v_mov_b32_e32 v1, s5
+; GFX908-NEXT:    v_mov_b32_e32 v2, s6
+; GFX908-NEXT:    v_accvgpr_write_b32 a20, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a21, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a22, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s7
+; GFX908-NEXT:    v_mov_b32_e32 v1, s8
+; GFX908-NEXT:    v_mov_b32_e32 v2, s9
+; GFX908-NEXT:    v_mov_b32_e32 v3, s19
+; GFX908-NEXT:    v_accvgpr_write_b32 a23, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a24, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a25, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s10
+; GFX908-NEXT:    v_mov_b32_e32 v1, s11
+; GFX908-NEXT:    v_mov_b32_e32 v2, s12
+; GFX908-NEXT:    v_mov_b32_e32 v5, s20
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a26, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a27, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a28, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s13
+; GFX908-NEXT:    v_mov_b32_e32 v1, s14
+; GFX908-NEXT:    v_mov_b32_e32 v2, s15
+; GFX908-NEXT:    v_mov_b32_e32 v3, 1.0
+; GFX908-NEXT:    v_accvgpr_write_b32 a4, v5
+; GFX908-NEXT:    v_accvgpr_write_b32 a29, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a30, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a31, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, 2.0
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_mfma_f32_32x32x1f32 a[0:31], v3, v0, a[0:31]
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a27
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a26
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a25
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a24
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:96
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a31
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a30
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a29
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a28
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:112
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a19
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a18
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a17
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a16
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:64
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a23
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a22
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a21
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a20
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:80
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a11
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a10
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a9
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a8
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:32
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a15
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a14
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a13
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a12
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:48
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a3
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a2
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a1
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a0
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35]
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a7
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a6
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a5
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a4
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:16
+; GFX908-NEXT:    s_endpgm
 bb:
   %in.1 = load <32 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <32 x float> @llvm.amdgcn.mfma.f32.32x32x1f32(float 1.0, float 2.0, <32 x float> %in.1, i32 0, i32 0, i32 0)
@@ -17,9 +150,142 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_32x32x1f32_agpr:
-; GCN: v_mfma_f32_32x32x1{{.*}} a[{{[0-9:]+}}], v{{[0-9]+}}, v{{[0-9:]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_32x32x1f32_agpr(ptr addrspace(1) %arg) #2 {
+; GFX908-LABEL: test_mfma_f32_32x32x1f32_agpr:
+; GFX908:       ; %bb.0: ; %bb
+; GFX908-NEXT:    s_load_dwordx2 s[34:35], s[4:5], 0x24
+; GFX908-NEXT:    v_mov_b32_e32 v4, 0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    s_load_dwordx16 s[16:31], s[34:35], 0x0
+; GFX908-NEXT:    s_load_dwordx16 s[0:15], s[34:35], 0x40
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    v_mov_b32_e32 v0, s16
+; GFX908-NEXT:    v_mov_b32_e32 v1, s17
+; GFX908-NEXT:    v_mov_b32_e32 v2, s18
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s21
+; GFX908-NEXT:    v_mov_b32_e32 v1, s22
+; GFX908-NEXT:    v_mov_b32_e32 v2, s23
+; GFX908-NEXT:    v_accvgpr_write_b32 a5, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a6, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a7, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s24
+; GFX908-NEXT:    v_mov_b32_e32 v1, s25
+; GFX908-NEXT:    v_mov_b32_e32 v2, s26
+; GFX908-NEXT:    v_accvgpr_write_b32 a8, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a9, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a10, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s27
+; GFX908-NEXT:    v_mov_b32_e32 v1, s28
+; GFX908-NEXT:    v_mov_b32_e32 v2, s29
+; GFX908-NEXT:    v_accvgpr_write_b32 a11, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a12, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a13, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s30
+; GFX908-NEXT:    v_mov_b32_e32 v1, s31
+; GFX908-NEXT:    v_mov_b32_e32 v2, s0
+; GFX908-NEXT:    v_accvgpr_write_b32 a14, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a15, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a16, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s1
+; GFX908-NEXT:    v_mov_b32_e32 v1, s2
+; GFX908-NEXT:    v_mov_b32_e32 v2, s3
+; GFX908-NEXT:    v_accvgpr_write_b32 a17, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a18, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a19, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s4
+; GFX908-NEXT:    v_mov_b32_e32 v1, s5
+; GFX908-NEXT:    v_mov_b32_e32 v2, s6
+; GFX908-NEXT:    v_accvgpr_write_b32 a20, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a21, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a22, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s7
+; GFX908-NEXT:    v_mov_b32_e32 v1, s8
+; GFX908-NEXT:    v_mov_b32_e32 v2, s9
+; GFX908-NEXT:    v_mov_b32_e32 v3, s19
+; GFX908-NEXT:    v_accvgpr_write_b32 a23, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a24, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a25, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s10
+; GFX908-NEXT:    v_mov_b32_e32 v1, s11
+; GFX908-NEXT:    v_mov_b32_e32 v2, s12
+; GFX908-NEXT:    v_mov_b32_e32 v5, s20
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a26, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a27, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a28, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, s13
+; GFX908-NEXT:    v_mov_b32_e32 v1, s14
+; GFX908-NEXT:    v_mov_b32_e32 v2, s15
+; GFX908-NEXT:    v_mov_b32_e32 v3, 1.0
+; GFX908-NEXT:    v_accvgpr_write_b32 a4, v5
+; GFX908-NEXT:    v_accvgpr_write_b32 a29, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a30, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a31, v2
+; GFX908-NEXT:    v_mov_b32_e32 v0, 2.0
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_mfma_f32_32x32x1f32 a[0:31], v3, v0, a[0:31]
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a27
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a26
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a25
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a24
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:96
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a31
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a30
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a29
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a28
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:112
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a19
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a18
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a17
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a16
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:64
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a23
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a22
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a21
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a20
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:80
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a11
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a10
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a9
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a8
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:32
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a15
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a14
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a13
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a12
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:48
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a3
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a2
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a1
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a0
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35]
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a7
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a6
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a5
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a4
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[34:35] offset:16
+; GFX908-NEXT:    s_endpgm
 bb:
   %in.1 = load <32 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <32 x float> @llvm.amdgcn.mfma.f32.32x32x1f32(float 1.0, float 2.0, <32 x float> %in.1, i32 0, i32 0, i32 0)
@@ -27,9 +293,105 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_32x32x1f32_inline_asm_virtual_agpr:
-; GCN: v_mfma_f32_32x32x1{{.*}} a[{{[0-9:]+}}], v{{[0-9]+}}, v{{[0-9:]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_32x32x1f32_inline_asm_virtual_agpr(ptr addrspace(1) %arg) {
+; GFX908-LABEL: test_mfma_f32_32x32x1f32_inline_asm_virtual_agpr:
+; GFX908:       ; %bb.0: ; %bb
+; GFX908-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX908-NEXT:    v_mov_b32_e32 v32, 0
+; GFX908-NEXT:    ;;#ASMSTART
+; GFX908-NEXT:    ; def a0
+; GFX908-NEXT:    ;;#ASMEND
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    global_load_dwordx4 v[28:31], v32, s[0:1] offset:112
+; GFX908-NEXT:    global_load_dwordx4 v[24:27], v32, s[0:1] offset:96
+; GFX908-NEXT:    global_load_dwordx4 v[20:23], v32, s[0:1] offset:80
+; GFX908-NEXT:    global_load_dwordx4 v[16:19], v32, s[0:1] offset:64
+; GFX908-NEXT:    global_load_dwordx4 v[12:15], v32, s[0:1] offset:48
+; GFX908-NEXT:    global_load_dwordx4 v[8:11], v32, s[0:1] offset:32
+; GFX908-NEXT:    global_load_dwordx4 v[4:7], v32, s[0:1] offset:16
+; GFX908-NEXT:    global_load_dwordx4 v[0:3], v32, s[0:1]
+; GFX908-NEXT:    s_waitcnt vmcnt(0)
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a4, v4
+; GFX908-NEXT:    v_accvgpr_write_b32 a5, v5
+; GFX908-NEXT:    v_accvgpr_write_b32 a6, v6
+; GFX908-NEXT:    v_accvgpr_write_b32 a7, v7
+; GFX908-NEXT:    v_accvgpr_write_b32 a8, v8
+; GFX908-NEXT:    v_accvgpr_write_b32 a9, v9
+; GFX908-NEXT:    v_accvgpr_write_b32 a10, v10
+; GFX908-NEXT:    v_accvgpr_write_b32 a11, v11
+; GFX908-NEXT:    v_accvgpr_write_b32 a12, v12
+; GFX908-NEXT:    v_accvgpr_write_b32 a13, v13
+; GFX908-NEXT:    v_accvgpr_write_b32 a14, v14
+; GFX908-NEXT:    v_accvgpr_write_b32 a15, v15
+; GFX908-NEXT:    v_accvgpr_write_b32 a16, v16
+; GFX908-NEXT:    v_accvgpr_write_b32 a17, v17
+; GFX908-NEXT:    v_accvgpr_write_b32 a18, v18
+; GFX908-NEXT:    v_accvgpr_write_b32 a19, v19
+; GFX908-NEXT:    v_accvgpr_write_b32 a20, v20
+; GFX908-NEXT:    v_accvgpr_write_b32 a21, v21
+; GFX908-NEXT:    v_accvgpr_write_b32 a22, v22
+; GFX908-NEXT:    v_accvgpr_write_b32 a23, v23
+; GFX908-NEXT:    v_accvgpr_write_b32 a24, v24
+; GFX908-NEXT:    v_accvgpr_write_b32 a25, v25
+; GFX908-NEXT:    v_accvgpr_write_b32 a26, v26
+; GFX908-NEXT:    v_accvgpr_write_b32 a27, v27
+; GFX908-NEXT:    v_accvgpr_write_b32 a28, v28
+; GFX908-NEXT:    v_accvgpr_write_b32 a29, v29
+; GFX908-NEXT:    v_accvgpr_write_b32 a30, v30
+; GFX908-NEXT:    v_accvgpr_write_b32 a31, v31
+; GFX908-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX908-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[0:31]
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a27
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a26
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a25
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a24
+; GFX908-NEXT:    v_accvgpr_read_b32 v7, a31
+; GFX908-NEXT:    v_accvgpr_read_b32 v6, a30
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a29
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a28
+; GFX908-NEXT:    v_accvgpr_read_b32 v11, a19
+; GFX908-NEXT:    v_accvgpr_read_b32 v10, a18
+; GFX908-NEXT:    v_accvgpr_read_b32 v9, a17
+; GFX908-NEXT:    v_accvgpr_read_b32 v8, a16
+; GFX908-NEXT:    v_accvgpr_read_b32 v15, a23
+; GFX908-NEXT:    v_accvgpr_read_b32 v14, a22
+; GFX908-NEXT:    v_accvgpr_read_b32 v13, a21
+; GFX908-NEXT:    v_accvgpr_read_b32 v12, a20
+; GFX908-NEXT:    v_accvgpr_read_b32 v19, a11
+; GFX908-NEXT:    v_accvgpr_read_b32 v18, a10
+; GFX908-NEXT:    v_accvgpr_read_b32 v17, a9
+; GFX908-NEXT:    v_accvgpr_read_b32 v16, a8
+; GFX908-NEXT:    v_accvgpr_read_b32 v23, a15
+; GFX908-NEXT:    v_accvgpr_read_b32 v22, a14
+; GFX908-NEXT:    v_accvgpr_read_b32 v21, a13
+; GFX908-NEXT:    v_accvgpr_read_b32 v20, a12
+; GFX908-NEXT:    v_accvgpr_read_b32 v27, a3
+; GFX908-NEXT:    v_accvgpr_read_b32 v26, a2
+; GFX908-NEXT:    v_accvgpr_read_b32 v25, a1
+; GFX908-NEXT:    v_accvgpr_read_b32 v24, a0
+; GFX908-NEXT:    global_store_dwordx4 v32, v[0:3], s[0:1] offset:96
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a7
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a6
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a5
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a4
+; GFX908-NEXT:    global_store_dwordx4 v32, v[4:7], s[0:1] offset:112
+; GFX908-NEXT:    global_store_dwordx4 v32, v[8:11], s[0:1] offset:64
+; GFX908-NEXT:    global_store_dwordx4 v32, v[12:15], s[0:1] offset:80
+; GFX908-NEXT:    global_store_dwordx4 v32, v[16:19], s[0:1] offset:32
+; GFX908-NEXT:    global_store_dwordx4 v32, v[20:23], s[0:1] offset:48
+; GFX908-NEXT:    global_store_dwordx4 v32, v[24:27], s[0:1]
+; GFX908-NEXT:    global_store_dwordx4 v32, v[0:3], s[0:1] offset:16
+; GFX908-NEXT:    s_endpgm
 bb:
   %acc = call i32 asm sideeffect "; def $0", "={a0}"()
   %in.1 = load <32 x float>, ptr addrspace(1) %arg
@@ -38,9 +400,105 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_32x32x1f32_inline_asm_phys_agpr:
-; GCN: v_mfma_f32_32x32x1{{.*}} a[{{[0-9:]+}}], v{{[0-9]+}}, v{{[0-9:]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_32x32x1f32_inline_asm_phys_agpr(ptr addrspace(1) %arg) {
+; GFX908-LABEL: test_mfma_f32_32x32x1f32_inline_asm_phys_agpr:
+; GFX908:       ; %bb.0: ; %bb
+; GFX908-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX908-NEXT:    v_mov_b32_e32 v32, 0
+; GFX908-NEXT:    ;;#ASMSTART
+; GFX908-NEXT:    ; use a[100:131]
+; GFX908-NEXT:    ;;#ASMEND
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    global_load_dwordx4 v[28:31], v32, s[0:1] offset:112
+; GFX908-NEXT:    global_load_dwordx4 v[24:27], v32, s[0:1] offset:96
+; GFX908-NEXT:    global_load_dwordx4 v[20:23], v32, s[0:1] offset:80
+; GFX908-NEXT:    global_load_dwordx4 v[16:19], v32, s[0:1] offset:64
+; GFX908-NEXT:    global_load_dwordx4 v[12:15], v32, s[0:1] offset:48
+; GFX908-NEXT:    global_load_dwordx4 v[8:11], v32, s[0:1] offset:32
+; GFX908-NEXT:    global_load_dwordx4 v[4:7], v32, s[0:1] offset:16
+; GFX908-NEXT:    global_load_dwordx4 v[0:3], v32, s[0:1]
+; GFX908-NEXT:    s_waitcnt vmcnt(0)
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a4, v4
+; GFX908-NEXT:    v_accvgpr_write_b32 a5, v5
+; GFX908-NEXT:    v_accvgpr_write_b32 a6, v6
+; GFX908-NEXT:    v_accvgpr_write_b32 a7, v7
+; GFX908-NEXT:    v_accvgpr_write_b32 a8, v8
+; GFX908-NEXT:    v_accvgpr_write_b32 a9, v9
+; GFX908-NEXT:    v_accvgpr_write_b32 a10, v10
+; GFX908-NEXT:    v_accvgpr_write_b32 a11, v11
+; GFX908-NEXT:    v_accvgpr_write_b32 a12, v12
+; GFX908-NEXT:    v_accvgpr_write_b32 a13, v13
+; GFX908-NEXT:    v_accvgpr_write_b32 a14, v14
+; GFX908-NEXT:    v_accvgpr_write_b32 a15, v15
+; GFX908-NEXT:    v_accvgpr_write_b32 a16, v16
+; GFX908-NEXT:    v_accvgpr_write_b32 a17, v17
+; GFX908-NEXT:    v_accvgpr_write_b32 a18, v18
+; GFX908-NEXT:    v_accvgpr_write_b32 a19, v19
+; GFX908-NEXT:    v_accvgpr_write_b32 a20, v20
+; GFX908-NEXT:    v_accvgpr_write_b32 a21, v21
+; GFX908-NEXT:    v_accvgpr_write_b32 a22, v22
+; GFX908-NEXT:    v_accvgpr_write_b32 a23, v23
+; GFX908-NEXT:    v_accvgpr_write_b32 a24, v24
+; GFX908-NEXT:    v_accvgpr_write_b32 a25, v25
+; GFX908-NEXT:    v_accvgpr_write_b32 a26, v26
+; GFX908-NEXT:    v_accvgpr_write_b32 a27, v27
+; GFX908-NEXT:    v_accvgpr_write_b32 a28, v28
+; GFX908-NEXT:    v_accvgpr_write_b32 a29, v29
+; GFX908-NEXT:    v_accvgpr_write_b32 a30, v30
+; GFX908-NEXT:    v_accvgpr_write_b32 a31, v31
+; GFX908-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX908-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[0:31]
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a27
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a26
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a25
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a24
+; GFX908-NEXT:    v_accvgpr_read_b32 v7, a31
+; GFX908-NEXT:    v_accvgpr_read_b32 v6, a30
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a29
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a28
+; GFX908-NEXT:    v_accvgpr_read_b32 v11, a19
+; GFX908-NEXT:    v_accvgpr_read_b32 v10, a18
+; GFX908-NEXT:    v_accvgpr_read_b32 v9, a17
+; GFX908-NEXT:    v_accvgpr_read_b32 v8, a16
+; GFX908-NEXT:    v_accvgpr_read_b32 v15, a23
+; GFX908-NEXT:    v_accvgpr_read_b32 v14, a22
+; GFX908-NEXT:    v_accvgpr_read_b32 v13, a21
+; GFX908-NEXT:    v_accvgpr_read_b32 v12, a20
+; GFX908-NEXT:    v_accvgpr_read_b32 v19, a11
+; GFX908-NEXT:    v_accvgpr_read_b32 v18, a10
+; GFX908-NEXT:    v_accvgpr_read_b32 v17, a9
+; GFX908-NEXT:    v_accvgpr_read_b32 v16, a8
+; GFX908-NEXT:    v_accvgpr_read_b32 v23, a15
+; GFX908-NEXT:    v_accvgpr_read_b32 v22, a14
+; GFX908-NEXT:    v_accvgpr_read_b32 v21, a13
+; GFX908-NEXT:    v_accvgpr_read_b32 v20, a12
+; GFX908-NEXT:    v_accvgpr_read_b32 v27, a3
+; GFX908-NEXT:    v_accvgpr_read_b32 v26, a2
+; GFX908-NEXT:    v_accvgpr_read_b32 v25, a1
+; GFX908-NEXT:    v_accvgpr_read_b32 v24, a0
+; GFX908-NEXT:    global_store_dwordx4 v32, v[0:3], s[0:1] offset:96
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a7
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a6
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a5
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a4
+; GFX908-NEXT:    global_store_dwordx4 v32, v[4:7], s[0:1] offset:112
+; GFX908-NEXT:    global_store_dwordx4 v32, v[8:11], s[0:1] offset:64
+; GFX908-NEXT:    global_store_dwordx4 v32, v[12:15], s[0:1] offset:80
+; GFX908-NEXT:    global_store_dwordx4 v32, v[16:19], s[0:1] offset:32
+; GFX908-NEXT:    global_store_dwordx4 v32, v[20:23], s[0:1] offset:48
+; GFX908-NEXT:    global_store_dwordx4 v32, v[24:27], s[0:1]
+; GFX908-NEXT:    global_store_dwordx4 v32, v[0:3], s[0:1] offset:16
+; GFX908-NEXT:    s_endpgm
 bb:
   call void asm sideeffect "; use $0", "{a[100:131]}"(<32 x float> poison)
   %in.1 = load <32 x float>, ptr addrspace(1) %arg
@@ -49,10 +507,105 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_32x32x1f32_inline_asm_no_agprs:
-; GFX908: v_mfma_f32_32x32x1{{.*}} a[{{[0-9:]+}}], v{{[0-9]+}}, v{{[0-9:]+}}, a[{{[0-9:]+}}]
-; GFX90A: v_mfma_f32_32x32x1{{.*}} v[{{[0-9:]+}}], v{{[0-9]+}}, v{{[0-9:]+}}, v[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_32x32x1f32_inline_asm_no_agprs(ptr addrspace(1) %arg) #0 {
+; GFX908-LABEL: test_mfma_f32_32x32x1f32_inline_asm_no_agprs:
+; GFX908:       ; %bb.0: ; %bb
+; GFX908-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX908-NEXT:    v_mov_b32_e32 v32, 0
+; GFX908-NEXT:    ;;#ASMSTART
+; GFX908-NEXT:    ; def v0
+; GFX908-NEXT:    ;;#ASMEND
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    global_load_dwordx4 v[28:31], v32, s[0:1] offset:112
+; GFX908-NEXT:    global_load_dwordx4 v[24:27], v32, s[0:1] offset:96
+; GFX908-NEXT:    global_load_dwordx4 v[20:23], v32, s[0:1] offset:80
+; GFX908-NEXT:    global_load_dwordx4 v[16:19], v32, s[0:1] offset:64
+; GFX908-NEXT:    global_load_dwordx4 v[12:15], v32, s[0:1] offset:48
+; GFX908-NEXT:    global_load_dwordx4 v[8:11], v32, s[0:1] offset:32
+; GFX908-NEXT:    global_load_dwordx4 v[4:7], v32, s[0:1] offset:16
+; GFX908-NEXT:    global_load_dwordx4 v[0:3], v32, s[0:1]
+; GFX908-NEXT:    s_waitcnt vmcnt(0)
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a4, v4
+; GFX908-NEXT:    v_accvgpr_write_b32 a5, v5
+; GFX908-NEXT:    v_accvgpr_write_b32 a6, v6
+; GFX908-NEXT:    v_accvgpr_write_b32 a7, v7
+; GFX908-NEXT:    v_accvgpr_write_b32 a8, v8
+; GFX908-NEXT:    v_accvgpr_write_b32 a9, v9
+; GFX908-NEXT:    v_accvgpr_write_b32 a10, v10
+; GFX908-NEXT:    v_accvgpr_write_b32 a11, v11
+; GFX908-NEXT:    v_accvgpr_write_b32 a12, v12
+; GFX908-NEXT:    v_accvgpr_write_b32 a13, v13
+; GFX908-NEXT:    v_accvgpr_write_b32 a14, v14
+; GFX908-NEXT:    v_accvgpr_write_b32 a15, v15
+; GFX908-NEXT:    v_accvgpr_write_b32 a16, v16
+; GFX908-NEXT:    v_accvgpr_write_b32 a17, v17
+; GFX908-NEXT:    v_accvgpr_write_b32 a18, v18
+; GFX908-NEXT:    v_accvgpr_write_b32 a19, v19
+; GFX908-NEXT:    v_accvgpr_write_b32 a20, v20
+; GFX908-NEXT:    v_accvgpr_write_b32 a21, v21
+; GFX908-NEXT:    v_accvgpr_write_b32 a22, v22
+; GFX908-NEXT:    v_accvgpr_write_b32 a23, v23
+; GFX908-NEXT:    v_accvgpr_write_b32 a24, v24
+; GFX908-NEXT:    v_accvgpr_write_b32 a25, v25
+; GFX908-NEXT:    v_accvgpr_write_b32 a26, v26
+; GFX908-NEXT:    v_accvgpr_write_b32 a27, v27
+; GFX908-NEXT:    v_accvgpr_write_b32 a28, v28
+; GFX908-NEXT:    v_accvgpr_write_b32 a29, v29
+; GFX908-NEXT:    v_accvgpr_write_b32 a30, v30
+; GFX908-NEXT:    v_accvgpr_write_b32 a31, v31
+; GFX908-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX908-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[0:31]
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a27
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a26
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a25
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a24
+; GFX908-NEXT:    v_accvgpr_read_b32 v7, a31
+; GFX908-NEXT:    v_accvgpr_read_b32 v6, a30
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a29
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a28
+; GFX908-NEXT:    v_accvgpr_read_b32 v11, a19
+; GFX908-NEXT:    v_accvgpr_read_b32 v10, a18
+; GFX908-NEXT:    v_accvgpr_read_b32 v9, a17
+; GFX908-NEXT:    v_accvgpr_read_b32 v8, a16
+; GFX908-NEXT:    v_accvgpr_read_b32 v15, a23
+; GFX908-NEXT:    v_accvgpr_read_b32 v14, a22
+; GFX908-NEXT:    v_accvgpr_read_b32 v13, a21
+; GFX908-NEXT:    v_accvgpr_read_b32 v12, a20
+; GFX908-NEXT:    v_accvgpr_read_b32 v19, a11
+; GFX908-NEXT:    v_accvgpr_read_b32 v18, a10
+; GFX908-NEXT:    v_accvgpr_read_b32 v17, a9
+; GFX908-NEXT:    v_accvgpr_read_b32 v16, a8
+; GFX908-NEXT:    v_accvgpr_read_b32 v23, a15
+; GFX908-NEXT:    v_accvgpr_read_b32 v22, a14
+; GFX908-NEXT:    v_accvgpr_read_b32 v21, a13
+; GFX908-NEXT:    v_accvgpr_read_b32 v20, a12
+; GFX908-NEXT:    v_accvgpr_read_b32 v27, a3
+; GFX908-NEXT:    v_accvgpr_read_b32 v26, a2
+; GFX908-NEXT:    v_accvgpr_read_b32 v25, a1
+; GFX908-NEXT:    v_accvgpr_read_b32 v24, a0
+; GFX908-NEXT:    global_store_dwordx4 v32, v[0:3], s[0:1] offset:96
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a7
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a6
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a5
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a4
+; GFX908-NEXT:    global_store_dwordx4 v32, v[4:7], s[0:1] offset:112
+; GFX908-NEXT:    global_store_dwordx4 v32, v[8:11], s[0:1] offset:64
+; GFX908-NEXT:    global_store_dwordx4 v32, v[12:15], s[0:1] offset:80
+; GFX908-NEXT:    global_store_dwordx4 v32, v[16:19], s[0:1] offset:32
+; GFX908-NEXT:    global_store_dwordx4 v32, v[20:23], s[0:1] offset:48
+; GFX908-NEXT:    global_store_dwordx4 v32, v[24:27], s[0:1]
+; GFX908-NEXT:    global_store_dwordx4 v32, v[0:3], s[0:1] offset:16
+; GFX908-NEXT:    s_endpgm
 bb:
   %acc = call i32 asm sideeffect "; def $0", "={v0}"()
   %in.1 = load <32 x float>, ptr addrspace(1) %arg
@@ -61,9 +614,127 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_32x32x1f32_call:
-; GCN: v_mfma_f32_32x32x1{{.*}} a[{{[0-9:]+}}], v{{[0-9]+}}, v{{[0-9:]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_32x32x1f32_call(ptr addrspace(1) %arg) #1 {
+; GFX908-LABEL: test_mfma_f32_32x32x1f32_call:
+; GFX908:       ; %bb.0: ; %bb
+; GFX908-NEXT:    s_mov_b32 s36, SCRATCH_RSRC_DWORD0
+; GFX908-NEXT:    s_mov_b32 s37, SCRATCH_RSRC_DWORD1
+; GFX908-NEXT:    s_mov_b32 s38, -1
+; GFX908-NEXT:    s_mov_b32 s39, 0xe00000
+; GFX908-NEXT:    s_add_u32 s36, s36, s11
+; GFX908-NEXT:    s_addc_u32 s37, s37, 0
+; GFX908-NEXT:    s_mov_b32 s12, s8
+; GFX908-NEXT:    s_add_u32 s8, s4, 44
+; GFX908-NEXT:    s_mov_b32 s13, s9
+; GFX908-NEXT:    s_addc_u32 s9, s5, 0
+; GFX908-NEXT:    s_load_dwordx2 s[34:35], s[4:5], 0x24
+; GFX908-NEXT:    s_getpc_b64 s[4:5]
+; GFX908-NEXT:    s_add_u32 s4, s4, foo@gotpcrel32@lo+4
+; GFX908-NEXT:    s_addc_u32 s5, s5, foo@gotpcrel32@hi+12
+; GFX908-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x0
+; GFX908-NEXT:    s_mov_b32 s14, s10
+; GFX908-NEXT:    s_mov_b64 s[10:11], s[6:7]
+; GFX908-NEXT:    v_lshlrev_b32_e32 v2, 20, v2
+; GFX908-NEXT:    v_lshlrev_b32_e32 v1, 10, v1
+; GFX908-NEXT:    s_mov_b64 s[4:5], s[0:1]
+; GFX908-NEXT:    s_mov_b64 s[6:7], s[2:3]
+; GFX908-NEXT:    s_mov_b64 s[0:1], s[36:37]
+; GFX908-NEXT:    v_or3_b32 v31, v0, v1, v2
+; GFX908-NEXT:    s_mov_b64 s[2:3], s[38:39]
+; GFX908-NEXT:    s_mov_b32 s32, 0
+; GFX908-NEXT:    v_mov_b32_e32 v40, 0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    s_swappc_b64 s[30:31], s[16:17]
+; GFX908-NEXT:    global_load_dwordx4 v[28:31], v40, s[34:35] offset:112
+; GFX908-NEXT:    global_load_dwordx4 v[24:27], v40, s[34:35] offset:96
+; GFX908-NEXT:    global_load_dwordx4 v[20:23], v40, s[34:35] offset:80
+; GFX908-NEXT:    global_load_dwordx4 v[16:19], v40, s[34:35] offset:64
+; GFX908-NEXT:    global_load_dwordx4 v[12:15], v40, s[34:35] offset:48
+; GFX908-NEXT:    global_load_dwordx4 v[8:11], v40, s[34:35] offset:32
+; GFX908-NEXT:    global_load_dwordx4 v[4:7], v40, s[34:35] offset:16
+; GFX908-NEXT:    global_load_dwordx4 v[0:3], v40, s[34:35]
+; GFX908-NEXT:    s_waitcnt vmcnt(0)
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a4, v4
+; GFX908-NEXT:    v_accvgpr_write_b32 a5, v5
+; GFX908-NEXT:    v_accvgpr_write_b32 a6, v6
+; GFX908-NEXT:    v_accvgpr_write_b32 a7, v7
+; GFX908-NEXT:    v_accvgpr_write_b32 a8, v8
+; GFX908-NEXT:    v_accvgpr_write_b32 a9, v9
+; GFX908-NEXT:    v_accvgpr_write_b32 a10, v10
+; GFX908-NEXT:    v_accvgpr_write_b32 a11, v11
+; GFX908-NEXT:    v_accvgpr_write_b32 a12, v12
+; GFX908-NEXT:    v_accvgpr_write_b32 a13, v13
+; GFX908-NEXT:    v_accvgpr_write_b32 a14, v14
+; GFX908-NEXT:    v_accvgpr_write_b32 a15, v15
+; GFX908-NEXT:    v_accvgpr_write_b32 a16, v16
+; GFX908-NEXT:    v_accvgpr_write_b32 a17, v17
+; GFX908-NEXT:    v_accvgpr_write_b32 a18, v18
+; GFX908-NEXT:    v_accvgpr_write_b32 a19, v19
+; GFX908-NEXT:    v_accvgpr_write_b32 a20, v20
+; GFX908-NEXT:    v_accvgpr_write_b32 a21, v21
+; GFX908-NEXT:    v_accvgpr_write_b32 a22, v22
+; GFX908-NEXT:    v_accvgpr_write_b32 a23, v23
+; GFX908-NEXT:    v_accvgpr_write_b32 a24, v24
+; GFX908-NEXT:    v_accvgpr_write_b32 a25, v25
+; GFX908-NEXT:    v_accvgpr_write_b32 a26, v26
+; GFX908-NEXT:    v_accvgpr_write_b32 a27, v27
+; GFX908-NEXT:    v_accvgpr_write_b32 a28, v28
+; GFX908-NEXT:    v_accvgpr_write_b32 a29, v29
+; GFX908-NEXT:    v_accvgpr_write_b32 a30, v30
+; GFX908-NEXT:    v_accvgpr_write_b32 a31, v31
+; GFX908-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX908-NEXT:    v_mov_b32_e32 v1, 2.0
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[0:31]
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a27
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a26
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a25
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a24
+; GFX908-NEXT:    v_accvgpr_read_b32 v7, a31
+; GFX908-NEXT:    v_accvgpr_read_b32 v6, a30
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a29
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a28
+; GFX908-NEXT:    v_accvgpr_read_b32 v11, a19
+; GFX908-NEXT:    v_accvgpr_read_b32 v10, a18
+; GFX908-NEXT:    v_accvgpr_read_b32 v9, a17
+; GFX908-NEXT:    v_accvgpr_read_b32 v8, a16
+; GFX908-NEXT:    v_accvgpr_read_b32 v15, a23
+; GFX908-NEXT:    v_accvgpr_read_b32 v14, a22
+; GFX908-NEXT:    v_accvgpr_read_b32 v13, a21
+; GFX908-NEXT:    v_accvgpr_read_b32 v12, a20
+; GFX908-NEXT:    v_accvgpr_read_b32 v19, a11
+; GFX908-NEXT:    v_accvgpr_read_b32 v18, a10
+; GFX908-NEXT:    v_accvgpr_read_b32 v17, a9
+; GFX908-NEXT:    v_accvgpr_read_b32 v16, a8
+; GFX908-NEXT:    v_accvgpr_read_b32 v23, a15
+; GFX908-NEXT:    v_accvgpr_read_b32 v22, a14
+; GFX908-NEXT:    v_accvgpr_read_b32 v21, a13
+; GFX908-NEXT:    v_accvgpr_read_b32 v20, a12
+; GFX908-NEXT:    v_accvgpr_read_b32 v27, a3
+; GFX908-NEXT:    v_accvgpr_read_b32 v26, a2
+; GFX908-NEXT:    v_accvgpr_read_b32 v25, a1
+; GFX908-NEXT:    v_accvgpr_read_b32 v24, a0
+; GFX908-NEXT:    global_store_dwordx4 v40, v[0:3], s[34:35] offset:96
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a7
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a6
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a5
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a4
+; GFX908-NEXT:    global_store_dwordx4 v40, v[4:7], s[34:35] offset:112
+; GFX908-NEXT:    global_store_dwordx4 v40, v[8:11], s[34:35] offset:64
+; GFX908-NEXT:    global_store_dwordx4 v40, v[12:15], s[34:35] offset:80
+; GFX908-NEXT:    global_store_dwordx4 v40, v[16:19], s[34:35] offset:32
+; GFX908-NEXT:    global_store_dwordx4 v40, v[20:23], s[34:35] offset:48
+; GFX908-NEXT:    global_store_dwordx4 v40, v[24:27], s[34:35]
+; GFX908-NEXT:    global_store_dwordx4 v40, v[0:3], s[34:35] offset:16
+; GFX908-NEXT:    s_endpgm
 bb:
   call void @foo()
   %in.1 = load <32 x float>, ptr addrspace(1) %arg
@@ -75,10 +746,173 @@ bb:
 ; We could avoid scan to find calls since we see these during lowering before selection.
 ; However, in SDag lowering and selection is done block by block, so it would only work
 ; in Global ISel.
-
-; GCN-LABEL: {{^}}test_mfma_f32_32x32x1f32_call_multi_bb:
-; GCN: v_mfma_f32_32x32x1{{.*}} a[{{[0-9:]+}}], v{{[0-9]+}}, v{{[0-9:]+}}, a[{{[0-9:]+}}]
 define amdgpu_kernel void @test_mfma_f32_32x32x1f32_call_multi_bb(ptr addrspace(1) %arg, i1 %c0) #1 {
+; GFX908-LABEL: test_mfma_f32_32x32x1f32_call_multi_bb:
+; GFX908:       ; %bb.0: ; %bb1
+; GFX908-NEXT:    s_mov_b32 s52, SCRATCH_RSRC_DWORD0
+; GFX908-NEXT:    s_mov_b32 s53, SCRATCH_RSRC_DWORD1
+; GFX908-NEXT:    s_mov_b32 s54, -1
+; GFX908-NEXT:    s_mov_b32 s55, 0xe00000
+; GFX908-NEXT:    s_add_u32 s52, s52, s11
+; GFX908-NEXT:    s_mov_b32 s14, s10
+; GFX908-NEXT:    s_mov_b32 s12, s8
+; GFX908-NEXT:    s_mov_b64 s[10:11], s[6:7]
+; GFX908-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x24
+; GFX908-NEXT:    s_load_dword s8, s[4:5], 0x2c
+; GFX908-NEXT:    v_mov_b32_e32 v6, 1.0
+; GFX908-NEXT:    v_mov_b32_e32 v7, 0
+; GFX908-NEXT:    s_addc_u32 s53, s53, 0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    s_load_dwordx16 s[36:51], s[6:7], 0x0
+; GFX908-NEXT:    s_load_dwordx16 s[16:31], s[6:7], 0x40
+; GFX908-NEXT:    s_bitcmp0_b32 s8, 0
+; GFX908-NEXT:    s_mov_b32 s32, 0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    v_mov_b32_e32 v3, s36
+; GFX908-NEXT:    v_mov_b32_e32 v4, s37
+; GFX908-NEXT:    v_mov_b32_e32 v5, s40
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v4
+; GFX908-NEXT:    v_mov_b32_e32 v3, s38
+; GFX908-NEXT:    v_mov_b32_e32 v4, s39
+; GFX908-NEXT:    v_accvgpr_write_b32 a4, v5
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, v4
+; GFX908-NEXT:    v_mov_b32_e32 v3, s41
+; GFX908-NEXT:    v_mov_b32_e32 v4, s42
+; GFX908-NEXT:    v_mov_b32_e32 v5, s43
+; GFX908-NEXT:    v_accvgpr_write_b32 a5, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a6, v4
+; GFX908-NEXT:    v_accvgpr_write_b32 a7, v5
+; GFX908-NEXT:    v_mov_b32_e32 v3, s44
+; GFX908-NEXT:    v_mov_b32_e32 v4, s45
+; GFX908-NEXT:    v_mov_b32_e32 v5, s46
+; GFX908-NEXT:    v_accvgpr_write_b32 a8, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a9, v4
+; GFX908-NEXT:    v_accvgpr_write_b32 a10, v5
+; GFX908-NEXT:    v_mov_b32_e32 v3, s47
+; GFX908-NEXT:    v_mov_b32_e32 v4, s48
+; GFX908-NEXT:    v_mov_b32_e32 v5, s49
+; GFX908-NEXT:    v_accvgpr_write_b32 a11, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a12, v4
+; GFX908-NEXT:    v_accvgpr_write_b32 a13, v5
+; GFX908-NEXT:    v_mov_b32_e32 v3, s50
+; GFX908-NEXT:    v_mov_b32_e32 v4, s51
+; GFX908-NEXT:    v_mov_b32_e32 v5, s16
+; GFX908-NEXT:    v_accvgpr_write_b32 a14, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a15, v4
+; GFX908-NEXT:    v_accvgpr_write_b32 a16, v5
+; GFX908-NEXT:    v_mov_b32_e32 v3, s17
+; GFX908-NEXT:    v_mov_b32_e32 v4, s18
+; GFX908-NEXT:    v_mov_b32_e32 v5, s19
+; GFX908-NEXT:    v_accvgpr_write_b32 a17, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a18, v4
+; GFX908-NEXT:    v_accvgpr_write_b32 a19, v5
+; GFX908-NEXT:    v_mov_b32_e32 v3, s20
+; GFX908-NEXT:    v_mov_b32_e32 v4, s21
+; GFX908-NEXT:    v_mov_b32_e32 v5, s22
+; GFX908-NEXT:    v_accvgpr_write_b32 a20, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a21, v4
+; GFX908-NEXT:    v_accvgpr_write_b32 a22, v5
+; GFX908-NEXT:    v_mov_b32_e32 v3, s23
+; GFX908-NEXT:    v_mov_b32_e32 v4, s24
+; GFX908-NEXT:    v_mov_b32_e32 v5, s25
+; GFX908-NEXT:    v_accvgpr_write_b32 a23, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a24, v4
+; GFX908-NEXT:    v_accvgpr_write_b32 a25, v5
+; GFX908-NEXT:    v_mov_b32_e32 v3, s26
+; GFX908-NEXT:    v_mov_b32_e32 v4, s27
+; GFX908-NEXT:    v_mov_b32_e32 v5, s28
+; GFX908-NEXT:    v_accvgpr_write_b32 a26, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a27, v4
+; GFX908-NEXT:    v_accvgpr_write_b32 a28, v5
+; GFX908-NEXT:    v_mov_b32_e32 v3, s29
+; GFX908-NEXT:    v_mov_b32_e32 v4, s30
+; GFX908-NEXT:    v_mov_b32_e32 v5, s31
+; GFX908-NEXT:    v_accvgpr_write_b32 a29, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a30, v4
+; GFX908-NEXT:    v_accvgpr_write_b32 a31, v5
+; GFX908-NEXT:    v_mov_b32_e32 v3, 2.0
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_mfma_f32_32x32x1f32 a[0:31], v6, v3, a[0:31] cbsz:1 abid:2 blgp:3
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_accvgpr_read_b32 v6, a27
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a26
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a25
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a24
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v7, v[3:6], s[6:7] offset:96
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v6, a31
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a30
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a29
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a28
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v7, v[3:6], s[6:7] offset:112
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v6, a19
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a18
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a17
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a16
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v7, v[3:6], s[6:7] offset:64
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v6, a23
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a22
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a21
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a20
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v7, v[3:6], s[6:7] offset:80
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v6, a11
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a10
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a9
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a8
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v7, v[3:6], s[6:7] offset:32
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v6, a15
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a14
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a13
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a12
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v7, v[3:6], s[6:7] offset:48
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v6, a3
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a2
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a1
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a0
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v7, v[3:6], s[6:7]
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v6, a7
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a6
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a5
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a4
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v7, v[3:6], s[6:7] offset:16
+; GFX908-NEXT:    s_cbranch_scc1 .LBB6_2
+; GFX908-NEXT:  ; %bb.1: ; %bb2
+; GFX908-NEXT:    s_add_u32 s8, s4, 48
+; GFX908-NEXT:    s_mov_b32 s13, s9
+; GFX908-NEXT:    s_addc_u32 s9, s5, 0
+; GFX908-NEXT:    s_getpc_b64 s[4:5]
+; GFX908-NEXT:    s_add_u32 s4, s4, foo@gotpcrel32@lo+4
+; GFX908-NEXT:    s_addc_u32 s5, s5, foo@gotpcrel32@hi+12
+; GFX908-NEXT:    s_load_dwordx2 s[16:17], s[4:5], 0x0
+; GFX908-NEXT:    v_lshlrev_b32_e32 v2, 20, v2
+; GFX908-NEXT:    v_lshlrev_b32_e32 v1, 10, v1
+; GFX908-NEXT:    s_mov_b64 s[4:5], s[0:1]
+; GFX908-NEXT:    s_mov_b64 s[6:7], s[2:3]
+; GFX908-NEXT:    s_mov_b64 s[0:1], s[52:53]
+; GFX908-NEXT:    v_or3_b32 v31, v0, v1, v2
+; GFX908-NEXT:    s_mov_b64 s[2:3], s[54:55]
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    s_swappc_b64 s[30:31], s[16:17]
+; GFX908-NEXT:  .LBB6_2: ; %bb3
+; GFX908-NEXT:    s_endpgm
 bb1:
   %in.1 = load <32 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <32 x float> @llvm.amdgcn.mfma.f32.32x32x1f32(float 1.0, float 2.0, <32 x float> %in.1, i32 1, i32 2, i32 3)
@@ -94,10 +928,101 @@ bb3:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_32x32x1f32_nonentry_noagpr:
-; GFX908: v_mfma_f32_32x32x1{{.*}} a[{{[0-9:]+}}], v{{[0-9]+}}, v{{[0-9:]+}}, a[{{[0-9:]+}}]
-; GFX90A: v_mfma_f32_32x32x1{{.*}} v[{{[0-9:]+}}], v{{[0-9]+}}, v{{[0-9:]+}}, v[{{[0-9:]+}}]
 define void @test_mfma_f32_32x32x1f32_nonentry_noagpr(ptr addrspace(1) %arg) #0 {
+; GFX908-LABEL: test_mfma_f32_32x32x1f32_nonentry_noagpr:
+; GFX908:       ; %bb.0: ; %bb
+; GFX908-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX908-NEXT:    global_load_dwordx4 v[30:33], v[0:1], off offset:112
+; GFX908-NEXT:    global_load_dwordx4 v[26:29], v[0:1], off offset:96
+; GFX908-NEXT:    global_load_dwordx4 v[22:25], v[0:1], off offset:80
+; GFX908-NEXT:    global_load_dwordx4 v[18:21], v[0:1], off offset:64
+; GFX908-NEXT:    global_load_dwordx4 v[14:17], v[0:1], off offset:48
+; GFX908-NEXT:    global_load_dwordx4 v[10:13], v[0:1], off offset:32
+; GFX908-NEXT:    global_load_dwordx4 v[6:9], v[0:1], off offset:16
+; GFX908-NEXT:    global_load_dwordx4 v[2:5], v[0:1], off
+; GFX908-NEXT:    s_waitcnt vmcnt(0)
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v4
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, v5
+; GFX908-NEXT:    v_accvgpr_write_b32 a4, v6
+; GFX908-NEXT:    v_accvgpr_write_b32 a5, v7
+; GFX908-NEXT:    v_accvgpr_write_b32 a6, v8
+; GFX908-NEXT:    v_accvgpr_write_b32 a7, v9
+; GFX908-NEXT:    v_accvgpr_write_b32 a8, v10
+; GFX908-NEXT:    v_accvgpr_write_b32 a9, v11
+; GFX908-NEXT:    v_accvgpr_write_b32 a10, v12
+; GFX908-NEXT:    v_accvgpr_write_b32 a11, v13
+; GFX908-NEXT:    v_accvgpr_write_b32 a12, v14
+; GFX908-NEXT:    v_accvgpr_write_b32 a13, v15
+; GFX908-NEXT:    v_accvgpr_write_b32 a14, v16
+; GFX908-NEXT:    v_accvgpr_write_b32 a15, v17
+; GFX908-NEXT:    v_accvgpr_write_b32 a16, v18
+; GFX908-NEXT:    v_accvgpr_write_b32 a17, v19
+; GFX908-NEXT:    v_accvgpr_write_b32 a18, v20
+; GFX908-NEXT:    v_accvgpr_write_b32 a19, v21
+; GFX908-NEXT:    v_accvgpr_write_b32 a20, v22
+; GFX908-NEXT:    v_accvgpr_write_b32 a21, v23
+; GFX908-NEXT:    v_accvgpr_write_b32 a22, v24
+; GFX908-NEXT:    v_accvgpr_write_b32 a23, v25
+; GFX908-NEXT:    v_accvgpr_write_b32 a24, v26
+; GFX908-NEXT:    v_accvgpr_write_b32 a25, v27
+; GFX908-NEXT:    v_accvgpr_write_b32 a26, v28
+; GFX908-NEXT:    v_accvgpr_write_b32 a27, v29
+; GFX908-NEXT:    v_accvgpr_write_b32 a28, v30
+; GFX908-NEXT:    v_accvgpr_write_b32 a29, v31
+; GFX908-NEXT:    v_accvgpr_write_b32 a30, v32
+; GFX908-NEXT:    v_accvgpr_write_b32 a31, v33
+; GFX908-NEXT:    v_mov_b32_e32 v2, 1.0
+; GFX908-NEXT:    v_mov_b32_e32 v3, 2.0
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_mfma_f32_32x32x1f32 a[0:31], v2, v3, a[0:31]
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a27
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a26
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a25
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a24
+; GFX908-NEXT:    v_accvgpr_read_b32 v9, a31
+; GFX908-NEXT:    v_accvgpr_read_b32 v8, a30
+; GFX908-NEXT:    v_accvgpr_read_b32 v7, a29
+; GFX908-NEXT:    v_accvgpr_read_b32 v6, a28
+; GFX908-NEXT:    v_accvgpr_read_b32 v13, a19
+; GFX908-NEXT:    v_accvgpr_read_b32 v12, a18
+; GFX908-NEXT:    v_accvgpr_read_b32 v11, a17
+; GFX908-NEXT:    v_accvgpr_read_b32 v10, a16
+; GFX908-NEXT:    v_accvgpr_read_b32 v17, a23
+; GFX908-NEXT:    v_accvgpr_read_b32 v16, a22
+; GFX908-NEXT:    v_accvgpr_read_b32 v15, a21
+; GFX908-NEXT:    v_accvgpr_read_b32 v14, a20
+; GFX908-NEXT:    v_accvgpr_read_b32 v21, a11
+; GFX908-NEXT:    v_accvgpr_read_b32 v20, a10
+; GFX908-NEXT:    v_accvgpr_read_b32 v19, a9
+; GFX908-NEXT:    v_accvgpr_read_b32 v18, a8
+; GFX908-NEXT:    v_accvgpr_read_b32 v25, a15
+; GFX908-NEXT:    v_accvgpr_read_b32 v24, a14
+; GFX908-NEXT:    v_accvgpr_read_b32 v23, a13
+; GFX908-NEXT:    v_accvgpr_read_b32 v22, a12
+; GFX908-NEXT:    v_accvgpr_read_b32 v29, a3
+; GFX908-NEXT:    v_accvgpr_read_b32 v28, a2
+; GFX908-NEXT:    v_accvgpr_read_b32 v27, a1
+; GFX908-NEXT:    v_accvgpr_read_b32 v26, a0
+; GFX908-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off offset:96
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a7
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a6
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a5
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a4
+; GFX908-NEXT:    global_store_dwordx4 v[0:1], v[6:9], off offset:112
+; GFX908-NEXT:    global_store_dwordx4 v[0:1], v[10:13], off offset:64
+; GFX908-NEXT:    global_store_dwordx4 v[0:1], v[14:17], off offset:80
+; GFX908-NEXT:    global_store_dwordx4 v[0:1], v[18:21], off offset:32
+; GFX908-NEXT:    global_store_dwordx4 v[0:1], v[22:25], off offset:48
+; GFX908-NEXT:    global_store_dwordx4 v[0:1], v[26:29], off
+; GFX908-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off offset:16
+; GFX908-NEXT:    s_waitcnt vmcnt(0)
+; GFX908-NEXT:    s_setpc_b64 s[30:31]
 bb:
   %in.1 = load <32 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <32 x float> @llvm.amdgcn.mfma.f32.32x32x1f32(float 1.0, float 2.0, <32 x float> %in.1, i32 0, i32 0, i32 0)
@@ -105,9 +1030,101 @@ bb:
   ret void
 }
 
-; GCN-LABEL: {{^}}test_mfma_f32_32x32x1f32_nonentry_with_agpr:
-; GCN: v_mfma_f32_32x32x1{{.*}} a[{{[0-9:]+}}], v{{[0-9]+}}, v{{[0-9:]+}}, a[{{[0-9:]+}}]
 define void @test_mfma_f32_32x32x1f32_nonentry_with_agpr(ptr addrspace(1) %arg) #3 {
+; GFX908-LABEL: test_mfma_f32_32x32x1f32_nonentry_with_agpr:
+; GFX908:       ; %bb.0: ; %bb
+; GFX908-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX908-NEXT:    global_load_dwordx4 v[30:33], v[0:1], off offset:112
+; GFX908-NEXT:    global_load_dwordx4 v[26:29], v[0:1], off offset:96
+; GFX908-NEXT:    global_load_dwordx4 v[22:25], v[0:1], off offset:80
+; GFX908-NEXT:    global_load_dwordx4 v[18:21], v[0:1], off offset:64
+; GFX908-NEXT:    global_load_dwordx4 v[14:17], v[0:1], off offset:48
+; GFX908-NEXT:    global_load_dwordx4 v[10:13], v[0:1], off offset:32
+; GFX908-NEXT:    global_load_dwordx4 v[6:9], v[0:1], off offset:16
+; GFX908-NEXT:    global_load_dwordx4 v[2:5], v[0:1], off
+; GFX908-NEXT:    s_waitcnt vmcnt(0)
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v4
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, v5
+; GFX908-NEXT:    v_accvgpr_write_b32 a4, v6
+; GFX908-NEXT:    v_accvgpr_write_b32 a5, v7
+; GFX908-NEXT:    v_accvgpr_write_b32 a6, v8
+; GFX908-NEXT:    v_accvgpr_write_b32 a7, v9
+; GFX908-NEXT:    v_accvgpr_write_b32 a8, v10
+; GFX908-NEXT:    v_accvgpr_write_b32 a9, v11
+; GFX908-NEXT:    v_accvgpr_write_b32 a10, v12
+; GFX908-NEXT:    v_accvgpr_write_b32 a11, v13
+; GFX908-NEXT:    v_accvgpr_write_b32 a12, v14
+; GFX908-NEXT:    v_accvgpr_write_b32 a13, v15
+; GFX908-NEXT:    v_accvgpr_write_b32 a14, v16
+; GFX908-NEXT:    v_accvgpr_write_b32 a15, v17
+; GFX908-NEXT:    v_accvgpr_write_b32 a16, v18
+; GFX908-NEXT:    v_accvgpr_write_b32 a17, v19
+; GFX908-NEXT:    v_accvgpr_write_b32 a18, v20
+; GFX908-NEXT:    v_accvgpr_write_b32 a19, v21
+; GFX908-NEXT:    v_accvgpr_write_b32 a20, v22
+; GFX908-NEXT:    v_accvgpr_write_b32 a21, v23
+; GFX908-NEXT:    v_accvgpr_write_b32 a22, v24
+; GFX908-NEXT:    v_accvgpr_write_b32 a23, v25
+; GFX908-NEXT:    v_accvgpr_write_b32 a24, v26
+; GFX908-NEXT:    v_accvgpr_write_b32 a25, v27
+; GFX908-NEXT:    v_accvgpr_write_b32 a26, v28
+; GFX908-NEXT:    v_accvgpr_write_b32 a27, v29
+; GFX908-NEXT:    v_accvgpr_write_b32 a28, v30
+; GFX908-NEXT:    v_accvgpr_write_b32 a29, v31
+; GFX908-NEXT:    v_accvgpr_write_b32 a30, v32
+; GFX908-NEXT:    v_accvgpr_write_b32 a31, v33
+; GFX908-NEXT:    v_mov_b32_e32 v2, 1.0
+; GFX908-NEXT:    v_mov_b32_e32 v3, 2.0
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_mfma_f32_32x32x1f32 a[0:31], v2, v3, a[0:31]
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a27
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a26
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a25
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a24
+; GFX908-NEXT:    v_accvgpr_read_b32 v9, a31
+; GFX908-NEXT:    v_accvgpr_read_b32 v8, a30
+; GFX908-NEXT:    v_accvgpr_read_b32 v7, a29
+; GFX908-NEXT:    v_accvgpr_read_b32 v6, a28
+; GFX908-NEXT:    v_accvgpr_read_b32 v13, a19
+; GFX908-NEXT:    v_accvgpr_read_b32 v12, a18
+; GFX908-NEXT:    v_accvgpr_read_b32 v11, a17
+; GFX908-NEXT:    v_accvgpr_read_b32 v10, a16
+; GFX908-NEXT:    v_accvgpr_read_b32 v17, a23
+; GFX908-NEXT:    v_accvgpr_read_b32 v16, a22
+; GFX908-NEXT:    v_accvgpr_read_b32 v15, a21
+; GFX908-NEXT:    v_accvgpr_read_b32 v14, a20
+; GFX908-NEXT:    v_accvgpr_read_b32 v21, a11
+; GFX908-NEXT:    v_accvgpr_read_b32 v20, a10
+; GFX908-NEXT:    v_accvgpr_read_b32 v19, a9
+; GFX908-NEXT:    v_accvgpr_read_b32 v18, a8
+; GFX908-NEXT:    v_accvgpr_read_b32 v25, a15
+; GFX908-NEXT:    v_accvgpr_read_b32 v24, a14
+; GFX908-NEXT:    v_accvgpr_read_b32 v23, a13
+; GFX908-NEXT:    v_accvgpr_read_b32 v22, a12
+; GFX908-NEXT:    v_accvgpr_read_b32 v29, a3
+; GFX908-NEXT:    v_accvgpr_read_b32 v28, a2
+; GFX908-NEXT:    v_accvgpr_read_b32 v27, a1
+; GFX908-NEXT:    v_accvgpr_read_b32 v26, a0
+; GFX908-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off offset:96
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a7
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a6
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a5
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a4
+; GFX908-NEXT:    global_store_dwordx4 v[0:1], v[6:9], off offset:112
+; GFX908-NEXT:    global_store_dwordx4 v[0:1], v[10:13], off offset:64
+; GFX908-NEXT:    global_store_dwordx4 v[0:1], v[14:17], off offset:80
+; GFX908-NEXT:    global_store_dwordx4 v[0:1], v[18:21], off offset:32
+; GFX908-NEXT:    global_store_dwordx4 v[0:1], v[22:25], off offset:48
+; GFX908-NEXT:    global_store_dwordx4 v[0:1], v[26:29], off
+; GFX908-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off offset:16
+; GFX908-NEXT:    s_waitcnt vmcnt(0)
+; GFX908-NEXT:    s_setpc_b64 s[30:31]
 bb:
   %in.1 = load <32 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <32 x float> @llvm.amdgcn.mfma.f32.32x32x1f32(float 1.0, float 2.0, <32 x float> %in.1, i32 0, i32 0, i32 0)
@@ -121,3 +1138,6 @@ attributes #0 = { "amdgpu-flat-work-group-size"="1,256" "amdgpu-waves-per-eu"="2
 attributes #1 = { "amdgpu-flat-work-group-size"="1,256" "amdgpu-waves-per-eu"="2" }
 attributes #2 = { "amdgpu-flat-work-group-size"="1,256" "amdgpu-agpr-alloc"="0" }
 attributes #3 = { "amdgpu-flat-work-group-size"="1,256" "amdgpu-waves-per-eu"="2" }
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; GCN: {{.*}}
+; GFX90A: {{.*}}
diff --git a/llvm/test/CodeGen/AMDGPU/packed-fp32.ll b/llvm/test/CodeGen/AMDGPU/packed-fp32.ll
index 9c38d7f..5b0d2d2 100644
--- a/llvm/test/CodeGen/AMDGPU/packed-fp32.ll
+++ b/llvm/test/CodeGen/AMDGPU/packed-fp32.ll
@@ -3565,13 +3565,13 @@ define amdgpu_kernel void @fneg_v2f32_scalar(ptr addrspace(1) %a, <2 x float> %x
 ; PACKED-SDAG-LABEL: fneg_v2f32_scalar:
 ; PACKED-SDAG:       ; %bb.0:
 ; PACKED-SDAG-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
-; PACKED-SDAG-NEXT:    v_mov_b32_e32 v2, 0
+; PACKED-SDAG-NEXT:    v_mov_b32_e32 v0, 0
 ; PACKED-SDAG-NEXT:    s_waitcnt lgkmcnt(0)
 ; PACKED-SDAG-NEXT:    s_xor_b32 s3, s3, 0x80000000
 ; PACKED-SDAG-NEXT:    s_xor_b32 s2, s2, 0x80000000
-; PACKED-SDAG-NEXT:    v_mov_b32_e32 v0, s2
-; PACKED-SDAG-NEXT:    v_mov_b32_e32 v1, s3
-; PACKED-SDAG-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; PACKED-SDAG-NEXT:    v_mov_b32_e32 v2, s2
+; PACKED-SDAG-NEXT:    v_mov_b32_e32 v3, s3
+; PACKED-SDAG-NEXT:    global_store_dwordx2 v0, v[2:3], s[0:1]
 ; PACKED-SDAG-NEXT:    s_endpgm
 ;
 ; PACKED-GISEL-LABEL: fneg_v2f32_scalar:
diff --git a/llvm/test/CodeGen/AMDGPU/partial-regcopy-and-spill-missed-at-regalloc.ll b/llvm/test/CodeGen/AMDGPU/partial-regcopy-and-spill-missed-at-regalloc.ll
index a5c8f04..f54a383 100644
--- a/llvm/test/CodeGen/AMDGPU/partial-regcopy-and-spill-missed-at-regalloc.ll
+++ b/llvm/test/CodeGen/AMDGPU/partial-regcopy-and-spill-missed-at-regalloc.ll
@@ -1,7 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py
-;RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx908 --stop-after=greedy,1 < %s | FileCheck -check-prefix=REGALLOC-GFX908 %s
+;RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx908 --stop-after=greedy,2 < %s | FileCheck -check-prefix=REGALLOC-GFX908 %s
 ;RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx908 --stop-after=prologepilog < %s | FileCheck -check-prefix=PEI-GFX908 %s
-;RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx90a --stop-after=greedy,1 < %s | FileCheck -check-prefix=REGALLOC-GFX90A %s
+;RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx90a --stop-after=greedy,2 < %s | FileCheck -check-prefix=REGALLOC-GFX90A %s
 ;RUN: llc -mtriple=amdgcn-amd-amdhsa -mcpu=gfx90a --stop-after=prologepilog < %s | FileCheck -check-prefix=PEI-GFX90A %s
 
 ; Partial reg copy and spill missed during regalloc handled later at frame lowering.
@@ -12,17 +12,21 @@ define amdgpu_kernel void @partial_copy(<4 x i32> %arg) #0 {
   ; REGALLOC-GFX908-NEXT:   liveins: $sgpr4_sgpr5
   ; REGALLOC-GFX908-NEXT: {{  $}}
   ; REGALLOC-GFX908-NEXT:   INLINEASM &"; use $0", 1 /* sideeffect attdialect */, 2162697 /* reguse:AGPR_32 */, undef %6:agpr_32
-  ; REGALLOC-GFX908-NEXT:   INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 6094858 /* regdef:VReg_128 */, def %7
-  ; REGALLOC-GFX908-NEXT:   INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 3538954 /* regdef:VReg_64 */, def %8
-  ; REGALLOC-GFX908-NEXT:   GLOBAL_STORE_DWORDX4 undef %15:vreg_64, %7, 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1)
+  ; REGALLOC-GFX908-NEXT:   INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 6094858 /* regdef:VReg_128 */, def %25
+  ; REGALLOC-GFX908-NEXT:   [[COPY:%[0-9]+]]:av_128 = COPY %25
+  ; REGALLOC-GFX908-NEXT:   INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 3538954 /* regdef:VReg_64 */, def %27
+  ; REGALLOC-GFX908-NEXT:   SI_SPILL_AV64_SAVE %27, %stack.0, $sgpr32, 0, implicit $exec :: (store (s64) into %stack.0, align 4, addrspace 5)
+  ; REGALLOC-GFX908-NEXT:   [[COPY1:%[0-9]+]]:vreg_128 = COPY [[COPY]]
+  ; REGALLOC-GFX908-NEXT:   GLOBAL_STORE_DWORDX4 undef %15:vreg_64, [[COPY1]], 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1)
   ; REGALLOC-GFX908-NEXT:   renamable $sgpr0_sgpr1_sgpr2_sgpr3 = S_LOAD_DWORDX4_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (s128) from %ir.arg.kernarg.offset1, addrspace 4)
-  ; REGALLOC-GFX908-NEXT:   [[COPY:%[0-9]+]]:areg_128 = COPY killed renamable $sgpr0_sgpr1_sgpr2_sgpr3
+  ; REGALLOC-GFX908-NEXT:   [[COPY2:%[0-9]+]]:areg_128 = COPY killed renamable $sgpr0_sgpr1_sgpr2_sgpr3
   ; REGALLOC-GFX908-NEXT:   [[AV_MOV_:%[0-9]+]]:vgpr_32 = AV_MOV_B32_IMM_PSEUDO 1, implicit $exec
   ; REGALLOC-GFX908-NEXT:   [[AV_MOV_1:%[0-9]+]]:vgpr_32 = AV_MOV_B32_IMM_PSEUDO 2, implicit $exec
-  ; REGALLOC-GFX908-NEXT:   [[V_MFMA_I32_4X4X4I8_e64_:%[0-9]+]]:areg_128 = V_MFMA_I32_4X4X4I8_e64 [[AV_MOV_]], [[AV_MOV_1]], [[COPY]], 0, 0, 0, implicit $mode, implicit $exec
-  ; REGALLOC-GFX908-NEXT:   GLOBAL_STORE_DWORDX2 undef %17:vreg_64, %8, 0, 0, implicit $exec :: (volatile store (s64) into `ptr addrspace(1) poison`, addrspace 1)
-  ; REGALLOC-GFX908-NEXT:   [[COPY1:%[0-9]+]]:vreg_128 = COPY [[V_MFMA_I32_4X4X4I8_e64_]]
-  ; REGALLOC-GFX908-NEXT:   GLOBAL_STORE_DWORDX4 undef %19:vreg_64, [[COPY1]], 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1)
+  ; REGALLOC-GFX908-NEXT:   [[V_MFMA_I32_4X4X4I8_e64_:%[0-9]+]]:areg_128 = V_MFMA_I32_4X4X4I8_e64 [[AV_MOV_]], [[AV_MOV_1]], [[COPY2]], 0, 0, 0, implicit $mode, implicit $exec
+  ; REGALLOC-GFX908-NEXT:   [[SI_SPILL_AV64_RESTORE:%[0-9]+]]:vreg_64 = SI_SPILL_AV64_RESTORE %stack.0, $sgpr32, 0, implicit $exec :: (load (s64) from %stack.0, align 4, addrspace 5)
+  ; REGALLOC-GFX908-NEXT:   GLOBAL_STORE_DWORDX2 undef %17:vreg_64, [[SI_SPILL_AV64_RESTORE]], 0, 0, implicit $exec :: (volatile store (s64) into `ptr addrspace(1) poison`, addrspace 1)
+  ; REGALLOC-GFX908-NEXT:   [[COPY3:%[0-9]+]]:vreg_128 = COPY [[V_MFMA_I32_4X4X4I8_e64_]]
+  ; REGALLOC-GFX908-NEXT:   GLOBAL_STORE_DWORDX4 undef %19:vreg_64, [[COPY3]], 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1)
   ; REGALLOC-GFX908-NEXT:   S_ENDPGM 0
   ;
   ; PEI-GFX908-LABEL: name: partial_copy
@@ -57,40 +61,35 @@ define amdgpu_kernel void @partial_copy(<4 x i32> %arg) #0 {
   ; REGALLOC-GFX90A-NEXT:   liveins: $sgpr4_sgpr5
   ; REGALLOC-GFX90A-NEXT: {{  $}}
   ; REGALLOC-GFX90A-NEXT:   INLINEASM &"; use $0", 1 /* sideeffect attdialect */, 2162697 /* reguse:AGPR_32 */, undef %6:agpr_32
-  ; REGALLOC-GFX90A-NEXT:   INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 6422538 /* regdef:VReg_128_Align2 */, def %7
-  ; REGALLOC-GFX90A-NEXT:   INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 3866634 /* regdef:VReg_64_Align2 */, def %8
-  ; REGALLOC-GFX90A-NEXT:   GLOBAL_STORE_DWORDX4 undef %15:vreg_64_align2, %7, 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1)
+  ; REGALLOC-GFX90A-NEXT:   INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 6422538 /* regdef:VReg_128_Align2 */, def %24
+  ; REGALLOC-GFX90A-NEXT:   [[COPY:%[0-9]+]]:av_128_align2 = COPY %24
+  ; REGALLOC-GFX90A-NEXT:   INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 3866634 /* regdef:VReg_64_Align2 */, def %22
+  ; REGALLOC-GFX90A-NEXT:   [[COPY1:%[0-9]+]]:av_64_align2 = COPY %22
+  ; REGALLOC-GFX90A-NEXT:   GLOBAL_STORE_DWORDX4 undef %15:vreg_64_align2, [[COPY]], 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1)
   ; REGALLOC-GFX90A-NEXT:   renamable $sgpr0_sgpr1_sgpr2_sgpr3 = S_LOAD_DWORDX4_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (s128) from %ir.arg.kernarg.offset1, addrspace 4)
-  ; REGALLOC-GFX90A-NEXT:   [[COPY:%[0-9]+]]:areg_128_align2 = COPY killed renamable $sgpr0_sgpr1_sgpr2_sgpr3
+  ; REGALLOC-GFX90A-NEXT:   [[COPY2:%[0-9]+]]:areg_128_align2 = COPY killed renamable $sgpr0_sgpr1_sgpr2_sgpr3
   ; REGALLOC-GFX90A-NEXT:   [[AV_MOV_:%[0-9]+]]:vgpr_32 = AV_MOV_B32_IMM_PSEUDO 1, implicit $exec
   ; REGALLOC-GFX90A-NEXT:   [[AV_MOV_1:%[0-9]+]]:vgpr_32 = AV_MOV_B32_IMM_PSEUDO 2, implicit $exec
-  ; REGALLOC-GFX90A-NEXT:   [[V_MFMA_I32_4X4X4I8_e64_:%[0-9]+]]:areg_128_align2 = V_MFMA_I32_4X4X4I8_e64 [[AV_MOV_]], [[AV_MOV_1]], [[COPY]], 0, 0, 0, implicit $mode, implicit $exec
-  ; REGALLOC-GFX90A-NEXT:   GLOBAL_STORE_DWORDX2 undef %17:vreg_64_align2, %8, 0, 0, implicit $exec :: (volatile store (s64) into `ptr addrspace(1) poison`, addrspace 1)
+  ; REGALLOC-GFX90A-NEXT:   [[V_MFMA_I32_4X4X4I8_e64_:%[0-9]+]]:areg_128_align2 = V_MFMA_I32_4X4X4I8_e64 [[AV_MOV_]], [[AV_MOV_1]], [[COPY2]], 0, 0, 0, implicit $mode, implicit $exec
+  ; REGALLOC-GFX90A-NEXT:   GLOBAL_STORE_DWORDX2 undef %17:vreg_64_align2, [[COPY1]], 0, 0, implicit $exec :: (volatile store (s64) into `ptr addrspace(1) poison`, addrspace 1)
   ; REGALLOC-GFX90A-NEXT:   GLOBAL_STORE_DWORDX4 undef %19:vreg_64_align2, [[V_MFMA_I32_4X4X4I8_e64_]], 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1)
   ; REGALLOC-GFX90A-NEXT:   S_ENDPGM 0
   ;
   ; PEI-GFX90A-LABEL: name: partial_copy
   ; PEI-GFX90A: bb.0 (%ir-block.0):
-  ; PEI-GFX90A-NEXT:   liveins: $agpr4, $sgpr4_sgpr5, $sgpr0_sgpr1_sgpr2_sgpr3, $sgpr9
+  ; PEI-GFX90A-NEXT:   liveins: $sgpr4_sgpr5
   ; PEI-GFX90A-NEXT: {{  $}}
-  ; PEI-GFX90A-NEXT:   $sgpr12_sgpr13_sgpr14_sgpr15 = COPY killed $sgpr0_sgpr1_sgpr2_sgpr3
-  ; PEI-GFX90A-NEXT:   $sgpr12 = S_ADD_U32 $sgpr12, $sgpr9, implicit-def $scc, implicit-def $sgpr12_sgpr13_sgpr14_sgpr15
-  ; PEI-GFX90A-NEXT:   $sgpr13 = S_ADDC_U32 $sgpr13, 0, implicit-def dead $scc, implicit $scc, implicit-def $sgpr12_sgpr13_sgpr14_sgpr15
   ; PEI-GFX90A-NEXT:   INLINEASM &"; use $0", 1 /* sideeffect attdialect */, 2162697 /* reguse:AGPR_32 */, undef renamable $agpr0
   ; PEI-GFX90A-NEXT:   INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 6422538 /* regdef:VReg_128_Align2 */, def renamable $vgpr0_vgpr1_vgpr2_vgpr3
   ; PEI-GFX90A-NEXT:   renamable $agpr0_agpr1_agpr2_agpr3 = COPY killed renamable $vgpr0_vgpr1_vgpr2_vgpr3, implicit $exec
-  ; PEI-GFX90A-NEXT:   INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 3866634 /* regdef:VReg_64_Align2 */, def renamable $vgpr0_vgpr1
-  ; PEI-GFX90A-NEXT:   BUFFER_STORE_DWORD_OFFSET killed $vgpr0, $sgpr12_sgpr13_sgpr14_sgpr15, 0, 0, 0, 0, implicit $exec, implicit-def $vgpr0_vgpr1, implicit $vgpr0_vgpr1 :: (store (s32) into %stack.0, addrspace 5)
-  ; PEI-GFX90A-NEXT:   $agpr4 = V_ACCVGPR_WRITE_B32_e64 killed $vgpr1, implicit $exec, implicit killed $vgpr0_vgpr1
+  ; PEI-GFX90A-NEXT:   INLINEASM &"; def $0", 1 /* sideeffect attdialect */, 3866634 /* regdef:VReg_64_Align2 */, def renamable $vgpr2_vgpr3
   ; PEI-GFX90A-NEXT:   GLOBAL_STORE_DWORDX4 undef renamable $vgpr0_vgpr1, killed renamable $agpr0_agpr1_agpr2_agpr3, 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1)
   ; PEI-GFX90A-NEXT:   renamable $sgpr0_sgpr1_sgpr2_sgpr3 = S_LOAD_DWORDX4_IMM killed renamable $sgpr4_sgpr5, 0, 0 :: (dereferenceable invariant load (s128) from %ir.arg.kernarg.offset1, addrspace 4)
   ; PEI-GFX90A-NEXT:   renamable $agpr0_agpr1_agpr2_agpr3 = COPY killed renamable $sgpr0_sgpr1_sgpr2_sgpr3, implicit $exec
   ; PEI-GFX90A-NEXT:   renamable $vgpr0 = AV_MOV_B32_IMM_PSEUDO 1, implicit $exec
   ; PEI-GFX90A-NEXT:   renamable $vgpr1 = AV_MOV_B32_IMM_PSEUDO 2, implicit $exec
   ; PEI-GFX90A-NEXT:   renamable $agpr0_agpr1_agpr2_agpr3 = V_MFMA_I32_4X4X4I8_e64 killed $vgpr0, killed $vgpr1, killed $agpr0_agpr1_agpr2_agpr3, 0, 0, 0, implicit $mode, implicit $exec
-  ; PEI-GFX90A-NEXT:   $vgpr0 = BUFFER_LOAD_DWORD_OFFSET $sgpr12_sgpr13_sgpr14_sgpr15, 0, 0, 0, 0, implicit $exec, implicit-def $vgpr0_vgpr1 :: (load (s32) from %stack.0, addrspace 5)
-  ; PEI-GFX90A-NEXT:   $vgpr1 = V_ACCVGPR_READ_B32_e64 $agpr4, implicit $exec, implicit $vgpr0_vgpr1
-  ; PEI-GFX90A-NEXT:   GLOBAL_STORE_DWORDX2 undef renamable $vgpr0_vgpr1, killed renamable $vgpr0_vgpr1, 0, 0, implicit $exec :: (volatile store (s64) into `ptr addrspace(1) poison`, addrspace 1)
+  ; PEI-GFX90A-NEXT:   GLOBAL_STORE_DWORDX2 undef renamable $vgpr0_vgpr1, killed renamable $vgpr2_vgpr3, 0, 0, implicit $exec :: (volatile store (s64) into `ptr addrspace(1) poison`, addrspace 1)
   ; PEI-GFX90A-NEXT:   GLOBAL_STORE_DWORDX4 undef renamable $vgpr0_vgpr1, killed renamable $agpr0_agpr1_agpr2_agpr3, 0, 0, implicit $exec :: (volatile store (s128) into `ptr addrspace(1) poison`, addrspace 1)
   ; PEI-GFX90A-NEXT:   S_ENDPGM 0
   call void asm sideeffect "; use $0", "a" (i32 poison)
diff --git a/llvm/test/CodeGen/AMDGPU/ptradd-sdag-optimizations.ll b/llvm/test/CodeGen/AMDGPU/ptradd-sdag-optimizations.ll
index d48bfe0..68ef30a9 100644
--- a/llvm/test/CodeGen/AMDGPU/ptradd-sdag-optimizations.ll
+++ b/llvm/test/CodeGen/AMDGPU/ptradd-sdag-optimizations.ll
@@ -53,31 +53,31 @@ define amdgpu_kernel void @store_v16i32(ptr addrspace(1) %out, <16 x i32> %a) {
 ; GFX942:       ; %bb.0: ; %entry
 ; GFX942-NEXT:    s_load_dwordx16 s[8:23], s[4:5], 0x40
 ; GFX942-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x0
-; GFX942-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b32_e32 v0, s20
-; GFX942-NEXT:    v_mov_b32_e32 v1, s21
-; GFX942-NEXT:    v_mov_b32_e32 v2, s22
-; GFX942-NEXT:    v_mov_b32_e32 v3, s23
-; GFX942-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:48
+; GFX942-NEXT:    v_mov_b32_e32 v2, s20
+; GFX942-NEXT:    v_mov_b32_e32 v3, s21
+; GFX942-NEXT:    v_mov_b32_e32 v4, s22
+; GFX942-NEXT:    v_mov_b32_e32 v5, s23
+; GFX942-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1] offset:48
 ; GFX942-NEXT:    s_nop 1
-; GFX942-NEXT:    v_mov_b32_e32 v0, s16
-; GFX942-NEXT:    v_mov_b32_e32 v1, s17
-; GFX942-NEXT:    v_mov_b32_e32 v2, s18
-; GFX942-NEXT:    v_mov_b32_e32 v3, s19
-; GFX942-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:32
+; GFX942-NEXT:    v_mov_b32_e32 v2, s16
+; GFX942-NEXT:    v_mov_b32_e32 v3, s17
+; GFX942-NEXT:    v_mov_b32_e32 v4, s18
+; GFX942-NEXT:    v_mov_b32_e32 v5, s19
+; GFX942-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1] offset:32
 ; GFX942-NEXT:    s_nop 1
-; GFX942-NEXT:    v_mov_b32_e32 v0, s12
-; GFX942-NEXT:    v_mov_b32_e32 v1, s13
-; GFX942-NEXT:    v_mov_b32_e32 v2, s14
-; GFX942-NEXT:    v_mov_b32_e32 v3, s15
-; GFX942-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1] offset:16
+; GFX942-NEXT:    v_mov_b32_e32 v2, s12
+; GFX942-NEXT:    v_mov_b32_e32 v3, s13
+; GFX942-NEXT:    v_mov_b32_e32 v4, s14
+; GFX942-NEXT:    v_mov_b32_e32 v5, s15
+; GFX942-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1] offset:16
 ; GFX942-NEXT:    s_nop 1
-; GFX942-NEXT:    v_mov_b32_e32 v0, s8
-; GFX942-NEXT:    v_mov_b32_e32 v1, s9
-; GFX942-NEXT:    v_mov_b32_e32 v2, s10
-; GFX942-NEXT:    v_mov_b32_e32 v3, s11
-; GFX942-NEXT:    global_store_dwordx4 v4, v[0:3], s[0:1]
+; GFX942-NEXT:    v_mov_b32_e32 v2, s8
+; GFX942-NEXT:    v_mov_b32_e32 v3, s9
+; GFX942-NEXT:    v_mov_b32_e32 v4, s10
+; GFX942-NEXT:    v_mov_b32_e32 v5, s11
+; GFX942-NEXT:    global_store_dwordx4 v0, v[2:5], s[0:1]
 ; GFX942-NEXT:    s_endpgm
 entry:
   store <16 x i32> %a, ptr addrspace(1) %out
diff --git a/llvm/test/CodeGen/AMDGPU/read-write-register-illegal-type.ll b/llvm/test/CodeGen/AMDGPU/read-write-register-illegal-type.ll
new file mode 100644
index 0000000..2324f3f
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/read-write-register-illegal-type.ll
@@ -0,0 +1,29 @@
+; RUN: not llc -mtriple=amdgcn -mcpu=gfx900 -filetype=null %s 2>&1 | FileCheck %s
+
+; CHECK: error: <unknown>:0:0: cannot use llvm.read_register with illegal type
+define amdgpu_kernel void @test_read_register_i9(ptr addrspace(1) %out) nounwind {
+  %reg = call i9 @llvm.read_register.i9(metadata !0)
+  store i9 %reg, ptr addrspace(1) %out
+  ret void
+}
+
+; CHECK: error: <unknown>:0:0: cannot use llvm.write_register with illegal type
+define amdgpu_kernel void @test_write_register_i9(ptr addrspace(1) %out) nounwind {
+ call void @llvm.write_register.i9(metadata !0, i9 42)
+ ret void
+}
+
+; CHECK: error: <unknown>:0:0: cannot use llvm.read_register with illegal type
+define amdgpu_kernel void @test_read_register_i128(ptr addrspace(1) %out) nounwind {
+  %reg = call i128 @llvm.read_register.i128(metadata !0)
+  store i128 %reg, ptr addrspace(1) %out
+  ret void
+}
+
+; CHECK: error: <unknown>:0:0: cannot use llvm.write_register with illegal type
+define amdgpu_kernel void @test_write_register_i128(ptr addrspace(1) %out) nounwind {
+ call void @llvm.write_register.i128(metadata !0, i128 42)
+ ret void
+}
+
+!0 = !{!"m0"}
diff --git a/llvm/test/CodeGen/AMDGPU/regalloc-illegal-eviction-assert.ll b/llvm/test/CodeGen/AMDGPU/regalloc-illegal-eviction-assert.ll
index f2fd3a8..c035e9f 100644
--- a/llvm/test/CodeGen/AMDGPU/regalloc-illegal-eviction-assert.ll
+++ b/llvm/test/CodeGen/AMDGPU/regalloc-illegal-eviction-assert.ll
@@ -9,9 +9,9 @@
 %asm.output = type { <16 x i32>, <8 x i32>, <5 x i32>, <4 x i32>, <16 x i32> }
 
 ; CHECK-LABEL: {{^}}illegal_eviction_assert:
-; CHECK: ; def v[4:19] v[20:27] v[0:4] v[0:3] a[0:15]
+; CHECK: ; def v[13:28] v[0:7] v[8:12] v[0:3] a[0:15]
 ; CHECK: ; clobber
-; CHECK: ; use v[4:19] v[20:27] v[0:4] v[0:3] a[1:16]
+; CHECK: ; use v[13:28] v[0:7] v[8:12] v[0:3] a[1:16]
 define void @illegal_eviction_assert(ptr addrspace(1) %arg) #0 {
   ;%agpr0 = call i32 asm sideeffect "; def $0","=${a0}"()
   %asm = call %asm.output asm sideeffect "; def $0 $1 $2 $3 $4","=v,=v,=v,=v,={a[0:15]}"()
diff --git a/llvm/test/CodeGen/AMDGPU/scalar_to_vector.v8i16.ll b/llvm/test/CodeGen/AMDGPU/scalar_to_vector.v8i16.ll
index e8e122e..bbb9df9 100644
--- a/llvm/test/CodeGen/AMDGPU/scalar_to_vector.v8i16.ll
+++ b/llvm/test/CodeGen/AMDGPU/scalar_to_vector.v8i16.ll
@@ -59,19 +59,19 @@ define amdgpu_kernel void @scalar_to_vector_v8i16(<2 x i32> %in, ptr %out) #0 {
 ; GFX90A-LABEL: scalar_to_vector_v8i16:
 ; GFX90A:       ; %bb.0: ; %entry
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
-; GFX90A-NEXT:    v_and_b32_e32 v4, 0x3ff, v0
-; GFX90A-NEXT:    v_lshlrev_b32_e32 v4, 4, v4
+; GFX90A-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
+; GFX90A-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
 ; GFX90A-NEXT:    s_add_u32 flat_scratch_lo, s12, s17
 ; GFX90A-NEXT:    s_addc_u32 flat_scratch_hi, s13, 0
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_mov_b32_e32 v5, s3
-; GFX90A-NEXT:    v_add_co_u32_e32 v4, vcc, s2, v4
-; GFX90A-NEXT:    v_mov_b32_e32 v0, s0
-; GFX90A-NEXT:    v_mov_b32_e32 v1, s1
+; GFX90A-NEXT:    v_mov_b32_e32 v1, s3
+; GFX90A-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v0
 ; GFX90A-NEXT:    v_mov_b32_e32 v2, s0
-; GFX90A-NEXT:    v_mov_b32_e32 v3, s0
-; GFX90A-NEXT:    v_addc_co_u32_e32 v5, vcc, 0, v5, vcc
-; GFX90A-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX90A-NEXT:    v_mov_b32_e32 v3, s1
+; GFX90A-NEXT:    v_mov_b32_e32 v4, s0
+; GFX90A-NEXT:    v_mov_b32_e32 v5, s0
+; GFX90A-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX90A-NEXT:    flat_store_dwordx4 v[0:1], v[2:5]
 ; GFX90A-NEXT:    s_endpgm
 entry:
   %val.1.i32 = extractelement <2 x i32> %in, i64 0
@@ -146,19 +146,19 @@ define amdgpu_kernel void @scalar_to_vector_v8f16(<2 x float> %in, ptr %out) #0
 ; GFX90A-LABEL: scalar_to_vector_v8f16:
 ; GFX90A:       ; %bb.0: ; %entry
 ; GFX90A-NEXT:    s_load_dwordx4 s[0:3], s[8:9], 0x0
-; GFX90A-NEXT:    v_and_b32_e32 v4, 0x3ff, v0
-; GFX90A-NEXT:    v_lshlrev_b32_e32 v4, 4, v4
+; GFX90A-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
+; GFX90A-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
 ; GFX90A-NEXT:    s_add_u32 flat_scratch_lo, s12, s17
 ; GFX90A-NEXT:    s_addc_u32 flat_scratch_hi, s13, 0
 ; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX90A-NEXT:    v_mov_b32_e32 v5, s3
-; GFX90A-NEXT:    v_add_co_u32_e32 v4, vcc, s2, v4
-; GFX90A-NEXT:    v_mov_b32_e32 v0, s0
-; GFX90A-NEXT:    v_mov_b32_e32 v1, s1
-; GFX90A-NEXT:    v_mov_b32_e32 v3, s0
+; GFX90A-NEXT:    v_mov_b32_e32 v1, s3
+; GFX90A-NEXT:    v_add_co_u32_e32 v0, vcc, s2, v0
 ; GFX90A-NEXT:    v_mov_b32_e32 v2, s0
-; GFX90A-NEXT:    v_addc_co_u32_e32 v5, vcc, 0, v5, vcc
-; GFX90A-NEXT:    flat_store_dwordx4 v[4:5], v[0:3]
+; GFX90A-NEXT:    v_mov_b32_e32 v3, s1
+; GFX90A-NEXT:    v_mov_b32_e32 v5, s0
+; GFX90A-NEXT:    v_mov_b32_e32 v4, s0
+; GFX90A-NEXT:    v_addc_co_u32_e32 v1, vcc, 0, v1, vcc
+; GFX90A-NEXT:    flat_store_dwordx4 v[0:1], v[2:5]
 ; GFX90A-NEXT:    s_endpgm
 entry:
   %val.1.float = extractelement <2 x float> %in, i64 0
diff --git a/llvm/test/CodeGen/AMDGPU/shufflevector-physreg-copy.ll b/llvm/test/CodeGen/AMDGPU/shufflevector-physreg-copy.ll
index 9361187..4d864ad 100644
--- a/llvm/test/CodeGen/AMDGPU/shufflevector-physreg-copy.ll
+++ b/llvm/test/CodeGen/AMDGPU/shufflevector-physreg-copy.ll
@@ -25,27 +25,27 @@ define void @shufflevector_v2i32_10_physreg_even_vgpr_pair_copy(ptr addrspace(1)
 ; GFX90A-LABEL: shufflevector_v2i32_10_physreg_even_vgpr_pair_copy:
 ; GFX90A:       ; %bb.0:
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; def v4, v5
 ; GFX90A-NEXT:    ;;#ASMEND
-; GFX90A-NEXT:    v_mov_b32_e32 v0, v5
-; GFX90A-NEXT:    v_mov_b32_e32 v1, v4
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[16:17]
+; GFX90A-NEXT:    v_mov_b32_e32 v2, v5
+; GFX90A-NEXT:    v_mov_b32_e32 v3, v4
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[16:17]
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX940-LABEL: shufflevector_v2i32_10_physreg_even_vgpr_pair_copy:
 ; GFX940:       ; %bb.0:
 ; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX940-NEXT:    v_mov_b32_e32 v2, 0
+; GFX940-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NEXT:    ;;#ASMSTART
 ; GFX940-NEXT:    ; def v4, v5
 ; GFX940-NEXT:    ;;#ASMEND
 ; GFX940-NEXT:    s_nop 0
-; GFX940-NEXT:    v_mov_b32_e32 v0, v5
-; GFX940-NEXT:    v_mov_b32_e32 v1, v4
-; GFX940-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX940-NEXT:    v_mov_b32_e32 v2, v5
+; GFX940-NEXT:    v_mov_b32_e32 v3, v4
+; GFX940-NEXT:    global_store_dwordx2 v0, v[2:3], s[0:1]
 ; GFX940-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NEXT:    s_setpc_b64 s[30:31]
   %asm = call { i32, i32 } asm "; def $0, $1", "={v4},={v5}"()
@@ -214,27 +214,27 @@ define void @shufflevector_v2i32_11_physreg_even_vgpr_pair_copy(ptr addrspace(1)
 ; GFX90A-LABEL: shufflevector_v2i32_11_physreg_even_vgpr_pair_copy:
 ; GFX90A:       ; %bb.0:
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX90A-NEXT:    v_mov_b32_e32 v2, 0
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; def v4, v5
 ; GFX90A-NEXT:    ;;#ASMEND
-; GFX90A-NEXT:    v_mov_b32_e32 v0, v5
-; GFX90A-NEXT:    v_mov_b32_e32 v1, v5
-; GFX90A-NEXT:    global_store_dwordx2 v2, v[0:1], s[16:17]
+; GFX90A-NEXT:    v_mov_b32_e32 v2, v5
+; GFX90A-NEXT:    v_mov_b32_e32 v3, v5
+; GFX90A-NEXT:    global_store_dwordx2 v0, v[2:3], s[16:17]
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX940-LABEL: shufflevector_v2i32_11_physreg_even_vgpr_pair_copy:
 ; GFX940:       ; %bb.0:
 ; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX940-NEXT:    v_mov_b32_e32 v2, 0
+; GFX940-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NEXT:    ;;#ASMSTART
 ; GFX940-NEXT:    ; def v4, v5
 ; GFX940-NEXT:    ;;#ASMEND
 ; GFX940-NEXT:    s_nop 0
-; GFX940-NEXT:    v_mov_b32_e32 v0, v5
-; GFX940-NEXT:    v_mov_b32_e32 v1, v5
-; GFX940-NEXT:    global_store_dwordx2 v2, v[0:1], s[0:1]
+; GFX940-NEXT:    v_mov_b32_e32 v2, v5
+; GFX940-NEXT:    v_mov_b32_e32 v3, v5
+; GFX940-NEXT:    global_store_dwordx2 v0, v[2:3], s[0:1]
 ; GFX940-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NEXT:    s_setpc_b64 s[30:31]
   %asm = call { i32, i32 } asm "; def $0, $1", "={v4},={v5}"()
@@ -265,31 +265,31 @@ define void @shufflevector_v4i32_3210_physreg_even_vgpr_quad_copy(ptr addrspace(
 ; GFX90A-LABEL: shufflevector_v4i32_3210_physreg_even_vgpr_quad_copy:
 ; GFX90A:       ; %bb.0:
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX90A-NEXT:    v_mov_b32_e32 v8, 0
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; def v4, v5, v6, v7
 ; GFX90A-NEXT:    ;;#ASMEND
-; GFX90A-NEXT:    v_mov_b32_e32 v2, v5
-; GFX90A-NEXT:    v_mov_b32_e32 v1, v6
-; GFX90A-NEXT:    v_mov_b32_e32 v0, v7
-; GFX90A-NEXT:    v_mov_b32_e32 v3, v4
-; GFX90A-NEXT:    global_store_dwordx4 v8, v[0:3], s[16:17]
+; GFX90A-NEXT:    v_mov_b32_e32 v10, v5
+; GFX90A-NEXT:    v_mov_b32_e32 v9, v6
+; GFX90A-NEXT:    v_mov_b32_e32 v8, v7
+; GFX90A-NEXT:    v_mov_b32_e32 v11, v4
+; GFX90A-NEXT:    global_store_dwordx4 v0, v[8:11], s[16:17]
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX940-LABEL: shufflevector_v4i32_3210_physreg_even_vgpr_quad_copy:
 ; GFX940:       ; %bb.0:
 ; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX940-NEXT:    v_mov_b32_e32 v8, 0
+; GFX940-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NEXT:    ;;#ASMSTART
 ; GFX940-NEXT:    ; def v4, v5, v6, v7
 ; GFX940-NEXT:    ;;#ASMEND
 ; GFX940-NEXT:    s_nop 0
-; GFX940-NEXT:    v_mov_b32_e32 v2, v5
-; GFX940-NEXT:    v_mov_b32_e32 v1, v6
-; GFX940-NEXT:    v_mov_b32_e32 v0, v7
-; GFX940-NEXT:    v_mov_b32_e32 v3, v4
-; GFX940-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX940-NEXT:    v_mov_b32_e32 v10, v5
+; GFX940-NEXT:    v_mov_b32_e32 v9, v6
+; GFX940-NEXT:    v_mov_b32_e32 v8, v7
+; GFX940-NEXT:    v_mov_b32_e32 v11, v4
+; GFX940-NEXT:    global_store_dwordx4 v0, v[8:11], s[0:1]
 ; GFX940-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NEXT:    s_setpc_b64 s[30:31]
   %asm = call { i32, i32, i32, i32 } asm "; def $0, $1, $2, $3", "={v4},={v5},={v6},={v7}"()
@@ -327,31 +327,31 @@ define void @shufflevector_v4i32_1032_physreg_even_vgpr_quad_copy(ptr addrspace(
 ; GFX90A-LABEL: shufflevector_v4i32_1032_physreg_even_vgpr_quad_copy:
 ; GFX90A:       ; %bb.0:
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX90A-NEXT:    v_mov_b32_e32 v8, 0
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; def v4, v5, v6, v7
 ; GFX90A-NEXT:    ;;#ASMEND
-; GFX90A-NEXT:    v_mov_b32_e32 v0, v5
-; GFX90A-NEXT:    v_mov_b32_e32 v3, v6
-; GFX90A-NEXT:    v_mov_b32_e32 v2, v7
-; GFX90A-NEXT:    v_mov_b32_e32 v1, v4
-; GFX90A-NEXT:    global_store_dwordx4 v8, v[0:3], s[16:17]
+; GFX90A-NEXT:    v_mov_b32_e32 v8, v5
+; GFX90A-NEXT:    v_mov_b32_e32 v11, v6
+; GFX90A-NEXT:    v_mov_b32_e32 v10, v7
+; GFX90A-NEXT:    v_mov_b32_e32 v9, v4
+; GFX90A-NEXT:    global_store_dwordx4 v0, v[8:11], s[16:17]
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX940-LABEL: shufflevector_v4i32_1032_physreg_even_vgpr_quad_copy:
 ; GFX940:       ; %bb.0:
 ; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX940-NEXT:    v_mov_b32_e32 v8, 0
+; GFX940-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NEXT:    ;;#ASMSTART
 ; GFX940-NEXT:    ; def v4, v5, v6, v7
 ; GFX940-NEXT:    ;;#ASMEND
 ; GFX940-NEXT:    s_nop 0
-; GFX940-NEXT:    v_mov_b32_e32 v0, v5
-; GFX940-NEXT:    v_mov_b32_e32 v3, v6
-; GFX940-NEXT:    v_mov_b32_e32 v2, v7
-; GFX940-NEXT:    v_mov_b32_e32 v1, v4
-; GFX940-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
+; GFX940-NEXT:    v_mov_b32_e32 v8, v5
+; GFX940-NEXT:    v_mov_b32_e32 v11, v6
+; GFX940-NEXT:    v_mov_b32_e32 v10, v7
+; GFX940-NEXT:    v_mov_b32_e32 v9, v4
+; GFX940-NEXT:    global_store_dwordx4 v0, v[8:11], s[0:1]
 ; GFX940-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NEXT:    s_setpc_b64 s[30:31]
   %asm = call { i32, i32, i32, i32 } asm "; def $0, $1, $2, $3", "={v4},={v5},={v6},={v7}"()
@@ -746,16 +746,15 @@ define i32 @shufflevector_v4i32_3210_physreg_even_vgpr_quad_copy_other_use_elt(p
 ; GFX90A-LABEL: shufflevector_v4i32_3210_physreg_even_vgpr_quad_copy_other_use_elt:
 ; GFX90A:       ; %bb.0:
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX90A-NEXT:    v_mov_b32_e32 v8, 0
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX90A-NEXT:    ;;#ASMSTART
 ; GFX90A-NEXT:    ; def v4, v5, v6, v7
 ; GFX90A-NEXT:    ;;#ASMEND
-; GFX90A-NEXT:    v_mov_b32_e32 v2, v5
-; GFX90A-NEXT:    v_mov_b32_e32 v1, v6
-; GFX90A-NEXT:    v_mov_b32_e32 v0, v7
-; GFX90A-NEXT:    v_mov_b32_e32 v3, v4
-; GFX90A-NEXT:    global_store_dwordx4 v8, v[0:3], s[16:17]
-; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    v_mov_b32_e32 v10, v5
+; GFX90A-NEXT:    v_mov_b32_e32 v9, v6
+; GFX90A-NEXT:    v_mov_b32_e32 v8, v7
+; GFX90A-NEXT:    v_mov_b32_e32 v11, v4
+; GFX90A-NEXT:    global_store_dwordx4 v0, v[8:11], s[16:17]
 ; GFX90A-NEXT:    v_mov_b32_e32 v0, v6
 ; GFX90A-NEXT:    s_waitcnt vmcnt(0)
 ; GFX90A-NEXT:    s_setpc_b64 s[30:31]
@@ -763,17 +762,16 @@ define i32 @shufflevector_v4i32_3210_physreg_even_vgpr_quad_copy_other_use_elt(p
 ; GFX940-LABEL: shufflevector_v4i32_3210_physreg_even_vgpr_quad_copy_other_use_elt:
 ; GFX940:       ; %bb.0:
 ; GFX940-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX940-NEXT:    v_mov_b32_e32 v8, 0
+; GFX940-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX940-NEXT:    ;;#ASMSTART
 ; GFX940-NEXT:    ; def v4, v5, v6, v7
 ; GFX940-NEXT:    ;;#ASMEND
 ; GFX940-NEXT:    s_nop 0
-; GFX940-NEXT:    v_mov_b32_e32 v2, v5
-; GFX940-NEXT:    v_mov_b32_e32 v1, v6
-; GFX940-NEXT:    v_mov_b32_e32 v0, v7
-; GFX940-NEXT:    v_mov_b32_e32 v3, v4
-; GFX940-NEXT:    global_store_dwordx4 v8, v[0:3], s[0:1]
-; GFX940-NEXT:    s_nop 1
+; GFX940-NEXT:    v_mov_b32_e32 v10, v5
+; GFX940-NEXT:    v_mov_b32_e32 v9, v6
+; GFX940-NEXT:    v_mov_b32_e32 v8, v7
+; GFX940-NEXT:    v_mov_b32_e32 v11, v4
+; GFX940-NEXT:    global_store_dwordx4 v0, v[8:11], s[0:1]
 ; GFX940-NEXT:    v_mov_b32_e32 v0, v6
 ; GFX940-NEXT:    s_waitcnt vmcnt(0)
 ; GFX940-NEXT:    s_setpc_b64 s[30:31]
diff --git a/llvm/test/CodeGen/AMDGPU/spill-agpr.ll b/llvm/test/CodeGen/AMDGPU/spill-agpr.ll
index 5484f77..eb0d546 100644
--- a/llvm/test/CodeGen/AMDGPU/spill-agpr.ll
+++ b/llvm/test/CodeGen/AMDGPU/spill-agpr.ll
@@ -1,15 +1,107 @@
-; RUN: llc -mtriple=amdgcn -mcpu=gfx908 < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GFX908 %s
-; RUN: llc -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck -enable-var-scope -check-prefixes=GCN,GFX90A %s
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -mtriple=amdgcn -mcpu=gfx908 < %s | FileCheck -check-prefixes=GCN,GFX908 %s
+; RUN: llc -mtriple=amdgcn -mcpu=gfx90a < %s | FileCheck -check-prefixes=GCN,GFX90A %s
 
-; GCN-LABEL: {{^}}max_12regs_13a_used:
-; GCN-NOT: s_mov_b32 s{{[0-9]+}}, SCRATCH_RSRC_DWORD0
-; GCN-NOT: s_mov_b32 s{{[0-9]+}}, SCRATCH_RSRC_DWORD1
-; GCN:     v_accvgpr_read_b32 v[[VSPILL:[0-9]+]], a{{[0-9]+}}
-; GCN-NOT: buffer_store_dword
-; GCN-NOT: buffer_load_dword
-; GCN:     v_accvgpr_write_b32 a{{[0-9]+}}, v[[VSPILL]]
-; GCN:     ScratchSize: 0
 define amdgpu_kernel void @max_12regs_13a_used(i32 %cond, ptr addrspace(1) %arg, ptr addrspace(1) %out) #2 {
+; GFX908-LABEL: max_12regs_13a_used:
+; GFX908:       ; %bb.0: ; %bb
+; GFX908-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x2c
+; GFX908-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX908-NEXT:    s_load_dword s0, s[0:1], 0x24
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    s_cmp_lg_u32 s0, 0
+; GFX908-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    v_mov_b32_e32 v5, s8
+; GFX908-NEXT:    v_mov_b32_e32 v1, s9
+; GFX908-NEXT:    v_mov_b32_e32 v2, s10
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, v5
+; GFX908-NEXT:    v_mov_b32_e32 v5, s11
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, v5
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    v_mfma_f32_4x4x1f32 a[0:3], v0, v0, a[0:3]
+; GFX908-NEXT:    v_mfma_f32_4x4x1f32 a[4:7], v0, v0, a[0:3]
+; GFX908-NEXT:    s_cbranch_scc0 .LBB0_2
+; GFX908-NEXT:  ; %bb.1: ; %st
+; GFX908-NEXT:    ;;#ASMSTART
+; GFX908-NEXT:    ;;#ASMEND
+; GFX908-NEXT:    s_endpgm
+; GFX908-NEXT:  .LBB0_2: ; %use
+; GFX908-NEXT:    s_nop 2
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a4
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a7
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a5
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a6
+; GFX908-NEXT:    v_accvgpr_write_b32 a4, 4
+; GFX908-NEXT:    v_accvgpr_write_b32 a8, 5
+; GFX908-NEXT:    v_accvgpr_write_b32 a9, 1
+; GFX908-NEXT:    v_accvgpr_write_b32 a10, 2
+; GFX908-NEXT:    v_accvgpr_write_b32 a11, 3
+; GFX908-NEXT:    ;;#ASMSTART
+; GFX908-NEXT:    ;;#ASMEND
+; GFX908-NEXT:    v_accvgpr_write_b32 a7, v4
+; GFX908-NEXT:    v_accvgpr_write_b32 a6, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a5, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a4, v1
+; GFX908-NEXT:    v_mov_b32_e32 v1, v0
+; GFX908-NEXT:    v_mov_b32_e32 v2, v0
+; GFX908-NEXT:    v_mov_b32_e32 v3, v0
+; GFX908-NEXT:    v_mov_b32_e32 v4, 0
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX908-NEXT:    s_waitcnt vmcnt(0)
+; GFX908-NEXT:    ;;#ASMSTART
+; GFX908-NEXT:    ;;#ASMEND
+; GFX908-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: max_12regs_13a_used:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    s_load_dwordx4 s[4:7], s[0:1], 0x2c
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX90A-NEXT:    s_load_dword s0, s[0:1], 0x24
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    s_cmp_lg_u32 s0, 0
+; GFX90A-NEXT:    s_load_dwordx4 s[8:11], s[4:5], 0x0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, s8
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, s9
+; GFX90A-NEXT:    v_accvgpr_write_b32 a2, s10
+; GFX90A-NEXT:    v_accvgpr_write_b32 a3, s11
+; GFX90A-NEXT:    s_nop 1
+; GFX90A-NEXT:    v_mfma_f32_4x4x1f32 a[0:3], v0, v0, a[0:3]
+; GFX90A-NEXT:    v_mfma_f32_4x4x1f32 a[4:7], v0, v0, a[0:3]
+; GFX90A-NEXT:    s_cbranch_scc0 .LBB0_2
+; GFX90A-NEXT:  ; %bb.1: ; %st
+; GFX90A-NEXT:    ;;#ASMSTART
+; GFX90A-NEXT:    ;;#ASMEND
+; GFX90A-NEXT:    s_endpgm
+; GFX90A-NEXT:  .LBB0_2: ; %use
+; GFX90A-NEXT:    s_nop 3
+; GFX90A-NEXT:    v_accvgpr_read_b32 v9, a7
+; GFX90A-NEXT:    v_accvgpr_read_b32 v8, a6
+; GFX90A-NEXT:    v_accvgpr_read_b32 v7, a5
+; GFX90A-NEXT:    v_accvgpr_read_b32 v6, a4
+; GFX90A-NEXT:    v_accvgpr_write_b32 a4, 4
+; GFX90A-NEXT:    v_accvgpr_write_b32 a8, 5
+; GFX90A-NEXT:    v_accvgpr_write_b32 a9, 1
+; GFX90A-NEXT:    v_accvgpr_write_b32 a10, 2
+; GFX90A-NEXT:    v_accvgpr_write_b32 a11, 3
+; GFX90A-NEXT:    ;;#ASMSTART
+; GFX90A-NEXT:    ;;#ASMEND
+; GFX90A-NEXT:    v_accvgpr_write_b32 a4, v6
+; GFX90A-NEXT:    v_mov_b32_e32 v4, 0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a5, v7
+; GFX90A-NEXT:    v_accvgpr_write_b32 a6, v8
+; GFX90A-NEXT:    v_accvgpr_write_b32 a7, v9
+; GFX90A-NEXT:    v_mov_b32_e32 v1, v0
+; GFX90A-NEXT:    v_mov_b32_e32 v2, v0
+; GFX90A-NEXT:    v_mov_b32_e32 v3, v0
+; GFX90A-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    ;;#ASMSTART
+; GFX90A-NEXT:    ;;#ASMEND
+; GFX90A-NEXT:    s_endpgm
 bb:
   %in.1 = load <4 x float>, ptr addrspace(1) %arg
   %mai.1 = tail call <4 x float> @llvm.amdgcn.mfma.f32.4x4x1f32(float 1.0, float 1.0, <4 x float> %in.1, i32 0, i32 0, i32 0)
@@ -28,16 +120,64 @@ st:
   call void asm sideeffect "", "a,a"(<4 x float> %mai.1, <4 x float> %mai.2)
   ret void
 }
-
-; GCN-LABEL: {{^}}max_10_vgprs_used_9a:
-; GCN-NOT: s_mov_b32 s{{[0-9]+}}, SCRATCH_RSRC_DWORD0
-; GCN-NOT: s_mov_b32 s{{[0-9]+}}, SCRATCH_RSRC_DWORD1
-; GCN:     v_accvgpr_read_b32 v[[VSPILL:[0-9]+]], a{{[0-9]+}}
-; GCN-NOT: buffer_store_dword
-; GCN-NOT: buffer_load_dword
-; GCN:     v_accvgpr_write_b32 a{{[0-9]+}}, v[[VSPILL]]
 ; GCN:     ScratchSize: 0
+
 define amdgpu_kernel void @max_10_vgprs_used_9a() #1 {
+; GFX908-LABEL: max_10_vgprs_used_9a:
+; GFX908:       ; %bb.0:
+; GFX908-NEXT:    ;;#ASMSTART
+; GFX908-NEXT:    ;;#ASMEND
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a0
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a3
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a1
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a2
+; GFX908-NEXT:    ;;#ASMSTART
+; GFX908-NEXT:    ;;#ASMEND
+; GFX908-NEXT:    ;;#ASMSTART
+; GFX908-NEXT:    ;;#ASMEND
+; GFX908-NEXT:    ;;#ASMSTART
+; GFX908-NEXT:    ;;#ASMEND
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a2
+; GFX908-NEXT:    v_accvgpr_read_b32 v4, a1
+; GFX908-NEXT:    v_accvgpr_write_b32 a4, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v1
+; GFX908-NEXT:    ;;#ASMSTART
+; GFX908-NEXT:    ;;#ASMEND
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, v4
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v5
+; GFX908-NEXT:    ;;#ASMSTART
+; GFX908-NEXT:    ;;#ASMEND
+; GFX908-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: max_10_vgprs_used_9a:
+; GFX90A:       ; %bb.0:
+; GFX90A-NEXT:    ;;#ASMSTART
+; GFX90A-NEXT:    ;;#ASMEND
+; GFX90A-NEXT:    v_accvgpr_read_b32 v0, a0
+; GFX90A-NEXT:    v_accvgpr_read_b32 v1, a1
+; GFX90A-NEXT:    v_accvgpr_read_b32 v2, a2
+; GFX90A-NEXT:    v_accvgpr_read_b32 v3, a3
+; GFX90A-NEXT:    ;;#ASMSTART
+; GFX90A-NEXT:    ;;#ASMEND
+; GFX90A-NEXT:    ;;#ASMSTART
+; GFX90A-NEXT:    ;;#ASMEND
+; GFX90A-NEXT:    ;;#ASMSTART
+; GFX90A-NEXT:    ;;#ASMEND
+; GFX90A-NEXT:    v_accvgpr_read_b32 v5, a3
+; GFX90A-NEXT:    v_accvgpr_read_b32 v4, a2
+; GFX90A-NEXT:    v_accvgpr_write_b32 a5, v3
+; GFX90A-NEXT:    v_accvgpr_write_b32 a4, v2
+; GFX90A-NEXT:    v_accvgpr_write_b32 a3, v1
+; GFX90A-NEXT:    v_accvgpr_write_b32 a2, v0
+; GFX90A-NEXT:    ;;#ASMSTART
+; GFX90A-NEXT:    ;;#ASMEND
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, v4
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v5
+; GFX90A-NEXT:    ;;#ASMSTART
+; GFX90A-NEXT:    ;;#ASMEND
+; GFX90A-NEXT:    s_endpgm
   %a1 = call <4 x i32> asm sideeffect "", "=a"()
   %a2 = call <4 x i32> asm sideeffect "", "=a"()
   %a3 = call i32 asm sideeffect "", "=a"()
@@ -46,17 +186,168 @@ define amdgpu_kernel void @max_10_vgprs_used_9a() #1 {
   call void asm sideeffect "", "a"(<2 x i32> %a4)
   ret void
 }
-
-; GCN-LABEL: {{^}}max_32regs_mfma32:
-; GCN-NOT: s_mov_b32 s{{[0-9]+}}, SCRATCH_RSRC_DWORD0
-; GCN-NOT: s_mov_b32 s{{[0-9]+}}, SCRATCH_RSRC_DWORD1
-; GCN-NOT: buffer_store_dword
-; GCN:     v_accvgpr_read_b32
-; GCN:     v_mfma_f32_32x32x1f32
-; GCN-NOT: buffer_load_dword
-; GCN:     v_accvgpr_write_b32
 ; GCN:     ScratchSize: 0
+
 define amdgpu_kernel void @max_32regs_mfma32(ptr addrspace(1) %arg) #3 {
+; GFX908-LABEL: max_32regs_mfma32:
+; GFX908:       ; %bb.0: ; %bb
+; GFX908-NEXT:    v_mov_b32_e32 v2, 0x40400000
+; GFX908-NEXT:    v_mov_b32_e32 v3, 0x40c00000
+; GFX908-NEXT:    v_mov_b32_e32 v4, 0x40e00000
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v2
+; GFX908-NEXT:    v_mov_b32_e32 v2, 0x40a00000
+; GFX908-NEXT:    v_accvgpr_write_b32 a5, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a6, v4
+; GFX908-NEXT:    v_accvgpr_write_b32 a4, v2
+; GFX908-NEXT:    v_mov_b32_e32 v2, 0x41000000
+; GFX908-NEXT:    v_mov_b32_e32 v3, 0x41100000
+; GFX908-NEXT:    v_mov_b32_e32 v4, 0x41200000
+; GFX908-NEXT:    v_accvgpr_write_b32 a7, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a8, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a9, v4
+; GFX908-NEXT:    v_mov_b32_e32 v2, 0x41300000
+; GFX908-NEXT:    v_mov_b32_e32 v3, 0x41400000
+; GFX908-NEXT:    v_mov_b32_e32 v4, 0x41500000
+; GFX908-NEXT:    v_accvgpr_write_b32 a10, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a11, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a12, v4
+; GFX908-NEXT:    v_mov_b32_e32 v2, 0x41600000
+; GFX908-NEXT:    v_mov_b32_e32 v3, 0x41700000
+; GFX908-NEXT:    v_mov_b32_e32 v4, 0x41800000
+; GFX908-NEXT:    v_accvgpr_write_b32 a13, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a14, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a15, v4
+; GFX908-NEXT:    v_mov_b32_e32 v2, 0x41880000
+; GFX908-NEXT:    v_mov_b32_e32 v3, 0x41900000
+; GFX908-NEXT:    v_mov_b32_e32 v4, 0x41980000
+; GFX908-NEXT:    v_accvgpr_write_b32 a16, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a17, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a18, v4
+; GFX908-NEXT:    v_mov_b32_e32 v2, 0x41a00000
+; GFX908-NEXT:    v_mov_b32_e32 v3, 0x41a80000
+; GFX908-NEXT:    v_mov_b32_e32 v4, 0x41b00000
+; GFX908-NEXT:    v_accvgpr_write_b32 a19, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a20, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a21, v4
+; GFX908-NEXT:    v_mov_b32_e32 v2, 0x41b80000
+; GFX908-NEXT:    v_mov_b32_e32 v3, 0x41c00000
+; GFX908-NEXT:    v_mov_b32_e32 v4, 0x41c80000
+; GFX908-NEXT:    v_accvgpr_write_b32 a22, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a23, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a24, v4
+; GFX908-NEXT:    v_mov_b32_e32 v2, 0x41d00000
+; GFX908-NEXT:    v_mov_b32_e32 v3, 0x41d80000
+; GFX908-NEXT:    v_mov_b32_e32 v4, 0x41e00000
+; GFX908-NEXT:    v_mov_b32_e32 v1, 1.0
+; GFX908-NEXT:    v_accvgpr_write_b32 a25, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a26, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a27, v4
+; GFX908-NEXT:    v_mov_b32_e32 v2, 0x41e80000
+; GFX908-NEXT:    v_mov_b32_e32 v3, 0x41f00000
+; GFX908-NEXT:    v_mov_b32_e32 v4, 0x41f80000
+; GFX908-NEXT:    ;;#ASMSTART
+; GFX908-NEXT:    ;;#ASMEND
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a0
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, 1.0
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, 2.0
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, 4.0
+; GFX908-NEXT:    v_accvgpr_write_b32 a28, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a29, v3
+; GFX908-NEXT:    v_accvgpr_write_b32 a30, v4
+; GFX908-NEXT:    v_accvgpr_write_b32 a31, 2.0
+; GFX908-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX908-NEXT:    v_mfma_f32_32x32x1f32 a[0:31], v1, v1, a[0:31]
+; GFX908-NEXT:    v_mov_b32_e32 v0, 0
+; GFX908-NEXT:    s_nop 7
+; GFX908-NEXT:    s_nop 5
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v5
+; GFX908-NEXT:    ;;#ASMSTART
+; GFX908-NEXT:    ;;#ASMEND
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a0
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    global_store_dword v0, v1, s[2:3]
+; GFX908-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: max_32regs_mfma32:
+; GFX90A:       ; %bb.0: ; %bb
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x40400000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a2, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x40a00000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a4, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x40c00000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a5, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x40e00000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a6, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41000000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a7, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41100000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a8, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41200000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a9, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41300000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a10, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41400000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a11, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41500000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a12, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41600000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a13, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41700000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a14, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41800000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a15, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41880000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a16, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41900000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a17, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41980000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a18, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41a00000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a19, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41a80000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a20, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41b00000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a21, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41b80000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a22, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41c00000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a23, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41c80000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a24, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41d00000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a25, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41d80000
+; GFX90A-NEXT:    v_mov_b32_e32 v1, 1.0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a26, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41e00000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a27, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41e80000
+; GFX90A-NEXT:    v_accvgpr_write_b32 a28, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41f00000
+; GFX90A-NEXT:    ;;#ASMSTART
+; GFX90A-NEXT:    ;;#ASMEND
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, 2.0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a29, v2
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 0x41f80000
+; GFX90A-NEXT:    v_accvgpr_read_b32 v3, a0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a0, 1.0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a3, 4.0
+; GFX90A-NEXT:    v_accvgpr_write_b32 a30, v2
+; GFX90A-NEXT:    v_accvgpr_mov_b32 a31, a1
+; GFX90A-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX90A-NEXT:    v_mov_b32_e32 v0, 0
+; GFX90A-NEXT:    v_mfma_f32_32x32x1f32 a[0:31], v1, v1, a[0:31]
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 7
+; GFX90A-NEXT:    s_nop 2
+; GFX90A-NEXT:    v_accvgpr_write_b32 a1, v3
+; GFX90A-NEXT:    ;;#ASMSTART
+; GFX90A-NEXT:    ;;#ASMEND
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    global_store_dword v0, a0, s[2:3]
+; GFX90A-NEXT:    s_endpgm
 bb:
   %v = call i32 asm sideeffect "", "=a"()
   br label %use
@@ -68,42 +359,110 @@ use:
   store float %elt1, ptr addrspace(1) %arg
   ret void
 }
+; GCN:     ScratchSize: 0
 
 ; Should spill agprs to memory for both gfx908 and gfx90a.
-; GCN-LABEL: {{^}}max_6regs_used_8a:
-; GCN: s_mov_b32 s{{[0-9]+}}, SCRATCH_RSRC_DWORD0
-; GCN: s_mov_b32 s{{[0-9]+}}, SCRATCH_RSRC_DWORD1
-
-; GFX908-DAG:  v_accvgpr_read_b32 v5, a0 ; Reload Reuse
-; GFX908-DAG:  buffer_store_dword v5, off, s[{{[0-9:]+}}], 0 ; 4-byte Folded Spill
-; GFX908-DAG:  v_accvgpr_read_b32 v5, a1 ; Reload Reuse
-; GFX908-DAG:  buffer_store_dword v5, off, s[{{[0-9:]+}}], 0 offset:4 ; 4-byte Folded Spill
-; GFX908-DAG:  v_accvgpr_read_b32 v5, a2 ; Reload Reuse
-; GFX908-DAG:  buffer_store_dword v5, off, s[{{[0-9:]+}}], 0 offset:8 ; 4-byte Folded Spill
-; GFX908-DAG:  v_accvgpr_read_b32 v5, a3 ; Reload Reuse
-; GFX908-DAG:  buffer_store_dword v5, off, s[{{[0-9:]+}}], 0 offset:12 ; 4-byte Folded Spill
-
-; GFX90A-DAG:  buffer_store_dword a0, off, s[{{[0-9:]+}}], 0 ; 4-byte Folded Spill
-; GFX90A-DAG:  buffer_store_dword a1, off, s[{{[0-9:]+}}], 0 offset:4 ; 4-byte Folded Spill
-; GFX90A-DAG:  buffer_store_dword a2, off, s[{{[0-9:]+}}], 0 offset:8 ; 4-byte Folded Spill
-; GFX90A-DAG:  buffer_store_dword a3, off, s[{{[0-9:]+}}], 0 offset:12 ; 4-byte Folded Spill
-
-; GCN:  v_mfma_f32_4x4x1f32 a[0:3], v{{[0-9]+}}, v{{[0-9]+}}, a[0:3]
-
-; GFX908-DAG:  buffer_load_dword v0, off, s[{{[0-9:]+}}], 0 ; 4-byte Folded Reload
-; GFX908-DAG:  buffer_load_dword v1, off, s[{{[0-9:]+}}], 0 offset:4 ; 4-byte Folded Reload
-; GFX908-DAG:  buffer_load_dword v2, off, s[{{[0-9:]+}}], 0 offset:8 ; 4-byte Folded Reload
-; GFX908-DAG:  buffer_load_dword v3, off, s[{{[0-9:]+}}], 0 offset:12 ; 4-byte Folded Reload
-; GFX908: global_store_dwordx4 v[{{[0-9:]+}}], v[0:3], off
-
-; GFX90A-DAG:  buffer_load_dword v2, off, s[4:7], 0 ; 4-byte Folded Reload
-; GFX90A-DAG:  buffer_load_dword v3, off, s[4:7], 0 offset:4 ; 4-byte Folded Reload
-; GFX90A-DAG:  buffer_load_dword v4, off, s[4:7], 0 offset:8 ; 4-byte Folded Reload
-; GFX90A-DAG:  buffer_load_dword v5, off, s[4:7], 0 offset:12 ; 4-byte Folded Reload
-; GFX90A:  global_store_dwordx4 v[0:1], v[2:5], off
-
-; GCN: ScratchSize: 20
 define amdgpu_kernel void @max_6regs_used_8a(ptr addrspace(1) %arg) #4 {
+; GFX908-LABEL: max_6regs_used_8a:
+; GFX908:       ; %bb.0:
+; GFX908-NEXT:    s_mov_b32 s4, SCRATCH_RSRC_DWORD0
+; GFX908-NEXT:    s_mov_b32 s5, SCRATCH_RSRC_DWORD1
+; GFX908-NEXT:    s_mov_b32 s6, -1
+; GFX908-NEXT:    s_mov_b32 s7, 0xe00000
+; GFX908-NEXT:    s_add_u32 s4, s4, s3
+; GFX908-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX908-NEXT:    ;;#ASMSTART
+; GFX908-NEXT:    ; def v1
+; GFX908-NEXT:    ;;#ASMEND
+; GFX908-NEXT:    v_lshlrev_b32_e32 v4, 4, v0
+; GFX908-NEXT:    ;;#ASMSTART
+; GFX908-NEXT:    ; def a[0:3]
+; GFX908-NEXT:    ;;#ASMEND
+; GFX908-NEXT:    s_addc_u32 s5, s5, 0
+; GFX908-NEXT:    v_accvgpr_write_b32 a4, v1
+; GFX908-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX908-NEXT:    global_load_dwordx4 v[0:3], v4, s[2:3]
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a0 ; Reload Reuse
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    buffer_store_dword v5, off, s[4:7], 0 ; 4-byte Folded Spill
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a1 ; Reload Reuse
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    buffer_store_dword v5, off, s[4:7], 0 offset:4 ; 4-byte Folded Spill
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a2 ; Reload Reuse
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    buffer_store_dword v5, off, s[4:7], 0 offset:8 ; 4-byte Folded Spill
+; GFX908-NEXT:    v_accvgpr_read_b32 v5, a3 ; Reload Reuse
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    buffer_store_dword v5, off, s[4:7], 0 offset:12 ; 4-byte Folded Spill
+; GFX908-NEXT:    s_waitcnt vmcnt(4)
+; GFX908-NEXT:    v_accvgpr_write_b32 a0, v0
+; GFX908-NEXT:    v_accvgpr_write_b32 a1, v1
+; GFX908-NEXT:    v_accvgpr_write_b32 a2, v2
+; GFX908-NEXT:    v_accvgpr_write_b32 a3, v3
+; GFX908-NEXT:    v_mov_b32_e32 v0, 1.0
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    v_mfma_f32_4x4x1f32 a[0:3], v0, v0, a[0:3]
+; GFX908-NEXT:    s_nop 3
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a0
+; GFX908-NEXT:    v_accvgpr_read_b32 v1, a1
+; GFX908-NEXT:    v_accvgpr_read_b32 v2, a2
+; GFX908-NEXT:    v_accvgpr_read_b32 v3, a3
+; GFX908-NEXT:    s_nop 1
+; GFX908-NEXT:    global_store_dwordx4 v4, v[0:3], s[2:3]
+; GFX908-NEXT:    buffer_load_dword v0, off, s[4:7], 0 ; 4-byte Folded Reload
+; GFX908-NEXT:    s_nop 0
+; GFX908-NEXT:    buffer_load_dword v1, off, s[4:7], 0 offset:4 ; 4-byte Folded Reload
+; GFX908-NEXT:    buffer_load_dword v2, off, s[4:7], 0 offset:8 ; 4-byte Folded Reload
+; GFX908-NEXT:    buffer_load_dword v3, off, s[4:7], 0 offset:12 ; 4-byte Folded Reload
+; GFX908-NEXT:    s_waitcnt vmcnt(0)
+; GFX908-NEXT:    global_store_dwordx4 v[0:1], v[0:3], off
+; GFX908-NEXT:    s_waitcnt vmcnt(0)
+; GFX908-NEXT:    v_accvgpr_read_b32 v0, a4
+; GFX908-NEXT:    ;;#ASMSTART
+; GFX908-NEXT:    ; use v0
+; GFX908-NEXT:    ;;#ASMEND
+; GFX908-NEXT:    s_endpgm
+;
+; GFX90A-LABEL: max_6regs_used_8a:
+; GFX90A:       ; %bb.0:
+; GFX90A-NEXT:    s_mov_b32 s4, SCRATCH_RSRC_DWORD0
+; GFX90A-NEXT:    s_mov_b32 s5, SCRATCH_RSRC_DWORD1
+; GFX90A-NEXT:    s_mov_b32 s6, -1
+; GFX90A-NEXT:    s_mov_b32 s7, 0xe00000
+; GFX90A-NEXT:    s_add_u32 s4, s4, s3
+; GFX90A-NEXT:    s_load_dwordx2 s[2:3], s[0:1], 0x24
+; GFX90A-NEXT:    s_addc_u32 s5, s5, 0
+; GFX90A-NEXT:    ;;#ASMSTART
+; GFX90A-NEXT:    ; def v1
+; GFX90A-NEXT:    ;;#ASMEND
+; GFX90A-NEXT:    ;;#ASMSTART
+; GFX90A-NEXT:    ; def a[0:3]
+; GFX90A-NEXT:    ;;#ASMEND
+; GFX90A-NEXT:    buffer_store_dword a0, off, s[4:7], 0 ; 4-byte Folded Spill
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    buffer_store_dword a1, off, s[4:7], 0 offset:4 ; 4-byte Folded Spill
+; GFX90A-NEXT:    buffer_store_dword a2, off, s[4:7], 0 offset:8 ; 4-byte Folded Spill
+; GFX90A-NEXT:    buffer_store_dword a3, off, s[4:7], 0 offset:12 ; 4-byte Folded Spill
+; GFX90A-NEXT:    v_lshlrev_b32_e32 v0, 4, v0
+; GFX90A-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX90A-NEXT:    global_load_dwordx4 a[0:3], v0, s[2:3]
+; GFX90A-NEXT:    v_mov_b32_e32 v2, 1.0
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    s_nop 0
+; GFX90A-NEXT:    v_mfma_f32_4x4x1f32 a[0:3], v2, v2, a[0:3]
+; GFX90A-NEXT:    s_nop 4
+; GFX90A-NEXT:    global_store_dwordx4 v0, a[0:3], s[2:3]
+; GFX90A-NEXT:    buffer_load_dword v2, off, s[4:7], 0 ; 4-byte Folded Reload
+; GFX90A-NEXT:    buffer_load_dword v3, off, s[4:7], 0 offset:4 ; 4-byte Folded Reload
+; GFX90A-NEXT:    buffer_load_dword v4, off, s[4:7], 0 offset:8 ; 4-byte Folded Reload
+; GFX90A-NEXT:    buffer_load_dword v5, off, s[4:7], 0 offset:12 ; 4-byte Folded Reload
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    global_store_dwordx4 v[0:1], v[2:5], off
+; GFX90A-NEXT:    s_waitcnt vmcnt(0)
+; GFX90A-NEXT:    ;;#ASMSTART
+; GFX90A-NEXT:    ; use v1
+; GFX90A-NEXT:    ;;#ASMEND
+; GFX90A-NEXT:    s_endpgm
   %tid = call i32 @llvm.amdgcn.workitem.id.x()
   %v0 = call float asm sideeffect "; def $0", "=v"()
   %a4 = call <4 x float> asm sideeffect "; def $0", "=a"()
@@ -115,6 +474,7 @@ define amdgpu_kernel void @max_6regs_used_8a(ptr addrspace(1) %arg) #4 {
   call void asm sideeffect "; use $0", "v"(float %v0);
   ret void
 }
+; GCN: ScratchSize: 20
 
 declare i32 @llvm.amdgcn.workitem.id.x()
 declare <16 x float> @llvm.amdgcn.mfma.f32.16x16x1f32(float, float, <16 x float>, i32, i32, i32)
@@ -125,3 +485,5 @@ attributes #1 = { nounwind "amdgpu-num-vgpr"="10" "amdgpu-no-dispatch-id" "amdgp
 attributes #2 = { nounwind "amdgpu-num-vgpr"="12" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" }
 attributes #3 = { nounwind "amdgpu-num-vgpr"="32" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" }
 attributes #4 = { nounwind "amdgpu-num-vgpr"="6" "amdgpu-no-dispatch-id" "amdgpu-no-dispatch-ptr" "amdgpu-no-implicitarg-ptr" "amdgpu-no-lds-kernel-id" "amdgpu-no-queue-ptr" "amdgpu-no-workgroup-id-x" "amdgpu-no-workgroup-id-y" "amdgpu-no-workgroup-id-z" "amdgpu-no-workitem-id-y" "amdgpu-no-workitem-id-z" }
+;; NOTE: These prefixes are unused and the list is autogenerated. Do not add tests below this line:
+; GCN: {{.*}}
diff --git a/llvm/test/CodeGen/AMDGPU/structurize-hoist.ll b/llvm/test/CodeGen/AMDGPU/structurize-hoist.ll
new file mode 100644
index 0000000..42436a1
--- /dev/null
+++ b/llvm/test/CodeGen/AMDGPU/structurize-hoist.ll
@@ -0,0 +1,180 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -mtriple=amdgcn -mcpu=gfx900 < %s | FileCheck -check-prefix=GFX900 %s
+
+
+%pair = type { i32, i32 }
+
+define void @test_extractvalue_then_else(ptr %ptr, i1 %cond) {
+; GFX900-LABEL: test_extractvalue_then_else:
+; GFX900:       ; %bb.0: ; %if
+; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX900-NEXT:    flat_load_dword v3, v[0:1]
+; GFX900-NEXT:    v_and_b32_e32 v2, 1, v2
+; GFX900-NEXT:    v_cmp_ne_u32_e32 vcc, 1, v2
+; GFX900-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX900-NEXT:    s_xor_b64 s[4:5], exec, s[4:5]
+; GFX900-NEXT:    s_cbranch_execz .LBB0_2
+; GFX900-NEXT:  ; %bb.1: ; %else
+; GFX900-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX900-NEXT:    v_add_u32_e32 v3, 1, v3
+; GFX900-NEXT:  .LBB0_2: ; %Flow
+; GFX900-NEXT:    s_andn2_saveexec_b64 s[4:5], s[4:5]
+; GFX900-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX900-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX900-NEXT:    flat_store_dword v[0:1], v3
+; GFX900-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX900-NEXT:    s_setpc_b64 s[30:31]
+if:
+  %load_then = load %pair, ptr %ptr
+  br i1 %cond, label %then, label %else
+
+then:
+  %a_then = extractvalue %pair %load_then, 0
+  br label %merge
+
+else:
+  %a_else = extractvalue %pair %load_then, 0
+  %sum_else = add i32 %a_else, 1
+  br label %merge
+
+merge:
+  %phi = phi i32  [ %a_then, %then ], [ %sum_else, %else ]
+  store i32 %phi, ptr  %ptr
+  ret void
+}
+
+define void @test_extractvalue_else_then(ptr %ptr, i1 %cond) {
+; GFX900-LABEL: test_extractvalue_else_then:
+; GFX900:       ; %bb.0: ; %if
+; GFX900-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
+; GFX900-NEXT:    flat_load_dword v3, v[0:1]
+; GFX900-NEXT:    v_and_b32_e32 v2, 1, v2
+; GFX900-NEXT:    v_cmp_ne_u32_e32 vcc, 1, v2
+; GFX900-NEXT:    s_and_saveexec_b64 s[4:5], vcc
+; GFX900-NEXT:    s_xor_b64 s[4:5], exec, s[4:5]
+; GFX900-NEXT:    s_andn2_saveexec_b64 s[4:5], s[4:5]
+; GFX900-NEXT:    s_cbranch_execz .LBB1_2
+; GFX900-NEXT:  ; %bb.1: ; %else
+; GFX900-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX900-NEXT:    v_add_u32_e32 v3, 1, v3
+; GFX900-NEXT:  .LBB1_2: ; %merge
+; GFX900-NEXT:    s_or_b64 exec, exec, s[4:5]
+; GFX900-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX900-NEXT:    flat_store_dword v[0:1], v3
+; GFX900-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX900-NEXT:    s_setpc_b64 s[30:31]
+if:
+  %load_then = load %pair, ptr %ptr
+  br i1 %cond, label %else, label %then
+
+else:
+  %a_else = extractvalue %pair %load_then, 0
+  %sum_else = add i32 %a_else, 1
+  br label %merge
+
+then:
+  %a_then = extractvalue %pair %load_then, 0
+  br label %merge
+
+merge:
+  %phi = phi i32  [ %a_then, %then ], [ %sum_else, %else ]
+  store i32 %phi, ptr  %ptr
+  ret void
+}
+
+define amdgpu_kernel void @test_loop_with_if( ptr %ptr, i1 %cond) #0 {
+; GFX900-LABEL: test_loop_with_if:
+; GFX900:       ; %bb.0: ; %entry
+; GFX900-NEXT:    s_load_dword s2, s[4:5], 0x2c
+; GFX900-NEXT:    s_load_dwordx2 s[0:1], s[4:5], 0x24
+; GFX900-NEXT:    v_mov_b32_e32 v5, 0
+; GFX900-NEXT:    s_mov_b64 s[4:5], 0
+; GFX900-NEXT:    s_movk_i32 s10, 0xfe
+; GFX900-NEXT:    s_waitcnt lgkmcnt(0)
+; GFX900-NEXT:    s_bitcmp1_b32 s2, 0
+; GFX900-NEXT:    s_cselect_b64 s[2:3], -1, 0
+; GFX900-NEXT:    v_cndmask_b32_e64 v3, 0, 1, s[2:3]
+; GFX900-NEXT:    v_mov_b32_e32 v2, s1
+; GFX900-NEXT:    s_xor_b64 s[2:3], s[2:3], -1
+; GFX900-NEXT:    v_mov_b32_e32 v1, s0
+; GFX900-NEXT:    v_cmp_ne_u32_e64 s[0:1], 1, v3
+; GFX900-NEXT:    s_branch .LBB2_2
+; GFX900-NEXT:  .LBB2_1: ; %latch
+; GFX900-NEXT:    ; in Loop: Header=BB2_2 Depth=1
+; GFX900-NEXT:    s_or_b64 exec, exec, s[8:9]
+; GFX900-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX900-NEXT:    v_add_u32_e32 v5, 20, v3
+; GFX900-NEXT:    v_cmp_lt_i32_e32 vcc, s10, v5
+; GFX900-NEXT:    s_or_b64 s[4:5], vcc, s[4:5]
+; GFX900-NEXT:    flat_store_dword v[1:2], v3
+; GFX900-NEXT:    s_andn2_b64 exec, exec, s[4:5]
+; GFX900-NEXT:    s_cbranch_execz .LBB2_8
+; GFX900-NEXT:  .LBB2_2: ; %loop
+; GFX900-NEXT:    ; =>This Inner Loop Header: Depth=1
+; GFX900-NEXT:    flat_load_dwordx2 v[3:4], v[1:2]
+; GFX900-NEXT:    s_and_b64 vcc, exec, s[0:1]
+; GFX900-NEXT:    s_mov_b64 s[8:9], s[2:3]
+; GFX900-NEXT:    s_mov_b64 s[6:7], 0
+; GFX900-NEXT:    s_cbranch_vccnz .LBB2_4
+; GFX900-NEXT:  ; %bb.3: ; %if
+; GFX900-NEXT:    ; in Loop: Header=BB2_2 Depth=1
+; GFX900-NEXT:    v_cmp_gt_i32_e32 vcc, 11, v5
+; GFX900-NEXT:    s_andn2_b64 s[8:9], s[2:3], exec
+; GFX900-NEXT:    s_and_b64 s[12:13], vcc, exec
+; GFX900-NEXT:    s_mov_b64 s[6:7], -1
+; GFX900-NEXT:    s_or_b64 s[8:9], s[8:9], s[12:13]
+; GFX900-NEXT:  .LBB2_4: ; %Flow
+; GFX900-NEXT:    ; in Loop: Header=BB2_2 Depth=1
+; GFX900-NEXT:    s_and_saveexec_b64 s[12:13], s[8:9]
+; GFX900-NEXT:    s_xor_b64 s[8:9], exec, s[12:13]
+; GFX900-NEXT:    s_cbranch_execz .LBB2_6
+; GFX900-NEXT:  ; %bb.5: ; %else
+; GFX900-NEXT:    ; in Loop: Header=BB2_2 Depth=1
+; GFX900-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
+; GFX900-NEXT:    v_add_u32_e32 v3, v3, v4
+; GFX900-NEXT:    s_andn2_b64 s[6:7], s[6:7], exec
+; GFX900-NEXT:  .LBB2_6: ; %Flow1
+; GFX900-NEXT:    ; in Loop: Header=BB2_2 Depth=1
+; GFX900-NEXT:    s_or_b64 exec, exec, s[8:9]
+; GFX900-NEXT:    s_and_saveexec_b64 s[8:9], s[6:7]
+; GFX900-NEXT:    s_cbranch_execz .LBB2_1
+; GFX900-NEXT:  ; %bb.7: ; %then
+; GFX900-NEXT:    ; in Loop: Header=BB2_2 Depth=1
+; GFX900-NEXT:    flat_store_dword v[1:2], v0
+; GFX900-NEXT:    s_branch .LBB2_1
+; GFX900-NEXT:  .LBB2_8: ; %end
+; GFX900-NEXT:    s_endpgm
+entry:
+  %a = tail call i32 @llvm.amdgcn.workitem.id.x()
+  br label %loop
+
+loop:
+  %entry_phi = phi i32 [ 0, %entry ], [ %a15, %latch ]
+  %load = load %pair, ptr %ptr
+  br i1 %cond, label %if, label %else
+
+if:
+  %cmp = icmp sgt i32 %entry_phi, 10
+  br i1 %cmp, label %then, label %else
+
+then:
+  %a_then = extractvalue %pair %load, 0
+  store i32 %a, ptr %ptr, align 4
+  br label %latch
+
+else:
+  %a2 = extractvalue %pair %load, 1
+  %y = extractvalue %pair %load, 0
+  %a_else = add i32 %y, %a2
+  br label %latch
+
+latch:
+  %a_test = phi i32 [ %a_then, %then ], [ %a_else, %else ]
+  store i32 %a_test, ptr  %ptr
+  %a15 = add nsw i32 %a_test, 20
+  %a16 = icmp slt i32  %a15, 255
+  br i1 %a16, label %loop, label %end
+
+end:
+  ret void
+}
diff --git a/llvm/test/CodeGen/AMDGPU/tuple-allocation-failure.ll b/llvm/test/CodeGen/AMDGPU/tuple-allocation-failure.ll
index d23e314..f6c357d 100644
--- a/llvm/test/CodeGen/AMDGPU/tuple-allocation-failure.ll
+++ b/llvm/test/CodeGen/AMDGPU/tuple-allocation-failure.ll
@@ -70,12 +70,12 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS1-NEXT:    s_xor_b64 s[4:5], s[4:5], -1
 ; GLOBALNESS1-NEXT:    s_mov_b64 s[38:39], s[8:9]
 ; GLOBALNESS1-NEXT:    v_cmp_ne_u32_e64 s[8:9], 1, v1
-; GLOBALNESS1-NEXT:    ; implicit-def: $vgpr59 : SGPR spill to VGPR lane
+; GLOBALNESS1-NEXT:    ; implicit-def: $vgpr57 : SGPR spill to VGPR lane
 ; GLOBALNESS1-NEXT:    v_cmp_ne_u32_e64 s[66:67], 1, v0
 ; GLOBALNESS1-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[4:5]
-; GLOBALNESS1-NEXT:    v_writelane_b32 v59, s8, 0
+; GLOBALNESS1-NEXT:    v_writelane_b32 v57, s8, 0
 ; GLOBALNESS1-NEXT:    v_cmp_ne_u32_e64 s[68:69], 1, v0
-; GLOBALNESS1-NEXT:    v_writelane_b32 v59, s9, 1
+; GLOBALNESS1-NEXT:    v_writelane_b32 v57, s9, 1
 ; GLOBALNESS1-NEXT:    v_cmp_ne_u32_e64 s[70:71], 1, v3
 ; GLOBALNESS1-NEXT:    v_mov_b32_e32 v46, 0x80
 ; GLOBALNESS1-NEXT:    s_mov_b32 s82, s16
@@ -84,7 +84,7 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS1-NEXT:    s_mov_b64 s[34:35], s[10:11]
 ; GLOBALNESS1-NEXT:    v_mov_b32_e32 v47, 0
 ; GLOBALNESS1-NEXT:    s_mov_b32 s32, 0
-; GLOBALNESS1-NEXT:    ; implicit-def: $vgpr56_vgpr57
+; GLOBALNESS1-NEXT:    ; implicit-def: $vgpr58_vgpr59
 ; GLOBALNESS1-NEXT:    s_waitcnt vmcnt(0)
 ; GLOBALNESS1-NEXT:    v_cmp_gt_i32_e32 vcc, 0, v2
 ; GLOBALNESS1-NEXT:    v_cndmask_b32_e64 v0, 0, 1, vcc
@@ -93,24 +93,24 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS1-NEXT:    v_cmp_eq_u32_e32 vcc, 1, v2
 ; GLOBALNESS1-NEXT:    v_cmp_ne_u32_e64 s[4:5], 1, v0
 ; GLOBALNESS1-NEXT:    v_cndmask_b32_e64 v3, 0, 1, vcc
-; GLOBALNESS1-NEXT:    v_writelane_b32 v59, s4, 2
+; GLOBALNESS1-NEXT:    v_writelane_b32 v57, s4, 2
 ; GLOBALNESS1-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v2
-; GLOBALNESS1-NEXT:    v_writelane_b32 v59, s5, 3
+; GLOBALNESS1-NEXT:    v_writelane_b32 v57, s5, 3
 ; GLOBALNESS1-NEXT:    v_cmp_ne_u32_e64 s[4:5], 1, v3
 ; GLOBALNESS1-NEXT:    v_cndmask_b32_e64 v2, 0, 1, vcc
-; GLOBALNESS1-NEXT:    v_writelane_b32 v59, s4, 4
-; GLOBALNESS1-NEXT:    v_writelane_b32 v59, s5, 5
+; GLOBALNESS1-NEXT:    v_writelane_b32 v57, s4, 4
+; GLOBALNESS1-NEXT:    v_writelane_b32 v57, s5, 5
 ; GLOBALNESS1-NEXT:    v_cmp_ne_u32_e64 s[4:5], 1, v2
-; GLOBALNESS1-NEXT:    v_writelane_b32 v59, s4, 6
-; GLOBALNESS1-NEXT:    v_writelane_b32 v59, s5, 7
+; GLOBALNESS1-NEXT:    v_writelane_b32 v57, s4, 6
+; GLOBALNESS1-NEXT:    v_writelane_b32 v57, s5, 7
 ; GLOBALNESS1-NEXT:    v_cmp_ne_u32_e64 s[80:81], 1, v1
-; GLOBALNESS1-NEXT:    v_writelane_b32 v59, s70, 8
-; GLOBALNESS1-NEXT:    v_writelane_b32 v59, s71, 9
+; GLOBALNESS1-NEXT:    v_writelane_b32 v57, s70, 8
+; GLOBALNESS1-NEXT:    v_writelane_b32 v57, s71, 9
 ; GLOBALNESS1-NEXT:    s_branch .LBB1_4
 ; GLOBALNESS1-NEXT:  .LBB1_1: ; %bb70.i
 ; GLOBALNESS1-NEXT:    ; in Loop: Header=BB1_4 Depth=1
-; GLOBALNESS1-NEXT:    v_readlane_b32 s6, v59, 6
-; GLOBALNESS1-NEXT:    v_readlane_b32 s7, v59, 7
+; GLOBALNESS1-NEXT:    v_readlane_b32 s6, v57, 6
+; GLOBALNESS1-NEXT:    v_readlane_b32 s7, v57, 7
 ; GLOBALNESS1-NEXT:    s_and_b64 vcc, exec, s[6:7]
 ; GLOBALNESS1-NEXT:    s_cbranch_vccz .LBB1_28
 ; GLOBALNESS1-NEXT:  .LBB1_2: ; %Flow15
@@ -120,7 +120,7 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS1-NEXT:  .LBB1_3: ; %Flow28
 ; GLOBALNESS1-NEXT:    ; in Loop: Header=BB1_4 Depth=1
 ; GLOBALNESS1-NEXT:    s_and_b64 vcc, exec, s[6:7]
-; GLOBALNESS1-NEXT:    v_pk_mov_b32 v[56:57], v[0:1], v[0:1] op_sel:[0,1]
+; GLOBALNESS1-NEXT:    v_pk_mov_b32 v[58:59], v[0:1], v[0:1] op_sel:[0,1]
 ; GLOBALNESS1-NEXT:    s_cbranch_vccnz .LBB1_29
 ; GLOBALNESS1-NEXT:  .LBB1_4: ; %bb5
 ; GLOBALNESS1-NEXT:    ; =>This Loop Header: Depth=1
@@ -128,7 +128,7 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS1-NEXT:    flat_load_dword v40, v[46:47]
 ; GLOBALNESS1-NEXT:    s_add_u32 s8, s38, 40
 ; GLOBALNESS1-NEXT:    buffer_store_dword v42, off, s[0:3], 0
-; GLOBALNESS1-NEXT:    flat_load_dword v58, v[46:47]
+; GLOBALNESS1-NEXT:    flat_load_dword v56, v[46:47]
 ; GLOBALNESS1-NEXT:    s_addc_u32 s9, s39, 0
 ; GLOBALNESS1-NEXT:    s_getpc_b64 s[4:5]
 ; GLOBALNESS1-NEXT:    s_add_u32 s4, s4, wobble@gotpcrel32@lo+4
@@ -186,10 +186,10 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS1-NEXT:  ; %bb.11: ; %bb33.i
 ; GLOBALNESS1-NEXT:    ; in Loop: Header=BB1_4 Depth=1
 ; GLOBALNESS1-NEXT:    global_load_dwordx2 v[0:1], v[44:45], off
-; GLOBALNESS1-NEXT:    v_writelane_b32 v59, s8, 10
-; GLOBALNESS1-NEXT:    v_writelane_b32 v59, s9, 11
-; GLOBALNESS1-NEXT:    v_readlane_b32 s4, v59, 2
-; GLOBALNESS1-NEXT:    v_readlane_b32 s5, v59, 3
+; GLOBALNESS1-NEXT:    v_writelane_b32 v57, s8, 10
+; GLOBALNESS1-NEXT:    v_writelane_b32 v57, s9, 11
+; GLOBALNESS1-NEXT:    v_readlane_b32 s4, v57, 2
+; GLOBALNESS1-NEXT:    v_readlane_b32 s5, v57, 3
 ; GLOBALNESS1-NEXT:    s_and_b64 vcc, exec, s[4:5]
 ; GLOBALNESS1-NEXT:    s_cbranch_vccnz .LBB1_13
 ; GLOBALNESS1-NEXT:  ; %bb.12: ; %bb39.i
@@ -198,7 +198,7 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS1-NEXT:    global_store_dwordx2 v[44:45], v[42:43], off
 ; GLOBALNESS1-NEXT:  .LBB1_13: ; %bb44.lr.ph.i
 ; GLOBALNESS1-NEXT:    ; in Loop: Header=BB1_4 Depth=1
-; GLOBALNESS1-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v58
+; GLOBALNESS1-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v56
 ; GLOBALNESS1-NEXT:    v_cndmask_b32_e32 v2, 0, v40, vcc
 ; GLOBALNESS1-NEXT:    s_waitcnt vmcnt(0)
 ; GLOBALNESS1-NEXT:    v_cmp_nlt_f64_e32 vcc, 0, v[0:1]
@@ -228,8 +228,8 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS1-NEXT:    s_cbranch_vccnz .LBB1_21
 ; GLOBALNESS1-NEXT:  ; %bb.19: ; %bb3.i.i
 ; GLOBALNESS1-NEXT:    ; in Loop: Header=BB1_16 Depth=2
-; GLOBALNESS1-NEXT:    v_readlane_b32 s4, v59, 0
-; GLOBALNESS1-NEXT:    v_readlane_b32 s5, v59, 1
+; GLOBALNESS1-NEXT:    v_readlane_b32 s4, v57, 0
+; GLOBALNESS1-NEXT:    v_readlane_b32 s5, v57, 1
 ; GLOBALNESS1-NEXT:    s_and_b64 vcc, exec, s[4:5]
 ; GLOBALNESS1-NEXT:    s_cbranch_vccnz .LBB1_21
 ; GLOBALNESS1-NEXT:  ; %bb.20: ; %bb6.i.i
@@ -265,7 +265,7 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS1-NEXT:    s_mov_b32 s13, s83
 ; GLOBALNESS1-NEXT:    s_mov_b32 s14, s82
 ; GLOBALNESS1-NEXT:    v_mov_b32_e32 v31, v41
-; GLOBALNESS1-NEXT:    global_store_dwordx2 v[44:45], v[56:57], off
+; GLOBALNESS1-NEXT:    global_store_dwordx2 v[44:45], v[58:59], off
 ; GLOBALNESS1-NEXT:    s_swappc_b64 s[30:31], s[54:55]
 ; GLOBALNESS1-NEXT:    s_and_saveexec_b64 s[4:5], s[96:97]
 ; GLOBALNESS1-NEXT:    s_cbranch_execz .LBB1_14
@@ -277,13 +277,13 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS1-NEXT:  .LBB1_24: ; %Flow23
 ; GLOBALNESS1-NEXT:    ; in Loop: Header=BB1_4 Depth=1
 ; GLOBALNESS1-NEXT:    s_load_dwordx4 s[4:7], s[38:39], 0x0
-; GLOBALNESS1-NEXT:    v_readlane_b32 s70, v59, 8
-; GLOBALNESS1-NEXT:    v_readlane_b32 s8, v59, 10
+; GLOBALNESS1-NEXT:    v_readlane_b32 s70, v57, 8
+; GLOBALNESS1-NEXT:    v_readlane_b32 s8, v57, 10
 ; GLOBALNESS1-NEXT:    v_pk_mov_b32 v[0:1], 0, 0
-; GLOBALNESS1-NEXT:    v_readlane_b32 s71, v59, 9
+; GLOBALNESS1-NEXT:    v_readlane_b32 s71, v57, 9
 ; GLOBALNESS1-NEXT:    s_waitcnt lgkmcnt(0)
 ; GLOBALNESS1-NEXT:    s_mov_b32 s55, s7
-; GLOBALNESS1-NEXT:    v_readlane_b32 s9, v59, 11
+; GLOBALNESS1-NEXT:    v_readlane_b32 s9, v57, 11
 ; GLOBALNESS1-NEXT:  .LBB1_25: ; %Flow24
 ; GLOBALNESS1-NEXT:    ; in Loop: Header=BB1_4 Depth=1
 ; GLOBALNESS1-NEXT:    s_or_b64 exec, exec, s[52:53]
@@ -291,8 +291,8 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS1-NEXT:    s_cbranch_execz .LBB1_2
 ; GLOBALNESS1-NEXT:  ; %bb.26: ; %bb67.i
 ; GLOBALNESS1-NEXT:    ; in Loop: Header=BB1_4 Depth=1
-; GLOBALNESS1-NEXT:    v_readlane_b32 s6, v59, 4
-; GLOBALNESS1-NEXT:    v_readlane_b32 s7, v59, 5
+; GLOBALNESS1-NEXT:    v_readlane_b32 s6, v57, 4
+; GLOBALNESS1-NEXT:    v_readlane_b32 s7, v57, 5
 ; GLOBALNESS1-NEXT:    s_and_b64 vcc, exec, s[6:7]
 ; GLOBALNESS1-NEXT:    s_cbranch_vccnz .LBB1_1
 ; GLOBALNESS1-NEXT:  ; %bb.27: ; %bb69.i
@@ -384,12 +384,12 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS0-NEXT:    s_xor_b64 s[4:5], s[4:5], -1
 ; GLOBALNESS0-NEXT:    s_mov_b64 s[38:39], s[8:9]
 ; GLOBALNESS0-NEXT:    v_cmp_ne_u32_e64 s[8:9], 1, v1
-; GLOBALNESS0-NEXT:    ; implicit-def: $vgpr59 : SGPR spill to VGPR lane
+; GLOBALNESS0-NEXT:    ; implicit-def: $vgpr57 : SGPR spill to VGPR lane
 ; GLOBALNESS0-NEXT:    v_cmp_ne_u32_e64 s[66:67], 1, v0
 ; GLOBALNESS0-NEXT:    v_cndmask_b32_e64 v0, 0, 1, s[4:5]
-; GLOBALNESS0-NEXT:    v_writelane_b32 v59, s8, 0
+; GLOBALNESS0-NEXT:    v_writelane_b32 v57, s8, 0
 ; GLOBALNESS0-NEXT:    v_cmp_ne_u32_e64 s[68:69], 1, v0
-; GLOBALNESS0-NEXT:    v_writelane_b32 v59, s9, 1
+; GLOBALNESS0-NEXT:    v_writelane_b32 v57, s9, 1
 ; GLOBALNESS0-NEXT:    v_cmp_ne_u32_e64 s[84:85], 1, v3
 ; GLOBALNESS0-NEXT:    v_mov_b32_e32 v46, 0x80
 ; GLOBALNESS0-NEXT:    s_mov_b32 s70, s16
@@ -398,7 +398,7 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS0-NEXT:    s_mov_b64 s[34:35], s[10:11]
 ; GLOBALNESS0-NEXT:    v_mov_b32_e32 v47, 0
 ; GLOBALNESS0-NEXT:    s_mov_b32 s32, 0
-; GLOBALNESS0-NEXT:    ; implicit-def: $vgpr56_vgpr57
+; GLOBALNESS0-NEXT:    ; implicit-def: $vgpr58_vgpr59
 ; GLOBALNESS0-NEXT:    s_waitcnt vmcnt(0)
 ; GLOBALNESS0-NEXT:    v_cmp_gt_i32_e32 vcc, 0, v2
 ; GLOBALNESS0-NEXT:    v_cndmask_b32_e64 v0, 0, 1, vcc
@@ -407,24 +407,24 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS0-NEXT:    v_cmp_eq_u32_e32 vcc, 1, v2
 ; GLOBALNESS0-NEXT:    v_cmp_ne_u32_e64 s[4:5], 1, v0
 ; GLOBALNESS0-NEXT:    v_cndmask_b32_e64 v3, 0, 1, vcc
-; GLOBALNESS0-NEXT:    v_writelane_b32 v59, s4, 2
+; GLOBALNESS0-NEXT:    v_writelane_b32 v57, s4, 2
 ; GLOBALNESS0-NEXT:    v_cmp_eq_u32_e32 vcc, 0, v2
-; GLOBALNESS0-NEXT:    v_writelane_b32 v59, s5, 3
+; GLOBALNESS0-NEXT:    v_writelane_b32 v57, s5, 3
 ; GLOBALNESS0-NEXT:    v_cmp_ne_u32_e64 s[4:5], 1, v3
 ; GLOBALNESS0-NEXT:    v_cndmask_b32_e64 v2, 0, 1, vcc
-; GLOBALNESS0-NEXT:    v_writelane_b32 v59, s4, 4
-; GLOBALNESS0-NEXT:    v_writelane_b32 v59, s5, 5
+; GLOBALNESS0-NEXT:    v_writelane_b32 v57, s4, 4
+; GLOBALNESS0-NEXT:    v_writelane_b32 v57, s5, 5
 ; GLOBALNESS0-NEXT:    v_cmp_ne_u32_e64 s[4:5], 1, v2
-; GLOBALNESS0-NEXT:    v_writelane_b32 v59, s4, 6
-; GLOBALNESS0-NEXT:    v_writelane_b32 v59, s5, 7
+; GLOBALNESS0-NEXT:    v_writelane_b32 v57, s4, 6
+; GLOBALNESS0-NEXT:    v_writelane_b32 v57, s5, 7
 ; GLOBALNESS0-NEXT:    v_cmp_ne_u32_e64 s[80:81], 1, v1
-; GLOBALNESS0-NEXT:    v_writelane_b32 v59, s84, 8
-; GLOBALNESS0-NEXT:    v_writelane_b32 v59, s85, 9
+; GLOBALNESS0-NEXT:    v_writelane_b32 v57, s84, 8
+; GLOBALNESS0-NEXT:    v_writelane_b32 v57, s85, 9
 ; GLOBALNESS0-NEXT:    s_branch .LBB1_4
 ; GLOBALNESS0-NEXT:  .LBB1_1: ; %bb70.i
 ; GLOBALNESS0-NEXT:    ; in Loop: Header=BB1_4 Depth=1
-; GLOBALNESS0-NEXT:    v_readlane_b32 s6, v59, 6
-; GLOBALNESS0-NEXT:    v_readlane_b32 s7, v59, 7
+; GLOBALNESS0-NEXT:    v_readlane_b32 s6, v57, 6
+; GLOBALNESS0-NEXT:    v_readlane_b32 s7, v57, 7
 ; GLOBALNESS0-NEXT:    s_and_b64 vcc, exec, s[6:7]
 ; GLOBALNESS0-NEXT:    s_cbranch_vccz .LBB1_28
 ; GLOBALNESS0-NEXT:  .LBB1_2: ; %Flow15
@@ -434,7 +434,7 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS0-NEXT:  .LBB1_3: ; %Flow28
 ; GLOBALNESS0-NEXT:    ; in Loop: Header=BB1_4 Depth=1
 ; GLOBALNESS0-NEXT:    s_and_b64 vcc, exec, s[6:7]
-; GLOBALNESS0-NEXT:    v_pk_mov_b32 v[56:57], v[0:1], v[0:1] op_sel:[0,1]
+; GLOBALNESS0-NEXT:    v_pk_mov_b32 v[58:59], v[0:1], v[0:1] op_sel:[0,1]
 ; GLOBALNESS0-NEXT:    s_cbranch_vccnz .LBB1_29
 ; GLOBALNESS0-NEXT:  .LBB1_4: ; %bb5
 ; GLOBALNESS0-NEXT:    ; =>This Loop Header: Depth=1
@@ -442,7 +442,7 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS0-NEXT:    flat_load_dword v40, v[46:47]
 ; GLOBALNESS0-NEXT:    s_add_u32 s8, s38, 40
 ; GLOBALNESS0-NEXT:    buffer_store_dword v42, off, s[0:3], 0
-; GLOBALNESS0-NEXT:    flat_load_dword v58, v[46:47]
+; GLOBALNESS0-NEXT:    flat_load_dword v56, v[46:47]
 ; GLOBALNESS0-NEXT:    s_addc_u32 s9, s39, 0
 ; GLOBALNESS0-NEXT:    s_getpc_b64 s[4:5]
 ; GLOBALNESS0-NEXT:    s_add_u32 s4, s4, wobble@gotpcrel32@lo+4
@@ -500,10 +500,10 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS0-NEXT:  ; %bb.11: ; %bb33.i
 ; GLOBALNESS0-NEXT:    ; in Loop: Header=BB1_4 Depth=1
 ; GLOBALNESS0-NEXT:    global_load_dwordx2 v[0:1], v[44:45], off
-; GLOBALNESS0-NEXT:    v_writelane_b32 v59, s8, 10
-; GLOBALNESS0-NEXT:    v_writelane_b32 v59, s9, 11
-; GLOBALNESS0-NEXT:    v_readlane_b32 s4, v59, 2
-; GLOBALNESS0-NEXT:    v_readlane_b32 s5, v59, 3
+; GLOBALNESS0-NEXT:    v_writelane_b32 v57, s8, 10
+; GLOBALNESS0-NEXT:    v_writelane_b32 v57, s9, 11
+; GLOBALNESS0-NEXT:    v_readlane_b32 s4, v57, 2
+; GLOBALNESS0-NEXT:    v_readlane_b32 s5, v57, 3
 ; GLOBALNESS0-NEXT:    s_mov_b32 s83, s55
 ; GLOBALNESS0-NEXT:    s_and_b64 vcc, exec, s[4:5]
 ; GLOBALNESS0-NEXT:    s_cbranch_vccnz .LBB1_13
@@ -513,7 +513,7 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS0-NEXT:    global_store_dwordx2 v[44:45], v[42:43], off
 ; GLOBALNESS0-NEXT:  .LBB1_13: ; %bb44.lr.ph.i
 ; GLOBALNESS0-NEXT:    ; in Loop: Header=BB1_4 Depth=1
-; GLOBALNESS0-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v58
+; GLOBALNESS0-NEXT:    v_cmp_ne_u32_e32 vcc, 0, v56
 ; GLOBALNESS0-NEXT:    v_cndmask_b32_e32 v2, 0, v40, vcc
 ; GLOBALNESS0-NEXT:    s_waitcnt vmcnt(0)
 ; GLOBALNESS0-NEXT:    v_cmp_nlt_f64_e32 vcc, 0, v[0:1]
@@ -543,8 +543,8 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS0-NEXT:    s_cbranch_vccnz .LBB1_21
 ; GLOBALNESS0-NEXT:  ; %bb.19: ; %bb3.i.i
 ; GLOBALNESS0-NEXT:    ; in Loop: Header=BB1_16 Depth=2
-; GLOBALNESS0-NEXT:    v_readlane_b32 s4, v59, 0
-; GLOBALNESS0-NEXT:    v_readlane_b32 s5, v59, 1
+; GLOBALNESS0-NEXT:    v_readlane_b32 s4, v57, 0
+; GLOBALNESS0-NEXT:    v_readlane_b32 s5, v57, 1
 ; GLOBALNESS0-NEXT:    s_and_b64 vcc, exec, s[4:5]
 ; GLOBALNESS0-NEXT:    s_cbranch_vccnz .LBB1_21
 ; GLOBALNESS0-NEXT:  ; %bb.20: ; %bb6.i.i
@@ -580,7 +580,7 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS0-NEXT:    s_mov_b32 s13, s71
 ; GLOBALNESS0-NEXT:    s_mov_b32 s14, s70
 ; GLOBALNESS0-NEXT:    v_mov_b32_e32 v31, v41
-; GLOBALNESS0-NEXT:    global_store_dwordx2 v[44:45], v[56:57], off
+; GLOBALNESS0-NEXT:    global_store_dwordx2 v[44:45], v[58:59], off
 ; GLOBALNESS0-NEXT:    s_swappc_b64 s[30:31], s[54:55]
 ; GLOBALNESS0-NEXT:    s_and_saveexec_b64 s[4:5], s[96:97]
 ; GLOBALNESS0-NEXT:    s_cbranch_execz .LBB1_14
@@ -591,12 +591,12 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS0-NEXT:    s_branch .LBB1_14
 ; GLOBALNESS0-NEXT:  .LBB1_24: ; %Flow23
 ; GLOBALNESS0-NEXT:    ; in Loop: Header=BB1_4 Depth=1
-; GLOBALNESS0-NEXT:    v_readlane_b32 s84, v59, 8
-; GLOBALNESS0-NEXT:    v_readlane_b32 s8, v59, 10
+; GLOBALNESS0-NEXT:    v_readlane_b32 s84, v57, 8
+; GLOBALNESS0-NEXT:    v_readlane_b32 s8, v57, 10
 ; GLOBALNESS0-NEXT:    v_pk_mov_b32 v[0:1], 0, 0
 ; GLOBALNESS0-NEXT:    s_mov_b32 s55, s83
-; GLOBALNESS0-NEXT:    v_readlane_b32 s85, v59, 9
-; GLOBALNESS0-NEXT:    v_readlane_b32 s9, v59, 11
+; GLOBALNESS0-NEXT:    v_readlane_b32 s85, v57, 9
+; GLOBALNESS0-NEXT:    v_readlane_b32 s9, v57, 11
 ; GLOBALNESS0-NEXT:  .LBB1_25: ; %Flow24
 ; GLOBALNESS0-NEXT:    ; in Loop: Header=BB1_4 Depth=1
 ; GLOBALNESS0-NEXT:    s_or_b64 exec, exec, s[52:53]
@@ -604,8 +604,8 @@ define amdgpu_kernel void @kernel(ptr addrspace(1) %arg1.global, i1 %tmp3.i.i, i
 ; GLOBALNESS0-NEXT:    s_cbranch_execz .LBB1_2
 ; GLOBALNESS0-NEXT:  ; %bb.26: ; %bb67.i
 ; GLOBALNESS0-NEXT:    ; in Loop: Header=BB1_4 Depth=1
-; GLOBALNESS0-NEXT:    v_readlane_b32 s6, v59, 4
-; GLOBALNESS0-NEXT:    v_readlane_b32 s7, v59, 5
+; GLOBALNESS0-NEXT:    v_readlane_b32 s6, v57, 4
+; GLOBALNESS0-NEXT:    v_readlane_b32 s7, v57, 5
 ; GLOBALNESS0-NEXT:    s_and_b64 vcc, exec, s[6:7]
 ; GLOBALNESS0-NEXT:    s_cbranch_vccnz .LBB1_1
 ; GLOBALNESS0-NEXT:  ; %bb.27: ; %bb69.i
diff --git a/llvm/test/CodeGen/AMDGPU/undef-handling-crash-in-ra.ll b/llvm/test/CodeGen/AMDGPU/undef-handling-crash-in-ra.ll
index d0d1ba8..b3166fa 100644
--- a/llvm/test/CodeGen/AMDGPU/undef-handling-crash-in-ra.ll
+++ b/llvm/test/CodeGen/AMDGPU/undef-handling-crash-in-ra.ll
@@ -8,9 +8,8 @@ define amdgpu_kernel void @foo(ptr addrspace(5) %ptr5, ptr %p0, double %v0, <4 x
 ; CHECK:       ; %bb.0: ; %entry
 ; CHECK-NEXT:    s_add_u32 flat_scratch_lo, s12, s17
 ; CHECK-NEXT:    s_addc_u32 flat_scratch_hi, s13, 0
-; CHECK-NEXT:    v_mov_b32_e32 v40, v0
-; CHECK-NEXT:    v_pk_mov_b32 v[0:1], 0, 0
-; CHECK-NEXT:    flat_load_dword v42, v[0:1]
+; CHECK-NEXT:    v_pk_mov_b32 v[44:45], 0, 0
+; CHECK-NEXT:    flat_load_dword v42, v[44:45]
 ; CHECK-NEXT:    s_mov_b64 s[38:39], s[6:7]
 ; CHECK-NEXT:    s_mov_b64 s[48:49], s[4:5]
 ; CHECK-NEXT:    s_load_dwordx4 s[4:7], s[8:9], 0x8
@@ -19,48 +18,44 @@ define amdgpu_kernel void @foo(ptr addrspace(5) %ptr5, ptr %p0, double %v0, <4 x
 ; CHECK-NEXT:    s_addc_u32 s1, s1, 0
 ; CHECK-NEXT:    s_mov_b64 s[34:35], s[8:9]
 ; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
-; CHECK-NEXT:    v_mov_b32_e32 v46, s6
-; CHECK-NEXT:    v_mov_b32_e32 v47, s7
+; CHECK-NEXT:    v_accvgpr_write_b32 a32, s6
+; CHECK-NEXT:    v_accvgpr_write_b32 a33, s7
 ; CHECK-NEXT:    s_mov_b64 s[6:7], src_private_base
 ; CHECK-NEXT:    s_cmp_lg_u32 s64, -1
 ; CHECK-NEXT:    s_cselect_b32 s7, s7, 0
 ; CHECK-NEXT:    s_cselect_b32 s8, s64, 0
 ; CHECK-NEXT:    s_add_u32 s50, s34, 48
 ; CHECK-NEXT:    s_addc_u32 s51, s35, 0
-; CHECK-NEXT:    v_pk_mov_b32 v[58:59], s[4:5], s[4:5] op_sel:[0,1]
+; CHECK-NEXT:    v_pk_mov_b32 v[56:57], s[4:5], s[4:5] op_sel:[0,1]
 ; CHECK-NEXT:    s_getpc_b64 s[4:5]
 ; CHECK-NEXT:    s_add_u32 s4, s4, G@gotpcrel32@lo+4
 ; CHECK-NEXT:    s_addc_u32 s5, s5, G@gotpcrel32@hi+12
 ; CHECK-NEXT:    s_load_dwordx2 s[54:55], s[4:5], 0x0
 ; CHECK-NEXT:    s_mov_b32 s6, 0
-; CHECK-NEXT:    v_pk_mov_b32 v[0:1], 0, 0
-; CHECK-NEXT:    v_mov_b32_e32 v57, s7
+; CHECK-NEXT:    v_mov_b32_e32 v47, s7
 ; CHECK-NEXT:    s_mov_b32 s7, s6
 ; CHECK-NEXT:    s_mov_b32 s53, s14
-; CHECK-NEXT:    v_accvgpr_write_b32 a33, v1
-; CHECK-NEXT:    v_mov_b32_e32 v56, s8
-; CHECK-NEXT:    v_pk_mov_b32 v[60:61], s[6:7], s[6:7] op_sel:[0,1]
+; CHECK-NEXT:    v_mov_b32_e32 v46, s8
+; CHECK-NEXT:    v_pk_mov_b32 v[58:59], s[6:7], s[6:7] op_sel:[0,1]
 ; CHECK-NEXT:    s_mov_b64 s[4:5], s[48:49]
 ; CHECK-NEXT:    s_mov_b64 s[6:7], s[38:39]
 ; CHECK-NEXT:    s_mov_b64 s[8:9], s[50:51]
 ; CHECK-NEXT:    s_mov_b32 s12, s14
 ; CHECK-NEXT:    s_mov_b32 s13, s15
 ; CHECK-NEXT:    s_mov_b32 s14, s16
-; CHECK-NEXT:    v_mov_b32_e32 v31, v40
+; CHECK-NEXT:    v_mov_b32_e32 v31, v0
 ; CHECK-NEXT:    s_mov_b32 s32, 0
 ; CHECK-NEXT:    s_mov_b32 s33, s16
 ; CHECK-NEXT:    s_mov_b32 s52, s15
 ; CHECK-NEXT:    s_mov_b64 s[36:37], s[10:11]
-; CHECK-NEXT:    v_accvgpr_write_b32 a32, v0
-; CHECK-NEXT:    flat_store_dwordx2 v[58:59], v[60:61]
+; CHECK-NEXT:    v_mov_b32_e32 v40, v0
+; CHECK-NEXT:    flat_store_dwordx2 v[56:57], v[58:59]
 ; CHECK-NEXT:    ; kill: def $sgpr15 killed $sgpr15
 ; CHECK-NEXT:    s_waitcnt lgkmcnt(0)
 ; CHECK-NEXT:    s_swappc_b64 s[30:31], s[54:55]
-; CHECK-NEXT:    flat_load_dwordx2 v[62:63], v[58:59]
-; CHECK-NEXT:    v_accvgpr_read_b32 v0, a32
-; CHECK-NEXT:    v_mov_b32_e32 v44, 0
-; CHECK-NEXT:    v_mov_b32_e32 v45, 0x3ff00000
-; CHECK-NEXT:    v_accvgpr_read_b32 v1, a33
+; CHECK-NEXT:    flat_load_dwordx2 v[60:61], v[56:57]
+; CHECK-NEXT:    v_mov_b32_e32 v62, 0
+; CHECK-NEXT:    v_mov_b32_e32 v63, 0x3ff00000
 ; CHECK-NEXT:    s_mov_b64 s[4:5], s[48:49]
 ; CHECK-NEXT:    s_mov_b64 s[6:7], s[38:39]
 ; CHECK-NEXT:    s_mov_b64 s[8:9], s[50:51]
@@ -69,20 +64,20 @@ define amdgpu_kernel void @foo(ptr addrspace(5) %ptr5, ptr %p0, double %v0, <4 x
 ; CHECK-NEXT:    s_mov_b32 s13, s52
 ; CHECK-NEXT:    s_mov_b32 s14, s33
 ; CHECK-NEXT:    v_mov_b32_e32 v31, v40
-; CHECK-NEXT:    flat_store_dwordx2 v[0:1], v[44:45]
-; CHECK-NEXT:    flat_store_dwordx2 v[58:59], v[60:61]
+; CHECK-NEXT:    flat_store_dwordx2 v[44:45], v[62:63]
+; CHECK-NEXT:    flat_store_dwordx2 v[56:57], v[58:59]
 ; CHECK-NEXT:    s_waitcnt vmcnt(0)
 ; CHECK-NEXT:    ; kill: def $sgpr15 killed $sgpr15
 ; CHECK-NEXT:    s_swappc_b64 s[30:31], s[54:55]
-; CHECK-NEXT:    flat_load_dwordx2 v[0:1], v[56:57] glc
+; CHECK-NEXT:    flat_load_dwordx2 v[0:1], v[46:47] glc
 ; CHECK-NEXT:    s_waitcnt vmcnt(0) lgkmcnt(0)
 ; CHECK-NEXT:    v_mov_b32_e32 v0, s64
 ; CHECK-NEXT:    v_cmp_lt_i32_e32 vcc, 0, v42
-; CHECK-NEXT:    flat_store_dwordx2 v[58:59], v[62:63]
+; CHECK-NEXT:    flat_store_dwordx2 v[56:57], v[60:61]
 ; CHECK-NEXT:    s_waitcnt vmcnt(0)
-; CHECK-NEXT:    flat_store_dwordx2 v[58:59], v[46:47]
-; CHECK-NEXT:    buffer_store_dword v47, v0, s[0:3], 0 offen offset:4
-; CHECK-NEXT:    buffer_store_dword v44, v0, s[0:3], 0 offen
+; CHECK-NEXT:    flat_store_dwordx2 v[56:57], a[32:33]
+; CHECK-NEXT:    buffer_store_dword a33, v0, s[0:3], 0 offen offset:4
+; CHECK-NEXT:    buffer_store_dword v62, v0, s[0:3], 0 offen
 ; CHECK-NEXT:    ; implicit-def: $vgpr4
 ; CHECK-NEXT:    s_and_saveexec_b64 s[4:5], vcc
 ; CHECK-NEXT:    s_xor_b64 s[4:5], exec, s[4:5]
diff --git a/llvm/test/CodeGen/AMDGPU/vector_shuffle.packed.ll b/llvm/test/CodeGen/AMDGPU/vector_shuffle.packed.ll
index 2f25a93..fe7def8a 100644
--- a/llvm/test/CodeGen/AMDGPU/vector_shuffle.packed.ll
+++ b/llvm/test/CodeGen/AMDGPU/vector_shuffle.packed.ll
@@ -1961,16 +1961,15 @@ define <6 x half> @shuffle_v6f16_452367(ptr addrspace(1) %arg0, ptr addrspace(1)
 ; GFX942-LABEL: shuffle_v6f16_452367:
 ; GFX942:       ; %bb.0:
 ; GFX942-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b32_e32 v7, v1
-; GFX942-NEXT:    v_mov_b32_e32 v6, v0
-; GFX942-NEXT:    v_mov_b32_e32 v5, v3
-; GFX942-NEXT:    v_mov_b32_e32 v4, v2
-; GFX942-NEXT:    global_load_dwordx3 v[0:2], v[6:7], off
-; GFX942-NEXT:    global_load_dword v3, v[4:5], off
+; GFX942-NEXT:    global_load_dwordx3 v[4:6], v[0:1], off
+; GFX942-NEXT:    global_load_dword v4, v[2:3], off
+; GFX942-NEXT:    ; kill: killed $vgpr0 killed $vgpr1
+; GFX942-NEXT:    ; kill: killed $vgpr2 killed $vgpr3
 ; GFX942-NEXT:    s_waitcnt vmcnt(1)
-; GFX942-NEXT:    v_mov_b32_e32 v0, v2
+; GFX942-NEXT:    v_mov_b32_e32 v0, v6
+; GFX942-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    v_mov_b32_e32 v2, v3
+; GFX942-NEXT:    v_mov_b32_e32 v2, v4
 ; GFX942-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-LABEL: shuffle_v6f16_452367:
@@ -5151,16 +5150,15 @@ define <6 x bfloat> @shuffle_v6bf16_452367(ptr addrspace(1) %arg0, ptr addrspace
 ; GFX942-LABEL: shuffle_v6bf16_452367:
 ; GFX942:       ; %bb.0:
 ; GFX942-NEXT:    s_waitcnt vmcnt(0) expcnt(0) lgkmcnt(0)
-; GFX942-NEXT:    v_mov_b32_e32 v7, v1
-; GFX942-NEXT:    v_mov_b32_e32 v6, v0
-; GFX942-NEXT:    v_mov_b32_e32 v5, v3
-; GFX942-NEXT:    v_mov_b32_e32 v4, v2
-; GFX942-NEXT:    global_load_dwordx3 v[0:2], v[6:7], off
-; GFX942-NEXT:    global_load_dword v3, v[4:5], off
+; GFX942-NEXT:    global_load_dwordx3 v[4:6], v[0:1], off
+; GFX942-NEXT:    global_load_dword v4, v[2:3], off
+; GFX942-NEXT:    ; kill: killed $vgpr0 killed $vgpr1
+; GFX942-NEXT:    ; kill: killed $vgpr2 killed $vgpr3
 ; GFX942-NEXT:    s_waitcnt vmcnt(1)
-; GFX942-NEXT:    v_mov_b32_e32 v0, v2
+; GFX942-NEXT:    v_mov_b32_e32 v0, v6
+; GFX942-NEXT:    v_mov_b32_e32 v1, v5
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    v_mov_b32_e32 v2, v3
+; GFX942-NEXT:    v_mov_b32_e32 v2, v4
 ; GFX942-NEXT:    s_setpc_b64 s[30:31]
 ;
 ; GFX10-LABEL: shuffle_v6bf16_452367:
diff --git a/llvm/test/CodeGen/AMDGPU/vni8-across-blocks.ll b/llvm/test/CodeGen/AMDGPU/vni8-across-blocks.ll
index a401f989..d8264b5a 100644
--- a/llvm/test/CodeGen/AMDGPU/vni8-across-blocks.ll
+++ b/llvm/test/CodeGen/AMDGPU/vni8-across-blocks.ll
@@ -58,19 +58,19 @@ define amdgpu_kernel void @v4i8_liveout(ptr addrspace(1) %src1, ptr addrspace(1)
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
 ; GFX942-NEXT:    v_and_b32_e32 v3, 0x3ff, v0
-; GFX942-NEXT:    v_lshlrev_b32_e32 v2, 2, v3
+; GFX942-NEXT:    v_lshlrev_b32_e32 v1, 2, v3
 ; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    global_load_dword v1, v2, s[0:1]
+; GFX942-NEXT:    global_load_dword v2, v1, s[0:1]
 ; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 15, v3
 ; GFX942-NEXT:    s_and_saveexec_b64 s[0:1], vcc
 ; GFX942-NEXT:    s_cbranch_execz .LBB1_2
 ; GFX942-NEXT:  ; %bb.1: ; %bb.1
-; GFX942-NEXT:    global_load_dword v1, v2, s[2:3]
+; GFX942-NEXT:    global_load_dword v2, v1, s[2:3]
 ; GFX942-NEXT:  .LBB1_2: ; %bb.2
 ; GFX942-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dword v0, v1, s[6:7]
+; GFX942-NEXT:    global_store_dword v0, v2, s[6:7]
 ; GFX942-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
@@ -136,19 +136,19 @@ define amdgpu_kernel void @v8i8_liveout(ptr addrspace(1) %src1, ptr addrspace(1)
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
 ; GFX942-NEXT:    v_and_b32_e32 v4, 0x3ff, v0
-; GFX942-NEXT:    v_lshlrev_b32_e32 v3, 3, v4
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_lshlrev_b32_e32 v1, 3, v4
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    global_load_dwordx2 v[0:1], v3, s[0:1]
+; GFX942-NEXT:    global_load_dwordx2 v[2:3], v1, s[0:1]
 ; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 15, v4
 ; GFX942-NEXT:    s_and_saveexec_b64 s[0:1], vcc
 ; GFX942-NEXT:    s_cbranch_execz .LBB3_2
 ; GFX942-NEXT:  ; %bb.1: ; %bb.1
-; GFX942-NEXT:    global_load_dwordx2 v[0:1], v3, s[2:3]
+; GFX942-NEXT:    global_load_dwordx2 v[2:3], v1, s[2:3]
 ; GFX942-NEXT:  .LBB3_2: ; %bb.2
 ; GFX942-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[6:7]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[6:7]
 ; GFX942-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
@@ -173,19 +173,19 @@ define amdgpu_kernel void @v16i8_liveout(ptr addrspace(1) %src1, ptr addrspace(1
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
 ; GFX942-NEXT:    v_and_b32_e32 v6, 0x3ff, v0
-; GFX942-NEXT:    v_lshlrev_b32_e32 v5, 4, v6
-; GFX942-NEXT:    v_mov_b32_e32 v4, 0
+; GFX942-NEXT:    v_lshlrev_b32_e32 v1, 4, v6
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    global_load_dwordx4 v[0:3], v5, s[0:1]
+; GFX942-NEXT:    global_load_dwordx4 v[2:5], v1, s[0:1]
 ; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 15, v6
 ; GFX942-NEXT:    s_and_saveexec_b64 s[0:1], vcc
 ; GFX942-NEXT:    s_cbranch_execz .LBB4_2
 ; GFX942-NEXT:  ; %bb.1: ; %bb.1
-; GFX942-NEXT:    global_load_dwordx4 v[0:3], v5, s[2:3]
+; GFX942-NEXT:    global_load_dwordx4 v[2:5], v1, s[2:3]
 ; GFX942-NEXT:  .LBB4_2: ; %bb.2
 ; GFX942-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx4 v4, v[0:3], s[6:7]
+; GFX942-NEXT:    global_store_dwordx4 v0, v[2:5], s[6:7]
 ; GFX942-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
@@ -210,23 +210,23 @@ define amdgpu_kernel void @v32i8_liveout(ptr addrspace(1) %src1, ptr addrspace(1
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
 ; GFX942-NEXT:    v_and_b32_e32 v10, 0x3ff, v0
-; GFX942-NEXT:    v_lshlrev_b32_e32 v9, 5, v10
-; GFX942-NEXT:    v_mov_b32_e32 v8, 0
+; GFX942-NEXT:    v_lshlrev_b32_e32 v1, 5, v10
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    global_load_dwordx4 v[4:7], v9, s[0:1] offset:16
-; GFX942-NEXT:    global_load_dwordx4 v[0:3], v9, s[0:1]
+; GFX942-NEXT:    global_load_dwordx4 v[6:9], v1, s[0:1] offset:16
+; GFX942-NEXT:    global_load_dwordx4 v[2:5], v1, s[0:1]
 ; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 15, v10
 ; GFX942-NEXT:    s_and_saveexec_b64 s[0:1], vcc
 ; GFX942-NEXT:    s_cbranch_execz .LBB5_2
 ; GFX942-NEXT:  ; %bb.1: ; %bb.1
-; GFX942-NEXT:    global_load_dwordx4 v[4:7], v9, s[2:3] offset:16
-; GFX942-NEXT:    global_load_dwordx4 v[0:3], v9, s[2:3]
+; GFX942-NEXT:    global_load_dwordx4 v[6:9], v1, s[2:3] offset:16
+; GFX942-NEXT:    global_load_dwordx4 v[2:5], v1, s[2:3]
 ; GFX942-NEXT:  .LBB5_2: ; %bb.2
 ; GFX942-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; GFX942-NEXT:    s_waitcnt vmcnt(1)
-; GFX942-NEXT:    global_store_dwordx4 v8, v[4:7], s[6:7] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v0, v[6:9], s[6:7] offset:16
 ; GFX942-NEXT:    s_waitcnt vmcnt(1)
-; GFX942-NEXT:    global_store_dwordx4 v8, v[0:3], s[6:7]
+; GFX942-NEXT:    global_store_dwordx4 v0, v[2:5], s[6:7]
 ; GFX942-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
@@ -250,72 +250,72 @@ define amdgpu_kernel void @v256i8_liveout(ptr addrspace(1) %src1, ptr addrspace(
 ; GFX942:       ; %bb.0: ; %entry
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[4:5], 0x24
 ; GFX942-NEXT:    s_load_dwordx2 s[6:7], s[4:5], 0x34
-; GFX942-NEXT:    v_and_b32_e32 v2, 0x3ff, v0
-; GFX942-NEXT:    v_lshlrev_b32_e32 v1, 3, v2
+; GFX942-NEXT:    v_and_b32_e32 v62, 0x3ff, v0
+; GFX942-NEXT:    v_lshlrev_b32_e32 v1, 3, v62
 ; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    global_load_dwordx4 v[28:31], v1, s[0:1] offset:240
-; GFX942-NEXT:    global_load_dwordx4 v[24:27], v1, s[0:1] offset:224
-; GFX942-NEXT:    global_load_dwordx4 v[20:23], v1, s[0:1] offset:208
-; GFX942-NEXT:    global_load_dwordx4 v[16:19], v1, s[0:1] offset:192
-; GFX942-NEXT:    global_load_dwordx4 v[12:15], v1, s[0:1] offset:176
-; GFX942-NEXT:    global_load_dwordx4 v[8:11], v1, s[0:1] offset:160
-; GFX942-NEXT:    global_load_dwordx4 v[4:7], v1, s[0:1] offset:144
-; GFX942-NEXT:    global_load_dwordx4 a[0:3], v1, s[0:1] offset:128
-; GFX942-NEXT:    global_load_dwordx4 v[60:63], v1, s[0:1] offset:112
-; GFX942-NEXT:    global_load_dwordx4 v[56:59], v1, s[0:1] offset:96
-; GFX942-NEXT:    global_load_dwordx4 v[52:55], v1, s[0:1] offset:80
-; GFX942-NEXT:    global_load_dwordx4 v[48:51], v1, s[0:1] offset:64
-; GFX942-NEXT:    global_load_dwordx4 v[44:47], v1, s[0:1] offset:48
-; GFX942-NEXT:    global_load_dwordx4 v[40:43], v1, s[0:1] offset:32
-; GFX942-NEXT:    global_load_dwordx4 v[36:39], v1, s[0:1] offset:16
-; GFX942-NEXT:    global_load_dwordx4 v[32:35], v1, s[0:1]
-; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 15, v2
+; GFX942-NEXT:    global_load_dwordx4 v[30:33], v1, s[0:1] offset:240
+; GFX942-NEXT:    global_load_dwordx4 v[26:29], v1, s[0:1] offset:224
+; GFX942-NEXT:    global_load_dwordx4 v[22:25], v1, s[0:1] offset:208
+; GFX942-NEXT:    global_load_dwordx4 v[18:21], v1, s[0:1] offset:192
+; GFX942-NEXT:    global_load_dwordx4 v[14:17], v1, s[0:1] offset:176
+; GFX942-NEXT:    global_load_dwordx4 v[10:13], v1, s[0:1] offset:160
+; GFX942-NEXT:    global_load_dwordx4 v[6:9], v1, s[0:1] offset:144
+; GFX942-NEXT:    global_load_dwordx4 v[2:5], v1, s[0:1] offset:128
+; GFX942-NEXT:    global_load_dwordx4 a[0:3], v1, s[0:1] offset:112
+; GFX942-NEXT:    global_load_dwordx4 v[58:61], v1, s[0:1] offset:96
+; GFX942-NEXT:    global_load_dwordx4 v[54:57], v1, s[0:1] offset:80
+; GFX942-NEXT:    global_load_dwordx4 v[50:53], v1, s[0:1] offset:64
+; GFX942-NEXT:    global_load_dwordx4 v[46:49], v1, s[0:1] offset:48
+; GFX942-NEXT:    global_load_dwordx4 v[42:45], v1, s[0:1] offset:32
+; GFX942-NEXT:    global_load_dwordx4 v[38:41], v1, s[0:1] offset:16
+; GFX942-NEXT:    global_load_dwordx4 v[34:37], v1, s[0:1]
+; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 15, v62
 ; GFX942-NEXT:    s_and_saveexec_b64 s[0:1], vcc
 ; GFX942-NEXT:    s_cbranch_execz .LBB6_2
 ; GFX942-NEXT:  ; %bb.1: ; %bb.1
-; GFX942-NEXT:    global_load_dwordx4 v[28:31], v1, s[2:3] offset:240
-; GFX942-NEXT:    global_load_dwordx4 v[24:27], v1, s[2:3] offset:224
-; GFX942-NEXT:    global_load_dwordx4 v[20:23], v1, s[2:3] offset:208
-; GFX942-NEXT:    global_load_dwordx4 v[16:19], v1, s[2:3] offset:192
-; GFX942-NEXT:    global_load_dwordx4 v[12:15], v1, s[2:3] offset:176
-; GFX942-NEXT:    global_load_dwordx4 v[8:11], v1, s[2:3] offset:160
-; GFX942-NEXT:    global_load_dwordx4 v[4:7], v1, s[2:3] offset:144
-; GFX942-NEXT:    global_load_dwordx4 a[0:3], v1, s[2:3] offset:128
-; GFX942-NEXT:    global_load_dwordx4 v[60:63], v1, s[2:3] offset:112
-; GFX942-NEXT:    global_load_dwordx4 v[56:59], v1, s[2:3] offset:96
-; GFX942-NEXT:    global_load_dwordx4 v[52:55], v1, s[2:3] offset:80
-; GFX942-NEXT:    global_load_dwordx4 v[48:51], v1, s[2:3] offset:64
-; GFX942-NEXT:    global_load_dwordx4 v[44:47], v1, s[2:3] offset:48
-; GFX942-NEXT:    global_load_dwordx4 v[40:43], v1, s[2:3] offset:32
-; GFX942-NEXT:    global_load_dwordx4 v[36:39], v1, s[2:3] offset:16
-; GFX942-NEXT:    global_load_dwordx4 v[32:35], v1, s[2:3]
+; GFX942-NEXT:    global_load_dwordx4 v[30:33], v1, s[2:3] offset:240
+; GFX942-NEXT:    global_load_dwordx4 v[26:29], v1, s[2:3] offset:224
+; GFX942-NEXT:    global_load_dwordx4 v[22:25], v1, s[2:3] offset:208
+; GFX942-NEXT:    global_load_dwordx4 v[18:21], v1, s[2:3] offset:192
+; GFX942-NEXT:    global_load_dwordx4 v[14:17], v1, s[2:3] offset:176
+; GFX942-NEXT:    global_load_dwordx4 v[10:13], v1, s[2:3] offset:160
+; GFX942-NEXT:    global_load_dwordx4 v[6:9], v1, s[2:3] offset:144
+; GFX942-NEXT:    global_load_dwordx4 v[2:5], v1, s[2:3] offset:128
+; GFX942-NEXT:    global_load_dwordx4 a[0:3], v1, s[2:3] offset:112
+; GFX942-NEXT:    global_load_dwordx4 v[58:61], v1, s[2:3] offset:96
+; GFX942-NEXT:    global_load_dwordx4 v[54:57], v1, s[2:3] offset:80
+; GFX942-NEXT:    global_load_dwordx4 v[50:53], v1, s[2:3] offset:64
+; GFX942-NEXT:    global_load_dwordx4 v[46:49], v1, s[2:3] offset:48
+; GFX942-NEXT:    global_load_dwordx4 v[42:45], v1, s[2:3] offset:32
+; GFX942-NEXT:    global_load_dwordx4 v[38:41], v1, s[2:3] offset:16
+; GFX942-NEXT:    global_load_dwordx4 v[34:37], v1, s[2:3]
 ; GFX942-NEXT:  .LBB6_2: ; %bb.2
 ; GFX942-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; GFX942-NEXT:    s_waitcnt vmcnt(7)
-; GFX942-NEXT:    global_store_dwordx4 v0, v[60:63], s[6:7] offset:112
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7] offset:112
 ; GFX942-NEXT:    s_waitcnt vmcnt(7)
-; GFX942-NEXT:    global_store_dwordx4 v0, v[56:59], s[6:7] offset:96
+; GFX942-NEXT:    global_store_dwordx4 v0, v[58:61], s[6:7] offset:96
 ; GFX942-NEXT:    s_waitcnt vmcnt(7)
-; GFX942-NEXT:    global_store_dwordx4 v0, v[52:55], s[6:7] offset:80
+; GFX942-NEXT:    global_store_dwordx4 v0, v[54:57], s[6:7] offset:80
 ; GFX942-NEXT:    s_waitcnt vmcnt(7)
-; GFX942-NEXT:    global_store_dwordx4 v0, v[48:51], s[6:7] offset:64
+; GFX942-NEXT:    global_store_dwordx4 v0, v[50:53], s[6:7] offset:64
 ; GFX942-NEXT:    s_waitcnt vmcnt(7)
-; GFX942-NEXT:    global_store_dwordx4 v0, v[44:47], s[6:7] offset:48
+; GFX942-NEXT:    global_store_dwordx4 v0, v[46:49], s[6:7] offset:48
 ; GFX942-NEXT:    s_waitcnt vmcnt(7)
-; GFX942-NEXT:    global_store_dwordx4 v0, v[40:43], s[6:7] offset:32
+; GFX942-NEXT:    global_store_dwordx4 v0, v[42:45], s[6:7] offset:32
 ; GFX942-NEXT:    s_waitcnt vmcnt(7)
-; GFX942-NEXT:    global_store_dwordx4 v0, v[36:39], s[6:7] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v0, v[38:41], s[6:7] offset:16
 ; GFX942-NEXT:    s_waitcnt vmcnt(7)
-; GFX942-NEXT:    global_store_dwordx4 v0, v[32:35], s[6:7]
-; GFX942-NEXT:    global_store_dwordx4 v0, v[28:31], s[6:7] offset:240
-; GFX942-NEXT:    global_store_dwordx4 v0, v[24:27], s[6:7] offset:224
-; GFX942-NEXT:    global_store_dwordx4 v0, v[20:23], s[6:7] offset:208
-; GFX942-NEXT:    global_store_dwordx4 v0, v[16:19], s[6:7] offset:192
-; GFX942-NEXT:    global_store_dwordx4 v0, v[12:15], s[6:7] offset:176
-; GFX942-NEXT:    global_store_dwordx4 v0, v[8:11], s[6:7] offset:160
-; GFX942-NEXT:    global_store_dwordx4 v0, v[4:7], s[6:7] offset:144
-; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[6:7] offset:128
+; GFX942-NEXT:    global_store_dwordx4 v0, v[34:37], s[6:7]
+; GFX942-NEXT:    global_store_dwordx4 v0, v[30:33], s[6:7] offset:240
+; GFX942-NEXT:    global_store_dwordx4 v0, v[26:29], s[6:7] offset:224
+; GFX942-NEXT:    global_store_dwordx4 v0, v[22:25], s[6:7] offset:208
+; GFX942-NEXT:    global_store_dwordx4 v0, v[18:21], s[6:7] offset:192
+; GFX942-NEXT:    global_store_dwordx4 v0, v[14:17], s[6:7] offset:176
+; GFX942-NEXT:    global_store_dwordx4 v0, v[10:13], s[6:7] offset:160
+; GFX942-NEXT:    global_store_dwordx4 v0, v[6:9], s[6:7] offset:144
+; GFX942-NEXT:    global_store_dwordx4 v0, v[2:5], s[6:7] offset:128
 ; GFX942-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
@@ -391,17 +391,17 @@ define amdgpu_kernel void @v8i8_phi_chain(ptr addrspace(1) %src1, ptr addrspace(
 ; GFX942-LABEL: v8i8_phi_chain:
 ; GFX942:       ; %bb.0: ; %entry
 ; GFX942-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x24
-; GFX942-NEXT:    v_and_b32_e32 v2, 0x3ff, v0
-; GFX942-NEXT:    v_lshlrev_b32_e32 v3, 3, v2
-; GFX942-NEXT:    v_cmp_lt_u32_e64 s[0:1], 14, v2
-; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 15, v2
+; GFX942-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
+; GFX942-NEXT:    v_lshlrev_b32_e32 v1, 3, v0
+; GFX942-NEXT:    v_cmp_lt_u32_e64 s[0:1], 14, v0
+; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 15, v0
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    global_load_dwordx2 v[0:1], v3, s[8:9]
+; GFX942-NEXT:    global_load_dwordx2 v[2:3], v1, s[8:9]
 ; GFX942-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX942-NEXT:    s_cbranch_execz .LBB8_2
 ; GFX942-NEXT:  ; %bb.1: ; %bb.1
-; GFX942-NEXT:    global_load_dwordx2 v[0:1], v3, s[10:11]
-; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 7, v2
+; GFX942-NEXT:    global_load_dwordx2 v[2:3], v1, s[10:11]
+; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 7, v0
 ; GFX942-NEXT:    s_andn2_b64 s[0:1], s[0:1], exec
 ; GFX942-NEXT:    s_and_b64 s[4:5], vcc, exec
 ; GFX942-NEXT:    s_or_b64 s[0:1], s[0:1], s[4:5]
@@ -410,14 +410,14 @@ define amdgpu_kernel void @v8i8_phi_chain(ptr addrspace(1) %src1, ptr addrspace(
 ; GFX942-NEXT:    s_and_saveexec_b64 s[2:3], s[0:1]
 ; GFX942-NEXT:    s_cbranch_execz .LBB8_4
 ; GFX942-NEXT:  ; %bb.3: ; %bb.2
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[12:13]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[12:13]
 ; GFX942-NEXT:  .LBB8_4: ; %bb.3
 ; GFX942-NEXT:    s_or_b64 exec, exec, s[2:3]
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[14:15]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[14:15]
 ; GFX942-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
@@ -447,38 +447,38 @@ define amdgpu_kernel void @v8i8_phi_zeroinit(ptr addrspace(1) %src1, ptr addrspa
 ; GFX942-LABEL: v8i8_phi_zeroinit:
 ; GFX942:       ; %bb.0: ; %entry
 ; GFX942-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x24
-; GFX942-NEXT:    v_and_b32_e32 v4, 0x3ff, v0
-; GFX942-NEXT:    v_lshlrev_b32_e32 v5, 3, v4
-; GFX942-NEXT:    v_cmp_lt_u32_e64 s[0:1], 14, v4
-; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 15, v4
+; GFX942-NEXT:    v_and_b32_e32 v0, 0x3ff, v0
+; GFX942-NEXT:    v_lshlrev_b32_e32 v1, 3, v0
+; GFX942-NEXT:    v_cmp_lt_u32_e64 s[0:1], 14, v0
+; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 15, v0
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    global_load_dwordx2 v[0:1], v5, s[8:9]
-; GFX942-NEXT:    ; implicit-def: $vgpr2_vgpr3
+; GFX942-NEXT:    global_load_dwordx2 v[2:3], v1, s[8:9]
+; GFX942-NEXT:    ; implicit-def: $vgpr4_vgpr5
 ; GFX942-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX942-NEXT:    s_cbranch_execz .LBB9_2
 ; GFX942-NEXT:  ; %bb.1: ; %bb.1
-; GFX942-NEXT:    global_load_dwordx2 v[2:3], v5, s[10:11]
-; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 7, v4
+; GFX942-NEXT:    global_load_dwordx2 v[4:5], v1, s[10:11]
+; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 7, v0
 ; GFX942-NEXT:    s_waitcnt vmcnt(1)
-; GFX942-NEXT:    v_mov_b32_e32 v0, 0
+; GFX942-NEXT:    v_mov_b32_e32 v2, 0
 ; GFX942-NEXT:    s_andn2_b64 s[0:1], s[0:1], exec
 ; GFX942-NEXT:    s_and_b64 s[4:5], vcc, exec
-; GFX942-NEXT:    v_mov_b32_e32 v1, v0
+; GFX942-NEXT:    v_mov_b32_e32 v3, v2
 ; GFX942-NEXT:    s_or_b64 s[0:1], s[0:1], s[4:5]
 ; GFX942-NEXT:  .LBB9_2: ; %Flow
 ; GFX942-NEXT:    s_or_b64 exec, exec, s[2:3]
 ; GFX942-NEXT:    s_and_saveexec_b64 s[2:3], s[0:1]
 ; GFX942-NEXT:    s_cbranch_execz .LBB9_4
 ; GFX942-NEXT:  ; %bb.3: ; %bb.2
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
-; GFX942-NEXT:    global_store_dwordx2 v2, v[0:1], s[12:13]
-; GFX942-NEXT:    v_mov_b64_e32 v[2:3], v[0:1]
+; GFX942-NEXT:    v_mov_b64_e32 v[4:5], v[2:3]
+; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[12:13]
 ; GFX942-NEXT:  .LBB9_4: ; %bb.3
 ; GFX942-NEXT:    s_or_b64 exec, exec, s[2:3]
-; GFX942-NEXT:    s_waitcnt vmcnt(0)
 ; GFX942-NEXT:    v_mov_b32_e32 v0, 0
-; GFX942-NEXT:    global_store_dwordx2 v0, v[2:3], s[14:15]
+; GFX942-NEXT:    s_waitcnt vmcnt(0)
+; GFX942-NEXT:    global_store_dwordx2 v0, v[4:5], s[14:15]
 ; GFX942-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
@@ -617,30 +617,30 @@ define amdgpu_kernel void @v8i8_multi_block(ptr addrspace(1) %src1, ptr addrspac
 ; GFX942-LABEL: v8i8_multi_block:
 ; GFX942:       ; %bb.0: ; %entry
 ; GFX942-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x24
-; GFX942-NEXT:    v_and_b32_e32 v5, 0x3ff, v0
-; GFX942-NEXT:    v_lshlrev_b32_e32 v6, 3, v5
-; GFX942-NEXT:    v_mov_b32_e32 v4, 0
-; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 15, v5
+; GFX942-NEXT:    v_and_b32_e32 v3, 0x3ff, v0
+; GFX942-NEXT:    v_lshlrev_b32_e32 v4, 3, v3
+; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 15, v3
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    global_load_dwordx2 v[2:3], v6, s[8:9]
+; GFX942-NEXT:    global_load_dwordx2 v[0:1], v4, s[8:9]
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    v_mov_b64_e32 v[0:1], v[2:3]
+; GFX942-NEXT:    v_mov_b64_e32 v[6:7], v[0:1]
 ; GFX942-NEXT:    s_and_saveexec_b64 s[0:1], vcc
 ; GFX942-NEXT:    s_cbranch_execz .LBB11_4
 ; GFX942-NEXT:  ; %bb.1: ; %bb.1
-; GFX942-NEXT:    global_load_dwordx2 v[0:1], v6, s[10:11]
-; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 7, v5
+; GFX942-NEXT:    global_load_dwordx2 v[6:7], v4, s[10:11]
+; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 7, v3
 ; GFX942-NEXT:    s_and_saveexec_b64 s[2:3], vcc
 ; GFX942-NEXT:    s_cbranch_execz .LBB11_3
 ; GFX942-NEXT:  ; %bb.2: ; %bb.2
-; GFX942-NEXT:    v_mov_b32_e32 v5, 0
-; GFX942-NEXT:    global_store_dwordx2 v5, v[2:3], s[12:13]
+; GFX942-NEXT:    v_mov_b32_e32 v3, 0
+; GFX942-NEXT:    global_store_dwordx2 v3, v[0:1], s[12:13]
 ; GFX942-NEXT:  .LBB11_3: ; %Flow
 ; GFX942-NEXT:    s_or_b64 exec, exec, s[2:3]
 ; GFX942-NEXT:  .LBB11_4: ; %bb.3
 ; GFX942-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
-; GFX942-NEXT:    global_store_dwordx2 v4, v[0:1], s[14:15]
+; GFX942-NEXT:    global_store_dwordx2 v2, v[6:7], s[14:15]
 ; GFX942-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
@@ -859,15 +859,15 @@ define amdgpu_kernel void @v8i8_mfma_i8(ptr addrspace(1) %src1, ptr addrspace(1)
 ; GFX942:       ; %bb.0: ; %entry
 ; GFX942-NEXT:    s_load_dwordx8 s[8:15], s[4:5], 0x24
 ; GFX942-NEXT:    v_and_b32_e32 v4, 0x3ff, v0
-; GFX942-NEXT:    v_lshlrev_b32_e32 v3, 3, v4
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_lshlrev_b32_e32 v1, 3, v4
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 15, v4
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    global_load_dwordx2 v[0:1], v3, s[8:9]
+; GFX942-NEXT:    global_load_dwordx2 v[2:3], v1, s[8:9]
 ; GFX942-NEXT:    s_and_saveexec_b64 s[0:1], vcc
 ; GFX942-NEXT:    s_cbranch_execz .LBB14_2
 ; GFX942-NEXT:  ; %bb.1: ; %bb.1
-; GFX942-NEXT:    global_load_dwordx2 v[0:1], v3, s[10:11]
+; GFX942-NEXT:    global_load_dwordx2 v[2:3], v1, s[10:11]
 ; GFX942-NEXT:  .LBB14_2: ; %bb.2
 ; GFX942-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; GFX942-NEXT:    s_load_dwordx4 s[0:3], s[14:15], 0x0
@@ -878,9 +878,9 @@ define amdgpu_kernel void @v8i8_mfma_i8(ptr addrspace(1) %src1, ptr addrspace(1)
 ; GFX942-NEXT:    v_accvgpr_write_b32 a3, s3
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
 ; GFX942-NEXT:    s_nop 0
-; GFX942-NEXT:    v_mfma_i32_16x16x32_i8 a[0:3], v[0:1], v[0:1], a[0:3] cbsz:1 abid:2 blgp:3
+; GFX942-NEXT:    v_mfma_i32_16x16x32_i8 a[0:3], v[2:3], v[2:3], a[0:3] cbsz:1 abid:2 blgp:3
 ; GFX942-NEXT:    s_nop 6
-; GFX942-NEXT:    global_store_dwordx4 v2, a[0:3], s[12:13]
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[12:13]
 ; GFX942-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
@@ -909,15 +909,15 @@ define amdgpu_kernel void @v8i8_mfma_half(ptr addrspace(1) %src1, ptr addrspace(
 ; GFX942:       ; %bb.0: ; %entry
 ; GFX942-NEXT:    s_load_dwordx8 s[36:43], s[4:5], 0x24
 ; GFX942-NEXT:    v_and_b32_e32 v4, 0x3ff, v0
-; GFX942-NEXT:    v_lshlrev_b32_e32 v3, 3, v4
-; GFX942-NEXT:    v_mov_b32_e32 v2, 0
+; GFX942-NEXT:    v_lshlrev_b32_e32 v1, 3, v4
+; GFX942-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX942-NEXT:    v_cmp_gt_u32_e32 vcc, 15, v4
 ; GFX942-NEXT:    s_waitcnt lgkmcnt(0)
-; GFX942-NEXT:    global_load_dwordx2 v[0:1], v3, s[36:37]
+; GFX942-NEXT:    global_load_dwordx2 v[2:3], v1, s[36:37]
 ; GFX942-NEXT:    s_and_saveexec_b64 s[0:1], vcc
 ; GFX942-NEXT:    s_cbranch_execz .LBB15_2
 ; GFX942-NEXT:  ; %bb.1: ; %bb.1
-; GFX942-NEXT:    global_load_dwordx2 v[0:1], v3, s[38:39]
+; GFX942-NEXT:    global_load_dwordx2 v[2:3], v1, s[38:39]
 ; GFX942-NEXT:  .LBB15_2: ; %bb.2
 ; GFX942-NEXT:    s_or_b64 exec, exec, s[0:1]
 ; GFX942-NEXT:    s_load_dwordx16 s[16:31], s[42:43], 0x0
@@ -957,18 +957,18 @@ define amdgpu_kernel void @v8i8_mfma_half(ptr addrspace(1) %src1, ptr addrspace(
 ; GFX942-NEXT:    v_accvgpr_write_b32 a31, s15
 ; GFX942-NEXT:    s_waitcnt vmcnt(0)
 ; GFX942-NEXT:    s_nop 0
-; GFX942-NEXT:    v_mfma_f32_32x32x4_2b_f16 a[0:31], v[0:1], v[0:1], a[0:31] cbsz:1 abid:2 blgp:3
+; GFX942-NEXT:    v_mfma_f32_32x32x4_2b_f16 a[0:31], v[2:3], v[2:3], a[0:31] cbsz:1 abid:2 blgp:3
 ; GFX942-NEXT:    s_nop 7
 ; GFX942-NEXT:    s_nop 7
 ; GFX942-NEXT:    s_nop 2
-; GFX942-NEXT:    global_store_dwordx4 v2, a[28:31], s[40:41] offset:112
-; GFX942-NEXT:    global_store_dwordx4 v2, a[24:27], s[40:41] offset:96
-; GFX942-NEXT:    global_store_dwordx4 v2, a[20:23], s[40:41] offset:80
-; GFX942-NEXT:    global_store_dwordx4 v2, a[16:19], s[40:41] offset:64
-; GFX942-NEXT:    global_store_dwordx4 v2, a[12:15], s[40:41] offset:48
-; GFX942-NEXT:    global_store_dwordx4 v2, a[8:11], s[40:41] offset:32
-; GFX942-NEXT:    global_store_dwordx4 v2, a[4:7], s[40:41] offset:16
-; GFX942-NEXT:    global_store_dwordx4 v2, a[0:3], s[40:41]
+; GFX942-NEXT:    global_store_dwordx4 v0, a[28:31], s[40:41] offset:112
+; GFX942-NEXT:    global_store_dwordx4 v0, a[24:27], s[40:41] offset:96
+; GFX942-NEXT:    global_store_dwordx4 v0, a[20:23], s[40:41] offset:80
+; GFX942-NEXT:    global_store_dwordx4 v0, a[16:19], s[40:41] offset:64
+; GFX942-NEXT:    global_store_dwordx4 v0, a[12:15], s[40:41] offset:48
+; GFX942-NEXT:    global_store_dwordx4 v0, a[8:11], s[40:41] offset:32
+; GFX942-NEXT:    global_store_dwordx4 v0, a[4:7], s[40:41] offset:16
+; GFX942-NEXT:    global_store_dwordx4 v0, a[0:3], s[40:41]
 ; GFX942-NEXT:    s_endpgm
 entry:
   %idx = call i32 @llvm.amdgcn.workitem.id.x()
diff --git a/llvm/test/CodeGen/AMDGPU/wwm-regalloc-error.ll b/llvm/test/CodeGen/AMDGPU/wwm-regalloc-error.ll
index 145f1e4..ff18b32 100644
--- a/llvm/test/CodeGen/AMDGPU/wwm-regalloc-error.ll
+++ b/llvm/test/CodeGen/AMDGPU/wwm-regalloc-error.ll
@@ -2,7 +2,7 @@
 
 ; A negative test to capture the expected error when the VGPRs are insufficient for wwm-regalloc.
 
-; CHECK: error: can't find enough VGPRs for wwm-regalloc
+; CHECK: error: cannot find enough VGPRs for wwm-regalloc
 
 define amdgpu_kernel void @test(i32 %in) {
 entry:
diff --git a/llvm/test/CodeGen/DirectX/Binding/binding-overlap-6.ll b/llvm/test/CodeGen/DirectX/Binding/binding-overlap-6.ll
new file mode 100644
index 0000000..3c37e63
--- /dev/null
+++ b/llvm/test/CodeGen/DirectX/Binding/binding-overlap-6.ll
@@ -0,0 +1,24 @@
+; RUN: not opt -S -passes='dxil-post-optimization-validation' -mtriple=dxil-pc-shadermodel6.3-library %s 2>&1 | FileCheck %s
+
+; Check overlap with unbounded array
+
+; A overlaps with B
+; RWBuffer<float> A[3] : register(u0);
+; RWBuffer<float> B[] : register(u4);
+; RWBuffer<float> C : register(u17);
+
+; CHECK: error: resource B at register 4 overlaps with resource C at register 17 in space 0
+
+target triple = "dxil-pc-shadermodel6.3-library"
+
+@A.str = private unnamed_addr constant [2 x i8] c"A\00", align 1
+@B.str = private unnamed_addr constant [2 x i8] c"B\00", align 1
+@C.str = private unnamed_addr constant [2 x i8] c"C\00", align 1
+
+define void @test_overlapping() {
+entry:
+  %h1 = call target("dx.TypedBuffer", float, 1, 0, 0) @llvm.dx.resource.handlefrombinding(i32 0, i32 0, i32 3, i32 0, i1 false, ptr @A.str)
+  %h2 = call target("dx.TypedBuffer", float, 1, 0, 0) @llvm.dx.resource.handlefrombinding(i32 0, i32 4, i32 -1, i32 0, i1 false, ptr @B.str)
+  %h3 = call target("dx.TypedBuffer", float, 1, 0, 0) @llvm.dx.resource.handlefrombinding(i32 0, i32 17, i32 1, i32 0, i1 false, ptr @C.str)
+  ret void
+}
diff --git a/llvm/test/CodeGen/LoongArch/lasx/build-vector.ll b/llvm/test/CodeGen/LoongArch/lasx/build-vector.ll
index 8aa38f0..c18c637 100644
--- a/llvm/test/CodeGen/LoongArch/lasx/build-vector.ll
+++ b/llvm/test/CodeGen/LoongArch/lasx/build-vector.ll
@@ -355,40 +355,46 @@ entry:
 define void @buildvector_v32i8_partial(ptr %dst, i8 %a0, i8 %a1, i8 %a2, i8 %a5, i8 %a7, i8 %a8, i8 %a15, i8 %a17, i8 %a18, i8 %a20, i8 %a22, i8 %a23, i8 %a27, i8 %a28, i8 %a31) nounwind {
 ; CHECK-LABEL: buildvector_v32i8_partial:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    addi.d $sp, $sp, -96
-; CHECK-NEXT:    st.d $ra, $sp, 88 # 8-byte Folded Spill
-; CHECK-NEXT:    st.d $fp, $sp, 80 # 8-byte Folded Spill
-; CHECK-NEXT:    addi.d $fp, $sp, 96
-; CHECK-NEXT:    bstrins.d $sp, $zero, 4, 0
-; CHECK-NEXT:    ld.b $t0, $fp, 0
-; CHECK-NEXT:    ld.b $t1, $fp, 8
-; CHECK-NEXT:    ld.b $t2, $fp, 16
-; CHECK-NEXT:    ld.b $t3, $fp, 24
-; CHECK-NEXT:    ld.b $t4, $fp, 56
-; CHECK-NEXT:    ld.b $t5, $fp, 48
-; CHECK-NEXT:    ld.b $t6, $fp, 40
-; CHECK-NEXT:    ld.b $t7, $fp, 32
-; CHECK-NEXT:    st.b $t4, $sp, 63
-; CHECK-NEXT:    st.b $t5, $sp, 60
-; CHECK-NEXT:    st.b $t6, $sp, 59
-; CHECK-NEXT:    st.b $t7, $sp, 55
-; CHECK-NEXT:    st.b $t3, $sp, 54
-; CHECK-NEXT:    st.b $t2, $sp, 52
-; CHECK-NEXT:    st.b $t1, $sp, 50
-; CHECK-NEXT:    st.b $t0, $sp, 49
-; CHECK-NEXT:    st.b $a7, $sp, 47
-; CHECK-NEXT:    st.b $a6, $sp, 40
-; CHECK-NEXT:    st.b $a5, $sp, 39
-; CHECK-NEXT:    st.b $a4, $sp, 37
-; CHECK-NEXT:    st.b $a3, $sp, 34
-; CHECK-NEXT:    st.b $a2, $sp, 33
-; CHECK-NEXT:    st.b $a1, $sp, 32
-; CHECK-NEXT:    xvld $xr0, $sp, 32
-; CHECK-NEXT:    xvst $xr0, $a0, 0
-; CHECK-NEXT:    addi.d $sp, $fp, -96
-; CHECK-NEXT:    ld.d $fp, $sp, 80 # 8-byte Folded Reload
-; CHECK-NEXT:    ld.d $ra, $sp, 88 # 8-byte Folded Reload
-; CHECK-NEXT:    addi.d $sp, $sp, 96
+; CHECK-NEXT:    ld.b $t0, $sp, 56
+; CHECK-NEXT:    ld.b $t1, $sp, 48
+; CHECK-NEXT:    ld.b $t2, $sp, 40
+; CHECK-NEXT:    ld.b $t3, $sp, 32
+; CHECK-NEXT:    ld.b $t4, $sp, 24
+; CHECK-NEXT:    ld.b $t5, $sp, 16
+; CHECK-NEXT:    ld.b $t6, $sp, 8
+; CHECK-NEXT:    ld.b $t7, $sp, 0
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a1, 0
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a2, 1
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a3, 2
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a4, 5
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a5, 7
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a6, 8
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a7, 15
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.b $vr1, $t7, 1
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.b $vr1, $t6, 2
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.b $vr1, $t5, 4
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.b $vr1, $t4, 6
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.b $vr1, $t3, 7
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.b $vr1, $t2, 11
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.b $vr1, $t1, 12
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.b $vr1, $t0, 15
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvst $xr0, $a0, 0
 ; CHECK-NEXT:    ret
 entry:
   %ins0  = insertelement <32 x i8> undef,  i8   %a0,  i32 0
@@ -430,49 +436,47 @@ entry:
 define void @buildvector_v32i8_with_constant(ptr %dst, i8 %a0, i8 %a1, i8 %a2, i8 %a5, i8 %a8, i8 %a9, i8 %a15, i8 %a17, i8 %a18, i8 %a20, i8 %a22, i8 %a23, i8 %a27, i8 %a28, i8 %a31) nounwind {
 ; CHECK-LABEL: buildvector_v32i8_with_constant:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    addi.d $sp, $sp, -96
-; CHECK-NEXT:    st.d $ra, $sp, 88 # 8-byte Folded Spill
-; CHECK-NEXT:    st.d $fp, $sp, 80 # 8-byte Folded Spill
-; CHECK-NEXT:    addi.d $fp, $sp, 96
-; CHECK-NEXT:    bstrins.d $sp, $zero, 4, 0
-; CHECK-NEXT:    ld.b $t0, $fp, 0
-; CHECK-NEXT:    ld.b $t1, $fp, 8
-; CHECK-NEXT:    ld.b $t2, $fp, 16
-; CHECK-NEXT:    ld.b $t3, $fp, 24
-; CHECK-NEXT:    ld.b $t4, $fp, 56
-; CHECK-NEXT:    ld.b $t5, $fp, 32
-; CHECK-NEXT:    ld.b $t6, $fp, 48
-; CHECK-NEXT:    ld.b $t7, $fp, 40
-; CHECK-NEXT:    st.b $t4, $sp, 63
-; CHECK-NEXT:    st.b $zero, $sp, 61
-; CHECK-NEXT:    st.b $t6, $sp, 60
-; CHECK-NEXT:    st.b $t7, $sp, 59
-; CHECK-NEXT:    st.b $zero, $sp, 56
-; CHECK-NEXT:    st.b $t5, $sp, 55
-; CHECK-NEXT:    st.b $t3, $sp, 54
-; CHECK-NEXT:    st.b $zero, $sp, 53
-; CHECK-NEXT:    st.b $t2, $sp, 52
-; CHECK-NEXT:    st.b $zero, $sp, 51
-; CHECK-NEXT:    st.b $t1, $sp, 50
-; CHECK-NEXT:    st.b $t0, $sp, 49
-; CHECK-NEXT:    st.b $zero, $sp, 48
-; CHECK-NEXT:    st.b $a7, $sp, 47
-; CHECK-NEXT:    st.h $zero, $sp, 44
-; CHECK-NEXT:    st.b $zero, $sp, 42
-; CHECK-NEXT:    st.b $a6, $sp, 41
-; CHECK-NEXT:    st.b $a5, $sp, 40
-; CHECK-NEXT:    st.b $zero, $sp, 39
-; CHECK-NEXT:    st.b $a4, $sp, 37
-; CHECK-NEXT:    st.h $zero, $sp, 35
-; CHECK-NEXT:    st.b $a3, $sp, 34
-; CHECK-NEXT:    st.b $a2, $sp, 33
-; CHECK-NEXT:    st.b $a1, $sp, 32
-; CHECK-NEXT:    xvld $xr0, $sp, 32
-; CHECK-NEXT:    xvst $xr0, $a0, 0
-; CHECK-NEXT:    addi.d $sp, $fp, -96
-; CHECK-NEXT:    ld.d $fp, $sp, 80 # 8-byte Folded Reload
-; CHECK-NEXT:    ld.d $ra, $sp, 88 # 8-byte Folded Reload
-; CHECK-NEXT:    addi.d $sp, $sp, 96
+; CHECK-NEXT:    ld.b $t0, $sp, 56
+; CHECK-NEXT:    ld.b $t1, $sp, 48
+; CHECK-NEXT:    ld.b $t2, $sp, 40
+; CHECK-NEXT:    ld.b $t3, $sp, 32
+; CHECK-NEXT:    ld.b $t4, $sp, 24
+; CHECK-NEXT:    ld.b $t5, $sp, 16
+; CHECK-NEXT:    ld.b $t6, $sp, 8
+; CHECK-NEXT:    ld.b $t7, $sp, 0
+; CHECK-NEXT:    xvrepli.b $xr0, 0
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a1, 0
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a2, 1
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a3, 2
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a4, 5
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a5, 8
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a6, 9
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a7, 15
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.b $vr1, $t7, 1
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.b $vr1, $t6, 2
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.b $vr1, $t5, 4
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.b $vr1, $t4, 6
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.b $vr1, $t3, 7
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.b $vr1, $t2, 11
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.b $vr1, $t1, 12
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.b $vr1, $t0, 15
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvst $xr0, $a0, 0
 ; CHECK-NEXT:    ret
 entry:
   %ins0  = insertelement <32 x i8> undef,  i8   %a0,  i32 0
@@ -581,24 +585,18 @@ entry:
 define void @buildvector_v16i16_partial(ptr %dst, i16 %a0, i16 %a2, i16 %a5, i16 %a6, i16 %a7, i16 %a12, i16 %a13) nounwind {
 ; CHECK-LABEL: buildvector_v16i16_partial:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    addi.d $sp, $sp, -96
-; CHECK-NEXT:    st.d $ra, $sp, 88 # 8-byte Folded Spill
-; CHECK-NEXT:    st.d $fp, $sp, 80 # 8-byte Folded Spill
-; CHECK-NEXT:    addi.d $fp, $sp, 96
-; CHECK-NEXT:    bstrins.d $sp, $zero, 4, 0
-; CHECK-NEXT:    st.h $a7, $sp, 58
-; CHECK-NEXT:    st.h $a6, $sp, 56
-; CHECK-NEXT:    st.h $a5, $sp, 46
-; CHECK-NEXT:    st.h $a4, $sp, 44
-; CHECK-NEXT:    st.h $a3, $sp, 42
-; CHECK-NEXT:    st.h $a2, $sp, 36
-; CHECK-NEXT:    st.h $a1, $sp, 32
-; CHECK-NEXT:    xvld $xr0, $sp, 32
-; CHECK-NEXT:    xvst $xr0, $a0, 0
-; CHECK-NEXT:    addi.d $sp, $fp, -96
-; CHECK-NEXT:    ld.d $fp, $sp, 80 # 8-byte Folded Reload
-; CHECK-NEXT:    ld.d $ra, $sp, 88 # 8-byte Folded Reload
-; CHECK-NEXT:    addi.d $sp, $sp, 96
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a1, 0
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a2, 2
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a3, 5
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a4, 6
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a5, 7
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.h $vr1, $a6, 4
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.h $vr1, $a7, 5
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvst $xr0, $a0, 0
 ; CHECK-NEXT:    ret
 entry:
   %ins0  = insertelement <16 x i16> undef,  i16   %a0,  i32 0
@@ -624,32 +622,19 @@ entry:
 define void @buildvector_v16i16_with_constant(ptr %dst, i16 %a2, i16 %a3, i16 %a5, i16 %a6, i16 %a7, i16 %a12, i16 %a13) nounwind {
 ; CHECK-LABEL: buildvector_v16i16_with_constant:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    addi.d $sp, $sp, -96
-; CHECK-NEXT:    st.d $ra, $sp, 88 # 8-byte Folded Spill
-; CHECK-NEXT:    st.d $fp, $sp, 80 # 8-byte Folded Spill
-; CHECK-NEXT:    addi.d $fp, $sp, 96
-; CHECK-NEXT:    bstrins.d $sp, $zero, 4, 0
-; CHECK-NEXT:    st.h $a7, $sp, 58
-; CHECK-NEXT:    st.h $a6, $sp, 56
-; CHECK-NEXT:    st.h $a5, $sp, 46
-; CHECK-NEXT:    st.h $a4, $sp, 44
-; CHECK-NEXT:    st.h $a3, $sp, 42
-; CHECK-NEXT:    ori $a3, $zero, 2
-; CHECK-NEXT:    st.h $a3, $sp, 40
-; CHECK-NEXT:    st.h $a2, $sp, 38
-; CHECK-NEXT:    st.h $a1, $sp, 36
-; CHECK-NEXT:    lu12i.w $a1, 32
-; CHECK-NEXT:    ori $a1, $a1, 2
-; CHECK-NEXT:    st.w $a1, $sp, 60
-; CHECK-NEXT:    st.w $a1, $sp, 32
-; CHECK-NEXT:    lu32i.d $a1, 131074
-; CHECK-NEXT:    st.d $a1, $sp, 48
-; CHECK-NEXT:    xvld $xr0, $sp, 32
-; CHECK-NEXT:    xvst $xr0, $a0, 0
-; CHECK-NEXT:    addi.d $sp, $fp, -96
-; CHECK-NEXT:    ld.d $fp, $sp, 80 # 8-byte Folded Reload
-; CHECK-NEXT:    ld.d $ra, $sp, 88 # 8-byte Folded Reload
-; CHECK-NEXT:    addi.d $sp, $sp, 96
+; CHECK-NEXT:    xvrepli.h $xr0, 2
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a1, 2
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a2, 3
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a3, 5
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a4, 6
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a5, 7
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.h $vr1, $a6, 4
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvpermi.d $xr1, $xr0, 14
+; CHECK-NEXT:    vinsgr2vr.h $vr1, $a7, 5
+; CHECK-NEXT:    xvpermi.q $xr0, $xr1, 2
+; CHECK-NEXT:    xvst $xr0, $a0, 0
 ; CHECK-NEXT:    ret
 entry:
   %ins0  = insertelement <16 x i16> undef,  i16    2,  i32 0
@@ -702,21 +687,11 @@ entry:
 define void @buildvector_v8i32_partial(ptr %dst, i32 %a2, i32 %a4, i32 %a5, i32 %a6) nounwind {
 ; CHECK-LABEL: buildvector_v8i32_partial:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    addi.d $sp, $sp, -96
-; CHECK-NEXT:    st.d $ra, $sp, 88 # 8-byte Folded Spill
-; CHECK-NEXT:    st.d $fp, $sp, 80 # 8-byte Folded Spill
-; CHECK-NEXT:    addi.d $fp, $sp, 96
-; CHECK-NEXT:    bstrins.d $sp, $zero, 4, 0
-; CHECK-NEXT:    st.w $a4, $sp, 56
-; CHECK-NEXT:    st.w $a3, $sp, 52
-; CHECK-NEXT:    st.w $a2, $sp, 48
-; CHECK-NEXT:    st.w $a1, $sp, 40
-; CHECK-NEXT:    xvld $xr0, $sp, 32
+; CHECK-NEXT:    xvinsgr2vr.w $xr0, $a1, 2
+; CHECK-NEXT:    xvinsgr2vr.w $xr0, $a2, 4
+; CHECK-NEXT:    xvinsgr2vr.w $xr0, $a3, 5
+; CHECK-NEXT:    xvinsgr2vr.w $xr0, $a4, 6
 ; CHECK-NEXT:    xvst $xr0, $a0, 0
-; CHECK-NEXT:    addi.d $sp, $fp, -96
-; CHECK-NEXT:    ld.d $fp, $sp, 80 # 8-byte Folded Reload
-; CHECK-NEXT:    ld.d $ra, $sp, 88 # 8-byte Folded Reload
-; CHECK-NEXT:    addi.d $sp, $sp, 96
 ; CHECK-NEXT:    ret
 entry:
   %ins0 = insertelement <8 x i32> undef, i32 undef, i32 0
@@ -734,24 +709,12 @@ entry:
 define void @buildvector_v8i32_with_constant(ptr %dst, i32 %a2, i32 %a4, i32 %a5, i32 %a6) nounwind {
 ; CHECK-LABEL: buildvector_v8i32_with_constant:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    addi.d $sp, $sp, -96
-; CHECK-NEXT:    st.d $ra, $sp, 88 # 8-byte Folded Spill
-; CHECK-NEXT:    st.d $fp, $sp, 80 # 8-byte Folded Spill
-; CHECK-NEXT:    addi.d $fp, $sp, 96
-; CHECK-NEXT:    bstrins.d $sp, $zero, 4, 0
-; CHECK-NEXT:    st.w $zero, $sp, 60
-; CHECK-NEXT:    st.w $a4, $sp, 56
-; CHECK-NEXT:    st.w $a3, $sp, 52
-; CHECK-NEXT:    st.w $a2, $sp, 48
-; CHECK-NEXT:    st.w $zero, $sp, 44
-; CHECK-NEXT:    st.w $a1, $sp, 40
-; CHECK-NEXT:    st.d $zero, $sp, 32
-; CHECK-NEXT:    xvld $xr0, $sp, 32
-; CHECK-NEXT:    xvst $xr0, $a0, 0
-; CHECK-NEXT:    addi.d $sp, $fp, -96
-; CHECK-NEXT:    ld.d $fp, $sp, 80 # 8-byte Folded Reload
-; CHECK-NEXT:    ld.d $ra, $sp, 88 # 8-byte Folded Reload
-; CHECK-NEXT:    addi.d $sp, $sp, 96
+; CHECK-NEXT:    xvrepli.b $xr0, 0
+; CHECK-NEXT:    xvinsgr2vr.w $xr0, $a1, 2
+; CHECK-NEXT:    xvinsgr2vr.w $xr0, $a2, 4
+; CHECK-NEXT:    xvinsgr2vr.w $xr0, $a3, 5
+; CHECK-NEXT:    xvinsgr2vr.w $xr0, $a4, 6
+; CHECK-NEXT:    xvst $xr0, $a0, 0
 ; CHECK-NEXT:    ret
 entry:
   %ins0 = insertelement <8 x i32> undef, i32   0, i32 0
@@ -787,11 +750,8 @@ entry:
 define void @buildvector_v4i64_partial(ptr %dst, i64 %a1, i64 %a2) nounwind {
 ; CHECK-LABEL: buildvector_v4i64_partial:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    xvinsgr2vr.d $xr0, $a2, 0
-; CHECK-NEXT:    xvpermi.d $xr0, $xr0, 68
-; CHECK-NEXT:    xvinsgr2vr.d $xr1, $a1, 0
-; CHECK-NEXT:    xvpermi.d $xr1, $xr1, 68
-; CHECK-NEXT:    xvpackev.d $xr0, $xr1, $xr0
+; CHECK-NEXT:    xvinsgr2vr.d $xr0, $a1, 1
+; CHECK-NEXT:    xvinsgr2vr.d $xr0, $a2, 2
 ; CHECK-NEXT:    xvst $xr0, $a0, 0
 ; CHECK-NEXT:    ret
 entry:
@@ -806,21 +766,10 @@ entry:
 define void @buildvector_v4i64_with_constant(ptr %dst, i64 %a0, i64 %a2) nounwind {
 ; CHECK-LABEL: buildvector_v4i64_with_constant:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    addi.d $sp, $sp, -96
-; CHECK-NEXT:    st.d $ra, $sp, 88 # 8-byte Folded Spill
-; CHECK-NEXT:    st.d $fp, $sp, 80 # 8-byte Folded Spill
-; CHECK-NEXT:    addi.d $fp, $sp, 96
-; CHECK-NEXT:    bstrins.d $sp, $zero, 4, 0
-; CHECK-NEXT:    st.d $zero, $sp, 56
-; CHECK-NEXT:    st.d $a2, $sp, 48
-; CHECK-NEXT:    st.d $zero, $sp, 40
-; CHECK-NEXT:    st.d $a1, $sp, 32
-; CHECK-NEXT:    xvld $xr0, $sp, 32
+; CHECK-NEXT:    xvrepli.b $xr0, 0
+; CHECK-NEXT:    xvinsgr2vr.d $xr0, $a1, 0
+; CHECK-NEXT:    xvinsgr2vr.d $xr0, $a2, 2
 ; CHECK-NEXT:    xvst $xr0, $a0, 0
-; CHECK-NEXT:    addi.d $sp, $fp, -96
-; CHECK-NEXT:    ld.d $fp, $sp, 80 # 8-byte Folded Reload
-; CHECK-NEXT:    ld.d $ra, $sp, 88 # 8-byte Folded Reload
-; CHECK-NEXT:    addi.d $sp, $sp, 96
 ; CHECK-NEXT:    ret
 entry:
   %ins0 = insertelement <4 x i64> undef, i64 %a0, i32 0
@@ -867,21 +816,15 @@ entry:
 define void @buildvector_v8f32_partial(ptr %dst, float %a1, float %a2, float %a5, float %a7) nounwind {
 ; CHECK-LABEL: buildvector_v8f32_partial:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    addi.d $sp, $sp, -96
-; CHECK-NEXT:    st.d $ra, $sp, 88 # 8-byte Folded Spill
-; CHECK-NEXT:    st.d $fp, $sp, 80 # 8-byte Folded Spill
-; CHECK-NEXT:    addi.d $fp, $sp, 96
-; CHECK-NEXT:    bstrins.d $sp, $zero, 4, 0
-; CHECK-NEXT:    fst.s $fa3, $sp, 60
-; CHECK-NEXT:    fst.s $fa2, $sp, 52
-; CHECK-NEXT:    fst.s $fa1, $sp, 40
-; CHECK-NEXT:    fst.s $fa0, $sp, 36
-; CHECK-NEXT:    xvld $xr0, $sp, 32
+; CHECK-NEXT:    # kill: def $f3 killed $f3 def $xr3
+; CHECK-NEXT:    # kill: def $f2 killed $f2 def $xr2
+; CHECK-NEXT:    # kill: def $f1 killed $f1 def $xr1
+; CHECK-NEXT:    # kill: def $f0 killed $f0 def $xr0
+; CHECK-NEXT:    xvinsve0.w $xr0, $xr0, 1
+; CHECK-NEXT:    xvinsve0.w $xr0, $xr1, 2
+; CHECK-NEXT:    xvinsve0.w $xr0, $xr2, 5
+; CHECK-NEXT:    xvinsve0.w $xr0, $xr3, 7
 ; CHECK-NEXT:    xvst $xr0, $a0, 0
-; CHECK-NEXT:    addi.d $sp, $fp, -96
-; CHECK-NEXT:    ld.d $fp, $sp, 80 # 8-byte Folded Reload
-; CHECK-NEXT:    ld.d $ra, $sp, 88 # 8-byte Folded Reload
-; CHECK-NEXT:    addi.d $sp, $sp, 96
 ; CHECK-NEXT:    ret
 entry:
   %ins0 = insertelement <8 x float> undef, float undef, i32 0
@@ -899,27 +842,17 @@ entry:
 define void @buildvector_v8f32_with_constant(ptr %dst, float %a1, float %a2, float %a5, float %a7) nounwind {
 ; CHECK-LABEL: buildvector_v8f32_with_constant:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    addi.d $sp, $sp, -96
-; CHECK-NEXT:    st.d $ra, $sp, 88 # 8-byte Folded Spill
-; CHECK-NEXT:    st.d $fp, $sp, 80 # 8-byte Folded Spill
-; CHECK-NEXT:    addi.d $fp, $sp, 96
-; CHECK-NEXT:    bstrins.d $sp, $zero, 4, 0
-; CHECK-NEXT:    fst.s $fa3, $sp, 60
-; CHECK-NEXT:    fst.s $fa2, $sp, 52
-; CHECK-NEXT:    fst.s $fa1, $sp, 40
-; CHECK-NEXT:    fst.s $fa0, $sp, 36
-; CHECK-NEXT:    vldi $vr0, -1280
-; CHECK-NEXT:    fst.s $fa0, $sp, 56
+; CHECK-NEXT:    # kill: def $f3 killed $f3 def $xr3
+; CHECK-NEXT:    # kill: def $f2 killed $f2 def $xr2
+; CHECK-NEXT:    # kill: def $f1 killed $f1 def $xr1
+; CHECK-NEXT:    # kill: def $f0 killed $f0 def $xr0
 ; CHECK-NEXT:    lu12i.w $a1, 262144
-; CHECK-NEXT:    lu52i.d $a1, $a1, 1024
-; CHECK-NEXT:    st.d $a1, $sp, 44
-; CHECK-NEXT:    fst.s $fa0, $sp, 32
-; CHECK-NEXT:    xvld $xr0, $sp, 32
-; CHECK-NEXT:    xvst $xr0, $a0, 0
-; CHECK-NEXT:    addi.d $sp, $fp, -96
-; CHECK-NEXT:    ld.d $fp, $sp, 80 # 8-byte Folded Reload
-; CHECK-NEXT:    ld.d $ra, $sp, 88 # 8-byte Folded Reload
-; CHECK-NEXT:    addi.d $sp, $sp, 96
+; CHECK-NEXT:    xvreplgr2vr.w $xr4, $a1
+; CHECK-NEXT:    xvinsve0.w $xr4, $xr0, 1
+; CHECK-NEXT:    xvinsve0.w $xr4, $xr1, 2
+; CHECK-NEXT:    xvinsve0.w $xr4, $xr2, 5
+; CHECK-NEXT:    xvinsve0.w $xr4, $xr3, 7
+; CHECK-NEXT:    xvst $xr4, $a0, 0
 ; CHECK-NEXT:    ret
 entry:
   %ins0 = insertelement <8 x float> undef, float 2.0, i32 0
@@ -960,9 +893,7 @@ define void @buildvector_v4f64_partial(ptr %dst, double %a0, double %a3) nounwin
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    # kill: def $f1_64 killed $f1_64 def $xr1
 ; CHECK-NEXT:    # kill: def $f0_64 killed $f0_64 def $xr0
-; CHECK-NEXT:    xvpermi.d $xr0, $xr0, 68
-; CHECK-NEXT:    xvpermi.d $xr1, $xr1, 68
-; CHECK-NEXT:    xvpackev.d $xr0, $xr1, $xr0
+; CHECK-NEXT:    xvinsve0.d $xr0, $xr1, 3
 ; CHECK-NEXT:    xvst $xr0, $a0, 0
 ; CHECK-NEXT:    ret
 entry:
@@ -977,21 +908,12 @@ entry:
 define void @buildvector_v4f64_with_constant(ptr %dst, double %a0, double %a3) nounwind {
 ; CHECK-LABEL: buildvector_v4f64_with_constant:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    addi.d $sp, $sp, -96
-; CHECK-NEXT:    st.d $ra, $sp, 88 # 8-byte Folded Spill
-; CHECK-NEXT:    st.d $fp, $sp, 80 # 8-byte Folded Spill
-; CHECK-NEXT:    addi.d $fp, $sp, 96
-; CHECK-NEXT:    bstrins.d $sp, $zero, 4, 0
-; CHECK-NEXT:    fst.d $fa1, $sp, 56
-; CHECK-NEXT:    vrepli.b $vr1, 0
-; CHECK-NEXT:    vst $vr1, $sp, 40
-; CHECK-NEXT:    fst.d $fa0, $sp, 32
-; CHECK-NEXT:    xvld $xr0, $sp, 32
-; CHECK-NEXT:    xvst $xr0, $a0, 0
-; CHECK-NEXT:    addi.d $sp, $fp, -96
-; CHECK-NEXT:    ld.d $fp, $sp, 80 # 8-byte Folded Reload
-; CHECK-NEXT:    ld.d $ra, $sp, 88 # 8-byte Folded Reload
-; CHECK-NEXT:    addi.d $sp, $sp, 96
+; CHECK-NEXT:    # kill: def $f1_64 killed $f1_64 def $xr1
+; CHECK-NEXT:    # kill: def $f0_64 killed $f0_64 def $xr0
+; CHECK-NEXT:    xvrepli.b $xr2, 0
+; CHECK-NEXT:    xvinsve0.d $xr2, $xr0, 0
+; CHECK-NEXT:    xvinsve0.d $xr2, $xr1, 3
+; CHECK-NEXT:    xvst $xr2, $a0, 0
 ; CHECK-NEXT:    ret
 entry:
   %ins0 = insertelement <4 x double> undef, double %a0, i32 0
diff --git a/llvm/test/CodeGen/LoongArch/lasx/xvmskcond.ll b/llvm/test/CodeGen/LoongArch/lasx/xvmskcond.ll
index c61b784..06d4a5d 100644
--- a/llvm/test/CodeGen/LoongArch/lasx/xvmskcond.ll
+++ b/llvm/test/CodeGen/LoongArch/lasx/xvmskcond.ll
@@ -524,9 +524,8 @@ define i8 @xvmsk_eq_v2i64_concat_poison(<2 x i64> %vec) {
 ; CHECK-NEXT:    vpickve2gr.d $a0, $vr0, 0
 ; CHECK-NEXT:    vinsgr2vr.h $vr1, $a0, 0
 ; CHECK-NEXT:    vpickve2gr.d $a0, $vr0, 1
-; CHECK-NEXT:    vinsgr2vr.h $vr0, $a0, 0
-; CHECK-NEXT:    vpackev.h $vr0, $vr0, $vr1
-; CHECK-NEXT:    vslli.h $vr0, $vr0, 15
+; CHECK-NEXT:    vinsgr2vr.h $vr1, $a0, 1
+; CHECK-NEXT:    vslli.h $vr0, $vr1, 15
 ; CHECK-NEXT:    vmskltz.h $vr0, $vr0
 ; CHECK-NEXT:    vpickve2gr.hu $a0, $vr0, 0
 ; CHECK-NEXT:    ret
@@ -539,24 +538,20 @@ define i8 @xvmsk_eq_v2i64_concat_poison(<2 x i64> %vec) {
 define i8 @xvmsk_ne_v4i32_concat_poison(<4 x i32> %vec) {
 ; CHECK-LABEL: xvmsk_ne_v4i32_concat_poison:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    addi.d $sp, $sp, -16
-; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    vseqi.w $vr0, $vr0, 0
 ; CHECK-NEXT:    vrepli.b $vr1, -1
 ; CHECK-NEXT:    vxor.v $vr0, $vr0, $vr1
-; CHECK-NEXT:    vpickve2gr.w $a0, $vr0, 3
-; CHECK-NEXT:    st.h $a0, $sp, 6
-; CHECK-NEXT:    vpickve2gr.w $a0, $vr0, 2
-; CHECK-NEXT:    st.h $a0, $sp, 4
-; CHECK-NEXT:    vpickve2gr.w $a0, $vr0, 1
-; CHECK-NEXT:    st.h $a0, $sp, 2
 ; CHECK-NEXT:    vpickve2gr.w $a0, $vr0, 0
-; CHECK-NEXT:    st.h $a0, $sp, 0
-; CHECK-NEXT:    vld $vr0, $sp, 0
-; CHECK-NEXT:    vslli.h $vr0, $vr0, 15
+; CHECK-NEXT:    vinsgr2vr.h $vr1, $a0, 0
+; CHECK-NEXT:    vpickve2gr.w $a0, $vr0, 1
+; CHECK-NEXT:    vinsgr2vr.h $vr1, $a0, 1
+; CHECK-NEXT:    vpickve2gr.w $a0, $vr0, 2
+; CHECK-NEXT:    vinsgr2vr.h $vr1, $a0, 2
+; CHECK-NEXT:    vpickve2gr.w $a0, $vr0, 3
+; CHECK-NEXT:    vinsgr2vr.h $vr1, $a0, 3
+; CHECK-NEXT:    vslli.h $vr0, $vr1, 15
 ; CHECK-NEXT:    vmskltz.h $vr0, $vr0
 ; CHECK-NEXT:    vpickve2gr.hu $a0, $vr0, 0
-; CHECK-NEXT:    addi.d $sp, $sp, 16
 ; CHECK-NEXT:    ret
   %tobool = icmp ne <4 x i32> %vec, zeroinitializer
   %insertvec = shufflevector <4 x i1> %tobool, <4 x i1> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
@@ -567,23 +562,19 @@ define i8 @xvmsk_ne_v4i32_concat_poison(<4 x i32> %vec) {
 define i8 @xvmsk_ogt_v4f64_concat_poison(<4 x double> %vec) {
 ; CHECK-LABEL: xvmsk_ogt_v4f64_concat_poison:
 ; CHECK:       # %bb.0:
-; CHECK-NEXT:    addi.d $sp, $sp, -16
-; CHECK-NEXT:    .cfi_def_cfa_offset 16
 ; CHECK-NEXT:    xvrepli.b $xr1, 0
 ; CHECK-NEXT:    xvfcmp.clt.d $xr0, $xr1, $xr0
-; CHECK-NEXT:    xvpickve2gr.d $a0, $xr0, 0
-; CHECK-NEXT:    xvpickve2gr.d $a1, $xr0, 1
-; CHECK-NEXT:    xvpickve2gr.d $a2, $xr0, 2
-; CHECK-NEXT:    xvpickve2gr.d $a3, $xr0, 3
-; CHECK-NEXT:    st.h $a3, $sp, 6
-; CHECK-NEXT:    st.h $a2, $sp, 4
-; CHECK-NEXT:    st.h $a1, $sp, 2
-; CHECK-NEXT:    st.h $a0, $sp, 0
-; CHECK-NEXT:    vld $vr0, $sp, 0
+; CHECK-NEXT:    xvpickve2gr.d $a0, $xr0, 3
+; CHECK-NEXT:    xvpickve2gr.d $a1, $xr0, 2
+; CHECK-NEXT:    xvpickve2gr.d $a2, $xr0, 1
+; CHECK-NEXT:    xvpickve2gr.d $a3, $xr0, 0
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a3, 0
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a2, 1
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a1, 2
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a0, 3
 ; CHECK-NEXT:    vslli.h $vr0, $vr0, 15
 ; CHECK-NEXT:    vmskltz.h $vr0, $vr0
 ; CHECK-NEXT:    vpickve2gr.hu $a0, $vr0, 0
-; CHECK-NEXT:    addi.d $sp, $sp, 16
 ; CHECK-NEXT:    ret
   %tobool = fcmp ogt <4 x double> %vec, zeroinitializer
   %insertvec = shufflevector <4 x i1> %tobool, <4 x i1> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
diff --git a/llvm/test/CodeGen/LoongArch/llvm.exp10.ll b/llvm/test/CodeGen/LoongArch/llvm.exp10.ll
index 62ea5cb..030b822c 100644
--- a/llvm/test/CodeGen/LoongArch/llvm.exp10.ll
+++ b/llvm/test/CodeGen/LoongArch/llvm.exp10.ll
@@ -137,20 +137,20 @@ define <2 x float> @exp10_v2f32(<2 x float> %x) #0 {
 ; LA64-NEXT:    addi.d $sp, $sp, -48
 ; LA64-NEXT:    st.d $ra, $sp, 40 # 8-byte Folded Spill
 ; LA64-NEXT:    vst $vr0, $sp, 0 # 16-byte Folded Spill
-; LA64-NEXT:    vreplvei.w $vr0, $vr0, 0
+; LA64-NEXT:    vreplvei.w $vr0, $vr0, 1
 ; LA64-NEXT:    # kill: def $f0 killed $f0 killed $vr0
 ; LA64-NEXT:    pcaddu18i $ra, %call36(exp10f)
 ; LA64-NEXT:    jirl $ra, $ra, 0
 ; LA64-NEXT:    # kill: def $f0 killed $f0 def $vr0
 ; LA64-NEXT:    vst $vr0, $sp, 16 # 16-byte Folded Spill
 ; LA64-NEXT:    vld $vr0, $sp, 0 # 16-byte Folded Reload
-; LA64-NEXT:    vreplvei.w $vr0, $vr0, 1
+; LA64-NEXT:    vreplvei.w $vr0, $vr0, 0
 ; LA64-NEXT:    # kill: def $f0 killed $f0 killed $vr0
 ; LA64-NEXT:    pcaddu18i $ra, %call36(exp10f)
 ; LA64-NEXT:    jirl $ra, $ra, 0
 ; LA64-NEXT:    # kill: def $f0 killed $f0 def $vr0
 ; LA64-NEXT:    vld $vr1, $sp, 16 # 16-byte Folded Reload
-; LA64-NEXT:    vpackev.w $vr0, $vr0, $vr1
+; LA64-NEXT:    vextrins.w $vr0, $vr1, 16
 ; LA64-NEXT:    ld.d $ra, $sp, 40 # 8-byte Folded Reload
 ; LA64-NEXT:    addi.d $sp, $sp, 48
 ; LA64-NEXT:    ret
diff --git a/llvm/test/CodeGen/LoongArch/llvm.sincos.ll b/llvm/test/CodeGen/LoongArch/llvm.sincos.ll
index 383d63c..4ac38a9 100644
--- a/llvm/test/CodeGen/LoongArch/llvm.sincos.ll
+++ b/llvm/test/CodeGen/LoongArch/llvm.sincos.ll
@@ -350,7 +350,7 @@ define { <2 x float>, <2 x float> } @test_sincos_v2f32(<2 x float> %a) #0 {
 ; LA64-NEXT:    addi.d $sp, $sp, -80
 ; LA64-NEXT:    st.d $ra, $sp, 72 # 8-byte Folded Spill
 ; LA64-NEXT:    vst $vr0, $sp, 16 # 16-byte Folded Spill
-; LA64-NEXT:    vreplvei.w $vr0, $vr0, 0
+; LA64-NEXT:    vreplvei.w $vr0, $vr0, 1
 ; LA64-NEXT:    vst $vr0, $sp, 48 # 16-byte Folded Spill
 ; LA64-NEXT:    # kill: def $f0 killed $f0 killed $vr0
 ; LA64-NEXT:    pcaddu18i $ra, %call36(sinf)
@@ -358,14 +358,14 @@ define { <2 x float>, <2 x float> } @test_sincos_v2f32(<2 x float> %a) #0 {
 ; LA64-NEXT:    # kill: def $f0 killed $f0 def $vr0
 ; LA64-NEXT:    vst $vr0, $sp, 32 # 16-byte Folded Spill
 ; LA64-NEXT:    vld $vr0, $sp, 16 # 16-byte Folded Reload
-; LA64-NEXT:    vreplvei.w $vr0, $vr0, 1
+; LA64-NEXT:    vreplvei.w $vr0, $vr0, 0
 ; LA64-NEXT:    vst $vr0, $sp, 16 # 16-byte Folded Spill
 ; LA64-NEXT:    # kill: def $f0 killed $f0 killed $vr0
 ; LA64-NEXT:    pcaddu18i $ra, %call36(sinf)
 ; LA64-NEXT:    jirl $ra, $ra, 0
 ; LA64-NEXT:    # kill: def $f0 killed $f0 def $vr0
 ; LA64-NEXT:    vld $vr1, $sp, 32 # 16-byte Folded Reload
-; LA64-NEXT:    vpackev.w $vr0, $vr0, $vr1
+; LA64-NEXT:    vextrins.w $vr0, $vr1, 16
 ; LA64-NEXT:    vst $vr0, $sp, 32 # 16-byte Folded Spill
 ; LA64-NEXT:    vld $vr0, $sp, 48 # 16-byte Folded Reload
 ; LA64-NEXT:    # kill: def $f0 killed $f0 killed $vr0
@@ -377,9 +377,9 @@ define { <2 x float>, <2 x float> } @test_sincos_v2f32(<2 x float> %a) #0 {
 ; LA64-NEXT:    # kill: def $f0 killed $f0 killed $vr0
 ; LA64-NEXT:    pcaddu18i $ra, %call36(cosf)
 ; LA64-NEXT:    jirl $ra, $ra, 0
-; LA64-NEXT:    # kill: def $f0 killed $f0 def $vr0
-; LA64-NEXT:    vld $vr1, $sp, 48 # 16-byte Folded Reload
-; LA64-NEXT:    vpackev.w $vr1, $vr0, $vr1
+; LA64-NEXT:    fmov.s $fa1, $fa0
+; LA64-NEXT:    vld $vr0, $sp, 48 # 16-byte Folded Reload
+; LA64-NEXT:    vextrins.w $vr1, $vr0, 16
 ; LA64-NEXT:    vld $vr0, $sp, 32 # 16-byte Folded Reload
 ; LA64-NEXT:    ld.d $ra, $sp, 72 # 8-byte Folded Reload
 ; LA64-NEXT:    addi.d $sp, $sp, 80
@@ -439,48 +439,60 @@ define { <3 x float>, <3 x float> } @test_sincos_v3f32(<3 x float> %a) #0 {
 ;
 ; LA64-LABEL: test_sincos_v3f32:
 ; LA64:       # %bb.0:
-; LA64-NEXT:    addi.d $sp, $sp, -112
-; LA64-NEXT:    st.d $ra, $sp, 104 # 8-byte Folded Spill
+; LA64-NEXT:    addi.d $sp, $sp, -96
+; LA64-NEXT:    st.d $ra, $sp, 88 # 8-byte Folded Spill
 ; LA64-NEXT:    vst $vr0, $sp, 48 # 16-byte Folded Spill
-; LA64-NEXT:    vreplvei.w $vr0, $vr0, 2
+; LA64-NEXT:    vreplvei.w $vr0, $vr0, 1
 ; LA64-NEXT:    vst $vr0, $sp, 32 # 16-byte Folded Spill
 ; LA64-NEXT:    # kill: def $f0 killed $f0 killed $vr0
 ; LA64-NEXT:    pcaddu18i $ra, %call36(sinf)
 ; LA64-NEXT:    jirl $ra, $ra, 0
-; LA64-NEXT:    fst.s $fa0, $sp, 88
+; LA64-NEXT:    # kill: def $f0 killed $f0 def $vr0
+; LA64-NEXT:    vst $vr0, $sp, 64 # 16-byte Folded Spill
 ; LA64-NEXT:    vld $vr0, $sp, 48 # 16-byte Folded Reload
-; LA64-NEXT:    vreplvei.w $vr0, $vr0, 1
+; LA64-NEXT:    vreplvei.w $vr0, $vr0, 0
 ; LA64-NEXT:    vst $vr0, $sp, 16 # 16-byte Folded Spill
 ; LA64-NEXT:    # kill: def $f0 killed $f0 killed $vr0
 ; LA64-NEXT:    pcaddu18i $ra, %call36(sinf)
 ; LA64-NEXT:    jirl $ra, $ra, 0
-; LA64-NEXT:    fst.s $fa0, $sp, 84
+; LA64-NEXT:    # kill: def $f0 killed $f0 def $vr0
+; LA64-NEXT:    vld $vr1, $sp, 64 # 16-byte Folded Reload
+; LA64-NEXT:    vextrins.w $vr0, $vr1, 16
+; LA64-NEXT:    vst $vr0, $sp, 64 # 16-byte Folded Spill
 ; LA64-NEXT:    vld $vr0, $sp, 48 # 16-byte Folded Reload
-; LA64-NEXT:    vreplvei.w $vr0, $vr0, 0
+; LA64-NEXT:    vreplvei.w $vr0, $vr0, 2
 ; LA64-NEXT:    vst $vr0, $sp, 48 # 16-byte Folded Spill
 ; LA64-NEXT:    # kill: def $f0 killed $f0 killed $vr0
 ; LA64-NEXT:    pcaddu18i $ra, %call36(sinf)
 ; LA64-NEXT:    jirl $ra, $ra, 0
-; LA64-NEXT:    fst.s $fa0, $sp, 80
+; LA64-NEXT:    # kill: def $f0 killed $f0 def $vr0
+; LA64-NEXT:    vld $vr1, $sp, 64 # 16-byte Folded Reload
+; LA64-NEXT:    vextrins.w $vr1, $vr0, 32
+; LA64-NEXT:    vst $vr1, $sp, 64 # 16-byte Folded Spill
 ; LA64-NEXT:    vld $vr0, $sp, 32 # 16-byte Folded Reload
 ; LA64-NEXT:    # kill: def $f0 killed $f0 killed $vr0
 ; LA64-NEXT:    pcaddu18i $ra, %call36(cosf)
 ; LA64-NEXT:    jirl $ra, $ra, 0
-; LA64-NEXT:    fst.s $fa0, $sp, 72
+; LA64-NEXT:    # kill: def $f0 killed $f0 def $vr0
+; LA64-NEXT:    vst $vr0, $sp, 32 # 16-byte Folded Spill
 ; LA64-NEXT:    vld $vr0, $sp, 16 # 16-byte Folded Reload
 ; LA64-NEXT:    # kill: def $f0 killed $f0 killed $vr0
 ; LA64-NEXT:    pcaddu18i $ra, %call36(cosf)
 ; LA64-NEXT:    jirl $ra, $ra, 0
-; LA64-NEXT:    fst.s $fa0, $sp, 68
+; LA64-NEXT:    # kill: def $f0 killed $f0 def $vr0
+; LA64-NEXT:    vld $vr1, $sp, 32 # 16-byte Folded Reload
+; LA64-NEXT:    vextrins.w $vr0, $vr1, 16
+; LA64-NEXT:    vst $vr0, $sp, 32 # 16-byte Folded Spill
 ; LA64-NEXT:    vld $vr0, $sp, 48 # 16-byte Folded Reload
 ; LA64-NEXT:    # kill: def $f0 killed $f0 killed $vr0
 ; LA64-NEXT:    pcaddu18i $ra, %call36(cosf)
 ; LA64-NEXT:    jirl $ra, $ra, 0
-; LA64-NEXT:    fst.s $fa0, $sp, 64
-; LA64-NEXT:    vld $vr0, $sp, 80
-; LA64-NEXT:    vld $vr1, $sp, 64
-; LA64-NEXT:    ld.d $ra, $sp, 104 # 8-byte Folded Reload
-; LA64-NEXT:    addi.d $sp, $sp, 112
+; LA64-NEXT:    # kill: def $f0 killed $f0 def $vr0
+; LA64-NEXT:    vld $vr1, $sp, 32 # 16-byte Folded Reload
+; LA64-NEXT:    vextrins.w $vr1, $vr0, 32
+; LA64-NEXT:    vld $vr0, $sp, 64 # 16-byte Folded Reload
+; LA64-NEXT:    ld.d $ra, $sp, 88 # 8-byte Folded Reload
+; LA64-NEXT:    addi.d $sp, $sp, 96
 ; LA64-NEXT:    ret
   %result = call { <3 x float>, <3 x float> } @llvm.sincos.v3f32(<3 x float> %a)
   ret { <3 x float>, <3 x float> } %result
diff --git a/llvm/test/CodeGen/LoongArch/lsx/build-vector.ll b/llvm/test/CodeGen/LoongArch/lsx/build-vector.ll
index 36d337b..9517558 100644
--- a/llvm/test/CodeGen/LoongArch/lsx/build-vector.ll
+++ b/llvm/test/CodeGen/LoongArch/lsx/build-vector.ll
@@ -275,16 +275,13 @@ entry:
 define void @buildvector_v16i8_partial(ptr %dst, i8 %a2, i8 %a6, i8 %a8, i8 %a11, i8 %a12, i8 %a15) nounwind {
 ; CHECK-LABEL: buildvector_v16i8_partial:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    addi.d $sp, $sp, -16
-; CHECK-NEXT:    st.b $a6, $sp, 15
-; CHECK-NEXT:    st.b $a5, $sp, 12
-; CHECK-NEXT:    st.b $a4, $sp, 11
-; CHECK-NEXT:    st.b $a3, $sp, 8
-; CHECK-NEXT:    st.b $a2, $sp, 6
-; CHECK-NEXT:    st.b $a1, $sp, 2
-; CHECK-NEXT:    vld $vr0, $sp, 0
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a1, 2
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a2, 6
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a3, 8
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a4, 11
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a5, 12
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a6, 15
 ; CHECK-NEXT:    vst $vr0, $a0, 0
-; CHECK-NEXT:    addi.d $sp, $sp, 16
 ; CHECK-NEXT:    ret
 entry:
   %ins0  = insertelement <16 x i8> undef,  i8 undef, i32 0
@@ -310,23 +307,15 @@ entry:
 define void @buildvector_v16i8_with_constant(ptr %dst, i8 %a0, i8 %a4, i8 %a6, i8 %a8, i8 %a11, i8 %a12, i8 %a15) nounwind {
 ; CHECK-LABEL: buildvector_v16i8_with_constant:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    addi.d $sp, $sp, -16
-; CHECK-NEXT:    st.b $a7, $sp, 15
-; CHECK-NEXT:    st.h $zero, $sp, 13
-; CHECK-NEXT:    st.b $a6, $sp, 12
-; CHECK-NEXT:    st.b $a5, $sp, 11
-; CHECK-NEXT:    st.h $zero, $sp, 9
-; CHECK-NEXT:    st.b $a4, $sp, 8
-; CHECK-NEXT:    st.b $zero, $sp, 7
-; CHECK-NEXT:    st.b $a3, $sp, 6
-; CHECK-NEXT:    st.b $zero, $sp, 5
-; CHECK-NEXT:    st.b $a2, $sp, 4
-; CHECK-NEXT:    st.b $zero, $sp, 3
-; CHECK-NEXT:    st.h $zero, $sp, 1
-; CHECK-NEXT:    st.b $a1, $sp, 0
-; CHECK-NEXT:    vld $vr0, $sp, 0
-; CHECK-NEXT:    vst $vr0, $a0, 0
-; CHECK-NEXT:    addi.d $sp, $sp, 16
+; CHECK-NEXT:    vrepli.b $vr0, 0
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a1, 0
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a2, 4
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a3, 6
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a4, 8
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a5, 11
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a6, 12
+; CHECK-NEXT:    vinsgr2vr.b $vr0, $a7, 15
+; CHECK-NEXT:    vst $vr0, $a0, 0
 ; CHECK-NEXT:    ret
 entry:
   %ins0  = insertelement <16 x i8> undef,  i8  %a0, i32 0
@@ -379,14 +368,11 @@ entry:
 define void @buildvector_v8i16_partial(ptr %dst, i16 %a1, i16 %a3, i16 %a4, i16 %a5) nounwind {
 ; CHECK-LABEL: buildvector_v8i16_partial:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    addi.d $sp, $sp, -16
-; CHECK-NEXT:    st.h $a4, $sp, 10
-; CHECK-NEXT:    st.h $a3, $sp, 8
-; CHECK-NEXT:    st.h $a2, $sp, 6
-; CHECK-NEXT:    st.h $a1, $sp, 2
-; CHECK-NEXT:    vld $vr0, $sp, 0
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a1, 1
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a2, 3
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a3, 4
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a4, 5
 ; CHECK-NEXT:    vst $vr0, $a0, 0
-; CHECK-NEXT:    addi.d $sp, $sp, 16
 ; CHECK-NEXT:    ret
 entry:
   %ins0 = insertelement <8 x i16> undef, i16 undef, i32 0
@@ -404,16 +390,12 @@ entry:
 define void @buildvector_v8i16_with_constant(ptr %dst, i16 %a0, i16 %a3, i16 %a4, i16 %a5) nounwind {
 ; CHECK-LABEL: buildvector_v8i16_with_constant:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    addi.d $sp, $sp, -16
-; CHECK-NEXT:    st.h $zero, $sp, 12
-; CHECK-NEXT:    st.h $a4, $sp, 10
-; CHECK-NEXT:    st.h $a3, $sp, 8
-; CHECK-NEXT:    st.h $a2, $sp, 6
-; CHECK-NEXT:    st.h $zero, $sp, 2
-; CHECK-NEXT:    st.h $a1, $sp, 0
-; CHECK-NEXT:    vld $vr0, $sp, 0
+; CHECK-NEXT:    vrepli.b $vr0, 0
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a1, 0
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a2, 3
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a3, 4
+; CHECK-NEXT:    vinsgr2vr.h $vr0, $a4, 5
 ; CHECK-NEXT:    vst $vr0, $a0, 0
-; CHECK-NEXT:    addi.d $sp, $sp, 16
 ; CHECK-NEXT:    ret
 entry:
   %ins0 = insertelement <8 x i16> undef, i16   %a0, i32 0
@@ -449,11 +431,8 @@ entry:
 define void @buildvector_v4i32_partial(ptr %dst, i32 %a0, i32 %a3) nounwind {
 ; CHECK-LABEL: buildvector_v4i32_partial:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    pcalau12i $a3, %pc_hi20(.LCPI25_0)
-; CHECK-NEXT:    vld $vr0, $a3, %pc_lo12(.LCPI25_0)
-; CHECK-NEXT:    vinsgr2vr.w $vr1, $a1, 0
-; CHECK-NEXT:    vinsgr2vr.w $vr2, $a2, 0
-; CHECK-NEXT:    vshuf.w $vr0, $vr2, $vr1
+; CHECK-NEXT:    vinsgr2vr.w $vr0, $a1, 0
+; CHECK-NEXT:    vinsgr2vr.w $vr0, $a2, 3
 ; CHECK-NEXT:    vst $vr0, $a0, 0
 ; CHECK-NEXT:    ret
 entry:
@@ -468,15 +447,11 @@ entry:
 define void @buildvector_v4i32_with_constant(ptr %dst, i32 %a0, i32 %a2, i32 %a3) nounwind {
 ; CHECK-LABEL: buildvector_v4i32_with_constant:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    addi.d $sp, $sp, -16
-; CHECK-NEXT:    st.w $a3, $sp, 12
-; CHECK-NEXT:    st.w $a2, $sp, 8
-; CHECK-NEXT:    ori $a2, $zero, 2
-; CHECK-NEXT:    st.w $a2, $sp, 4
-; CHECK-NEXT:    st.w $a1, $sp, 0
-; CHECK-NEXT:    vld $vr0, $sp, 0
+; CHECK-NEXT:    vrepli.w $vr0, 2
+; CHECK-NEXT:    vinsgr2vr.w $vr0, $a1, 0
+; CHECK-NEXT:    vinsgr2vr.w $vr0, $a2, 2
+; CHECK-NEXT:    vinsgr2vr.w $vr0, $a3, 3
 ; CHECK-NEXT:    vst $vr0, $a0, 0
-; CHECK-NEXT:    addi.d $sp, $sp, 16
 ; CHECK-NEXT:    ret
 entry:
   %ins0 = insertelement <4 x i32> undef, i32 %a0, i32 0
@@ -517,9 +492,8 @@ entry:
 define void @buildvector_v2i64_with_constant(ptr %dst, i64 %a1) nounwind {
 ; CHECK-LABEL: buildvector_v2i64_with_constant:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    vinsgr2vr.d $vr0, $zero, 0
-; CHECK-NEXT:    vinsgr2vr.d $vr1, $a1, 0
-; CHECK-NEXT:    vpackev.d $vr0, $vr1, $vr0
+; CHECK-NEXT:    vrepli.b $vr0, 0
+; CHECK-NEXT:    vinsgr2vr.d $vr0, $a1, 1
 ; CHECK-NEXT:    vst $vr0, $a0, 0
 ; CHECK-NEXT:    ret
 entry:
@@ -553,12 +527,10 @@ entry:
 define void @buildvector_v4f32_partial(ptr %dst, float %a0, float %a3) nounwind {
 ; CHECK-LABEL: buildvector_v4f32_partial:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    pcalau12i $a1, %pc_hi20(.LCPI31_0)
-; CHECK-NEXT:    vld $vr2, $a1, %pc_lo12(.LCPI31_0)
 ; CHECK-NEXT:    # kill: def $f1 killed $f1 def $vr1
 ; CHECK-NEXT:    # kill: def $f0 killed $f0 def $vr0
-; CHECK-NEXT:    vshuf.w $vr2, $vr1, $vr0
-; CHECK-NEXT:    vst $vr2, $a0, 0
+; CHECK-NEXT:    vextrins.w $vr0, $vr1, 48
+; CHECK-NEXT:    vst $vr0, $a0, 0
 ; CHECK-NEXT:    ret
 entry:
   %ins0 = insertelement <4 x float> undef, float   %a0, i32 0
@@ -572,15 +544,14 @@ entry:
 define void @buildvector_v4f32_with_constant(ptr %dst, float %a1, float %a2, float %a3) nounwind {
 ; CHECK-LABEL: buildvector_v4f32_with_constant:
 ; CHECK:       # %bb.0: # %entry
-; CHECK-NEXT:    addi.d $sp, $sp, -16
-; CHECK-NEXT:    fst.s $fa2, $sp, 12
-; CHECK-NEXT:    fst.s $fa1, $sp, 8
-; CHECK-NEXT:    fst.s $fa0, $sp, 4
-; CHECK-NEXT:    movgr2fr.w $fa0, $zero
-; CHECK-NEXT:    fst.s $fa0, $sp, 0
-; CHECK-NEXT:    vld $vr0, $sp, 0
-; CHECK-NEXT:    vst $vr0, $a0, 0
-; CHECK-NEXT:    addi.d $sp, $sp, 16
+; CHECK-NEXT:    # kill: def $f2 killed $f2 def $vr2
+; CHECK-NEXT:    # kill: def $f1 killed $f1 def $vr1
+; CHECK-NEXT:    # kill: def $f0 killed $f0 def $vr0
+; CHECK-NEXT:    vrepli.b $vr3, 0
+; CHECK-NEXT:    vextrins.w $vr3, $vr0, 16
+; CHECK-NEXT:    vextrins.w $vr3, $vr1, 32
+; CHECK-NEXT:    vextrins.w $vr3, $vr2, 48
+; CHECK-NEXT:    vst $vr3, $a0, 0
 ; CHECK-NEXT:    ret
 entry:
   %ins0 = insertelement <4 x float> undef, float 0.0, i32 0
@@ -610,7 +581,7 @@ define void @buildvector_v2f64_partial(ptr %dst, double %a1) nounwind {
 ; CHECK-LABEL: buildvector_v2f64_partial:
 ; CHECK:       # %bb.0: # %entry
 ; CHECK-NEXT:    # kill: def $f0_64 killed $f0_64 def $vr0
-; CHECK-NEXT:    vreplvei.d $vr0, $vr0, 0
+; CHECK-NEXT:    vextrins.d $vr0, $vr0, 16
 ; CHECK-NEXT:    vst $vr0, $a0, 0
 ; CHECK-NEXT:    ret
 entry:
diff --git a/llvm/test/CodeGen/Mips/abiflags-soft-float.ll b/llvm/test/CodeGen/Mips/abiflags-soft-float.ll
new file mode 100644
index 0000000..01821f2
--- /dev/null
+++ b/llvm/test/CodeGen/Mips/abiflags-soft-float.ll
@@ -0,0 +1,12 @@
+; RUN: llc -filetype=obj -mtriple mipsel-unknown-linux -mcpu=mips32 %s -o tmp.o
+; RUN: llvm-readobj -A tmp.o | FileCheck %s -check-prefix=OBJ
+; RUN: llc -filetype=asm -mtriple mipsel-unknown-linux -mcpu=mips32 %s -o - | \
+; RUN: FileCheck %s -check-prefix=ASM
+
+; OBJ: FP ABI: Soft float
+; ASM: .module	softfloat 
+
+define dso_local void @asm_is_null() "use-soft-float"="true" {
+  call void asm sideeffect "", ""()
+  ret void
+}
diff --git a/llvm/test/CodeGen/Mips/llvm.frexp.ll b/llvm/test/CodeGen/Mips/llvm.frexp.ll
new file mode 100644
index 0000000..3226766
--- /dev/null
+++ b/llvm/test/CodeGen/Mips/llvm.frexp.ll
@@ -0,0 +1,651 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -mtriple=mipsel < %s | FileCheck -check-prefix=MIPSEL %s
+; RUN: llc -mtriple=mips < %s | FileCheck %s -check-prefixes=SOFT-FLOAT-32
+; RUN: llc -mtriple=mips64 < %s | FileCheck %s -check-prefixes=SOFT-FLOAT-64
+
+define { half, i32 } @test_frexp_f16_i32(half %a) nounwind {
+; MIPSEL-LABEL: test_frexp_f16_i32:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -24
+; MIPSEL-NEXT:    sw $ra, 20($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    jal __extendhfsf2
+; MIPSEL-NEXT:    nop
+; MIPSEL-NEXT:    addiu $5, $sp, 16
+; MIPSEL-NEXT:    jal frexpf
+; MIPSEL-NEXT:    mov.s $f12, $f0
+; MIPSEL-NEXT:    jal __truncsfhf2
+; MIPSEL-NEXT:    mov.s $f12, $f0
+; MIPSEL-NEXT:    lw $3, 16($sp)
+; MIPSEL-NEXT:    lw $ra, 20($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 24
+;
+; SOFT-FLOAT-32-LABEL: test_frexp_f16_i32:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -24
+; SOFT-FLOAT-32-NEXT:    sw $ra, 20($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    jal __extendhfsf2
+; SOFT-FLOAT-32-NEXT:    nop
+; SOFT-FLOAT-32-NEXT:    addiu $5, $sp, 16
+; SOFT-FLOAT-32-NEXT:    jal frexpf
+; SOFT-FLOAT-32-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-32-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-32-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-32-NEXT:    lw $3, 16($sp)
+; SOFT-FLOAT-32-NEXT:    lw $ra, 20($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 24
+;
+; SOFT-FLOAT-64-LABEL: test_frexp_f16_i32:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -16
+; SOFT-FLOAT-64-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    jal __extendhfsf2
+; SOFT-FLOAT-64-NEXT:    sll $4, $4, 0
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $sp, 4
+; SOFT-FLOAT-64-NEXT:    jal frexpf
+; SOFT-FLOAT-64-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-64-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-64-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-64-NEXT:    lw $3, 4($sp)
+; SOFT-FLOAT-64-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 16
+%result = call { half, i32 } @llvm.frexp.f16.i32(half %a)
+  ret { half, i32 } %result
+}
+
+define { <2 x half>, <2 x i32> } @test_frexp_v2f16_v2i32(<2 x half> %a) nounwind {
+; MIPSEL-LABEL: test_frexp_v2f16_v2i32:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -32
+; MIPSEL-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $18, 24($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $17, 20($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $16, 16($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    move $16, $5
+; MIPSEL-NEXT:    move $17, $4
+; MIPSEL-NEXT:    jal __extendhfsf2
+; MIPSEL-NEXT:    srl $4, $5, 16
+; MIPSEL-NEXT:    addiu $5, $17, 12
+; MIPSEL-NEXT:    jal frexpf
+; MIPSEL-NEXT:    mov.s $f12, $f0
+; MIPSEL-NEXT:    jal __truncsfhf2
+; MIPSEL-NEXT:    mov.s $f12, $f0
+; MIPSEL-NEXT:    move $18, $2
+; MIPSEL-NEXT:    jal __extendhfsf2
+; MIPSEL-NEXT:    move $4, $16
+; MIPSEL-NEXT:    addiu $5, $17, 8
+; MIPSEL-NEXT:    jal frexpf
+; MIPSEL-NEXT:    mov.s $f12, $f0
+; MIPSEL-NEXT:    sh $18, 2($17)
+; MIPSEL-NEXT:    jal __truncsfhf2
+; MIPSEL-NEXT:    mov.s $f12, $f0
+; MIPSEL-NEXT:    sh $2, 0($17)
+; MIPSEL-NEXT:    lw $16, 16($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $17, 20($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $18, 24($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 32
+;
+; SOFT-FLOAT-32-LABEL: test_frexp_v2f16_v2i32:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -32
+; SOFT-FLOAT-32-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $18, 24($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $17, 20($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $16, 16($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    move $16, $5
+; SOFT-FLOAT-32-NEXT:    move $17, $4
+; SOFT-FLOAT-32-NEXT:    jal __extendhfsf2
+; SOFT-FLOAT-32-NEXT:    move $4, $5
+; SOFT-FLOAT-32-NEXT:    addiu $5, $17, 12
+; SOFT-FLOAT-32-NEXT:    jal frexpf
+; SOFT-FLOAT-32-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-32-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-32-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-32-NEXT:    move $18, $2
+; SOFT-FLOAT-32-NEXT:    jal __extendhfsf2
+; SOFT-FLOAT-32-NEXT:    srl $4, $16, 16
+; SOFT-FLOAT-32-NEXT:    addiu $5, $17, 8
+; SOFT-FLOAT-32-NEXT:    jal frexpf
+; SOFT-FLOAT-32-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-32-NEXT:    sh $18, 2($17)
+; SOFT-FLOAT-32-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-32-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-32-NEXT:    sh $2, 0($17)
+; SOFT-FLOAT-32-NEXT:    lw $16, 16($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $17, 20($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $18, 24($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 32
+;
+; SOFT-FLOAT-64-LABEL: test_frexp_v2f16_v2i32:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -32
+; SOFT-FLOAT-64-NEXT:    sd $ra, 24($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $18, 16($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $17, 8($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $16, 0($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    move $16, $5
+; SOFT-FLOAT-64-NEXT:    move $17, $4
+; SOFT-FLOAT-64-NEXT:    jal __extendhfsf2
+; SOFT-FLOAT-64-NEXT:    sll $4, $5, 0
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $17, 12
+; SOFT-FLOAT-64-NEXT:    jal frexpf
+; SOFT-FLOAT-64-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-64-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-64-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-64-NEXT:    move $18, $2
+; SOFT-FLOAT-64-NEXT:    sll $1, $16, 0
+; SOFT-FLOAT-64-NEXT:    jal __extendhfsf2
+; SOFT-FLOAT-64-NEXT:    srl $4, $1, 16
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $17, 8
+; SOFT-FLOAT-64-NEXT:    jal frexpf
+; SOFT-FLOAT-64-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-64-NEXT:    sh $18, 2($17)
+; SOFT-FLOAT-64-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-64-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-64-NEXT:    sh $2, 0($17)
+; SOFT-FLOAT-64-NEXT:    ld $16, 0($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $17, 8($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $18, 16($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $ra, 24($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 32
+  %result = call { <2 x half>, <2 x i32> } @llvm.frexp.v2f16.v2i32(<2 x half> %a)
+  ret { <2 x half>, <2 x i32> } %result
+}
+
+define { float, i32 } @test_frexp_f32_i32(float %a) nounwind {
+; MIPSEL-LABEL: test_frexp_f32_i32:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -24
+; MIPSEL-NEXT:    sw $ra, 20($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    jal frexpf
+; MIPSEL-NEXT:    addiu $5, $sp, 16
+; MIPSEL-NEXT:    lw $2, 16($sp)
+; MIPSEL-NEXT:    lw $ra, 20($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 24
+;
+; SOFT-FLOAT-32-LABEL: test_frexp_f32_i32:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -24
+; SOFT-FLOAT-32-NEXT:    sw $ra, 20($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    jal frexpf
+; SOFT-FLOAT-32-NEXT:    addiu $5, $sp, 16
+; SOFT-FLOAT-32-NEXT:    lw $2, 16($sp)
+; SOFT-FLOAT-32-NEXT:    lw $ra, 20($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 24
+;
+; SOFT-FLOAT-64-LABEL: test_frexp_f32_i32:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -16
+; SOFT-FLOAT-64-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    jal frexpf
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $sp, 4
+; SOFT-FLOAT-64-NEXT:    lw $2, 4($sp)
+; SOFT-FLOAT-64-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 16
+  %result = call { float, i32 } @llvm.frexp.f32.i32(float %a)
+  ret { float, i32 } %result
+}
+
+define { float, i32 } @test_frexp_f32_i32_tailcall(float %a) nounwind {
+; MIPSEL-LABEL: test_frexp_f32_i32_tailcall:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -24
+; MIPSEL-NEXT:    sw $ra, 20($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    jal frexpf
+; MIPSEL-NEXT:    addiu $5, $sp, 16
+; MIPSEL-NEXT:    lw $2, 16($sp)
+; MIPSEL-NEXT:    lw $ra, 20($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 24
+;
+; SOFT-FLOAT-32-LABEL: test_frexp_f32_i32_tailcall:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -24
+; SOFT-FLOAT-32-NEXT:    sw $ra, 20($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    jal frexpf
+; SOFT-FLOAT-32-NEXT:    addiu $5, $sp, 16
+; SOFT-FLOAT-32-NEXT:    lw $2, 16($sp)
+; SOFT-FLOAT-32-NEXT:    lw $ra, 20($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 24
+;
+; SOFT-FLOAT-64-LABEL: test_frexp_f32_i32_tailcall:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -16
+; SOFT-FLOAT-64-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    jal frexpf
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $sp, 4
+; SOFT-FLOAT-64-NEXT:    lw $2, 4($sp)
+; SOFT-FLOAT-64-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 16
+  %result = tail call { float, i32 } @llvm.frexp.f32.i32(float %a)
+  ret { float, i32 } %result
+}
+
+define { <2 x float>, <2 x i32> } @test_frexp_v2f32_v2i32(<2 x float> %a) nounwind {
+; MIPSEL-LABEL: test_frexp_v2f32_v2i32:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -32
+; MIPSEL-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $17, 24($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $16, 20($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    move $16, $6
+; MIPSEL-NEXT:    move $17, $4
+; MIPSEL-NEXT:    mtc1 $7, $f12
+; MIPSEL-NEXT:    jal frexpf
+; MIPSEL-NEXT:    addiu $5, $4, 12
+; MIPSEL-NEXT:    swc1 $f0, 4($17)
+; MIPSEL-NEXT:    mtc1 $16, $f12
+; MIPSEL-NEXT:    jal frexpf
+; MIPSEL-NEXT:    addiu $5, $17, 8
+; MIPSEL-NEXT:    swc1 $f0, 0($17)
+; MIPSEL-NEXT:    lw $16, 20($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $17, 24($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 32
+;
+; SOFT-FLOAT-32-LABEL: test_frexp_v2f32_v2i32:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -32
+; SOFT-FLOAT-32-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $17, 24($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $16, 20($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    move $16, $6
+; SOFT-FLOAT-32-NEXT:    move $17, $4
+; SOFT-FLOAT-32-NEXT:    mtc1 $7, $f12
+; SOFT-FLOAT-32-NEXT:    jal frexpf
+; SOFT-FLOAT-32-NEXT:    addiu $5, $4, 12
+; SOFT-FLOAT-32-NEXT:    swc1 $f0, 4($17)
+; SOFT-FLOAT-32-NEXT:    mtc1 $16, $f12
+; SOFT-FLOAT-32-NEXT:    jal frexpf
+; SOFT-FLOAT-32-NEXT:    addiu $5, $17, 8
+; SOFT-FLOAT-32-NEXT:    swc1 $f0, 0($17)
+; SOFT-FLOAT-32-NEXT:    lw $16, 20($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $17, 24($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 32
+;
+; SOFT-FLOAT-64-LABEL: test_frexp_v2f32_v2i32:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -32
+; SOFT-FLOAT-64-NEXT:    sd $ra, 24($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $16, 16($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    move $16, $4
+; SOFT-FLOAT-64-NEXT:    sll $1, $16, 0
+; SOFT-FLOAT-64-NEXT:    mtc1 $1, $f12
+; SOFT-FLOAT-64-NEXT:    jal frexpf
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $sp, 8
+; SOFT-FLOAT-64-NEXT:    mfc1 $1, $f0
+; SOFT-FLOAT-64-NEXT:    dsll $1, $1, 32
+; SOFT-FLOAT-64-NEXT:    dsrl $2, $16, 32
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $sp, 12
+; SOFT-FLOAT-64-NEXT:    dsrl $16, $1, 32
+; SOFT-FLOAT-64-NEXT:    sll $1, $2, 0
+; SOFT-FLOAT-64-NEXT:    jal frexpf
+; SOFT-FLOAT-64-NEXT:    mtc1 $1, $f12
+; SOFT-FLOAT-64-NEXT:    mfc1 $1, $f0
+; SOFT-FLOAT-64-NEXT:    dsll $1, $1, 32
+; SOFT-FLOAT-64-NEXT:    or $2, $16, $1
+; SOFT-FLOAT-64-NEXT:    lw $1, 12($sp)
+; SOFT-FLOAT-64-NEXT:    dsll $1, $1, 32
+; SOFT-FLOAT-64-NEXT:    lw $3, 8($sp)
+; SOFT-FLOAT-64-NEXT:    dsll $3, $3, 32
+; SOFT-FLOAT-64-NEXT:    dsrl $3, $3, 32
+; SOFT-FLOAT-64-NEXT:    or $3, $3, $1
+; SOFT-FLOAT-64-NEXT:    ld $16, 16($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $ra, 24($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 32
+  %result = call { <2 x float>, <2 x i32> } @llvm.frexp.v2f32.v2i32(<2 x float> %a)
+  ret { <2 x float>, <2 x i32> } %result
+}
+
+define { double, i32 } @test_frexp_f64_i32(double %a) nounwind {
+; MIPSEL-LABEL: test_frexp_f64_i32:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -24
+; MIPSEL-NEXT:    sw $ra, 20($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    jal frexp
+; MIPSEL-NEXT:    addiu $6, $sp, 16
+; MIPSEL-NEXT:    lw $2, 16($sp)
+; MIPSEL-NEXT:    lw $ra, 20($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 24
+;
+; SOFT-FLOAT-32-LABEL: test_frexp_f64_i32:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -24
+; SOFT-FLOAT-32-NEXT:    sw $ra, 20($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    jal frexp
+; SOFT-FLOAT-32-NEXT:    addiu $6, $sp, 16
+; SOFT-FLOAT-32-NEXT:    lw $2, 16($sp)
+; SOFT-FLOAT-32-NEXT:    lw $ra, 20($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 24
+;
+; SOFT-FLOAT-64-LABEL: test_frexp_f64_i32:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -16
+; SOFT-FLOAT-64-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    jal frexp
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $sp, 4
+; SOFT-FLOAT-64-NEXT:    lw $2, 4($sp)
+; SOFT-FLOAT-64-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 16
+  %result = call { double, i32 } @llvm.frexp.f64.i32(double %a)
+  ret { double, i32 } %result
+}
+
+define { <2 x double>, <2 x i32> } @test_frexp_v2f64_v2i32(<2 x double> %a) nounwind {
+; MIPSEL-LABEL: test_frexp_v2f64_v2i32:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -48
+; MIPSEL-NEXT:    sw $ra, 44($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $18, 40($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $17, 36($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $16, 32($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    move $16, $7
+; MIPSEL-NEXT:    move $17, $6
+; MIPSEL-NEXT:    move $18, $4
+; MIPSEL-NEXT:    lw $1, 64($sp)
+; MIPSEL-NEXT:    lw $2, 68($sp)
+; MIPSEL-NEXT:    sw $2, 28($sp)
+; MIPSEL-NEXT:    sw $1, 24($sp)
+; MIPSEL-NEXT:    addiu $6, $4, 20
+; MIPSEL-NEXT:    jal frexp
+; MIPSEL-NEXT:    ldc1 $f12, 24($sp)
+; MIPSEL-NEXT:    sdc1 $f0, 8($18)
+; MIPSEL-NEXT:    sw $16, 20($sp)
+; MIPSEL-NEXT:    sw $17, 16($sp)
+; MIPSEL-NEXT:    addiu $6, $18, 16
+; MIPSEL-NEXT:    jal frexp
+; MIPSEL-NEXT:    ldc1 $f12, 16($sp)
+; MIPSEL-NEXT:    sdc1 $f0, 0($18)
+; MIPSEL-NEXT:    lw $16, 32($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $17, 36($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $18, 40($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $ra, 44($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 48
+;
+; SOFT-FLOAT-32-LABEL: test_frexp_v2f64_v2i32:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -48
+; SOFT-FLOAT-32-NEXT:    sw $ra, 44($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $18, 40($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $17, 36($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $16, 32($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    move $16, $7
+; SOFT-FLOAT-32-NEXT:    move $17, $6
+; SOFT-FLOAT-32-NEXT:    move $18, $4
+; SOFT-FLOAT-32-NEXT:    lw $1, 64($sp)
+; SOFT-FLOAT-32-NEXT:    lw $2, 68($sp)
+; SOFT-FLOAT-32-NEXT:    sw $2, 28($sp)
+; SOFT-FLOAT-32-NEXT:    sw $1, 24($sp)
+; SOFT-FLOAT-32-NEXT:    addiu $6, $4, 20
+; SOFT-FLOAT-32-NEXT:    jal frexp
+; SOFT-FLOAT-32-NEXT:    ldc1 $f12, 24($sp)
+; SOFT-FLOAT-32-NEXT:    sdc1 $f0, 8($18)
+; SOFT-FLOAT-32-NEXT:    sw $16, 20($sp)
+; SOFT-FLOAT-32-NEXT:    sw $17, 16($sp)
+; SOFT-FLOAT-32-NEXT:    addiu $6, $18, 16
+; SOFT-FLOAT-32-NEXT:    jal frexp
+; SOFT-FLOAT-32-NEXT:    ldc1 $f12, 16($sp)
+; SOFT-FLOAT-32-NEXT:    sdc1 $f0, 0($18)
+; SOFT-FLOAT-32-NEXT:    lw $16, 32($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $17, 36($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $18, 40($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $ra, 44($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 48
+;
+; SOFT-FLOAT-64-LABEL: test_frexp_v2f64_v2i32:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -32
+; SOFT-FLOAT-64-NEXT:    sd $ra, 24($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $17, 16($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $16, 8($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    move $16, $5
+; SOFT-FLOAT-64-NEXT:    move $17, $4
+; SOFT-FLOAT-64-NEXT:    dmtc1 $6, $f12
+; SOFT-FLOAT-64-NEXT:    jal frexp
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $4, 20
+; SOFT-FLOAT-64-NEXT:    sdc1 $f0, 8($17)
+; SOFT-FLOAT-64-NEXT:    dmtc1 $16, $f12
+; SOFT-FLOAT-64-NEXT:    jal frexp
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $17, 16
+; SOFT-FLOAT-64-NEXT:    sdc1 $f0, 0($17)
+; SOFT-FLOAT-64-NEXT:    ld $16, 8($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $17, 16($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $ra, 24($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 32
+  %result = call { <2 x double>, <2 x i32> } @llvm.frexp.v2f64.v2i32(<2 x double> %a)
+  ret { <2 x double>, <2 x i32> } %result
+}
+
+define { fp128, i32 } @test_frexp_fp128_i32(fp128 %a) nounwind {
+; MIPSEL-LABEL: test_frexp_fp128_i32:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -40
+; MIPSEL-NEXT:    sw $ra, 36($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $16, 32($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    move $1, $7
+; MIPSEL-NEXT:    move $16, $4
+; MIPSEL-NEXT:    addiu $2, $sp, 28
+; MIPSEL-NEXT:    sw $2, 16($sp)
+; MIPSEL-NEXT:    lw $7, 56($sp)
+; MIPSEL-NEXT:    move $4, $5
+; MIPSEL-NEXT:    move $5, $6
+; MIPSEL-NEXT:    jal frexpl
+; MIPSEL-NEXT:    move $6, $1
+; MIPSEL-NEXT:    sw $5, 12($16)
+; MIPSEL-NEXT:    sw $4, 8($16)
+; MIPSEL-NEXT:    sw $3, 4($16)
+; MIPSEL-NEXT:    sw $2, 0($16)
+; MIPSEL-NEXT:    lw $1, 28($sp)
+; MIPSEL-NEXT:    sw $1, 16($16)
+; MIPSEL-NEXT:    lw $16, 32($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $ra, 36($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 40
+;
+; SOFT-FLOAT-32-LABEL: test_frexp_fp128_i32:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -40
+; SOFT-FLOAT-32-NEXT:    sw $ra, 36($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $16, 32($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    move $1, $7
+; SOFT-FLOAT-32-NEXT:    move $16, $4
+; SOFT-FLOAT-32-NEXT:    addiu $2, $sp, 28
+; SOFT-FLOAT-32-NEXT:    sw $2, 16($sp)
+; SOFT-FLOAT-32-NEXT:    lw $7, 56($sp)
+; SOFT-FLOAT-32-NEXT:    move $4, $5
+; SOFT-FLOAT-32-NEXT:    move $5, $6
+; SOFT-FLOAT-32-NEXT:    jal frexpl
+; SOFT-FLOAT-32-NEXT:    move $6, $1
+; SOFT-FLOAT-32-NEXT:    sw $5, 12($16)
+; SOFT-FLOAT-32-NEXT:    sw $4, 8($16)
+; SOFT-FLOAT-32-NEXT:    sw $3, 4($16)
+; SOFT-FLOAT-32-NEXT:    sw $2, 0($16)
+; SOFT-FLOAT-32-NEXT:    lw $1, 28($sp)
+; SOFT-FLOAT-32-NEXT:    sw $1, 16($16)
+; SOFT-FLOAT-32-NEXT:    lw $16, 32($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $ra, 36($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 40
+;
+; SOFT-FLOAT-64-LABEL: test_frexp_fp128_i32:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -16
+; SOFT-FLOAT-64-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    dmfc1 $4, $f12
+; SOFT-FLOAT-64-NEXT:    dmfc1 $5, $f13
+; SOFT-FLOAT-64-NEXT:    jal frexpl
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $sp, 4
+; SOFT-FLOAT-64-NEXT:    lw $4, 4($sp)
+; SOFT-FLOAT-64-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 16
+  %result = call { fp128, i32 } @llvm.frexp.fp128.i32(fp128 %a)
+  ret { fp128, i32 } %result
+}
+
+define { <2 x fp128>, <2 x i32> } @test_frexp_v2fp128_v2i32(<2 x fp128> %a) nounwind {
+; MIPSEL-LABEL: test_frexp_v2fp128_v2i32:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -48
+; MIPSEL-NEXT:    sw $ra, 44($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $18, 40($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $17, 36($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $16, 32($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    move $16, $7
+; MIPSEL-NEXT:    move $17, $6
+; MIPSEL-NEXT:    move $18, $4
+; MIPSEL-NEXT:    addiu $1, $sp, 28
+; MIPSEL-NEXT:    sw $1, 16($sp)
+; MIPSEL-NEXT:    lw $4, 72($sp)
+; MIPSEL-NEXT:    lw $5, 76($sp)
+; MIPSEL-NEXT:    lw $6, 80($sp)
+; MIPSEL-NEXT:    lw $7, 84($sp)
+; MIPSEL-NEXT:    jal frexpl
+; MIPSEL-NEXT:    nop
+; MIPSEL-NEXT:    addiu $1, $sp, 24
+; MIPSEL-NEXT:    sw $1, 16($sp)
+; MIPSEL-NEXT:    lw $7, 68($sp)
+; MIPSEL-NEXT:    lw $6, 64($sp)
+; MIPSEL-NEXT:    sw $5, 28($18)
+; MIPSEL-NEXT:    sw $4, 24($18)
+; MIPSEL-NEXT:    sw $3, 20($18)
+; MIPSEL-NEXT:    sw $2, 16($18)
+; MIPSEL-NEXT:    move $4, $17
+; MIPSEL-NEXT:    jal frexpl
+; MIPSEL-NEXT:    move $5, $16
+; MIPSEL-NEXT:    sw $5, 12($18)
+; MIPSEL-NEXT:    sw $4, 8($18)
+; MIPSEL-NEXT:    sw $3, 4($18)
+; MIPSEL-NEXT:    sw $2, 0($18)
+; MIPSEL-NEXT:    lw $1, 28($sp)
+; MIPSEL-NEXT:    sw $1, 36($18)
+; MIPSEL-NEXT:    lw $1, 24($sp)
+; MIPSEL-NEXT:    sw $1, 32($18)
+; MIPSEL-NEXT:    lw $16, 32($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $17, 36($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $18, 40($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $ra, 44($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 48
+;
+; SOFT-FLOAT-32-LABEL: test_frexp_v2fp128_v2i32:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -48
+; SOFT-FLOAT-32-NEXT:    sw $ra, 44($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $18, 40($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $17, 36($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $16, 32($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    move $16, $7
+; SOFT-FLOAT-32-NEXT:    move $17, $6
+; SOFT-FLOAT-32-NEXT:    move $18, $4
+; SOFT-FLOAT-32-NEXT:    addiu $1, $sp, 28
+; SOFT-FLOAT-32-NEXT:    sw $1, 16($sp)
+; SOFT-FLOAT-32-NEXT:    lw $4, 72($sp)
+; SOFT-FLOAT-32-NEXT:    lw $5, 76($sp)
+; SOFT-FLOAT-32-NEXT:    lw $6, 80($sp)
+; SOFT-FLOAT-32-NEXT:    lw $7, 84($sp)
+; SOFT-FLOAT-32-NEXT:    jal frexpl
+; SOFT-FLOAT-32-NEXT:    nop
+; SOFT-FLOAT-32-NEXT:    addiu $1, $sp, 24
+; SOFT-FLOAT-32-NEXT:    sw $1, 16($sp)
+; SOFT-FLOAT-32-NEXT:    lw $7, 68($sp)
+; SOFT-FLOAT-32-NEXT:    lw $6, 64($sp)
+; SOFT-FLOAT-32-NEXT:    sw $5, 28($18)
+; SOFT-FLOAT-32-NEXT:    sw $4, 24($18)
+; SOFT-FLOAT-32-NEXT:    sw $3, 20($18)
+; SOFT-FLOAT-32-NEXT:    sw $2, 16($18)
+; SOFT-FLOAT-32-NEXT:    move $4, $17
+; SOFT-FLOAT-32-NEXT:    jal frexpl
+; SOFT-FLOAT-32-NEXT:    move $5, $16
+; SOFT-FLOAT-32-NEXT:    sw $5, 12($18)
+; SOFT-FLOAT-32-NEXT:    sw $4, 8($18)
+; SOFT-FLOAT-32-NEXT:    sw $3, 4($18)
+; SOFT-FLOAT-32-NEXT:    sw $2, 0($18)
+; SOFT-FLOAT-32-NEXT:    lw $1, 28($sp)
+; SOFT-FLOAT-32-NEXT:    sw $1, 36($18)
+; SOFT-FLOAT-32-NEXT:    lw $1, 24($sp)
+; SOFT-FLOAT-32-NEXT:    sw $1, 32($18)
+; SOFT-FLOAT-32-NEXT:    lw $16, 32($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $17, 36($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $18, 40($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $ra, 44($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 48
+;
+; SOFT-FLOAT-64-LABEL: test_frexp_v2fp128_v2i32:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -64
+; SOFT-FLOAT-64-NEXT:    sd $ra, 56($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $20, 48($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $19, 40($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $18, 32($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $17, 24($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $16, 16($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    move $16, $6
+; SOFT-FLOAT-64-NEXT:    move $17, $5
+; SOFT-FLOAT-64-NEXT:    move $18, $4
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $sp, 12
+; SOFT-FLOAT-64-NEXT:    move $4, $7
+; SOFT-FLOAT-64-NEXT:    jal frexpl
+; SOFT-FLOAT-64-NEXT:    move $5, $8
+; SOFT-FLOAT-64-NEXT:    move $19, $2
+; SOFT-FLOAT-64-NEXT:    move $20, $3
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $sp, 8
+; SOFT-FLOAT-64-NEXT:    lw $1, 12($sp)
+; SOFT-FLOAT-64-NEXT:    sw $1, 36($18)
+; SOFT-FLOAT-64-NEXT:    move $4, $17
+; SOFT-FLOAT-64-NEXT:    jal frexpl
+; SOFT-FLOAT-64-NEXT:    move $5, $16
+; SOFT-FLOAT-64-NEXT:    lw $1, 8($sp)
+; SOFT-FLOAT-64-NEXT:    sw $1, 32($18)
+; SOFT-FLOAT-64-NEXT:    sd $20, 24($18)
+; SOFT-FLOAT-64-NEXT:    sd $19, 16($18)
+; SOFT-FLOAT-64-NEXT:    sd $3, 8($18)
+; SOFT-FLOAT-64-NEXT:    sd $2, 0($18)
+; SOFT-FLOAT-64-NEXT:    ld $16, 16($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $17, 24($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $18, 32($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $19, 40($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $20, 48($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $ra, 56($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 64
+  %result = call { <2 x fp128>, <2 x i32> } @llvm.frexp.v2fp128.v2i32(<2 x fp128> %a)
+  ret { <2 x fp128>, <2 x i32> } %result
+}
+
+declare { half, i32 } @llvm.frexp.f16.i32(half) #0
+declare { <2 x half>, <2 x i32> } @llvm.frexp.v2f16.v2i32(<2 x half>) #0
+
+declare { float, i32 } @llvm.frexp.f32.i32(float) #0
+declare { <2 x float>, <2 x i32> } @llvm.frexp.v2f32.v2i32(<2 x float>) #0
+
+declare { double, i32 } @llvm.frexp.f64.i32(double) #0
+declare { <2 x double>, <2 x i32> } @llvm.frexp.v2f64.v2i32(<2 x double>) #0
+
+declare { fp128, i32 } @llvm.frexp.fp128.i32(fp128) #0
+declare { <2 x fp128>, <2 x i32> } @llvm.frexp.v2fp128.v2i32(<2 x fp128>) #0
+
+attributes #0 = { nocallback nofree nosync nounwind speculatable willreturn memory(none) }
diff --git a/llvm/test/CodeGen/Mips/llvm.sincos.ll b/llvm/test/CodeGen/Mips/llvm.sincos.ll
new file mode 100644
index 0000000..046be12
--- /dev/null
+++ b/llvm/test/CodeGen/Mips/llvm.sincos.ll
@@ -0,0 +1,1044 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -mtriple=mipsel < %s | FileCheck -check-prefix=MIPSEL %s
+; RUN: llc -mtriple=mips < %s | FileCheck %s -check-prefixes=SOFT-FLOAT-32
+; RUN: llc -mtriple=mips64 < %s | FileCheck %s -check-prefixes=SOFT-FLOAT-64
+
+define { half, half } @test_sincos_f16(half %a) #0 {
+; MIPSEL-LABEL: test_sincos_f16:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -32
+; MIPSEL-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $16, 24($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    jal __extendhfsf2
+; MIPSEL-NEXT:    nop
+; MIPSEL-NEXT:    addiu $5, $sp, 20
+; MIPSEL-NEXT:    addiu $6, $sp, 16
+; MIPSEL-NEXT:    jal sincosf
+; MIPSEL-NEXT:    mov.s $f12, $f0
+; MIPSEL-NEXT:    jal __truncsfhf2
+; MIPSEL-NEXT:    lwc1 $f12, 20($sp)
+; MIPSEL-NEXT:    move $16, $2
+; MIPSEL-NEXT:    jal __truncsfhf2
+; MIPSEL-NEXT:    lwc1 $f12, 16($sp)
+; MIPSEL-NEXT:    move $3, $2
+; MIPSEL-NEXT:    move $2, $16
+; MIPSEL-NEXT:    lw $16, 24($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 32
+;
+; SOFT-FLOAT-32-LABEL: test_sincos_f16:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -32
+; SOFT-FLOAT-32-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $16, 24($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    jal __extendhfsf2
+; SOFT-FLOAT-32-NEXT:    nop
+; SOFT-FLOAT-32-NEXT:    addiu $5, $sp, 20
+; SOFT-FLOAT-32-NEXT:    addiu $6, $sp, 16
+; SOFT-FLOAT-32-NEXT:    jal sincosf
+; SOFT-FLOAT-32-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-32-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-32-NEXT:    lwc1 $f12, 20($sp)
+; SOFT-FLOAT-32-NEXT:    move $16, $2
+; SOFT-FLOAT-32-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-32-NEXT:    lwc1 $f12, 16($sp)
+; SOFT-FLOAT-32-NEXT:    move $3, $2
+; SOFT-FLOAT-32-NEXT:    move $2, $16
+; SOFT-FLOAT-32-NEXT:    lw $16, 24($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 32
+;
+; SOFT-FLOAT-64-LABEL: test_sincos_f16:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -32
+; SOFT-FLOAT-64-NEXT:    sd $ra, 24($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $16, 16($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    jal __extendhfsf2
+; SOFT-FLOAT-64-NEXT:    sll $4, $4, 0
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $sp, 12
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $sp, 8
+; SOFT-FLOAT-64-NEXT:    jal sincosf
+; SOFT-FLOAT-64-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-64-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-64-NEXT:    lwc1 $f12, 12($sp)
+; SOFT-FLOAT-64-NEXT:    move $16, $2
+; SOFT-FLOAT-64-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-64-NEXT:    lwc1 $f12, 8($sp)
+; SOFT-FLOAT-64-NEXT:    move $3, $2
+; SOFT-FLOAT-64-NEXT:    move $2, $16
+; SOFT-FLOAT-64-NEXT:    ld $16, 16($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $ra, 24($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 32
+  %result = call { half, half } @llvm.sincos.f16(half %a)
+  ret { half, half } %result
+}
+
+define half @test_sincos_f16_only_use_sin(half %a) #0 {
+; MIPSEL-LABEL: test_sincos_f16_only_use_sin:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -32
+; MIPSEL-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    jal __extendhfsf2
+; MIPSEL-NEXT:    nop
+; MIPSEL-NEXT:    addiu $5, $sp, 24
+; MIPSEL-NEXT:    addiu $6, $sp, 20
+; MIPSEL-NEXT:    jal sincosf
+; MIPSEL-NEXT:    mov.s $f12, $f0
+; MIPSEL-NEXT:    jal __truncsfhf2
+; MIPSEL-NEXT:    lwc1 $f12, 24($sp)
+; MIPSEL-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 32
+;
+; SOFT-FLOAT-32-LABEL: test_sincos_f16_only_use_sin:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -32
+; SOFT-FLOAT-32-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    jal __extendhfsf2
+; SOFT-FLOAT-32-NEXT:    nop
+; SOFT-FLOAT-32-NEXT:    addiu $5, $sp, 24
+; SOFT-FLOAT-32-NEXT:    addiu $6, $sp, 20
+; SOFT-FLOAT-32-NEXT:    jal sincosf
+; SOFT-FLOAT-32-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-32-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-32-NEXT:    lwc1 $f12, 24($sp)
+; SOFT-FLOAT-32-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 32
+;
+; SOFT-FLOAT-64-LABEL: test_sincos_f16_only_use_sin:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -16
+; SOFT-FLOAT-64-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    jal __extendhfsf2
+; SOFT-FLOAT-64-NEXT:    sll $4, $4, 0
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $sp, 4
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $sp, 0
+; SOFT-FLOAT-64-NEXT:    jal sincosf
+; SOFT-FLOAT-64-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-64-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-64-NEXT:    lwc1 $f12, 4($sp)
+; SOFT-FLOAT-64-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 16
+  %result = call { half, half } @llvm.sincos.f16(half %a)
+  %result.0 = extractvalue { half, half } %result, 0
+  ret half %result.0
+}
+
+define half @test_sincos_f16_only_use_cos(half %a) #0 {
+; MIPSEL-LABEL: test_sincos_f16_only_use_cos:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -32
+; MIPSEL-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    jal __extendhfsf2
+; MIPSEL-NEXT:    nop
+; MIPSEL-NEXT:    addiu $5, $sp, 24
+; MIPSEL-NEXT:    addiu $6, $sp, 20
+; MIPSEL-NEXT:    jal sincosf
+; MIPSEL-NEXT:    mov.s $f12, $f0
+; MIPSEL-NEXT:    jal __truncsfhf2
+; MIPSEL-NEXT:    lwc1 $f12, 20($sp)
+; MIPSEL-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 32
+;
+; SOFT-FLOAT-32-LABEL: test_sincos_f16_only_use_cos:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -32
+; SOFT-FLOAT-32-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    jal __extendhfsf2
+; SOFT-FLOAT-32-NEXT:    nop
+; SOFT-FLOAT-32-NEXT:    addiu $5, $sp, 24
+; SOFT-FLOAT-32-NEXT:    addiu $6, $sp, 20
+; SOFT-FLOAT-32-NEXT:    jal sincosf
+; SOFT-FLOAT-32-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-32-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-32-NEXT:    lwc1 $f12, 20($sp)
+; SOFT-FLOAT-32-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 32
+;
+; SOFT-FLOAT-64-LABEL: test_sincos_f16_only_use_cos:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -16
+; SOFT-FLOAT-64-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    jal __extendhfsf2
+; SOFT-FLOAT-64-NEXT:    sll $4, $4, 0
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $sp, 4
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $sp, 0
+; SOFT-FLOAT-64-NEXT:    jal sincosf
+; SOFT-FLOAT-64-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-64-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-64-NEXT:    lwc1 $f12, 0($sp)
+; SOFT-FLOAT-64-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 16
+  %result = call { half, half } @llvm.sincos.f16(half %a)
+  %result.1 = extractvalue { half, half } %result, 1
+  ret half %result.1
+}
+
+define { <2 x half>, <2 x half> } @test_sincos_v2f16(<2 x half> %a) #0 {
+; MIPSEL-LABEL: test_sincos_v2f16:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -48
+; MIPSEL-NEXT:    sw $ra, 44($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $17, 40($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $16, 36($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    move $16, $5
+; MIPSEL-NEXT:    move $17, $4
+; MIPSEL-NEXT:    jal __extendhfsf2
+; MIPSEL-NEXT:    move $4, $5
+; MIPSEL-NEXT:    addiu $5, $sp, 24
+; MIPSEL-NEXT:    addiu $6, $sp, 20
+; MIPSEL-NEXT:    jal sincosf
+; MIPSEL-NEXT:    mov.s $f12, $f0
+; MIPSEL-NEXT:    jal __extendhfsf2
+; MIPSEL-NEXT:    srl $4, $16, 16
+; MIPSEL-NEXT:    addiu $5, $sp, 32
+; MIPSEL-NEXT:    addiu $6, $sp, 28
+; MIPSEL-NEXT:    jal sincosf
+; MIPSEL-NEXT:    mov.s $f12, $f0
+; MIPSEL-NEXT:    jal __truncsfhf2
+; MIPSEL-NEXT:    lwc1 $f12, 20($sp)
+; MIPSEL-NEXT:    lwc1 $f12, 24($sp)
+; MIPSEL-NEXT:    jal __truncsfhf2
+; MIPSEL-NEXT:    sh $2, 4($17)
+; MIPSEL-NEXT:    sh $2, 0($17)
+; MIPSEL-NEXT:    jal __truncsfhf2
+; MIPSEL-NEXT:    lwc1 $f12, 28($sp)
+; MIPSEL-NEXT:    sh $2, 6($17)
+; MIPSEL-NEXT:    jal __truncsfhf2
+; MIPSEL-NEXT:    lwc1 $f12, 32($sp)
+; MIPSEL-NEXT:    sh $2, 2($17)
+; MIPSEL-NEXT:    lw $16, 36($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $17, 40($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $ra, 44($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 48
+;
+; SOFT-FLOAT-32-LABEL: test_sincos_v2f16:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -48
+; SOFT-FLOAT-32-NEXT:    sw $ra, 44($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $17, 40($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $16, 36($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    move $16, $5
+; SOFT-FLOAT-32-NEXT:    move $17, $4
+; SOFT-FLOAT-32-NEXT:    jal __extendhfsf2
+; SOFT-FLOAT-32-NEXT:    move $4, $5
+; SOFT-FLOAT-32-NEXT:    addiu $5, $sp, 24
+; SOFT-FLOAT-32-NEXT:    addiu $6, $sp, 20
+; SOFT-FLOAT-32-NEXT:    jal sincosf
+; SOFT-FLOAT-32-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-32-NEXT:    jal __extendhfsf2
+; SOFT-FLOAT-32-NEXT:    srl $4, $16, 16
+; SOFT-FLOAT-32-NEXT:    addiu $5, $sp, 32
+; SOFT-FLOAT-32-NEXT:    addiu $6, $sp, 28
+; SOFT-FLOAT-32-NEXT:    jal sincosf
+; SOFT-FLOAT-32-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-32-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-32-NEXT:    lwc1 $f12, 20($sp)
+; SOFT-FLOAT-32-NEXT:    lwc1 $f12, 24($sp)
+; SOFT-FLOAT-32-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-32-NEXT:    sh $2, 6($17)
+; SOFT-FLOAT-32-NEXT:    sh $2, 2($17)
+; SOFT-FLOAT-32-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-32-NEXT:    lwc1 $f12, 28($sp)
+; SOFT-FLOAT-32-NEXT:    sh $2, 4($17)
+; SOFT-FLOAT-32-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-32-NEXT:    lwc1 $f12, 32($sp)
+; SOFT-FLOAT-32-NEXT:    sh $2, 0($17)
+; SOFT-FLOAT-32-NEXT:    lw $16, 36($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $17, 40($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $ra, 44($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 48
+;
+; SOFT-FLOAT-64-LABEL: test_sincos_v2f16:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -48
+; SOFT-FLOAT-64-NEXT:    sd $ra, 40($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $17, 32($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $16, 24($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    move $17, $5
+; SOFT-FLOAT-64-NEXT:    move $16, $4
+; SOFT-FLOAT-64-NEXT:    jal __extendhfsf2
+; SOFT-FLOAT-64-NEXT:    sll $4, $5, 0
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $sp, 12
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $sp, 8
+; SOFT-FLOAT-64-NEXT:    jal sincosf
+; SOFT-FLOAT-64-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-64-NEXT:    sll $1, $17, 0
+; SOFT-FLOAT-64-NEXT:    jal __extendhfsf2
+; SOFT-FLOAT-64-NEXT:    srl $4, $1, 16
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $sp, 20
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $sp, 16
+; SOFT-FLOAT-64-NEXT:    jal sincosf
+; SOFT-FLOAT-64-NEXT:    mov.s $f12, $f0
+; SOFT-FLOAT-64-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-64-NEXT:    lwc1 $f12, 8($sp)
+; SOFT-FLOAT-64-NEXT:    lwc1 $f12, 12($sp)
+; SOFT-FLOAT-64-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-64-NEXT:    sh $2, 6($16)
+; SOFT-FLOAT-64-NEXT:    sh $2, 2($16)
+; SOFT-FLOAT-64-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-64-NEXT:    lwc1 $f12, 16($sp)
+; SOFT-FLOAT-64-NEXT:    sh $2, 4($16)
+; SOFT-FLOAT-64-NEXT:    jal __truncsfhf2
+; SOFT-FLOAT-64-NEXT:    lwc1 $f12, 20($sp)
+; SOFT-FLOAT-64-NEXT:    sh $2, 0($16)
+; SOFT-FLOAT-64-NEXT:    ld $16, 24($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $17, 32($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $ra, 40($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 48
+  %result = call { <2 x half>, <2 x half> } @llvm.sincos.v2f16(<2 x half> %a)
+  ret { <2 x half>, <2 x half> } %result
+}
+
+define { float, float } @test_sincos_f32(float %a) #0 {
+; MIPSEL-LABEL: test_sincos_f32:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -32
+; MIPSEL-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    addiu $5, $sp, 24
+; MIPSEL-NEXT:    jal sincosf
+; MIPSEL-NEXT:    addiu $6, $sp, 20
+; MIPSEL-NEXT:    lwc1 $f0, 24($sp)
+; MIPSEL-NEXT:    lwc1 $f2, 20($sp)
+; MIPSEL-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 32
+;
+; SOFT-FLOAT-32-LABEL: test_sincos_f32:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -32
+; SOFT-FLOAT-32-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    addiu $5, $sp, 24
+; SOFT-FLOAT-32-NEXT:    jal sincosf
+; SOFT-FLOAT-32-NEXT:    addiu $6, $sp, 20
+; SOFT-FLOAT-32-NEXT:    lwc1 $f0, 24($sp)
+; SOFT-FLOAT-32-NEXT:    lwc1 $f2, 20($sp)
+; SOFT-FLOAT-32-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 32
+;
+; SOFT-FLOAT-64-LABEL: test_sincos_f32:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -16
+; SOFT-FLOAT-64-NEXT:    sd $ra, 8($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $sp, 4
+; SOFT-FLOAT-64-NEXT:    jal sincosf
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $sp, 0
+; SOFT-FLOAT-64-NEXT:    lwc1 $f0, 4($sp)
+; SOFT-FLOAT-64-NEXT:    lwc1 $f2, 0($sp)
+; SOFT-FLOAT-64-NEXT:    ld $ra, 8($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 16
+  %result = call { float, float } @llvm.sincos.f32(float %a)
+  ret { float, float } %result
+}
+
+define { <2 x float>, <2 x float> } @test_sincos_v2f32(<2 x float> %a) #0 {
+; MIPSEL-LABEL: test_sincos_v2f32:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -32
+; MIPSEL-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $17, 24($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $16, 20($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    move $16, $6
+; MIPSEL-NEXT:    move $17, $4
+; MIPSEL-NEXT:    mtc1 $7, $f12
+; MIPSEL-NEXT:    addiu $5, $4, 4
+; MIPSEL-NEXT:    jal sincosf
+; MIPSEL-NEXT:    addiu $6, $4, 12
+; MIPSEL-NEXT:    mtc1 $16, $f12
+; MIPSEL-NEXT:    addiu $6, $17, 8
+; MIPSEL-NEXT:    jal sincosf
+; MIPSEL-NEXT:    move $5, $17
+; MIPSEL-NEXT:    lw $16, 20($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $17, 24($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 32
+;
+; SOFT-FLOAT-32-LABEL: test_sincos_v2f32:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -32
+; SOFT-FLOAT-32-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $17, 24($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $16, 20($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    move $16, $6
+; SOFT-FLOAT-32-NEXT:    move $17, $4
+; SOFT-FLOAT-32-NEXT:    mtc1 $7, $f12
+; SOFT-FLOAT-32-NEXT:    addiu $5, $4, 4
+; SOFT-FLOAT-32-NEXT:    jal sincosf
+; SOFT-FLOAT-32-NEXT:    addiu $6, $4, 12
+; SOFT-FLOAT-32-NEXT:    mtc1 $16, $f12
+; SOFT-FLOAT-32-NEXT:    addiu $6, $17, 8
+; SOFT-FLOAT-32-NEXT:    jal sincosf
+; SOFT-FLOAT-32-NEXT:    move $5, $17
+; SOFT-FLOAT-32-NEXT:    lw $16, 20($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $17, 24($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 32
+;
+; SOFT-FLOAT-64-LABEL: test_sincos_v2f32:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -32
+; SOFT-FLOAT-64-NEXT:    sd $ra, 24($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $16, 16($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    move $16, $4
+; SOFT-FLOAT-64-NEXT:    dsrl $1, $4, 32
+; SOFT-FLOAT-64-NEXT:    sll $1, $1, 0
+; SOFT-FLOAT-64-NEXT:    mtc1 $1, $f12
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $sp, 12
+; SOFT-FLOAT-64-NEXT:    jal sincosf
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $sp, 8
+; SOFT-FLOAT-64-NEXT:    sll $1, $16, 0
+; SOFT-FLOAT-64-NEXT:    mtc1 $1, $f12
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $sp, 4
+; SOFT-FLOAT-64-NEXT:    jal sincosf
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $sp, 0
+; SOFT-FLOAT-64-NEXT:    lwc1 $f0, 12($sp)
+; SOFT-FLOAT-64-NEXT:    mfc1 $1, $f0
+; SOFT-FLOAT-64-NEXT:    dsll $1, $1, 32
+; SOFT-FLOAT-64-NEXT:    lwc1 $f0, 8($sp)
+; SOFT-FLOAT-64-NEXT:    mfc1 $3, $f0
+; SOFT-FLOAT-64-NEXT:    lwc1 $f0, 4($sp)
+; SOFT-FLOAT-64-NEXT:    mfc1 $2, $f0
+; SOFT-FLOAT-64-NEXT:    dsll $2, $2, 32
+; SOFT-FLOAT-64-NEXT:    dsrl $2, $2, 32
+; SOFT-FLOAT-64-NEXT:    or $2, $2, $1
+; SOFT-FLOAT-64-NEXT:    dsll $1, $3, 32
+; SOFT-FLOAT-64-NEXT:    lwc1 $f0, 0($sp)
+; SOFT-FLOAT-64-NEXT:    mfc1 $3, $f0
+; SOFT-FLOAT-64-NEXT:    dsll $3, $3, 32
+; SOFT-FLOAT-64-NEXT:    dsrl $3, $3, 32
+; SOFT-FLOAT-64-NEXT:    or $3, $3, $1
+; SOFT-FLOAT-64-NEXT:    ld $16, 16($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $ra, 24($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 32
+; SOFT-FLOAT-64R2-LABEL: test_sincos_v2f32:
+; SOFT-FLOAT-64R2:       # %bb.0:
+; SOFT-FLOAT-64R2-NEXT:    daddiu $sp, $sp, -32
+; SOFT-FLOAT-64R2-NEXT:    sd $ra, 24($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64R2-NEXT:    sd $16, 16($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64R2-NEXT:    move $16, $4
+; SOFT-FLOAT-64R2-NEXT:    dsrl $1, $4, 32
+; SOFT-FLOAT-64R2-NEXT:    sll $1, $1, 0
+; SOFT-FLOAT-64R2-NEXT:    mtc1 $1, $f12
+; SOFT-FLOAT-64R2-NEXT:    daddiu $5, $sp, 12
+; SOFT-FLOAT-64R2-NEXT:    jal sincosf
+; SOFT-FLOAT-64R2-NEXT:    daddiu $6, $sp, 8
+; SOFT-FLOAT-64R2-NEXT:    sll $1, $16, 0
+; SOFT-FLOAT-64R2-NEXT:    mtc1 $1, $f12
+; SOFT-FLOAT-64R2-NEXT:    daddiu $5, $sp, 4
+; SOFT-FLOAT-64R2-NEXT:    jal sincosf
+; SOFT-FLOAT-64R2-NEXT:    daddiu $6, $sp, 0
+; SOFT-FLOAT-64R2-NEXT:    lwc1 $f0, 12($sp)
+; SOFT-FLOAT-64R2-NEXT:    mfc1 $1, $f0
+; SOFT-FLOAT-64R2-NEXT:    dsll $1, $1, 32
+; SOFT-FLOAT-64R2-NEXT:    lwc1 $f0, 4($sp)
+; SOFT-FLOAT-64R2-NEXT:    mfc1 $2, $f0
+; SOFT-FLOAT-64R2-NEXT:    dext $2, $2, 0, 32
+; SOFT-FLOAT-64R2-NEXT:    lwc1 $f0, 8($sp)
+; SOFT-FLOAT-64R2-NEXT:    or $2, $2, $1
+; SOFT-FLOAT-64R2-NEXT:    mfc1 $1, $f0
+; SOFT-FLOAT-64R2-NEXT:    dsll $1, $1, 32
+; SOFT-FLOAT-64R2-NEXT:    lwc1 $f0, 0($sp)
+; SOFT-FLOAT-64R2-NEXT:    mfc1 $3, $f0
+; SOFT-FLOAT-64R2-NEXT:    dext $3, $3, 0, 32
+; SOFT-FLOAT-64R2-NEXT:    or $3, $3, $1
+; SOFT-FLOAT-64R2-NEXT:    ld $16, 16($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64R2-NEXT:    ld $ra, 24($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64R2-NEXT:    jr $ra
+; SOFT-FLOAT-64R2-NEXT:    daddiu $sp, $sp, 32
+  %result = call { <2 x float>, <2 x float> } @llvm.sincos.v2f32(<2 x float> %a)
+  ret { <2 x float>, <2 x float> } %result
+}
+
+define { <3 x float>, <3 x float> } @test_sincos_v3f32(<3 x float> %a) #0 {
+; MIPSEL-LABEL: test_sincos_v3f32:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -32
+; MIPSEL-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $18, 24($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $17, 20($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $16, 16($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    move $16, $6
+; MIPSEL-NEXT:    move $17, $5
+; MIPSEL-NEXT:    move $18, $4
+; MIPSEL-NEXT:    mtc1 $7, $f12
+; MIPSEL-NEXT:    addiu $5, $4, 8
+; MIPSEL-NEXT:    jal sincosf
+; MIPSEL-NEXT:    addiu $6, $4, 24
+; MIPSEL-NEXT:    mtc1 $16, $f12
+; MIPSEL-NEXT:    addiu $5, $18, 4
+; MIPSEL-NEXT:    jal sincosf
+; MIPSEL-NEXT:    addiu $6, $18, 20
+; MIPSEL-NEXT:    mtc1 $17, $f12
+; MIPSEL-NEXT:    addiu $6, $18, 16
+; MIPSEL-NEXT:    jal sincosf
+; MIPSEL-NEXT:    move $5, $18
+; MIPSEL-NEXT:    lw $16, 16($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $17, 20($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $18, 24($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 32
+;
+; SOFT-FLOAT-32-LABEL: test_sincos_v3f32:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -32
+; SOFT-FLOAT-32-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $18, 24($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $17, 20($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $16, 16($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    move $16, $6
+; SOFT-FLOAT-32-NEXT:    move $17, $5
+; SOFT-FLOAT-32-NEXT:    move $18, $4
+; SOFT-FLOAT-32-NEXT:    mtc1 $7, $f12
+; SOFT-FLOAT-32-NEXT:    addiu $5, $4, 8
+; SOFT-FLOAT-32-NEXT:    jal sincosf
+; SOFT-FLOAT-32-NEXT:    addiu $6, $4, 24
+; SOFT-FLOAT-32-NEXT:    mtc1 $16, $f12
+; SOFT-FLOAT-32-NEXT:    addiu $5, $18, 4
+; SOFT-FLOAT-32-NEXT:    jal sincosf
+; SOFT-FLOAT-32-NEXT:    addiu $6, $18, 20
+; SOFT-FLOAT-32-NEXT:    mtc1 $17, $f12
+; SOFT-FLOAT-32-NEXT:    addiu $6, $18, 16
+; SOFT-FLOAT-32-NEXT:    jal sincosf
+; SOFT-FLOAT-32-NEXT:    move $5, $18
+; SOFT-FLOAT-32-NEXT:    lw $16, 16($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $17, 20($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $18, 24($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 32
+;
+; SOFT-FLOAT-64-LABEL: test_sincos_v3f32:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -48
+; SOFT-FLOAT-64-NEXT:    sdc1 $f25, 40($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sdc1 $f24, 32($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $ra, 24($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $16, 16($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    mov.s $f24, $f15
+; SOFT-FLOAT-64-NEXT:    mov.s $f25, $f14
+; SOFT-FLOAT-64-NEXT:    mov.s $f12, $f13
+; SOFT-FLOAT-64-NEXT:    move $16, $4
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $sp, 4
+; SOFT-FLOAT-64-NEXT:    jal sincosf
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $sp, 0
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $sp, 12
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $sp, 8
+; SOFT-FLOAT-64-NEXT:    jal sincosf
+; SOFT-FLOAT-64-NEXT:    mov.s $f12, $f25
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $16, 8
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $16, 24
+; SOFT-FLOAT-64-NEXT:    jal sincosf
+; SOFT-FLOAT-64-NEXT:    mov.s $f12, $f24
+; SOFT-FLOAT-64-NEXT:    lwc1 $f0, 0($sp)
+; SOFT-FLOAT-64-NEXT:    mfc1 $1, $f0
+; SOFT-FLOAT-64-NEXT:    lwc1 $f0, 4($sp)
+; SOFT-FLOAT-64-NEXT:    mfc1 $2, $f0
+; SOFT-FLOAT-64-NEXT:    dsll $1, $1, 32
+; SOFT-FLOAT-64-NEXT:    lwc1 $f0, 8($sp)
+; SOFT-FLOAT-64-NEXT:    mfc1 $3, $f0
+; SOFT-FLOAT-64-NEXT:    dsll $3, $3, 32
+; SOFT-FLOAT-64-NEXT:    dsrl $3, $3, 32
+; SOFT-FLOAT-64-NEXT:    or $1, $3, $1
+; SOFT-FLOAT-64-NEXT:    sd $1, 16($16)
+; SOFT-FLOAT-64-NEXT:    dsll $1, $2, 32
+; SOFT-FLOAT-64-NEXT:    lwc1 $f0, 12($sp)
+; SOFT-FLOAT-64-NEXT:    mfc1 $2, $f0
+; SOFT-FLOAT-64-NEXT:    dsll $2, $2, 32
+; SOFT-FLOAT-64-NEXT:    dsrl $2, $2, 32
+; SOFT-FLOAT-64-NEXT:    or $1, $2, $1
+; SOFT-FLOAT-64-NEXT:    sd $1, 0($16)
+; SOFT-FLOAT-64-NEXT:    ld $16, 16($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $ra, 24($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ldc1 $f24, 32($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ldc1 $f25, 40($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 48
+; SOFT-FLOAT-64R2-LABEL: test_sincos_v3f32:
+; SOFT-FLOAT-64R2:       # %bb.0:
+; SOFT-FLOAT-64R2-NEXT:    daddiu $sp, $sp, -48
+; SOFT-FLOAT-64R2-NEXT:    sdc1 $f25, 40($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64R2-NEXT:    sdc1 $f24, 32($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64R2-NEXT:    sd $ra, 24($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64R2-NEXT:    sd $16, 16($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64R2-NEXT:    mov.s $f24, $f15
+; SOFT-FLOAT-64R2-NEXT:    mov.s $f25, $f14
+; SOFT-FLOAT-64R2-NEXT:    mov.s $f12, $f13
+; SOFT-FLOAT-64R2-NEXT:    move $16, $4
+; SOFT-FLOAT-64R2-NEXT:    daddiu $5, $sp, 4
+; SOFT-FLOAT-64R2-NEXT:    jal sincosf
+; SOFT-FLOAT-64R2-NEXT:    daddiu $6, $sp, 0
+; SOFT-FLOAT-64R2-NEXT:    daddiu $5, $sp, 12
+; SOFT-FLOAT-64R2-NEXT:    daddiu $6, $sp, 8
+; SOFT-FLOAT-64R2-NEXT:    jal sincosf
+; SOFT-FLOAT-64R2-NEXT:    mov.s $f12, $f25
+; SOFT-FLOAT-64R2-NEXT:    daddiu $5, $16, 8
+; SOFT-FLOAT-64R2-NEXT:    daddiu $6, $16, 24
+; SOFT-FLOAT-64R2-NEXT:    jal sincosf
+; SOFT-FLOAT-64R2-NEXT:    mov.s $f12, $f24
+; SOFT-FLOAT-64R2-NEXT:    lwc1 $f0, 0($sp)
+; SOFT-FLOAT-64R2-NEXT:    mfc1 $1, $f0
+; SOFT-FLOAT-64R2-NEXT:    dsll $1, $1, 32
+; SOFT-FLOAT-64R2-NEXT:    lwc1 $f0, 8($sp)
+; SOFT-FLOAT-64R2-NEXT:    mfc1 $2, $f0
+; SOFT-FLOAT-64R2-NEXT:    dext $2, $2, 0, 32
+; SOFT-FLOAT-64R2-NEXT:    or $1, $2, $1
+; SOFT-FLOAT-64R2-NEXT:    lwc1 $f0, 4($sp)
+; SOFT-FLOAT-64R2-NEXT:    mfc1 $2, $f0
+; SOFT-FLOAT-64R2-NEXT:    sd $1, 16($16)
+; SOFT-FLOAT-64R2-NEXT:    dsll $1, $2, 32
+; SOFT-FLOAT-64R2-NEXT:    lwc1 $f0, 12($sp)
+; SOFT-FLOAT-64R2-NEXT:    mfc1 $2, $f0
+; SOFT-FLOAT-64R2-NEXT:    dext $2, $2, 0, 32
+; SOFT-FLOAT-64R2-NEXT:    or $1, $2, $1
+; SOFT-FLOAT-64R2-NEXT:    sd $1, 0($16)
+; SOFT-FLOAT-64R2-NEXT:    ld $16, 16($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64R2-NEXT:    ld $ra, 24($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64R2-NEXT:    ldc1 $f24, 32($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64R2-NEXT:    ldc1 $f25, 40($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64R2-NEXT:    jr $ra
+; SOFT-FLOAT-64R2-NEXT:    daddiu $sp, $sp, 48
+  %result = call { <3 x float>, <3 x float> } @llvm.sincos.v3f32(<3 x float> %a)
+  ret { <3 x float>, <3 x float> } %result
+}
+
+define { double, double } @test_sincos_f64(double %a) #0 {
+; MIPSEL-LABEL: test_sincos_f64:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -40
+; MIPSEL-NEXT:    sw $ra, 36($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    addiu $6, $sp, 24
+; MIPSEL-NEXT:    jal sincos
+; MIPSEL-NEXT:    addiu $7, $sp, 16
+; MIPSEL-NEXT:    ldc1 $f0, 24($sp)
+; MIPSEL-NEXT:    ldc1 $f2, 16($sp)
+; MIPSEL-NEXT:    lw $ra, 36($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 40
+;
+; SOFT-FLOAT-32-LABEL: test_sincos_f64:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -40
+; SOFT-FLOAT-32-NEXT:    sw $ra, 36($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    addiu $6, $sp, 24
+; SOFT-FLOAT-32-NEXT:    jal sincos
+; SOFT-FLOAT-32-NEXT:    addiu $7, $sp, 16
+; SOFT-FLOAT-32-NEXT:    ldc1 $f0, 24($sp)
+; SOFT-FLOAT-32-NEXT:    ldc1 $f2, 16($sp)
+; SOFT-FLOAT-32-NEXT:    lw $ra, 36($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 40
+;
+; SOFT-FLOAT-64-LABEL: test_sincos_f64:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -32
+; SOFT-FLOAT-64-NEXT:    sd $ra, 24($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $sp, 16
+; SOFT-FLOAT-64-NEXT:    jal sincos
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $sp, 8
+; SOFT-FLOAT-64-NEXT:    ldc1 $f0, 16($sp)
+; SOFT-FLOAT-64-NEXT:    ldc1 $f2, 8($sp)
+; SOFT-FLOAT-64-NEXT:    ld $ra, 24($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 32
+  %result = call { double, double } @llvm.sincos.f64(double %a)
+  ret { double, double } %result
+}
+
+define { <2 x double>, <2 x double> } @test_sincos_v2f64(<2 x double> %a) #0 {
+; MIPSEL-LABEL: test_sincos_v2f64:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -40
+; MIPSEL-NEXT:    sw $ra, 36($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $16, 32($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    move $16, $4
+; MIPSEL-NEXT:    lw $1, 56($sp)
+; MIPSEL-NEXT:    lw $2, 60($sp)
+; MIPSEL-NEXT:    sw $2, 28($sp)
+; MIPSEL-NEXT:    sw $1, 24($sp)
+; MIPSEL-NEXT:    sw $7, 20($sp)
+; MIPSEL-NEXT:    sw $6, 16($sp)
+; MIPSEL-NEXT:    addiu $6, $4, 8
+; MIPSEL-NEXT:    addiu $7, $4, 24
+; MIPSEL-NEXT:    jal sincos
+; MIPSEL-NEXT:    ldc1 $f12, 24($sp)
+; MIPSEL-NEXT:    addiu $7, $16, 16
+; MIPSEL-NEXT:    ldc1 $f12, 16($sp)
+; MIPSEL-NEXT:    jal sincos
+; MIPSEL-NEXT:    move $6, $16
+; MIPSEL-NEXT:    lw $16, 32($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $ra, 36($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 40
+;
+; SOFT-FLOAT-32-LABEL: test_sincos_v2f64:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -40
+; SOFT-FLOAT-32-NEXT:    sw $ra, 36($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $16, 32($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    move $16, $4
+; SOFT-FLOAT-32-NEXT:    lw $1, 56($sp)
+; SOFT-FLOAT-32-NEXT:    lw $2, 60($sp)
+; SOFT-FLOAT-32-NEXT:    sw $2, 28($sp)
+; SOFT-FLOAT-32-NEXT:    sw $1, 24($sp)
+; SOFT-FLOAT-32-NEXT:    sw $7, 20($sp)
+; SOFT-FLOAT-32-NEXT:    sw $6, 16($sp)
+; SOFT-FLOAT-32-NEXT:    addiu $6, $4, 8
+; SOFT-FLOAT-32-NEXT:    addiu $7, $4, 24
+; SOFT-FLOAT-32-NEXT:    jal sincos
+; SOFT-FLOAT-32-NEXT:    ldc1 $f12, 24($sp)
+; SOFT-FLOAT-32-NEXT:    addiu $7, $16, 16
+; SOFT-FLOAT-32-NEXT:    ldc1 $f12, 16($sp)
+; SOFT-FLOAT-32-NEXT:    jal sincos
+; SOFT-FLOAT-32-NEXT:    move $6, $16
+; SOFT-FLOAT-32-NEXT:    lw $16, 32($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $ra, 36($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 40
+;
+; SOFT-FLOAT-64-LABEL: test_sincos_v2f64:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -32
+; SOFT-FLOAT-64-NEXT:    sd $ra, 24($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $17, 16($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $16, 8($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    move $16, $5
+; SOFT-FLOAT-64-NEXT:    move $17, $4
+; SOFT-FLOAT-64-NEXT:    dmtc1 $6, $f12
+; SOFT-FLOAT-64-NEXT:    daddiu $5, $4, 8
+; SOFT-FLOAT-64-NEXT:    jal sincos
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $4, 24
+; SOFT-FLOAT-64-NEXT:    dmtc1 $16, $f12
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $17, 16
+; SOFT-FLOAT-64-NEXT:    jal sincos
+; SOFT-FLOAT-64-NEXT:    move $5, $17
+; SOFT-FLOAT-64-NEXT:    ld $16, 8($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $17, 16($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $ra, 24($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 32
+; SOFT-FLOAT-32R2-LABEL: test_sincos_v2f64:
+; SOFT-FLOAT-32R2:       # %bb.0:
+; SOFT-FLOAT-32R2-NEXT:    addiu $sp, $sp, -32
+; SOFT-FLOAT-32R2-NEXT:    sw $ra, 28($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32R2-NEXT:    sw $18, 24($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32R2-NEXT:    sw $17, 20($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32R2-NEXT:    sw $16, 16($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32R2-NEXT:    move $16, $7
+; SOFT-FLOAT-32R2-NEXT:    move $17, $6
+; SOFT-FLOAT-32R2-NEXT:    move $18, $4
+; SOFT-FLOAT-32R2-NEXT:    lw $1, 48($sp)
+; SOFT-FLOAT-32R2-NEXT:    lw $2, 52($sp)
+; SOFT-FLOAT-32R2-NEXT:    mtc1 $2, $f12
+; SOFT-FLOAT-32R2-NEXT:    mthc1 $1, $f12
+; SOFT-FLOAT-32R2-NEXT:    addiu $6, $4, 8
+; SOFT-FLOAT-32R2-NEXT:    jal sincos
+; SOFT-FLOAT-32R2-NEXT:    addiu $7, $4, 24
+; SOFT-FLOAT-32R2-NEXT:    mtc1 $16, $f12
+; SOFT-FLOAT-32R2-NEXT:    mthc1 $17, $f12
+; SOFT-FLOAT-32R2-NEXT:    addiu $7, $18, 16
+; SOFT-FLOAT-32R2-NEXT:    jal sincos
+; SOFT-FLOAT-32R2-NEXT:    move $6, $18
+; SOFT-FLOAT-32R2-NEXT:    lw $16, 16($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32R2-NEXT:    lw $17, 20($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32R2-NEXT:    lw $18, 24($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32R2-NEXT:    lw $ra, 28($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32R2-NEXT:    jr $ra
+; SOFT-FLOAT-32R2-NEXT:    addiu $sp, $sp, 32
+  %result = call { <2 x double>, <2 x double> } @llvm.sincos.v2f64(<2 x double> %a)
+  ret { <2 x double>, <2 x double> } %result
+}
+
+define { fp128, fp128 } @test_sincos_f128(fp128 %a) #0 {
+; MIPSEL-LABEL: test_sincos_f128:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -64
+; MIPSEL-NEXT:    sw $ra, 60($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $16, 56($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    move $1, $7
+; MIPSEL-NEXT:    move $16, $4
+; MIPSEL-NEXT:    addiu $2, $sp, 24
+; MIPSEL-NEXT:    sw $2, 20($sp)
+; MIPSEL-NEXT:    addiu $2, $sp, 40
+; MIPSEL-NEXT:    sw $2, 16($sp)
+; MIPSEL-NEXT:    lw $7, 80($sp)
+; MIPSEL-NEXT:    move $4, $5
+; MIPSEL-NEXT:    move $5, $6
+; MIPSEL-NEXT:    jal sincosl
+; MIPSEL-NEXT:    move $6, $1
+; MIPSEL-NEXT:    lw $1, 52($sp)
+; MIPSEL-NEXT:    lw $2, 24($sp)
+; MIPSEL-NEXT:    lw $3, 28($sp)
+; MIPSEL-NEXT:    lw $4, 32($sp)
+; MIPSEL-NEXT:    lw $5, 36($sp)
+; MIPSEL-NEXT:    sw $5, 28($16)
+; MIPSEL-NEXT:    sw $4, 24($16)
+; MIPSEL-NEXT:    sw $3, 20($16)
+; MIPSEL-NEXT:    sw $2, 16($16)
+; MIPSEL-NEXT:    sw $1, 12($16)
+; MIPSEL-NEXT:    lw $1, 48($sp)
+; MIPSEL-NEXT:    sw $1, 8($16)
+; MIPSEL-NEXT:    lw $1, 44($sp)
+; MIPSEL-NEXT:    sw $1, 4($16)
+; MIPSEL-NEXT:    lw $1, 40($sp)
+; MIPSEL-NEXT:    sw $1, 0($16)
+; MIPSEL-NEXT:    lw $16, 56($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $ra, 60($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 64
+;
+; SOFT-FLOAT-32-LABEL: test_sincos_f128:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -64
+; SOFT-FLOAT-32-NEXT:    sw $ra, 60($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $16, 56($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    move $1, $7
+; SOFT-FLOAT-32-NEXT:    move $16, $4
+; SOFT-FLOAT-32-NEXT:    addiu $2, $sp, 24
+; SOFT-FLOAT-32-NEXT:    sw $2, 20($sp)
+; SOFT-FLOAT-32-NEXT:    addiu $2, $sp, 40
+; SOFT-FLOAT-32-NEXT:    sw $2, 16($sp)
+; SOFT-FLOAT-32-NEXT:    lw $7, 80($sp)
+; SOFT-FLOAT-32-NEXT:    move $4, $5
+; SOFT-FLOAT-32-NEXT:    move $5, $6
+; SOFT-FLOAT-32-NEXT:    jal sincosl
+; SOFT-FLOAT-32-NEXT:    move $6, $1
+; SOFT-FLOAT-32-NEXT:    lw $1, 52($sp)
+; SOFT-FLOAT-32-NEXT:    lw $2, 24($sp)
+; SOFT-FLOAT-32-NEXT:    lw $3, 28($sp)
+; SOFT-FLOAT-32-NEXT:    lw $4, 32($sp)
+; SOFT-FLOAT-32-NEXT:    lw $5, 36($sp)
+; SOFT-FLOAT-32-NEXT:    sw $5, 28($16)
+; SOFT-FLOAT-32-NEXT:    sw $4, 24($16)
+; SOFT-FLOAT-32-NEXT:    sw $3, 20($16)
+; SOFT-FLOAT-32-NEXT:    sw $2, 16($16)
+; SOFT-FLOAT-32-NEXT:    sw $1, 12($16)
+; SOFT-FLOAT-32-NEXT:    lw $1, 48($sp)
+; SOFT-FLOAT-32-NEXT:    sw $1, 8($16)
+; SOFT-FLOAT-32-NEXT:    lw $1, 44($sp)
+; SOFT-FLOAT-32-NEXT:    sw $1, 4($16)
+; SOFT-FLOAT-32-NEXT:    lw $1, 40($sp)
+; SOFT-FLOAT-32-NEXT:    sw $1, 0($16)
+; SOFT-FLOAT-32-NEXT:    lw $16, 56($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $ra, 60($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 64
+;
+; SOFT-FLOAT-64-LABEL: test_sincos_f128:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -48
+; SOFT-FLOAT-64-NEXT:    sd $ra, 40($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $16, 32($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    move $16, $4
+; SOFT-FLOAT-64-NEXT:    dmfc1 $4, $f13
+; SOFT-FLOAT-64-NEXT:    dmfc1 $5, $f14
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $sp, 16
+; SOFT-FLOAT-64-NEXT:    jal sincosl
+; SOFT-FLOAT-64-NEXT:    daddiu $7, $sp, 0
+; SOFT-FLOAT-64-NEXT:    ld $1, 8($sp)
+; SOFT-FLOAT-64-NEXT:    sd $1, 24($16)
+; SOFT-FLOAT-64-NEXT:    ld $1, 0($sp)
+; SOFT-FLOAT-64-NEXT:    sd $1, 16($16)
+; SOFT-FLOAT-64-NEXT:    ld $1, 24($sp)
+; SOFT-FLOAT-64-NEXT:    sd $1, 8($16)
+; SOFT-FLOAT-64-NEXT:    ld $1, 16($sp)
+; SOFT-FLOAT-64-NEXT:    sd $1, 0($16)
+; SOFT-FLOAT-64-NEXT:    ld $16, 32($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $ra, 40($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 48
+  %result = call { fp128, fp128 } @llvm.sincos.f128(fp128 %a)
+  ret { fp128, fp128 } %result
+}
+
+define { <2 x fp128>, <2 x fp128> } @test_sincos_v2f128(<2 x fp128> %a) #0 {
+; MIPSEL-LABEL: test_sincos_v2f128:
+; MIPSEL:       # %bb.0:
+; MIPSEL-NEXT:    addiu $sp, $sp, -96
+; MIPSEL-NEXT:    sw $ra, 92($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    sw $16, 88($sp) # 4-byte Folded Spill
+; MIPSEL-NEXT:    move $5, $7
+; MIPSEL-NEXT:    move $1, $6
+; MIPSEL-NEXT:    move $16, $4
+; MIPSEL-NEXT:    addiu $2, $sp, 24
+; MIPSEL-NEXT:    sw $2, 20($sp)
+; MIPSEL-NEXT:    addiu $2, $sp, 40
+; MIPSEL-NEXT:    sw $2, 16($sp)
+; MIPSEL-NEXT:    lw $6, 112($sp)
+; MIPSEL-NEXT:    lw $7, 116($sp)
+; MIPSEL-NEXT:    jal sincosl
+; MIPSEL-NEXT:    move $4, $1
+; MIPSEL-NEXT:    addiu $1, $sp, 56
+; MIPSEL-NEXT:    sw $1, 20($sp)
+; MIPSEL-NEXT:    addiu $1, $sp, 72
+; MIPSEL-NEXT:    sw $1, 16($sp)
+; MIPSEL-NEXT:    lw $4, 120($sp)
+; MIPSEL-NEXT:    lw $5, 124($sp)
+; MIPSEL-NEXT:    lw $6, 128($sp)
+; MIPSEL-NEXT:    lw $7, 132($sp)
+; MIPSEL-NEXT:    jal sincosl
+; MIPSEL-NEXT:    nop
+; MIPSEL-NEXT:    lw $1, 36($sp)
+; MIPSEL-NEXT:    lw $2, 56($sp)
+; MIPSEL-NEXT:    lw $3, 60($sp)
+; MIPSEL-NEXT:    lw $4, 64($sp)
+; MIPSEL-NEXT:    lw $5, 52($sp)
+; MIPSEL-NEXT:    lw $6, 72($sp)
+; MIPSEL-NEXT:    lw $7, 76($sp)
+; MIPSEL-NEXT:    lw $8, 80($sp)
+; MIPSEL-NEXT:    lw $9, 84($sp)
+; MIPSEL-NEXT:    lw $10, 24($sp)
+; MIPSEL-NEXT:    lw $11, 28($sp)
+; MIPSEL-NEXT:    lw $12, 32($sp)
+; MIPSEL-NEXT:    lw $13, 68($sp)
+; MIPSEL-NEXT:    sw $13, 60($16)
+; MIPSEL-NEXT:    sw $4, 56($16)
+; MIPSEL-NEXT:    sw $3, 52($16)
+; MIPSEL-NEXT:    sw $2, 48($16)
+; MIPSEL-NEXT:    sw $1, 44($16)
+; MIPSEL-NEXT:    sw $12, 40($16)
+; MIPSEL-NEXT:    sw $11, 36($16)
+; MIPSEL-NEXT:    sw $10, 32($16)
+; MIPSEL-NEXT:    sw $9, 28($16)
+; MIPSEL-NEXT:    sw $8, 24($16)
+; MIPSEL-NEXT:    sw $7, 20($16)
+; MIPSEL-NEXT:    sw $6, 16($16)
+; MIPSEL-NEXT:    sw $5, 12($16)
+; MIPSEL-NEXT:    lw $1, 48($sp)
+; MIPSEL-NEXT:    sw $1, 8($16)
+; MIPSEL-NEXT:    lw $1, 44($sp)
+; MIPSEL-NEXT:    sw $1, 4($16)
+; MIPSEL-NEXT:    lw $1, 40($sp)
+; MIPSEL-NEXT:    sw $1, 0($16)
+; MIPSEL-NEXT:    lw $16, 88($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    lw $ra, 92($sp) # 4-byte Folded Reload
+; MIPSEL-NEXT:    jr $ra
+; MIPSEL-NEXT:    addiu $sp, $sp, 96
+;
+; SOFT-FLOAT-32-LABEL: test_sincos_v2f128:
+; SOFT-FLOAT-32:       # %bb.0:
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, -96
+; SOFT-FLOAT-32-NEXT:    sw $ra, 92($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    sw $16, 88($sp) # 4-byte Folded Spill
+; SOFT-FLOAT-32-NEXT:    move $5, $7
+; SOFT-FLOAT-32-NEXT:    move $1, $6
+; SOFT-FLOAT-32-NEXT:    move $16, $4
+; SOFT-FLOAT-32-NEXT:    addiu $2, $sp, 24
+; SOFT-FLOAT-32-NEXT:    sw $2, 20($sp)
+; SOFT-FLOAT-32-NEXT:    addiu $2, $sp, 40
+; SOFT-FLOAT-32-NEXT:    sw $2, 16($sp)
+; SOFT-FLOAT-32-NEXT:    lw $6, 112($sp)
+; SOFT-FLOAT-32-NEXT:    lw $7, 116($sp)
+; SOFT-FLOAT-32-NEXT:    jal sincosl
+; SOFT-FLOAT-32-NEXT:    move $4, $1
+; SOFT-FLOAT-32-NEXT:    addiu $1, $sp, 56
+; SOFT-FLOAT-32-NEXT:    sw $1, 20($sp)
+; SOFT-FLOAT-32-NEXT:    addiu $1, $sp, 72
+; SOFT-FLOAT-32-NEXT:    sw $1, 16($sp)
+; SOFT-FLOAT-32-NEXT:    lw $4, 120($sp)
+; SOFT-FLOAT-32-NEXT:    lw $5, 124($sp)
+; SOFT-FLOAT-32-NEXT:    lw $6, 128($sp)
+; SOFT-FLOAT-32-NEXT:    lw $7, 132($sp)
+; SOFT-FLOAT-32-NEXT:    jal sincosl
+; SOFT-FLOAT-32-NEXT:    nop
+; SOFT-FLOAT-32-NEXT:    lw $1, 36($sp)
+; SOFT-FLOAT-32-NEXT:    lw $2, 56($sp)
+; SOFT-FLOAT-32-NEXT:    lw $3, 60($sp)
+; SOFT-FLOAT-32-NEXT:    lw $4, 64($sp)
+; SOFT-FLOAT-32-NEXT:    lw $5, 52($sp)
+; SOFT-FLOAT-32-NEXT:    lw $6, 72($sp)
+; SOFT-FLOAT-32-NEXT:    lw $7, 76($sp)
+; SOFT-FLOAT-32-NEXT:    lw $8, 80($sp)
+; SOFT-FLOAT-32-NEXT:    lw $9, 84($sp)
+; SOFT-FLOAT-32-NEXT:    lw $10, 24($sp)
+; SOFT-FLOAT-32-NEXT:    lw $11, 28($sp)
+; SOFT-FLOAT-32-NEXT:    lw $12, 32($sp)
+; SOFT-FLOAT-32-NEXT:    lw $13, 68($sp)
+; SOFT-FLOAT-32-NEXT:    sw $13, 60($16)
+; SOFT-FLOAT-32-NEXT:    sw $4, 56($16)
+; SOFT-FLOAT-32-NEXT:    sw $3, 52($16)
+; SOFT-FLOAT-32-NEXT:    sw $2, 48($16)
+; SOFT-FLOAT-32-NEXT:    sw $1, 44($16)
+; SOFT-FLOAT-32-NEXT:    sw $12, 40($16)
+; SOFT-FLOAT-32-NEXT:    sw $11, 36($16)
+; SOFT-FLOAT-32-NEXT:    sw $10, 32($16)
+; SOFT-FLOAT-32-NEXT:    sw $9, 28($16)
+; SOFT-FLOAT-32-NEXT:    sw $8, 24($16)
+; SOFT-FLOAT-32-NEXT:    sw $7, 20($16)
+; SOFT-FLOAT-32-NEXT:    sw $6, 16($16)
+; SOFT-FLOAT-32-NEXT:    sw $5, 12($16)
+; SOFT-FLOAT-32-NEXT:    lw $1, 48($sp)
+; SOFT-FLOAT-32-NEXT:    sw $1, 8($16)
+; SOFT-FLOAT-32-NEXT:    lw $1, 44($sp)
+; SOFT-FLOAT-32-NEXT:    sw $1, 4($16)
+; SOFT-FLOAT-32-NEXT:    lw $1, 40($sp)
+; SOFT-FLOAT-32-NEXT:    sw $1, 0($16)
+; SOFT-FLOAT-32-NEXT:    lw $16, 88($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    lw $ra, 92($sp) # 4-byte Folded Reload
+; SOFT-FLOAT-32-NEXT:    jr $ra
+; SOFT-FLOAT-32-NEXT:    addiu $sp, $sp, 96
+;
+; SOFT-FLOAT-64-LABEL: test_sincos_v2f128:
+; SOFT-FLOAT-64:       # %bb.0:
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, -96
+; SOFT-FLOAT-64-NEXT:    sd $ra, 88($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $18, 80($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $17, 72($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    sd $16, 64($sp) # 8-byte Folded Spill
+; SOFT-FLOAT-64-NEXT:    move $1, $7
+; SOFT-FLOAT-64-NEXT:    move $16, $6
+; SOFT-FLOAT-64-NEXT:    move $17, $5
+; SOFT-FLOAT-64-NEXT:    move $18, $4
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $sp, 48
+; SOFT-FLOAT-64-NEXT:    daddiu $7, $sp, 32
+; SOFT-FLOAT-64-NEXT:    move $4, $1
+; SOFT-FLOAT-64-NEXT:    jal sincosl
+; SOFT-FLOAT-64-NEXT:    move $5, $8
+; SOFT-FLOAT-64-NEXT:    daddiu $6, $sp, 16
+; SOFT-FLOAT-64-NEXT:    daddiu $7, $sp, 0
+; SOFT-FLOAT-64-NEXT:    move $4, $17
+; SOFT-FLOAT-64-NEXT:    jal sincosl
+; SOFT-FLOAT-64-NEXT:    move $5, $16
+; SOFT-FLOAT-64-NEXT:    ld $1, 56($sp)
+; SOFT-FLOAT-64-NEXT:    ld $2, 0($sp)
+; SOFT-FLOAT-64-NEXT:    ld $3, 8($sp)
+; SOFT-FLOAT-64-NEXT:    ld $4, 32($sp)
+; SOFT-FLOAT-64-NEXT:    ld $5, 40($sp)
+; SOFT-FLOAT-64-NEXT:    sd $5, 56($18)
+; SOFT-FLOAT-64-NEXT:    sd $4, 48($18)
+; SOFT-FLOAT-64-NEXT:    sd $3, 40($18)
+; SOFT-FLOAT-64-NEXT:    sd $2, 32($18)
+; SOFT-FLOAT-64-NEXT:    sd $1, 24($18)
+; SOFT-FLOAT-64-NEXT:    ld $1, 48($sp)
+; SOFT-FLOAT-64-NEXT:    sd $1, 16($18)
+; SOFT-FLOAT-64-NEXT:    ld $1, 24($sp)
+; SOFT-FLOAT-64-NEXT:    sd $1, 8($18)
+; SOFT-FLOAT-64-NEXT:    ld $1, 16($sp)
+; SOFT-FLOAT-64-NEXT:    sd $1, 0($18)
+; SOFT-FLOAT-64-NEXT:    ld $16, 64($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $17, 72($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $18, 80($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    ld $ra, 88($sp) # 8-byte Folded Reload
+; SOFT-FLOAT-64-NEXT:    jr $ra
+; SOFT-FLOAT-64-NEXT:    daddiu $sp, $sp, 96
+  %result = call { <2 x fp128>, <2 x fp128> } @llvm.sincos.v2f128(<2 x fp128> %a)
+  ret { <2 x fp128>, <2 x fp128> } %result
+}
+
+attributes #0 = { nounwind }
diff --git a/llvm/test/CodeGen/Mips/nan_lowering.ll b/llvm/test/CodeGen/Mips/nan_lowering.ll
new file mode 100644
index 0000000..2a11278
--- /dev/null
+++ b/llvm/test/CodeGen/Mips/nan_lowering.ll
@@ -0,0 +1,25 @@
+; RUN: llc -mtriple=mips-linux-gnu -mattr=-nan2008 < %s | FileCheck %s
+; RUN: llc -mtriple=mips-linux-gnu -mattr=+nan2008 < %s | FileCheck %s
+
+; Make sure that lowering does not corrupt the value of NaN values,
+; regardless of what the NaN mode is.
+
+define float @test1() {
+; CHECK: .4byte 0x7fc00000
+  ret float bitcast (i32 u0x7fc00000 to float)
+}
+
+define float @test2() {
+; CHECK: .4byte 0x7fc00001
+  ret float bitcast (i32 u0x7fc00001 to float)
+}
+
+define float @test3() {
+; CHECK: .4byte 0x7f800000
+  ret float bitcast (i32 u0x7f800000 to float)
+}
+
+define float @test4() {
+; CHECK: .4byte 0x7f800001
+  ret float bitcast (i32 u0x7f800001 to float)
+}
diff --git a/llvm/test/CodeGen/Mips/qnan.ll b/llvm/test/CodeGen/Mips/qnan.ll
deleted file mode 100644
index e5b4aa1..0000000
--- a/llvm/test/CodeGen/Mips/qnan.ll
+++ /dev/null
@@ -1,14 +0,0 @@
-; RUN: llc -O3 -mcpu=mips32r2 -mtriple=mips-linux-gnu < %s -o - | FileCheck %s -check-prefixes=MIPS_Legacy
-; RUN: llc -O3 -mcpu=mips32r2 -mtriple=mips-linux-gnu -mattr=+nan2008 < %s -o - | FileCheck %s -check-prefixes=MIPS_NaN2008
-
-define dso_local float @nan(float noundef %a, float noundef %b) local_unnamed_addr #0 {
-; MIPS_Legacy: $CPI0_0:
-; MIPS_Legacy-NEXT: .4byte  0x7fa00000 # float NaN
-
-; MIPS_NaN2008: $CPI0_0:
-; MIPS_NaN2008-NEXT: .4byte  0x7fc00000 # float NaN
-
-entry:
-  %0 = tail call float @llvm.minimum.f32(float %a, float %b)
-  ret float %0
-}
diff --git a/llvm/test/CodeGen/PowerPC/froundeven-legalization.ll b/llvm/test/CodeGen/PowerPC/froundeven-legalization.ll
new file mode 100644
index 0000000..238e200
--- /dev/null
+++ b/llvm/test/CodeGen/PowerPC/froundeven-legalization.ll
@@ -0,0 +1,111 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc -mtriple=powerpc64le < %s | FileCheck %s
+
+define void @test(ptr %p1, ptr %p2) nounwind {
+; CHECK-LABEL: test:
+; CHECK:       # %bb.0:
+; CHECK-NEXT:    mflr 0
+; CHECK-NEXT:    stdu 1, -224(1)
+; CHECK-NEXT:    li 5, 48
+; CHECK-NEXT:    std 0, 240(1)
+; CHECK-NEXT:    std 27, 184(1) # 8-byte Folded Spill
+; CHECK-NEXT:    li 27, 16
+; CHECK-NEXT:    std 28, 192(1) # 8-byte Folded Spill
+; CHECK-NEXT:    std 29, 200(1) # 8-byte Folded Spill
+; CHECK-NEXT:    li 29, 32
+; CHECK-NEXT:    li 28, 48
+; CHECK-NEXT:    stxvd2x 56, 1, 5 # 16-byte Folded Spill
+; CHECK-NEXT:    li 5, 64
+; CHECK-NEXT:    std 30, 208(1) # 8-byte Folded Spill
+; CHECK-NEXT:    mr 30, 4
+; CHECK-NEXT:    stxvd2x 57, 1, 5 # 16-byte Folded Spill
+; CHECK-NEXT:    li 5, 80
+; CHECK-NEXT:    stxvd2x 58, 1, 5 # 16-byte Folded Spill
+; CHECK-NEXT:    li 5, 96
+; CHECK-NEXT:    lxvd2x 58, 0, 3
+; CHECK-NEXT:    stxvd2x 59, 1, 5 # 16-byte Folded Spill
+; CHECK-NEXT:    li 5, 112
+; CHECK-NEXT:    lxvd2x 59, 3, 27
+; CHECK-NEXT:    stxvd2x 60, 1, 5 # 16-byte Folded Spill
+; CHECK-NEXT:    li 5, 128
+; CHECK-NEXT:    stxvd2x 61, 1, 5 # 16-byte Folded Spill
+; CHECK-NEXT:    li 5, 144
+; CHECK-NEXT:    stxvd2x 62, 1, 5 # 16-byte Folded Spill
+; CHECK-NEXT:    li 5, 160
+; CHECK-NEXT:    lxvd2x 62, 3, 28
+; CHECK-NEXT:    stxvd2x 63, 1, 5 # 16-byte Folded Spill
+; CHECK-NEXT:    lxvd2x 63, 3, 29
+; CHECK-NEXT:    xxswapd 57, 58
+; CHECK-NEXT:    xxswapd 1, 59
+; CHECK-NEXT:    xxswapd 60, 62
+; CHECK-NEXT:    xxswapd 61, 63
+; CHECK-NEXT:    bl roundeven
+; CHECK-NEXT:    nop
+; CHECK-NEXT:    xxswapd 56, 1
+; CHECK-NEXT:    xxlor 1, 59, 59
+; CHECK-NEXT:    bl roundeven
+; CHECK-NEXT:    nop
+; CHECK-NEXT:    xxswapd 0, 1
+; CHECK-NEXT:    xxlor 1, 60, 60
+; CHECK-NEXT:    xxmrgld 59, 0, 56
+; CHECK-NEXT:    bl roundeven
+; CHECK-NEXT:    nop
+; CHECK-NEXT:    xxswapd 60, 1
+; CHECK-NEXT:    xxlor 1, 62, 62
+; CHECK-NEXT:    bl roundeven
+; CHECK-NEXT:    nop
+; CHECK-NEXT:    xxswapd 0, 1
+; CHECK-NEXT:    xxlor 1, 61, 61
+; CHECK-NEXT:    xxmrgld 62, 0, 60
+; CHECK-NEXT:    bl roundeven
+; CHECK-NEXT:    nop
+; CHECK-NEXT:    xxswapd 61, 1
+; CHECK-NEXT:    xxlor 1, 63, 63
+; CHECK-NEXT:    bl roundeven
+; CHECK-NEXT:    nop
+; CHECK-NEXT:    xxswapd 0, 1
+; CHECK-NEXT:    xxlor 1, 57, 57
+; CHECK-NEXT:    xxmrgld 63, 0, 61
+; CHECK-NEXT:    bl roundeven
+; CHECK-NEXT:    nop
+; CHECK-NEXT:    xxswapd 61, 1
+; CHECK-NEXT:    xxlor 1, 58, 58
+; CHECK-NEXT:    bl roundeven
+; CHECK-NEXT:    nop
+; CHECK-NEXT:    li 3, 160
+; CHECK-NEXT:    stxvd2x 63, 30, 29
+; CHECK-NEXT:    xxswapd 0, 1
+; CHECK-NEXT:    stxvd2x 62, 30, 28
+; CHECK-NEXT:    stxvd2x 59, 30, 27
+; CHECK-NEXT:    ld 29, 200(1) # 8-byte Folded Reload
+; CHECK-NEXT:    ld 28, 192(1) # 8-byte Folded Reload
+; CHECK-NEXT:    ld 27, 184(1) # 8-byte Folded Reload
+; CHECK-NEXT:    lxvd2x 63, 1, 3 # 16-byte Folded Reload
+; CHECK-NEXT:    li 3, 144
+; CHECK-NEXT:    xxmrgld 0, 0, 61
+; CHECK-NEXT:    lxvd2x 62, 1, 3 # 16-byte Folded Reload
+; CHECK-NEXT:    li 3, 128
+; CHECK-NEXT:    stxvd2x 0, 0, 30
+; CHECK-NEXT:    ld 30, 208(1) # 8-byte Folded Reload
+; CHECK-NEXT:    lxvd2x 61, 1, 3 # 16-byte Folded Reload
+; CHECK-NEXT:    li 3, 112
+; CHECK-NEXT:    lxvd2x 60, 1, 3 # 16-byte Folded Reload
+; CHECK-NEXT:    li 3, 96
+; CHECK-NEXT:    lxvd2x 59, 1, 3 # 16-byte Folded Reload
+; CHECK-NEXT:    li 3, 80
+; CHECK-NEXT:    lxvd2x 58, 1, 3 # 16-byte Folded Reload
+; CHECK-NEXT:    li 3, 64
+; CHECK-NEXT:    lxvd2x 57, 1, 3 # 16-byte Folded Reload
+; CHECK-NEXT:    li 3, 48
+; CHECK-NEXT:    lxvd2x 56, 1, 3 # 16-byte Folded Reload
+; CHECK-NEXT:    addi 1, 1, 224
+; CHECK-NEXT:    ld 0, 16(1)
+; CHECK-NEXT:    mtlr 0
+; CHECK-NEXT:    blr
+  %v = load <8 x double>, ptr %p1, align 64
+  %res = call <8 x double> @llvm.roundeven.v8f64(<8 x double> %v)
+  store <8 x double> %res, ptr %p2, align 64
+  ret void
+}
+
+declare <8 x double> @llvm.roundeven.v8f64(<8 x double>)
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-deinterleave-load.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-deinterleave-load.ll
index 807651c..dc80225 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-deinterleave-load.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-deinterleave-load.ll
@@ -261,7 +261,7 @@ define { <8 x i8>, <8 x i8>, <8 x i8> } @vector_deinterleave_load_factor3(ptr %p
 ; CHECK-LABEL: vector_deinterleave_load_factor3:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
-; CHECK-NEXT:    vlseg3e8.v v6, (a0)
+; CHECK-NEXT:    vlseg3e8.v v8, (a0)
 ; CHECK-NEXT:    ret
   %vec = load <24 x i8>, ptr %p
   %d0 = call {<8 x i8>, <8 x i8>, <8 x i8>} @llvm.vector.deinterleave3(<24 x i8> %vec)
@@ -269,8 +269,8 @@ define { <8 x i8>, <8 x i8>, <8 x i8> } @vector_deinterleave_load_factor3(ptr %p
   %t1 = extractvalue {<8 x i8>, <8 x i8>, <8 x i8>} %d0, 1
   %t2 = extractvalue {<8 x i8>, <8 x i8>, <8 x i8>} %d0, 2
   %res0 = insertvalue { <8 x i8>, <8 x i8>, <8 x i8> } poison, <8 x i8> %t0, 0
-  %res1 = insertvalue { <8 x i8>, <8 x i8>, <8 x i8> } %res0, <8 x i8> %t1, 0
-  %res2 = insertvalue { <8 x i8>, <8 x i8>, <8 x i8> } %res1, <8 x i8> %t2, 0
+  %res1 = insertvalue { <8 x i8>, <8 x i8>, <8 x i8> } %res0, <8 x i8> %t1, 1
+  %res2 = insertvalue { <8 x i8>, <8 x i8>, <8 x i8> } %res1, <8 x i8> %t2, 2
   ret { <8 x i8>, <8 x i8>, <8 x i8> } %res2
 }
 
diff --git a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-segN-load.ll b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-segN-load.ll
index 4eed3df..8c3ebb9 100644
--- a/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-segN-load.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/fixed-vectors-segN-load.ll
@@ -1,107 +1,72 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple riscv64 -mattr=+zve64x,+zvl128b < %s | FileCheck %s
 
-define <8 x i8> @load_factor2(ptr %ptr) {
+define {<8 x i8>, <8 x i8>} @load_factor2(ptr %ptr) {
 ; CHECK-LABEL: load_factor2:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
-; CHECK-NEXT:    vlseg2e8.v v7, (a0)
+; CHECK-NEXT:    vlseg2e8.v v8, (a0)
 ; CHECK-NEXT:    ret
     %1 = call { <8 x i8>, <8 x i8> } @llvm.riscv.seg2.load.mask.v8i8.i64(ptr %ptr, <8 x i1> splat (i1 true), i64 8)
-    %2 = extractvalue { <8 x i8>, <8 x i8> } %1, 0
-    %3 = extractvalue { <8 x i8>, <8 x i8> } %1, 1
-    ret <8 x i8> %3
+    ret {<8 x i8>, <8 x i8>} %1
 }
 
-define <8 x i8> @load_factor3(ptr %ptr) {
+define {<8 x i8>, <8 x i8>, <8 x i8>} @load_factor3(ptr %ptr) {
 ; CHECK-LABEL: load_factor3:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
-; CHECK-NEXT:    vlseg3e8.v v6, (a0)
+; CHECK-NEXT:    vlseg3e8.v v8, (a0)
 ; CHECK-NEXT:    ret
     %1 = call { <8 x i8>, <8 x i8>, <8 x i8> } @llvm.riscv.seg3.load.mask.v8i8.i64(ptr %ptr, <8 x i1> splat (i1 true), i64 8)
-    %2 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8> } %1, 0
-    %3 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8> } %1, 1
-    %4 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8> } %1, 2
-    ret <8 x i8> %4
+    ret { <8 x i8>, <8 x i8>, <8 x i8> } %1
 }
 
-define <8 x i8> @load_factor4(ptr %ptr) {
+define {<8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>} @load_factor4(ptr %ptr) {
 ; CHECK-LABEL: load_factor4:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
-; CHECK-NEXT:    vlseg4e8.v v5, (a0)
+; CHECK-NEXT:    vlseg4e8.v v8, (a0)
 ; CHECK-NEXT:    ret
     %1 = call { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } @llvm.riscv.seg4.load.mask.v8i8.i64(ptr %ptr, <8 x i1> splat (i1 true), i64 8)
-    %2 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 0
-    %3 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 1
-    %4 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 2
-    %5 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 3
-    ret <8 x i8> %5
+    ret { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1
 }
 
-define <8 x i8> @load_factor5(ptr %ptr) {
+define {<8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>} @load_factor5(ptr %ptr) {
 ; CHECK-LABEL: load_factor5:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
-; CHECK-NEXT:    vlseg5e8.v v4, (a0)
+; CHECK-NEXT:    vlseg5e8.v v8, (a0)
 ; CHECK-NEXT:    ret
     %1 = call { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } @llvm.riscv.seg5.load.mask.v8i8.i64(ptr %ptr, <8 x i1> splat (i1 true), i64 8)
-    %2 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 0
-    %3 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 1
-    %4 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 2
-    %5 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 3
-    %6 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 4
-    ret <8 x i8> %6
+    ret { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1
 }
 
-define <8 x i8> @load_factor6(ptr %ptr) {
+define {<8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>} @load_factor6(ptr %ptr) {
 ; CHECK-LABEL: load_factor6:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
-; CHECK-NEXT:    vlseg6e8.v v3, (a0)
+; CHECK-NEXT:    vlseg6e8.v v8, (a0)
 ; CHECK-NEXT:    ret
     %1 = call { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } @llvm.riscv.seg6.load.mask.v8i8.i64(ptr %ptr, <8 x i1> splat (i1 true), i64 8)
-    %2 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 0
-    %3 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 1
-    %4 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 2
-    %5 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 3
-    %6 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 4
-    %7 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 5
-    ret <8 x i8> %7
+    ret { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1
 }
 
-define <8 x i8> @load_factor7(ptr %ptr) {
+define {<8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>} @load_factor7(ptr %ptr) {
 ; CHECK-LABEL: load_factor7:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
-; CHECK-NEXT:    vlseg7e8.v v2, (a0)
+; CHECK-NEXT:    vlseg7e8.v v8, (a0)
 ; CHECK-NEXT:    ret
     %1 = call { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } @llvm.riscv.seg7.load.mask.v8i8.i64(ptr %ptr, <8 x i1> splat (i1 true), i64 8)
-    %2 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 0
-    %3 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 1
-    %4 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 2
-    %5 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 3
-    %6 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 4
-    %7 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 5
-    %8 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 6
-    ret <8 x i8> %8
+    ret { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1
 }
 
-define <8 x i8> @load_factor8(ptr %ptr) {
+define {<8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>} @load_factor8(ptr %ptr) {
 ; CHECK-LABEL: load_factor8:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetivli zero, 8, e8, mf2, ta, ma
-; CHECK-NEXT:    vlseg8e8.v v1, (a0)
+; CHECK-NEXT:    vlseg8e8.v v8, (a0)
 ; CHECK-NEXT:    ret
     %1 = call { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } @llvm.riscv.seg8.load.mask.v8i8.i64(ptr %ptr, <8 x i1> splat (i1 true), i64 8)
-    %2 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 0
-    %3 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 1
-    %4 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 2
-    %5 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 3
-    %6 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 4
-    %7 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 5
-    %8 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 6
-    %9 = extractvalue { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1, 7
-    ret <8 x i8> %9
+    ret { <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8>, <8 x i8> } %1
 }
diff --git a/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave-load.ll b/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave-load.ll
index f9f0aa6..fba592d 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave-load.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave-load.ll
@@ -372,7 +372,7 @@ define { <vscale x 8 x i8>, <vscale x 8 x i8>, <vscale x 8 x i8> } @vector_deint
 ; CHECK-LABEL: vector_deinterleave_load_factor3:
 ; CHECK:       # %bb.0:
 ; CHECK-NEXT:    vsetvli a1, zero, e8, m1, ta, ma
-; CHECK-NEXT:    vlseg3e8.v v6, (a0)
+; CHECK-NEXT:    vlseg3e8.v v8, (a0)
 ; CHECK-NEXT:    ret
   %vec = load <vscale x 24 x i8>, ptr %p
   %d0 = call {<vscale x 8 x i8>, <vscale x 8 x i8>, <vscale x 8 x i8>} @llvm.vector.deinterleave3(<vscale x 24 x i8> %vec)
@@ -380,8 +380,8 @@ define { <vscale x 8 x i8>, <vscale x 8 x i8>, <vscale x 8 x i8> } @vector_deint
   %t1 = extractvalue {<vscale x 8 x i8>, <vscale x 8 x i8>, <vscale x 8 x i8>} %d0, 1
   %t2 = extractvalue {<vscale x 8 x i8>, <vscale x 8 x i8>, <vscale x 8 x i8>} %d0, 2
   %res0 = insertvalue { <vscale x 8 x i8>, <vscale x 8 x i8>, <vscale x 8 x i8> } poison, <vscale x 8 x i8> %t0, 0
-  %res1 = insertvalue { <vscale x 8 x i8>, <vscale x 8 x i8>, <vscale x 8 x i8> } %res0, <vscale x 8 x i8> %t1, 0
-  %res2 = insertvalue { <vscale x 8 x i8>, <vscale x 8 x i8>, <vscale x 8 x i8> } %res1, <vscale x 8 x i8> %t2, 0
+  %res1 = insertvalue { <vscale x 8 x i8>, <vscale x 8 x i8>, <vscale x 8 x i8> } %res0, <vscale x 8 x i8> %t1, 1
+  %res2 = insertvalue { <vscale x 8 x i8>, <vscale x 8 x i8>, <vscale x 8 x i8> } %res1, <vscale x 8 x i8> %t2, 2
   ret { <vscale x 8 x i8>, <vscale x 8 x i8>, <vscale x 8 x i8> } %res2
 }
 
@@ -634,3 +634,19 @@ define {<vscale x 8 x i8>, <vscale x 8 x i8>, <vscale x 8 x i8>, <vscale x 8 x i
   %deinterleaved.results = call {<vscale x 8 x i8>, <vscale x 8 x i8>, <vscale x 8 x i8>, <vscale x 8 x i8>} @llvm.vector.deinterleave4.nxv32i8(<vscale x 32 x i8> %vec)
   ret {<vscale x 8 x i8>, <vscale x 8 x i8>, <vscale x 8 x i8>, <vscale x 8 x i8>} %deinterleaved.results
 }
+
+define { <8 x float>, <8 x float> } @deinterleave_unrelated(<16 x float> %arg) {
+; CHECK-LABEL: deinterleave_unrelated:
+; CHECK:       # %bb.0: # %entry
+; CHECK-NEXT:    vsetivli zero, 16, e32, m4, ta, ma
+; CHECK-NEXT:    vfabs.v v12, v8
+; CHECK-NEXT:    li a0, 32
+; CHECK-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
+; CHECK-NEXT:    vnsrl.wx v10, v12, a0
+; CHECK-NEXT:    vnsrl.wi v8, v12, 0
+; CHECK-NEXT:    ret
+entry:
+  %abs = call <16 x float> @llvm.fabs(<16 x float> %arg)
+  %res = call { <8 x float>, <8 x float> } @llvm.vector.deinterleave2.v16f32(<16 x float> %abs)
+  ret { <8 x float>, <8 x float> } %res
+}
diff --git a/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave.ll b/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave.ll
index 0a96e4f..5b1746d 100644
--- a/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave.ll
+++ b/llvm/test/CodeGen/RISCV/rvv/vector-deinterleave.ll
@@ -3744,3 +3744,61 @@ define <vscale x 1 x float> @vector_deinterleave_nxv1f32_nxv8f32_oneactive2(<vsc
   %ext = extractvalue {<vscale x 1 x float>, <vscale x 1 x float>, <vscale x 1 x float>, <vscale x 1 x float>, <vscale x 1 x float>, <vscale x 1 x float>, <vscale x 1 x float>, <vscale x 1 x float>} %res, 5
   ret <vscale x 1 x float> %ext
 }
+
+
+define { <8 x float>, <8 x float> } @interleave_deinterleave2(<8 x float> %a, <8 x float> %b) {
+; V-LABEL: interleave_deinterleave2:
+; V:       # %bb.0: # %entry
+; V-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
+; V-NEXT:    vwaddu.vv v12, v8, v10
+; V-NEXT:    li a0, -1
+; V-NEXT:    vwmaccu.vx v12, a0, v10
+; V-NEXT:    li a0, 32
+; V-NEXT:    vnsrl.wx v10, v12, a0
+; V-NEXT:    vnsrl.wi v8, v12, 0
+; V-NEXT:    ret
+;
+; ZIP-LABEL: interleave_deinterleave2:
+; ZIP:       # %bb.0: # %entry
+; ZIP-NEXT:    vsetivli zero, 16, e32, m4, ta, ma
+; ZIP-NEXT:    vmv2r.v v12, v10
+; ZIP-NEXT:    li a0, 32
+; ZIP-NEXT:    ri.vzip2a.vv v16, v8, v12
+; ZIP-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
+; ZIP-NEXT:    vnsrl.wx v10, v16, a0
+; ZIP-NEXT:    vnsrl.wi v8, v16, 0
+; ZIP-NEXT:    ret
+entry:
+  %0 = call <16 x float> @llvm.vector.interleave2.v16f32(<8 x float> %a, <8 x float> %b)
+  %1 = call { <8 x float>, <8 x float> } @llvm.vector.deinterleave2.v16f32(<16 x float> %0)
+  ret { <8 x float>, <8 x float> } %1
+}
+
+define <16 x float> @deinterleave_interleave2(<16 x float> %arg) {
+; V-LABEL: deinterleave_interleave2:
+; V:       # %bb.0: # %entry
+; V-NEXT:    li a0, 32
+; V-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
+; V-NEXT:    vnsrl.wi v12, v8, 0
+; V-NEXT:    vnsrl.wx v14, v8, a0
+; V-NEXT:    vwaddu.vv v8, v12, v14
+; V-NEXT:    li a0, -1
+; V-NEXT:    vwmaccu.vx v8, a0, v14
+; V-NEXT:    ret
+;
+; ZIP-LABEL: deinterleave_interleave2:
+; ZIP:       # %bb.0: # %entry
+; ZIP-NEXT:    vsetivli zero, 8, e32, m2, ta, ma
+; ZIP-NEXT:    vnsrl.wi v12, v8, 0
+; ZIP-NEXT:    li a0, 32
+; ZIP-NEXT:    vnsrl.wx v16, v8, a0
+; ZIP-NEXT:    vsetivli zero, 16, e32, m4, ta, ma
+; ZIP-NEXT:    ri.vzip2a.vv v8, v12, v16
+; ZIP-NEXT:    ret
+entry:
+  %0 = call { <8 x float>, <8 x float> } @llvm.vector.deinterleave2.v16f32(<16 x float> %arg)
+  %a = extractvalue { <8 x float>, <8 x float> } %0, 0
+  %b = extractvalue { <8 x float>, <8 x float> } %0, 1
+  %res = call <16 x float> @llvm.vector.interleave2.v16f32(<8 x float> %a, <8 x float> %b)
+  ret <16 x float> %res
+}
diff --git a/llvm/test/CodeGen/WebAssembly/libcall_vectorized.ll b/llvm/test/CodeGen/WebAssembly/libcall_vectorized.ll
new file mode 100644
index 0000000..2d1056f
--- /dev/null
+++ b/llvm/test/CodeGen/WebAssembly/libcall_vectorized.ll
@@ -0,0 +1,33 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+
+; RUN: llc < %s -disable-wasm-fallthrough-return-opt -wasm-keep-registers  -mattr=+simd128 | FileCheck %s
+
+target triple = "wasm32-unknown-unknown"
+
+declare <4 x float> @llvm.exp10.v4f32(<4 x float>)
+
+define <4 x float> @exp10_f32v4(<4 x float> %v) {
+; CHECK-LABEL: exp10_f32v4:
+; CHECK:         .functype exp10_f32v4 (v128) -> (v128)
+; CHECK-NEXT:  # %bb.0: # %entry
+; CHECK-NEXT:    local.get $push12=, 0
+; CHECK-NEXT:    f32x4.extract_lane $push0=, $pop12, 0
+; CHECK-NEXT:    call $push1=, exp10f, $pop0
+; CHECK-NEXT:    f32x4.splat $push2=, $pop1
+; CHECK-NEXT:    local.get $push13=, 0
+; CHECK-NEXT:    f32x4.extract_lane $push3=, $pop13, 1
+; CHECK-NEXT:    call $push4=, exp10f, $pop3
+; CHECK-NEXT:    f32x4.replace_lane $push5=, $pop2, 1, $pop4
+; CHECK-NEXT:    local.get $push14=, 0
+; CHECK-NEXT:    f32x4.extract_lane $push6=, $pop14, 2
+; CHECK-NEXT:    call $push7=, exp10f, $pop6
+; CHECK-NEXT:    f32x4.replace_lane $push8=, $pop5, 2, $pop7
+; CHECK-NEXT:    local.get $push15=, 0
+; CHECK-NEXT:    f32x4.extract_lane $push9=, $pop15, 3
+; CHECK-NEXT:    call $push10=, exp10f, $pop9
+; CHECK-NEXT:    f32x4.replace_lane $push11=, $pop8, 3, $pop10
+; CHECK-NEXT:    return $pop11
+entry:
+  %r = call <4 x float> @llvm.exp10.v4f32(<4 x float> %v)
+  ret <4 x float> %r
+}
diff --git a/llvm/test/CodeGen/WebAssembly/ref-test-func.ll b/llvm/test/CodeGen/WebAssembly/ref-test-func.ll
index e4014ba..ea2453f 100644
--- a/llvm/test/CodeGen/WebAssembly/ref-test-func.ll
+++ b/llvm/test/CodeGen/WebAssembly/ref-test-func.ll
@@ -1,6 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
-; RUN: llc < %s --mtriple=wasm32-unknown-unknown -mcpu=mvp -mattr=+reference-types -verify-machineinstrs | FileCheck --check-prefixes CHECK,CHK32 %s
-; RUN: llc < %s --mtriple=wasm64-unknown-unknown -mcpu=mvp -mattr=+reference-types -verify-machineinstrs | FileCheck --check-prefixes CHECK,CHK64 %s
+; RUN: llc < %s --mtriple=wasm32-unknown-unknown -mcpu=mvp -mattr=+reference-types -mattr=+gc -verify-machineinstrs | FileCheck --check-prefixes CHECK,CHK32 %s
+; RUN: llc < %s --mtriple=wasm64-unknown-unknown -mcpu=mvp -mattr=+reference-types -mattr=+gc -verify-machineinstrs | FileCheck --check-prefixes CHECK,CHK64 %s
 
 define void @test_fpsig_void_void(ptr noundef %func) local_unnamed_addr #0 {
 ; CHECK-LABEL: test_fpsig_void_void:
diff --git a/llvm/test/CodeGen/WebAssembly/target-features-cpus.ll b/llvm/test/CodeGen/WebAssembly/target-features-cpus.ll
index 1c77ad5..60cfc27 100644
--- a/llvm/test/CodeGen/WebAssembly/target-features-cpus.ll
+++ b/llvm/test/CodeGen/WebAssembly/target-features-cpus.ll
@@ -70,9 +70,9 @@ target triple = "wasm32-unknown-unknown"
 ;                +call-indirect-overlong, +exception-handling,
 ;                +extended-const, +fp16, +multimemory, +multivalue,
 ;                +mutable-globals, +nontrapping-fptoint, +relaxed-simd,
-;                +reference-types, +simd128, +sign-ext, +tail-call
+;                +reference-types, +simd128, +sign-ext, +tail-call, +gc
 ; BLEEDING-EDGE-LABEL: .section  .custom_section.target_features,"",@
-; BLEEDING-EDGE-NEXT: .int8  16
+; BLEEDING-EDGE-NEXT: .int8  17
 ; BLEEDING-EDGE-NEXT: .int8  43
 ; BLEEDING-EDGE-NEXT: .int8  7
 ; BLEEDING-EDGE-NEXT: .ascii  "atomics"
@@ -95,6 +95,9 @@ target triple = "wasm32-unknown-unknown"
 ; BLEEDING-EDGE-NEXT: .int8  4
 ; BLEEDING-EDGE-NEXT: .ascii  "fp16"
 ; BLEEDING-EDGE-NEXT: .int8  43
+; BLEEDING-EDGE-NEXT: .int8  2
+; BLEEDING-EDGE-NEXT: .ascii  "gc"
+; BLEEDING-EDGE-NEXT: .int8  43
 ; BLEEDING-EDGE-NEXT: .int8  11
 ; BLEEDING-EDGE-NEXT: .ascii  "multimemory"
 ; BLEEDING-EDGE-NEXT: .int8  43
diff --git a/llvm/test/CodeGen/X86/embed-bitcode.ll b/llvm/test/CodeGen/X86/embed-bitcode.ll
index 0d66ba8..d4af954 100644
--- a/llvm/test/CodeGen/X86/embed-bitcode.ll
+++ b/llvm/test/CodeGen/X86/embed-bitcode.ll
@@ -1,10 +1,23 @@
 ; RUN: llc -filetype=obj -mtriple=x86_64 %s -o %t
 ; RUN: llvm-readelf -S %t | FileCheck %s
+; RUN: llc -filetype=obj -mtriple=x86_64-pc-windows-msvc %s -o %t
+; RUN: llvm-readobj -S %t | FileCheck %s --check-prefix=COFF
 
 ; CHECK:      .text    PROGBITS 0000000000000000 [[#%x,OFF:]] 000000 00 AX 0
 ; CHECK-NEXT: .llvmbc  PROGBITS 0000000000000000 [[#%x,OFF:]] 000004 00    0
 ; CHECK-NEXT: .llvmcmd PROGBITS 0000000000000000 [[#%x,OFF:]] 000005 00    0
 
+; COFF:      Name: .llvmbc (2E 6C 6C 76 6D 62 63 00)
+; COFF:      Characteristics [
+; COFF-NEXT:   IMAGE_SCN_ALIGN_1BYTES
+; COFF-NEXT:   IMAGE_SCN_MEM_DISCARDABLE
+; COFF-NEXT: ]
+; COFF:      Name: .llvmcmd (2E 6C 6C 76 6D 63 6D 64)
+; COFF:      Characteristics [
+; COFF-NEXT:   IMAGE_SCN_ALIGN_1BYTES
+; COFF-NEXT:   IMAGE_SCN_MEM_DISCARDABLE
+; COFF-NEXT: ]
+
 @llvm.embedded.module = private constant [4 x i8] c"BC\C0\DE", section ".llvmbc", align 1
 @llvm.cmdline = private constant [5 x i8] c"-cc1\00", section ".llvmcmd", align 1
 @llvm.compiler.used = appending global [2 x ptr] [ptr @llvm.embedded.module, ptr @llvm.cmdline], section "llvm.metadata"
diff --git a/llvm/test/CodeGen/X86/isel-fpclass.ll b/llvm/test/CodeGen/X86/isel-fpclass.ll
new file mode 100644
index 0000000..960bbf5
--- /dev/null
+++ b/llvm/test/CodeGen/X86/isel-fpclass.ll
@@ -0,0 +1,526 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc < %s -mtriple=i686-linux | FileCheck %s -check-prefixes=X86-SDAGISEL
+; RUN: llc < %s -mtriple=x86_64-linux | FileCheck %s -check-prefixes=X64,X64-SDAGISEL
+; RUN: llc < %s -mtriple=i686-linux -fast-isel -fast-isel-abort=1  | FileCheck %s -check-prefixes=X86-FASTISEL
+; RUN: llc < %s -mtriple=x86_64-linux -fast-isel -fast-isel-abort=1  | FileCheck %s -check-prefixes=X64,X64-FASTISEL
+
+; FIXME: We can reuse/delete llvm/test/CodeGen/X86/is_fpclass.ll when all patches are included.
+
+define i1 @isnone_f(float %x) {
+; X86-SDAGISEL-LABEL: isnone_f:
+; X86-SDAGISEL:       # %bb.0: # %entry
+; X86-SDAGISEL-NEXT:    xorl %eax, %eax
+; X86-SDAGISEL-NEXT:    retl
+;
+; X64-LABEL: isnone_f:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    xorl %eax, %eax
+; X64-NEXT:    retq
+;
+; X86-FASTISEL-LABEL: isnone_f:
+; X86-FASTISEL:       # %bb.0: # %entry
+; X86-FASTISEL-NEXT:    flds {{[0-9]+}}(%esp)
+; X86-FASTISEL-NEXT:    fstp %st(0)
+; X86-FASTISEL-NEXT:    xorl %eax, %eax
+; X86-FASTISEL-NEXT:    retl
+entry:
+  %0 = tail call i1 @llvm.is.fpclass.f32(float %x, i32 0)
+  ret i1 %0
+}
+
+define i1 @isany_f(float %x) {
+; X86-SDAGISEL-LABEL: isany_f:
+; X86-SDAGISEL:       # %bb.0: # %entry
+; X86-SDAGISEL-NEXT:    movb $1, %al
+; X86-SDAGISEL-NEXT:    retl
+;
+; X64-LABEL: isany_f:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movb $1, %al
+; X64-NEXT:    retq
+;
+; X86-FASTISEL-LABEL: isany_f:
+; X86-FASTISEL:       # %bb.0: # %entry
+; X86-FASTISEL-NEXT:    flds {{[0-9]+}}(%esp)
+; X86-FASTISEL-NEXT:    fstp %st(0)
+; X86-FASTISEL-NEXT:    movb $1, %al
+; X86-FASTISEL-NEXT:    retl
+entry:
+  %0 = tail call i1 @llvm.is.fpclass.f32(float %x, i32 1023)
+  ret i1 %0
+}
+
+define i1 @issignaling_f(float %x) {
+; X86-SDAGISEL-LABEL: issignaling_f:
+; X86-SDAGISEL:       # %bb.0:
+; X86-SDAGISEL-NEXT:    movl $2147483647, %eax # imm = 0x7FFFFFFF
+; X86-SDAGISEL-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-SDAGISEL-NEXT:    cmpl $2143289344, %eax # imm = 0x7FC00000
+; X86-SDAGISEL-NEXT:    setl %cl
+; X86-SDAGISEL-NEXT:    cmpl $2139095041, %eax # imm = 0x7F800001
+; X86-SDAGISEL-NEXT:    setge %al
+; X86-SDAGISEL-NEXT:    andb %cl, %al
+; X86-SDAGISEL-NEXT:    retl
+;
+; X64-LABEL: issignaling_f:
+; X64:       # %bb.0:
+; X64-NEXT:    movd %xmm0, %eax
+; X64-NEXT:    andl $2147483647, %eax # imm = 0x7FFFFFFF
+; X64-NEXT:    cmpl $2143289344, %eax # imm = 0x7FC00000
+; X64-NEXT:    setl %cl
+; X64-NEXT:    cmpl $2139095041, %eax # imm = 0x7F800001
+; X64-NEXT:    setge %al
+; X64-NEXT:    andb %cl, %al
+; X64-NEXT:    retq
+;
+; X86-FASTISEL-LABEL: issignaling_f:
+; X86-FASTISEL:       # %bb.0:
+; X86-FASTISEL-NEXT:    pushl %eax
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 8
+; X86-FASTISEL-NEXT:    flds {{[0-9]+}}(%esp)
+; X86-FASTISEL-NEXT:    fstps (%esp)
+; X86-FASTISEL-NEXT:    movl $2147483647, %eax # imm = 0x7FFFFFFF
+; X86-FASTISEL-NEXT:    andl (%esp), %eax
+; X86-FASTISEL-NEXT:    cmpl $2143289344, %eax # imm = 0x7FC00000
+; X86-FASTISEL-NEXT:    setl %cl
+; X86-FASTISEL-NEXT:    cmpl $2139095041, %eax # imm = 0x7F800001
+; X86-FASTISEL-NEXT:    setge %al
+; X86-FASTISEL-NEXT:    andb %cl, %al
+; X86-FASTISEL-NEXT:    popl %ecx
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 4
+; X86-FASTISEL-NEXT:    retl
+   %a0 = tail call i1 @llvm.is.fpclass.f32(float %x, i32 1)  ; "snan"
+   ret i1 %a0
+}
+
+ define i1 @isquiet_f(float %x) {
+; X86-SDAGISEL-LABEL: isquiet_f:
+; X86-SDAGISEL:       # %bb.0: # %entry
+; X86-SDAGISEL-NEXT:    movl $2147483647, %eax # imm = 0x7FFFFFFF
+; X86-SDAGISEL-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-SDAGISEL-NEXT:    cmpl $2143289344, %eax # imm = 0x7FC00000
+; X86-SDAGISEL-NEXT:    setge %al
+; X86-SDAGISEL-NEXT:    retl
+;
+; X64-LABEL: isquiet_f:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movd %xmm0, %eax
+; X64-NEXT:    andl $2147483647, %eax # imm = 0x7FFFFFFF
+; X64-NEXT:    cmpl $2143289344, %eax # imm = 0x7FC00000
+; X64-NEXT:    setge %al
+; X64-NEXT:    retq
+;
+; X86-FASTISEL-LABEL: isquiet_f:
+; X86-FASTISEL:       # %bb.0: # %entry
+; X86-FASTISEL-NEXT:    pushl %eax
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 8
+; X86-FASTISEL-NEXT:    flds {{[0-9]+}}(%esp)
+; X86-FASTISEL-NEXT:    fstps (%esp)
+; X86-FASTISEL-NEXT:    movl $2147483647, %eax # imm = 0x7FFFFFFF
+; X86-FASTISEL-NEXT:    andl (%esp), %eax
+; X86-FASTISEL-NEXT:    cmpl $2143289344, %eax # imm = 0x7FC00000
+; X86-FASTISEL-NEXT:    setge %al
+; X86-FASTISEL-NEXT:    popl %ecx
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 4
+; X86-FASTISEL-NEXT:    retl
+ entry:
+   %0 = tail call i1 @llvm.is.fpclass.f32(float %x, i32 2)  ; "qnan"
+   ret i1 %0
+}
+
+define i1 @not_isquiet_f(float %x) {
+; X86-SDAGISEL-LABEL: not_isquiet_f:
+; X86-SDAGISEL:       # %bb.0: # %entry
+; X86-SDAGISEL-NEXT:    movl $2147483647, %eax # imm = 0x7FFFFFFF
+; X86-SDAGISEL-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-SDAGISEL-NEXT:    cmpl $2143289344, %eax # imm = 0x7FC00000
+; X86-SDAGISEL-NEXT:    setl %al
+; X86-SDAGISEL-NEXT:    retl
+;
+; X64-LABEL: not_isquiet_f:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movd %xmm0, %eax
+; X64-NEXT:    andl $2147483647, %eax # imm = 0x7FFFFFFF
+; X64-NEXT:    cmpl $2143289344, %eax # imm = 0x7FC00000
+; X64-NEXT:    setl %al
+; X64-NEXT:    retq
+;
+; X86-FASTISEL-LABEL: not_isquiet_f:
+; X86-FASTISEL:       # %bb.0: # %entry
+; X86-FASTISEL-NEXT:    pushl %eax
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 8
+; X86-FASTISEL-NEXT:    flds {{[0-9]+}}(%esp)
+; X86-FASTISEL-NEXT:    fstps (%esp)
+; X86-FASTISEL-NEXT:    movl $2147483647, %eax # imm = 0x7FFFFFFF
+; X86-FASTISEL-NEXT:    andl (%esp), %eax
+; X86-FASTISEL-NEXT:    cmpl $2143289344, %eax # imm = 0x7FC00000
+; X86-FASTISEL-NEXT:    setl %al
+; X86-FASTISEL-NEXT:    popl %ecx
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 4
+; X86-FASTISEL-NEXT:    retl
+entry:
+  %0 = tail call i1 @llvm.is.fpclass.f32(float %x, i32 1021)  ; ~"qnan"
+  ret i1 %0
+}
+
+define i1 @isinf_f(float %x) {
+; X86-SDAGISEL-LABEL: isinf_f:
+; X86-SDAGISEL:       # %bb.0: # %entry
+; X86-SDAGISEL-NEXT:    movl $2147483647, %eax # imm = 0x7FFFFFFF
+; X86-SDAGISEL-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-SDAGISEL-NEXT:    cmpl $2139095040, %eax # imm = 0x7F800000
+; X86-SDAGISEL-NEXT:    sete %al
+; X86-SDAGISEL-NEXT:    retl
+;
+; X64-LABEL: isinf_f:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movd %xmm0, %eax
+; X64-NEXT:    andl $2147483647, %eax # imm = 0x7FFFFFFF
+; X64-NEXT:    cmpl $2139095040, %eax # imm = 0x7F800000
+; X64-NEXT:    sete %al
+; X64-NEXT:    retq
+;
+; X86-FASTISEL-LABEL: isinf_f:
+; X86-FASTISEL:       # %bb.0: # %entry
+; X86-FASTISEL-NEXT:    pushl %eax
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 8
+; X86-FASTISEL-NEXT:    flds {{[0-9]+}}(%esp)
+; X86-FASTISEL-NEXT:    fstps (%esp)
+; X86-FASTISEL-NEXT:    movl $2147483647, %eax # imm = 0x7FFFFFFF
+; X86-FASTISEL-NEXT:    andl (%esp), %eax
+; X86-FASTISEL-NEXT:    cmpl $2139095040, %eax # imm = 0x7F800000
+; X86-FASTISEL-NEXT:    sete %al
+; X86-FASTISEL-NEXT:    popl %ecx
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 4
+; X86-FASTISEL-NEXT:    retl
+entry:
+  %0 = tail call i1 @llvm.is.fpclass.f32(float %x, i32 516)  ; 0x204 = "inf"
+  ret i1 %0
+}
+
+define i1 @not_isinf_f(float %x) {
+; X86-SDAGISEL-LABEL: not_isinf_f:
+; X86-SDAGISEL:       # %bb.0: # %entry
+; X86-SDAGISEL-NEXT:    movl $2147483647, %eax # imm = 0x7FFFFFFF
+; X86-SDAGISEL-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-SDAGISEL-NEXT:    cmpl $2139095040, %eax # imm = 0x7F800000
+; X86-SDAGISEL-NEXT:    setne %al
+; X86-SDAGISEL-NEXT:    retl
+;
+; X64-LABEL: not_isinf_f:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movd %xmm0, %eax
+; X64-NEXT:    andl $2147483647, %eax # imm = 0x7FFFFFFF
+; X64-NEXT:    cmpl $2139095040, %eax # imm = 0x7F800000
+; X64-NEXT:    setne %al
+; X64-NEXT:    retq
+;
+; X86-FASTISEL-LABEL: not_isinf_f:
+; X86-FASTISEL:       # %bb.0: # %entry
+; X86-FASTISEL-NEXT:    pushl %eax
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 8
+; X86-FASTISEL-NEXT:    flds {{[0-9]+}}(%esp)
+; X86-FASTISEL-NEXT:    fstps (%esp)
+; X86-FASTISEL-NEXT:    movl $2147483647, %eax # imm = 0x7FFFFFFF
+; X86-FASTISEL-NEXT:    andl (%esp), %eax
+; X86-FASTISEL-NEXT:    cmpl $2139095040, %eax # imm = 0x7F800000
+; X86-FASTISEL-NEXT:    setne %al
+; X86-FASTISEL-NEXT:    popl %ecx
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 4
+; X86-FASTISEL-NEXT:    retl
+entry:
+  %0 = tail call i1 @llvm.is.fpclass.f32(float %x, i32 507)  ; ~0x204 = "~inf"
+  ret i1 %0
+}
+
+define i1 @is_plus_inf_f(float %x) {
+; X86-SDAGISEL-LABEL: is_plus_inf_f:
+; X86-SDAGISEL:       # %bb.0: # %entry
+; X86-SDAGISEL-NEXT:    cmpl $2139095040, {{[0-9]+}}(%esp) # imm = 0x7F800000
+; X86-SDAGISEL-NEXT:    sete %al
+; X86-SDAGISEL-NEXT:    retl
+;
+; X64-LABEL: is_plus_inf_f:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movd %xmm0, %eax
+; X64-NEXT:    cmpl $2139095040, %eax # imm = 0x7F800000
+; X64-NEXT:    sete %al
+; X64-NEXT:    retq
+;
+; X86-FASTISEL-LABEL: is_plus_inf_f:
+; X86-FASTISEL:       # %bb.0: # %entry
+; X86-FASTISEL-NEXT:    pushl %eax
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 8
+; X86-FASTISEL-NEXT:    flds {{[0-9]+}}(%esp)
+; X86-FASTISEL-NEXT:    fstps (%esp)
+; X86-FASTISEL-NEXT:    cmpl $2139095040, (%esp) # imm = 0x7F800000
+; X86-FASTISEL-NEXT:    sete %al
+; X86-FASTISEL-NEXT:    popl %ecx
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 4
+; X86-FASTISEL-NEXT:    retl
+entry:
+  %0 = tail call i1 @llvm.is.fpclass.f32(float %x, i32 512)  ; 0x200 = "+inf"
+  ret i1 %0
+}
+
+define i1 @is_minus_inf_f(float %x) {
+; X86-SDAGISEL-LABEL: is_minus_inf_f:
+; X86-SDAGISEL:       # %bb.0: # %entry
+; X86-SDAGISEL-NEXT:    cmpl $-8388608, {{[0-9]+}}(%esp) # imm = 0xFF800000
+; X86-SDAGISEL-NEXT:    sete %al
+; X86-SDAGISEL-NEXT:    retl
+;
+; X64-LABEL: is_minus_inf_f:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movd %xmm0, %eax
+; X64-NEXT:    cmpl $-8388608, %eax # imm = 0xFF800000
+; X64-NEXT:    sete %al
+; X64-NEXT:    retq
+;
+; X86-FASTISEL-LABEL: is_minus_inf_f:
+; X86-FASTISEL:       # %bb.0: # %entry
+; X86-FASTISEL-NEXT:    pushl %eax
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 8
+; X86-FASTISEL-NEXT:    flds {{[0-9]+}}(%esp)
+; X86-FASTISEL-NEXT:    fstps (%esp)
+; X86-FASTISEL-NEXT:    cmpl $-8388608, (%esp) # imm = 0xFF800000
+; X86-FASTISEL-NEXT:    sete %al
+; X86-FASTISEL-NEXT:    popl %ecx
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 4
+; X86-FASTISEL-NEXT:    retl
+entry:
+  %0 = tail call i1 @llvm.is.fpclass.f32(float %x, i32 4)  ; "-inf"
+  ret i1 %0
+}
+
+define i1 @not_is_minus_inf_f(float %x) {
+; X86-SDAGISEL-LABEL: not_is_minus_inf_f:
+; X86-SDAGISEL:       # %bb.0: # %entry
+; X86-SDAGISEL-NEXT:    cmpl $-8388608, {{[0-9]+}}(%esp) # imm = 0xFF800000
+; X86-SDAGISEL-NEXT:    setne %al
+; X86-SDAGISEL-NEXT:    retl
+;
+; X64-LABEL: not_is_minus_inf_f:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movd %xmm0, %eax
+; X64-NEXT:    cmpl $-8388608, %eax # imm = 0xFF800000
+; X64-NEXT:    setne %al
+; X64-NEXT:    retq
+;
+; X86-FASTISEL-LABEL: not_is_minus_inf_f:
+; X86-FASTISEL:       # %bb.0: # %entry
+; X86-FASTISEL-NEXT:    pushl %eax
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 8
+; X86-FASTISEL-NEXT:    flds {{[0-9]+}}(%esp)
+; X86-FASTISEL-NEXT:    fstps (%esp)
+; X86-FASTISEL-NEXT:    cmpl $-8388608, (%esp) # imm = 0xFF800000
+; X86-FASTISEL-NEXT:    setne %al
+; X86-FASTISEL-NEXT:    popl %ecx
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 4
+; X86-FASTISEL-NEXT:    retl
+entry:
+  %0 = tail call i1 @llvm.is.fpclass.f32(float %x, i32 1019)  ; ~"-inf"
+  ret i1 %0
+}
+
+define i1 @isfinite_f(float %x) {
+; X86-SDAGISEL-LABEL: isfinite_f:
+; X86-SDAGISEL:       # %bb.0: # %entry
+; X86-SDAGISEL-NEXT:    movl $2147483647, %eax # imm = 0x7FFFFFFF
+; X86-SDAGISEL-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-SDAGISEL-NEXT:    cmpl $2139095040, %eax # imm = 0x7F800000
+; X86-SDAGISEL-NEXT:    setl %al
+; X86-SDAGISEL-NEXT:    retl
+;
+; X64-LABEL: isfinite_f:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movd %xmm0, %eax
+; X64-NEXT:    andl $2147483647, %eax # imm = 0x7FFFFFFF
+; X64-NEXT:    cmpl $2139095040, %eax # imm = 0x7F800000
+; X64-NEXT:    setl %al
+; X64-NEXT:    retq
+;
+; X86-FASTISEL-LABEL: isfinite_f:
+; X86-FASTISEL:       # %bb.0: # %entry
+; X86-FASTISEL-NEXT:    pushl %eax
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 8
+; X86-FASTISEL-NEXT:    flds {{[0-9]+}}(%esp)
+; X86-FASTISEL-NEXT:    fstps (%esp)
+; X86-FASTISEL-NEXT:    movl $2147483647, %eax # imm = 0x7FFFFFFF
+; X86-FASTISEL-NEXT:    andl (%esp), %eax
+; X86-FASTISEL-NEXT:    cmpl $2139095040, %eax # imm = 0x7F800000
+; X86-FASTISEL-NEXT:    setl %al
+; X86-FASTISEL-NEXT:    popl %ecx
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 4
+; X86-FASTISEL-NEXT:    retl
+entry:
+  %0 = tail call i1 @llvm.is.fpclass.f32(float %x, i32 504)  ; 0x1f8 = "finite"
+  ret i1 %0
+}
+
+define i1 @not_isfinite_f(float %x) {
+; X86-SDAGISEL-LABEL: not_isfinite_f:
+; X86-SDAGISEL:       # %bb.0: # %entry
+; X86-SDAGISEL-NEXT:    movl $2147483647, %eax # imm = 0x7FFFFFFF
+; X86-SDAGISEL-NEXT:    andl {{[0-9]+}}(%esp), %eax
+; X86-SDAGISEL-NEXT:    cmpl $2139095040, %eax # imm = 0x7F800000
+; X86-SDAGISEL-NEXT:    setge %al
+; X86-SDAGISEL-NEXT:    retl
+;
+; X64-LABEL: not_isfinite_f:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movd %xmm0, %eax
+; X64-NEXT:    andl $2147483647, %eax # imm = 0x7FFFFFFF
+; X64-NEXT:    cmpl $2139095040, %eax # imm = 0x7F800000
+; X64-NEXT:    setge %al
+; X64-NEXT:    retq
+;
+; X86-FASTISEL-LABEL: not_isfinite_f:
+; X86-FASTISEL:       # %bb.0: # %entry
+; X86-FASTISEL-NEXT:    pushl %eax
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 8
+; X86-FASTISEL-NEXT:    flds {{[0-9]+}}(%esp)
+; X86-FASTISEL-NEXT:    fstps (%esp)
+; X86-FASTISEL-NEXT:    movl $2147483647, %eax # imm = 0x7FFFFFFF
+; X86-FASTISEL-NEXT:    andl (%esp), %eax
+; X86-FASTISEL-NEXT:    cmpl $2139095040, %eax # imm = 0x7F800000
+; X86-FASTISEL-NEXT:    setge %al
+; X86-FASTISEL-NEXT:    popl %ecx
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 4
+; X86-FASTISEL-NEXT:    retl
+entry:
+  %0 = tail call i1 @llvm.is.fpclass.f32(float %x, i32 519)  ; ~0x1f8 = "~finite"
+  ret i1 %0
+}
+
+define i1 @is_plus_finite_f(float %x) {
+; X86-SDAGISEL-LABEL: is_plus_finite_f:
+; X86-SDAGISEL:       # %bb.0: # %entry
+; X86-SDAGISEL-NEXT:    cmpl $2139095040, {{[0-9]+}}(%esp) # imm = 0x7F800000
+; X86-SDAGISEL-NEXT:    setb %al
+; X86-SDAGISEL-NEXT:    retl
+;
+; X64-LABEL: is_plus_finite_f:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movd %xmm0, %eax
+; X64-NEXT:    cmpl $2139095040, %eax # imm = 0x7F800000
+; X64-NEXT:    setb %al
+; X64-NEXT:    retq
+;
+; X86-FASTISEL-LABEL: is_plus_finite_f:
+; X86-FASTISEL:       # %bb.0: # %entry
+; X86-FASTISEL-NEXT:    pushl %eax
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 8
+; X86-FASTISEL-NEXT:    flds {{[0-9]+}}(%esp)
+; X86-FASTISEL-NEXT:    fstps (%esp)
+; X86-FASTISEL-NEXT:    cmpl $2139095040, (%esp) # imm = 0x7F800000
+; X86-FASTISEL-NEXT:    setb %al
+; X86-FASTISEL-NEXT:    popl %ecx
+; X86-FASTISEL-NEXT:    .cfi_def_cfa_offset 4
+; X86-FASTISEL-NEXT:    retl
+entry:
+  %0 = tail call i1 @llvm.is.fpclass.f32(float %x, i32 448)  ; 0x1c0 = "+finite"
+  ret i1 %0
+}
+
+define i1 @isnone_d(double %x) nounwind {
+; X86-SDAGISEL-LABEL: isnone_d:
+; X86-SDAGISEL:       # %bb.0: # %entry
+; X86-SDAGISEL-NEXT:    xorl %eax, %eax
+; X86-SDAGISEL-NEXT:    retl
+;
+; X64-LABEL: isnone_d:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    xorl %eax, %eax
+; X64-NEXT:    retq
+;
+; X86-FASTISEL-LABEL: isnone_d:
+; X86-FASTISEL:       # %bb.0: # %entry
+; X86-FASTISEL-NEXT:    fldl {{[0-9]+}}(%esp)
+; X86-FASTISEL-NEXT:    fstp %st(0)
+; X86-FASTISEL-NEXT:    xorl %eax, %eax
+; X86-FASTISEL-NEXT:    retl
+entry:
+    %0 = tail call i1 @llvm.is.fpclass.f64(double %x, i32 0)
+    ret i1 %0
+}
+
+define i1 @isany_d(double %x) nounwind {
+; X86-SDAGISEL-LABEL: isany_d:
+; X86-SDAGISEL:       # %bb.0: # %entry
+; X86-SDAGISEL-NEXT:    movb $1, %al
+; X86-SDAGISEL-NEXT:    retl
+;
+; X64-LABEL: isany_d:
+; X64:       # %bb.0: # %entry
+; X64-NEXT:    movb $1, %al
+; X64-NEXT:    retq
+;
+; X86-FASTISEL-LABEL: isany_d:
+; X86-FASTISEL:       # %bb.0: # %entry
+; X86-FASTISEL-NEXT:    fldl {{[0-9]+}}(%esp)
+; X86-FASTISEL-NEXT:    fstp %st(0)
+; X86-FASTISEL-NEXT:    movb $1, %al
+; X86-FASTISEL-NEXT:    retl
+entry:
+    %0 = tail call i1 @llvm.is.fpclass.f64(double %x, i32 1023)
+    ret i1 %0
+}
+
+define i1 @isnone_f80(x86_fp80 %x) nounwind {
+; X86-SDAGISEL-LABEL: isnone_f80:
+; X86-SDAGISEL:       # %bb.0: # %entry
+; X86-SDAGISEL-NEXT:    xorl %eax, %eax
+; X86-SDAGISEL-NEXT:    retl
+;
+; X64-SDAGISEL-LABEL: isnone_f80:
+; X64-SDAGISEL:       # %bb.0: # %entry
+; X64-SDAGISEL-NEXT:    xorl %eax, %eax
+; X64-SDAGISEL-NEXT:    retq
+;
+; X86-FASTISEL-LABEL: isnone_f80:
+; X86-FASTISEL:       # %bb.0: # %entry
+; X86-FASTISEL-NEXT:    fldt {{[0-9]+}}(%esp)
+; X86-FASTISEL-NEXT:    fstp %st(0)
+; X86-FASTISEL-NEXT:    xorl %eax, %eax
+; X86-FASTISEL-NEXT:    retl
+;
+; X64-FASTISEL-LABEL: isnone_f80:
+; X64-FASTISEL:       # %bb.0: # %entry
+; X64-FASTISEL-NEXT:    fldt {{[0-9]+}}(%rsp)
+; X64-FASTISEL-NEXT:    fstp %st(0)
+; X64-FASTISEL-NEXT:    xorl %eax, %eax
+; X64-FASTISEL-NEXT:    retq
+entry:
+%0 = tail call i1 @llvm.is.fpclass.f80(x86_fp80 %x, i32 0)
+ret i1 %0
+}
+
+define i1 @isany_f80(x86_fp80 %x) nounwind {
+; X86-SDAGISEL-LABEL: isany_f80:
+; X86-SDAGISEL:       # %bb.0: # %entry
+; X86-SDAGISEL-NEXT:    movb $1, %al
+; X86-SDAGISEL-NEXT:    retl
+;
+; X64-SDAGISEL-LABEL: isany_f80:
+; X64-SDAGISEL:       # %bb.0: # %entry
+; X64-SDAGISEL-NEXT:    movb $1, %al
+; X64-SDAGISEL-NEXT:    retq
+;
+; X86-FASTISEL-LABEL: isany_f80:
+; X86-FASTISEL:       # %bb.0: # %entry
+; X86-FASTISEL-NEXT:    fldt {{[0-9]+}}(%esp)
+; X86-FASTISEL-NEXT:    fstp %st(0)
+; X86-FASTISEL-NEXT:    movb $1, %al
+; X86-FASTISEL-NEXT:    retl
+;
+; X64-FASTISEL-LABEL: isany_f80:
+; X64-FASTISEL:       # %bb.0: # %entry
+; X64-FASTISEL-NEXT:    fldt {{[0-9]+}}(%rsp)
+; X64-FASTISEL-NEXT:    fstp %st(0)
+; X64-FASTISEL-NEXT:    movb $1, %al
+; X64-FASTISEL-NEXT:    retq
+entry:
+    %0 = tail call i1 @llvm.is.fpclass.f80(x86_fp80 %x, i32 1023)
+    ret i1 %0
+}
diff --git a/llvm/test/CodeGen/X86/late-tail-dup-computed-goto.mir b/llvm/test/CodeGen/X86/late-tail-dup-computed-goto.mir
new file mode 100644
index 0000000..e272e7e
--- /dev/null
+++ b/llvm/test/CodeGen/X86/late-tail-dup-computed-goto.mir
@@ -0,0 +1,128 @@
+# NOTE: Assertions have been autogenerated by utils/update_mir_test_checks.py UTC_ARGS: --version 5
+# RUN: llc -mtriple=x86_64-unknown-linux-gnu -run-pass=tailduplication -tail-dup-pred-size=1 -tail-dup-succ-size=1 %s -o - | FileCheck %s
+#
+# Check that only the computed gotos are duplicated aggressively.
+#
+--- |
+  @computed_goto.dispatch = constant [5 x ptr] [ptr null, ptr blockaddress(@computed_goto, %bb1), ptr blockaddress(@computed_goto, %bb2), ptr blockaddress(@computed_goto, %bb3), ptr blockaddress(@computed_goto, %bb4)]
+  declare i64 @f0()
+  declare i64 @f1()
+  declare i64 @f2()
+  declare i64 @f3()
+  declare i64 @f4()
+  declare i64 @f5()
+  define void @computed_goto() {
+    start:
+      ret void
+    bb1:
+      ret void
+    bb2:
+      ret void
+    bb3:
+      ret void
+    bb4:
+      ret void
+  }
+  define void @jump_table() { ret void }
+  define void @jump_table_pic() { ret void }
+...
+---
+name:            computed_goto
+alignment:       1
+exposesReturnsTwice: false
+legalized:       false
+regBankSelected: false
+selected:        false
+failedISel:      false
+tracksRegLiveness: true
+body:             |
+  ; CHECK-LABEL: name: computed_goto
+  ; CHECK: bb.0:
+  ; CHECK-NEXT:   successors: %bb.1(0x20000000), %bb.2(0x20000000), %bb.3(0x20000000), %bb.4(0x20000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   CALL64pcrel32 target-flags(x86-plt) @f0, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+  ; CHECK-NEXT:   [[COPY:%[0-9]+]]:gr64 = COPY $rax
+  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:gr64_nosp = COPY [[COPY]]
+  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:gr64_nosp = COPY [[COPY1]]
+  ; CHECK-NEXT:   JMP64m $noreg, 8, [[COPY2]], @computed_goto.dispatch, $noreg
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.1.bb1 (ir-block-address-taken %ir-block.bb1):
+  ; CHECK-NEXT:   successors: %bb.1(0x20000000), %bb.2(0x20000000), %bb.3(0x20000000), %bb.4(0x20000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   CALL64pcrel32 target-flags(x86-plt) @f1, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+  ; CHECK-NEXT:   [[COPY3:%[0-9]+]]:gr64 = COPY $rax
+  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:gr64_nosp = COPY [[COPY3]]
+  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:gr64_nosp = COPY [[COPY1]]
+  ; CHECK-NEXT:   JMP64m $noreg, 8, [[COPY2]], @computed_goto.dispatch, $noreg
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.2.bb2 (ir-block-address-taken %ir-block.bb2):
+  ; CHECK-NEXT:   successors: %bb.1(0x20000000), %bb.2(0x20000000), %bb.3(0x20000000), %bb.4(0x20000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   CALL64pcrel32 target-flags(x86-plt) @f2, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+  ; CHECK-NEXT:   [[COPY4:%[0-9]+]]:gr64 = COPY $rax
+  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:gr64_nosp = COPY [[COPY4]]
+  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:gr64_nosp = COPY [[COPY1]]
+  ; CHECK-NEXT:   JMP64m $noreg, 8, [[COPY2]], @computed_goto.dispatch, $noreg
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.3.bb3 (ir-block-address-taken %ir-block.bb3):
+  ; CHECK-NEXT:   successors: %bb.1(0x20000000), %bb.2(0x20000000), %bb.3(0x20000000), %bb.4(0x20000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   CALL64pcrel32 target-flags(x86-plt) @f3, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+  ; CHECK-NEXT:   [[COPY5:%[0-9]+]]:gr64 = COPY $rax
+  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:gr64_nosp = COPY [[COPY5]]
+  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:gr64_nosp = COPY [[COPY1]]
+  ; CHECK-NEXT:   JMP64m $noreg, 8, [[COPY2]], @computed_goto.dispatch, $noreg
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT: bb.4.bb4 (ir-block-address-taken %ir-block.bb4):
+  ; CHECK-NEXT:   successors: %bb.1(0x20000000), %bb.2(0x20000000), %bb.3(0x20000000), %bb.4(0x20000000)
+  ; CHECK-NEXT: {{  $}}
+  ; CHECK-NEXT:   CALL64pcrel32 target-flags(x86-plt) @f4, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+  ; CHECK-NEXT:   [[COPY6:%[0-9]+]]:gr64 = COPY $rax
+  ; CHECK-NEXT:   [[COPY1:%[0-9]+]]:gr64_nosp = COPY [[COPY6]]
+  ; CHECK-NEXT:   [[COPY2:%[0-9]+]]:gr64_nosp = COPY [[COPY1]]
+  ; CHECK-NEXT:   JMP64m $noreg, 8, [[COPY2]], @computed_goto.dispatch, $noreg
+  bb.0:
+    successors: %bb.5(0x80000000)
+
+    CALL64pcrel32 target-flags(x86-plt) @f0, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+    %0:gr64 = COPY $rax
+    %6:gr64_nosp = COPY %0
+    JMP_1 %bb.5
+
+  bb.1.bb1 (ir-block-address-taken %ir-block.bb1):
+    successors: %bb.5(0x80000000)
+
+    CALL64pcrel32 target-flags(x86-plt) @f1, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+    %1:gr64 = COPY $rax
+    %6:gr64_nosp = COPY %1
+    JMP_1 %bb.5
+
+  bb.2.bb2 (ir-block-address-taken %ir-block.bb2):
+    successors: %bb.5(0x80000000)
+
+    CALL64pcrel32 target-flags(x86-plt) @f2, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+    %2:gr64 = COPY $rax
+    %6:gr64_nosp = COPY %2
+    JMP_1 %bb.5
+
+  bb.3.bb3 (ir-block-address-taken %ir-block.bb3):
+    successors: %bb.5(0x80000000)
+
+    CALL64pcrel32 target-flags(x86-plt) @f3, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+    %3:gr64 = COPY $rax
+    %6:gr64_nosp = COPY %3
+    JMP_1 %bb.5
+
+  bb.4.bb4 (ir-block-address-taken %ir-block.bb4):
+    successors: %bb.5(0x80000000)
+
+    CALL64pcrel32 target-flags(x86-plt) @f4, csr_64, implicit $rsp, implicit $ssp, implicit-def $rsp, implicit-def $ssp, implicit-def $rax
+    %4:gr64 = COPY $rax
+    %6:gr64_nosp = COPY %4
+
+  bb.5:
+    successors: %bb.1(0x20000000), %bb.2(0x20000000), %bb.3(0x20000000), %bb.4(0x20000000)
+
+    %5:gr64_nosp = COPY %6
+    JMP64m $noreg, 8, %5, @computed_goto.dispatch, $noreg
+...
diff --git a/llvm/test/DebugInfo/X86/stringpool.ll b/llvm/test/DebugInfo/X86/stringpool.ll
index 219e672..1c3a57a 100644
--- a/llvm/test/DebugInfo/X86/stringpool.ll
+++ b/llvm/test/DebugInfo/X86/stringpool.ll
@@ -28,8 +28,12 @@ source_filename = "test/DebugInfo/X86/stringpool.ll"
 ; LINUX: .section .debug_str,"MS",@progbits,1
 ; LINUX: yyyy
 
+; DARWIN:      .section __DWARF,__debug_abbrev,regular,debug
+; DARWIN-NEXT: Lsection_abbrev:
+
 ; Verify that we refer to 'yyyy' with a direct offset.
 ; DARWIN: .section        __DWARF,__debug_info,regular,debug
+; DARWIN: Lset1 = Lsection_abbrev-Lsection_abbrev
 ; DARWIN: DW_TAG_variable
 ; DARWIN:             .long   [[YYYY:[0-9]+]]
 ; DARWIN-NEXT:        .long   {{[0-9]+}}              ## DW_AT_type
diff --git a/llvm/test/MC/Disassembler/RISCV/c_slli.txt b/llvm/test/MC/Disassembler/RISCV/c_slli.txt
index 2520455..d8d65ef 100644
--- a/llvm/test/MC/Disassembler/RISCV/c_slli.txt
+++ b/llvm/test/MC/Disassembler/RISCV/c_slli.txt
@@ -14,7 +14,7 @@
 # RUN:     -M no-aliases --show-encoding < %s 2>&1 | \
 # RUN:   FileCheck --check-prefix=NOHINTS %s
 
-# GOOD: c.slli64 zero
+# GOOD: c.slli zero, 0
 # NOHINTS: invalid instruction encoding
 0x02 0x00
 
@@ -302,7 +302,7 @@
 # NOHINTS: invalid instruction encoding
 0x7E 0x10
 
-# GOOD: c.slli64 ra
+# GOOD: c.slli ra, 0
 0x82 0x00
 
 # GOOD: c.slli ra, 1
@@ -526,7 +526,7 @@
 # GOOD64: c.slli ra, 63
 0xFE 0x10
 
-# GOOD: c.slli64 sp
+# GOOD: c.slli sp, 0
 0x02 0x01
 
 # GOOD: c.slli sp, 1
@@ -750,7 +750,7 @@
 # GOOD64: c.slli sp, 63
 0x7E 0x11
 
-# GOOD: c.slli64 gp
+# GOOD: c.slli gp, 0
 0x82 0x01
 
 # GOOD: c.slli gp, 1
@@ -974,7 +974,7 @@
 # GOOD64: c.slli gp, 63
 0xFE 0x11
 
-# GOOD: c.slli64 tp
+# GOOD: c.slli tp, 0
 0x02 0x02
 
 # GOOD: c.slli tp, 1
@@ -1198,7 +1198,7 @@
 # GOOD64: c.slli tp, 63
 0x7E 0x12
 
-# GOOD: c.slli64 t0
+# GOOD: c.slli t0, 0
 0x82 0x02
 
 # GOOD: c.slli t0, 1
@@ -1422,7 +1422,7 @@
 # GOOD64: c.slli t0, 63
 0xFE 0x12
 
-# GOOD: c.slli64 t1
+# GOOD: c.slli t1, 0
 0x02 0x03
 
 # GOOD: c.slli t1, 1
@@ -1646,7 +1646,7 @@
 # GOOD64: c.slli t1, 63
 0x7E 0x13
 
-# GOOD: c.slli64 t2
+# GOOD: c.slli t2, 0
 0x82 0x03
 
 # GOOD: c.slli t2, 1
@@ -1870,7 +1870,7 @@
 # GOOD64: c.slli t2, 63
 0xFE 0x13
 
-# GOOD: c.slli64 s0
+# GOOD: c.slli s0, 0
 0x02 0x04
 
 # GOOD: c.slli s0, 1
@@ -2094,7 +2094,7 @@
 # GOOD64: c.slli s0, 63
 0x7E 0x14
 
-# GOOD: c.slli64 s1
+# GOOD: c.slli s1, 0
 0x82 0x04
 
 # GOOD: c.slli s1, 1
@@ -2318,7 +2318,7 @@
 # GOOD64: c.slli s1, 63
 0xFE 0x14
 
-# GOOD: c.slli64 a0
+# GOOD: c.slli a0, 0
 0x02 0x05
 
 # GOOD: c.slli a0, 1
@@ -2542,7 +2542,7 @@
 # GOOD64: c.slli a0, 63
 0x7E 0x15
 
-# GOOD: c.slli64 a1
+# GOOD: c.slli a1, 0
 0x82 0x05
 
 # GOOD: c.slli a1, 1
@@ -2766,7 +2766,7 @@
 # GOOD64: c.slli a1, 63
 0xFE 0x15
 
-# GOOD: c.slli64 a2
+# GOOD: c.slli a2, 0
 0x02 0x06
 
 # GOOD: c.slli a2, 1
@@ -2990,7 +2990,7 @@
 # GOOD64: c.slli a2, 63
 0x7E 0x16
 
-# GOOD: c.slli64 a3
+# GOOD: c.slli a3, 0
 0x82 0x06
 
 # GOOD: c.slli a3, 1
@@ -3214,7 +3214,7 @@
 # GOOD64: c.slli a3, 63
 0xFE 0x16
 
-# GOOD: c.slli64 a4
+# GOOD: c.slli a4, 0
 0x02 0x07
 
 # GOOD: c.slli a4, 1
@@ -3438,7 +3438,7 @@
 # GOOD64: c.slli a4, 63
 0x7E 0x17
 
-# GOOD: c.slli64 a5
+# GOOD: c.slli a5, 0
 0x82 0x07
 
 # GOOD: c.slli a5, 1
@@ -3662,7 +3662,7 @@
 # GOOD64: c.slli a5, 63
 0xFE 0x17
 
-# GOOD: c.slli64 a6
+# GOOD: c.slli a6, 0
 0x02 0x08
 
 # GOOD: c.slli a6, 1
@@ -3886,7 +3886,7 @@
 # GOOD64: c.slli a6, 63
 0x7E 0x18
 
-# GOOD: c.slli64 a7
+# GOOD: c.slli a7, 0
 0x82 0x08
 
 # GOOD: c.slli a7, 1
@@ -4110,7 +4110,7 @@
 # GOOD64: c.slli a7, 63
 0xFE 0x18
 
-# GOOD: c.slli64 s2
+# GOOD: c.slli s2, 0
 0x02 0x09
 
 # GOOD: c.slli s2, 1
@@ -4334,7 +4334,7 @@
 # GOOD64: c.slli s2, 63
 0x7E 0x19
 
-# GOOD: c.slli64 s3
+# GOOD: c.slli s3, 0
 0x82 0x09
 
 # GOOD: c.slli s3, 1
@@ -4558,7 +4558,7 @@
 # GOOD64: c.slli s3, 63
 0xFE 0x19
 
-# GOOD: c.slli64 s4
+# GOOD: c.slli s4, 0
 0x02 0x0A
 
 # GOOD: c.slli s4, 1
@@ -4782,7 +4782,7 @@
 # GOOD64: c.slli s4, 63
 0x7E 0x1A
 
-# GOOD: c.slli64 s5
+# GOOD: c.slli s5, 0
 0x82 0x0A
 
 # GOOD: c.slli s5, 1
@@ -5006,7 +5006,7 @@
 # GOOD64: c.slli s5, 63
 0xFE 0x1A
 
-# GOOD: c.slli64 s6
+# GOOD: c.slli s6, 0
 0x02 0x0B
 
 # GOOD: c.slli s6, 1
@@ -5230,7 +5230,7 @@
 # GOOD64: c.slli s6, 63
 0x7E 0x1B
 
-# GOOD: c.slli64 s7
+# GOOD: c.slli s7, 0
 0x82 0x0B
 
 # GOOD: c.slli s7, 1
@@ -5454,7 +5454,7 @@
 # GOOD64: c.slli s7, 63
 0xFE 0x1B
 
-# GOOD: c.slli64 s8
+# GOOD: c.slli s8, 0
 0x02 0x0C
 
 # GOOD: c.slli s8, 1
@@ -5678,7 +5678,7 @@
 # GOOD64: c.slli s8, 63
 0x7E 0x1C
 
-# GOOD: c.slli64 s9
+# GOOD: c.slli s9, 0
 0x82 0x0C
 
 # GOOD: c.slli s9, 1
@@ -5902,7 +5902,7 @@
 # GOOD64: c.slli s9, 63
 0xFE 0x1C
 
-# GOOD: c.slli64 s10
+# GOOD: c.slli s10, 0
 0x02 0x0D
 
 # GOOD: c.slli s10, 1
@@ -6126,7 +6126,7 @@
 # GOOD64: c.slli s10, 63
 0x7E 0x1D
 
-# GOOD: c.slli64 s11
+# GOOD: c.slli s11, 0
 0x82 0x0D
 
 # GOOD: c.slli s11, 1
@@ -6350,7 +6350,7 @@
 # GOOD64: c.slli s11, 63
 0xFE 0x1D
 
-# GOOD: c.slli64 t3
+# GOOD: c.slli t3, 0
 0x02 0x0E
 
 # GOOD: c.slli t3, 1
@@ -6574,7 +6574,7 @@
 # GOOD64: c.slli t3, 63
 0x7E 0x1E
 
-# GOOD: c.slli64 t4
+# GOOD: c.slli t4, 0
 0x82 0x0E
 
 # GOOD: c.slli t4, 1
@@ -6798,7 +6798,7 @@
 # GOOD64: c.slli t4, 63
 0xFE 0x1E
 
-# GOOD: c.slli64 t5
+# GOOD: c.slli t5, 0
 0x02 0x0F
 
 # GOOD: c.slli t5, 1
@@ -7022,7 +7022,7 @@
 # GOOD64: c.slli t5, 63
 0x7E 0x1F
 
-# GOOD: c.slli64 t6
+# GOOD: c.slli t6, 0
 0x82 0x0F
 
 # GOOD: c.slli t6, 1
diff --git a/llvm/test/MC/RISCV/Relocations/mc-dump.s b/llvm/test/MC/RISCV/Relocations/mc-dump.s
index e8f4b14..ddc0c7d 100644
--- a/llvm/test/MC/RISCV/Relocations/mc-dump.s
+++ b/llvm/test/MC/RISCV/Relocations/mc-dump.s
@@ -9,10 +9,10 @@
 # CHECK-NEXT:0 Data LinkerRelaxable Size:8 [97,00,00,00,e7,80,00,00]
 # CHECK-NEXT:  Fixup @0 Value:specifier(19,ext) Kind:4023
 # CHECK-NEXT:  Symbol @0 $x
-# CHECK-NEXT:8 Align Size:0+4 []
+# CHECK-NEXT:8 Align LinkerRelaxable Size:0+4 []
 # CHECK-NEXT:  Align:8 Fill:0 FillLen:1 MaxBytesToEmit:8 Nops
 # CHECK-NEXT:  Fixup @0 Value:4 Kind:[[#]]
-# CHECK-NEXT:12 Align Size:4+4 [13,05,30,00]
+# CHECK-NEXT:12 Align LinkerRelaxable Size:4+4 [13,05,30,00]
 # CHECK-NEXT:  Align:8 Fill:0 FillLen:1 MaxBytesToEmit:8 Nops
 # CHECK-NEXT:  Fixup @4 Value:4 Kind:[[#]]
 # CHECK-NEXT:]
diff --git a/llvm/test/MC/RISCV/align.s b/llvm/test/MC/RISCV/align.s
index 1dab1c6..da3b1aa 100644
--- a/llvm/test/MC/RISCV/align.s
+++ b/llvm/test/MC/RISCV/align.s
@@ -154,3 +154,41 @@ data2:
 2:
 	bnez t1, 1b
 	bnez t1, 2b
+
+## .text3 with a call at the start
+# NORELAX-RELOC:    .rela.text3a
+# C-OR-ZCA-EXT-NORELAX-RELOC: .rela.text3a
+# RELAX-RELOC:      .rela.text3a {
+# RELAX-RELOC-NEXT:    0x0  R_RISCV_CALL_PLT foo 0x0
+# RELAX-RELOC-NEXT:    0x0  R_RISCV_RELAX - 0x0
+# RELAX-RELOC-NEXT:    0xC  R_RISCV_BRANCH .Ltmp[[#]] 0x0
+# RELAX-RELOC-NEXT:    0x10 R_RISCV_ALIGN - 0x4
+# RELAX-RELOC-NEXT:    0x14 R_RISCV_BRANCH .Ltmp[[#]] 0x0
+# RELAX-RELOC-NEXT: }
+.section .text3a, "ax"
+call foo
+bnez t1, 1f
+bnez t2, 2f
+1:
+.p2align 3
+2:
+bnez t1, 1b
+bnez t1, 2b
+
+## .text3 with a call at the end
+# RELAX-RELOC:      .rela.text3b {
+# RELAX-RELOC-NEXT:    0x4  R_RISCV_BRANCH .Ltmp[[#]] 0x0
+# RELAX-RELOC-NEXT:    0x8  R_RISCV_ALIGN - 0x4
+# RELAX-RELOC-NEXT:    0xC  R_RISCV_BRANCH .Ltmp[[#]] 0x0
+# RELAX-RELOC-NEXT:    0x14 R_RISCV_CALL_PLT foo 0x0
+# RELAX-RELOC-NEXT:    0x14 R_RISCV_RELAX - 0x0
+# RELAX-RELOC-NEXT: }
+.section .text3b, "ax"
+bnez t1, 1f
+bnez t2, 2f
+1:
+.p2align 3
+2:
+bnez t1, 1b
+bnez t1, 2b
+call foo
diff --git a/llvm/test/MC/RISCV/rv32c-invalid.s b/llvm/test/MC/RISCV/rv32c-invalid.s
index 413573a..5214747 100644
--- a/llvm/test/MC/RISCV/rv32c-invalid.s
+++ b/llvm/test/MC/RISCV/rv32c-invalid.s
@@ -26,8 +26,8 @@ c.jalr  zero # CHECK: :[[@LINE]]:9: error: register must be a GPR excluding zero
 c.mv  ra, x0 # CHECK: :[[@LINE]]:11: error: register must be a GPR excluding zero (x0)
 c.add  ra, ra, x0 # CHECK: :[[@LINE]]:16: error: invalid operand for instruction
 
-## GPRNoX0X2
-c.lui x2, 4 # CHECK: :[[@LINE]]:7: error: register must be a GPR excluding zero (x0) and sp (x2){{$}}
+## GPRNoX2
+c.lui x2, 4 # CHECK: :[[@LINE]]:7: error: register must be a GPR excluding sp (x2){{$}}
 
 ## SP
 c.addi4spn  a0, a0, 12 # CHECK: :[[@LINE]]:17: error: register must be sp (x2)
@@ -35,10 +35,9 @@ c.addi16sp  t0, 16 # CHECK: :[[@LINE]]:13: error: register must be sp (x2)
 
 # Out of range immediates
 
-## uimmlog2xlennonzero
-c.slli t0, 64 # CHECK: :[[@LINE]]:12: error: immediate must be an integer in the range [1, 31]
-c.srli a0, 32 # CHECK: :[[@LINE]]:12: error: immediate must be an integer in the range [1, 31]
-c.srai a0, 0  # CHECK: :[[@LINE]]:12: error: immediate must be an integer in the range [1, 31]
+## uimmlog2xlenn
+c.slli t0, 64 # CHECK: :[[@LINE]]:12: error: immediate must be an integer in the range [0, 31]
+c.srli a0, 32 # CHECK: :[[@LINE]]:12: error: immediate must be an integer in the range [0, 31]
 
 ## simm6
 c.li t0, 128 # CHECK: :[[@LINE]]:10: error: immediate must be an integer in the range [-32, 31]
@@ -49,13 +48,14 @@ c.andi a0, -33 # CHECK: :[[@LINE]]:12: error: immediate must be an integer in th
 c.andi a0, foo # CHECK: :[[@LINE]]:12: error: immediate must be an integer in the range [-32, 31]
 c.andi a0, %lo(foo) # CHECK: :[[@LINE]]:12: error: immediate must be an integer in the range [-32, 31]
 c.andi a0, %hi(foo) # CHECK: :[[@LINE]]:12: error: immediate must be an integer in the range [-32, 31]
+c.addi t0, -33 # CHECK: :[[@LINE]]:12: error: immediate must be an integer in the range [-32, 31]
+c.addi t0, 32 # CHECK: :[[@LINE]]:12: error: immediate must be an integer in the range [-32, 31]
+c.addi t0, foo # CHECK: :[[@LINE]]:12: error: immediate must be an integer in the range [-32, 31]
+c.addi t0, %lo(foo) # CHECK: :[[@LINE]]:12: error: immediate must be an integer in the range [-32, 31]
+c.addi t0, %hi(foo) # CHECK: :[[@LINE]]:12: error: immediate must be an integer in the range [-32, 31]
 
 ## simm6nonzero
-c.addi t0, -33 # CHECK: :[[@LINE]]:12: error: immediate must be non-zero in the range [-32, 31]
-c.addi t0, 32 # CHECK: :[[@LINE]]:12: error: immediate must be non-zero in the range [-32, 31]
-c.addi t0, foo # CHECK: :[[@LINE]]:12: error: immediate must be non-zero in the range [-32, 31]
-c.addi t0, %lo(foo) # CHECK: :[[@LINE]]:12: error: immediate must be non-zero in the range [-32, 31]
-c.addi t0, %hi(foo) # CHECK: :[[@LINE]]:12: error: immediate must be non-zero in the range [-32, 31]
+c.nop 32 # CHECK: :[[@LINE]]:7: error: immediate must be non-zero in the range [-32, 31]
 
 ## c_lui_imm
 c.lui t0, 0 # CHECK: :[[@LINE]]:11: error: immediate must be in [0xfffe0, 0xfffff] or [1, 31]
diff --git a/llvm/test/MC/RISCV/rv64c-invalid.s b/llvm/test/MC/RISCV/rv64c-invalid.s
index 9b0a324..404e02cf 100644
--- a/llvm/test/MC/RISCV/rv64c-invalid.s
+++ b/llvm/test/MC/RISCV/rv64c-invalid.s
@@ -13,10 +13,9 @@ c.ldsp  zero, 4(sp) # CHECK: :[[@LINE]]:9: error: register must be a GPR excludi
 
 # Out of range immediates
 
-## uimmlog2xlennonzero
-c.slli t0, 64 # CHECK: :[[@LINE]]:12: error: immediate must be an integer in the range [1, 63]
-c.srli a0, -1 # CHECK: :[[@LINE]]:12: error: immediate must be an integer in the range [1, 63]
-c.srai a0, 0  # CHECK: :[[@LINE]]:12: error: immediate must be an integer in the range [1, 63]
+## uimmlog2xlen
+c.slli t0, 64 # CHECK: :[[@LINE]]:12: error: immediate must be an integer in the range [0, 63]
+c.srli a0, -1 # CHECK: :[[@LINE]]:12: error: immediate must be an integer in the range [0, 63]
 
 ## simm6
 c.addiw t0, -33 # CHECK: :[[@LINE]]:13: error: immediate must be an integer in the range [-32, 31]
diff --git a/llvm/test/MC/RISCV/rvc-hints-invalid.s b/llvm/test/MC/RISCV/rvc-hints-invalid.s
index 2a7a6ad..540d2c6 100644
--- a/llvm/test/MC/RISCV/rvc-hints-invalid.s
+++ b/llvm/test/MC/RISCV/rvc-hints-invalid.s
@@ -1,11 +1,11 @@
 # RUN: not llvm-mc -triple=riscv32 -mattr=+c < %s 2>&1 \
 # RUN:     | FileCheck -check-prefixes=CHECK,CHECK-RV32 %s
 # RUN: not llvm-mc -triple=riscv64 -mattr=+c < %s 2>&1 \
-# RUN:     | FileCheck -check-prefixes=CHECK,CHECK-RV64 %s
+# RUN:     | FileCheck -check-prefixes=CHECK %s
 
 c.nop 0 # CHECK: :[[@LINE]]:7: error: immediate must be non-zero in the range [-32, 31]
 
-c.addi x0, 33 # CHECK: :[[@LINE]]:12: error: immediate must be non-zero in the range [-32, 31]
+c.addi x0, 33 # CHECK: :[[@LINE]]:12: error: immediate must be an integer in the range [-32, 31]
 
 c.li x0, 42 # CHECK: :[[@LINE]]:10: error: immediate must be an integer in the range [-32, 31]
 
@@ -15,10 +15,7 @@ c.mv x0, x0 # CHECK: :[[@LINE]]:10: error: register must be a GPR excluding zero
 
 c.add x0, x0 # CHECK: :[[@LINE]]:11: error: register must be a GPR excluding zero (x0)
 
-c.slli x0, 0 # CHECK-RV32: :[[@LINE]]:12: error: immediate must be an integer in the range [1, 31]
-c.slli x0, 32 # CHECK-RV32: :[[@LINE]]:12: error: immediate must be an integer in the range [1, 31]
-
-c.slli x0, 0 # CHECK-RV64: :[[@LINE]]:12: error: immediate must be an integer in the range [1, 63]
+c.slli x0, 32 # CHECK-RV32: :[[@LINE]]:12: error: immediate must be an integer in the range [0, 31]
 
 c.srli64 x30 # CHECK: :[[@LINE]]:10: error: invalid operand for instruction
 
diff --git a/llvm/test/MC/RISCV/rvc-hints-valid.s b/llvm/test/MC/RISCV/rvc-hints-valid.s
index 5dc86d1..7fba864 100644
--- a/llvm/test/MC/RISCV/rvc-hints-valid.s
+++ b/llvm/test/MC/RISCV/rvc-hints-valid.s
@@ -48,18 +48,34 @@ c.add x0, a0
 # CHECK-ASM: encoding: [0x06,0x00]
 c.slli x0, 1
 
-# CHECK-ASM-AND-OBJ: c.slli64 zero
+# CHECK-ASM-AND-OBJ: c.slli zero, 0
 # CHECK-ASM: encoding: [0x02,0x00]
 c.slli64 x0
 
-# CHECK-ASM-AND-OBJ: c.slli64 a0
+# CHECK-ASM-AND-OBJ: c.slli zero, 0
+# CHECK-ASM: encoding: [0x02,0x00]
+c.slli x0, 0
+
+# CHECK-ASM-AND-OBJ: c.slli a0, 0
 # CHECK-ASM: encoding: [0x02,0x05]
 c.slli64 a0
 
-# CHECK-ASM-AND-OBJ: c.srli64 a1
+# CHECK-ASM-AND-OBJ: c.slli a0, 0
+# CHECK-ASM: encoding: [0x02,0x05]
+c.slli a0, 0
+
+# CHECK-ASM-AND-OBJ: c.srli a1, 0
 # CHECK-ASM: encoding: [0x81,0x81]
 c.srli64 a1
 
-# CHECK-ASM-AND-OBJ: c.srai64 a0
+# CHECK-ASM-AND-OBJ: c.srli a1, 0
+# CHECK-ASM: encoding: [0x81,0x81]
+c.srli a1, 0
+
+# CHECK-ASM-AND-OBJ: c.srai a0, 0
 # CHECK-ASM: encoding: [0x01,0x85]
 c.srai64 a0
+
+# CHECK-ASM-AND-OBJ: c.srai a0, 0
+# CHECK-ASM: encoding: [0x01,0x85]
+c.srai a0, 0
diff --git a/llvm/test/MC/RISCV/rve-invalid.s b/llvm/test/MC/RISCV/rve-invalid.s
index 0b1e896..92e0239a 100644
--- a/llvm/test/MC/RISCV/rve-invalid.s
+++ b/llvm/test/MC/RISCV/rve-invalid.s
@@ -115,7 +115,7 @@ auipc t6, 32
 # CHECK: :[[@LINE+1]]:8: error: register must be a GPR excluding zero (x0)
 c.addi x31, 0
 # CHECK-DIS: 9846 <unknown>
-# CHECK: :[[@LINE+1]]:7: error: register must be a GPR excluding zero (x0)
+# CHECK: :[[@LINE+1]]:7: error: invalid operand for instruction
 c.add x16, x17
 # CHECK-DIS: 8046 <unknown>
 # CHECK: :[[@LINE+1]]:10: error: register must be a GPR excluding zero (x0)
diff --git a/llvm/test/MC/WebAssembly/reference-types.s b/llvm/test/MC/WebAssembly/reference-types.s
index 08aafb2..7a838fc 100644
--- a/llvm/test/MC/WebAssembly/reference-types.s
+++ b/llvm/test/MC/WebAssembly/reference-types.s
@@ -1,5 +1,5 @@
-# RUN: llvm-mc -show-encoding -triple=wasm32-unknown-unknown -mattr=+reference-types < %s | FileCheck %s
-# RUN: llvm-mc -show-encoding -triple=wasm64-unknown-unknown -mattr=+reference-types < %s | FileCheck %s
+# RUN: llvm-mc -show-encoding -triple=wasm32-unknown-unknown -mattr=+reference-types -mattr=+gc < %s | FileCheck %s
+# RUN: llvm-mc -show-encoding -triple=wasm64-unknown-unknown -mattr=+reference-types -mattr=+gc < %s | FileCheck %s
 
 # CHECK-LABEL:ref_is_null:
 # CHECK: ref.is_null     # encoding: [0xd1]
diff --git a/llvm/test/TableGen/CompressInstEmitter/suboperands.td b/llvm/test/TableGen/CompressInstEmitter/suboperands.td
index d83cc04..cd724e9 100644
--- a/llvm/test/TableGen/CompressInstEmitter/suboperands.td
+++ b/llvm/test/TableGen/CompressInstEmitter/suboperands.td
@@ -161,15 +161,15 @@ def : CompressPat<(BigInst3 RegsC:$dst, RegsC:$src, simm6:$imm),
 // CHECK-NEXT: } // if
 
 // CHECK-LABEL: ArchValidateMCOperandForUncompress
-// CHECK: // simm12
-// CHECK: return isInt<12>(Imm);
+// CHECK: // simm6
+// CHECK: return isInt<6>(Imm);
 
 // CHECK-LABEL: uncompressInst
 // CHECK: case Arch::SmallInst:
 // CHECK-NEXT:  if (MI.getOperand(0).isReg() &&
-// CHECK-NEXT: ArchMCRegisterClasses[Arch::RegsRegClassID].contains(MI.getOperand(0).getReg()) &&
+// CHECK-NEXT: ArchMCRegisterClasses[Arch::RegsCRegClassID].contains(MI.getOperand(0).getReg()) &&
 // CHECK-NEXT: MI.getOperand(1).isReg() &&
-// CHECK-NEXT: ArchMCRegisterClasses[Arch::RegsRegClassID].contains(MI.getOperand(1).getReg()) &&
+// CHECK-NEXT: ArchMCRegisterClasses[Arch::RegsCRegClassID].contains(MI.getOperand(1).getReg()) &&
 // CHECK-NEXT: ArchValidateMCOperandForUncompress(MI.getOperand(2), STI, 1)) {
 // CHECK-NEXT: // big $dst, $addr
 // CHECK-NEXT: OutInst.setOpcode(Arch::BigInst);
@@ -183,9 +183,9 @@ def : CompressPat<(BigInst3 RegsC:$dst, RegsC:$src, simm6:$imm),
 // CHECK-NEXT: } // if
 // CHECK: case Arch::SmallInst2:
 // CHECK-NEXT:  if (MI.getOperand(0).isReg() &&
-// CHECK-NEXT: ArchMCRegisterClasses[Arch::RegsRegClassID].contains(MI.getOperand(0).getReg()) &&
+// CHECK-NEXT: ArchMCRegisterClasses[Arch::RegsCRegClassID].contains(MI.getOperand(0).getReg()) &&
 // CHECK-NEXT: MI.getOperand(1).isReg() &&
-// CHECK-NEXT: ArchMCRegisterClasses[Arch::RegsRegClassID].contains(MI.getOperand(1).getReg()) &&
+// CHECK-NEXT: ArchMCRegisterClasses[Arch::RegsCRegClassID].contains(MI.getOperand(1).getReg()) &&
 // CHECK-NEXT: ArchValidateMCOperandForUncompress(MI.getOperand(2), STI, 1)) {
 // CHECK-NEXT: // big $dst, $addr
 // CHECK-NEXT: OutInst.setOpcode(Arch::BigInst2);
@@ -199,9 +199,9 @@ def : CompressPat<(BigInst3 RegsC:$dst, RegsC:$src, simm6:$imm),
 // CHECK-NEXT: } // if
 // CHECK: case Arch::SmallInst3:
 // CHECK-NEXT:  if (MI.getOperand(0).isReg() &&
-// CHECK-NEXT: ArchMCRegisterClasses[Arch::RegsRegClassID].contains(MI.getOperand(0).getReg()) &&
+// CHECK-NEXT: ArchMCRegisterClasses[Arch::RegsCRegClassID].contains(MI.getOperand(0).getReg()) &&
 // CHECK-NEXT: MI.getOperand(1).isReg() &&
-// CHECK-NEXT: ArchMCRegisterClasses[Arch::RegsRegClassID].contains(MI.getOperand(1).getReg()) &&
+// CHECK-NEXT: ArchMCRegisterClasses[Arch::RegsCRegClassID].contains(MI.getOperand(1).getReg()) &&
 // CHECK-NEXT: ArchValidateMCOperandForUncompress(MI.getOperand(2), STI, 1)) {
 // CHECK-NEXT: // big $dst, $src, $imm
 // CHECK-NEXT: OutInst.setOpcode(Arch::BigInst3);
diff --git a/llvm/test/TableGen/RuntimeLibcallEmitter.td b/llvm/test/TableGen/RuntimeLibcallEmitter.td
index 579e3c7..783a861 100644
--- a/llvm/test/TableGen/RuntimeLibcallEmitter.td
+++ b/llvm/test/TableGen/RuntimeLibcallEmitter.td
@@ -95,8 +95,8 @@ def BlahLibrary : SystemRuntimeLibrary<isBlahArch, (add calloc, LibraryWithCondi
 // CHECK-NEXT:  __lshrdi3 = 4, // __lshrdi3
 // CHECK-NEXT:  bzero = 5, // bzero
 // CHECK-NEXT:  calloc = 6, // calloc
-// CHECK-NEXT:  sqrtl_f80 = 7, // sqrtl
-// CHECK-NEXT:  sqrtl_f128 = 8, // sqrtl
+// CHECK-NEXT:  sqrtl_f128 = 7, // sqrtl
+// CHECK-NEXT:  sqrtl_f80 = 8, // sqrtl
 // CHECK-NEXT:  NumLibcallImpls = 9
 // CHECK-NEXT: };
 // CHECK-NEXT: } // End namespace RTLIB
@@ -157,8 +157,8 @@ def BlahLibrary : SystemRuntimeLibrary<isBlahArch, (add calloc, LibraryWithCondi
 // CHECK-NEXT: RTLIB::SRL_I64, // RTLIB::__lshrdi3
 // CHECK-NEXT: RTLIB::BZERO, // RTLIB::bzero
 // CHECK-NEXT: RTLIB::CALLOC, // RTLIB::calloc
-// CHECK-NEXT: RTLIB::SQRT_F80, // RTLIB::sqrtl_f80
 // CHECK-NEXT: RTLIB::SQRT_F128, // RTLIB::sqrtl_f128
+// CHECK-NEXT: RTLIB::SQRT_F80, // RTLIB::sqrtl_f80
 // CHECK-NEXT: };
 
 
diff --git a/llvm/test/ThinLTO/X86/memprof-basic.ll b/llvm/test/ThinLTO/X86/memprof-basic.ll
index c5eedb2..757f6e9 100644
--- a/llvm/test/ThinLTO/X86/memprof-basic.ll
+++ b/llvm/test/ThinLTO/X86/memprof-basic.ll
@@ -52,6 +52,7 @@
 ; RUN:	cat %t.ccg.postbuild.dot | FileCheck %s --check-prefix=DOT
 ;; We should have cloned bar, baz, and foo, for the cold memory allocation.
 ; RUN:	cat %t.ccg.cloned.dot | FileCheck %s --check-prefix=DOTCLONED
+; RUN:	cat %t.ccg.clonefuncassign.dot | FileCheck %s --check-prefix=DOTFUNCASSIGN
 
 ; RUN: llvm-dis %t.out.1.4.opt.bc -o - | FileCheck %s --check-prefix=IR
 
@@ -370,6 +371,10 @@ attributes #0 = { noinline optnone }
 ; DOTCLONED: 	Node[[BAR2]] [shape=record,tooltip="N[[BAR2]] ContextIds: 2",fillcolor="cyan",color="blue",style="filled,bold,dashed",label="{OrigId: Alloc0\n_Z3barv -\> alloc}"];
 ; DOTCLONED: }
 
+;; Here we are just ensuring that the post-function assign dot graph includes
+;; clone information for both the caller and callee in the node labels.
+; DOTFUNCASSIGN: _Z3bazv.memprof.1 -\> _Z3barv.memprof.1
+; DOTFUNCASSIGN: _Z3barv.memprof.1 -\> alloc
 
 ; DISTRIB: ^[[BAZ:[0-9]+]] = gv: (guid: 1807954217441101578, {{.*}} callsites: ((callee: ^[[BAR:[0-9]+]], clones: (0, 1)
 ; DISTRIB: ^[[FOO:[0-9]+]] = gv: (guid: 8107868197919466657, {{.*}} callsites: ((callee: ^[[BAZ]], clones: (0, 1)
diff --git a/llvm/test/Transforms/CodeGenPrepare/X86/fold-loop-of-urem.ll b/llvm/test/Transforms/CodeGenPrepare/X86/fold-loop-of-urem.ll
index 7abc32e..f53127f 100644
--- a/llvm/test/Transforms/CodeGenPrepare/X86/fold-loop-of-urem.ll
+++ b/llvm/test/Transforms/CodeGenPrepare/X86/fold-loop-of-urem.ll
@@ -1065,3 +1065,37 @@ for.body:
   %exitcond.not = icmp eq i32 %inc, %N
   br i1 %exitcond.not, label %for.cond.cleanup, label %for.body
 }
+
+define i64 @pr150611_add_offset_is_not_loop_invariant(i1 %cond) {
+; CHECK-LABEL: define i64 @pr150611_add_offset_is_not_loop_invariant(
+; CHECK-SAME: i1 [[COND:%.*]]) {
+; CHECK-NEXT:  [[ENTRY:.*]]:
+; CHECK-NEXT:    [[REMAMT:%.*]] = select i1 [[COND]], i64 2, i64 0
+; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
+; CHECK:       [[FOR_BODY]]:
+; CHECK-NEXT:    [[INDVARS:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[INDVARS_NEXT:%.*]], %[[FOR_BODY]] ]
+; CHECK-NEXT:    [[ADD_OFFSET:%.*]] = zext i1 [[COND]] to i64
+; CHECK-NEXT:    [[ADD:%.*]] = add nuw i64 [[INDVARS]], [[ADD_OFFSET]]
+; CHECK-NEXT:    [[REM:%.*]] = urem i64 [[ADD]], [[REMAMT]]
+; CHECK-NEXT:    [[INDVARS_NEXT]] = add nuw i64 [[INDVARS]], 1
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_NEXT]], 3
+; CHECK-NEXT:    br i1 [[EXITCOND]], label %[[FOR_EXIT:.*]], label %[[FOR_BODY]]
+; CHECK:       [[FOR_EXIT]]:
+; CHECK-NEXT:    ret i64 [[REM]]
+;
+entry:
+  %remamt = select i1 %cond, i64 2, i64 0
+  br label %for.body
+
+for.body:
+  %indvars = phi i64 [ 0, %entry ], [ %indvars.next, %for.body ]
+  %add.offset = zext i1 %cond to i64
+  %add = add nuw i64 %indvars, %add.offset
+  %rem = urem i64 %add, %remamt
+  %indvars.next = add nuw i64 %indvars, 1
+  %exitcond = icmp eq i64 %indvars.next, 3
+  br i1 %exitcond, label %for.exit, label %for.body
+
+for.exit:
+  ret i64 %rem
+}
diff --git a/llvm/test/Transforms/InstCombine/trunc-inseltpoison.ll b/llvm/test/Transforms/InstCombine/trunc-inseltpoison.ll
index 33fa2c3..f83352c 100644
--- a/llvm/test/Transforms/InstCombine/trunc-inseltpoison.ll
+++ b/llvm/test/Transforms/InstCombine/trunc-inseltpoison.ll
@@ -959,8 +959,8 @@ define <3 x i31> @wide_splat3(<3 x i33> %x) {
 
 define <8 x i8> @wide_lengthening_splat(<4 x i16> %v) {
 ; CHECK-LABEL: @wide_lengthening_splat(
-; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i16> [[V:%.*]], <4 x i16> poison, <8 x i32> zeroinitializer
-; CHECK-NEXT:    [[TR:%.*]] = trunc <8 x i16> [[SHUF]] to <8 x i8>
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc <4 x i16> [[V:%.*]] to <4 x i8>
+; CHECK-NEXT:    [[TR:%.*]] = shufflevector <4 x i8> [[TMP1]], <4 x i8> poison, <8 x i32> zeroinitializer
 ; CHECK-NEXT:    ret <8 x i8> [[TR]]
 ;
   %shuf = shufflevector <4 x i16> %v, <4 x i16> %v, <8 x i32> zeroinitializer
diff --git a/llvm/test/Transforms/InstCombine/trunc.ll b/llvm/test/Transforms/InstCombine/trunc.ll
index a85ce71..dfe9d94 100644
--- a/llvm/test/Transforms/InstCombine/trunc.ll
+++ b/llvm/test/Transforms/InstCombine/trunc.ll
@@ -960,8 +960,8 @@ define <3 x i31> @wide_splat3(<3 x i33> %x) {
 
 define <8 x i8> @wide_lengthening_splat(<4 x i16> %v) {
 ; CHECK-LABEL: @wide_lengthening_splat(
-; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <4 x i16> [[V:%.*]], <4 x i16> poison, <8 x i32> zeroinitializer
-; CHECK-NEXT:    [[TR:%.*]] = trunc <8 x i16> [[SHUF]] to <8 x i8>
+; CHECK-NEXT:    [[TMP1:%.*]] = trunc <4 x i16> [[V:%.*]] to <4 x i8>
+; CHECK-NEXT:    [[TR:%.*]] = shufflevector <4 x i8> [[TMP1]], <4 x i8> poison, <8 x i32> zeroinitializer
 ; CHECK-NEXT:    ret <8 x i8> [[TR]]
 ;
   %shuf = shufflevector <4 x i16> %v, <4 x i16> %v, <8 x i32> zeroinitializer
@@ -969,6 +969,19 @@ define <8 x i8> @wide_lengthening_splat(<4 x i16> %v) {
   ret <8 x i8> %tr
 }
 
+; This is a negative test, we expect the trunc to remain after the shuffle as it
+; might not be beneficial to preform trunc on a wider type
+define <4 x i8> @wide_shortening_splat(<8 x i16> %v) {
+; CHECK-LABEL: @wide_shortening_splat(
+; CHECK-NEXT:    [[SHUF:%.*]] = shufflevector <8 x i16> [[V:%.*]], <8 x i16> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[TR:%.*]] = trunc <4 x i16> [[SHUF]] to <4 x i8>
+; CHECK-NEXT:    ret <4 x i8> [[TR]]
+;
+  %shuf = shufflevector <8 x i16> %v, <8 x i16> %v, <4 x i32> zeroinitializer
+  %tr = trunc <4 x i16> %shuf to <4 x i8>
+  ret <4 x i8> %tr
+}
+
 define <2 x i8> @narrow_add_vec_constant(<2 x i32> %x) {
 ; CHECK-LABEL: @narrow_add_vec_constant(
 ; CHECK-NEXT:    [[TMP1:%.*]] = trunc <2 x i32> [[X:%.*]] to <2 x i8>
diff --git a/llvm/test/Transforms/InstSimplify/ConstProp/gep.ll b/llvm/test/Transforms/InstSimplify/ConstProp/gep.ll
index 7632579..5b24906 100644
--- a/llvm/test/Transforms/InstSimplify/ConstProp/gep.ll
+++ b/llvm/test/Transforms/InstSimplify/ConstProp/gep.ll
@@ -29,3 +29,31 @@ define ptr @f2() {
 ;
   ret ptr getelementptr (ptr, ptr getelementptr inbounds inrange(0, 8) ([3 x ptr], ptr @vt, i64 0, i64 1), i64 3)
 }
+
+define ptr @f3() {
+; CHECK-LABEL: @f3(
+; CHECK-NEXT:    ret ptr getelementptr inbounds nuw inrange(-8, 0) (i8, ptr @vt, i64 16)
+;
+  ret ptr getelementptr inrange(-16, 8) (ptr, ptr getelementptr inbounds inrange(0, 8) ([3 x ptr], ptr @vt, i64 0, i64 1), i64 1)
+}
+
+define ptr @f4() {
+; CHECK-LABEL: @f4(
+; CHECK-NEXT:    ret ptr getelementptr inbounds nuw inrange(-8, 8) (i8, ptr @vt, i64 16)
+;
+  ret ptr getelementptr inrange(-16, 8) (ptr, ptr getelementptr inbounds inrange(0, 24) ([3 x ptr], ptr @vt, i64 0, i64 1), i64 1)
+}
+
+define ptr @f5() {
+; CHECK-LABEL: @f5(
+; CHECK-NEXT:    ret ptr getelementptr inbounds nuw inrange(0, 0) (i8, ptr @vt, i64 16)
+;
+  ret ptr getelementptr inrange(0, 8) (ptr, ptr getelementptr inbounds inrange(0, 8) ([3 x ptr], ptr @vt, i64 0, i64 0), i64 2)
+}
+
+define ptr @f6() {
+; CHECK-LABEL: @f6(
+; CHECK-NEXT:    ret ptr getelementptr inbounds nuw inrange(-8, 8) (i8, ptr @vt, i64 16)
+;
+  ret ptr getelementptr inrange(-8, 8) (ptr, ptr getelementptr inbounds inrange(-8, 16) ([3 x ptr], ptr @vt, i64 0, i64 1), i64 1)
+}
diff --git a/llvm/test/Transforms/LoopFusion/sunk-phi-nodes.ll b/llvm/test/Transforms/LoopFusion/sunk-phi-nodes.ll
new file mode 100644
index 0000000..36c6bdd
--- /dev/null
+++ b/llvm/test/Transforms/LoopFusion/sunk-phi-nodes.ll
@@ -0,0 +1,65 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -passes=loop-fusion -S < %s 2>&1 | FileCheck %s
+define dso_local i32 @check_sunk_phi_nodes() {
+; CHECK-LABEL: define dso_local i32 @check_sunk_phi_nodes() {
+; CHECK-NEXT:  [[ENTRY:.*]]:
+; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
+; CHECK:       [[FOR_BODY]]:
+; CHECK-NEXT:    [[SUM1_02:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[ADD:%.*]], %[[FOR_INC6:.*]] ]
+; CHECK-NEXT:    [[I_01:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[INC:%.*]], %[[FOR_INC6]] ]
+; CHECK-NEXT:    [[I1_04:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[INC7:%.*]], %[[FOR_INC6]] ]
+; CHECK-NEXT:    [[SUM2_03:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[ADD5:%.*]], %[[FOR_INC6]] ]
+; CHECK-NEXT:    [[ADD]] = add nsw i32 [[SUM1_02]], [[I_01]]
+; CHECK-NEXT:    br label %[[FOR_INC:.*]]
+; CHECK:       [[FOR_INC]]:
+; CHECK-NEXT:    [[MUL:%.*]] = mul nsw i32 [[I1_04]], [[I1_04]]
+; CHECK-NEXT:    [[ADD5]] = add nsw i32 [[SUM2_03]], [[MUL]]
+; CHECK-NEXT:    br label %[[FOR_INC6]]
+; CHECK:       [[FOR_INC6]]:
+; CHECK-NEXT:    [[INC]] = add nsw i32 [[I_01]], 1
+; CHECK-NEXT:    [[CMP:%.*]] = icmp slt i32 [[INC]], 10
+; CHECK-NEXT:    [[INC7]] = add nsw i32 [[I1_04]], 1
+; CHECK-NEXT:    [[CMP3:%.*]] = icmp slt i32 [[INC7]], 10
+; CHECK-NEXT:    br i1 [[CMP3]], label %[[FOR_BODY]], label %[[FOR_END8:.*]]
+; CHECK:       [[FOR_END8]]:
+; CHECK-NEXT:    [[SUM2_0_LCSSA:%.*]] = phi i32 [ [[ADD5]], %[[FOR_INC6]] ]
+; CHECK-NEXT:    [[SUM1_0_LCSSA:%.*]] = phi i32 [ [[ADD]], %[[FOR_INC6]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = add i32 [[SUM1_0_LCSSA]], [[SUM2_0_LCSSA]]
+; CHECK-NEXT:    ret i32 [[TMP0]]
+;
+entry:
+  br label %for.body
+
+for.body:                                         ; preds = %entry, %for.inc
+  %sum1.02 = phi i32 [ 0, %entry ], [ %add, %for.inc ]
+  %i.01 = phi i32 [ 0, %entry ], [ %inc, %for.inc ]
+  %add = add nsw i32 %sum1.02, %i.01
+  br label %for.inc
+
+for.inc:                                          ; preds = %for.body
+  %inc = add nsw i32 %i.01, 1
+  %cmp = icmp slt i32 %inc, 10
+  br i1 %cmp, label %for.body, label %for.end
+
+for.end:                                          ; preds = %for.inc
+  %sum1.0.lcssa = phi i32 [ %add, %for.inc ]
+  br label %for.body4
+
+for.body4:                                        ; preds = %for.end, %for.inc6
+  %i1.04 = phi i32 [ 0, %for.end ], [ %inc7, %for.inc6 ]
+  %sum2.03 = phi i32 [ 0, %for.end ], [ %add5, %for.inc6 ]
+  %mul = mul nsw i32 %i1.04, %i1.04
+  %add5 = add nsw i32 %sum2.03, %mul
+  br label %for.inc6
+
+for.inc6:                                         ; preds = %for.body4
+  %inc7 = add nsw i32 %i1.04, 1
+  %cmp3 = icmp slt i32 %inc7, 10
+  br i1 %cmp3, label %for.body4, label %for.end8
+
+for.end8:                                         ; preds = %for.inc6
+  %sum2.0.lcssa = phi i32 [ %add5, %for.inc6 ]
+  %0 = add i32 %sum1.0.lcssa, %sum2.0.lcssa
+  ret i32 %0
+}
+
diff --git a/llvm/test/Transforms/LoopIdiom/reuse-lcssa-phi-scev-expansion.ll b/llvm/test/Transforms/LoopIdiom/reuse-lcssa-phi-scev-expansion.ll
new file mode 100644
index 0000000..357a7b6
--- /dev/null
+++ b/llvm/test/Transforms/LoopIdiom/reuse-lcssa-phi-scev-expansion.ll
@@ -0,0 +1,98 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -p loop-idiom -S %s | FileCheck %s
+
+declare void @foo()
+declare void @bar()
+
+define void @scev_expand_ptrtoint(i8 %x, ptr %start) {
+; CHECK-LABEL: define void @scev_expand_ptrtoint(
+; CHECK-SAME: i8 [[X:%.*]], ptr [[START:%.*]]) {
+; CHECK-NEXT:  [[ENTRY:.*]]:
+; CHECK-NEXT:    [[START1:%.*]] = ptrtoint ptr [[START]] to i64
+; CHECK-NEXT:    br label %[[LOOP_1_HEADER:.*]]
+; CHECK:       [[LOOP_1_HEADER]]:
+; CHECK-NEXT:    [[PTR_IV_1:%.*]] = phi ptr [ [[START]], %[[ENTRY]] ], [ [[PTR_IV_1_NEXT:%.*]], %[[LOOP_1_LATCH:.*]] ]
+; CHECK-NEXT:    [[C:%.*]] = icmp ule i8 [[X]], 1
+; CHECK-NEXT:    br i1 [[C]], label %[[LOOP_1_LATCH]], label %[[MIDDLE:.*]]
+; CHECK:       [[LOOP_1_LATCH]]:
+; CHECK-NEXT:    [[PTR_IV_1_NEXT]] = getelementptr i8, ptr [[PTR_IV_1]], i64 1
+; CHECK-NEXT:    br label %[[LOOP_1_HEADER]]
+; CHECK:       [[MIDDLE]]:
+; CHECK-NEXT:    [[PTR_IV_1_LCSSA:%.*]] = phi ptr [ [[PTR_IV_1]], %[[LOOP_1_HEADER]] ]
+; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i8 [[X]], 0
+; CHECK-NEXT:    [[CMP_EXT:%.*]] = zext i1 [[CMP]] to i64
+; CHECK-NEXT:    [[GEP_START:%.*]] = getelementptr i8, ptr [[PTR_IV_1_LCSSA]], i64 [[CMP_EXT]]
+; CHECK-NEXT:    br label %[[LOOP_2_HEADER:.*]]
+; CHECK:       [[LOOP_2_HEADER]]:
+; CHECK-NEXT:    [[INDVAR:%.*]] = phi i64 [ [[INDVAR_NEXT:%.*]], %[[LOOP_2_LATCH:.*]] ], [ 0, %[[MIDDLE]] ]
+; CHECK-NEXT:    [[PTR_IV_2:%.*]] = phi ptr [ [[GEP_START]], %[[MIDDLE]] ], [ [[PTR_IV_2_NEXT:%.*]], %[[LOOP_2_LATCH]] ]
+; CHECK-NEXT:    switch i8 [[X]], label %[[LOOP_2_LATCH]] [
+; CHECK-NEXT:      i8 1, label %[[LOOP_3_PREHEADER:.*]]
+; CHECK-NEXT:      i8 4, label %[[LOOP_3_PREHEADER]]
+; CHECK-NEXT:    ]
+; CHECK:       [[LOOP_3_PREHEADER]]:
+; CHECK-NEXT:    [[INDVAR_LCSSA:%.*]] = phi i64 [ [[INDVAR]], %[[LOOP_2_HEADER]] ], [ [[INDVAR]], %[[LOOP_2_HEADER]] ]
+; CHECK-NEXT:    [[PTR_IV_2_LCSSA:%.*]] = phi ptr [ [[PTR_IV_2]], %[[LOOP_2_HEADER]] ], [ [[PTR_IV_2]], %[[LOOP_2_HEADER]] ]
+; CHECK-NEXT:    [[TMP0:%.*]] = sub i64 0, [[START1]]
+; CHECK-NEXT:    [[TMP1:%.*]] = ptrtoint ptr [[PTR_IV_1_LCSSA]] to i64
+; CHECK-NEXT:    [[TMP2:%.*]] = add i64 [[TMP1]], [[TMP0]]
+; CHECK-NEXT:    [[TMP3:%.*]] = add i64 [[TMP2]], 1
+; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[CMP_EXT]], [[TMP3]]
+; CHECK-NEXT:    [[TMP5:%.*]] = add i64 [[INDVAR_LCSSA]], [[TMP4]]
+; CHECK-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[START]], i64 [[TMP5]]
+; CHECK-NEXT:    [[STRLEN:%.*]] = call i64 @strlen(ptr [[SCEVGEP]])
+; CHECK-NEXT:    br label %[[LOOP_3:.*]]
+; CHECK:       [[LOOP_2_LATCH]]:
+; CHECK-NEXT:    [[PTR_IV_2_NEXT]] = getelementptr i8, ptr [[PTR_IV_2]], i64 1
+; CHECK-NEXT:    [[INDVAR_NEXT]] = add i64 [[INDVAR]], 1
+; CHECK-NEXT:    br label %[[LOOP_2_HEADER]]
+; CHECK:       [[LOOP_3]]:
+; CHECK-NEXT:    [[IV_3:%.*]] = phi i64 [ [[IV_3_NEXT:%.*]], %[[LOOP_3]] ], [ 1, %[[LOOP_3_PREHEADER]] ]
+; CHECK-NEXT:    [[GEP:%.*]] = getelementptr i8, ptr [[PTR_IV_2_LCSSA]], i64 [[IV_3]]
+; CHECK-NEXT:    [[TMP6:%.*]] = load i8, ptr [[GEP]], align 1
+; CHECK-NEXT:    [[EC:%.*]] = icmp eq i8 [[TMP6]], 0
+; CHECK-NEXT:    [[IV_3_NEXT]] = add i64 [[IV_3]], 1
+; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[LOOP_3]]
+; CHECK:       [[EXIT]]:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br label %loop.1.header
+
+loop.1.header:
+  %ptr.iv.1 = phi ptr [ %start, %entry ], [ %ptr.iv.1.next, %loop.1.latch ]
+  %c = icmp ule i8 %x, 1
+  br i1 %c, label %loop.1.latch, label %middle
+
+loop.1.latch:
+  %ptr.iv.1.next = getelementptr i8, ptr %ptr.iv.1, i64 1
+  br label %loop.1.header
+
+middle:
+  %cmp = icmp eq i8 %x, 0
+  %cmp.ext = zext i1 %cmp to i64
+  %gep.start = getelementptr i8, ptr %ptr.iv.1, i64 %cmp.ext
+  br label %loop.2.header
+
+loop.2.header:
+  %ptr.iv.2 = phi ptr [ %gep.start, %middle ], [ %ptr.iv.2.next, %loop.2.latch ]
+  switch i8 %x, label %loop.2.latch [
+  i8 1, label %loop.3
+  i8 4, label %loop.3
+  ]
+
+loop.2.latch:
+  %ptr.iv.2.next = getelementptr i8, ptr %ptr.iv.2, i64 1
+  br label %loop.2.header
+
+loop.3:
+  %iv.3 = phi i64 [ 1, %loop.2.header ], [ 1, %loop.2.header ], [ %iv.3.next, %loop.3 ]
+  %gep = getelementptr i8, ptr %ptr.iv.2, i64 %iv.3
+  %1 = load i8, ptr %gep, align 1
+  %ec = icmp eq i8 %1, 0
+  %iv.3.next = add i64 %iv.3, 1
+  br i1 %ec, label %exit, label %loop.3
+
+exit:
+  ret void
+}
diff --git a/llvm/test/Transforms/LoopInterchange/profitability-vectorization-heuristic.ll b/llvm/test/Transforms/LoopInterchange/profitability-vectorization-heuristic.ll
index 9c113d4..4194849 100644
--- a/llvm/test/Transforms/LoopInterchange/profitability-vectorization-heuristic.ll
+++ b/llvm/test/Transforms/LoopInterchange/profitability-vectorization-heuristic.ll
@@ -5,6 +5,8 @@
 @A = dso_local global [256 x [256 x float]] zeroinitializer
 @B = dso_local global [256 x [256 x float]] zeroinitializer
 @C = dso_local global [256 x [256 x float]] zeroinitializer
+@D = global [256 x [256 x [256 x float]]] zeroinitializer
+@E = global [256 x [256 x [256 x float]]] zeroinitializer
 
 ; Check that the below loops are exchanged for vectorization.
 ;
@@ -64,15 +66,13 @@ exit:
 ;   for (int j = 1; j < 256; j++)
 ;     A[i][j-1] = A[i][j] + B[i][j];
 ;
-; FIXME: These loops are exchanged at this time due to the problem in
-; profitability heuristic calculation for vectorization.
 
-; CHECK:      --- !Passed
+; CHECK:      --- !Missed
 ; CHECK-NEXT: Pass:            loop-interchange
-; CHECK-NEXT: Name:            Interchanged
+; CHECK-NEXT: Name:            InterchangeNotProfitable
 ; CHECK-NEXT: Function:        interchange_unnecesasry_for_vectorization
 ; CHECK-NEXT: Args:
-; CHECK-NEXT:   - String:          Loop interchanged with enclosing loop.
+; CHECK-NEXT:   - String:          Insufficient information to calculate the cost of loop for interchange.
 define void @interchange_unnecesasry_for_vectorization() {
 entry:
   br label %for.i.header
@@ -103,3 +103,134 @@ for.i.inc:
 exit:
   ret void
 }
+
+; Check that the below loops are exchanged to allow innermost loop
+; vectorization. We cannot vectorize the j-loop because it has a lexically
+; backward dependency, but the i-loop can be vectorized because all the
+; loop-carried dependencies are lexically forward. LoopVectorize currently only
+; vectorizes innermost loop, hence move the i-loop to that position.
+;
+; for (int i = 0; i < 255; i++) {
+;   for (int j = 1; j < 256; j++) {
+;     A[i][j] = A[i][j-1] + B[i][j];
+;     C[i][j] += C[i+1][j];
+;   }
+; }
+;
+
+; CHECK:      --- !Passed
+; CHECK-NEXT: Pass:            loop-interchange
+; CHECK-NEXT: Name:            Interchanged
+; CHECK-NEXT: Function:        interchange_necessary_for_vectorization2
+; CHECK-NEXT: Args:
+; CHECK-NEXT:   - String:          Loop interchanged with enclosing loop.
+define void @interchange_necessary_for_vectorization2() {
+entry:
+  br label %for.i.header
+
+for.i.header:
+  %i = phi i64 [ 1, %entry ], [ %i.next, %for.i.inc ]
+  %i.inc = add i64 %i, 1
+  br label %for.j.body
+
+for.j.body:
+  %j = phi i64 [ 1, %for.i.header ], [ %j.next, %for.j.body ]
+  %j.dec = add i64 %j, -1
+  %a.load.index = getelementptr [256 x [256 x float]], ptr @A, i64 0, i64 %i, i64 %j.dec
+  %b.index = getelementptr [256 x [256 x float]], ptr @B, i64 0, i64 %i, i64 %j
+  %c.load.index = getelementptr [256 x [256 x float]], ptr @C, i64 0, i64 %i.inc, i64 %j
+  %c.store.index = getelementptr [256 x [256 x float]], ptr @C, i64 0, i64 %i, i64 %j
+  %a = load float, ptr %a.load.index
+  %b = load float, ptr %b.index
+  %c0 = load float, ptr %c.load.index
+  %c1 = load float, ptr %c.store.index
+  %add.0 = fadd float %a, %b
+  %a.store.index = getelementptr [256 x [256 x float]], ptr @A, i64 0, i64 %i, i64 %j
+  store float %add.0, ptr %a.store.index
+  %add.1 = fadd float %c0, %c1
+  store float %add.1, ptr %c.store.index
+  %j.next = add i64 %j, 1
+  %cmp.j = icmp eq i64 %j.next, 256
+  br i1 %cmp.j, label %for.i.inc, label %for.j.body
+
+for.i.inc:
+  %i.next = add i64 %i, 1
+  %cmp.i = icmp eq i64 %i.next, 255
+  br i1 %cmp.i, label %exit, label %for.i.header
+
+exit:
+  ret void
+}
+
+; Check that no interchange is performed for the following loop. Interchanging
+; the j-loop and k-loop makes the innermost loop vectorizble, since the j-loop
+; has only forward dependencies. However, at the moment, a loop body consisting
+; of multiple BBs is handled pesimistically. Hence the j-loop isn't moved to
+; the innermost place.
+;
+; for (int i = 0; i < 255; i++) {
+;   for (int j = 0; j < 255; j++) {
+;     for (int k = 0; k < 128; k++) {
+;       E[i][j][k] = D[i+1][j+1][2*k];
+;       if (cond)
+;         D[i][j][k+1] = 1.0;
+;   }
+; }
+
+; CHECK:      --- !Missed
+; CHECK-NEXT: Pass:            loop-interchange
+; CHECK-NEXT: Name:            InterchangeNotProfitable
+; CHECK-NEXT: Function:        multiple_BBs_in_loop
+; CHECK-NEXT: Args:
+; CHECK-NEXT:   - String:          Interchanging loops is not considered to improve cache locality nor vectorization.
+; CHECK:      --- !Missed
+; CHECK-NEXT: Pass:            loop-interchange
+; CHECK-NEXT: Name:            InterchangeNotProfitable
+; CHECK-NEXT: Function:        multiple_BBs_in_loop
+; CHECK-NEXT: Args:
+; CHECK-NEXT:   - String:          Interchanging loops is not considered to improve cache locality nor vectorization.
+define void @multiple_BBs_in_loop() {
+entry:
+  br label %for.i.header
+
+for.i.header:
+  %i = phi i64 [ 0, %entry ], [ %i.inc, %for.i.inc ]
+  %i.inc = add i64 %i, 1
+  br label %for.j.header
+
+for.j.header:
+  %j = phi i64 [ 0, %for.i.header ], [ %j.inc, %for.j.inc ]
+  %j.inc = add i64 %j, 1
+  br label %for.k.body
+
+for.k.body:
+  %k = phi i64 [ 0, %for.j.header ], [ %k.inc, %for.k.inc ]
+  %k.inc = add i64 %k, 1
+  %k.2 = mul i64 %k, 2
+  %d.index = getelementptr [256 x [256 x [256 x float]]], ptr @D, i64 0, i64 %i.inc, i64 %j.inc, i64 %k.2
+  %e.index = getelementptr [256 x [256 x [256 x float]]], ptr @E, i64 0, i64 %i, i64 %j, i64 %k
+  %d.load = load float, ptr %d.index
+  store float %d.load, ptr %e.index
+  %cond = freeze i1 undef
+  br i1 %cond, label %if.then, label %for.k.inc
+
+if.then:
+  %d.index2 = getelementptr [256 x [256 x [256 x float]]], ptr @D, i64 0, i64 %i, i64 %j, i64 %k.inc
+  store float 1.0, ptr %d.index2
+  br label %for.k.inc
+
+for.k.inc:
+  %cmp.k = icmp eq i64 %k.inc, 128
+  br i1 %cmp.k, label %for.j.inc, label %for.k.body
+
+for.j.inc:
+  %cmp.j = icmp eq i64 %j.inc, 255
+  br i1 %cmp.j, label %for.i.inc, label %for.j.header
+
+for.i.inc:
+  %cmp.i = icmp eq i64 %i.inc, 255
+  br i1 %cmp.i, label %exit, label %for.i.header
+
+exit:
+  ret void
+}
diff --git a/llvm/test/Transforms/LoopLoadElim/invalidate-laa-after-versioning.ll b/llvm/test/Transforms/LoopLoadElim/invalidate-laa-after-versioning.ll
index 10e1065..037851f 100644
--- a/llvm/test/Transforms/LoopLoadElim/invalidate-laa-after-versioning.ll
+++ b/llvm/test/Transforms/LoopLoadElim/invalidate-laa-after-versioning.ll
@@ -59,19 +59,14 @@ define void @test(ptr %arg, i64 %arg1) {
 ; CHECK-NEXT:    [[GEP_5:%.*]] = getelementptr inbounds double, ptr [[LCSSA_PTR_IV_1]], i64 1
 ; CHECK-NEXT:    br label [[INNER_2:%.*]]
 ; CHECK:       inner.2:
-; CHECK-NEXT:    [[INDVAR:%.*]] = phi i64 [ [[INDVAR_NEXT:%.*]], [[INNER_2]] ], [ 0, [[INNER_1_EXIT]] ]
 ; CHECK-NEXT:    [[PTR_IV_2:%.*]] = phi ptr [ [[GEP_5]], [[INNER_1_EXIT]] ], [ [[PTR_IV_2_NEXT:%.*]], [[INNER_2]] ]
 ; CHECK-NEXT:    [[PTR_IV_2_NEXT]] = getelementptr inbounds double, ptr [[PTR_IV_2]], i64 1
-; CHECK-NEXT:    [[INDVAR_NEXT]] = add i64 [[INDVAR]], 1
 ; CHECK-NEXT:    br i1 false, label [[INNER_3_LVER_CHECK:%.*]], label [[INNER_2]]
 ; CHECK:       inner.3.lver.check:
-; CHECK-NEXT:    [[INDVAR_LCSSA:%.*]] = phi i64 [ [[INDVAR]], [[INNER_2]] ]
 ; CHECK-NEXT:    [[LCSSA_PTR_IV_2:%.*]] = phi ptr [ [[PTR_IV_2]], [[INNER_2]] ]
 ; CHECK-NEXT:    [[GEP_6:%.*]] = getelementptr inbounds double, ptr [[PTR_PHI]], i64 1
 ; CHECK-NEXT:    [[GEP_7:%.*]] = getelementptr inbounds double, ptr [[LCSSA_PTR_IV_2]], i64 1
-; CHECK-NEXT:    [[TMP0:%.*]] = shl i64 [[INDVAR_LCSSA]], 3
-; CHECK-NEXT:    [[TMP1:%.*]] = add i64 [[TMP0]], 24
-; CHECK-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[LCSSA_PTR_IV_1]], i64 [[TMP1]]
+; CHECK-NEXT:    [[SCEVGEP3:%.*]] = getelementptr i8, ptr [[LCSSA_PTR_IV_2]], i64 16
 ; CHECK-NEXT:    [[BOUND0:%.*]] = icmp ult ptr [[GEP_7]], [[GEP_1]]
 ; CHECK-NEXT:    [[BOUND1:%.*]] = icmp ult ptr [[PTR_PHI]], [[SCEVGEP3]]
 ; CHECK-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
@@ -104,7 +99,7 @@ define void @test(ptr %arg, i64 %arg1) {
 ; CHECK-NEXT:    br i1 [[C_2]], label [[OUTER_LATCH_LOOPEXIT4:%.*]], label [[INNER_3]]
 ; CHECK:       outer.latch.loopexit:
 ; CHECK-NEXT:    br label [[OUTER_LATCH]]
-; CHECK:       outer.latch.loopexit4:
+; CHECK:       outer.latch.loopexit3:
 ; CHECK-NEXT:    br label [[OUTER_LATCH]]
 ; CHECK:       outer.latch:
 ; CHECK-NEXT:    br label [[INNER_1_LVER_CHECK]]
diff --git a/llvm/test/Transforms/LoopLoadElim/versioning-scev-invalidation.ll b/llvm/test/Transforms/LoopLoadElim/versioning-scev-invalidation.ll
index 7472680..abad40d 100644
--- a/llvm/test/Transforms/LoopLoadElim/versioning-scev-invalidation.ll
+++ b/llvm/test/Transforms/LoopLoadElim/versioning-scev-invalidation.ll
@@ -65,8 +65,7 @@ define void @g(ptr %dst.1, ptr %start, i64 %N) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[LCSSA_PTR_IV_1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr double, ptr [[NEXT_GEP]], i32 0
-; CHECK-NEXT:    store <4 x double> zeroinitializer, ptr [[TMP5]], align 8
+; CHECK-NEXT:    store <4 x double> zeroinitializer, ptr [[NEXT_GEP]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -84,7 +83,7 @@ define void @g(ptr %dst.1, ptr %start, i64 %N) {
 ; CHECK-NEXT:    [[PTR_IV_2_NEXT]] = getelementptr inbounds double, ptr [[PTR_IV_2]], i64 1
 ; CHECK-NEXT:    [[IV_2_NEXT]] = add nuw nsw i64 [[IV_2]], 1
 ; CHECK-NEXT:    [[EXITCOND_1_NOT:%.*]] = icmp eq i64 [[IV_2_NEXT]], [[N]]
-; CHECK-NEXT:    br i1 [[EXITCOND_1_NOT]], label [[EXIT_LOOPEXIT:%.*]], label [[LOOP_2]], !llvm.loop [[LOOP2:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND_1_NOT]], label [[EXIT_LOOPEXIT:%.*]], label [[LOOP_2]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       exit.loopexit:
 ; CHECK-NEXT:    br label [[EXIT]]
 ; CHECK:       exit:
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/blend-costs.ll b/llvm/test/Transforms/LoopVectorize/AArch64/blend-costs.ll
index 43b9424..e44ddbc 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/blend-costs.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/blend-costs.ll
@@ -22,8 +22,7 @@ define void @test_blend_feeding_replicated_store_1(i64 %N, ptr noalias %src, ptr
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE30:.*]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i32>, ptr [[TMP5]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i32>, ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp slt <16 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP7:%.*]] = select <16 x i1> [[TMP6]], <16 x i1> zeroinitializer, <16 x i1> zeroinitializer
 ; CHECK-NEXT:    [[TMP8:%.*]] = xor <16 x i1> [[TMP6]], splat (i1 true)
@@ -213,8 +212,7 @@ define void @test_blend_feeding_replicated_store_2(ptr noalias %src, ptr %dst, i
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE30:.*]] ]
 ; CHECK-NEXT:    [[GEP_SRC:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i32 [[IV]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[GEP_SRC]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[GEP_SRC]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq <16 x i8> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = xor <16 x i1> [[TMP3]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP6:%.*]] = select <16 x i1> [[TMP4]], <16 x i1> [[TMP5]], <16 x i1> zeroinitializer
@@ -368,7 +366,7 @@ define void @test_blend_feeding_replicated_store_2(ptr noalias %src, ptr %dst, i
 ; CHECK-NEXT:    [[TMP71:%.*]] = icmp eq i32 [[INDEX_NEXT]], 96
 ; CHECK-NEXT:    br i1 [[TMP71]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 96, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
@@ -388,7 +386,7 @@ define void @test_blend_feeding_replicated_store_2(ptr noalias %src, ptr %dst, i
 ; CHECK:       [[LOOP_LATCH]]:
 ; CHECK-NEXT:    [[IV_NEXT]] = add i32 [[IV1]], 1
 ; CHECK-NEXT:    [[EC:%.*]] = icmp eq i32 [[IV_NEXT]], 100
-; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/call-costs.ll b/llvm/test/Transforms/LoopVectorize/AArch64/call-costs.ll
index 8c2a48a..b7706da 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/call-costs.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/call-costs.ll
@@ -14,25 +14,23 @@ define void @fshl_operand_first_order_recurrence(ptr %dst, ptr noalias %src) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <2 x i64> [ <i64 poison, i64 0>, %[[VECTOR_PH]] ], [ [[WIDE_LOAD1:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[SRC]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 2
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP4]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD1]] = load <2 x i64>, ptr [[TMP5]], align 8
 ; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <2 x i64> [[VECTOR_RECUR]], <2 x i64> [[WIDE_LOAD]], <2 x i32> <i32 1, i32 2>
 ; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <2 x i64> [[WIDE_LOAD]], <2 x i64> [[WIDE_LOAD1]], <2 x i32> <i32 1, i32 2>
 ; CHECK-NEXT:    [[TMP8:%.*]] = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> splat (i64 1), <2 x i64> [[TMP6]], <2 x i64> splat (i64 1))
 ; CHECK-NEXT:    [[TMP9:%.*]] = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> splat (i64 1), <2 x i64> [[TMP7]], <2 x i64> splat (i64 1))
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i32 0
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i32 2
-; CHECK-NEXT:    store <2 x i64> [[TMP8]], ptr [[TMP12]], align 8
+; CHECK-NEXT:    store <2 x i64> [[TMP8]], ptr [[TMP10]], align 8
 ; CHECK-NEXT:    store <2 x i64> [[TMP9]], ptr [[TMP13]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i64> [[WIDE_LOAD1]], i32 1
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
@@ -47,7 +45,7 @@ define void @fshl_operand_first_order_recurrence(ptr %dst, ptr noalias %src) {
 ; CHECK-NEXT:    store i64 [[OR]], ptr [[GEP_DST]], align 8
 ; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV]], 100
-; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    ret void
 ;
@@ -79,16 +77,14 @@ define void @powi_call(ptr %P) {
 ; CHECK:       [[VECTOR_PH]]:
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds double, ptr [[P]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP2]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[P]], align 8
 ; CHECK-NEXT:    [[TMP3:%.*]] = call <2 x double> @llvm.powi.v2f64.i32(<2 x double> [[WIDE_LOAD]], i32 3)
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds double, ptr [[P]], i32 0
-; CHECK-NEXT:    store <2 x double> [[TMP3]], ptr [[TMP4]], align 8
+; CHECK-NEXT:    store <2 x double> [[TMP3]], ptr [[P]], align 8
 ; CHECK-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 2, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/check-prof-info.ll b/llvm/test/Transforms/LoopVectorize/AArch64/check-prof-info.ll
index 812bca9..027a88d 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/check-prof-info.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/check-prof-info.ll
@@ -10,9 +10,9 @@ target triple = "aarch64-unknown-linux-gnu"
 
 ; We expect the branch weight computations after vectorisation to use
 ; vscale=2 for neoverse-v1 and vscale=1 for neoverse-v2.
-define void @_Z3foov() {
+define void @_Z3foov(i64 %n) {
 ; CHECK-V1-IC1-LABEL: define void @_Z3foov(
-; CHECK-V1-IC1-SAME: ) #[[ATTR0:[0-9]+]] {
+; CHECK-V1-IC1-SAME: i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
 ; CHECK-V1-IC1:  [[ENTRY:.*:]]
 ; CHECK-V1-IC1:    br i1 [[MIN_ITERS_CHECK:%.*]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]], !prof [[PROF0:![0-9]+]]
 ; CHECK-V1-IC1:  [[VECTOR_PH]]:
@@ -28,35 +28,45 @@ define void @_Z3foov() {
 ; CHECK-V1-IC1:  [[FOR_COND_CLEANUP]]:
 ;
 ; CHECK-V2-IC1-LABEL: define void @_Z3foov(
-; CHECK-V2-IC1-SAME: ) #[[ATTR0:[0-9]+]] {
+; CHECK-V2-IC1-SAME: i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
 ; CHECK-V2-IC1:  [[ENTRY:.*:]]
-; CHECK-V2-IC1:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]], !prof [[PROF0:![0-9]+]]
+; CHECK-V2-IC1:    br i1 [[MIN_ITERS_CHECK:%.*]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]], !prof [[PROF0:![0-9]+]]
 ; CHECK-V2-IC1:  [[VECTOR_PH]]:
 ; CHECK-V2-IC1:    br label %[[VECTOR_BODY:.*]]
 ; CHECK-V2-IC1:  [[VECTOR_BODY]]:
 ; CHECK-V2-IC1:    br i1 [[TMP4:%.*]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !prof [[PROF1:![0-9]+]], !llvm.loop [[LOOP2:![0-9]+]]
 ; CHECK-V2-IC1:  [[MIDDLE_BLOCK]]:
-; CHECK-V2-IC1:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]], !prof [[PROF5:![0-9]+]]
+; CHECK-V2-IC1:    br i1 [[CMP_N:%.*]], label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]], !prof [[PROF5:![0-9]+]]
 ; CHECK-V2-IC1:  [[SCALAR_PH]]:
 ; CHECK-V2-IC1:    br label %[[FOR_BODY:.*]]
 ; CHECK-V2-IC1:  [[FOR_BODY]]:
-; CHECK-V2-IC1:    br i1 [[EXITCOND:%.*]], label %[[FOR_COND_CLEANUP]], label %[[FOR_BODY]], !prof [[PROF6:![0-9]+]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK-V2-IC1:    br i1 [[EXITCOND:%.*]], label %[[FOR_COND_CLEANUP]], label %[[FOR_BODY]], !prof [[PROF5:![0-9]+]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK-V2-IC1:  [[FOR_COND_CLEANUP]]:
 ;
 ; CHECK-V2-IC4-LABEL: define void @_Z3foov(
-; CHECK-V2-IC4-SAME: ) #[[ATTR0:[0-9]+]] {
+; CHECK-V2-IC4-SAME: i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
 ; CHECK-V2-IC4:  [[VEC_EPILOG_VECTOR_BODY1:.*:]]
-; CHECK-V2-IC4:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]], !prof [[PROF0:![0-9]+]]
+; CHECK-V2-IC4:    br i1 [[MIN_ITERS_CHECK:%.*]], label %[[VEC_EPILOG_SCALAR_PH:.*]], label %[[VECTOR_MAIN_LOOP_ITER_CHECK:.*]], !prof [[PROF0:![0-9]+]]
+; CHECK-V2-IC4:  [[VECTOR_MAIN_LOOP_ITER_CHECK]]:
+; CHECK-V2-IC4:    br i1 [[MIN_ITERS_CHECK1:%.*]], label %[[VEC_EPILOG_PH:.*]], label %[[VECTOR_PH:.*]], !prof [[PROF0]]
 ; CHECK-V2-IC4:  [[VECTOR_PH]]:
 ; CHECK-V2-IC4:    br label %[[VECTOR_BODY:.*]]
 ; CHECK-V2-IC4:  [[VECTOR_BODY]]:
 ; CHECK-V2-IC4:    br i1 [[TMP10:%.*]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !prof [[PROF1:![0-9]+]], !llvm.loop [[LOOP2:![0-9]+]]
 ; CHECK-V2-IC4:  [[MIDDLE_BLOCK]]:
-; CHECK-V2-IC4:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]], !prof [[PROF5:![0-9]+]]
-; CHECK-V2-IC4:  [[SCALAR_PH]]:
+; CHECK-V2-IC4:    br i1 [[CMP_N:%.*]], label %[[FOR_COND_CLEANUP:.*]], label %[[VEC_EPILOG_ITER_CHECK:.*]], !prof [[PROF5:![0-9]+]]
+; CHECK-V2-IC4:  [[VEC_EPILOG_ITER_CHECK]]:
+; CHECK-V2-IC4:    br i1 [[MIN_EPILOG_ITERS_CHECK:%.*]], label %[[VEC_EPILOG_SCALAR_PH]], label %[[VEC_EPILOG_PH]], !prof [[PROF6:![0-9]+]]
+; CHECK-V2-IC4:  [[VEC_EPILOG_PH]]:
+; CHECK-V2-IC4:    br label %[[VEC_EPILOG_VECTOR_BODY:.*]]
+; CHECK-V2-IC4:  [[VEC_EPILOG_VECTOR_BODY]]:
+; CHECK-V2-IC4:    br i1 [[TMP15:%.*]], label %[[VEC_EPILOG_MIDDLE_BLOCK:.*]], label %[[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK-V2-IC4:  [[VEC_EPILOG_MIDDLE_BLOCK]]:
+; CHECK-V2-IC4:    br i1 [[CMP_N10:%.*]], label %[[FOR_COND_CLEANUP]], label %[[VEC_EPILOG_SCALAR_PH]], !prof [[PROF8:![0-9]+]]
+; CHECK-V2-IC4:  [[VEC_EPILOG_SCALAR_PH]]:
 ; CHECK-V2-IC4:    br label %[[FOR_BODY:.*]]
 ; CHECK-V2-IC4:  [[FOR_BODY]]:
-; CHECK-V2-IC4:    br i1 [[EXITCOND:%.*]], label %[[FOR_COND_CLEANUP]], label %[[FOR_BODY]], !prof [[PROF6:![0-9]+]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK-V2-IC4:    br i1 [[EXITCOND:%.*]], label %[[FOR_COND_CLEANUP]], label %[[FOR_BODY]], !prof [[PROF9:![0-9]+]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK-V2-IC4:  [[FOR_COND_CLEANUP]]:
 ;
 entry:
@@ -69,7 +79,7 @@ for.body:                                         ; preds = %for.body, %entry
   %arrayidx2 = getelementptr inbounds [1024 x i32], ptr @a, i64 0, i64 %iv
   store i32 %load, ptr %arrayidx2, align 4
   %iv.next = add nuw nsw i64 %iv, 1
-  %exitcond = icmp eq i64 %iv.next, 1024
+  %exitcond = icmp eq i64 %iv.next, %n
   br i1 %exitcond, label %for.cond.cleanup, label %for.body, !prof !0
 
 for.cond.cleanup:                                 ; preds = %for.body
@@ -91,9 +101,8 @@ for.cond.cleanup:                                 ; preds = %for.body
 ; CHECK-V2-IC1: [[LOOP2]] = distinct !{[[LOOP2]], [[META3:![0-9]+]], [[META4:![0-9]+]]}
 ; CHECK-V2-IC1: [[META3]] = !{!"llvm.loop.isvectorized", i32 1}
 ; CHECK-V2-IC1: [[META4]] = !{!"llvm.loop.unroll.runtime.disable"}
-; CHECK-V2-IC1: [[PROF5]] = !{!"branch_weights", i32 1, i32 3}
-; CHECK-V2-IC1: [[PROF6]] = !{!"branch_weights", i32 0, i32 0}
-; CHECK-V2-IC1: [[LOOP7]] = distinct !{[[LOOP7]], [[META4]], [[META3]]}
+; CHECK-V2-IC1: [[PROF5]] = !{!"branch_weights", i32 0, i32 0}
+; CHECK-V2-IC1: [[LOOP6]] = distinct !{[[LOOP6]], [[META4]], [[META3]]}
 ;.
 ; CHECK-V2-IC4: [[PROF0]] = !{!"branch_weights", i32 1, i32 127}
 ; CHECK-V2-IC4: [[PROF1]] = !{!"branch_weights", i32 1, i32 63}
@@ -101,6 +110,9 @@ for.cond.cleanup:                                 ; preds = %for.body
 ; CHECK-V2-IC4: [[META3]] = !{!"llvm.loop.isvectorized", i32 1}
 ; CHECK-V2-IC4: [[META4]] = !{!"llvm.loop.unroll.runtime.disable"}
 ; CHECK-V2-IC4: [[PROF5]] = !{!"branch_weights", i32 1, i32 15}
-; CHECK-V2-IC4: [[PROF6]] = !{!"branch_weights", i32 0, i32 0}
-; CHECK-V2-IC4: [[LOOP7]] = distinct !{[[LOOP7]], [[META4]], [[META3]]}
+; CHECK-V2-IC4: [[PROF6]] = !{!"branch_weights", i32 4, i32 0}
+; CHECK-V2-IC4: [[LOOP7]] = distinct !{[[LOOP7]], [[META3]], [[META4]]}
+; CHECK-V2-IC4: [[PROF8]] = !{!"branch_weights", i32 1, i32 3}
+; CHECK-V2-IC4: [[PROF9]] = !{!"branch_weights", i32 0, i32 0}
+; CHECK-V2-IC4: [[LOOP10]] = distinct !{[[LOOP10]], [[META4]], [[META3]]}
 ;.
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/clamped-trip-count.ll b/llvm/test/Transforms/LoopVectorize/AArch64/clamped-trip-count.ll
index 95f3eb7..795de3d 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/clamped-trip-count.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/clamped-trip-count.ll
@@ -33,8 +33,7 @@ define void @clamped_tc_8(ptr nocapture %dst, i32 %n, i64 %val) vscale_range(1,1
 ; CHECK-NEXT:    [[TMP10:%.*]] = shl nuw nsw <vscale x 8 x i64> [[VEC_IND]], splat (i64 3)
 ; CHECK-NEXT:    [[TMP11:%.*]] = lshr <vscale x 8 x i64> [[BROADCAST_SPLAT]], [[TMP10]]
 ; CHECK-NEXT:    [[TMP14:%.*]] = trunc <vscale x 8 x i64> [[TMP11]] to <vscale x 8 x i8>
-; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.nxv8i8.p0(<vscale x 8 x i8> [[TMP14]], ptr [[TMP17]], i32 1, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv8i8.p0(<vscale x 8 x i8> [[TMP14]], ptr [[NEXT_GEP]], i32 1, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX_NEXT]], i64 8)
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 8 x i64> [[VEC_IND]], [[DOTSPLAT]]
@@ -117,8 +116,7 @@ define void @clamped_tc_max_8(ptr nocapture %dst, i32 %n, i64 %val) vscale_range
 ; CHECK-NEXT:    [[TMP10:%.*]] = shl nuw nsw <vscale x 8 x i64> [[VEC_IND]], splat (i64 3)
 ; CHECK-NEXT:    [[TMP11:%.*]] = lshr <vscale x 8 x i64> [[BROADCAST_SPLAT]], [[TMP10]]
 ; CHECK-NEXT:    [[TMP14:%.*]] = trunc <vscale x 8 x i64> [[TMP11]] to <vscale x 8 x i8>
-; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.nxv8i8.p0(<vscale x 8 x i8> [[TMP14]], ptr [[TMP17]], i32 1, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv8i8.p0(<vscale x 8 x i8> [[TMP14]], ptr [[NEXT_GEP]], i32 1, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX_NEXT]], i64 [[WIDE_TRIP_COUNT]])
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 8 x i64> [[VEC_IND]], [[DOTSPLAT]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/conditional-branches-cost.ll b/llvm/test/Transforms/LoopVectorize/AArch64/conditional-branches-cost.ll
index 46a194d..0232d88 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/conditional-branches-cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/conditional-branches-cost.ll
@@ -82,9 +82,8 @@ define void @loop_dependent_cond(ptr %src, ptr noalias %dst, i64 %N) {
 ; DEFAULT:       [[VECTOR_BODY]]:
 ; DEFAULT-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE7:.*]] ]
 ; DEFAULT-NEXT:    [[TMP3:%.*]] = getelementptr double, ptr [[SRC]], i64 [[INDEX]]
-; DEFAULT-NEXT:    [[TMP5:%.*]] = getelementptr double, ptr [[TMP3]], i32 0
 ; DEFAULT-NEXT:    [[TMP6:%.*]] = getelementptr double, ptr [[TMP3]], i32 2
-; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP5]], align 8
+; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP3]], align 8
 ; DEFAULT-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x double>, ptr [[TMP6]], align 8
 ; DEFAULT-NEXT:    [[TMP7:%.*]] = call <2 x double> @llvm.fabs.v2f64(<2 x double> [[WIDE_LOAD]])
 ; DEFAULT-NEXT:    [[TMP8:%.*]] = call <2 x double> @llvm.fabs.v2f64(<2 x double> [[WIDE_LOAD1]])
@@ -341,9 +340,8 @@ define void @latch_branch_cost(ptr %dst) {
 ; DEFAULT:       [[VECTOR_BODY]]:
 ; DEFAULT-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; DEFAULT-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INDEX]]
-; DEFAULT-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[TMP2]], i32 0
 ; DEFAULT-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP2]], i32 16
-; DEFAULT-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP6]], align 1
+; DEFAULT-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP2]], align 1
 ; DEFAULT-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP5]], align 1
 ; DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
 ; DEFAULT-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 96
@@ -358,8 +356,7 @@ define void @latch_branch_cost(ptr %dst) {
 ; DEFAULT:       [[VEC_EPILOG_VECTOR_BODY]]:
 ; DEFAULT-NEXT:    [[INDEX1:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], %[[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT2:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
 ; DEFAULT-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INDEX1]]
-; DEFAULT-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[TMP8]], i32 0
-; DEFAULT-NEXT:    store <4 x i8> zeroinitializer, ptr [[TMP9]], align 1
+; DEFAULT-NEXT:    store <4 x i8> zeroinitializer, ptr [[TMP8]], align 1
 ; DEFAULT-NEXT:    [[INDEX_NEXT2]] = add nuw i64 [[INDEX1]], 4
 ; DEFAULT-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT2]], 100
 ; DEFAULT-NEXT:    br i1 [[TMP10]], label %[[VEC_EPILOG_MIDDLE_BLOCK:.*]], label %[[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
@@ -575,8 +572,7 @@ define i32 @header_mask_and_invariant_compare(ptr %A, ptr %B, ptr %C, ptr %D, pt
 ; DEFAULT-NEXT:    store i32 [[TMP22]], ptr [[E]], align 4, !alias.scope [[META14]], !noalias [[META16]]
 ; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE37]]
 ; DEFAULT:       [[PRED_STORE_CONTINUE37]]:
-; DEFAULT-NEXT:    [[TMP17:%.*]] = getelementptr i32, ptr [[TMP16]], i32 0
-; DEFAULT-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> zeroinitializer, ptr [[TMP17]], i32 4, <4 x i1> [[TMP8]]), !alias.scope [[META18:![0-9]+]], !noalias [[META19:![0-9]+]]
+; DEFAULT-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> zeroinitializer, ptr [[TMP16]], i32 4, <4 x i1> [[TMP8]]), !alias.scope [[META18:![0-9]+]], !noalias [[META19:![0-9]+]]
 ; DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; DEFAULT-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; DEFAULT-NEXT:    br i1 [[TMP18]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
@@ -674,13 +670,12 @@ define void @multiple_exit_conditions(ptr %src, ptr noalias %dst) #1 {
 ; DEFAULT-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x i16> [[BROADCAST_SPLATINSERT]], <8 x i16> poison, <8 x i32> zeroinitializer
 ; DEFAULT-NEXT:    [[TMP2:%.*]] = or <8 x i16> [[BROADCAST_SPLAT]], splat (i16 1)
 ; DEFAULT-NEXT:    [[TMP3:%.*]] = uitofp <8 x i16> [[TMP2]] to <8 x double>
-; DEFAULT-NEXT:    [[TMP4:%.*]] = getelementptr double, ptr [[NEXT_GEP]], i32 0
-; DEFAULT-NEXT:    store <8 x double> [[TMP3]], ptr [[TMP4]], align 8
+; DEFAULT-NEXT:    store <8 x double> [[TMP3]], ptr [[NEXT_GEP]], align 8
 ; DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; DEFAULT-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; DEFAULT-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
 ; DEFAULT:       [[MIDDLE_BLOCK]]:
-; DEFAULT-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; DEFAULT-NEXT:    br label %[[SCALAR_PH]]
 ; DEFAULT:       [[SCALAR_PH]]:
 ; DEFAULT-NEXT:    [[BC_RESUME_VAL:%.*]] = phi ptr [ [[IND_END]], %[[MIDDLE_BLOCK]] ], [ [[DST]], %[[ENTRY]] ]
 ; DEFAULT-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i64 [ 512, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
@@ -696,7 +691,7 @@ define void @multiple_exit_conditions(ptr %src, ptr noalias %dst) #1 {
 ; DEFAULT-NEXT:    [[PTR_IV_NEXT]] = getelementptr i8, ptr [[PTR_IV]], i64 8
 ; DEFAULT-NEXT:    [[IV_CLAMP:%.*]] = and i64 [[IV]], 4294967294
 ; DEFAULT-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV_CLAMP]], 512
-; DEFAULT-NEXT:    br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP23:![0-9]+]]
+; DEFAULT-NEXT:    br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP]], !llvm.loop [[LOOP23:![0-9]+]]
 ; DEFAULT:       [[EXIT]]:
 ; DEFAULT-NEXT:    ret void
 ;
@@ -730,8 +725,7 @@ define void @multiple_exit_conditions(ptr %src, ptr noalias %dst) #1 {
 ; PRED-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i16> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i16> poison, <vscale x 2 x i32> zeroinitializer
 ; PRED-NEXT:    [[TMP13:%.*]] = or <vscale x 2 x i16> [[BROADCAST_SPLAT]], splat (i16 1)
 ; PRED-NEXT:    [[TMP14:%.*]] = uitofp <vscale x 2 x i16> [[TMP13]] to <vscale x 2 x double>
-; PRED-NEXT:    [[TMP15:%.*]] = getelementptr double, ptr [[NEXT_GEP]], i32 0
-; PRED-NEXT:    call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP14]], ptr [[TMP15]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; PRED-NEXT:    call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> [[TMP14]], ptr [[NEXT_GEP]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
 ; PRED-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP5]]
 ; PRED-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[INDEX]], i64 [[TMP10]])
 ; PRED-NEXT:    [[TMP16:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
@@ -1492,7 +1486,7 @@ define void @redundant_branch_and_tail_folding(ptr %dst, i1 %c) {
 ; DEFAULT-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 16
 ; DEFAULT-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
 ; DEFAULT:       [[MIDDLE_BLOCK]]:
-; DEFAULT-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; DEFAULT-NEXT:    br label %[[SCALAR_PH]]
 ; DEFAULT:       [[SCALAR_PH]]:
 ; DEFAULT-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 16, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; DEFAULT-NEXT:    br label %[[LOOP_HEADER:.*]]
@@ -1506,7 +1500,7 @@ define void @redundant_branch_and_tail_folding(ptr %dst, i1 %c) {
 ; DEFAULT-NEXT:    [[T:%.*]] = trunc nuw nsw i64 [[IV_NEXT]] to i32
 ; DEFAULT-NEXT:    store i32 [[T]], ptr [[DST]], align 4
 ; DEFAULT-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], 21
-; DEFAULT-NEXT:    br i1 [[EC]], label %[[EXIT]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP29:![0-9]+]]
+; DEFAULT-NEXT:    br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP29:![0-9]+]]
 ; DEFAULT:       [[EXIT]]:
 ; DEFAULT-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/deterministic-type-shrinkage.ll b/llvm/test/Transforms/LoopVectorize/AArch64/deterministic-type-shrinkage.ll
index 73ef8534b..06e6306 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/deterministic-type-shrinkage.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/deterministic-type-shrinkage.ll
@@ -469,7 +469,7 @@ define void @old_and_new_size_equalko(ptr noalias %src, ptr noalias %dst) {
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/divs-with-scalable-vfs.ll b/llvm/test/Transforms/LoopVectorize/AArch64/divs-with-scalable-vfs.ll
index d42be20..1ad1e426 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/divs-with-scalable-vfs.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/divs-with-scalable-vfs.ll
@@ -38,11 +38,10 @@ define void @sdiv_feeding_gep(ptr %dst, i32 %x, i64 %M, i64 %conv6, i64 %N) {
 ; CHECK-NEXT:    [[TMP30:%.*]] = add i32 [[TMP28]], [[TMP26]]
 ; CHECK-NEXT:    [[TMP32:%.*]] = sext i32 [[TMP30]] to i64
 ; CHECK-NEXT:    [[TMP34:%.*]] = getelementptr double, ptr [[DST]], i64 [[TMP32]]
-; CHECK-NEXT:    [[TMP36:%.*]] = getelementptr double, ptr [[TMP34]], i32 0
 ; CHECK-NEXT:    [[TMP37:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP38:%.*]] = mul nuw i64 [[TMP37]], 2
 ; CHECK-NEXT:    [[TMP39:%.*]] = getelementptr double, ptr [[TMP34]], i64 [[TMP38]]
-; CHECK-NEXT:    store <vscale x 2 x double> zeroinitializer, ptr [[TMP36]], align 8
+; CHECK-NEXT:    store <vscale x 2 x double> zeroinitializer, ptr [[TMP34]], align 8
 ; CHECK-NEXT:    store <vscale x 2 x double> zeroinitializer, ptr [[TMP39]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP11]]
 ; CHECK-NEXT:    [[TMP40:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -149,8 +148,7 @@ define void @sdiv_feeding_gep_predicated(ptr %dst, i32 %x, i64 %M, i64 %conv6, i
 ; CHECK-NEXT:    [[TMP32:%.*]] = add i32 [[TMP31]], [[TMP30]]
 ; CHECK-NEXT:    [[TMP33:%.*]] = sext i32 [[TMP32]] to i64
 ; CHECK-NEXT:    [[TMP34:%.*]] = getelementptr double, ptr [[DST]], i64 [[TMP33]]
-; CHECK-NEXT:    [[TMP35:%.*]] = getelementptr double, ptr [[TMP34]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> zeroinitializer, ptr [[TMP35]], i32 8, <vscale x 2 x i1> [[TMP23]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv2f64.p0(<vscale x 2 x double> zeroinitializer, ptr [[TMP34]], i32 8, <vscale x 2 x i1> [[TMP23]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP9]]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[INDEX]], i64 [[TMP14]])
 ; CHECK-NEXT:    [[TMP36:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
@@ -275,8 +273,7 @@ define void @udiv_urem_feeding_gep(i64 %x, ptr %dst, i64 %N) {
 ; CHECK-NEXT:    [[TMP36:%.*]] = shl i64 [[TMP35]], 32
 ; CHECK-NEXT:    [[TMP37:%.*]] = ashr i64 [[TMP36]], 32
 ; CHECK-NEXT:    [[TMP38:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP37]]
-; CHECK-NEXT:    [[TMP39:%.*]] = getelementptr i64, ptr [[TMP38]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP23]], ptr [[TMP39]], i32 4, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP23]], ptr [[TMP38]], i32 4, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP9]]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[INDEX]], i64 [[TMP14]])
 ; CHECK-NEXT:    [[TMP47:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/drop-poison-generating-flags.ll b/llvm/test/Transforms/LoopVectorize/AArch64/drop-poison-generating-flags.ll
index e28c79e..221d944 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/drop-poison-generating-flags.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/drop-poison-generating-flags.ll
@@ -15,15 +15,13 @@ define void @check_widen_intrinsic_with_nnan(ptr noalias %dst.0, ptr noalias %ds
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_LOAD_CONTINUE6:.*]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds double, ptr [[SRC_1]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds double, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP2]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP1]], align 8
 ; CHECK-NEXT:    [[TMP3:%.*]] = call <4 x double> @llvm.fabs.v4f64(<4 x double> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[TMP4:%.*]] = fcmp olt <4 x double> [[TMP3]], splat (double 1.000000e+00)
 ; CHECK-NEXT:    [[TMP5:%.*]] = xor <4 x i1> [[TMP4]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], -1
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr double, ptr [[DST_0]], i64 [[TMP6]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr double, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> zeroinitializer, ptr [[TMP8]], i32 8, <4 x i1> [[TMP5]])
+; CHECK-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> zeroinitializer, ptr [[TMP7]], i32 8, <4 x i1> [[TMP5]])
 ; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <4 x i1> [[TMP4]], i32 0
 ; CHECK-NEXT:    br i1 [[TMP9]], label %[[PRED_LOAD_IF:.*]], label %[[PRED_LOAD_CONTINUE:.*]]
 ; CHECK:       [[PRED_LOAD_IF]]:
@@ -58,23 +56,21 @@ define void @check_widen_intrinsic_with_nnan(ptr noalias %dst.0, ptr noalias %ds
 ; CHECK-NEXT:    [[TMP24:%.*]] = phi <4 x double> [ [[TMP20]], %[[PRED_LOAD_CONTINUE4]] ], [ [[TMP23]], %[[PRED_LOAD_IF5]] ]
 ; CHECK-NEXT:    [[TMP25:%.*]] = add i64 [[INDEX]], -1
 ; CHECK-NEXT:    [[TMP26:%.*]] = getelementptr double, ptr [[DST_0]], i64 [[TMP25]]
-; CHECK-NEXT:    [[TMP27:%.*]] = getelementptr double, ptr [[TMP26]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> zeroinitializer, ptr [[TMP27]], i32 8, <4 x i1> [[TMP4]])
+; CHECK-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> zeroinitializer, ptr [[TMP26]], i32 8, <4 x i1> [[TMP4]])
 ; CHECK-NEXT:    [[TMP28:%.*]] = fcmp oeq <4 x double> [[TMP24]], zeroinitializer
 ; CHECK-NEXT:    [[TMP29:%.*]] = select <4 x i1> [[TMP4]], <4 x i1> [[TMP28]], <4 x i1> zeroinitializer
 ; CHECK-NEXT:    [[TMP30:%.*]] = or <4 x i1> [[TMP5]], [[TMP29]]
 ; CHECK-NEXT:    [[TMP31:%.*]] = extractelement <4 x i1> [[TMP29]], i32 0
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select i1 [[TMP31]], i64 [[TMP25]], i64 [[TMP6]]
 ; CHECK-NEXT:    [[TMP32:%.*]] = getelementptr i32, ptr [[DST_1]], i64 [[PREDPHI]]
-; CHECK-NEXT:    [[TMP33:%.*]] = getelementptr i32, ptr [[TMP32]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> splat (i32 10), ptr [[TMP33]], i32 4, <4 x i1> [[TMP30]])
+; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> splat (i32 10), ptr [[TMP32]], i32 4, <4 x i1> [[TMP30]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP34:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP34]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/eliminate-tail-predication.ll b/llvm/test/Transforms/LoopVectorize/AArch64/eliminate-tail-predication.ll
index c824bee..ab008e7 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/eliminate-tail-predication.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/eliminate-tail-predication.ll
@@ -23,11 +23,10 @@ define void @f1(ptr %A) #0 {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <vscale x 4 x i32> splat (i32 1), ptr [[TMP7]], align 4
+; CHECK-NEXT:    store <vscale x 4 x i32> splat (i32 1), ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
-; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/epilog-iv-select-cmp.ll b/llvm/test/Transforms/LoopVectorize/AArch64/epilog-iv-select-cmp.ll
index 895781de..8013a8f 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/epilog-iv-select-cmp.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/epilog-iv-select-cmp.ll
@@ -27,9 +27,8 @@ define i8 @select_icmp_var_start(ptr %a, i8 %n, i8 %start) {
 ; CHECK-NEXT:    [[STEP_ADD:%.*]] = add <16 x i8> [[VEC_IND]], splat (i8 16)
 ; CHECK-NEXT:    [[INDEX4:%.*]] = trunc i32 [[INDEX]] to i8
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[A]], i8 [[INDEX4]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i32 0
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i32 16
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP9]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP8]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <16 x i8>, ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq <16 x i8> [[WIDE_LOAD]], splat (i8 3)
 ; CHECK-NEXT:    [[TMP23:%.*]] = icmp eq <16 x i8> [[WIDE_LOAD3]], splat (i8 3)
@@ -72,8 +71,7 @@ define i8 @select_icmp_var_start(ptr %a, i8 %n, i8 %start) {
 ; CHECK-NEXT:    [[VEC_PHI9:%.*]] = phi <8 x i8> [ [[DOTSPLAT]], %[[VEC_EPILOG_PH]] ], [ [[TMP20:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[IV:%.*]] = trunc i32 [[INDEX6]] to i8
 ; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i8, ptr [[A]], i8 [[IV]]
-; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[GEP]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD12:%.*]] = load <8 x i8>, ptr [[TMP18]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD12:%.*]] = load <8 x i8>, ptr [[GEP]], align 8
 ; CHECK-NEXT:    [[TMP19:%.*]] = icmp eq <8 x i8> [[WIDE_LOAD12]], splat (i8 3)
 ; CHECK-NEXT:    [[TMP20]] = select <8 x i1> [[TMP19]], <8 x i8> [[VEC_IND7]], <8 x i8> [[VEC_PHI9]]
 ; CHECK-NEXT:    [[INDEX_NEXT13]] = add nuw i32 [[INDEX6]], 8
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/epilog-vectorization-factors.ll b/llvm/test/Transforms/LoopVectorize/AArch64/epilog-vectorization-factors.ll
index e6fad4b..37b5f5d 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/epilog-vectorization-factors.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/epilog-vectorization-factors.ll
@@ -19,20 +19,18 @@ define void @add_i8(ptr noalias nocapture noundef writeonly %A, ptr nocapture no
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[B:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 16
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 32
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 48
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[C:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[TMP6]], i32 0
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[TMP6]], i32 16
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[TMP6]], i32 32
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[TMP6]], i32 48
-; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <16 x i8>, ptr [[TMP7]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <16 x i8>, ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <16 x i8>, ptr [[TMP8]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <16 x i8>, ptr [[TMP9]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <16 x i8>, ptr [[TMP10]], align 1
@@ -41,11 +39,10 @@ define void @add_i8(ptr noalias nocapture noundef writeonly %A, ptr nocapture no
 ; CHECK-NEXT:    [[TMP13:%.*]] = add <16 x i8> [[WIDE_LOAD7]], [[WIDE_LOAD3]]
 ; CHECK-NEXT:    [[TMP14:%.*]] = add <16 x i8> [[WIDE_LOAD8]], [[WIDE_LOAD4]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[TMP15]], i32 0
 ; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP15]], i32 16
 ; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[TMP15]], i32 32
 ; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[TMP15]], i32 48
-; CHECK-NEXT:    store <16 x i8> [[TMP11]], ptr [[TMP16]], align 1
+; CHECK-NEXT:    store <16 x i8> [[TMP11]], ptr [[TMP15]], align 1
 ; CHECK-NEXT:    store <16 x i8> [[TMP12]], ptr [[TMP17]], align 1
 ; CHECK-NEXT:    store <16 x i8> [[TMP13]], ptr [[TMP18]], align 1
 ; CHECK-NEXT:    store <16 x i8> [[TMP14]], ptr [[TMP19]], align 1
@@ -67,15 +64,12 @@ define void @add_i8(ptr noalias nocapture noundef writeonly %A, ptr nocapture no
 ; CHECK:       vec.epilog.vector.body:
 ; CHECK-NEXT:    [[INDEX11:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT14:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX11]]
-; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i8, ptr [[TMP22]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD12:%.*]] = load <8 x i8>, ptr [[TMP23]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD12:%.*]] = load <8 x i8>, ptr [[TMP22]], align 1
 ; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i8, ptr [[C]], i64 [[INDEX11]]
-; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i8, ptr [[TMP24]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD13:%.*]] = load <8 x i8>, ptr [[TMP25]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD13:%.*]] = load <8 x i8>, ptr [[TMP24]], align 1
 ; CHECK-NEXT:    [[TMP26:%.*]] = add <8 x i8> [[WIDE_LOAD13]], [[WIDE_LOAD12]]
 ; CHECK-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX11]]
-; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i8, ptr [[TMP27]], i32 0
-; CHECK-NEXT:    store <8 x i8> [[TMP26]], ptr [[TMP28]], align 1
+; CHECK-NEXT:    store <8 x i8> [[TMP26]], ptr [[TMP27]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT14]] = add nuw i64 [[INDEX11]], 8
 ; CHECK-NEXT:    [[TMP29:%.*]] = icmp eq i64 [[INDEX_NEXT14]], [[N_VEC10]]
 ; CHECK-NEXT:    br i1 [[TMP29]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
@@ -135,20 +129,18 @@ define void @add_i16(ptr noalias nocapture noundef writeonly %A, ptr nocapture n
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[B:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 8
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 16
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 24
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i16>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i16>, ptr [[TMP4]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <8 x i16>, ptr [[TMP5]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[C:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i16, ptr [[TMP6]], i32 0
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[TMP6]], i32 8
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i16, ptr [[TMP6]], i32 16
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i16, ptr [[TMP6]], i32 24
-; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <8 x i16>, ptr [[TMP7]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <8 x i16>, ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <8 x i16>, ptr [[TMP8]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <8 x i16>, ptr [[TMP9]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <8 x i16>, ptr [[TMP10]], align 1
@@ -157,11 +149,10 @@ define void @add_i16(ptr noalias nocapture noundef writeonly %A, ptr nocapture n
 ; CHECK-NEXT:    [[TMP13:%.*]] = add <8 x i16> [[WIDE_LOAD7]], [[WIDE_LOAD3]]
 ; CHECK-NEXT:    [[TMP14:%.*]] = add <8 x i16> [[WIDE_LOAD8]], [[WIDE_LOAD4]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i16, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i16, ptr [[TMP15]], i32 0
 ; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i16, ptr [[TMP15]], i32 8
 ; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i16, ptr [[TMP15]], i32 16
 ; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i16, ptr [[TMP15]], i32 24
-; CHECK-NEXT:    store <8 x i16> [[TMP11]], ptr [[TMP16]], align 1
+; CHECK-NEXT:    store <8 x i16> [[TMP11]], ptr [[TMP15]], align 1
 ; CHECK-NEXT:    store <8 x i16> [[TMP12]], ptr [[TMP17]], align 1
 ; CHECK-NEXT:    store <8 x i16> [[TMP13]], ptr [[TMP18]], align 1
 ; CHECK-NEXT:    store <8 x i16> [[TMP14]], ptr [[TMP19]], align 1
@@ -183,15 +174,12 @@ define void @add_i16(ptr noalias nocapture noundef writeonly %A, ptr nocapture n
 ; CHECK:       vec.epilog.vector.body:
 ; CHECK-NEXT:    [[INDEX11:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT14:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i16, ptr [[B]], i64 [[INDEX11]]
-; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i16, ptr [[TMP22]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD12:%.*]] = load <4 x i16>, ptr [[TMP23]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD12:%.*]] = load <4 x i16>, ptr [[TMP22]], align 1
 ; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i16, ptr [[C]], i64 [[INDEX11]]
-; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i16, ptr [[TMP24]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD13:%.*]] = load <4 x i16>, ptr [[TMP25]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD13:%.*]] = load <4 x i16>, ptr [[TMP24]], align 1
 ; CHECK-NEXT:    [[TMP26:%.*]] = add <4 x i16> [[WIDE_LOAD13]], [[WIDE_LOAD12]]
 ; CHECK-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[INDEX11]]
-; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i16, ptr [[TMP27]], i32 0
-; CHECK-NEXT:    store <4 x i16> [[TMP26]], ptr [[TMP28]], align 1
+; CHECK-NEXT:    store <4 x i16> [[TMP26]], ptr [[TMP27]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT14]] = add nuw i64 [[INDEX11]], 4
 ; CHECK-NEXT:    [[TMP29:%.*]] = icmp eq i64 [[INDEX_NEXT14]], [[N_VEC10]]
 ; CHECK-NEXT:    br i1 [[TMP29]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
@@ -251,20 +239,18 @@ define void @add_i32(ptr noalias nocapture noundef writeonly %A, ptr nocapture n
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 8
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 12
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i32>, ptr [[TMP4]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i32>, ptr [[TMP5]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 4
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 8
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 12
-; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i32>, ptr [[TMP7]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i32>, ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i32>, ptr [[TMP8]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i32>, ptr [[TMP9]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x i32>, ptr [[TMP10]], align 1
@@ -273,11 +259,10 @@ define void @add_i32(ptr noalias nocapture noundef writeonly %A, ptr nocapture n
 ; CHECK-NEXT:    [[TMP13:%.*]] = add <4 x i32> [[WIDE_LOAD7]], [[WIDE_LOAD3]]
 ; CHECK-NEXT:    [[TMP14:%.*]] = add <4 x i32> [[WIDE_LOAD8]], [[WIDE_LOAD4]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[TMP15]], i32 0
 ; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP15]], i32 4
 ; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[TMP15]], i32 8
 ; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[TMP15]], i32 12
-; CHECK-NEXT:    store <4 x i32> [[TMP11]], ptr [[TMP16]], align 1
+; CHECK-NEXT:    store <4 x i32> [[TMP11]], ptr [[TMP15]], align 1
 ; CHECK-NEXT:    store <4 x i32> [[TMP12]], ptr [[TMP17]], align 1
 ; CHECK-NEXT:    store <4 x i32> [[TMP13]], ptr [[TMP18]], align 1
 ; CHECK-NEXT:    store <4 x i32> [[TMP14]], ptr [[TMP19]], align 1
@@ -299,15 +284,12 @@ define void @add_i32(ptr noalias nocapture noundef writeonly %A, ptr nocapture n
 ; CHECK:       vec.epilog.vector.body:
 ; CHECK-NEXT:    [[INDEX11:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT14:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX11]]
-; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, ptr [[TMP22]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD12:%.*]] = load <4 x i32>, ptr [[TMP23]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD12:%.*]] = load <4 x i32>, ptr [[TMP22]], align 1
 ; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[INDEX11]]
-; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i32, ptr [[TMP24]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD13:%.*]] = load <4 x i32>, ptr [[TMP25]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD13:%.*]] = load <4 x i32>, ptr [[TMP24]], align 1
 ; CHECK-NEXT:    [[TMP26:%.*]] = add <4 x i32> [[WIDE_LOAD13]], [[WIDE_LOAD12]]
 ; CHECK-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX11]]
-; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i32, ptr [[TMP27]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP26]], ptr [[TMP28]], align 1
+; CHECK-NEXT:    store <4 x i32> [[TMP26]], ptr [[TMP27]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT14]] = add nuw i64 [[INDEX11]], 4
 ; CHECK-NEXT:    [[TMP29:%.*]] = icmp eq i64 [[INDEX_NEXT14]], [[N_VEC10]]
 ; CHECK-NEXT:    br i1 [[TMP29]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/epilog-vectorization-widen-inductions.ll b/llvm/test/Transforms/LoopVectorize/AArch64/epilog-vectorization-widen-inductions.ll
index 83b35309..5b15896 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/epilog-vectorization-widen-inductions.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/epilog-vectorization-widen-inductions.ll
@@ -36,9 +36,8 @@ define void @test_widen_ptr_induction(ptr %ptr.start.1) {
 ; CHECK-NEXT:    tail call void @llvm.assume(i1 [[TMP12]])
 ; CHECK-NEXT:    [[TMP13:%.*]] = extractelement <2 x i1> [[TMP9]], i32 1
 ; CHECK-NEXT:    tail call void @llvm.assume(i1 [[TMP13]])
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 2
-; CHECK-NEXT:    store <2 x i8> zeroinitializer, ptr [[TMP14]], align 1
+; CHECK-NEXT:    store <2 x i8> zeroinitializer, ptr [[NEXT_GEP]], align 1
 ; CHECK-NEXT:    store <2 x i8> zeroinitializer, ptr [[TMP15]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10000
@@ -65,8 +64,7 @@ define void @test_widen_ptr_induction(ptr %ptr.start.1) {
 ; CHECK-NEXT:    tail call void @llvm.assume(i1 [[TMP22]])
 ; CHECK-NEXT:    [[TMP23:%.*]] = extractelement <2 x i1> [[TMP21]], i32 1
 ; CHECK-NEXT:    tail call void @llvm.assume(i1 [[TMP23]])
-; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr i8, ptr [[NEXT_GEP7]], i32 0
-; CHECK-NEXT:    store <2 x i8> zeroinitializer, ptr [[TMP24]], align 1
+; CHECK-NEXT:    store <2 x i8> zeroinitializer, ptr [[NEXT_GEP7]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT9]] = add nuw i64 [[INDEX6]], 2
 ; CHECK-NEXT:    [[TMP25:%.*]] = icmp eq i64 [[INDEX_NEXT9]], 10000
 ; CHECK-NEXT:    br i1 [[TMP25]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
@@ -126,9 +124,8 @@ define void @test_widen_induction(ptr %A, i64 %N) {
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[STEP_ADD:%.*]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 2
-; CHECK-NEXT:    store <2 x i64> [[VEC_IND]], ptr [[TMP2]], align 4
+; CHECK-NEXT:    store <2 x i64> [[VEC_IND]], ptr [[TMP1]], align 4
 ; CHECK-NEXT:    store <2 x i64> [[STEP_ADD]], ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[STEP_ADD]], splat (i64 2)
@@ -153,8 +150,7 @@ define void @test_widen_induction(ptr %A, i64 %N) {
 ; CHECK-NEXT:    [[INDEX5:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT8:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_IND6:%.*]] = phi <2 x i64> [ [[INDUCTION]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT7:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX5]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <2 x i64> [[VEC_IND6]], ptr [[TMP7]], align 4
+; CHECK-NEXT:    store <2 x i64> [[VEC_IND6]], ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT8]] = add nuw i64 [[INDEX5]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT7]] = add <2 x i64> [[VEC_IND6]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT8]], [[N_VEC3]]
@@ -213,9 +209,8 @@ define void @test_widen_induction_variable_start(ptr %A, i64 %N, i64 %start) {
 ; CHECK-NEXT:    [[STEP_ADD:%.*]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 [[START]], [[INDEX]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 2
-; CHECK-NEXT:    store <2 x i64> [[VEC_IND]], ptr [[TMP3]], align 4
+; CHECK-NEXT:    store <2 x i64> [[VEC_IND]], ptr [[TMP2]], align 4
 ; CHECK-NEXT:    store <2 x i64> [[STEP_ADD]], ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[STEP_ADD]], splat (i64 2)
@@ -244,8 +239,7 @@ define void @test_widen_induction_variable_start(ptr %A, i64 %N, i64 %start) {
 ; CHECK-NEXT:    [[VEC_IND11:%.*]] = phi <2 x i64> [ [[INDUCTION10]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT12:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX13:%.*]] = add i64 [[START]], [[INDEX7]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[OFFSET_IDX13]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <2 x i64> [[VEC_IND11]], ptr [[TMP8]], align 4
+; CHECK-NEXT:    store <2 x i64> [[VEC_IND11]], ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT14]] = add nuw i64 [[INDEX7]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT12]] = add <2 x i64> [[VEC_IND11]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT14]], [[N_VEC3]]
@@ -300,9 +294,8 @@ define void @test_widen_induction_step_2(ptr %A, i64 %N, i32 %step) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = add <2 x i64> [[VEC_IND]], splat (i64 10)
 ; CHECK-NEXT:    [[TMP3:%.*]] = add <2 x i64> [[STEP_ADD]], splat (i64 10)
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 2
-; CHECK-NEXT:    store <2 x i64> [[TMP2]], ptr [[TMP4]], align 4
+; CHECK-NEXT:    store <2 x i64> [[TMP2]], ptr [[TMP1]], align 4
 ; CHECK-NEXT:    store <2 x i64> [[TMP3]], ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[STEP_ADD]], splat (i64 2)
@@ -328,8 +321,7 @@ define void @test_widen_induction_step_2(ptr %A, i64 %N, i32 %step) {
 ; CHECK-NEXT:    [[VEC_IND8:%.*]] = phi <2 x i64> [ [[INDUCTION]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT9:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX7]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = add <2 x i64> [[VEC_IND8]], splat (i64 10)
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <2 x i64> [[TMP9]], ptr [[TMP10]], align 4
+; CHECK-NEXT:    store <2 x i64> [[TMP9]], ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT10]] = add nuw i64 [[INDEX7]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT9]] = add <2 x i64> [[VEC_IND8]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT10]], [[IND_END]]
@@ -374,66 +366,17 @@ exit:
 
 define void @test_widen_extended_induction(ptr %dst) {
 ; CHECK-LABEL: @test_widen_extended_induction(
-; CHECK-NEXT:  iter.check:
-; CHECK-NEXT:    br i1 false, label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_SCEVCHECK:%.*]]
-; CHECK:       vector.scevcheck:
-; CHECK-NEXT:    br i1 true, label [[VEC_EPILOG_SCALAR_PH]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
-; CHECK:       vector.main.loop.iter.check:
-; CHECK-NEXT:    br i1 false, label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
-; CHECK:       vector.ph:
-; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
-; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i8> [ <i8 0, i8 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[STEP_ADD:%.*]] = add <2 x i8> [[VEC_IND]], splat (i8 2)
-; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i8
-; CHECK-NEXT:    [[TMP1:%.*]] = zext i8 [[OFFSET_IDX]] to i64
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [6 x i8], ptr [[DST:%.*]], i64 0, i64 [[TMP1]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 2
-; CHECK-NEXT:    store <2 x i8> [[VEC_IND]], ptr [[TMP3]], align 1
-; CHECK-NEXT:    store <2 x i8> [[STEP_ADD]], ptr [[TMP4]], align 1
-; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
-; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i8> [[STEP_ADD]], splat (i8 2)
-; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], 10000
-; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
-; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
-; CHECK:       vec.epilog.iter.check:
-; CHECK-NEXT:    br i1 true, label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
-; CHECK:       vec.epilog.ph:
-; CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i32 [ 10000, [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 16, [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
-; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x i8> poison, i8 [[BC_RESUME_VAL]], i64 0
-; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x i8> [[DOTSPLATINSERT]], <2 x i8> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[INDUCTION:%.*]] = add <2 x i8> [[DOTSPLAT]], <i8 0, i8 1>
-; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
-; CHECK:       vec.epilog.vector.body:
-; CHECK-NEXT:    [[INDEX2:%.*]] = phi i32 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT6:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_IND3:%.*]] = phi <2 x i8> [ [[INDUCTION]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT4:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
-; CHECK-NEXT:    [[OFFSET_IDX5:%.*]] = trunc i32 [[INDEX2]] to i8
-; CHECK-NEXT:    [[TMP7:%.*]] = zext i8 [[OFFSET_IDX5]] to i64
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [6 x i8], ptr [[DST]], i64 0, i64 [[TMP7]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <2 x i8> [[VEC_IND3]], ptr [[TMP9]], align 1
-; CHECK-NEXT:    [[INDEX_NEXT6]] = add nuw i32 [[INDEX2]], 2
-; CHECK-NEXT:    [[VEC_IND_NEXT4]] = add <2 x i8> [[VEC_IND3]], splat (i8 2)
-; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[INDEX_NEXT6]], 10000
-; CHECK-NEXT:    br i1 [[TMP10]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
-; CHECK:       vec.epilog.middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
-; CHECK:       vec.epilog.scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i8 [ 16, [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ 16, [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_SCEVCHECK]] ], [ 0, [[ITER_CHECK:%.*]] ]
+; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
-; CHECK-NEXT:    [[IV:%.*]] = phi i8 [ [[BC_RESUME_VAL1]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
+; CHECK-NEXT:    [[IV:%.*]] = phi i8 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
 ; CHECK-NEXT:    [[IV_EXT:%.*]] = zext i8 [[IV]] to i64
-; CHECK-NEXT:    [[ARRAYIDX1449:%.*]] = getelementptr inbounds [6 x i8], ptr [[DST]], i64 0, i64 [[IV_EXT]]
+; CHECK-NEXT:    [[ARRAYIDX1449:%.*]] = getelementptr inbounds [6 x i8], ptr [[DST:%.*]], i64 0, i64 [[IV_EXT]]
 ; CHECK-NEXT:    store i8 [[IV]], ptr [[ARRAYIDX1449]], align 1
 ; CHECK-NEXT:    [[IV_NEXT]] = add i8 [[IV]], 1
 ; CHECK-NEXT:    [[IV_NEXT_EXT:%.*]] = zext i8 [[IV_NEXT]] to i32
 ; CHECK-NEXT:    [[EC:%.*]] = icmp eq i32 [[IV_NEXT_EXT]], 10000
-; CHECK-NEXT:    br i1 [[EC]], label [[EXIT]], label [[LOOP]], {{!llvm.loop ![0-9]+}}
+; CHECK-NEXT:    br i1 [[EC]], label [[EXIT:%.*]], label [[LOOP]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
@@ -467,9 +410,8 @@ define void @test_widen_truncated_induction(ptr %A) {
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i8> [ <i8 0, i8 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[STEP_ADD:%.*]] = add <2 x i8> [[VEC_IND]], splat (i8 2)
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 2
-; CHECK-NEXT:    store <2 x i8> [[VEC_IND]], ptr [[TMP2]], align 1
+; CHECK-NEXT:    store <2 x i8> [[VEC_IND]], ptr [[TMP1]], align 1
 ; CHECK-NEXT:    store <2 x i8> [[STEP_ADD]], ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i8> [[STEP_ADD]], splat (i8 2)
@@ -490,8 +432,7 @@ define void @test_widen_truncated_induction(ptr %A) {
 ; CHECK-NEXT:    [[INDEX2:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT5:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_IND3:%.*]] = phi <2 x i8> [ [[INDUCTION]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT4:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX2]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <2 x i8> [[VEC_IND3]], ptr [[TMP8]], align 1
+; CHECK-NEXT:    store <2 x i8> [[VEC_IND3]], ptr [[TMP7]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT5]] = add nuw i64 [[INDEX2]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT4]] = add <2 x i8> [[VEC_IND3]], splat (i8 2)
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT5]], 10000
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/extractvalue-no-scalarization-required.ll b/llvm/test/Transforms/LoopVectorize/AArch64/extractvalue-no-scalarization-required.ll
index 07c060a..19f2a36 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/extractvalue-no-scalarization-required.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/extractvalue-no-scalarization-required.ll
@@ -33,8 +33,7 @@
 ; FORCED-LABEL: vector.body:                                      ; preds = %vector.body, %vector.ph
 ; FORCED-NEXT:    %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
 ; FORCED-NEXT:    [[GEP:%.+]] = getelementptr i64, ptr %dst, i32 %index
-; FORCED-NEXT:    [[GEP2:%.+]] = getelementptr i64, ptr [[GEP]], i32 0
-; FORCED-NEXT:    store <2 x i64> [[ADD]], ptr [[GEP2]], align 4
+; FORCED-NEXT:    store <2 x i64> [[ADD]], ptr [[GEP]], align 4
 ; FORCED-NEXT:    %index.next = add nuw i32 %index, 2
 ; FORCED-NEXT:    [[C:%.+]] = icmp eq i32 %index.next, 1000
 ; FORCED-NEXT:    br i1 [[C]], label %middle.block, label %vector.body
@@ -84,8 +83,7 @@ declare float @powf(float, float) readnone nounwind
 ; FORCED-NEXT:    %index = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
 ; FORCED-NEXT:    [[GEP1:%.+]] = getelementptr float, ptr %dst, i32 %index
 ; FORCED-NEXT:    [[POW:%.+]] = call <2 x float> @llvm.pow.v2f32(<2 x float> %broadcast.splat, <2 x float> %broadcast.splat2)
-; FORCED-NEXT:    [[GEP2:%.+]] = getelementptr float, ptr [[GEP1]], i32 0
-; FORCED-NEXT:    store <2 x float> [[POW]], ptr [[GEP2]], align 4
+; FORCED-NEXT:    store <2 x float> [[POW]], ptr [[GEP1]], align 4
 ; FORCED-NEXT:    %index.next = add nuw i32 %index, 2
 ; FORCED-NEXT:    [[C:%.+]] = icmp eq i32 %index.next, 1000
 ; FORCED-NEXT:    br i1 [[C]], label %middle.block, label %vector.body
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/first-order-recurrence-fold-tail.ll b/llvm/test/Transforms/LoopVectorize/AArch64/first-order-recurrence-fold-tail.ll
index 30e454d..fff99f1 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/first-order-recurrence-fold-tail.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/first-order-recurrence-fold-tail.ll
@@ -61,8 +61,7 @@ define i32 @test_phi_iterator_invalidation(ptr %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP26:%.*]] = sext <4 x i16> [[TMP25]] to <4 x i32>
 ; CHECK-NEXT:    [[TMP27:%.*]] = extractelement <4 x i64> [[TMP0]], i32 0
 ; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr i32, ptr [[B:%.*]], i64 [[TMP27]]
-; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr i32, ptr [[TMP28]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[TMP26]], ptr [[TMP29]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[TMP26]], ptr [[TMP28]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i64(i64 [[INDEX_NEXT]], i64 1002)
 ; CHECK-NEXT:    [[TMP30:%.*]] = xor <4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/fixed-order-recurrence.ll b/llvm/test/Transforms/LoopVectorize/AArch64/fixed-order-recurrence.ll
index a6c5692..c94b3a4 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/fixed-order-recurrence.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/fixed-order-recurrence.ll
@@ -26,18 +26,16 @@ define void @firstorderrec(ptr nocapture noundef readonly %x, ptr noalias nocapt
 ; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <16 x i8> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD1:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[X]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 16
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD1]] = load <16 x i8>, ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <16 x i8> [[VECTOR_RECUR]], <16 x i8> [[WIDE_LOAD]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
 ; CHECK-NEXT:    [[TMP8:%.*]] = shufflevector <16 x i8> [[WIDE_LOAD]], <16 x i8> [[WIDE_LOAD1]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
 ; CHECK-NEXT:    [[TMP9:%.*]] = add <16 x i8> [[WIDE_LOAD]], [[TMP7]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = add <16 x i8> [[WIDE_LOAD1]], [[TMP8]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[Y:%.*]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
 ; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 16
-; CHECK-NEXT:    store <16 x i8> [[TMP9]], ptr [[TMP13]], align 1
+; CHECK-NEXT:    store <16 x i8> [[TMP9]], ptr [[TMP11]], align 1
 ; CHECK-NEXT:    store <16 x i8> [[TMP10]], ptr [[TMP14]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
 ; CHECK-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -121,9 +119,8 @@ define void @thirdorderrec(ptr nocapture noundef readonly %x, ptr noalias nocapt
 ; CHECK-NEXT:    [[VECTOR_RECUR4:%.*]] = phi <16 x i8> [ [[VECTOR_RECUR_INIT3]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[X]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 16
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD5]] = load <16 x i8>, ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <16 x i8> [[VECTOR_RECUR]], <16 x i8> [[WIDE_LOAD]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
 ; CHECK-NEXT:    [[TMP8]] = shufflevector <16 x i8> [[WIDE_LOAD]], <16 x i8> [[WIDE_LOAD5]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
@@ -138,9 +135,8 @@ define void @thirdorderrec(ptr nocapture noundef readonly %x, ptr noalias nocapt
 ; CHECK-NEXT:    [[TMP17:%.*]] = add <16 x i8> [[TMP15]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP18:%.*]] = add <16 x i8> [[TMP16]], [[WIDE_LOAD5]]
 ; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[Y:%.*]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i8, ptr [[TMP19]], i32 0
 ; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i8, ptr [[TMP19]], i32 16
-; CHECK-NEXT:    store <16 x i8> [[TMP17]], ptr [[TMP21]], align 1
+; CHECK-NEXT:    store <16 x i8> [[TMP17]], ptr [[TMP19]], align 1
 ; CHECK-NEXT:    store <16 x i8> [[TMP18]], ptr [[TMP22]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
 ; CHECK-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/fmax-without-fast-math-flags.ll b/llvm/test/Transforms/LoopVectorize/AArch64/fmax-without-fast-math-flags.ll
index 427a05c..5de9d0e 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/fmax-without-fast-math-flags.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/fmax-without-fast-math-flags.ll
@@ -53,9 +53,8 @@ define float @fmaxnum(ptr %src, i64 %n) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ splat (float -1.000000e+07), %[[VECTOR_PH]] ], [ [[TMP7:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x float> [ splat (float -1.000000e+07), %[[VECTOR_PH]] ], [ [[TMP8:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw float, ptr [[GEP_SRC]], i32 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw float, ptr [[GEP_SRC]], i32 4
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[GEP_SRC]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP7]] = call <4 x float> @llvm.maxnum.v4f32(<4 x float> [[VEC_PHI]], <4 x float> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[TMP8]] = call <4 x float> @llvm.maxnum.v4f32(<4 x float> [[VEC_PHI1]], <4 x float> [[WIDE_LOAD2]])
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/fmin-without-fast-math-flags.ll b/llvm/test/Transforms/LoopVectorize/AArch64/fmin-without-fast-math-flags.ll
index 1a8e594..ea44fc3 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/fmin-without-fast-math-flags.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/fmin-without-fast-math-flags.ll
@@ -53,9 +53,8 @@ define float @fminnum(ptr %src, i64 %n) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ splat (float -1.000000e+07), %[[VECTOR_PH]] ], [ [[TMP7:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x float> [ splat (float -1.000000e+07), %[[VECTOR_PH]] ], [ [[TMP8:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw float, ptr [[GEP_SRC]], i32 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw float, ptr [[GEP_SRC]], i32 4
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[GEP_SRC]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP7]] = call <4 x float> @llvm.minnum.v4f32(<4 x float> [[VEC_PHI]], <4 x float> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[TMP8]] = call <4 x float> @llvm.minnum.v4f32(<4 x float> [[VEC_PHI1]], <4 x float> [[WIDE_LOAD2]])
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/fminimumnum.ll b/llvm/test/Transforms/LoopVectorize/AArch64/fminimumnum.ll
index 403a5f1..2ed2819 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/fminimumnum.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/fminimumnum.ll
@@ -20,29 +20,26 @@ define void @fmin32(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw float, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw float, ptr [[TMP2]], i32 4
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw float, ptr [[TMP5]], i32 0
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw float, ptr [[TMP5]], i32 4
-; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP6]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x float>, ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = call <4 x float> @llvm.minimumnum.v4f32(<4 x float> [[WIDE_LOAD]], <4 x float> [[WIDE_LOAD6]])
 ; CHECK-NEXT:    [[TMP9:%.*]] = call <4 x float> @llvm.minimumnum.v4f32(<4 x float> [[WIDE_LOAD5]], <4 x float> [[WIDE_LOAD7]])
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw float, ptr [[TMP10]], i32 0
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw float, ptr [[TMP10]], i32 4
-; CHECK-NEXT:    store <4 x float> [[TMP8]], ptr [[TMP11]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP8]], ptr [[TMP10]], align 4
 ; CHECK-NEXT:    store <4 x float> [[TMP9]], ptr [[TMP12]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; CHECK-NEXT:    br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -101,29 +98,26 @@ define void @fmax32(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw float, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw float, ptr [[TMP2]], i32 4
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw float, ptr [[TMP5]], i32 0
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw float, ptr [[TMP5]], i32 4
-; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP6]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x float>, ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = call <4 x float> @llvm.maximumnum.v4f32(<4 x float> [[WIDE_LOAD]], <4 x float> [[WIDE_LOAD6]])
 ; CHECK-NEXT:    [[TMP9:%.*]] = call <4 x float> @llvm.maximumnum.v4f32(<4 x float> [[WIDE_LOAD5]], <4 x float> [[WIDE_LOAD7]])
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw float, ptr [[TMP10]], i32 0
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw float, ptr [[TMP10]], i32 4
-; CHECK-NEXT:    store <4 x float> [[TMP8]], ptr [[TMP11]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP8]], ptr [[TMP10]], align 4
 ; CHECK-NEXT:    store <4 x float> [[TMP9]], ptr [[TMP12]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; CHECK-NEXT:    br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -182,29 +176,26 @@ define void @fmin64(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw double, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw double, ptr [[TMP2]], i32 2
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP3]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP2]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <2 x double>, ptr [[TMP4]], align 8
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw double, ptr [[TMP5]], i32 0
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw double, ptr [[TMP5]], i32 2
-; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <2 x double>, ptr [[TMP6]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <2 x double>, ptr [[TMP5]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <2 x double>, ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[TMP8:%.*]] = call <2 x double> @llvm.minimumnum.v2f64(<2 x double> [[WIDE_LOAD]], <2 x double> [[WIDE_LOAD6]])
 ; CHECK-NEXT:    [[TMP9:%.*]] = call <2 x double> @llvm.minimumnum.v2f64(<2 x double> [[WIDE_LOAD5]], <2 x double> [[WIDE_LOAD7]])
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw double, ptr [[TMP10]], i32 0
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw double, ptr [[TMP10]], i32 2
-; CHECK-NEXT:    store <2 x double> [[TMP8]], ptr [[TMP11]], align 8
+; CHECK-NEXT:    store <2 x double> [[TMP8]], ptr [[TMP10]], align 8
 ; CHECK-NEXT:    store <2 x double> [[TMP9]], ptr [[TMP12]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; CHECK-NEXT:    br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -263,29 +254,26 @@ define void @fmax64(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw double, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw double, ptr [[TMP2]], i32 2
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP3]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP2]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <2 x double>, ptr [[TMP4]], align 8
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw double, ptr [[TMP5]], i32 0
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw double, ptr [[TMP5]], i32 2
-; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <2 x double>, ptr [[TMP6]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <2 x double>, ptr [[TMP5]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <2 x double>, ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[TMP8:%.*]] = call <2 x double> @llvm.maximumnum.v2f64(<2 x double> [[WIDE_LOAD]], <2 x double> [[WIDE_LOAD6]])
 ; CHECK-NEXT:    [[TMP9:%.*]] = call <2 x double> @llvm.maximumnum.v2f64(<2 x double> [[WIDE_LOAD5]], <2 x double> [[WIDE_LOAD7]])
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw double, ptr [[TMP10]], i32 0
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw double, ptr [[TMP10]], i32 2
-; CHECK-NEXT:    store <2 x double> [[TMP8]], ptr [[TMP11]], align 8
+; CHECK-NEXT:    store <2 x double> [[TMP8]], ptr [[TMP10]], align 8
 ; CHECK-NEXT:    store <2 x double> [[TMP9]], ptr [[TMP12]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; CHECK-NEXT:    br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -344,29 +332,26 @@ define void @fmin16(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw half, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw half, ptr [[TMP2]], i32 8
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x half>, ptr [[TMP3]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x half>, ptr [[TMP2]], align 2
 ; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <8 x half>, ptr [[TMP6]], align 2
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw half, ptr [[TMP4]], i32 0
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw half, ptr [[TMP4]], i32 8
-; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <8 x half>, ptr [[TMP5]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <8 x half>, ptr [[TMP4]], align 2
 ; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <8 x half>, ptr [[TMP10]], align 2
 ; CHECK-NEXT:    [[TMP11:%.*]] = call <8 x half> @llvm.minimumnum.v8f16(<8 x half> [[WIDE_LOAD]], <8 x half> [[WIDE_LOAD6]])
 ; CHECK-NEXT:    [[TMP13:%.*]] = call <8 x half> @llvm.minimumnum.v8f16(<8 x half> [[WIDE_LOAD5]], <8 x half> [[WIDE_LOAD7]])
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw half, ptr [[TMP7]], i32 0
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw half, ptr [[TMP7]], i32 8
-; CHECK-NEXT:    store <8 x half> [[TMP11]], ptr [[TMP8]], align 2
+; CHECK-NEXT:    store <8 x half> [[TMP11]], ptr [[TMP7]], align 2
 ; CHECK-NEXT:    store <8 x half> [[TMP13]], ptr [[TMP12]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; CHECK-NEXT:    br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -425,29 +410,26 @@ define void @fmax16(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw half, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw half, ptr [[TMP2]], i32 8
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x half>, ptr [[TMP3]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x half>, ptr [[TMP2]], align 2
 ; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <8 x half>, ptr [[TMP6]], align 2
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw half, ptr [[TMP4]], i32 0
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw half, ptr [[TMP4]], i32 8
-; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <8 x half>, ptr [[TMP5]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <8 x half>, ptr [[TMP4]], align 2
 ; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <8 x half>, ptr [[TMP10]], align 2
 ; CHECK-NEXT:    [[TMP11:%.*]] = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> [[WIDE_LOAD]], <8 x half> [[WIDE_LOAD6]])
 ; CHECK-NEXT:    [[TMP13:%.*]] = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> [[WIDE_LOAD5]], <8 x half> [[WIDE_LOAD7]])
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw half, ptr [[TMP7]], i32 0
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw half, ptr [[TMP7]], i32 8
-; CHECK-NEXT:    store <8 x half> [[TMP11]], ptr [[TMP8]], align 2
+; CHECK-NEXT:    store <8 x half> [[TMP11]], ptr [[TMP7]], align 2
 ; CHECK-NEXT:    store <8 x half> [[TMP13]], ptr [[TMP12]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; CHECK-NEXT:    br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/force-target-instruction-cost.ll b/llvm/test/Transforms/LoopVectorize/AArch64/force-target-instruction-cost.ll
index 095ac22..c9cef14 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/force-target-instruction-cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/force-target-instruction-cost.ll
@@ -19,11 +19,11 @@ define double @test_reduction_costs() {
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 2, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi double [ [[TMP0]], %[[MIDDLE_BLOCK]] ], [ 0.000000e+00, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX2:%.*]] = phi double [ [[TMP1]], %[[MIDDLE_BLOCK]] ], [ 0.000000e+00, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi double [ 0.000000e+00, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX2:%.*]] = phi double [ 0.000000e+00, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_1:.*]]
 ; CHECK:       [[LOOP_1]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_1]] ]
@@ -79,9 +79,8 @@ define void @test_iv_cost(ptr %ptr.start, i8 %a, i64 %b) {
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[NEXT_GEP1:%.*]] = getelementptr i8, ptr [[PTR_START]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[NEXT_GEP1]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[NEXT_GEP1]], i32 16
-; CHECK-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP6]], align 1
+; CHECK-NEXT:    store <16 x i8> zeroinitializer, ptr [[NEXT_GEP1]], align 1
 ; CHECK-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -103,16 +102,15 @@ define void @test_iv_cost(ptr %ptr.start, i8 %a, i64 %b) {
 ; CHECK-NEXT:    [[IND_END5:%.*]] = getelementptr i8, ptr [[PTR_START]], i64 [[N_VEC3]]
 ; CHECK-NEXT:    br label %[[VEC_EPILOG_VECTOR_BODY:.*]]
 ; CHECK:       [[VEC_EPILOG_VECTOR_BODY]]:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], %[[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT10:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
-; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[PTR_START]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
-; CHECK-NEXT:    store <4 x i8> zeroinitializer, ptr [[TMP2]], align 1
-; CHECK-NEXT:    [[INDEX_NEXT10]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT10]], [[N_VEC3]]
-; CHECK-NEXT:    br i1 [[TMP7]], label %[[VEC_EPILOG_MIDDLE_BLOCK:.*]], label %[[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    [[INDEX4:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], %[[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT6:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
+; CHECK-NEXT:    [[NEXT_GEP5:%.*]] = getelementptr i8, ptr [[PTR_START]], i64 [[INDEX4]]
+; CHECK-NEXT:    store <4 x i8> zeroinitializer, ptr [[NEXT_GEP5]], align 1
+; CHECK-NEXT:    [[INDEX_NEXT6]] = add nuw i64 [[INDEX4]], 4
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT6]], [[N_VEC3]]
+; CHECK-NEXT:    br i1 [[TMP5]], label %[[VEC_EPILOG_MIDDLE_BLOCK:.*]], label %[[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       [[VEC_EPILOG_MIDDLE_BLOCK]]:
-; CHECK-NEXT:    [[CMP_N11:%.*]] = icmp eq i64 [[START]], [[N_VEC3]]
-; CHECK-NEXT:    br i1 [[CMP_N11]], label %[[EXIT_LOOPEXIT]], label %[[VEC_EPILOG_SCALAR_PH]]
+; CHECK-NEXT:    [[CMP_N7:%.*]] = icmp eq i64 [[START]], [[N_VEC3]]
+; CHECK-NEXT:    br i1 [[CMP_N7]], label %[[EXIT_LOOPEXIT]], label %[[VEC_EPILOG_SCALAR_PH]]
 ; CHECK:       [[VEC_EPILOG_SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END1]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[IND_END]], %[[VEC_EPILOG_ITER_CHECK]] ], [ [[START]], %[[ITER_CHECK]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL8:%.*]] = phi ptr [ [[IND_END5]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[IND_END2]], %[[VEC_EPILOG_ITER_CHECK]] ], [ [[PTR_START]], %[[ITER_CHECK]] ]
@@ -212,8 +210,7 @@ define void @test_exit_branch_cost(ptr %dst, ptr noalias %x.ptr, ptr noalias %y.
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE55:.*]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i64, ptr [[X_PTR]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i64, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP6]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP4]], align 8
 ; CHECK-NEXT:    [[TMP47:%.*]] = icmp eq <2 x i64> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP5:%.*]] = xor <2 x i1> [[TMP47]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i1> [[TMP5]], i32 0
@@ -279,7 +276,7 @@ define void @test_exit_branch_cost(ptr %dst, ptr noalias %x.ptr, ptr noalias %y.
 ; CHECK-NEXT:    [[TMP46:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; CHECK-NEXT:    br i1 [[TMP46]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 64, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
@@ -318,7 +315,7 @@ define void @test_exit_branch_cost(ptr %dst, ptr noalias %x.ptr, ptr noalias %y.
 ; CHECK:       [[LOOP_LATCH]]:
 ; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; CHECK-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV]], 64
-; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP22:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP22:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/gather-do-not-vectorize-addressing.ll b/llvm/test/Transforms/LoopVectorize/AArch64/gather-do-not-vectorize-addressing.ll
index 8471555..596a2ed 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/gather-do-not-vectorize-addressing.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/gather-do-not-vectorize-addressing.ll
@@ -87,8 +87,7 @@ define dso_local double @test(ptr nocapture noundef readonly %data, ptr nocaptur
 ; SVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; SVE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x double> [ insertelement (<vscale x 2 x double> splat (double -0.000000e+00), double 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; SVE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[OFFSET:%.*]], i64 [[INDEX]]
-; SVE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i32>, ptr [[TMP6]], align 4
+; SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i32>, ptr [[TMP5]], align 4
 ; SVE-NEXT:    [[TMP7:%.*]] = sext <vscale x 2 x i32> [[WIDE_LOAD]] to <vscale x 2 x i64>
 ; SVE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds double, ptr [[DATA:%.*]], <vscale x 2 x i64> [[TMP7]]
 ; SVE-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 2 x double> @llvm.masked.gather.nxv2f64.nxv2p0(<vscale x 2 x ptr> [[TMP8]], i32 8, <vscale x 2 x i1> splat (i1 true), <vscale x 2 x double> poison)
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/induction-costs-sve.ll b/llvm/test/Transforms/LoopVectorize/AArch64/induction-costs-sve.ll
index dab1428..8b354d9 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/induction-costs-sve.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/induction-costs-sve.ll
@@ -37,11 +37,10 @@ define void @iv_casts(ptr %dst, ptr %src, i32 %x, i64 %N) #0 {
 ; DEFAULT:       [[VECTOR_BODY]]:
 ; DEFAULT-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; DEFAULT-NEXT:    [[TMP20:%.*]] = getelementptr i8, ptr [[SRC]], i64 [[INDEX]]
-; DEFAULT-NEXT:    [[TMP22:%.*]] = getelementptr i8, ptr [[TMP20]], i32 0
 ; DEFAULT-NEXT:    [[TMP23:%.*]] = call i64 @llvm.vscale.i64()
 ; DEFAULT-NEXT:    [[TMP24:%.*]] = mul nuw i64 [[TMP23]], 8
 ; DEFAULT-NEXT:    [[TMP25:%.*]] = getelementptr i8, ptr [[TMP20]], i64 [[TMP24]]
-; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP22]], align 1
+; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP20]], align 1
 ; DEFAULT-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x i8>, ptr [[TMP25]], align 1
 ; DEFAULT-NEXT:    [[TMP26:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i16>
 ; DEFAULT-NEXT:    [[TMP27:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD4]] to <vscale x 8 x i16>
@@ -56,11 +55,10 @@ define void @iv_casts(ptr %dst, ptr %src, i32 %x, i64 %N) #0 {
 ; DEFAULT-NEXT:    [[TMP36:%.*]] = trunc <vscale x 8 x i16> [[TMP34]] to <vscale x 8 x i8>
 ; DEFAULT-NEXT:    [[TMP37:%.*]] = trunc <vscale x 8 x i16> [[TMP35]] to <vscale x 8 x i8>
 ; DEFAULT-NEXT:    [[TMP38:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INDEX]]
-; DEFAULT-NEXT:    [[TMP40:%.*]] = getelementptr i8, ptr [[TMP38]], i32 0
 ; DEFAULT-NEXT:    [[TMP41:%.*]] = call i64 @llvm.vscale.i64()
 ; DEFAULT-NEXT:    [[TMP42:%.*]] = mul nuw i64 [[TMP41]], 8
 ; DEFAULT-NEXT:    [[TMP43:%.*]] = getelementptr i8, ptr [[TMP38]], i64 [[TMP42]]
-; DEFAULT-NEXT:    store <vscale x 8 x i8> [[TMP36]], ptr [[TMP40]], align 1
+; DEFAULT-NEXT:    store <vscale x 8 x i8> [[TMP36]], ptr [[TMP38]], align 1
 ; DEFAULT-NEXT:    store <vscale x 8 x i8> [[TMP37]], ptr [[TMP43]], align 1
 ; DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
 ; DEFAULT-NEXT:    [[TMP44:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -125,8 +123,7 @@ define void @iv_casts(ptr %dst, ptr %src, i32 %x, i64 %N) #0 {
 ; PRED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; PRED-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], %[[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; PRED-NEXT:    [[TMP18:%.*]] = getelementptr i8, ptr [[SRC]], i64 [[INDEX]]
-; PRED-NEXT:    [[TMP19:%.*]] = getelementptr i8, ptr [[TMP18]], i32 0
-; PRED-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr [[TMP19]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i8> poison)
+; PRED-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr [[TMP18]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i8> poison)
 ; PRED-NEXT:    [[TMP17:%.*]] = zext <vscale x 16 x i8> [[WIDE_MASKED_LOAD]] to <vscale x 16 x i16>
 ; PRED-NEXT:    [[TMP22:%.*]] = mul <vscale x 16 x i16> [[TMP17]], [[TMP16]]
 ; PRED-NEXT:    [[TMP24:%.*]] = zext <vscale x 16 x i8> [[WIDE_MASKED_LOAD]] to <vscale x 16 x i16>
@@ -134,8 +131,7 @@ define void @iv_casts(ptr %dst, ptr %src, i32 %x, i64 %N) #0 {
 ; PRED-NEXT:    [[TMP21:%.*]] = lshr <vscale x 16 x i16> [[TMP20]], splat (i16 1)
 ; PRED-NEXT:    [[TMP23:%.*]] = trunc <vscale x 16 x i16> [[TMP21]] to <vscale x 16 x i8>
 ; PRED-NEXT:    [[TMP26:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INDEX]]
-; PRED-NEXT:    [[TMP27:%.*]] = getelementptr i8, ptr [[TMP26]], i32 0
-; PRED-NEXT:    call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP23]], ptr [[TMP27]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
+; PRED-NEXT:    call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP23]], ptr [[TMP26]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
 ; PRED-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP10]]
 ; PRED-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[INDEX]], i64 [[TMP15]])
 ; PRED-NEXT:    [[TMP28:%.*]] = xor <vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/induction-costs.ll b/llvm/test/Transforms/LoopVectorize/AArch64/induction-costs.ll
index 767bc35..aa2ec2de 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/induction-costs.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/induction-costs.ll
@@ -25,9 +25,8 @@ define i32 @multi_exit_iv_uniform(i32 %a, i64 %N, ptr %dst) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP11:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i64, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[TMP5]], i32 0
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr i64, ptr [[TMP5]], i32 4
-; CHECK-NEXT:    store <4 x i64> [[TMP7]], ptr [[TMP8]], align 8
+; CHECK-NEXT:    store <4 x i64> [[TMP7]], ptr [[TMP5]], align 8
 ; CHECK-NEXT:    store <4 x i64> [[TMP7]], ptr [[TMP9]], align 8
 ; CHECK-NEXT:    [[TMP10]] = add <4 x i32> [[VEC_PHI]], splat (i32 -1)
 ; CHECK-NEXT:    [[TMP11]] = add <4 x i32> [[VEC_PHI1]], splat (i32 -1)
@@ -157,7 +156,6 @@ define i64 @int_and_pointer_iv(ptr %start, i32 %N) {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK-NEXT:    [[IND_END:%.*]] = getelementptr i8, ptr [[START]], i64 4000
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
@@ -170,13 +168,12 @@ define i64 @int_and_pointer_iv(ptr %start, i32 %N) {
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i64> [[TMP5]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i64> [[TMP5]], i32 2
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi ptr [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi ptr [ [[START]], [[ENTRY]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -224,9 +221,8 @@ define void @wide_truncated_iv(ptr %dst) {
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <8 x i8> [ <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7>, [[VECTOR_PH1]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[STEP_ADD:%.*]] = add <8 x i8> [[VEC_IND]], splat (i8 8)
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP2]], i32 8
-; CHECK-NEXT:    store <8 x i8> [[VEC_IND]], ptr [[TMP4]], align 1
+; CHECK-NEXT:    store <8 x i8> [[VEC_IND]], ptr [[TMP2]], align 1
 ; CHECK-NEXT:    store <8 x i8> [[STEP_ADD]], ptr [[TMP5]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <8 x i8> [[STEP_ADD]], splat (i8 8)
@@ -247,8 +243,7 @@ define void @wide_truncated_iv(ptr %dst) {
 ; CHECK-NEXT:    [[INDEX3:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT7:%.*]], [[LOOP]] ]
 ; CHECK-NEXT:    [[VEC_IND4:%.*]] = phi <8 x i8> [ [[INDUCTION]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT6:%.*]], [[LOOP]] ]
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INDEX3]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[TMP9]], i32 0
-; CHECK-NEXT:    store <8 x i8> [[VEC_IND4]], ptr [[TMP10]], align 1
+; CHECK-NEXT:    store <8 x i8> [[VEC_IND4]], ptr [[TMP9]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT7]] = add nuw i64 [[INDEX3]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT6]] = add <8 x i8> [[VEC_IND4]], splat (i8 8)
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT7]], 200
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/interleave-allocsize-not-equal-typesize.ll b/llvm/test/Transforms/LoopVectorize/AArch64/interleave-allocsize-not-equal-typesize.ll
index 79d7ab8..f92aa06 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/interleave-allocsize-not-equal-typesize.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/interleave-allocsize-not-equal-typesize.ll
@@ -35,10 +35,10 @@ define void @pr58722_load_interleave_group(ptr %src, ptr %dst) {
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i64 1
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i64 1
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i64 1
-; CHECK-NEXT:    [[TMP13:%.*]] = load i24, ptr [[TMP9]], align 4, !alias.scope !0
-; CHECK-NEXT:    [[TMP14:%.*]] = load i24, ptr [[TMP10]], align 4, !alias.scope !0
-; CHECK-NEXT:    [[TMP15:%.*]] = load i24, ptr [[TMP11]], align 4, !alias.scope !0
-; CHECK-NEXT:    [[TMP16:%.*]] = load i24, ptr [[TMP12]], align 4, !alias.scope !0
+; CHECK-NEXT:    [[TMP13:%.*]] = load i24, ptr [[TMP9]], align 4, !alias.scope [[META0:![0-9]+]]
+; CHECK-NEXT:    [[TMP14:%.*]] = load i24, ptr [[TMP10]], align 4, !alias.scope [[META0]]
+; CHECK-NEXT:    [[TMP15:%.*]] = load i24, ptr [[TMP11]], align 4, !alias.scope [[META0]]
+; CHECK-NEXT:    [[TMP16:%.*]] = load i24, ptr [[TMP12]], align 4, !alias.scope [[META0]]
 ; CHECK-NEXT:    [[TMP17:%.*]] = insertelement <4 x i24> poison, i24 [[TMP13]], i32 0
 ; CHECK-NEXT:    [[TMP18:%.*]] = insertelement <4 x i24> [[TMP17]], i24 [[TMP14]], i32 1
 ; CHECK-NEXT:    [[TMP19:%.*]] = insertelement <4 x i24> [[TMP18]], i24 [[TMP15]], i32 2
@@ -46,8 +46,7 @@ define void @pr58722_load_interleave_group(ptr %src, ptr %dst) {
 ; CHECK-NEXT:    [[TMP21:%.*]] = zext <4 x i24> [[TMP20]] to <4 x i32>
 ; CHECK-NEXT:    [[TMP22:%.*]] = add <4 x i32> [[STRIDED_VEC]], [[TMP21]]
 ; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, ptr [[TMP23]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP22]], ptr [[TMP24]], align 4, !alias.scope !3, !noalias !0
+; CHECK-NEXT:    store <4 x i32> [[TMP22]], ptr [[TMP23]], align 4, !alias.scope [[META3:![0-9]+]], !noalias [[META0]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP25:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10000
 ; CHECK-NEXT:    br i1 [[TMP25]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/interleave-with-gaps.ll b/llvm/test/Transforms/LoopVectorize/AArch64/interleave-with-gaps.ll
index 5b8acee..649be65 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/interleave-with-gaps.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/interleave-with-gaps.ll
@@ -32,8 +32,7 @@ define i64 @vector_loop_with_remaining_iterations(ptr %src, ptr noalias %dst, i3
 ; CHECK-NEXT:    [[TMP6:%.*]] = call <16 x i32> @llvm.umin.v16i32(<16 x i32> [[TMP2]], <16 x i32> [[TMP5]])
 ; CHECK-NEXT:    [[TMP7:%.*]] = call <16 x i32> @llvm.umin.v16i32(<16 x i32> [[TMP3]], <16 x i32> [[TMP6]])
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP9]], align 1
+; CHECK-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP8]], align 1
 ; CHECK-NEXT:    [[TMP10:%.*]] = zext <16 x i32> [[TMP7]] to <16 x i64>
 ; CHECK-NEXT:    [[TMP11]] = or <16 x i64> [[VEC_PHI]], [[TMP10]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
@@ -82,8 +81,7 @@ define i64 @vector_loop_with_remaining_iterations(ptr %src, ptr noalias %dst, i3
 ; CHECK-NEXT:    [[TMP30:%.*]] = call <vscale x 2 x i32> @llvm.umin.nxv2i32(<vscale x 2 x i32> [[TMP23]], <vscale x 2 x i32> [[TMP29]])
 ; CHECK-NEXT:    [[TMP31:%.*]] = call <vscale x 2 x i32> @llvm.umin.nxv2i32(<vscale x 2 x i32> [[TMP24]], <vscale x 2 x i32> [[TMP30]])
 ; CHECK-NEXT:    [[TMP32:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[INDEX7]]
-; CHECK-NEXT:    [[TMP33:%.*]] = getelementptr inbounds i8, ptr [[TMP32]], i32 0
-; CHECK-NEXT:    store <vscale x 2 x i8> zeroinitializer, ptr [[TMP33]], align 1
+; CHECK-NEXT:    store <vscale x 2 x i8> zeroinitializer, ptr [[TMP32]], align 1
 ; CHECK-NEXT:    [[TMP34:%.*]] = zext <vscale x 2 x i32> [[TMP31]] to <vscale x 2 x i64>
 ; CHECK-NEXT:    [[TMP35]] = or <vscale x 2 x i64> [[VEC_PHI8]], [[TMP34]]
 ; CHECK-NEXT:    [[INDEX_NEXT9]] = add nuw i64 [[INDEX7]], [[TMP21]]
@@ -172,8 +170,7 @@ define i64 @main_vector_loop_fixed_with_no_remaining_iterations(ptr %src, ptr no
 ; CHECK-NEXT:    [[TMP6:%.*]] = call <16 x i32> @llvm.umin.v16i32(<16 x i32> [[TMP2]], <16 x i32> [[TMP5]])
 ; CHECK-NEXT:    [[TMP7:%.*]] = call <16 x i32> @llvm.umin.v16i32(<16 x i32> [[TMP3]], <16 x i32> [[TMP6]])
 ; CHECK-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i8, ptr [[TMP26]], i32 0
-; CHECK-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP27]], align 1
+; CHECK-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP26]], align 1
 ; CHECK-NEXT:    [[TMP10:%.*]] = zext <16 x i32> [[TMP7]] to <16 x i64>
 ; CHECK-NEXT:    [[TMP11]] = or <16 x i64> [[VEC_PHI]], [[TMP10]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
@@ -222,8 +219,7 @@ define i64 @main_vector_loop_fixed_with_no_remaining_iterations(ptr %src, ptr no
 ; CHECK-NEXT:    [[TMP30:%.*]] = call <vscale x 2 x i32> @llvm.umin.nxv2i32(<vscale x 2 x i32> [[TMP23]], <vscale x 2 x i32> [[TMP29]])
 ; CHECK-NEXT:    [[TMP31:%.*]] = call <vscale x 2 x i32> @llvm.umin.nxv2i32(<vscale x 2 x i32> [[TMP24]], <vscale x 2 x i32> [[TMP30]])
 ; CHECK-NEXT:    [[TMP32:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[INDEX7]]
-; CHECK-NEXT:    [[TMP33:%.*]] = getelementptr inbounds i8, ptr [[TMP32]], i32 0
-; CHECK-NEXT:    store <vscale x 2 x i8> zeroinitializer, ptr [[TMP33]], align 1
+; CHECK-NEXT:    store <vscale x 2 x i8> zeroinitializer, ptr [[TMP32]], align 1
 ; CHECK-NEXT:    [[TMP34:%.*]] = zext <vscale x 2 x i32> [[TMP31]] to <vscale x 2 x i64>
 ; CHECK-NEXT:    [[TMP35]] = or <vscale x 2 x i64> [[VEC_PHI8]], [[TMP34]]
 ; CHECK-NEXT:    [[INDEX_NEXT9]] = add nuw i64 [[INDEX7]], [[TMP21]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/interleaving-load-store.ll b/llvm/test/Transforms/LoopVectorize/AArch64/interleaving-load-store.ll
index 890e13c..9bd3d30 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/interleaving-load-store.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/interleaving-load-store.ll
@@ -35,11 +35,10 @@ define void @interleave_single_load_store(ptr %src, ptr %dst, i64 %N, i8 %a, i8
 ; INTERLEAVE-4:       vector.body:
 ; INTERLEAVE-4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; INTERLEAVE-4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[INDEX]]
-; INTERLEAVE-4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 0
 ; INTERLEAVE-4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 16
 ; INTERLEAVE-4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 32
 ; INTERLEAVE-4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 48
-; INTERLEAVE-4-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP9]], align 1
+; INTERLEAVE-4-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
 ; INTERLEAVE-4-NEXT:    [[WIDE_LOAD4:%.*]] = load <16 x i8>, ptr [[TMP10]], align 1
 ; INTERLEAVE-4-NEXT:    [[WIDE_LOAD5:%.*]] = load <16 x i8>, ptr [[TMP11]], align 1
 ; INTERLEAVE-4-NEXT:    [[WIDE_LOAD6:%.*]] = load <16 x i8>, ptr [[TMP12]], align 1
@@ -56,11 +55,10 @@ define void @interleave_single_load_store(ptr %src, ptr %dst, i64 %N, i8 %a, i8
 ; INTERLEAVE-4-NEXT:    [[TMP23:%.*]] = select <16 x i1> [[TMP15]], <16 x i8> [[BROADCAST_SPLAT]], <16 x i8> [[TMP19]]
 ; INTERLEAVE-4-NEXT:    [[TMP24:%.*]] = select <16 x i1> [[TMP16]], <16 x i8> [[BROADCAST_SPLAT]], <16 x i8> [[TMP20]]
 ; INTERLEAVE-4-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[INDEX]]
-; INTERLEAVE-4-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i8, ptr [[TMP25]], i32 0
 ; INTERLEAVE-4-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i8, ptr [[TMP25]], i32 16
 ; INTERLEAVE-4-NEXT:    [[TMP31:%.*]] = getelementptr inbounds i8, ptr [[TMP25]], i32 32
 ; INTERLEAVE-4-NEXT:    [[TMP32:%.*]] = getelementptr inbounds i8, ptr [[TMP25]], i32 48
-; INTERLEAVE-4-NEXT:    store <16 x i8> [[TMP21]], ptr [[TMP29]], align 1
+; INTERLEAVE-4-NEXT:    store <16 x i8> [[TMP21]], ptr [[TMP25]], align 1
 ; INTERLEAVE-4-NEXT:    store <16 x i8> [[TMP22]], ptr [[TMP30]], align 1
 ; INTERLEAVE-4-NEXT:    store <16 x i8> [[TMP23]], ptr [[TMP31]], align 1
 ; INTERLEAVE-4-NEXT:    store <16 x i8> [[TMP24]], ptr [[TMP32]], align 1
@@ -86,14 +84,12 @@ define void @interleave_single_load_store(ptr %src, ptr %dst, i64 %N, i8 %a, i8
 ; INTERLEAVE-4:       vec.epilog.vector.body:
 ; INTERLEAVE-4-NEXT:    [[INDEX12:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT18:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; INTERLEAVE-4-NEXT:    [[TMP35:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[INDEX12]]
-; INTERLEAVE-4-NEXT:    [[TMP36:%.*]] = getelementptr inbounds i8, ptr [[TMP35]], i32 0
-; INTERLEAVE-4-NEXT:    [[WIDE_LOAD13:%.*]] = load <8 x i8>, ptr [[TMP36]], align 1
+; INTERLEAVE-4-NEXT:    [[WIDE_LOAD13:%.*]] = load <8 x i8>, ptr [[TMP35]], align 1
 ; INTERLEAVE-4-NEXT:    [[TMP37:%.*]] = icmp sgt <8 x i8> [[WIDE_LOAD13]], [[BROADCAST_SPLAT15]]
 ; INTERLEAVE-4-NEXT:    [[TMP38:%.*]] = call <8 x i8> @llvm.smax.v8i8(<8 x i8> [[WIDE_LOAD13]], <8 x i8> [[BROADCAST_SPLAT17]])
 ; INTERLEAVE-4-NEXT:    [[TMP39:%.*]] = select <8 x i1> [[TMP37]], <8 x i8> [[BROADCAST_SPLAT15]], <8 x i8> [[TMP38]]
 ; INTERLEAVE-4-NEXT:    [[TMP40:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[INDEX12]]
-; INTERLEAVE-4-NEXT:    [[TMP41:%.*]] = getelementptr inbounds i8, ptr [[TMP40]], i32 0
-; INTERLEAVE-4-NEXT:    store <8 x i8> [[TMP39]], ptr [[TMP41]], align 1
+; INTERLEAVE-4-NEXT:    store <8 x i8> [[TMP39]], ptr [[TMP40]], align 1
 ; INTERLEAVE-4-NEXT:    [[INDEX_NEXT18]] = add nuw i64 [[INDEX12]], 8
 ; INTERLEAVE-4-NEXT:    [[TMP42:%.*]] = icmp eq i64 [[INDEX_NEXT18]], [[N_VEC10]]
 ; INTERLEAVE-4-NEXT:    br i1 [[TMP42]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
@@ -142,9 +138,8 @@ define void @interleave_single_load_store(ptr %src, ptr %dst, i64 %N, i8 %a, i8
 ; INTERLEAVE-2:       vector.body:
 ; INTERLEAVE-2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; INTERLEAVE-2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[INDEX]]
-; INTERLEAVE-2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
 ; INTERLEAVE-2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 16
-; INTERLEAVE-2-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
+; INTERLEAVE-2-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; INTERLEAVE-2-NEXT:    [[WIDE_LOAD4:%.*]] = load <16 x i8>, ptr [[TMP6]], align 1
 ; INTERLEAVE-2-NEXT:    [[TMP7:%.*]] = icmp sgt <16 x i8> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
 ; INTERLEAVE-2-NEXT:    [[TMP8:%.*]] = icmp sgt <16 x i8> [[WIDE_LOAD4]], [[BROADCAST_SPLAT]]
@@ -153,9 +148,8 @@ define void @interleave_single_load_store(ptr %src, ptr %dst, i64 %N, i8 %a, i8
 ; INTERLEAVE-2-NEXT:    [[TMP11:%.*]] = select <16 x i1> [[TMP7]], <16 x i8> [[BROADCAST_SPLAT]], <16 x i8> [[TMP9]]
 ; INTERLEAVE-2-NEXT:    [[TMP12:%.*]] = select <16 x i1> [[TMP8]], <16 x i8> [[BROADCAST_SPLAT]], <16 x i8> [[TMP10]]
 ; INTERLEAVE-2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[INDEX]]
-; INTERLEAVE-2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 0
 ; INTERLEAVE-2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 16
-; INTERLEAVE-2-NEXT:    store <16 x i8> [[TMP11]], ptr [[TMP15]], align 1
+; INTERLEAVE-2-NEXT:    store <16 x i8> [[TMP11]], ptr [[TMP13]], align 1
 ; INTERLEAVE-2-NEXT:    store <16 x i8> [[TMP12]], ptr [[TMP16]], align 1
 ; INTERLEAVE-2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
 ; INTERLEAVE-2-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -179,14 +173,12 @@ define void @interleave_single_load_store(ptr %src, ptr %dst, i64 %N, i8 %a, i8
 ; INTERLEAVE-2:       vec.epilog.vector.body:
 ; INTERLEAVE-2-NEXT:    [[INDEX10:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT16:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; INTERLEAVE-2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[INDEX10]]
-; INTERLEAVE-2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i8, ptr [[TMP19]], i32 0
-; INTERLEAVE-2-NEXT:    [[WIDE_LOAD11:%.*]] = load <8 x i8>, ptr [[TMP20]], align 1
+; INTERLEAVE-2-NEXT:    [[WIDE_LOAD11:%.*]] = load <8 x i8>, ptr [[TMP19]], align 1
 ; INTERLEAVE-2-NEXT:    [[TMP21:%.*]] = icmp sgt <8 x i8> [[WIDE_LOAD11]], [[BROADCAST_SPLAT13]]
 ; INTERLEAVE-2-NEXT:    [[TMP22:%.*]] = call <8 x i8> @llvm.smax.v8i8(<8 x i8> [[WIDE_LOAD11]], <8 x i8> [[BROADCAST_SPLAT15]])
 ; INTERLEAVE-2-NEXT:    [[TMP23:%.*]] = select <8 x i1> [[TMP21]], <8 x i8> [[BROADCAST_SPLAT13]], <8 x i8> [[TMP22]]
 ; INTERLEAVE-2-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[INDEX10]]
-; INTERLEAVE-2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i8, ptr [[TMP24]], i32 0
-; INTERLEAVE-2-NEXT:    store <8 x i8> [[TMP23]], ptr [[TMP25]], align 1
+; INTERLEAVE-2-NEXT:    store <8 x i8> [[TMP23]], ptr [[TMP24]], align 1
 ; INTERLEAVE-2-NEXT:    [[INDEX_NEXT16]] = add nuw i64 [[INDEX10]], 8
 ; INTERLEAVE-2-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT16]], [[N_VEC8]]
 ; INTERLEAVE-2-NEXT:    br i1 [[TMP26]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/interleaving-reduction.ll b/llvm/test/Transforms/LoopVectorize/AArch64/interleaving-reduction.ll
index 449bcaa..f069347 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/interleaving-reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/interleaving-reduction.ll
@@ -29,11 +29,10 @@ define i32 @interleave_integer_reduction(ptr %src, i64 %N) {
 ; INTERLEAVE-4-NEXT:    [[VEC_PHI3:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
 ; INTERLEAVE-4-NEXT:    [[VEC_PHI4:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
 ; INTERLEAVE-4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i64 [[INDEX]]
-; INTERLEAVE-4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
 ; INTERLEAVE-4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 4
 ; INTERLEAVE-4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 8
 ; INTERLEAVE-4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 12
-; INTERLEAVE-4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP8]], align 1
+; INTERLEAVE-4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP4]], align 1
 ; INTERLEAVE-4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i32>, ptr [[TMP9]], align 1
 ; INTERLEAVE-4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i32>, ptr [[TMP10]], align 1
 ; INTERLEAVE-4-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i32>, ptr [[TMP11]], align 1
@@ -66,8 +65,7 @@ define i32 @interleave_integer_reduction(ptr %src, i64 %N) {
 ; INTERLEAVE-4-NEXT:    [[INDEX12:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT15:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; INTERLEAVE-4-NEXT:    [[VEC_PHI13:%.*]] = phi <4 x i32> [ [[TMP18]], [[VEC_EPILOG_PH]] ], [ [[TMP22:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; INTERLEAVE-4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[INDEX12]]
-; INTERLEAVE-4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, ptr [[TMP20]], i32 0
-; INTERLEAVE-4-NEXT:    [[WIDE_LOAD14:%.*]] = load <4 x i32>, ptr [[TMP21]], align 1
+; INTERLEAVE-4-NEXT:    [[WIDE_LOAD14:%.*]] = load <4 x i32>, ptr [[TMP20]], align 1
 ; INTERLEAVE-4-NEXT:    [[TMP22]] = add <4 x i32> [[VEC_PHI13]], [[WIDE_LOAD14]]
 ; INTERLEAVE-4-NEXT:    [[INDEX_NEXT15]] = add nuw i64 [[INDEX12]], 4
 ; INTERLEAVE-4-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT15]], [[N_VEC11]]
@@ -106,9 +104,8 @@ define i32 @interleave_integer_reduction(ptr %src, i64 %N) {
 ; INTERLEAVE-2-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
 ; INTERLEAVE-2-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; INTERLEAVE-2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i64 [[INDEX]]
-; INTERLEAVE-2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
 ; INTERLEAVE-2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 4
-; INTERLEAVE-2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP4]], align 1
+; INTERLEAVE-2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 1
 ; INTERLEAVE-2-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP5]], align 1
 ; INTERLEAVE-2-NEXT:    [[TMP6]] = add <4 x i32> [[VEC_PHI]], [[WIDE_LOAD]]
 ; INTERLEAVE-2-NEXT:    [[TMP7]] = add <4 x i32> [[VEC_PHI1]], [[WIDE_LOAD2]]
@@ -138,10 +135,69 @@ define i32 @interleave_integer_reduction(ptr %src, i64 %N) {
 ; INTERLEAVE-2-NEXT:    ret i32 [[RED_NEXT_LCSSA]]
 ;
 ; INTERLEAVE-4-VLA-LABEL: @interleave_integer_reduction(
-; INTERLEAVE-4-VLA:       add <vscale x 4 x i32>
-; INTERLEAVE-4-VLA-NEXT:  add <vscale x 4 x i32>
-; INTERLEAVE-4-VLA-NEXT:  add <vscale x 4 x i32>
-; INTERLEAVE-4-VLA-NEXT:  add <vscale x 4 x i32>
+; INTERLEAVE-4-VLA-NEXT:  entry:
+; INTERLEAVE-4-VLA-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; INTERLEAVE-4-VLA-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 16
+; INTERLEAVE-4-VLA-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP1]]
+; INTERLEAVE-4-VLA-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; INTERLEAVE-4-VLA:       vector.ph:
+; INTERLEAVE-4-VLA-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; INTERLEAVE-4-VLA-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16
+; INTERLEAVE-4-VLA-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
+; INTERLEAVE-4-VLA-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; INTERLEAVE-4-VLA-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; INTERLEAVE-4-VLA-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 16
+; INTERLEAVE-4-VLA-NEXT:    br label [[VECTOR_BODY:%.*]]
+; INTERLEAVE-4-VLA:       vector.body:
+; INTERLEAVE-4-VLA-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-4-VLA-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP16:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-4-VLA-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP17:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-4-VLA-NEXT:    [[VEC_PHI2:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-4-VLA-NEXT:    [[VEC_PHI3:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP19:%.*]], [[VECTOR_BODY]] ]
+; INTERLEAVE-4-VLA-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i64 [[INDEX]]
+; INTERLEAVE-4-VLA-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; INTERLEAVE-4-VLA-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4
+; INTERLEAVE-4-VLA-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i64 [[TMP8]]
+; INTERLEAVE-4-VLA-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; INTERLEAVE-4-VLA-NEXT:    [[TMP11:%.*]] = mul nuw i64 [[TMP10]], 8
+; INTERLEAVE-4-VLA-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i64 [[TMP11]]
+; INTERLEAVE-4-VLA-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; INTERLEAVE-4-VLA-NEXT:    [[TMP14:%.*]] = mul nuw i64 [[TMP13]], 12
+; INTERLEAVE-4-VLA-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i64 [[TMP14]]
+; INTERLEAVE-4-VLA-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP6]], align 1
+; INTERLEAVE-4-VLA-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 4 x i32>, ptr [[TMP9]], align 1
+; INTERLEAVE-4-VLA-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 4 x i32>, ptr [[TMP12]], align 1
+; INTERLEAVE-4-VLA-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 4 x i32>, ptr [[TMP15]], align 1
+; INTERLEAVE-4-VLA-NEXT:    [[TMP16]] = add <vscale x 4 x i32> [[VEC_PHI]], [[WIDE_LOAD]]
+; INTERLEAVE-4-VLA-NEXT:    [[TMP17]] = add <vscale x 4 x i32> [[VEC_PHI1]], [[WIDE_LOAD4]]
+; INTERLEAVE-4-VLA-NEXT:    [[TMP18]] = add <vscale x 4 x i32> [[VEC_PHI2]], [[WIDE_LOAD5]]
+; INTERLEAVE-4-VLA-NEXT:    [[TMP19]] = add <vscale x 4 x i32> [[VEC_PHI3]], [[WIDE_LOAD6]]
+; INTERLEAVE-4-VLA-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; INTERLEAVE-4-VLA-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; INTERLEAVE-4-VLA-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; INTERLEAVE-4-VLA:       middle.block:
+; INTERLEAVE-4-VLA-NEXT:    [[BIN_RDX:%.*]] = add <vscale x 4 x i32> [[TMP17]], [[TMP16]]
+; INTERLEAVE-4-VLA-NEXT:    [[BIN_RDX7:%.*]] = add <vscale x 4 x i32> [[TMP18]], [[BIN_RDX]]
+; INTERLEAVE-4-VLA-NEXT:    [[BIN_RDX8:%.*]] = add <vscale x 4 x i32> [[TMP19]], [[BIN_RDX7]]
+; INTERLEAVE-4-VLA-NEXT:    [[TMP21:%.*]] = call i32 @llvm.vector.reduce.add.nxv4i32(<vscale x 4 x i32> [[BIN_RDX8]])
+; INTERLEAVE-4-VLA-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; INTERLEAVE-4-VLA-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE-4-VLA:       scalar.ph:
+; INTERLEAVE-4-VLA-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; INTERLEAVE-4-VLA-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP21]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; INTERLEAVE-4-VLA-NEXT:    br label [[LOOP:%.*]]
+; INTERLEAVE-4-VLA:       loop:
+; INTERLEAVE-4-VLA-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
+; INTERLEAVE-4-VLA-NEXT:    [[RED:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[RED_NEXT:%.*]], [[LOOP]] ]
+; INTERLEAVE-4-VLA-NEXT:    [[GEP_SRC:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[IV]]
+; INTERLEAVE-4-VLA-NEXT:    [[L:%.*]] = load i32, ptr [[GEP_SRC]], align 1
+; INTERLEAVE-4-VLA-NEXT:    [[RED_NEXT]] = add i32 [[RED]], [[L]]
+; INTERLEAVE-4-VLA-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; INTERLEAVE-4-VLA-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; INTERLEAVE-4-VLA-NEXT:    br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
+; INTERLEAVE-4-VLA:       exit:
+; INTERLEAVE-4-VLA-NEXT:    [[RED_NEXT_LCSSA:%.*]] = phi i32 [ [[RED_NEXT]], [[LOOP]] ], [ [[TMP21]], [[MIDDLE_BLOCK]] ]
+; INTERLEAVE-4-VLA-NEXT:    ret i32 [[RED_NEXT_LCSSA]]
 ;
 entry:
   br label %loop
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/invariant-replicate-region.ll b/llvm/test/Transforms/LoopVectorize/AArch64/invariant-replicate-region.ll
index 26e96ca..d45dbcc 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/invariant-replicate-region.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/invariant-replicate-region.ll
@@ -50,9 +50,9 @@ define i32 @test_invariant_replicate_region(i32 %x, i1 %c) {
 ; CHECK-NEXT:    br i1 [[TMP16]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[TMP17:%.*]] = extractelement <4 x i32> [[PREDPHI]], i32 3
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 100, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/licm-calls.ll b/llvm/test/Transforms/LoopVectorize/AArch64/licm-calls.ll
index e6d5c54..8c4eba6 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/licm-calls.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/licm-calls.ll
@@ -23,16 +23,15 @@ define void @licm_replicate_call(double %x, ptr %dst) {
 ; CHECK-NEXT:    [[TMP6:%.*]] = fmul <2 x double> [[TMP3]], [[TMP4]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = fmul <2 x double> [[TMP3]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds double, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds double, ptr [[TMP8]], i32 0
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds double, ptr [[TMP8]], i32 2
-; CHECK-NEXT:    store <2 x double> [[TMP6]], ptr [[TMP9]], align 8
+; CHECK-NEXT:    store <2 x double> [[TMP6]], ptr [[TMP8]], align 8
 ; CHECK-NEXT:    store <2 x double> [[TMP7]], ptr [[TMP10]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], splat (i32 2)
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128
 ; CHECK-NEXT:    br i1 [[TMP11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 128, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
@@ -46,7 +45,7 @@ define void @licm_replicate_call(double %x, ptr %dst) {
 ; CHECK-NEXT:    store double [[MUL]], ptr [[GEP_DST]], align 8
 ; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; CHECK-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV]], 128
-; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/loop-vectorization-factors.ll b/llvm/test/Transforms/LoopVectorize/AArch64/loop-vectorization-factors.ll
index 57d5b43..5066a9b 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/loop-vectorization-factors.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/loop-vectorization-factors.ll
@@ -24,15 +24,13 @@ define void @add_a(ptr noalias nocapture readonly %p, ptr noalias nocapture %q,
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
-; CHECK-NEXT:    [[TMP3:%.*]] = add <16 x i8> [[WIDE_LOAD]], splat (i8 2)
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <16 x i8> [[TMP3]], ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEXT:    [[TMP2:%.*]] = add <16 x i8> [[WIDE_LOAD]], splat (i8 2)
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX]]
+; CHECK-NEXT:    store <16 x i8> [[TMP2]], ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
-; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
@@ -47,16 +45,14 @@ define void @add_a(ptr noalias nocapture readonly %p, ptr noalias nocapture %q,
 ; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
 ; CHECK:       vec.epilog.vector.body:
 ; CHECK-NEXT:    [[INDEX4:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT6:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDEX4]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i8>, ptr [[TMP8]], align 1
-; CHECK-NEXT:    [[TMP9:%.*]] = add <4 x i8> [[WIDE_LOAD5]], splat (i8 2)
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX4]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[TMP9]], ptr [[TMP11]], align 1
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDEX4]]
+; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i8>, ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[TMP6:%.*]] = add <4 x i8> [[WIDE_LOAD5]], splat (i8 2)
+; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX4]]
+; CHECK-NEXT:    store <4 x i8> [[TMP6]], ptr [[TMP7]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT6]] = add nuw i64 [[INDEX4]], 4
-; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT6]], [[N_VEC3]]
-; CHECK-NEXT:    br i1 [[TMP12]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT6]], [[N_VEC3]]
+; CHECK-NEXT:    br i1 [[TMP8]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       vec.epilog.middle.block:
 ; CHECK-NEXT:    [[CMP_N7:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC3]]
 ; CHECK-NEXT:    br i1 [[CMP_N7]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[VEC_EPILOG_SCALAR_PH]]
@@ -70,8 +66,8 @@ define void @add_a(ptr noalias nocapture readonly %p, ptr noalias nocapture %q,
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ]
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDVARS_IV]]
-; CHECK-NEXT:    [[TMP13:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
-; CHECK-NEXT:    [[CONV:%.*]] = zext i8 [[TMP13]] to i32
+; CHECK-NEXT:    [[TMP9:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
+; CHECK-NEXT:    [[CONV:%.*]] = zext i8 [[TMP9]] to i32
 ; CHECK-NEXT:    [[ADD:%.*]] = add nuw nsw i32 [[CONV]], 2
 ; CHECK-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i8
 ; CHECK-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDVARS_IV]]
@@ -125,15 +121,13 @@ define void @add_a1(ptr noalias nocapture readonly %p, ptr noalias nocapture %q,
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
-; CHECK-NEXT:    [[TMP3:%.*]] = add nuw nsw <16 x i8> [[WIDE_LOAD]], splat (i8 2)
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <16 x i8> [[TMP3]], ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEXT:    [[TMP2:%.*]] = add nuw nsw <16 x i8> [[WIDE_LOAD]], splat (i8 2)
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX]]
+; CHECK-NEXT:    store <16 x i8> [[TMP2]], ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
-; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
@@ -148,16 +142,14 @@ define void @add_a1(ptr noalias nocapture readonly %p, ptr noalias nocapture %q,
 ; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
 ; CHECK:       vec.epilog.vector.body:
 ; CHECK-NEXT:    [[INDEX4:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT6:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDEX4]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i8>, ptr [[TMP8]], align 1
-; CHECK-NEXT:    [[TMP9:%.*]] = add nuw nsw <4 x i8> [[WIDE_LOAD5]], splat (i8 2)
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX4]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[TMP9]], ptr [[TMP11]], align 1
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDEX4]]
+; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i8>, ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[TMP6:%.*]] = add nuw nsw <4 x i8> [[WIDE_LOAD5]], splat (i8 2)
+; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX4]]
+; CHECK-NEXT:    store <4 x i8> [[TMP6]], ptr [[TMP7]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT6]] = add nuw i64 [[INDEX4]], 4
-; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT6]], [[N_VEC3]]
-; CHECK-NEXT:    br i1 [[TMP12]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT6]], [[N_VEC3]]
+; CHECK-NEXT:    br i1 [[TMP8]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       vec.epilog.middle.block:
 ; CHECK-NEXT:    [[CMP_N7:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC3]]
 ; CHECK-NEXT:    br i1 [[CMP_N7]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[VEC_EPILOG_SCALAR_PH]]
@@ -171,8 +163,8 @@ define void @add_a1(ptr noalias nocapture readonly %p, ptr noalias nocapture %q,
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ]
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDVARS_IV]]
-; CHECK-NEXT:    [[TMP13:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
-; CHECK-NEXT:    [[ADD:%.*]] = add nuw nsw i8 [[TMP13]], 2
+; CHECK-NEXT:    [[TMP9:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
+; CHECK-NEXT:    [[ADD:%.*]] = add nuw nsw i8 [[TMP9]], 2
 ; CHECK-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDVARS_IV]]
 ; CHECK-NEXT:    store i8 [[ADD]], ptr [[ARRAYIDX3]], align 1
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
@@ -217,15 +209,13 @@ define void @add_b(ptr noalias nocapture readonly %p, ptr noalias nocapture %q,
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[P]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP2]], align 2
-; CHECK-NEXT:    [[TMP3:%.*]] = add <8 x i16> [[WIDE_LOAD]], splat (i16 2)
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, ptr [[Q]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <8 x i16> [[TMP3]], ptr [[TMP5]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP1]], align 2
+; CHECK-NEXT:    [[TMP2:%.*]] = add <8 x i16> [[WIDE_LOAD]], splat (i16 2)
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, ptr [[Q]], i64 [[INDEX]]
+; CHECK-NEXT:    store <8 x i16> [[TMP2]], ptr [[TMP3]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
-; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]
@@ -239,8 +229,8 @@ define void @add_b(ptr noalias nocapture readonly %p, ptr noalias nocapture %q,
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, ptr [[P]], i64 [[INDVARS_IV]]
-; CHECK-NEXT:    [[TMP7:%.*]] = load i16, ptr [[ARRAYIDX]], align 2
-; CHECK-NEXT:    [[CONV8:%.*]] = zext i16 [[TMP7]] to i32
+; CHECK-NEXT:    [[TMP5:%.*]] = load i16, ptr [[ARRAYIDX]], align 2
+; CHECK-NEXT:    [[CONV8:%.*]] = zext i16 [[TMP5]] to i32
 ; CHECK-NEXT:    [[ADD:%.*]] = add nuw nsw i32 [[CONV8]], 2
 ; CHECK-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
 ; CHECK-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, ptr [[Q]], i64 [[INDVARS_IV]]
@@ -292,16 +282,14 @@ define void @add_c(ptr noalias nocapture readonly %p, ptr noalias nocapture %q,
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
-; CHECK-NEXT:    [[TMP3:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i16>
-; CHECK-NEXT:    [[TMP4:%.*]] = add <16 x i16> [[TMP3]], splat (i16 2)
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[Q]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <16 x i16> [[TMP4]], ptr [[TMP6]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEXT:    [[TMP2:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i16>
+; CHECK-NEXT:    [[TMP3:%.*]] = add <16 x i16> [[TMP2]], splat (i16 2)
+; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, ptr [[Q]], i64 [[INDEX]]
+; CHECK-NEXT:    store <16 x i16> [[TMP3]], ptr [[TMP4]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
-; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
@@ -316,17 +304,15 @@ define void @add_c(ptr noalias nocapture readonly %p, ptr noalias nocapture %q,
 ; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
 ; CHECK:       vec.epilog.vector.body:
 ; CHECK-NEXT:    [[INDEX4:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT6:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDEX4]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i8>, ptr [[TMP9]], align 1
-; CHECK-NEXT:    [[TMP10:%.*]] = zext <4 x i8> [[WIDE_LOAD5]] to <4 x i16>
-; CHECK-NEXT:    [[TMP11:%.*]] = add <4 x i16> [[TMP10]], splat (i16 2)
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i16, ptr [[Q]], i64 [[INDEX4]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i16, ptr [[TMP12]], i32 0
-; CHECK-NEXT:    store <4 x i16> [[TMP11]], ptr [[TMP13]], align 2
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDEX4]]
+; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i8>, ptr [[TMP6]], align 1
+; CHECK-NEXT:    [[TMP7:%.*]] = zext <4 x i8> [[WIDE_LOAD5]] to <4 x i16>
+; CHECK-NEXT:    [[TMP8:%.*]] = add <4 x i16> [[TMP7]], splat (i16 2)
+; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i16, ptr [[Q]], i64 [[INDEX4]]
+; CHECK-NEXT:    store <4 x i16> [[TMP8]], ptr [[TMP9]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT6]] = add nuw i64 [[INDEX4]], 4
-; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT6]], [[N_VEC3]]
-; CHECK-NEXT:    br i1 [[TMP14]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
+; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT6]], [[N_VEC3]]
+; CHECK-NEXT:    br i1 [[TMP10]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ; CHECK:       vec.epilog.middle.block:
 ; CHECK-NEXT:    [[CMP_N7:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC3]]
 ; CHECK-NEXT:    br i1 [[CMP_N7]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[VEC_EPILOG_SCALAR_PH]]
@@ -340,8 +326,8 @@ define void @add_c(ptr noalias nocapture readonly %p, ptr noalias nocapture %q,
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ]
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDVARS_IV]]
-; CHECK-NEXT:    [[TMP15:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
-; CHECK-NEXT:    [[CONV:%.*]] = zext i8 [[TMP15]] to i32
+; CHECK-NEXT:    [[TMP11:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
+; CHECK-NEXT:    [[CONV:%.*]] = zext i8 [[TMP11]] to i32
 ; CHECK-NEXT:    [[ADD:%.*]] = add nuw nsw i32 [[CONV]], 2
 ; CHECK-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i16
 ; CHECK-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i16, ptr [[Q]], i64 [[INDVARS_IV]]
@@ -390,16 +376,14 @@ define void @add_d(ptr noalias nocapture readonly %p, ptr noalias nocapture %q,
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[P]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP2]], align 2
-; CHECK-NEXT:    [[TMP3:%.*]] = sext <8 x i16> [[WIDE_LOAD]] to <8 x i32>
-; CHECK-NEXT:    [[TMP4:%.*]] = add nsw <8 x i32> [[TMP3]], splat (i32 2)
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[Q]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <8 x i32> [[TMP4]], ptr [[TMP6]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP1]], align 2
+; CHECK-NEXT:    [[TMP2:%.*]] = sext <8 x i16> [[WIDE_LOAD]] to <8 x i32>
+; CHECK-NEXT:    [[TMP3:%.*]] = add nsw <8 x i32> [[TMP2]], splat (i32 2)
+; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[Q]], i64 [[INDEX]]
+; CHECK-NEXT:    store <8 x i32> [[TMP3]], ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
-; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]
@@ -413,8 +397,8 @@ define void @add_d(ptr noalias nocapture readonly %p, ptr noalias nocapture %q,
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, ptr [[P]], i64 [[INDVARS_IV]]
-; CHECK-NEXT:    [[TMP8:%.*]] = load i16, ptr [[ARRAYIDX]], align 2
-; CHECK-NEXT:    [[CONV:%.*]] = sext i16 [[TMP8]] to i32
+; CHECK-NEXT:    [[TMP6:%.*]] = load i16, ptr [[ARRAYIDX]], align 2
+; CHECK-NEXT:    [[CONV:%.*]] = sext i16 [[TMP6]] to i32
 ; CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 [[CONV]], 2
 ; CHECK-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds i32, ptr [[Q]], i64 [[INDVARS_IV]]
 ; CHECK-NEXT:    store i32 [[ADD]], ptr [[ARRAYIDX2]], align 4
@@ -470,22 +454,20 @@ define void @add_e(ptr noalias nocapture readonly %p, ptr noalias nocapture %q,
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
-; CHECK-NEXT:    [[TMP3:%.*]] = shl <16 x i8> [[WIDE_LOAD]], splat (i8 4)
-; CHECK-NEXT:    [[TMP4:%.*]] = add <16 x i8> [[TMP3]], splat (i8 32)
-; CHECK-NEXT:    [[TMP5:%.*]] = or <16 x i8> [[WIDE_LOAD]], splat (i8 51)
-; CHECK-NEXT:    [[TMP6:%.*]] = mul <16 x i8> [[TMP5]], splat (i8 60)
-; CHECK-NEXT:    [[TMP7:%.*]] = and <16 x i8> [[TMP4]], [[BROADCAST_SPLAT]]
-; CHECK-NEXT:    [[TMP8:%.*]] = and <16 x i8> [[TMP6]], splat (i8 -4)
-; CHECK-NEXT:    [[TMP9:%.*]] = xor <16 x i8> [[TMP8]], [[BROADCAST_SPLAT3]]
-; CHECK-NEXT:    [[TMP10:%.*]] = mul <16 x i8> [[TMP9]], [[TMP7]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    store <16 x i8> [[TMP10]], ptr [[TMP12]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEXT:    [[TMP2:%.*]] = shl <16 x i8> [[WIDE_LOAD]], splat (i8 4)
+; CHECK-NEXT:    [[TMP3:%.*]] = add <16 x i8> [[TMP2]], splat (i8 32)
+; CHECK-NEXT:    [[TMP4:%.*]] = or <16 x i8> [[WIDE_LOAD]], splat (i8 51)
+; CHECK-NEXT:    [[TMP5:%.*]] = mul <16 x i8> [[TMP4]], splat (i8 60)
+; CHECK-NEXT:    [[TMP6:%.*]] = and <16 x i8> [[TMP3]], [[BROADCAST_SPLAT]]
+; CHECK-NEXT:    [[TMP7:%.*]] = and <16 x i8> [[TMP5]], splat (i8 -4)
+; CHECK-NEXT:    [[TMP8:%.*]] = xor <16 x i8> [[TMP7]], [[BROADCAST_SPLAT3]]
+; CHECK-NEXT:    [[TMP9:%.*]] = mul <16 x i8> [[TMP8]], [[TMP6]]
+; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX]]
+; CHECK-NEXT:    store <16 x i8> [[TMP9]], ptr [[TMP10]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
-; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
+; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
@@ -504,23 +486,21 @@ define void @add_e(ptr noalias nocapture readonly %p, ptr noalias nocapture %q,
 ; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
 ; CHECK:       vec.epilog.vector.body:
 ; CHECK-NEXT:    [[INDEX10:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT12:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDEX10]]
-; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[TMP14]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x i8>, ptr [[TMP15]], align 1
-; CHECK-NEXT:    [[TMP16:%.*]] = shl <4 x i8> [[WIDE_LOAD11]], splat (i8 4)
-; CHECK-NEXT:    [[TMP17:%.*]] = add <4 x i8> [[TMP16]], splat (i8 32)
-; CHECK-NEXT:    [[TMP18:%.*]] = or <4 x i8> [[WIDE_LOAD11]], splat (i8 51)
-; CHECK-NEXT:    [[TMP19:%.*]] = mul <4 x i8> [[TMP18]], splat (i8 60)
-; CHECK-NEXT:    [[TMP20:%.*]] = and <4 x i8> [[TMP17]], [[BROADCAST_SPLAT7]]
-; CHECK-NEXT:    [[TMP21:%.*]] = and <4 x i8> [[TMP19]], splat (i8 -4)
-; CHECK-NEXT:    [[TMP22:%.*]] = xor <4 x i8> [[TMP21]], [[BROADCAST_SPLAT9]]
-; CHECK-NEXT:    [[TMP23:%.*]] = mul <4 x i8> [[TMP22]], [[TMP20]]
-; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX10]]
-; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i8, ptr [[TMP24]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[TMP23]], ptr [[TMP25]], align 1
+; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDEX10]]
+; CHECK-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x i8>, ptr [[TMP12]], align 1
+; CHECK-NEXT:    [[TMP13:%.*]] = shl <4 x i8> [[WIDE_LOAD11]], splat (i8 4)
+; CHECK-NEXT:    [[TMP14:%.*]] = add <4 x i8> [[TMP13]], splat (i8 32)
+; CHECK-NEXT:    [[TMP15:%.*]] = or <4 x i8> [[WIDE_LOAD11]], splat (i8 51)
+; CHECK-NEXT:    [[TMP16:%.*]] = mul <4 x i8> [[TMP15]], splat (i8 60)
+; CHECK-NEXT:    [[TMP17:%.*]] = and <4 x i8> [[TMP14]], [[BROADCAST_SPLAT7]]
+; CHECK-NEXT:    [[TMP18:%.*]] = and <4 x i8> [[TMP16]], splat (i8 -4)
+; CHECK-NEXT:    [[TMP19:%.*]] = xor <4 x i8> [[TMP18]], [[BROADCAST_SPLAT9]]
+; CHECK-NEXT:    [[TMP20:%.*]] = mul <4 x i8> [[TMP19]], [[TMP17]]
+; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX10]]
+; CHECK-NEXT:    store <4 x i8> [[TMP20]], ptr [[TMP21]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT12]] = add nuw i64 [[INDEX10]], 4
-; CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT12]], [[N_VEC5]]
-; CHECK-NEXT:    br i1 [[TMP26]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT12]], [[N_VEC5]]
+; CHECK-NEXT:    br i1 [[TMP22]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; CHECK:       vec.epilog.middle.block:
 ; CHECK-NEXT:    [[CMP_N13:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC5]]
 ; CHECK-NEXT:    br i1 [[CMP_N13]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[VEC_EPILOG_SCALAR_PH]]
@@ -534,8 +514,8 @@ define void @add_e(ptr noalias nocapture readonly %p, ptr noalias nocapture %q,
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDVARS_IV]]
-; CHECK-NEXT:    [[TMP27:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
-; CHECK-NEXT:    [[CONV:%.*]] = zext i8 [[TMP27]] to i32
+; CHECK-NEXT:    [[TMP23:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
+; CHECK-NEXT:    [[CONV:%.*]] = zext i8 [[TMP23]] to i32
 ; CHECK-NEXT:    [[ADD:%.*]] = shl i32 [[CONV]], 4
 ; CHECK-NEXT:    [[CONV2:%.*]] = add nuw nsw i32 [[ADD]], 32
 ; CHECK-NEXT:    [[OR:%.*]] = or i32 [[CONV]], 51
@@ -612,24 +592,22 @@ define void @add_f(ptr noalias nocapture readonly %p, ptr noalias nocapture %q,
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[P]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i16>, ptr [[TMP2]], align 2
-; CHECK-NEXT:    [[TMP3:%.*]] = trunc <16 x i16> [[WIDE_LOAD]] to <16 x i8>
-; CHECK-NEXT:    [[TMP4:%.*]] = shl <16 x i8> [[TMP3]], splat (i8 4)
-; CHECK-NEXT:    [[TMP5:%.*]] = add <16 x i8> [[TMP4]], splat (i8 32)
-; CHECK-NEXT:    [[TMP6:%.*]] = and <16 x i8> [[TMP3]], splat (i8 -52)
-; CHECK-NEXT:    [[TMP7:%.*]] = or <16 x i8> [[TMP6]], splat (i8 51)
-; CHECK-NEXT:    [[TMP8:%.*]] = mul <16 x i8> [[TMP7]], splat (i8 60)
-; CHECK-NEXT:    [[TMP9:%.*]] = and <16 x i8> [[TMP5]], [[BROADCAST_SPLAT]]
-; CHECK-NEXT:    [[TMP10:%.*]] = and <16 x i8> [[TMP8]], splat (i8 -4)
-; CHECK-NEXT:    [[TMP11:%.*]] = xor <16 x i8> [[TMP10]], [[BROADCAST_SPLAT3]]
-; CHECK-NEXT:    [[TMP12:%.*]] = mul <16 x i8> [[TMP11]], [[TMP9]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 0
-; CHECK-NEXT:    store <16 x i8> [[TMP12]], ptr [[TMP14]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i16>, ptr [[TMP1]], align 2
+; CHECK-NEXT:    [[TMP2:%.*]] = trunc <16 x i16> [[WIDE_LOAD]] to <16 x i8>
+; CHECK-NEXT:    [[TMP3:%.*]] = shl <16 x i8> [[TMP2]], splat (i8 4)
+; CHECK-NEXT:    [[TMP4:%.*]] = add <16 x i8> [[TMP3]], splat (i8 32)
+; CHECK-NEXT:    [[TMP5:%.*]] = and <16 x i8> [[TMP2]], splat (i8 -52)
+; CHECK-NEXT:    [[TMP6:%.*]] = or <16 x i8> [[TMP5]], splat (i8 51)
+; CHECK-NEXT:    [[TMP7:%.*]] = mul <16 x i8> [[TMP6]], splat (i8 60)
+; CHECK-NEXT:    [[TMP8:%.*]] = and <16 x i8> [[TMP4]], [[BROADCAST_SPLAT]]
+; CHECK-NEXT:    [[TMP9:%.*]] = and <16 x i8> [[TMP7]], splat (i8 -4)
+; CHECK-NEXT:    [[TMP10:%.*]] = xor <16 x i8> [[TMP9]], [[BROADCAST_SPLAT3]]
+; CHECK-NEXT:    [[TMP11:%.*]] = mul <16 x i8> [[TMP10]], [[TMP8]]
+; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX]]
+; CHECK-NEXT:    store <16 x i8> [[TMP11]], ptr [[TMP12]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
-; CHECK-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
@@ -648,25 +626,23 @@ define void @add_f(ptr noalias nocapture readonly %p, ptr noalias nocapture %q,
 ; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
 ; CHECK:       vec.epilog.vector.body:
 ; CHECK-NEXT:    [[INDEX10:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT12:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i16, ptr [[P]], i64 [[INDEX10]]
-; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i16, ptr [[TMP16]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x i16>, ptr [[TMP17]], align 2
-; CHECK-NEXT:    [[TMP18:%.*]] = trunc <4 x i16> [[WIDE_LOAD11]] to <4 x i8>
-; CHECK-NEXT:    [[TMP19:%.*]] = shl <4 x i8> [[TMP18]], splat (i8 4)
-; CHECK-NEXT:    [[TMP20:%.*]] = add <4 x i8> [[TMP19]], splat (i8 32)
-; CHECK-NEXT:    [[TMP21:%.*]] = and <4 x i8> [[TMP18]], splat (i8 -52)
-; CHECK-NEXT:    [[TMP22:%.*]] = or <4 x i8> [[TMP21]], splat (i8 51)
-; CHECK-NEXT:    [[TMP23:%.*]] = mul <4 x i8> [[TMP22]], splat (i8 60)
-; CHECK-NEXT:    [[TMP24:%.*]] = and <4 x i8> [[TMP20]], [[BROADCAST_SPLAT7]]
-; CHECK-NEXT:    [[TMP25:%.*]] = and <4 x i8> [[TMP23]], splat (i8 -4)
-; CHECK-NEXT:    [[TMP26:%.*]] = xor <4 x i8> [[TMP25]], [[BROADCAST_SPLAT9]]
-; CHECK-NEXT:    [[TMP27:%.*]] = mul <4 x i8> [[TMP26]], [[TMP24]]
-; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX10]]
-; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i8, ptr [[TMP28]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[TMP27]], ptr [[TMP29]], align 1
+; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i16, ptr [[P]], i64 [[INDEX10]]
+; CHECK-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x i16>, ptr [[TMP14]], align 2
+; CHECK-NEXT:    [[TMP15:%.*]] = trunc <4 x i16> [[WIDE_LOAD11]] to <4 x i8>
+; CHECK-NEXT:    [[TMP16:%.*]] = shl <4 x i8> [[TMP15]], splat (i8 4)
+; CHECK-NEXT:    [[TMP17:%.*]] = add <4 x i8> [[TMP16]], splat (i8 32)
+; CHECK-NEXT:    [[TMP18:%.*]] = and <4 x i8> [[TMP15]], splat (i8 -52)
+; CHECK-NEXT:    [[TMP19:%.*]] = or <4 x i8> [[TMP18]], splat (i8 51)
+; CHECK-NEXT:    [[TMP20:%.*]] = mul <4 x i8> [[TMP19]], splat (i8 60)
+; CHECK-NEXT:    [[TMP21:%.*]] = and <4 x i8> [[TMP17]], [[BROADCAST_SPLAT7]]
+; CHECK-NEXT:    [[TMP22:%.*]] = and <4 x i8> [[TMP20]], splat (i8 -4)
+; CHECK-NEXT:    [[TMP23:%.*]] = xor <4 x i8> [[TMP22]], [[BROADCAST_SPLAT9]]
+; CHECK-NEXT:    [[TMP24:%.*]] = mul <4 x i8> [[TMP23]], [[TMP21]]
+; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX10]]
+; CHECK-NEXT:    store <4 x i8> [[TMP24]], ptr [[TMP25]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT12]] = add nuw i64 [[INDEX10]], 4
-; CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT12]], [[N_VEC5]]
-; CHECK-NEXT:    br i1 [[TMP30]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
+; CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT12]], [[N_VEC5]]
+; CHECK-NEXT:    br i1 [[TMP26]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
 ; CHECK:       vec.epilog.middle.block:
 ; CHECK-NEXT:    [[CMP_N13:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC5]]
 ; CHECK-NEXT:    br i1 [[CMP_N13]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[VEC_EPILOG_SCALAR_PH]]
@@ -680,8 +656,8 @@ define void @add_f(ptr noalias nocapture readonly %p, ptr noalias nocapture %q,
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i16, ptr [[P]], i64 [[INDVARS_IV]]
-; CHECK-NEXT:    [[TMP31:%.*]] = load i16, ptr [[ARRAYIDX]], align 2
-; CHECK-NEXT:    [[CONV:%.*]] = sext i16 [[TMP31]] to i32
+; CHECK-NEXT:    [[TMP27:%.*]] = load i16, ptr [[ARRAYIDX]], align 2
+; CHECK-NEXT:    [[CONV:%.*]] = sext i16 [[TMP27]] to i32
 ; CHECK-NEXT:    [[ADD:%.*]] = shl i32 [[CONV]], 4
 ; CHECK-NEXT:    [[CONV2:%.*]] = add nsw i32 [[ADD]], 32
 ; CHECK-NEXT:    [[OR:%.*]] = and i32 [[CONV]], 204
@@ -751,19 +727,17 @@ define void @add_phifail(ptr noalias nocapture readonly %p, ptr noalias nocaptur
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
-; CHECK-NEXT:    [[TMP3:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
-; CHECK-NEXT:    [[TMP4:%.*]] = add nuw nsw <16 x i32> [[TMP3]], splat (i32 2)
-; CHECK-NEXT:    [[TMP5:%.*]] = trunc <16 x i32> [[TMP4]] to <16 x i8>
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <16 x i8> [[TMP5]], ptr [[TMP7]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEXT:    [[TMP2:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
+; CHECK-NEXT:    [[TMP3:%.*]] = add nuw nsw <16 x i32> [[TMP2]], splat (i32 2)
+; CHECK-NEXT:    [[TMP4:%.*]] = trunc <16 x i32> [[TMP3]] to <16 x i8>
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX]]
+; CHECK-NEXT:    store <16 x i8> [[TMP4]], ptr [[TMP5]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
-; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
+; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <16 x i32> [[TMP3]], i32 15
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <16 x i32> [[TMP2]], i32 15
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
@@ -778,8 +752,8 @@ define void @add_phifail(ptr noalias nocapture readonly %p, ptr noalias nocaptur
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
 ; CHECK-NEXT:    [[A_PHI:%.*]] = phi i32 [ [[CONV:%.*]], [[FOR_BODY]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDVARS_IV]]
-; CHECK-NEXT:    [[TMP9:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
-; CHECK-NEXT:    [[CONV]] = zext i8 [[TMP9]] to i32
+; CHECK-NEXT:    [[TMP7:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
+; CHECK-NEXT:    [[CONV]] = zext i8 [[TMP7]] to i32
 ; CHECK-NEXT:    [[ADD:%.*]] = add nuw nsw i32 [[CONV]], 2
 ; CHECK-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i8
 ; CHECK-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDVARS_IV]]
@@ -832,20 +806,18 @@ define i8 @add_phifail2(ptr noalias nocapture readonly %p, ptr noalias nocapture
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
-; CHECK-NEXT:    [[TMP5:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
-; CHECK-NEXT:    [[TMP6:%.*]] = add nuw nsw <16 x i32> [[TMP5]], splat (i32 2)
-; CHECK-NEXT:    [[TMP7:%.*]] = trunc <16 x i32> [[TMP6]] to <16 x i8>
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <16 x i8> [[TMP7]], ptr [[TMP9]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
+; CHECK-NEXT:    [[TMP4:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
+; CHECK-NEXT:    [[TMP5:%.*]] = add nuw nsw <16 x i32> [[TMP4]], splat (i32 2)
+; CHECK-NEXT:    [[TMP6:%.*]] = trunc <16 x i32> [[TMP5]] to <16 x i8>
+; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDEX]]
+; CHECK-NEXT:    store <16 x i8> [[TMP6]], ptr [[TMP7]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
-; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
+; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <16 x i32> [[TMP5]], i32 15
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <16 x i32> [[TMP5]], i32 14
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <16 x i32> [[TMP4]], i32 15
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <16 x i32> [[TMP4]], i32 14
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP2]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
@@ -860,8 +832,8 @@ define i8 @add_phifail2(ptr noalias nocapture readonly %p, ptr noalias nocapture
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
 ; CHECK-NEXT:    [[A_PHI]] = phi i32 [ [[CONV:%.*]], [[FOR_BODY]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDVARS_IV]]
-; CHECK-NEXT:    [[TMP11:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
-; CHECK-NEXT:    [[CONV]] = zext i8 [[TMP11]] to i32
+; CHECK-NEXT:    [[TMP9:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
+; CHECK-NEXT:    [[CONV]] = zext i8 [[TMP9]] to i32
 ; CHECK-NEXT:    [[ADD:%.*]] = add nuw nsw i32 [[CONV]], 2
 ; CHECK-NEXT:    [[CONV1:%.*]] = trunc i32 [[ADD]] to i8
 ; CHECK-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i8, ptr [[Q]], i64 [[INDVARS_IV]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/low_trip_count_predicates.ll b/llvm/test/Transforms/LoopVectorize/AArch64/low_trip_count_predicates.ll
index 6499a1f..5f5d326 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/low_trip_count_predicates.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/low_trip_count_predicates.ll
@@ -81,10 +81,9 @@ define void @low_vf_ic_is_better(ptr nocapture noundef %p, i32 %tc, i16 noundef
 ; CHECK-VS1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VS1-NEXT:    [[TMP20:%.*]] = add i64 [[TMP0]], [[INDEX]]
 ; CHECK-VS1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw i8, ptr [[V]], i64 [[TMP20]]
-; CHECK-VS1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP22]], i32 0
-; CHECK-VS1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP23]], align 1
+; CHECK-VS1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP22]], align 1
 ; CHECK-VS1-NEXT:    [[TMP24:%.*]] = add <vscale x 16 x i8> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
-; CHECK-VS1-NEXT:    store <vscale x 16 x i8> [[TMP24]], ptr [[TMP23]], align 1
+; CHECK-VS1-NEXT:    store <vscale x 16 x i8> [[TMP24]], ptr [[TMP22]], align 1
 ; CHECK-VS1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP19]]
 ; CHECK-VS1-NEXT:    [[TMP25:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-VS1-NEXT:    br i1 [[TMP25]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -114,10 +113,9 @@ define void @low_vf_ic_is_better(ptr nocapture noundef %p, i32 %tc, i16 noundef
 ; CHECK-VS1-NEXT:    [[INDEX5:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], %[[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT9:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-VS1-NEXT:    [[OFFSET_IDX:%.*]] = add i64 [[TMP0]], [[INDEX5]]
 ; CHECK-VS1-NEXT:    [[TMP33:%.*]] = getelementptr inbounds nuw i8, ptr [[V]], i64 [[OFFSET_IDX]]
-; CHECK-VS1-NEXT:    [[TMP34:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP33]], i32 0
-; CHECK-VS1-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x i8>, ptr [[TMP34]], align 1
+; CHECK-VS1-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x i8>, ptr [[TMP33]], align 1
 ; CHECK-VS1-NEXT:    [[TMP35:%.*]] = add <vscale x 8 x i8> [[WIDE_LOAD6]], [[BROADCAST_SPLAT8]]
-; CHECK-VS1-NEXT:    store <vscale x 8 x i8> [[TMP35]], ptr [[TMP34]], align 1
+; CHECK-VS1-NEXT:    store <vscale x 8 x i8> [[TMP35]], ptr [[TMP33]], align 1
 ; CHECK-VS1-NEXT:    [[INDEX_NEXT9]] = add nuw i64 [[INDEX5]], [[TMP31]]
 ; CHECK-VS1-NEXT:    [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT9]], [[N_VEC3]]
 ; CHECK-VS1-NEXT:    br i1 [[TMP36]], label %[[VEC_EPILOG_MIDDLE_BLOCK:.*]], label %[[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
@@ -187,10 +185,9 @@ define void @low_vf_ic_is_better(ptr nocapture noundef %p, i32 %tc, i16 noundef
 ; CHECK-VS2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VS2-NEXT:    [[TMP20:%.*]] = add i64 [[TMP0]], [[INDEX]]
 ; CHECK-VS2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw i8, ptr [[V]], i64 [[TMP20]]
-; CHECK-VS2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP22]], i32 0
-; CHECK-VS2-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP23]], align 1
+; CHECK-VS2-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP22]], align 1
 ; CHECK-VS2-NEXT:    [[TMP24:%.*]] = add <vscale x 8 x i8> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
-; CHECK-VS2-NEXT:    store <vscale x 8 x i8> [[TMP24]], ptr [[TMP23]], align 1
+; CHECK-VS2-NEXT:    store <vscale x 8 x i8> [[TMP24]], ptr [[TMP22]], align 1
 ; CHECK-VS2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP19]]
 ; CHECK-VS2-NEXT:    [[TMP25:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-VS2-NEXT:    br i1 [[TMP25]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -220,10 +217,9 @@ define void @low_vf_ic_is_better(ptr nocapture noundef %p, i32 %tc, i16 noundef
 ; CHECK-VS2-NEXT:    [[INDEX5:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], %[[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT9:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-VS2-NEXT:    [[OFFSET_IDX:%.*]] = add i64 [[TMP0]], [[INDEX5]]
 ; CHECK-VS2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds nuw i8, ptr [[V]], i64 [[OFFSET_IDX]]
-; CHECK-VS2-NEXT:    [[TMP34:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP33]], i32 0
-; CHECK-VS2-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 4 x i8>, ptr [[TMP34]], align 1
+; CHECK-VS2-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 4 x i8>, ptr [[TMP33]], align 1
 ; CHECK-VS2-NEXT:    [[TMP35:%.*]] = add <vscale x 4 x i8> [[WIDE_LOAD6]], [[BROADCAST_SPLAT8]]
-; CHECK-VS2-NEXT:    store <vscale x 4 x i8> [[TMP35]], ptr [[TMP34]], align 1
+; CHECK-VS2-NEXT:    store <vscale x 4 x i8> [[TMP35]], ptr [[TMP33]], align 1
 ; CHECK-VS2-NEXT:    [[INDEX_NEXT9]] = add nuw i64 [[INDEX5]], [[TMP31]]
 ; CHECK-VS2-NEXT:    [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT9]], [[N_VEC3]]
 ; CHECK-VS2-NEXT:    br i1 [[TMP36]], label %[[VEC_EPILOG_MIDDLE_BLOCK:.*]], label %[[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
@@ -435,10 +431,9 @@ define void @overflow_indvar_known_false(ptr nocapture noundef %p, i32 noundef %
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], %[[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 [[TMP0]], [[INDEX]]
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw i8, ptr [[V]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP13]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr [[TMP14]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i8> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr [[TMP13]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i8> poison)
 ; CHECK-NEXT:    [[TMP15:%.*]] = add <vscale x 16 x i8> [[WIDE_MASKED_LOAD]], [[BROADCAST_SPLAT]]
-; CHECK-NEXT:    call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP15]], ptr [[TMP14]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP15]], ptr [[TMP13]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP8]]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[INDEX_NEXT]], i64 [[TMP1]])
 ; CHECK-NEXT:    [[TMP16:%.*]] = xor <vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
@@ -501,17 +496,16 @@ define i32 @tc4(ptr noundef readonly captures(none) %tmp) vscale_range(1,16) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP3:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i32, ptr [[ARRAYIDX1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[ARRAYIDX1]], align 4
 ; CHECK-NEXT:    [[TMP3]] = add <4 x i32> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[TMP4:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP3]])
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP4]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/mul-simplification.ll b/llvm/test/Transforms/LoopVectorize/AArch64/mul-simplification.ll
index a8d4442..1159a64 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/mul-simplification.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/mul-simplification.ll
@@ -61,10 +61,10 @@ define i32 @add_reduction_select_operand_constant_but_non_uniform() {
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[TMP1]], [[TMP2]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 64, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP3]], %[[MIDDLE_BLOCK]] ], [ 42, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 42, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[ADD2_REASS:%.*]], %[[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/optsize_minsize.ll b/llvm/test/Transforms/LoopVectorize/AArch64/optsize_minsize.ll
index f0835b8..1471896 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/optsize_minsize.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/optsize_minsize.ll
@@ -24,16 +24,14 @@ define void @always_vectorize(ptr %p, i32 %x) {
 ; DEFAULT-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; DEFAULT-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; DEFAULT:       [[VECTOR_BODY]]:
-; DEFAULT-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[P]], i32 0
-; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[P]], align 4
 ; DEFAULT-NEXT:    [[TMP3:%.*]] = add nsw <4 x i32> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
-; DEFAULT-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[P]], i32 0
-; DEFAULT-NEXT:    store <4 x i32> [[TMP3]], ptr [[TMP5]], align 4
+; DEFAULT-NEXT:    store <4 x i32> [[TMP3]], ptr [[P]], align 4
 ; DEFAULT-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; DEFAULT:       [[MIDDLE_BLOCK]]:
-; DEFAULT-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
+; DEFAULT-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
 ; DEFAULT:       [[SCALAR_PH]]:
-; DEFAULT-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; DEFAULT-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; DEFAULT-NEXT:    br label %[[FOR_BODY:.*]]
 ; DEFAULT:       [[FOR_BODY]]:
 ; DEFAULT-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -56,16 +54,14 @@ define void @always_vectorize(ptr %p, i32 %x) {
 ; OPTSIZE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; OPTSIZE-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; OPTSIZE:       [[VECTOR_BODY]]:
-; OPTSIZE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[P]], i32 0
-; OPTSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; OPTSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[P]], align 4
 ; OPTSIZE-NEXT:    [[TMP3:%.*]] = add nsw <4 x i32> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
-; OPTSIZE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[P]], i32 0
-; OPTSIZE-NEXT:    store <4 x i32> [[TMP3]], ptr [[TMP5]], align 4
+; OPTSIZE-NEXT:    store <4 x i32> [[TMP3]], ptr [[P]], align 4
 ; OPTSIZE-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; OPTSIZE:       [[MIDDLE_BLOCK]]:
-; OPTSIZE-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
+; OPTSIZE-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
 ; OPTSIZE:       [[SCALAR_PH]]:
-; OPTSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; OPTSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; OPTSIZE-NEXT:    br label %[[FOR_BODY:.*]]
 ; OPTSIZE:       [[FOR_BODY]]:
 ; OPTSIZE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -88,16 +84,14 @@ define void @always_vectorize(ptr %p, i32 %x) {
 ; MINSIZE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; MINSIZE-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; MINSIZE:       [[VECTOR_BODY]]:
-; MINSIZE-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[P]], i32 0
-; MINSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
+; MINSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[P]], align 4
 ; MINSIZE-NEXT:    [[TMP2:%.*]] = add nsw <4 x i32> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
-; MINSIZE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[P]], i32 0
-; MINSIZE-NEXT:    store <4 x i32> [[TMP2]], ptr [[TMP3]], align 4
+; MINSIZE-NEXT:    store <4 x i32> [[TMP2]], ptr [[P]], align 4
 ; MINSIZE-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; MINSIZE:       [[MIDDLE_BLOCK]]:
-; MINSIZE-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
+; MINSIZE-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
 ; MINSIZE:       [[SCALAR_PH]]:
-; MINSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; MINSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; MINSIZE-NEXT:    br label %[[FOR_BODY:.*]]
 ; MINSIZE:       [[FOR_BODY]]:
 ; MINSIZE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -145,13 +139,12 @@ define void @vectorize_without_optsize(ptr %p, i32 %x, i64 %n) {
 ; DEFAULT:       [[VECTOR_BODY]]:
 ; DEFAULT-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; DEFAULT-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[P]], i64 [[INDEX]]
-; DEFAULT-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
 ; DEFAULT-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 4
-; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; DEFAULT-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
 ; DEFAULT-NEXT:    [[TMP4:%.*]] = add nsw <4 x i32> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
 ; DEFAULT-NEXT:    [[TMP5:%.*]] = add nsw <4 x i32> [[WIDE_LOAD1]], [[BROADCAST_SPLAT]]
-; DEFAULT-NEXT:    store <4 x i32> [[TMP4]], ptr [[TMP2]], align 4
+; DEFAULT-NEXT:    store <4 x i32> [[TMP4]], ptr [[TMP1]], align 4
 ; DEFAULT-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP3]], align 4
 ; DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; DEFAULT-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -239,9 +232,9 @@ define void @tail_predicate_without_optsize(ptr %p, i8 %a, i8 %b, i8 %c, i32 %n)
 ; DEFAULT-NEXT:    [[BROADCAST_SPLAT6:%.*]] = shufflevector <16 x i8> [[BROADCAST_SPLATINSERT5]], <16 x i8> poison, <16 x i32> zeroinitializer
 ; DEFAULT-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; DEFAULT:       [[VECTOR_BODY]]:
-; DEFAULT-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE36:.*]] ]
-; DEFAULT-NEXT:    [[VEC_IND:%.*]] = phi <16 x i8> [ <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[PRED_STORE_CONTINUE36]] ]
-; DEFAULT-NEXT:    [[VEC_IND1:%.*]] = phi <16 x i8> [ <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT2:%.*]], %[[PRED_STORE_CONTINUE36]] ]
+; DEFAULT-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE35:.*]] ]
+; DEFAULT-NEXT:    [[VEC_IND:%.*]] = phi <16 x i8> [ <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[PRED_STORE_CONTINUE35]] ]
+; DEFAULT-NEXT:    [[VEC_IND1:%.*]] = phi <16 x i8> [ <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT2:%.*]], %[[PRED_STORE_CONTINUE35]] ]
 ; DEFAULT-NEXT:    [[TMP0:%.*]] = icmp ule <16 x i8> [[VEC_IND]], splat (i8 14)
 ; DEFAULT-NEXT:    [[TMP1:%.*]] = mul <16 x i8> [[BROADCAST_SPLAT]], [[VEC_IND1]]
 ; DEFAULT-NEXT:    [[TMP2:%.*]] = lshr <16 x i8> [[VEC_IND1]], splat (i8 1)
@@ -260,140 +253,140 @@ define void @tail_predicate_without_optsize(ptr %p, i8 %a, i8 %b, i8 %c, i32 %n)
 ; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE]]
 ; DEFAULT:       [[PRED_STORE_CONTINUE]]:
 ; DEFAULT-NEXT:    [[TMP12:%.*]] = extractelement <16 x i1> [[TMP0]], i32 1
-; DEFAULT-NEXT:    br i1 [[TMP12]], label %[[PRED_STORE_IF7:.*]], label %[[PRED_STORE_CONTINUE8:.*]]
-; DEFAULT:       [[PRED_STORE_IF7]]:
+; DEFAULT-NEXT:    br i1 [[TMP12]], label %[[PRED_STORE_IF6:.*]], label %[[PRED_STORE_CONTINUE7:.*]]
+; DEFAULT:       [[PRED_STORE_IF6]]:
 ; DEFAULT-NEXT:    [[TMP13:%.*]] = add i64 [[INDEX]], 1
 ; DEFAULT-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP13]]
 ; DEFAULT-NEXT:    [[TMP15:%.*]] = extractelement <16 x i8> [[TMP7]], i32 1
 ; DEFAULT-NEXT:    store i8 [[TMP15]], ptr [[TMP14]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE8]]
-; DEFAULT:       [[PRED_STORE_CONTINUE8]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE7]]
+; DEFAULT:       [[PRED_STORE_CONTINUE7]]:
 ; DEFAULT-NEXT:    [[TMP16:%.*]] = extractelement <16 x i1> [[TMP0]], i32 2
-; DEFAULT-NEXT:    br i1 [[TMP16]], label %[[PRED_STORE_IF9:.*]], label %[[PRED_STORE_CONTINUE10:.*]]
-; DEFAULT:       [[PRED_STORE_IF9]]:
+; DEFAULT-NEXT:    br i1 [[TMP16]], label %[[PRED_STORE_IF8:.*]], label %[[PRED_STORE_CONTINUE9:.*]]
+; DEFAULT:       [[PRED_STORE_IF8]]:
 ; DEFAULT-NEXT:    [[TMP17:%.*]] = add i64 [[INDEX]], 2
 ; DEFAULT-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP17]]
 ; DEFAULT-NEXT:    [[TMP19:%.*]] = extractelement <16 x i8> [[TMP7]], i32 2
 ; DEFAULT-NEXT:    store i8 [[TMP19]], ptr [[TMP18]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE10]]
-; DEFAULT:       [[PRED_STORE_CONTINUE10]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE9]]
+; DEFAULT:       [[PRED_STORE_CONTINUE9]]:
 ; DEFAULT-NEXT:    [[TMP20:%.*]] = extractelement <16 x i1> [[TMP0]], i32 3
-; DEFAULT-NEXT:    br i1 [[TMP20]], label %[[PRED_STORE_IF11:.*]], label %[[PRED_STORE_CONTINUE12:.*]]
-; DEFAULT:       [[PRED_STORE_IF11]]:
+; DEFAULT-NEXT:    br i1 [[TMP20]], label %[[PRED_STORE_IF10:.*]], label %[[PRED_STORE_CONTINUE11:.*]]
+; DEFAULT:       [[PRED_STORE_IF10]]:
 ; DEFAULT-NEXT:    [[TMP21:%.*]] = add i64 [[INDEX]], 3
 ; DEFAULT-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP21]]
 ; DEFAULT-NEXT:    [[TMP23:%.*]] = extractelement <16 x i8> [[TMP7]], i32 3
 ; DEFAULT-NEXT:    store i8 [[TMP23]], ptr [[TMP22]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE12]]
-; DEFAULT:       [[PRED_STORE_CONTINUE12]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE11]]
+; DEFAULT:       [[PRED_STORE_CONTINUE11]]:
 ; DEFAULT-NEXT:    [[TMP24:%.*]] = extractelement <16 x i1> [[TMP0]], i32 4
-; DEFAULT-NEXT:    br i1 [[TMP24]], label %[[PRED_STORE_IF13:.*]], label %[[PRED_STORE_CONTINUE14:.*]]
-; DEFAULT:       [[PRED_STORE_IF13]]:
+; DEFAULT-NEXT:    br i1 [[TMP24]], label %[[PRED_STORE_IF12:.*]], label %[[PRED_STORE_CONTINUE13:.*]]
+; DEFAULT:       [[PRED_STORE_IF12]]:
 ; DEFAULT-NEXT:    [[TMP25:%.*]] = add i64 [[INDEX]], 4
 ; DEFAULT-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP25]]
 ; DEFAULT-NEXT:    [[TMP27:%.*]] = extractelement <16 x i8> [[TMP7]], i32 4
 ; DEFAULT-NEXT:    store i8 [[TMP27]], ptr [[TMP26]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE14]]
-; DEFAULT:       [[PRED_STORE_CONTINUE14]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE13]]
+; DEFAULT:       [[PRED_STORE_CONTINUE13]]:
 ; DEFAULT-NEXT:    [[TMP28:%.*]] = extractelement <16 x i1> [[TMP0]], i32 5
-; DEFAULT-NEXT:    br i1 [[TMP28]], label %[[PRED_STORE_IF15:.*]], label %[[PRED_STORE_CONTINUE16:.*]]
-; DEFAULT:       [[PRED_STORE_IF15]]:
+; DEFAULT-NEXT:    br i1 [[TMP28]], label %[[PRED_STORE_IF14:.*]], label %[[PRED_STORE_CONTINUE15:.*]]
+; DEFAULT:       [[PRED_STORE_IF14]]:
 ; DEFAULT-NEXT:    [[TMP29:%.*]] = add i64 [[INDEX]], 5
 ; DEFAULT-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP29]]
 ; DEFAULT-NEXT:    [[TMP31:%.*]] = extractelement <16 x i8> [[TMP7]], i32 5
 ; DEFAULT-NEXT:    store i8 [[TMP31]], ptr [[TMP30]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE16]]
-; DEFAULT:       [[PRED_STORE_CONTINUE16]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE15]]
+; DEFAULT:       [[PRED_STORE_CONTINUE15]]:
 ; DEFAULT-NEXT:    [[TMP32:%.*]] = extractelement <16 x i1> [[TMP0]], i32 6
-; DEFAULT-NEXT:    br i1 [[TMP32]], label %[[PRED_STORE_IF17:.*]], label %[[PRED_STORE_CONTINUE18:.*]]
-; DEFAULT:       [[PRED_STORE_IF17]]:
+; DEFAULT-NEXT:    br i1 [[TMP32]], label %[[PRED_STORE_IF16:.*]], label %[[PRED_STORE_CONTINUE17:.*]]
+; DEFAULT:       [[PRED_STORE_IF16]]:
 ; DEFAULT-NEXT:    [[TMP33:%.*]] = add i64 [[INDEX]], 6
 ; DEFAULT-NEXT:    [[TMP34:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP33]]
 ; DEFAULT-NEXT:    [[TMP35:%.*]] = extractelement <16 x i8> [[TMP7]], i32 6
 ; DEFAULT-NEXT:    store i8 [[TMP35]], ptr [[TMP34]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE18]]
-; DEFAULT:       [[PRED_STORE_CONTINUE18]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE17]]
+; DEFAULT:       [[PRED_STORE_CONTINUE17]]:
 ; DEFAULT-NEXT:    [[TMP36:%.*]] = extractelement <16 x i1> [[TMP0]], i32 7
-; DEFAULT-NEXT:    br i1 [[TMP36]], label %[[PRED_STORE_IF19:.*]], label %[[PRED_STORE_CONTINUE20:.*]]
-; DEFAULT:       [[PRED_STORE_IF19]]:
+; DEFAULT-NEXT:    br i1 [[TMP36]], label %[[PRED_STORE_IF18:.*]], label %[[PRED_STORE_CONTINUE19:.*]]
+; DEFAULT:       [[PRED_STORE_IF18]]:
 ; DEFAULT-NEXT:    [[TMP37:%.*]] = add i64 [[INDEX]], 7
 ; DEFAULT-NEXT:    [[TMP38:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP37]]
 ; DEFAULT-NEXT:    [[TMP39:%.*]] = extractelement <16 x i8> [[TMP7]], i32 7
 ; DEFAULT-NEXT:    store i8 [[TMP39]], ptr [[TMP38]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE20]]
-; DEFAULT:       [[PRED_STORE_CONTINUE20]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE19]]
+; DEFAULT:       [[PRED_STORE_CONTINUE19]]:
 ; DEFAULT-NEXT:    [[TMP40:%.*]] = extractelement <16 x i1> [[TMP0]], i32 8
-; DEFAULT-NEXT:    br i1 [[TMP40]], label %[[PRED_STORE_IF21:.*]], label %[[PRED_STORE_CONTINUE22:.*]]
-; DEFAULT:       [[PRED_STORE_IF21]]:
+; DEFAULT-NEXT:    br i1 [[TMP40]], label %[[PRED_STORE_IF20:.*]], label %[[PRED_STORE_CONTINUE21:.*]]
+; DEFAULT:       [[PRED_STORE_IF20]]:
 ; DEFAULT-NEXT:    [[TMP41:%.*]] = add i64 [[INDEX]], 8
 ; DEFAULT-NEXT:    [[TMP42:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP41]]
 ; DEFAULT-NEXT:    [[TMP43:%.*]] = extractelement <16 x i8> [[TMP7]], i32 8
 ; DEFAULT-NEXT:    store i8 [[TMP43]], ptr [[TMP42]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE22]]
-; DEFAULT:       [[PRED_STORE_CONTINUE22]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE21]]
+; DEFAULT:       [[PRED_STORE_CONTINUE21]]:
 ; DEFAULT-NEXT:    [[TMP44:%.*]] = extractelement <16 x i1> [[TMP0]], i32 9
-; DEFAULT-NEXT:    br i1 [[TMP44]], label %[[PRED_STORE_IF23:.*]], label %[[PRED_STORE_CONTINUE24:.*]]
-; DEFAULT:       [[PRED_STORE_IF23]]:
+; DEFAULT-NEXT:    br i1 [[TMP44]], label %[[PRED_STORE_IF22:.*]], label %[[PRED_STORE_CONTINUE23:.*]]
+; DEFAULT:       [[PRED_STORE_IF22]]:
 ; DEFAULT-NEXT:    [[TMP45:%.*]] = add i64 [[INDEX]], 9
 ; DEFAULT-NEXT:    [[TMP46:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP45]]
 ; DEFAULT-NEXT:    [[TMP47:%.*]] = extractelement <16 x i8> [[TMP7]], i32 9
 ; DEFAULT-NEXT:    store i8 [[TMP47]], ptr [[TMP46]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE24]]
-; DEFAULT:       [[PRED_STORE_CONTINUE24]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE23]]
+; DEFAULT:       [[PRED_STORE_CONTINUE23]]:
 ; DEFAULT-NEXT:    [[TMP48:%.*]] = extractelement <16 x i1> [[TMP0]], i32 10
-; DEFAULT-NEXT:    br i1 [[TMP48]], label %[[PRED_STORE_IF25:.*]], label %[[PRED_STORE_CONTINUE26:.*]]
-; DEFAULT:       [[PRED_STORE_IF25]]:
+; DEFAULT-NEXT:    br i1 [[TMP48]], label %[[PRED_STORE_IF24:.*]], label %[[PRED_STORE_CONTINUE25:.*]]
+; DEFAULT:       [[PRED_STORE_IF24]]:
 ; DEFAULT-NEXT:    [[TMP49:%.*]] = add i64 [[INDEX]], 10
 ; DEFAULT-NEXT:    [[TMP50:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP49]]
 ; DEFAULT-NEXT:    [[TMP51:%.*]] = extractelement <16 x i8> [[TMP7]], i32 10
 ; DEFAULT-NEXT:    store i8 [[TMP51]], ptr [[TMP50]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE26]]
-; DEFAULT:       [[PRED_STORE_CONTINUE26]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE25]]
+; DEFAULT:       [[PRED_STORE_CONTINUE25]]:
 ; DEFAULT-NEXT:    [[TMP52:%.*]] = extractelement <16 x i1> [[TMP0]], i32 11
-; DEFAULT-NEXT:    br i1 [[TMP52]], label %[[PRED_STORE_IF27:.*]], label %[[PRED_STORE_CONTINUE28:.*]]
-; DEFAULT:       [[PRED_STORE_IF27]]:
+; DEFAULT-NEXT:    br i1 [[TMP52]], label %[[PRED_STORE_IF26:.*]], label %[[PRED_STORE_CONTINUE27:.*]]
+; DEFAULT:       [[PRED_STORE_IF26]]:
 ; DEFAULT-NEXT:    [[TMP53:%.*]] = add i64 [[INDEX]], 11
 ; DEFAULT-NEXT:    [[TMP54:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP53]]
 ; DEFAULT-NEXT:    [[TMP55:%.*]] = extractelement <16 x i8> [[TMP7]], i32 11
 ; DEFAULT-NEXT:    store i8 [[TMP55]], ptr [[TMP54]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE28]]
-; DEFAULT:       [[PRED_STORE_CONTINUE28]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE27]]
+; DEFAULT:       [[PRED_STORE_CONTINUE27]]:
 ; DEFAULT-NEXT:    [[TMP56:%.*]] = extractelement <16 x i1> [[TMP0]], i32 12
-; DEFAULT-NEXT:    br i1 [[TMP56]], label %[[PRED_STORE_IF29:.*]], label %[[PRED_STORE_CONTINUE30:.*]]
-; DEFAULT:       [[PRED_STORE_IF29]]:
+; DEFAULT-NEXT:    br i1 [[TMP56]], label %[[PRED_STORE_IF28:.*]], label %[[PRED_STORE_CONTINUE29:.*]]
+; DEFAULT:       [[PRED_STORE_IF28]]:
 ; DEFAULT-NEXT:    [[TMP57:%.*]] = add i64 [[INDEX]], 12
 ; DEFAULT-NEXT:    [[TMP58:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP57]]
 ; DEFAULT-NEXT:    [[TMP59:%.*]] = extractelement <16 x i8> [[TMP7]], i32 12
 ; DEFAULT-NEXT:    store i8 [[TMP59]], ptr [[TMP58]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE30]]
-; DEFAULT:       [[PRED_STORE_CONTINUE30]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE29]]
+; DEFAULT:       [[PRED_STORE_CONTINUE29]]:
 ; DEFAULT-NEXT:    [[TMP60:%.*]] = extractelement <16 x i1> [[TMP0]], i32 13
-; DEFAULT-NEXT:    br i1 [[TMP60]], label %[[PRED_STORE_IF31:.*]], label %[[PRED_STORE_CONTINUE32:.*]]
-; DEFAULT:       [[PRED_STORE_IF31]]:
+; DEFAULT-NEXT:    br i1 [[TMP60]], label %[[PRED_STORE_IF30:.*]], label %[[PRED_STORE_CONTINUE31:.*]]
+; DEFAULT:       [[PRED_STORE_IF30]]:
 ; DEFAULT-NEXT:    [[TMP61:%.*]] = add i64 [[INDEX]], 13
 ; DEFAULT-NEXT:    [[TMP62:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP61]]
 ; DEFAULT-NEXT:    [[TMP63:%.*]] = extractelement <16 x i8> [[TMP7]], i32 13
 ; DEFAULT-NEXT:    store i8 [[TMP63]], ptr [[TMP62]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE32]]
-; DEFAULT:       [[PRED_STORE_CONTINUE32]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE31]]
+; DEFAULT:       [[PRED_STORE_CONTINUE31]]:
 ; DEFAULT-NEXT:    [[TMP64:%.*]] = extractelement <16 x i1> [[TMP0]], i32 14
-; DEFAULT-NEXT:    br i1 [[TMP64]], label %[[PRED_STORE_IF33:.*]], label %[[PRED_STORE_CONTINUE34:.*]]
-; DEFAULT:       [[PRED_STORE_IF33]]:
+; DEFAULT-NEXT:    br i1 [[TMP64]], label %[[PRED_STORE_IF32:.*]], label %[[PRED_STORE_CONTINUE33:.*]]
+; DEFAULT:       [[PRED_STORE_IF32]]:
 ; DEFAULT-NEXT:    [[TMP65:%.*]] = add i64 [[INDEX]], 14
 ; DEFAULT-NEXT:    [[TMP66:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP65]]
 ; DEFAULT-NEXT:    [[TMP67:%.*]] = extractelement <16 x i8> [[TMP7]], i32 14
 ; DEFAULT-NEXT:    store i8 [[TMP67]], ptr [[TMP66]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE34]]
-; DEFAULT:       [[PRED_STORE_CONTINUE34]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE33]]
+; DEFAULT:       [[PRED_STORE_CONTINUE33]]:
 ; DEFAULT-NEXT:    [[TMP68:%.*]] = extractelement <16 x i1> [[TMP0]], i32 15
-; DEFAULT-NEXT:    br i1 [[TMP68]], label %[[PRED_STORE_IF35:.*]], label %[[PRED_STORE_CONTINUE36]]
-; DEFAULT:       [[PRED_STORE_IF35]]:
+; DEFAULT-NEXT:    br i1 [[TMP68]], label %[[PRED_STORE_IF34:.*]], label %[[PRED_STORE_CONTINUE35]]
+; DEFAULT:       [[PRED_STORE_IF34]]:
 ; DEFAULT-NEXT:    [[TMP69:%.*]] = add i64 [[INDEX]], 15
 ; DEFAULT-NEXT:    [[TMP70:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP69]]
 ; DEFAULT-NEXT:    [[TMP71:%.*]] = extractelement <16 x i8> [[TMP7]], i32 15
 ; DEFAULT-NEXT:    store i8 [[TMP71]], ptr [[TMP70]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE36]]
-; DEFAULT:       [[PRED_STORE_CONTINUE36]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE35]]
+; DEFAULT:       [[PRED_STORE_CONTINUE35]]:
 ; DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; DEFAULT-NEXT:    [[VEC_IND_NEXT]] = add <16 x i8> [[VEC_IND]], splat (i8 16)
 ; DEFAULT-NEXT:    [[VEC_IND_NEXT2]] = add <16 x i8> [[VEC_IND1]], splat (i8 16)
@@ -540,8 +533,7 @@ define void @sve_tail_predicate_without_minsize(ptr %p, i8 %a, i8 %b, i8 %c, i32
 ; DEFAULT-NEXT:    [[TMP20:%.*]] = mul <vscale x 16 x i8> [[TMP19]], [[BROADCAST_SPLAT4]]
 ; DEFAULT-NEXT:    [[TMP21:%.*]] = add <vscale x 16 x i8> [[TMP18]], [[TMP20]]
 ; DEFAULT-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDEX]]
-; DEFAULT-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i8, ptr [[TMP22]], i32 0
-; DEFAULT-NEXT:    call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP21]], ptr [[TMP23]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
+; DEFAULT-NEXT:    call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP21]], ptr [[TMP22]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
 ; DEFAULT-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP4]]
 ; DEFAULT-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[INDEX]], i64 [[TMP9]])
 ; DEFAULT-NEXT:    [[TMP24:%.*]] = xor <vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
@@ -616,8 +608,7 @@ define void @sve_tail_predicate_without_minsize(ptr %p, i8 %a, i8 %b, i8 %c, i32
 ; OPTSIZE-NEXT:    [[TMP20:%.*]] = mul <vscale x 16 x i8> [[TMP19]], [[BROADCAST_SPLAT4]]
 ; OPTSIZE-NEXT:    [[TMP21:%.*]] = add <vscale x 16 x i8> [[TMP18]], [[TMP20]]
 ; OPTSIZE-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDEX]]
-; OPTSIZE-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i8, ptr [[TMP22]], i32 0
-; OPTSIZE-NEXT:    call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP21]], ptr [[TMP23]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
+; OPTSIZE-NEXT:    call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP21]], ptr [[TMP22]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
 ; OPTSIZE-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP4]]
 ; OPTSIZE-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[INDEX]], i64 [[TMP9]])
 ; OPTSIZE-NEXT:    [[TMP24:%.*]] = xor <vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
@@ -692,8 +683,7 @@ define void @sve_tail_predicate_without_minsize(ptr %p, i8 %a, i8 %b, i8 %c, i32
 ; MINSIZE-NEXT:    [[TMP20:%.*]] = mul <vscale x 16 x i8> [[TMP19]], [[BROADCAST_SPLAT4]]
 ; MINSIZE-NEXT:    [[TMP21:%.*]] = add <vscale x 16 x i8> [[TMP18]], [[TMP20]]
 ; MINSIZE-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[INDEX]]
-; MINSIZE-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i8, ptr [[TMP22]], i32 0
-; MINSIZE-NEXT:    call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP21]], ptr [[TMP23]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
+; MINSIZE-NEXT:    call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> [[TMP21]], ptr [[TMP22]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
 ; MINSIZE-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP4]]
 ; MINSIZE-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[INDEX]], i64 [[TMP9]])
 ; MINSIZE-NEXT:    [[TMP24:%.*]] = xor <vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
@@ -760,35 +750,32 @@ define void @dont_vectorize_with_minsize() {
 ; DEFAULT:       [[VECTOR_BODY]]:
 ; DEFAULT-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; DEFAULT-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @B, i64 0, i64 [[INDEX]]
-; DEFAULT-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i32 0
 ; DEFAULT-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i32 8
-; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP2]], align 4
+; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP1]], align 4
 ; DEFAULT-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i32>, ptr [[TMP3]], align 4
 ; DEFAULT-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @C, i64 0, i64 [[INDEX]]
-; DEFAULT-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP4]], i32 0
 ; DEFAULT-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP4]], i32 8
-; DEFAULT-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i32>, ptr [[TMP5]], align 4
+; DEFAULT-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i32>, ptr [[TMP4]], align 4
 ; DEFAULT-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i32>, ptr [[TMP6]], align 4
 ; DEFAULT-NEXT:    [[TMP7:%.*]] = mul nsw <8 x i32> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; DEFAULT-NEXT:    [[TMP8:%.*]] = mul nsw <8 x i32> [[WIDE_LOAD1]], [[WIDE_LOAD3]]
 ; DEFAULT-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [1000 x i16], ptr @A, i64 0, i64 [[INDEX]]
-; DEFAULT-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP9]], i32 0
 ; DEFAULT-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP9]], i32 8
-; DEFAULT-NEXT:    [[WIDE_LOAD4:%.*]] = load <8 x i16>, ptr [[TMP10]], align 2
+; DEFAULT-NEXT:    [[WIDE_LOAD4:%.*]] = load <8 x i16>, ptr [[TMP9]], align 2
 ; DEFAULT-NEXT:    [[WIDE_LOAD5:%.*]] = load <8 x i16>, ptr [[TMP11]], align 2
 ; DEFAULT-NEXT:    [[TMP12:%.*]] = trunc <8 x i32> [[TMP7]] to <8 x i16>
 ; DEFAULT-NEXT:    [[TMP13:%.*]] = trunc <8 x i32> [[TMP8]] to <8 x i16>
 ; DEFAULT-NEXT:    [[TMP14:%.*]] = add <8 x i16> [[TMP12]], [[WIDE_LOAD4]]
 ; DEFAULT-NEXT:    [[TMP15:%.*]] = add <8 x i16> [[TMP13]], [[WIDE_LOAD5]]
-; DEFAULT-NEXT:    store <8 x i16> [[TMP14]], ptr [[TMP10]], align 2
+; DEFAULT-NEXT:    store <8 x i16> [[TMP14]], ptr [[TMP9]], align 2
 ; DEFAULT-NEXT:    store <8 x i16> [[TMP15]], ptr [[TMP11]], align 2
 ; DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; DEFAULT-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; DEFAULT-NEXT:    br i1 [[TMP16]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
 ; DEFAULT:       [[MIDDLE_BLOCK]]:
-; DEFAULT-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
+; DEFAULT-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
 ; DEFAULT:       [[SCALAR_PH]]:
-; DEFAULT-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 64, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; DEFAULT-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; DEFAULT-NEXT:    br label %[[FOR_BODY:.*]]
 ; DEFAULT:       [[FOR_BODY]]:
 ; DEFAULT-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -817,25 +804,22 @@ define void @dont_vectorize_with_minsize() {
 ; OPTSIZE:       [[VECTOR_BODY]]:
 ; OPTSIZE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; OPTSIZE-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @B, i64 0, i64 [[INDEX]]
-; OPTSIZE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i32 0
-; OPTSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP2]], align 4
+; OPTSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP1]], align 4
 ; OPTSIZE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @C, i64 0, i64 [[INDEX]]
-; OPTSIZE-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP3]], i32 0
-; OPTSIZE-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i32>, ptr [[TMP4]], align 4
+; OPTSIZE-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i32>, ptr [[TMP3]], align 4
 ; OPTSIZE-NEXT:    [[TMP5:%.*]] = mul nsw <8 x i32> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; OPTSIZE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [1000 x i16], ptr @A, i64 0, i64 [[INDEX]]
-; OPTSIZE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP6]], i32 0
-; OPTSIZE-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i16>, ptr [[TMP7]], align 2
+; OPTSIZE-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i16>, ptr [[TMP6]], align 2
 ; OPTSIZE-NEXT:    [[TMP8:%.*]] = trunc <8 x i32> [[TMP5]] to <8 x i16>
 ; OPTSIZE-NEXT:    [[TMP9:%.*]] = add <8 x i16> [[TMP8]], [[WIDE_LOAD2]]
-; OPTSIZE-NEXT:    store <8 x i16> [[TMP9]], ptr [[TMP7]], align 2
+; OPTSIZE-NEXT:    store <8 x i16> [[TMP9]], ptr [[TMP6]], align 2
 ; OPTSIZE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; OPTSIZE-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; OPTSIZE-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; OPTSIZE:       [[MIDDLE_BLOCK]]:
-; OPTSIZE-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
+; OPTSIZE-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
 ; OPTSIZE:       [[SCALAR_PH]]:
-; OPTSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 64, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; OPTSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; OPTSIZE-NEXT:    br label %[[FOR_BODY:.*]]
 ; OPTSIZE:       [[FOR_BODY]]:
 ; OPTSIZE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -864,25 +848,22 @@ define void @dont_vectorize_with_minsize() {
 ; MINSIZE:       [[VECTOR_BODY]]:
 ; MINSIZE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; MINSIZE-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @B, i64 0, i64 [[INDEX]]
-; MINSIZE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i32 0
-; MINSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; MINSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; MINSIZE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @C, i64 0, i64 [[INDEX]]
-; MINSIZE-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP3]], i32 0
-; MINSIZE-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i32>, ptr [[TMP4]], align 4
+; MINSIZE-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i32>, ptr [[TMP3]], align 4
 ; MINSIZE-NEXT:    [[TMP5:%.*]] = mul nsw <2 x i32> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; MINSIZE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [1000 x i16], ptr @A, i64 0, i64 [[INDEX]]
-; MINSIZE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP6]], i32 0
-; MINSIZE-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i16>, ptr [[TMP7]], align 2
+; MINSIZE-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i16>, ptr [[TMP6]], align 2
 ; MINSIZE-NEXT:    [[TMP8:%.*]] = trunc <2 x i32> [[TMP5]] to <2 x i16>
 ; MINSIZE-NEXT:    [[TMP9:%.*]] = add <2 x i16> [[TMP8]], [[WIDE_LOAD2]]
-; MINSIZE-NEXT:    store <2 x i16> [[TMP9]], ptr [[TMP7]], align 2
+; MINSIZE-NEXT:    store <2 x i16> [[TMP9]], ptr [[TMP6]], align 2
 ; MINSIZE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; MINSIZE-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; MINSIZE-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; MINSIZE:       [[MIDDLE_BLOCK]]:
-; MINSIZE-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
+; MINSIZE-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
 ; MINSIZE:       [[SCALAR_PH]]:
-; MINSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 64, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; MINSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; MINSIZE-NEXT:    br label %[[FOR_BODY:.*]]
 ; MINSIZE:       [[FOR_BODY]]:
 ; MINSIZE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -939,35 +920,32 @@ define void @vectorization_forced_minsize_reduce_width() {
 ; DEFAULT:       [[VECTOR_BODY]]:
 ; DEFAULT-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; DEFAULT-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @B, i64 0, i64 [[INDEX]]
-; DEFAULT-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i32 0
 ; DEFAULT-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i32 8
-; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP2]], align 4
+; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP1]], align 4
 ; DEFAULT-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i32>, ptr [[TMP3]], align 4
 ; DEFAULT-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @C, i64 0, i64 [[INDEX]]
-; DEFAULT-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP4]], i32 0
 ; DEFAULT-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP4]], i32 8
-; DEFAULT-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i32>, ptr [[TMP5]], align 4
+; DEFAULT-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i32>, ptr [[TMP4]], align 4
 ; DEFAULT-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i32>, ptr [[TMP6]], align 4
 ; DEFAULT-NEXT:    [[TMP7:%.*]] = mul nsw <8 x i32> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; DEFAULT-NEXT:    [[TMP8:%.*]] = mul nsw <8 x i32> [[WIDE_LOAD1]], [[WIDE_LOAD3]]
 ; DEFAULT-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [1000 x i16], ptr @A, i64 0, i64 [[INDEX]]
-; DEFAULT-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP9]], i32 0
 ; DEFAULT-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP9]], i32 8
-; DEFAULT-NEXT:    [[WIDE_LOAD4:%.*]] = load <8 x i16>, ptr [[TMP10]], align 2
+; DEFAULT-NEXT:    [[WIDE_LOAD4:%.*]] = load <8 x i16>, ptr [[TMP9]], align 2
 ; DEFAULT-NEXT:    [[WIDE_LOAD5:%.*]] = load <8 x i16>, ptr [[TMP11]], align 2
 ; DEFAULT-NEXT:    [[TMP12:%.*]] = trunc <8 x i32> [[TMP7]] to <8 x i16>
 ; DEFAULT-NEXT:    [[TMP13:%.*]] = trunc <8 x i32> [[TMP8]] to <8 x i16>
 ; DEFAULT-NEXT:    [[TMP14:%.*]] = add <8 x i16> [[TMP12]], [[WIDE_LOAD4]]
 ; DEFAULT-NEXT:    [[TMP15:%.*]] = add <8 x i16> [[TMP13]], [[WIDE_LOAD5]]
-; DEFAULT-NEXT:    store <8 x i16> [[TMP14]], ptr [[TMP10]], align 2
+; DEFAULT-NEXT:    store <8 x i16> [[TMP14]], ptr [[TMP9]], align 2
 ; DEFAULT-NEXT:    store <8 x i16> [[TMP15]], ptr [[TMP11]], align 2
 ; DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; DEFAULT-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; DEFAULT-NEXT:    br i1 [[TMP16]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ; DEFAULT:       [[MIDDLE_BLOCK]]:
-; DEFAULT-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
+; DEFAULT-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
 ; DEFAULT:       [[SCALAR_PH]]:
-; DEFAULT-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 64, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; DEFAULT-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; DEFAULT-NEXT:    br label %[[FOR_BODY:.*]]
 ; DEFAULT:       [[FOR_BODY]]:
 ; DEFAULT-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -996,25 +974,22 @@ define void @vectorization_forced_minsize_reduce_width() {
 ; OPTSIZE:       [[VECTOR_BODY]]:
 ; OPTSIZE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; OPTSIZE-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @B, i64 0, i64 [[INDEX]]
-; OPTSIZE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i32 0
-; OPTSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP2]], align 4
+; OPTSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP1]], align 4
 ; OPTSIZE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @C, i64 0, i64 [[INDEX]]
-; OPTSIZE-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP3]], i32 0
-; OPTSIZE-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i32>, ptr [[TMP4]], align 4
+; OPTSIZE-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i32>, ptr [[TMP3]], align 4
 ; OPTSIZE-NEXT:    [[TMP5:%.*]] = mul nsw <8 x i32> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; OPTSIZE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [1000 x i16], ptr @A, i64 0, i64 [[INDEX]]
-; OPTSIZE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP6]], i32 0
-; OPTSIZE-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i16>, ptr [[TMP7]], align 2
+; OPTSIZE-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i16>, ptr [[TMP6]], align 2
 ; OPTSIZE-NEXT:    [[TMP8:%.*]] = trunc <8 x i32> [[TMP5]] to <8 x i16>
 ; OPTSIZE-NEXT:    [[TMP9:%.*]] = add <8 x i16> [[TMP8]], [[WIDE_LOAD2]]
-; OPTSIZE-NEXT:    store <8 x i16> [[TMP9]], ptr [[TMP7]], align 2
+; OPTSIZE-NEXT:    store <8 x i16> [[TMP9]], ptr [[TMP6]], align 2
 ; OPTSIZE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; OPTSIZE-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; OPTSIZE-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
 ; OPTSIZE:       [[MIDDLE_BLOCK]]:
-; OPTSIZE-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
+; OPTSIZE-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
 ; OPTSIZE:       [[SCALAR_PH]]:
-; OPTSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 64, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; OPTSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; OPTSIZE-NEXT:    br label %[[FOR_BODY:.*]]
 ; OPTSIZE:       [[FOR_BODY]]:
 ; OPTSIZE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -1043,25 +1018,22 @@ define void @vectorization_forced_minsize_reduce_width() {
 ; MINSIZE:       [[VECTOR_BODY]]:
 ; MINSIZE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; MINSIZE-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @B, i64 0, i64 [[INDEX]]
-; MINSIZE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i32 0
-; MINSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; MINSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; MINSIZE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @C, i64 0, i64 [[INDEX]]
-; MINSIZE-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP3]], i32 0
-; MINSIZE-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i32>, ptr [[TMP4]], align 4
+; MINSIZE-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i32>, ptr [[TMP3]], align 4
 ; MINSIZE-NEXT:    [[TMP5:%.*]] = mul nsw <2 x i32> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; MINSIZE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [1000 x i16], ptr @A, i64 0, i64 [[INDEX]]
-; MINSIZE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP6]], i32 0
-; MINSIZE-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i16>, ptr [[TMP7]], align 2
+; MINSIZE-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i16>, ptr [[TMP6]], align 2
 ; MINSIZE-NEXT:    [[TMP8:%.*]] = trunc <2 x i32> [[TMP5]] to <2 x i16>
 ; MINSIZE-NEXT:    [[TMP9:%.*]] = add <2 x i16> [[TMP8]], [[WIDE_LOAD2]]
-; MINSIZE-NEXT:    store <2 x i16> [[TMP9]], ptr [[TMP7]], align 2
+; MINSIZE-NEXT:    store <2 x i16> [[TMP9]], ptr [[TMP6]], align 2
 ; MINSIZE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; MINSIZE-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; MINSIZE-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
 ; MINSIZE:       [[MIDDLE_BLOCK]]:
-; MINSIZE-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
+; MINSIZE-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
 ; MINSIZE:       [[SCALAR_PH]]:
-; MINSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 64, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; MINSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; MINSIZE-NEXT:    br label %[[FOR_BODY:.*]]
 ; MINSIZE:       [[FOR_BODY]]:
 ; MINSIZE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-chained.ll b/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-chained.ll
index b02b314..787d63c 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-chained.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-chained.ll
@@ -26,12 +26,9 @@ define i32 @chained_partial_reduce_add_sub(ptr %a, ptr %b, ptr %c, i32 %N) #0 {
 ; CHECK-NEON-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-NEON-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-NEON-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP2]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
-; CHECK-NEON-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP3]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP6]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEON-NEXT:    [[TMP7:%.*]] = sext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-NEON-NEXT:    [[TMP8:%.*]] = sext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-NEON-NEXT:    [[TMP9:%.*]] = sext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
@@ -73,12 +70,9 @@ define i32 @chained_partial_reduce_add_sub(ptr %a, ptr %b, ptr %c, i32 %N) #0 {
 ; CHECK-SVE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-SVE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-SVE-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-SVE-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
-; CHECK-SVE-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP11]], align 1
-; CHECK-SVE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i8>, ptr [[TMP12]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP8]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i8>, ptr [[TMP9]], align 1
 ; CHECK-SVE-NEXT:    [[TMP13:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; CHECK-SVE-NEXT:    [[TMP14:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; CHECK-SVE-NEXT:    [[TMP15:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD2]] to <vscale x 4 x i32>
@@ -119,12 +113,9 @@ define i32 @chained_partial_reduce_add_sub(ptr %a, ptr %b, ptr %c, i32 %N) #0 {
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-SVE-MAXBW-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
-; CHECK-SVE-MAXBW-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP11]], align 1
-; CHECK-SVE-MAXBW-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP12]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP7]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP8]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP9]], align 1
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP13:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP14:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD1]] to <vscale x 8 x i32>
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP15:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD2]] to <vscale x 8 x i32>
@@ -192,12 +183,9 @@ define i32 @chained_partial_reduce_add_add(ptr %a, ptr %b, ptr %c, i32 %N) #0 {
 ; CHECK-NEON-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-NEON-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-NEON-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP2]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
-; CHECK-NEON-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP3]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP6]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEON-NEXT:    [[TMP7:%.*]] = sext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-NEON-NEXT:    [[TMP8:%.*]] = sext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-NEON-NEXT:    [[TMP9:%.*]] = sext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
@@ -238,12 +226,9 @@ define i32 @chained_partial_reduce_add_add(ptr %a, ptr %b, ptr %c, i32 %N) #0 {
 ; CHECK-SVE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-SVE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-SVE-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-SVE-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
-; CHECK-SVE-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP11]], align 1
-; CHECK-SVE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i8>, ptr [[TMP12]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP8]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i8>, ptr [[TMP9]], align 1
 ; CHECK-SVE-NEXT:    [[TMP13:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; CHECK-SVE-NEXT:    [[TMP14:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; CHECK-SVE-NEXT:    [[TMP15:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD2]] to <vscale x 4 x i32>
@@ -284,12 +269,9 @@ define i32 @chained_partial_reduce_add_add(ptr %a, ptr %b, ptr %c, i32 %N) #0 {
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-SVE-MAXBW-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
-; CHECK-SVE-MAXBW-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP11]], align 1
-; CHECK-SVE-MAXBW-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP12]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP7]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP8]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP9]], align 1
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP13:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP14:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD1]] to <vscale x 8 x i32>
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP15:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD2]] to <vscale x 8 x i32>
@@ -356,12 +338,9 @@ define i32 @chained_partial_reduce_sub_add(ptr %a, ptr %b, ptr %c, i32 %N) #0 {
 ; CHECK-NEON-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-NEON-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-NEON-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP2]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
-; CHECK-NEON-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP3]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP6]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEON-NEXT:    [[TMP7:%.*]] = sext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-NEON-NEXT:    [[TMP8:%.*]] = sext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-NEON-NEXT:    [[TMP9:%.*]] = sext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
@@ -403,12 +382,9 @@ define i32 @chained_partial_reduce_sub_add(ptr %a, ptr %b, ptr %c, i32 %N) #0 {
 ; CHECK-SVE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-SVE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-SVE-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-SVE-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
-; CHECK-SVE-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP11]], align 1
-; CHECK-SVE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i8>, ptr [[TMP12]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP8]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i8>, ptr [[TMP9]], align 1
 ; CHECK-SVE-NEXT:    [[TMP13:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; CHECK-SVE-NEXT:    [[TMP14:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; CHECK-SVE-NEXT:    [[TMP15:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD2]] to <vscale x 4 x i32>
@@ -449,12 +425,9 @@ define i32 @chained_partial_reduce_sub_add(ptr %a, ptr %b, ptr %c, i32 %N) #0 {
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-SVE-MAXBW-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
-; CHECK-SVE-MAXBW-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP11]], align 1
-; CHECK-SVE-MAXBW-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP12]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP7]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP8]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP9]], align 1
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP13:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP14:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD1]] to <vscale x 8 x i32>
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP15:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD2]] to <vscale x 8 x i32>
@@ -524,12 +497,9 @@ define i32 @chained_partial_reduce_sub_sub(ptr %a, ptr %b, ptr %c, i32 %N) #0 {
 ; CHECK-NEON-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-NEON-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-NEON-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP2]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
-; CHECK-NEON-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP3]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP6]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEON-NEXT:    [[TMP7:%.*]] = sext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-NEON-NEXT:    [[TMP8:%.*]] = sext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-NEON-NEXT:    [[TMP9:%.*]] = sext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
@@ -572,12 +542,9 @@ define i32 @chained_partial_reduce_sub_sub(ptr %a, ptr %b, ptr %c, i32 %N) #0 {
 ; CHECK-SVE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-SVE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-SVE-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-SVE-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
-; CHECK-SVE-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP11]], align 1
-; CHECK-SVE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i8>, ptr [[TMP12]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP8]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i8>, ptr [[TMP9]], align 1
 ; CHECK-SVE-NEXT:    [[TMP13:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; CHECK-SVE-NEXT:    [[TMP14:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; CHECK-SVE-NEXT:    [[TMP15:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD2]] to <vscale x 4 x i32>
@@ -618,12 +585,9 @@ define i32 @chained_partial_reduce_sub_sub(ptr %a, ptr %b, ptr %c, i32 %N) #0 {
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-SVE-MAXBW-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
-; CHECK-SVE-MAXBW-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP11]], align 1
-; CHECK-SVE-MAXBW-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP12]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP7]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP8]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP9]], align 1
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP13:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP14:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD1]] to <vscale x 8 x i32>
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP15:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD2]] to <vscale x 8 x i32>
@@ -695,12 +659,9 @@ define i32 @chained_partial_reduce_add_add_add(ptr %a, ptr %b, ptr %c, i32 %N) #
 ; CHECK-NEON-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-NEON-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-NEON-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP2]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
-; CHECK-NEON-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP3]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP6]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEON-NEXT:    [[TMP7:%.*]] = sext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-NEON-NEXT:    [[TMP8:%.*]] = sext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-NEON-NEXT:    [[TMP9:%.*]] = sext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
@@ -743,12 +704,9 @@ define i32 @chained_partial_reduce_add_add_add(ptr %a, ptr %b, ptr %c, i32 %N) #
 ; CHECK-SVE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-SVE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-SVE-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-SVE-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
-; CHECK-SVE-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP11]], align 1
-; CHECK-SVE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i8>, ptr [[TMP12]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP8]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i8>, ptr [[TMP9]], align 1
 ; CHECK-SVE-NEXT:    [[TMP13:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; CHECK-SVE-NEXT:    [[TMP14:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; CHECK-SVE-NEXT:    [[TMP15:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD2]] to <vscale x 4 x i32>
@@ -791,12 +749,9 @@ define i32 @chained_partial_reduce_add_add_add(ptr %a, ptr %b, ptr %c, i32 %N) #
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-SVE-MAXBW-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
-; CHECK-SVE-MAXBW-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP11]], align 1
-; CHECK-SVE-MAXBW-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP12]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP7]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP8]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP9]], align 1
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP13:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP14:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD1]] to <vscale x 8 x i32>
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP15:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD2]] to <vscale x 8 x i32>
@@ -870,12 +825,9 @@ define i32 @chained_partial_reduce_sub_add_sub(ptr %a, ptr %b, ptr %c, i32 %N) #
 ; CHECK-NEON-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-NEON-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-NEON-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP2]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
-; CHECK-NEON-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP3]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP6]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEON-NEXT:    [[TMP7:%.*]] = sext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-NEON-NEXT:    [[TMP8:%.*]] = sext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-NEON-NEXT:    [[TMP9:%.*]] = sext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
@@ -920,12 +872,9 @@ define i32 @chained_partial_reduce_sub_add_sub(ptr %a, ptr %b, ptr %c, i32 %N) #
 ; CHECK-SVE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-SVE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-SVE-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-SVE-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
-; CHECK-SVE-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP11]], align 1
-; CHECK-SVE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i8>, ptr [[TMP12]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP8]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i8>, ptr [[TMP9]], align 1
 ; CHECK-SVE-NEXT:    [[TMP13:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; CHECK-SVE-NEXT:    [[TMP14:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; CHECK-SVE-NEXT:    [[TMP15:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD2]] to <vscale x 4 x i32>
@@ -968,12 +917,9 @@ define i32 @chained_partial_reduce_sub_add_sub(ptr %a, ptr %b, ptr %c, i32 %N) #
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-SVE-MAXBW-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
-; CHECK-SVE-MAXBW-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP11]], align 1
-; CHECK-SVE-MAXBW-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP12]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP7]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP8]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP9]], align 1
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP13:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP14:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD1]] to <vscale x 8 x i32>
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP15:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD2]] to <vscale x 8 x i32>
@@ -1050,12 +996,9 @@ define i32 @chained_partial_reduce_madd_extadd(ptr %a, ptr %b, ptr %c, i32 %N) #
 ; CHECK-NEON-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-NEON-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-NEON-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP2]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
-; CHECK-NEON-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP3]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP6]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEON-NEXT:    [[TMP7:%.*]] = sext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-NEON-NEXT:    [[TMP8:%.*]] = sext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-NEON-NEXT:    [[TMP9:%.*]] = sext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
@@ -1095,12 +1038,9 @@ define i32 @chained_partial_reduce_madd_extadd(ptr %a, ptr %b, ptr %c, i32 %N) #
 ; CHECK-SVE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-SVE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-SVE-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-SVE-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
-; CHECK-SVE-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP11]], align 1
-; CHECK-SVE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i8>, ptr [[TMP12]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP8]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i8>, ptr [[TMP9]], align 1
 ; CHECK-SVE-NEXT:    [[TMP13:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; CHECK-SVE-NEXT:    [[TMP14:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; CHECK-SVE-NEXT:    [[TMP15:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD2]] to <vscale x 4 x i32>
@@ -1140,12 +1080,9 @@ define i32 @chained_partial_reduce_madd_extadd(ptr %a, ptr %b, ptr %c, i32 %N) #
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-SVE-MAXBW-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
-; CHECK-SVE-MAXBW-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP11]], align 1
-; CHECK-SVE-MAXBW-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP12]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP7]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP8]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP9]], align 1
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP13:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP14:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD1]] to <vscale x 8 x i32>
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP15:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD2]] to <vscale x 8 x i32>
@@ -1210,10 +1147,8 @@ define i32 @chained_partial_reduce_extadd_extadd(ptr %a, ptr %b, i32 %N) #0 {
 ; CHECK-NEON-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE2:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEON-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-NEON-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEON-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP2]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
 ; CHECK-NEON-NEXT:    [[TMP5:%.*]] = sext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-NEON-NEXT:    [[TMP6:%.*]] = sext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-NEON-NEXT:    [[PARTIAL_REDUCE:%.*]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI]], <16 x i32> [[TMP5]])
@@ -1250,10 +1185,8 @@ define i32 @chained_partial_reduce_extadd_extadd(ptr %a, ptr %b, i32 %N) #0 {
 ; CHECK-SVE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-SVE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-SVE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-SVE-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP9]], align 1
-; CHECK-SVE-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP8]], align 1
 ; CHECK-SVE-NEXT:    [[TMP11:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; CHECK-SVE-NEXT:    [[TMP12:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; CHECK-SVE-NEXT:    [[TMP13:%.*]] = add <vscale x 4 x i32> [[VEC_PHI]], [[TMP11]]
@@ -1290,10 +1223,8 @@ define i32 @chained_partial_reduce_extadd_extadd(ptr %a, ptr %b, i32 %N) #0 {
 ; CHECK-SVE-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE2:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-SVE-MAXBW-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP9]], align 1
-; CHECK-SVE-MAXBW-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP7]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP8]], align 1
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP11:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP12:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD1]] to <vscale x 8 x i32>
 ; CHECK-SVE-MAXBW-NEXT:    [[PARTIAL_REDUCE:%.*]] = call <vscale x 2 x i32> @llvm.experimental.vector.partial.reduce.add.nxv2i32.nxv8i32(<vscale x 2 x i32> [[VEC_PHI]], <vscale x 8 x i32> [[TMP11]])
@@ -1354,12 +1285,9 @@ define i32 @chained_partial_reduce_extadd_madd(ptr %a, ptr %b, ptr %c, i32 %N) #
 ; CHECK-NEON-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-NEON-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-NEON-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-NEON-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP1]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
-; CHECK-NEON-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP2]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
-; CHECK-NEON-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP3]], i32 0
-; CHECK-NEON-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP6]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEON-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEON-NEXT:    [[TMP7:%.*]] = sext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-NEON-NEXT:    [[TMP8:%.*]] = sext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-NEON-NEXT:    [[TMP9:%.*]] = sext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
@@ -1399,12 +1327,9 @@ define i32 @chained_partial_reduce_extadd_madd(ptr %a, ptr %b, ptr %c, i32 %N) #
 ; CHECK-SVE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-SVE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-SVE-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-SVE-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
-; CHECK-SVE-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP11]], align 1
-; CHECK-SVE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i32 0
-; CHECK-SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i8>, ptr [[TMP12]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP8]], align 1
+; CHECK-SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i8>, ptr [[TMP9]], align 1
 ; CHECK-SVE-NEXT:    [[TMP13:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; CHECK-SVE-NEXT:    [[TMP14:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; CHECK-SVE-NEXT:    [[TMP15:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD2]] to <vscale x 4 x i32>
@@ -1444,12 +1369,9 @@ define i32 @chained_partial_reduce_extadd_madd(ptr %a, ptr %b, ptr %c, i32 %N) #
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[INDEX]]
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[C]], i64 [[INDEX]]
-; CHECK-SVE-MAXBW-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
-; CHECK-SVE-MAXBW-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP11]], align 1
-; CHECK-SVE-MAXBW-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP9]], i32 0
-; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP12]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP7]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP8]], align 1
+; CHECK-SVE-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP9]], align 1
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP13:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP14:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD1]] to <vscale x 8 x i32>
 ; CHECK-SVE-MAXBW-NEXT:    [[TMP15:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD2]] to <vscale x 8 x i32>
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product-epilogue.ll b/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product-epilogue.ll
index 7090ae8..be8cfa2 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product-epilogue.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product-epilogue.ll
@@ -15,12 +15,10 @@ define i32 @dotp(ptr %a, ptr %b) #0 {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[TMP5:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP7]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[TMP8:%.*]] = zext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-NEXT:    [[TMP9:%.*]] = mul <16 x i32> [[TMP8]], [[TMP5]]
 ; CHECK-NEXT:    [[PARTIAL_REDUCE]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI]], <16 x i32> [[TMP9]])
@@ -29,8 +27,6 @@ define i32 @dotp(ptr %a, ptr %b) #0 {
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP11:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[PARTIAL_REDUCE]])
-; CHECK-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[VEC_EPILOG_PH]]
-; CHECK:       scalar.ph:
 ;
 entry:
   br label %for.body
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product-mixed.ll b/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product-mixed.ll
index 1d37414..d01effd 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product-mixed.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product-mixed.ll
@@ -25,20 +25,18 @@ define i32 @sudot(ptr %a, ptr %b) #0 {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
 ; CHECK-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 8
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[TMP6]], i64 [[TMP9]]
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP7]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
 ; CHECK-NEXT:    [[TMP11:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-NEXT:    [[TMP12:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD2]] to <vscale x 8 x i32>
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[TMP13]], i32 0
 ; CHECK-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP16:%.*]] = mul nuw i64 [[TMP15]], 8
 ; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i8, ptr [[TMP13]], i64 [[TMP16]]
-; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x i8>, ptr [[TMP14]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x i8>, ptr [[TMP13]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x i8>, ptr [[TMP17]], align 1
 ; CHECK-NEXT:    [[TMP18:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD3]] to <vscale x 8 x i32>
 ; CHECK-NEXT:    [[TMP19:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD4]] to <vscale x 8 x i32>
@@ -75,20 +73,18 @@ define i32 @sudot(ptr %a, ptr %b) #0 {
 ; CHECK-NOI8MM-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NOI8MM-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NOI8MM-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-NOI8MM-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
 ; CHECK-NOI8MM-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NOI8MM-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 8
 ; CHECK-NOI8MM-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[TMP6]], i64 [[TMP9]]
-; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP7]], align 1
+; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP6]], align 1
 ; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
 ; CHECK-NOI8MM-NEXT:    [[TMP11:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-NOI8MM-NEXT:    [[TMP12:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD2]] to <vscale x 8 x i32>
 ; CHECK-NOI8MM-NEXT:    [[TMP13:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-NOI8MM-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[TMP13]], i32 0
 ; CHECK-NOI8MM-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NOI8MM-NEXT:    [[TMP16:%.*]] = mul nuw i64 [[TMP15]], 8
 ; CHECK-NOI8MM-NEXT:    [[TMP17:%.*]] = getelementptr i8, ptr [[TMP13]], i64 [[TMP16]]
-; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x i8>, ptr [[TMP14]], align 1
+; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x i8>, ptr [[TMP13]], align 1
 ; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x i8>, ptr [[TMP17]], align 1
 ; CHECK-NOI8MM-NEXT:    [[TMP18:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD3]] to <vscale x 8 x i32>
 ; CHECK-NOI8MM-NEXT:    [[TMP19:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD4]] to <vscale x 8 x i32>
@@ -148,20 +144,18 @@ define i32 @usdot(ptr %a, ptr %b) #0 {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
 ; CHECK-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 8
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[TMP6]], i64 [[TMP9]]
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP7]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
 ; CHECK-NEXT:    [[TMP11:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-NEXT:    [[TMP12:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD2]] to <vscale x 8 x i32>
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[TMP13]], i32 0
 ; CHECK-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP16:%.*]] = mul nuw i64 [[TMP15]], 8
 ; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i8, ptr [[TMP13]], i64 [[TMP16]]
-; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x i8>, ptr [[TMP14]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x i8>, ptr [[TMP13]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x i8>, ptr [[TMP17]], align 1
 ; CHECK-NEXT:    [[TMP18:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD3]] to <vscale x 8 x i32>
 ; CHECK-NEXT:    [[TMP19:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD4]] to <vscale x 8 x i32>
@@ -198,20 +192,18 @@ define i32 @usdot(ptr %a, ptr %b) #0 {
 ; CHECK-NOI8MM-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NOI8MM-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NOI8MM-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-NOI8MM-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
 ; CHECK-NOI8MM-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NOI8MM-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 8
 ; CHECK-NOI8MM-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[TMP6]], i64 [[TMP9]]
-; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP7]], align 1
+; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP6]], align 1
 ; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
 ; CHECK-NOI8MM-NEXT:    [[TMP11:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-NOI8MM-NEXT:    [[TMP12:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD2]] to <vscale x 8 x i32>
 ; CHECK-NOI8MM-NEXT:    [[TMP13:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-NOI8MM-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[TMP13]], i32 0
 ; CHECK-NOI8MM-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NOI8MM-NEXT:    [[TMP16:%.*]] = mul nuw i64 [[TMP15]], 8
 ; CHECK-NOI8MM-NEXT:    [[TMP17:%.*]] = getelementptr i8, ptr [[TMP13]], i64 [[TMP16]]
-; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x i8>, ptr [[TMP14]], align 1
+; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x i8>, ptr [[TMP13]], align 1
 ; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x i8>, ptr [[TMP17]], align 1
 ; CHECK-NOI8MM-NEXT:    [[TMP18:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD3]] to <vscale x 8 x i32>
 ; CHECK-NOI8MM-NEXT:    [[TMP19:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD4]] to <vscale x 8 x i32>
@@ -263,16 +255,14 @@ define i32 @sudot_neon(ptr %a, ptr %b) #1 {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP0]], i32 16
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP0]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-NEXT:    [[TMP4:%.*]] = zext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[TMP5]], i32 0
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP5]], i32 16
-; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <16 x i8>, ptr [[TMP6]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <16 x i8>, ptr [[TMP7]], align 1
 ; CHECK-NEXT:    [[TMP8:%.*]] = sext <16 x i8> [[WIDE_LOAD3]] to <16 x i32>
 ; CHECK-NEXT:    [[TMP9:%.*]] = sext <16 x i8> [[WIDE_LOAD4]] to <16 x i32>
@@ -286,7 +276,7 @@ define i32 @sudot_neon(ptr %a, ptr %b) #1 {
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[PARTIAL_REDUCE5]], [[PARTIAL_REDUCE]]
 ; CHECK-NEXT:    [[TMP13:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
-; CHECK-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ;
 ; CHECK-NOI8MM-LABEL: define i32 @sudot_neon(
@@ -300,16 +290,14 @@ define i32 @sudot_neon(ptr %a, ptr %b) #1 {
 ; CHECK-NOI8MM-NEXT:    [[VEC_PHI:%.*]] = phi <16 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NOI8MM-NEXT:    [[VEC_PHI1:%.*]] = phi <16 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NOI8MM-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-NOI8MM-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
 ; CHECK-NOI8MM-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP0]], i32 16
-; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP0]], align 1
 ; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
 ; CHECK-NOI8MM-NEXT:    [[TMP3:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-NOI8MM-NEXT:    [[TMP4:%.*]] = zext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
 ; CHECK-NOI8MM-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-NOI8MM-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[TMP5]], i32 0
 ; CHECK-NOI8MM-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP5]], i32 16
-; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD3:%.*]] = load <16 x i8>, ptr [[TMP6]], align 1
+; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD3:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
 ; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD4:%.*]] = load <16 x i8>, ptr [[TMP7]], align 1
 ; CHECK-NOI8MM-NEXT:    [[TMP8:%.*]] = sext <16 x i8> [[WIDE_LOAD3]] to <16 x i32>
 ; CHECK-NOI8MM-NEXT:    [[TMP9:%.*]] = sext <16 x i8> [[WIDE_LOAD4]] to <16 x i32>
@@ -323,7 +311,7 @@ define i32 @sudot_neon(ptr %a, ptr %b) #1 {
 ; CHECK-NOI8MM:       middle.block:
 ; CHECK-NOI8MM-NEXT:    [[BIN_RDX:%.*]] = add <16 x i32> [[TMP13]], [[TMP12]]
 ; CHECK-NOI8MM-NEXT:    [[TMP15:%.*]] = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[BIN_RDX]])
-; CHECK-NOI8MM-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NOI8MM-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK-NOI8MM:       scalar.ph:
 ;
 entry:
@@ -360,16 +348,14 @@ define i32 @usdot_neon(ptr %a, ptr %b) #1 {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP0]], i32 16
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP0]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = sext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-NEXT:    [[TMP4:%.*]] = sext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[TMP5]], i32 0
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP5]], i32 16
-; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <16 x i8>, ptr [[TMP6]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <16 x i8>, ptr [[TMP7]], align 1
 ; CHECK-NEXT:    [[TMP8:%.*]] = zext <16 x i8> [[WIDE_LOAD3]] to <16 x i32>
 ; CHECK-NEXT:    [[TMP9:%.*]] = zext <16 x i8> [[WIDE_LOAD4]] to <16 x i32>
@@ -383,7 +369,7 @@ define i32 @usdot_neon(ptr %a, ptr %b) #1 {
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[PARTIAL_REDUCE5]], [[PARTIAL_REDUCE]]
 ; CHECK-NEXT:    [[TMP13:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
-; CHECK-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ;
 ; CHECK-NOI8MM-LABEL: define i32 @usdot_neon(
@@ -397,16 +383,14 @@ define i32 @usdot_neon(ptr %a, ptr %b) #1 {
 ; CHECK-NOI8MM-NEXT:    [[VEC_PHI:%.*]] = phi <16 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NOI8MM-NEXT:    [[VEC_PHI1:%.*]] = phi <16 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NOI8MM-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-NOI8MM-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
 ; CHECK-NOI8MM-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP0]], i32 16
-; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP0]], align 1
 ; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
 ; CHECK-NOI8MM-NEXT:    [[TMP3:%.*]] = sext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-NOI8MM-NEXT:    [[TMP4:%.*]] = sext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
 ; CHECK-NOI8MM-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-NOI8MM-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[TMP5]], i32 0
 ; CHECK-NOI8MM-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP5]], i32 16
-; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD3:%.*]] = load <16 x i8>, ptr [[TMP6]], align 1
+; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD3:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
 ; CHECK-NOI8MM-NEXT:    [[WIDE_LOAD4:%.*]] = load <16 x i8>, ptr [[TMP7]], align 1
 ; CHECK-NOI8MM-NEXT:    [[TMP8:%.*]] = zext <16 x i8> [[WIDE_LOAD3]] to <16 x i32>
 ; CHECK-NOI8MM-NEXT:    [[TMP9:%.*]] = zext <16 x i8> [[WIDE_LOAD4]] to <16 x i32>
@@ -420,7 +404,7 @@ define i32 @usdot_neon(ptr %a, ptr %b) #1 {
 ; CHECK-NOI8MM:       middle.block:
 ; CHECK-NOI8MM-NEXT:    [[BIN_RDX:%.*]] = add <16 x i32> [[TMP13]], [[TMP12]]
 ; CHECK-NOI8MM-NEXT:    [[TMP15:%.*]] = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[BIN_RDX]])
-; CHECK-NOI8MM-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NOI8MM-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK-NOI8MM:       scalar.ph:
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product-neon.ll b/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product-neon.ll
index 0c79086..25a4ab8 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product-neon.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product-neon.ll
@@ -17,12 +17,10 @@ define i32 @dotp(ptr %a, ptr %b) {
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP3:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP4]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP6:%.*]] = zext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP7:%.*]] = mul <16 x i32> [[TMP6]], [[TMP3]]
 ; CHECK-INTERLEAVE1-NEXT:    [[PARTIAL_REDUCE]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI]], <16 x i32> [[TMP7]])
@@ -31,7 +29,7 @@ define i32 @dotp(ptr %a, ptr %b) {
 ; CHECK-INTERLEAVE1-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK-INTERLEAVE1:       middle.block:
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[PARTIAL_REDUCE]])
-; CHECK-INTERLEAVE1-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVE1-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK-INTERLEAVE1:       scalar.ph:
 ;
 ; CHECK-INTERLEAVED-LABEL: define i32 @dotp(
@@ -45,16 +43,14 @@ define i32 @dotp(ptr %a, ptr %b) {
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[TMP1]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP4:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP5:%.*]] = zext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[TMP6]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD3:%.*]] = load <16 x i8>, ptr [[TMP7]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD3:%.*]] = load <16 x i8>, ptr [[TMP6]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD4:%.*]] = load <16 x i8>, ptr [[TMP8]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP9:%.*]] = zext <16 x i8> [[WIDE_LOAD3]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP10:%.*]] = zext <16 x i8> [[WIDE_LOAD4]] to <16 x i32>
@@ -68,7 +64,7 @@ define i32 @dotp(ptr %a, ptr %b) {
 ; CHECK-INTERLEAVED:       middle.block:
 ; CHECK-INTERLEAVED-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[PARTIAL_REDUCE5]], [[PARTIAL_REDUCE]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
-; CHECK-INTERLEAVED-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVED-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK-INTERLEAVED:       scalar.ph:
 ;
 ; CHECK-MAXBW-LABEL: define i32 @dotp(
@@ -81,12 +77,10 @@ define i32 @dotp(ptr %a, ptr %b) {
 ; CHECK-MAXBW-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP3:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP4]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP6:%.*]] = zext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP7:%.*]] = mul <16 x i32> [[TMP6]], [[TMP3]]
 ; CHECK-MAXBW-NEXT:    [[PARTIAL_REDUCE]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI]], <16 x i32> [[TMP7]])
@@ -95,7 +89,7 @@ define i32 @dotp(ptr %a, ptr %b) {
 ; CHECK-MAXBW-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK-MAXBW:       middle.block:
 ; CHECK-MAXBW-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[PARTIAL_REDUCE]])
-; CHECK-MAXBW-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-MAXBW-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK-MAXBW:       scalar.ph:
 ;
 entry:
@@ -147,8 +141,7 @@ define i32 @not_dotp_different_types(ptr %a, ptr %b) {
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP14:%.*]] = add i64 [[INDEX]], 14
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP15:%.*]] = add i64 [[INDEX]], 15
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP16:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP0]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP17:%.*]] = getelementptr i8, ptr [[TMP16]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP17]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP16]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP18:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP19:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP0]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP20:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP1]]
@@ -206,7 +199,7 @@ define i32 @not_dotp_different_types(ptr %a, ptr %b) {
 ; CHECK-INTERLEAVE1-NEXT:    br i1 [[TMP70]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK-INTERLEAVE1:       middle.block:
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP71:%.*]] = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[TMP69]])
-; CHECK-INTERLEAVE1-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVE1-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK-INTERLEAVE1:       scalar.ph:
 ;
 ; CHECK-INTERLEAVED-LABEL: define i32 @not_dotp_different_types(
@@ -252,9 +245,8 @@ define i32 @not_dotp_different_types(ptr %a, ptr %b) {
 ; CHECK-INTERLEAVED-NEXT:    [[TMP30:%.*]] = add i64 [[INDEX]], 30
 ; CHECK-INTERLEAVED-NEXT:    [[TMP31:%.*]] = add i64 [[INDEX]], 31
 ; CHECK-INTERLEAVED-NEXT:    [[TMP32:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP0]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP33:%.*]] = getelementptr i8, ptr [[TMP32]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP34:%.*]] = getelementptr i8, ptr [[TMP32]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP33]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP32]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP34]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP35:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP36:%.*]] = zext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
@@ -366,7 +358,7 @@ define i32 @not_dotp_different_types(ptr %a, ptr %b) {
 ; CHECK-INTERLEAVED:       middle.block:
 ; CHECK-INTERLEAVED-NEXT:    [[BIN_RDX:%.*]] = add <16 x i32> [[TMP138]], [[TMP137]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP140:%.*]] = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[BIN_RDX]])
-; CHECK-INTERLEAVED-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVED-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK-INTERLEAVED:       scalar.ph:
 ;
 ; CHECK-MAXBW-LABEL: define i32 @not_dotp_different_types(
@@ -395,8 +387,7 @@ define i32 @not_dotp_different_types(ptr %a, ptr %b) {
 ; CHECK-MAXBW-NEXT:    [[TMP14:%.*]] = add i64 [[INDEX]], 14
 ; CHECK-MAXBW-NEXT:    [[TMP15:%.*]] = add i64 [[INDEX]], 15
 ; CHECK-MAXBW-NEXT:    [[TMP16:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP0]]
-; CHECK-MAXBW-NEXT:    [[TMP17:%.*]] = getelementptr i8, ptr [[TMP16]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP17]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP16]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP18:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP19:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP0]]
 ; CHECK-MAXBW-NEXT:    [[TMP20:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP1]]
@@ -454,7 +445,7 @@ define i32 @not_dotp_different_types(ptr %a, ptr %b) {
 ; CHECK-MAXBW-NEXT:    br i1 [[TMP70]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK-MAXBW:       middle.block:
 ; CHECK-MAXBW-NEXT:    [[TMP71:%.*]] = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[TMP69]])
-; CHECK-MAXBW-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-MAXBW-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK-MAXBW:       scalar.ph:
 ;
 entry:
@@ -490,12 +481,10 @@ define i32 @not_dotp_not_loop_carried(ptr %a, ptr %b) {
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VECTOR_RECUR:%.*]] = phi <16 x i32> [ <i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP3:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP4]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP6:%.*]] = zext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP7]] = mul <16 x i32> [[TMP6]], [[TMP3]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP8:%.*]] = shufflevector <16 x i32> [[VECTOR_RECUR]], <16 x i32> [[TMP7]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
@@ -504,9 +493,6 @@ define i32 @not_dotp_not_loop_carried(ptr %a, ptr %b) {
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-INTERLEAVE1-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK-INTERLEAVE1:       middle.block:
-; CHECK-INTERLEAVE1-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <16 x i32> [[TMP7]], i32 15
-; CHECK-INTERLEAVE1-NEXT:    [[TMP11:%.*]] = extractelement <16 x i32> [[TMP9]], i32 15
-; CHECK-INTERLEAVE1-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
 ; CHECK-INTERLEAVE1:       scalar.ph:
 ;
 ; CHECK-INTERLEAVED-LABEL: define i32 @not_dotp_not_loop_carried(
@@ -519,12 +505,10 @@ define i32 @not_dotp_not_loop_carried(ptr %a, ptr %b) {
 ; CHECK-INTERLEAVED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VECTOR_RECUR:%.*]] = phi <16 x i32> [ <i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP3:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP4]], i32 0
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP6:%.*]] = zext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP7]] = mul <16 x i32> [[TMP6]], [[TMP3]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP8:%.*]] = shufflevector <16 x i32> [[VECTOR_RECUR]], <16 x i32> [[TMP7]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
@@ -533,9 +517,6 @@ define i32 @not_dotp_not_loop_carried(ptr %a, ptr %b) {
 ; CHECK-INTERLEAVED-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-INTERLEAVED-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK-INTERLEAVED:       middle.block:
-; CHECK-INTERLEAVED-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <16 x i32> [[TMP7]], i32 15
-; CHECK-INTERLEAVED-NEXT:    [[TMP11:%.*]] = extractelement <16 x i32> [[TMP9]], i32 15
-; CHECK-INTERLEAVED-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
 ; CHECK-INTERLEAVED:       scalar.ph:
 ;
 ; CHECK-MAXBW-LABEL: define i32 @not_dotp_not_loop_carried(
@@ -548,12 +529,10 @@ define i32 @not_dotp_not_loop_carried(ptr %a, ptr %b) {
 ; CHECK-MAXBW-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VECTOR_RECUR:%.*]] = phi <16 x i32> [ <i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP3:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP4]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP6:%.*]] = zext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP7]] = mul <16 x i32> [[TMP6]], [[TMP3]]
 ; CHECK-MAXBW-NEXT:    [[TMP8:%.*]] = shufflevector <16 x i32> [[VECTOR_RECUR]], <16 x i32> [[TMP7]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
@@ -562,9 +541,6 @@ define i32 @not_dotp_not_loop_carried(ptr %a, ptr %b) {
 ; CHECK-MAXBW-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-MAXBW-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK-MAXBW:       middle.block:
-; CHECK-MAXBW-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <16 x i32> [[TMP7]], i32 15
-; CHECK-MAXBW-NEXT:    [[TMP11:%.*]] = extractelement <16 x i32> [[TMP9]], i32 15
-; CHECK-MAXBW-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
 ; CHECK-MAXBW:       scalar.ph:
 ;
 entry:
@@ -600,26 +576,22 @@ define i32 @not_dotp_not_phi(ptr %a, ptr noalias %b, ptr noalias %c) {
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VECTOR_RECUR:%.*]] = phi <16 x i32> [ <i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP2:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP11:%.*]] = getelementptr i8, ptr [[TMP4]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP11]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP5:%.*]] = zext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP6:%.*]] = mul <16 x i32> [[TMP5]], [[TMP2]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP7]] = add <16 x i32> [[TMP6]], [[TMP5]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP8:%.*]] = shufflevector <16 x i32> [[VECTOR_RECUR]], <16 x i32> [[TMP7]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP13:%.*]] = getelementptr i32, ptr [[C]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[TMP13]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    store <16 x i32> [[TMP8]], ptr [[TMP10]], align 4
+; CHECK-INTERLEAVE1-NEXT:    store <16 x i32> [[TMP8]], ptr [[TMP13]], align 4
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-INTERLEAVE1-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK-INTERLEAVE1:       middle.block:
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP12:%.*]] = extractelement <16 x i32> [[TMP7]], i32 15
-; CHECK-INTERLEAVE1-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <16 x i32> [[TMP7]], i32 15
-; CHECK-INTERLEAVE1-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVE1-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK-INTERLEAVE1:       scalar.ph:
 ;
 ; CHECK-INTERLEAVED-LABEL: define i32 @not_dotp_not_phi(
@@ -632,26 +604,22 @@ define i32 @not_dotp_not_phi(ptr %a, ptr noalias %b, ptr noalias %c) {
 ; CHECK-INTERLEAVED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VECTOR_RECUR:%.*]] = phi <16 x i32> [ <i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP2:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP11:%.*]] = getelementptr i8, ptr [[TMP4]], i32 0
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP11]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP5:%.*]] = zext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP6:%.*]] = mul <16 x i32> [[TMP5]], [[TMP2]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP7]] = add <16 x i32> [[TMP6]], [[TMP5]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP8:%.*]] = shufflevector <16 x i32> [[VECTOR_RECUR]], <16 x i32> [[TMP7]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP13:%.*]] = getelementptr i32, ptr [[C]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[TMP13]], i32 0
-; CHECK-INTERLEAVED-NEXT:    store <16 x i32> [[TMP8]], ptr [[TMP10]], align 4
+; CHECK-INTERLEAVED-NEXT:    store <16 x i32> [[TMP8]], ptr [[TMP13]], align 4
 ; CHECK-INTERLEAVED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; CHECK-INTERLEAVED-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-INTERLEAVED-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK-INTERLEAVED:       middle.block:
 ; CHECK-INTERLEAVED-NEXT:    [[TMP12:%.*]] = extractelement <16 x i32> [[TMP7]], i32 15
-; CHECK-INTERLEAVED-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <16 x i32> [[TMP7]], i32 15
-; CHECK-INTERLEAVED-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVED-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK-INTERLEAVED:       scalar.ph:
 ;
 ; CHECK-MAXBW-LABEL: define i32 @not_dotp_not_phi(
@@ -664,26 +632,22 @@ define i32 @not_dotp_not_phi(ptr %a, ptr noalias %b, ptr noalias %c) {
 ; CHECK-MAXBW-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VECTOR_RECUR:%.*]] = phi <16 x i32> [ <i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP2:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP11:%.*]] = getelementptr i8, ptr [[TMP4]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP11]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP5:%.*]] = zext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP6:%.*]] = mul <16 x i32> [[TMP5]], [[TMP2]]
 ; CHECK-MAXBW-NEXT:    [[TMP7]] = add <16 x i32> [[TMP6]], [[TMP5]]
 ; CHECK-MAXBW-NEXT:    [[TMP8:%.*]] = shufflevector <16 x i32> [[VECTOR_RECUR]], <16 x i32> [[TMP7]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
 ; CHECK-MAXBW-NEXT:    [[TMP13:%.*]] = getelementptr i32, ptr [[C]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[TMP13]], i32 0
-; CHECK-MAXBW-NEXT:    store <16 x i32> [[TMP8]], ptr [[TMP10]], align 4
+; CHECK-MAXBW-NEXT:    store <16 x i32> [[TMP8]], ptr [[TMP13]], align 4
 ; CHECK-MAXBW-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; CHECK-MAXBW-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-MAXBW-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK-MAXBW:       middle.block:
 ; CHECK-MAXBW-NEXT:    [[TMP12:%.*]] = extractelement <16 x i32> [[TMP7]], i32 15
-; CHECK-MAXBW-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <16 x i32> [[TMP7]], i32 15
-; CHECK-MAXBW-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-MAXBW-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK-MAXBW:       scalar.ph:
 ;
 entry:
@@ -737,35 +701,27 @@ define i32 @dotp_unrolled(i32 %num_out, i64 %num_in, ptr %a, ptr %b) {
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP9:%.*]] = or disjoint i64 [[INDEX]], 3
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[TMP9]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[TMP9]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP12]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP13:%.*]] = sext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
-; CHECK-INTERLEAVE1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD4:%.*]] = load <16 x i8>, ptr [[TMP14]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD4:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP15:%.*]] = sext <16 x i8> [[WIDE_LOAD4]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP16:%.*]] = mul nsw <16 x i32> [[TMP15]], [[TMP13]]
 ; CHECK-INTERLEAVE1-NEXT:    [[PARTIAL_REDUCE]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI3]], <16 x i32> [[TMP16]])
-; CHECK-INTERLEAVE1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD5:%.*]] = load <16 x i8>, ptr [[TMP17]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD5:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP18:%.*]] = sext <16 x i8> [[WIDE_LOAD5]] to <16 x i32>
-; CHECK-INTERLEAVE1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD6:%.*]] = load <16 x i8>, ptr [[TMP19]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD6:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP20:%.*]] = sext <16 x i8> [[WIDE_LOAD6]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP21:%.*]] = mul nsw <16 x i32> [[TMP18]], [[TMP20]]
 ; CHECK-INTERLEAVE1-NEXT:    [[PARTIAL_REDUCE7]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI2]], <16 x i32> [[TMP21]])
-; CHECK-INTERLEAVE1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD8:%.*]] = load <16 x i8>, ptr [[TMP22]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD8:%.*]] = load <16 x i8>, ptr [[TMP7]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP23:%.*]] = sext <16 x i8> [[WIDE_LOAD8]] to <16 x i32>
-; CHECK-INTERLEAVE1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD9:%.*]] = load <16 x i8>, ptr [[TMP24]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD9:%.*]] = load <16 x i8>, ptr [[TMP8]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP25:%.*]] = sext <16 x i8> [[WIDE_LOAD9]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP26:%.*]] = mul nsw <16 x i32> [[TMP23]], [[TMP25]]
 ; CHECK-INTERLEAVE1-NEXT:    [[PARTIAL_REDUCE10]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI1]], <16 x i32> [[TMP26]])
-; CHECK-INTERLEAVE1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i8, ptr [[TMP10]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD11:%.*]] = load <16 x i8>, ptr [[TMP27]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD11:%.*]] = load <16 x i8>, ptr [[TMP10]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP28:%.*]] = sext <16 x i8> [[WIDE_LOAD11]] to <16 x i32>
-; CHECK-INTERLEAVE1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD12:%.*]] = load <16 x i8>, ptr [[TMP29]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD12:%.*]] = load <16 x i8>, ptr [[TMP11]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP30:%.*]] = sext <16 x i8> [[WIDE_LOAD12]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP31:%.*]] = mul nsw <16 x i32> [[TMP28]], [[TMP30]]
 ; CHECK-INTERLEAVE1-NEXT:    [[PARTIAL_REDUCE13]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI]], <16 x i32> [[TMP31]])
@@ -811,15 +767,13 @@ define i32 @dotp_unrolled(i32 %num_out, i64 %num_in, ptr %a, ptr %b) {
 ; CHECK-INTERLEAVED-NEXT:    [[TMP9:%.*]] = or disjoint i64 [[INDEX]], 3
 ; CHECK-INTERLEAVED-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[TMP9]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[TMP9]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP38:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP12]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD8:%.*]] = load <16 x i8>, ptr [[TMP38]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP13:%.*]] = sext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP14:%.*]] = sext <16 x i8> [[WIDE_LOAD8]] to <16 x i32>
-; CHECK-INTERLEAVED-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD9:%.*]] = load <16 x i8>, ptr [[TMP15]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD9:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD10:%.*]] = load <16 x i8>, ptr [[TMP16]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP17:%.*]] = sext <16 x i8> [[WIDE_LOAD9]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP18:%.*]] = sext <16 x i8> [[WIDE_LOAD10]] to <16 x i32>
@@ -827,15 +781,13 @@ define i32 @dotp_unrolled(i32 %num_out, i64 %num_in, ptr %a, ptr %b) {
 ; CHECK-INTERLEAVED-NEXT:    [[TMP20:%.*]] = mul nsw <16 x i32> [[TMP18]], [[TMP14]]
 ; CHECK-INTERLEAVED-NEXT:    [[PARTIAL_REDUCE1]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI6]], <16 x i32> [[TMP19]])
 ; CHECK-INTERLEAVED-NEXT:    [[PARTIAL_REDUCE11]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI7]], <16 x i32> [[TMP20]])
-; CHECK-INTERLEAVED-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD12:%.*]] = load <16 x i8>, ptr [[TMP21]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD12:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD13:%.*]] = load <16 x i8>, ptr [[TMP22]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP23:%.*]] = sext <16 x i8> [[WIDE_LOAD12]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP24:%.*]] = sext <16 x i8> [[WIDE_LOAD13]] to <16 x i32>
-; CHECK-INTERLEAVED-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD14:%.*]] = load <16 x i8>, ptr [[TMP25]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD14:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD15:%.*]] = load <16 x i8>, ptr [[TMP26]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP27:%.*]] = sext <16 x i8> [[WIDE_LOAD14]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP28:%.*]] = sext <16 x i8> [[WIDE_LOAD15]] to <16 x i32>
@@ -843,15 +795,13 @@ define i32 @dotp_unrolled(i32 %num_out, i64 %num_in, ptr %a, ptr %b) {
 ; CHECK-INTERLEAVED-NEXT:    [[TMP30:%.*]] = mul nsw <16 x i32> [[TMP24]], [[TMP28]]
 ; CHECK-INTERLEAVED-NEXT:    [[PARTIAL_REDUCE16]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI4]], <16 x i32> [[TMP29]])
 ; CHECK-INTERLEAVED-NEXT:    [[PARTIAL_REDUCE17]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI5]], <16 x i32> [[TMP30]])
-; CHECK-INTERLEAVED-NEXT:    [[TMP31:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP32:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD18:%.*]] = load <16 x i8>, ptr [[TMP31]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD18:%.*]] = load <16 x i8>, ptr [[TMP7]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD19:%.*]] = load <16 x i8>, ptr [[TMP32]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP33:%.*]] = sext <16 x i8> [[WIDE_LOAD18]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP34:%.*]] = sext <16 x i8> [[WIDE_LOAD19]] to <16 x i32>
-; CHECK-INTERLEAVED-NEXT:    [[TMP35:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP36:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD20:%.*]] = load <16 x i8>, ptr [[TMP35]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD20:%.*]] = load <16 x i8>, ptr [[TMP8]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD21:%.*]] = load <16 x i8>, ptr [[TMP36]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP37:%.*]] = sext <16 x i8> [[WIDE_LOAD20]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP56:%.*]] = sext <16 x i8> [[WIDE_LOAD21]] to <16 x i32>
@@ -859,15 +809,13 @@ define i32 @dotp_unrolled(i32 %num_out, i64 %num_in, ptr %a, ptr %b) {
 ; CHECK-INTERLEAVED-NEXT:    [[TMP40:%.*]] = mul nsw <16 x i32> [[TMP34]], [[TMP56]]
 ; CHECK-INTERLEAVED-NEXT:    [[PARTIAL_REDUCE7]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI2]], <16 x i32> [[TMP39]])
 ; CHECK-INTERLEAVED-NEXT:    [[PARTIAL_REDUCE]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI3]], <16 x i32> [[TMP40]])
-; CHECK-INTERLEAVED-NEXT:    [[TMP41:%.*]] = getelementptr inbounds i8, ptr [[TMP10]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP42:%.*]] = getelementptr inbounds i8, ptr [[TMP10]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD24:%.*]] = load <16 x i8>, ptr [[TMP41]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD24:%.*]] = load <16 x i8>, ptr [[TMP10]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD25:%.*]] = load <16 x i8>, ptr [[TMP42]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP43:%.*]] = sext <16 x i8> [[WIDE_LOAD24]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP44:%.*]] = sext <16 x i8> [[WIDE_LOAD25]] to <16 x i32>
-; CHECK-INTERLEAVED-NEXT:    [[TMP45:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP46:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD26:%.*]] = load <16 x i8>, ptr [[TMP45]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD26:%.*]] = load <16 x i8>, ptr [[TMP11]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD27:%.*]] = load <16 x i8>, ptr [[TMP46]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP47:%.*]] = sext <16 x i8> [[WIDE_LOAD26]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP48:%.*]] = sext <16 x i8> [[WIDE_LOAD27]] to <16 x i32>
@@ -917,35 +865,27 @@ define i32 @dotp_unrolled(i32 %num_out, i64 %num_in, ptr %a, ptr %b) {
 ; CHECK-MAXBW-NEXT:    [[TMP9:%.*]] = or disjoint i64 [[INDEX]], 3
 ; CHECK-MAXBW-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[TMP9]]
 ; CHECK-MAXBW-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[TMP9]]
-; CHECK-MAXBW-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP12]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP13:%.*]] = sext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
-; CHECK-MAXBW-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD4:%.*]] = load <16 x i8>, ptr [[TMP14]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD4:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP15:%.*]] = sext <16 x i8> [[WIDE_LOAD4]] to <16 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP16:%.*]] = mul nsw <16 x i32> [[TMP15]], [[TMP13]]
 ; CHECK-MAXBW-NEXT:    [[PARTIAL_REDUCE]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI3]], <16 x i32> [[TMP16]])
-; CHECK-MAXBW-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD5:%.*]] = load <16 x i8>, ptr [[TMP17]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD5:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP18:%.*]] = sext <16 x i8> [[WIDE_LOAD5]] to <16 x i32>
-; CHECK-MAXBW-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD6:%.*]] = load <16 x i8>, ptr [[TMP19]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD6:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP20:%.*]] = sext <16 x i8> [[WIDE_LOAD6]] to <16 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP21:%.*]] = mul nsw <16 x i32> [[TMP18]], [[TMP20]]
 ; CHECK-MAXBW-NEXT:    [[PARTIAL_REDUCE7]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI2]], <16 x i32> [[TMP21]])
-; CHECK-MAXBW-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD8:%.*]] = load <16 x i8>, ptr [[TMP22]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD8:%.*]] = load <16 x i8>, ptr [[TMP7]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP23:%.*]] = sext <16 x i8> [[WIDE_LOAD8]] to <16 x i32>
-; CHECK-MAXBW-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD9:%.*]] = load <16 x i8>, ptr [[TMP24]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD9:%.*]] = load <16 x i8>, ptr [[TMP8]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP25:%.*]] = sext <16 x i8> [[WIDE_LOAD9]] to <16 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP26:%.*]] = mul nsw <16 x i32> [[TMP23]], [[TMP25]]
 ; CHECK-MAXBW-NEXT:    [[PARTIAL_REDUCE10]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI1]], <16 x i32> [[TMP26]])
-; CHECK-MAXBW-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i8, ptr [[TMP10]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD11:%.*]] = load <16 x i8>, ptr [[TMP27]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD11:%.*]] = load <16 x i8>, ptr [[TMP10]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP28:%.*]] = sext <16 x i8> [[WIDE_LOAD11]] to <16 x i32>
-; CHECK-MAXBW-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD12:%.*]] = load <16 x i8>, ptr [[TMP29]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD12:%.*]] = load <16 x i8>, ptr [[TMP11]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP30:%.*]] = sext <16 x i8> [[WIDE_LOAD12]] to <16 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP31:%.*]] = mul nsw <16 x i32> [[TMP28]], [[TMP30]]
 ; CHECK-MAXBW-NEXT:    [[PARTIAL_REDUCE13]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI]], <16 x i32> [[TMP31]])
@@ -2055,12 +1995,10 @@ define i32 @not_dotp_extend_user(ptr %a, ptr %b) {
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <16 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP3:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP4]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP6:%.*]] = zext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP7:%.*]] = mul <16 x i32> [[TMP6]], [[TMP3]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP8]] = add <16 x i32> [[TMP7]], [[VEC_PHI]]
@@ -2070,7 +2008,7 @@ define i32 @not_dotp_extend_user(ptr %a, ptr %b) {
 ; CHECK-INTERLEAVE1:       middle.block:
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP10:%.*]] = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[TMP8]])
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP11:%.*]] = extractelement <16 x i32> [[TMP6]], i32 15
-; CHECK-INTERLEAVE1-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVE1-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK-INTERLEAVE1:       scalar.ph:
 ;
 ; CHECK-INTERLEAVED-LABEL: define i32 @not_dotp_extend_user(
@@ -2084,16 +2022,14 @@ define i32 @not_dotp_extend_user(ptr %a, ptr %b) {
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI:%.*]] = phi <16 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI1:%.*]] = phi <16 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[TMP1]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP4:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP5:%.*]] = zext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[TMP6]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD3:%.*]] = load <16 x i8>, ptr [[TMP7]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD3:%.*]] = load <16 x i8>, ptr [[TMP6]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD4:%.*]] = load <16 x i8>, ptr [[TMP8]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP9:%.*]] = zext <16 x i8> [[WIDE_LOAD3]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP10:%.*]] = zext <16 x i8> [[WIDE_LOAD4]] to <16 x i32>
@@ -2108,7 +2044,7 @@ define i32 @not_dotp_extend_user(ptr %a, ptr %b) {
 ; CHECK-INTERLEAVED-NEXT:    [[BIN_RDX:%.*]] = add <16 x i32> [[TMP14]], [[TMP13]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP16:%.*]] = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[BIN_RDX]])
 ; CHECK-INTERLEAVED-NEXT:    [[TMP17:%.*]] = extractelement <16 x i32> [[TMP10]], i32 15
-; CHECK-INTERLEAVED-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVED-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK-INTERLEAVED:       scalar.ph:
 ;
 ; CHECK-MAXBW-LABEL: define i32 @not_dotp_extend_user(
@@ -2121,12 +2057,10 @@ define i32 @not_dotp_extend_user(ptr %a, ptr %b) {
 ; CHECK-MAXBW-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <16 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP3:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP4]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP6:%.*]] = zext <16 x i8> [[WIDE_LOAD1]] to <16 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP7:%.*]] = mul <16 x i32> [[TMP6]], [[TMP3]]
 ; CHECK-MAXBW-NEXT:    [[TMP8]] = add <16 x i32> [[TMP7]], [[VEC_PHI]]
@@ -2136,7 +2070,7 @@ define i32 @not_dotp_extend_user(ptr %a, ptr %b) {
 ; CHECK-MAXBW:       middle.block:
 ; CHECK-MAXBW-NEXT:    [[TMP10:%.*]] = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[TMP8]])
 ; CHECK-MAXBW-NEXT:    [[TMP11:%.*]] = extractelement <16 x i32> [[TMP6]], i32 15
-; CHECK-MAXBW-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-MAXBW-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK-MAXBW:       scalar.ph:
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product.ll b/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product.ll
index 14a73db..5218e64 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product.ll
@@ -25,12 +25,10 @@ define i32 @dotp(ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT1:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP16:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX1]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP17:%.*]] = getelementptr i8, ptr [[TMP16]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP17]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP16]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP9:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP20:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX1]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP21:%.*]] = getelementptr i8, ptr [[TMP20]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP21]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP20]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP18:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP13:%.*]] = mul <vscale x 4 x i32> [[TMP18]], [[TMP9]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP14]] = add <vscale x 4 x i32> [[TMP13]], [[VEC_PHI]]
@@ -62,20 +60,18 @@ define i32 @dotp(ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP23:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP24:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP20:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX1]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP21:%.*]] = getelementptr i8, ptr [[TMP20]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP14]], 4
 ; CHECK-INTERLEAVED-NEXT:    [[TMP11:%.*]] = getelementptr i8, ptr [[TMP20]], i64 [[TMP10]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP21]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP20]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i8>, ptr [[TMP11]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP12:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP13:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD2]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP28:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX1]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[TMP28]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP26:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP27:%.*]] = mul nuw i64 [[TMP26]], 4
 ; CHECK-INTERLEAVED-NEXT:    [[TMP18:%.*]] = getelementptr i8, ptr [[TMP28]], i64 [[TMP27]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x i8>, ptr [[TMP9]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x i8>, ptr [[TMP28]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 4 x i8>, ptr [[TMP18]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP19:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD3]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP29:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD4]] to <vscale x 4 x i32>
@@ -111,12 +107,10 @@ define i32 @dotp(ptr %a, ptr %b) #0 {
 ; CHECK-MAXBW-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[TMP7]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP8]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP7]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP13:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD2]] to <vscale x 8 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP15:%.*]] = getelementptr i8, ptr [[TMP14]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x i8>, ptr [[TMP15]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x i8>, ptr [[TMP14]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP20:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD4]] to <vscale x 8 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP22:%.*]] = mul <vscale x 8 x i32> [[TMP20]], [[TMP13]]
 ; CHECK-MAXBW-NEXT:    [[PARTIAL_REDUCE5]] = call <vscale x 2 x i32> @llvm.experimental.vector.partial.reduce.add.nxv2i32.nxv8i32(<vscale x 2 x i32> [[VEC_PHI1]], <vscale x 8 x i32> [[TMP22]])
@@ -173,11 +167,9 @@ define i64 @not_dotp_i8_to_i64_has_neon_dotprod(ptr readonly %a, ptr readonly %b
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-INTERLEAVE1-NEXT:    [[NEXT_GEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i8>, ptr [[TMP4]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i8>, ptr [[NEXT_GEP]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP11:%.*]] = zext <vscale x 2 x i8> [[WIDE_LOAD]] to <vscale x 2 x i64>
-; CHECK-INTERLEAVE1-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[NEXT_GEP1]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 2 x i8>, ptr [[TMP6]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 2 x i8>, ptr [[NEXT_GEP1]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP13:%.*]] = zext <vscale x 2 x i8> [[WIDE_LOAD2]] to <vscale x 2 x i64>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP14:%.*]] = mul nuw nsw <vscale x 2 x i64> [[TMP13]], [[TMP11]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP15]] = add <vscale x 2 x i64> [[TMP14]], [[VEC_PHI]]
@@ -212,19 +204,17 @@ define i64 @not_dotp_i8_to_i64_has_neon_dotprod(ptr readonly %a, ptr readonly %b
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP25:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-INTERLEAVED-NEXT:    [[NEXT_GEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 2
 ; CHECK-INTERLEAVED-NEXT:    [[TMP13:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i64 [[TMP12]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i8>, ptr [[TMP4]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i8>, ptr [[NEXT_GEP]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 2 x i8>, ptr [[TMP13]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP14:%.*]] = zext <vscale x 2 x i8> [[WIDE_LOAD]] to <vscale x 2 x i64>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP15:%.*]] = zext <vscale x 2 x i8> [[WIDE_LOAD3]] to <vscale x 2 x i64>
-; CHECK-INTERLEAVED-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[NEXT_GEP1]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP17:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP18:%.*]] = mul nuw i64 [[TMP17]], 2
 ; CHECK-INTERLEAVED-NEXT:    [[TMP19:%.*]] = getelementptr i8, ptr [[NEXT_GEP1]], i64 [[TMP18]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 2 x i8>, ptr [[TMP6]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 2 x i8>, ptr [[NEXT_GEP1]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 2 x i8>, ptr [[TMP19]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP20:%.*]] = zext <vscale x 2 x i8> [[WIDE_LOAD4]] to <vscale x 2 x i64>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP21:%.*]] = zext <vscale x 2 x i8> [[WIDE_LOAD5]] to <vscale x 2 x i64>
@@ -263,11 +253,9 @@ define i64 @not_dotp_i8_to_i64_has_neon_dotprod(ptr readonly %a, ptr readonly %b
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-MAXBW-NEXT:    [[NEXT_GEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[NEXT_GEP]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP11:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i64>
-; CHECK-MAXBW-NEXT:    [[TMP12:%.*]] = getelementptr i8, ptr [[NEXT_GEP1]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP12]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[NEXT_GEP1]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP13:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD2]] to <vscale x 8 x i64>
 ; CHECK-MAXBW-NEXT:    [[TMP14:%.*]] = mul nuw nsw <vscale x 8 x i64> [[TMP13]], [[TMP11]]
 ; CHECK-MAXBW-NEXT:    [[TMP15]] = add <vscale x 8 x i64> [[TMP14]], [[VEC_PHI]]
@@ -330,11 +318,9 @@ define i64 @not_dotp_i16_to_i64_has_neon_dotprod(ptr readonly %a, ptr readonly %
 ; CHECK-INTERLEAVE1-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[OFFSET_IDX]]
 ; CHECK-INTERLEAVE1-NEXT:    [[OFFSET_IDX1:%.*]] = mul i64 [[INDEX]], 2
 ; CHECK-INTERLEAVE1-NEXT:    [[NEXT_GEP2:%.*]] = getelementptr i8, ptr [[B]], i64 [[OFFSET_IDX1]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP4:%.*]] = getelementptr i16, ptr [[NEXT_GEP]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i16>, ptr [[TMP4]], align 2
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i16>, ptr [[NEXT_GEP]], align 2
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP13:%.*]] = zext <vscale x 2 x i16> [[WIDE_LOAD]] to <vscale x 2 x i64>
-; CHECK-INTERLEAVE1-NEXT:    [[TMP6:%.*]] = getelementptr i16, ptr [[NEXT_GEP2]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 2 x i16>, ptr [[TMP6]], align 2
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 2 x i16>, ptr [[NEXT_GEP2]], align 2
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP15:%.*]] = zext <vscale x 2 x i16> [[WIDE_LOAD3]] to <vscale x 2 x i64>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP16:%.*]] = mul nuw nsw <vscale x 2 x i64> [[TMP15]], [[TMP13]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP17]] = add <vscale x 2 x i64> [[TMP16]], [[VEC_PHI]]
@@ -373,19 +359,17 @@ define i64 @not_dotp_i16_to_i64_has_neon_dotprod(ptr readonly %a, ptr readonly %
 ; CHECK-INTERLEAVED-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[OFFSET_IDX]]
 ; CHECK-INTERLEAVED-NEXT:    [[OFFSET_IDX2:%.*]] = mul i64 [[INDEX]], 2
 ; CHECK-INTERLEAVED-NEXT:    [[NEXT_GEP3:%.*]] = getelementptr i8, ptr [[B]], i64 [[OFFSET_IDX2]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP4:%.*]] = getelementptr i16, ptr [[NEXT_GEP]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP14:%.*]] = mul nuw i64 [[TMP13]], 2
 ; CHECK-INTERLEAVED-NEXT:    [[TMP30:%.*]] = getelementptr i16, ptr [[NEXT_GEP]], i64 [[TMP14]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i16>, ptr [[TMP4]], align 2
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i16>, ptr [[NEXT_GEP]], align 2
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 2 x i16>, ptr [[TMP30]], align 2
 ; CHECK-INTERLEAVED-NEXT:    [[TMP16:%.*]] = zext <vscale x 2 x i16> [[WIDE_LOAD]] to <vscale x 2 x i64>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP17:%.*]] = zext <vscale x 2 x i16> [[WIDE_LOAD4]] to <vscale x 2 x i64>
-; CHECK-INTERLEAVED-NEXT:    [[TMP8:%.*]] = getelementptr i16, ptr [[NEXT_GEP3]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP19:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP20:%.*]] = mul nuw i64 [[TMP19]], 2
 ; CHECK-INTERLEAVED-NEXT:    [[TMP21:%.*]] = getelementptr i16, ptr [[NEXT_GEP3]], i64 [[TMP20]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 2 x i16>, ptr [[TMP8]], align 2
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 2 x i16>, ptr [[NEXT_GEP3]], align 2
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 2 x i16>, ptr [[TMP21]], align 2
 ; CHECK-INTERLEAVED-NEXT:    [[TMP22:%.*]] = zext <vscale x 2 x i16> [[WIDE_LOAD5]] to <vscale x 2 x i64>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP23:%.*]] = zext <vscale x 2 x i16> [[WIDE_LOAD6]] to <vscale x 2 x i64>
@@ -428,11 +412,9 @@ define i64 @not_dotp_i16_to_i64_has_neon_dotprod(ptr readonly %a, ptr readonly %
 ; CHECK-MAXBW-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[OFFSET_IDX]]
 ; CHECK-MAXBW-NEXT:    [[OFFSET_IDX1:%.*]] = mul i64 [[INDEX]], 2
 ; CHECK-MAXBW-NEXT:    [[NEXT_GEP2:%.*]] = getelementptr i8, ptr [[B]], i64 [[OFFSET_IDX1]]
-; CHECK-MAXBW-NEXT:    [[TMP12:%.*]] = getelementptr i16, ptr [[NEXT_GEP]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP12]], align 2
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[NEXT_GEP]], align 2
 ; CHECK-MAXBW-NEXT:    [[TMP13:%.*]] = zext <vscale x 4 x i16> [[WIDE_LOAD]] to <vscale x 4 x i64>
-; CHECK-MAXBW-NEXT:    [[TMP14:%.*]] = getelementptr i16, ptr [[NEXT_GEP2]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x i16>, ptr [[TMP14]], align 2
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x i16>, ptr [[NEXT_GEP2]], align 2
 ; CHECK-MAXBW-NEXT:    [[TMP15:%.*]] = zext <vscale x 4 x i16> [[WIDE_LOAD3]] to <vscale x 4 x i64>
 ; CHECK-MAXBW-NEXT:    [[TMP16:%.*]] = mul nuw nsw <vscale x 4 x i64> [[TMP15]], [[TMP13]]
 ; CHECK-MAXBW-NEXT:    [[TMP17]] = add <vscale x 4 x i64> [[TMP16]], [[VEC_PHI]]
@@ -496,8 +478,7 @@ define i32 @not_dotp_different_types(ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP14:%.*]] = add i64 [[INDEX]], 14
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP15:%.*]] = add i64 [[INDEX]], 15
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP16:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP0]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP17:%.*]] = getelementptr i8, ptr [[TMP16]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP17]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP16]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP18:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP19:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP0]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP20:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP1]]
@@ -555,7 +536,7 @@ define i32 @not_dotp_different_types(ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    br i1 [[TMP70]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK-INTERLEAVE1:       middle.block:
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP71:%.*]] = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[TMP69]])
-; CHECK-INTERLEAVE1-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVE1-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK-INTERLEAVE1:       scalar.ph:
 ;
 ; CHECK-INTERLEAVED-LABEL: define i32 @not_dotp_different_types(
@@ -601,9 +582,8 @@ define i32 @not_dotp_different_types(ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[TMP30:%.*]] = add i64 [[INDEX]], 30
 ; CHECK-INTERLEAVED-NEXT:    [[TMP31:%.*]] = add i64 [[INDEX]], 31
 ; CHECK-INTERLEAVED-NEXT:    [[TMP32:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP0]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP33:%.*]] = getelementptr i8, ptr [[TMP32]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP34:%.*]] = getelementptr i8, ptr [[TMP32]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP33]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP32]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP34]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP35:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP36:%.*]] = zext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
@@ -715,7 +695,7 @@ define i32 @not_dotp_different_types(ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVED:       middle.block:
 ; CHECK-INTERLEAVED-NEXT:    [[BIN_RDX:%.*]] = add <16 x i32> [[TMP138]], [[TMP137]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP142:%.*]] = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[BIN_RDX]])
-; CHECK-INTERLEAVED-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVED-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK-INTERLEAVED:       scalar.ph:
 ;
 ; CHECK-MAXBW-LABEL: define i32 @not_dotp_different_types(
@@ -744,8 +724,7 @@ define i32 @not_dotp_different_types(ptr %a, ptr %b) #0 {
 ; CHECK-MAXBW-NEXT:    [[TMP14:%.*]] = add i64 [[INDEX]], 14
 ; CHECK-MAXBW-NEXT:    [[TMP15:%.*]] = add i64 [[INDEX]], 15
 ; CHECK-MAXBW-NEXT:    [[TMP32:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP0]]
-; CHECK-MAXBW-NEXT:    [[TMP33:%.*]] = getelementptr i8, ptr [[TMP32]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP33]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP32]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP36:%.*]] = zext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP37:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP0]]
 ; CHECK-MAXBW-NEXT:    [[TMP38:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP1]]
@@ -803,7 +782,7 @@ define i32 @not_dotp_different_types(ptr %a, ptr %b) #0 {
 ; CHECK-MAXBW-NEXT:    br i1 [[TMP70]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK-MAXBW:       middle.block:
 ; CHECK-MAXBW-NEXT:    [[TMP71:%.*]] = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[TMP138]])
-; CHECK-MAXBW-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-MAXBW-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK-MAXBW:       scalar.ph:
 ;
 entry:
@@ -851,12 +830,10 @@ define i32 @not_dotp_not_loop_carried(ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 8 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP16:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP11:%.*]] = getelementptr i8, ptr [[TMP10]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP11]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP12:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP13:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[TMP13]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP14]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP13]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP15:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD1]] to <vscale x 8 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP16]] = mul <vscale x 8 x i32> [[TMP15]], [[TMP12]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP17:%.*]] = call <vscale x 8 x i32> @llvm.vector.splice.nxv8i32(<vscale x 8 x i32> [[VECTOR_RECUR]], <vscale x 8 x i32> [[TMP16]], i32 -1)
@@ -894,20 +871,18 @@ define i32 @not_dotp_not_loop_carried(ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVED:       vector.body:
 ; CHECK-INTERLEAVED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP11:%.*]] = getelementptr i8, ptr [[TMP10]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP13:%.*]] = mul nuw i64 [[TMP12]], 8
 ; CHECK-INTERLEAVED-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[TMP10]], i64 [[TMP13]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP11]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP14]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP15:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP16:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD1]] to <vscale x 8 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP17:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP18:%.*]] = getelementptr i8, ptr [[TMP17]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP19:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP20:%.*]] = mul nuw i64 [[TMP19]], 8
 ; CHECK-INTERLEAVED-NEXT:    [[TMP21:%.*]] = getelementptr i8, ptr [[TMP17]], i64 [[TMP20]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP18]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP17]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x i8>, ptr [[TMP21]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP22:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD2]] to <vscale x 8 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP23:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD3]] to <vscale x 8 x i32>
@@ -953,12 +928,10 @@ define i32 @not_dotp_not_loop_carried(ptr %a, ptr %b) #0 {
 ; CHECK-MAXBW-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 8 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[TMP25:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP11:%.*]] = getelementptr i8, ptr [[TMP10]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP11]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP16:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD1]] to <vscale x 8 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP17:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP18:%.*]] = getelementptr i8, ptr [[TMP17]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x i8>, ptr [[TMP18]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x i8>, ptr [[TMP17]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP23:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD3]] to <vscale x 8 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP25]] = mul <vscale x 8 x i32> [[TMP23]], [[TMP16]]
 ; CHECK-MAXBW-NEXT:    [[TMP26:%.*]] = call <vscale x 8 x i32> @llvm.vector.splice.nxv8i32(<vscale x 8 x i32> [[VECTOR_RECUR]], <vscale x 8 x i32> [[TMP25]], i32 -1)
@@ -1019,12 +992,10 @@ define i32 @not_dotp_not_phi(ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVE1:       vector.body:
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP11:%.*]] = getelementptr i8, ptr [[TMP10]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP11]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP12:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP13:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[TMP13]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP14]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP13]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP15:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD1]] to <vscale x 8 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP16:%.*]] = mul <vscale x 8 x i32> [[TMP15]], [[TMP12]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP17:%.*]] = add <vscale x 8 x i32> [[TMP16]], [[TMP15]]
@@ -1107,12 +1078,10 @@ define i32 @not_dotp_not_phi(ptr %a, ptr %b) #0 {
 ; CHECK-MAXBW:       vector.body:
 ; CHECK-MAXBW-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP13:%.*]] = getelementptr i8, ptr [[TMP10]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP13]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP14:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP15:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP18:%.*]] = getelementptr i8, ptr [[TMP15]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP18]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP15]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP19:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD1]] to <vscale x 8 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP20:%.*]] = mul <vscale x 8 x i32> [[TMP19]], [[TMP14]]
 ; CHECK-MAXBW-NEXT:    [[TMP21:%.*]] = add <vscale x 8 x i32> [[TMP20]], [[TMP19]]
@@ -1187,35 +1156,27 @@ define i32 @dotp_unrolled(i32 %num_out, i64 %num_in, ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP9:%.*]] = or disjoint i64 [[INDEX]], 3
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[TMP9]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[TMP9]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP12]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP36:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
-; CHECK-INTERLEAVE1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 4 x i8>, ptr [[TMP14]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 4 x i8>, ptr [[TMP2]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP21:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD4]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP38:%.*]] = mul nsw <vscale x 4 x i32> [[TMP21]], [[TMP36]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP23]] = add <vscale x 4 x i32> [[TMP38]], [[VEC_PHI3]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 4 x i8>, ptr [[TMP17]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 4 x i8>, ptr [[TMP4]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP25:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD5]] to <vscale x 4 x i32>
-; CHECK-INTERLEAVE1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 4 x i8>, ptr [[TMP19]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 4 x i8>, ptr [[TMP5]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP42:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD6]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP28:%.*]] = mul nsw <vscale x 4 x i32> [[TMP25]], [[TMP42]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP30]] = add <vscale x 4 x i32> [[TMP28]], [[VEC_PHI2]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD7:%.*]] = load <vscale x 4 x i8>, ptr [[TMP22]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD7:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP31:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD7]] to <vscale x 4 x i32>
-; CHECK-INTERLEAVE1-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD8:%.*]] = load <vscale x 4 x i8>, ptr [[TMP24]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD8:%.*]] = load <vscale x 4 x i8>, ptr [[TMP8]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP33:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD8]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP34:%.*]] = mul nsw <vscale x 4 x i32> [[TMP31]], [[TMP33]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP35]] = add <vscale x 4 x i32> [[TMP34]], [[VEC_PHI1]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i8, ptr [[TMP10]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD9:%.*]] = load <vscale x 4 x i8>, ptr [[TMP27]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD9:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP37:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD9]] to <vscale x 4 x i32>
-; CHECK-INTERLEAVE1-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD10:%.*]] = load <vscale x 4 x i8>, ptr [[TMP29]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD10:%.*]] = load <vscale x 4 x i8>, ptr [[TMP11]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP39:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD10]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP40:%.*]] = mul nsw <vscale x 4 x i32> [[TMP37]], [[TMP39]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP41]] = add <vscale x 4 x i32> [[TMP40]], [[VEC_PHI]]
@@ -1267,19 +1228,17 @@ define i32 @dotp_unrolled(i32 %num_out, i64 %num_in, ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[TMP9:%.*]] = or disjoint i64 [[INDEX]], 3
 ; CHECK-INTERLEAVED-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[TMP9]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[TMP9]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP56:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP20:%.*]] = mul nuw i64 [[TMP56]], 4
 ; CHECK-INTERLEAVED-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i64 [[TMP20]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP12]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD8:%.*]] = load <vscale x 4 x i8>, ptr [[TMP21]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP66:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP23:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD8]] to <vscale x 4 x i32>
-; CHECK-INTERLEAVED-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP25:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP26:%.*]] = mul nuw i64 [[TMP25]], 4
 ; CHECK-INTERLEAVED-NEXT:    [[TMP72:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i64 [[TMP26]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD9:%.*]] = load <vscale x 4 x i8>, ptr [[TMP14]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD9:%.*]] = load <vscale x 4 x i8>, ptr [[TMP2]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD10:%.*]] = load <vscale x 4 x i8>, ptr [[TMP72]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP28:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD9]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP82:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD10]] to <vscale x 4 x i32>
@@ -1287,19 +1246,17 @@ define i32 @dotp_unrolled(i32 %num_out, i64 %num_in, ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[TMP31:%.*]] = mul nsw <vscale x 4 x i32> [[TMP82]], [[TMP23]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP50]] = add <vscale x 4 x i32> [[TMP30]], [[VEC_PHI6]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP33]] = add <vscale x 4 x i32> [[TMP31]], [[VEC_PHI7]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP35:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP36:%.*]] = mul nuw i64 [[TMP35]], 4
 ; CHECK-INTERLEAVED-NEXT:    [[TMP37:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i64 [[TMP36]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD11:%.*]] = load <vscale x 4 x i8>, ptr [[TMP17]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD11:%.*]] = load <vscale x 4 x i8>, ptr [[TMP4]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD12:%.*]] = load <vscale x 4 x i8>, ptr [[TMP37]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP38:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD11]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP39:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD12]] to <vscale x 4 x i32>
-; CHECK-INTERLEAVED-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP41:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP42:%.*]] = mul nuw i64 [[TMP41]], 4
 ; CHECK-INTERLEAVED-NEXT:    [[TMP43:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i64 [[TMP42]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD13:%.*]] = load <vscale x 4 x i8>, ptr [[TMP19]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD13:%.*]] = load <vscale x 4 x i8>, ptr [[TMP5]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD14:%.*]] = load <vscale x 4 x i8>, ptr [[TMP43]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP44:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD13]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP45:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD14]] to <vscale x 4 x i32>
@@ -1307,19 +1264,17 @@ define i32 @dotp_unrolled(i32 %num_out, i64 %num_in, ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[TMP47:%.*]] = mul nsw <vscale x 4 x i32> [[TMP39]], [[TMP45]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP48]] = add <vscale x 4 x i32> [[TMP46]], [[VEC_PHI4]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP49]] = add <vscale x 4 x i32> [[TMP47]], [[VEC_PHI5]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP51:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP52:%.*]] = mul nuw i64 [[TMP51]], 4
 ; CHECK-INTERLEAVED-NEXT:    [[TMP53:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i64 [[TMP52]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD15:%.*]] = load <vscale x 4 x i8>, ptr [[TMP22]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD15:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD16:%.*]] = load <vscale x 4 x i8>, ptr [[TMP53]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP54:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD15]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP55:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD16]] to <vscale x 4 x i32>
-; CHECK-INTERLEAVED-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP57:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP58:%.*]] = mul nuw i64 [[TMP57]], 4
 ; CHECK-INTERLEAVED-NEXT:    [[TMP59:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i64 [[TMP58]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD17:%.*]] = load <vscale x 4 x i8>, ptr [[TMP24]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD17:%.*]] = load <vscale x 4 x i8>, ptr [[TMP8]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD18:%.*]] = load <vscale x 4 x i8>, ptr [[TMP59]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP60:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD17]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP61:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD18]] to <vscale x 4 x i32>
@@ -1327,19 +1282,17 @@ define i32 @dotp_unrolled(i32 %num_out, i64 %num_in, ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[TMP63:%.*]] = mul nsw <vscale x 4 x i32> [[TMP55]], [[TMP61]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP64]] = add <vscale x 4 x i32> [[TMP62]], [[VEC_PHI2]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP65]] = add <vscale x 4 x i32> [[TMP63]], [[VEC_PHI3]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i8, ptr [[TMP10]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP67:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP68:%.*]] = mul nuw i64 [[TMP67]], 4
 ; CHECK-INTERLEAVED-NEXT:    [[TMP69:%.*]] = getelementptr inbounds i8, ptr [[TMP10]], i64 [[TMP68]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD19:%.*]] = load <vscale x 4 x i8>, ptr [[TMP27]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD19:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD20:%.*]] = load <vscale x 4 x i8>, ptr [[TMP69]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP70:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD19]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP71:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD20]] to <vscale x 4 x i32>
-; CHECK-INTERLEAVED-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP73:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP74:%.*]] = mul nuw i64 [[TMP73]], 4
 ; CHECK-INTERLEAVED-NEXT:    [[TMP75:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i64 [[TMP74]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD21:%.*]] = load <vscale x 4 x i8>, ptr [[TMP29]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD21:%.*]] = load <vscale x 4 x i8>, ptr [[TMP11]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD22:%.*]] = load <vscale x 4 x i8>, ptr [[TMP75]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP76:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD21]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP77:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD22]] to <vscale x 4 x i32>
@@ -1395,35 +1348,27 @@ define i32 @dotp_unrolled(i32 %num_out, i64 %num_in, ptr %a, ptr %b) #0 {
 ; CHECK-MAXBW-NEXT:    [[TMP15:%.*]] = or disjoint i64 [[INDEX]], 3
 ; CHECK-MAXBW-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[TMP15]]
 ; CHECK-MAXBW-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[TMP15]]
-; CHECK-MAXBW-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP18]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP7]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP23:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
-; CHECK-MAXBW-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD9:%.*]] = load <vscale x 8 x i8>, ptr [[TMP24]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD9:%.*]] = load <vscale x 8 x i8>, ptr [[TMP8]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP29:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD9]] to <vscale x 8 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP31:%.*]] = mul nsw <vscale x 8 x i32> [[TMP29]], [[TMP23]]
 ; CHECK-MAXBW-NEXT:    [[PARTIAL_REDUCE11]] = call <vscale x 2 x i32> @llvm.experimental.vector.partial.reduce.add.nxv2i32.nxv8i32(<vscale x 2 x i32> [[VEC_PHI7]], <vscale x 8 x i32> [[TMP31]])
-; CHECK-MAXBW-NEXT:    [[TMP32:%.*]] = getelementptr inbounds i8, ptr [[TMP10]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD12:%.*]] = load <vscale x 8 x i8>, ptr [[TMP32]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD12:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP37:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD12]] to <vscale x 8 x i32>
-; CHECK-MAXBW-NEXT:    [[TMP38:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD14:%.*]] = load <vscale x 8 x i8>, ptr [[TMP38]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD14:%.*]] = load <vscale x 8 x i8>, ptr [[TMP11]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP43:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD14]] to <vscale x 8 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP45:%.*]] = mul nsw <vscale x 8 x i32> [[TMP37]], [[TMP43]]
 ; CHECK-MAXBW-NEXT:    [[PARTIAL_REDUCE]] = call <vscale x 2 x i32> @llvm.experimental.vector.partial.reduce.add.nxv2i32.nxv8i32(<vscale x 2 x i32> [[VEC_PHI6]], <vscale x 8 x i32> [[TMP45]])
-; CHECK-MAXBW-NEXT:    [[TMP46:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD18:%.*]] = load <vscale x 8 x i8>, ptr [[TMP46]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD18:%.*]] = load <vscale x 8 x i8>, ptr [[TMP13]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP51:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD18]] to <vscale x 8 x i32>
-; CHECK-MAXBW-NEXT:    [[TMP52:%.*]] = getelementptr inbounds i8, ptr [[TMP14]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD20:%.*]] = load <vscale x 8 x i8>, ptr [[TMP52]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD20:%.*]] = load <vscale x 8 x i8>, ptr [[TMP14]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP57:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD20]] to <vscale x 8 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP59:%.*]] = mul nsw <vscale x 8 x i32> [[TMP51]], [[TMP57]]
 ; CHECK-MAXBW-NEXT:    [[PARTIAL_REDUCE17]] = call <vscale x 2 x i32> @llvm.experimental.vector.partial.reduce.add.nxv2i32.nxv8i32(<vscale x 2 x i32> [[VEC_PHI5]], <vscale x 8 x i32> [[TMP59]])
-; CHECK-MAXBW-NEXT:    [[TMP60:%.*]] = getelementptr inbounds i8, ptr [[TMP16]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD24:%.*]] = load <vscale x 8 x i8>, ptr [[TMP60]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD24:%.*]] = load <vscale x 8 x i8>, ptr [[TMP16]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP65:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD24]] to <vscale x 8 x i32>
-; CHECK-MAXBW-NEXT:    [[TMP66:%.*]] = getelementptr inbounds i8, ptr [[TMP17]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD26:%.*]] = load <vscale x 8 x i8>, ptr [[TMP66]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD26:%.*]] = load <vscale x 8 x i8>, ptr [[TMP17]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP71:%.*]] = sext <vscale x 8 x i8> [[WIDE_LOAD26]] to <vscale x 8 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP73:%.*]] = mul nsw <vscale x 8 x i32> [[TMP65]], [[TMP71]]
 ; CHECK-MAXBW-NEXT:    [[PARTIAL_REDUCE16]] = call <vscale x 2 x i32> @llvm.experimental.vector.partial.reduce.add.nxv2i32.nxv8i32(<vscale x 2 x i32> [[VEC_PHI4]], <vscale x 8 x i32> [[TMP73]])
@@ -1520,12 +1465,10 @@ define i32 @dotp_predicated(i64 %N, ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP19:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i8> @llvm.masked.load.nxv4i8.p0(ptr [[TMP12]], i32 1, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i8> poison)
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i8> @llvm.masked.load.nxv4i8.p0(ptr [[TMP11]], i32 1, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i8> poison)
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP13:%.*]] = sext <vscale x 4 x i8> [[WIDE_MASKED_LOAD]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[TMP14]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 4 x i8> @llvm.masked.load.nxv4i8.p0(ptr [[TMP15]], i32 1, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i8> poison)
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 4 x i8> @llvm.masked.load.nxv4i8.p0(ptr [[TMP14]], i32 1, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i8> poison)
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP16:%.*]] = sext <vscale x 4 x i8> [[WIDE_MASKED_LOAD1]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP17:%.*]] = mul nsw <vscale x 4 x i32> [[TMP16]], [[TMP13]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP18:%.*]] = add <vscale x 4 x i32> [[TMP17]], [[VEC_PHI]]
@@ -1565,12 +1508,10 @@ define i32 @dotp_predicated(i64 %N, ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP19:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i8> @llvm.masked.load.nxv4i8.p0(ptr [[TMP12]], i32 1, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i8> poison)
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i8> @llvm.masked.load.nxv4i8.p0(ptr [[TMP11]], i32 1, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i8> poison)
 ; CHECK-INTERLEAVED-NEXT:    [[TMP13:%.*]] = sext <vscale x 4 x i8> [[WIDE_MASKED_LOAD]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[TMP14]], i32 0
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 4 x i8> @llvm.masked.load.nxv4i8.p0(ptr [[TMP15]], i32 1, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i8> poison)
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 4 x i8> @llvm.masked.load.nxv4i8.p0(ptr [[TMP14]], i32 1, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i8> poison)
 ; CHECK-INTERLEAVED-NEXT:    [[TMP16:%.*]] = sext <vscale x 4 x i8> [[WIDE_MASKED_LOAD1]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP17:%.*]] = mul nsw <vscale x 4 x i32> [[TMP16]], [[TMP13]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP18:%.*]] = add <vscale x 4 x i32> [[TMP17]], [[VEC_PHI]]
@@ -1610,12 +1551,10 @@ define i32 @dotp_predicated(i64 %N, ptr %a, ptr %b) #0 {
 ; CHECK-MAXBW-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr [[TMP12]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i8> poison)
+; CHECK-MAXBW-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr [[TMP11]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i8> poison)
 ; CHECK-MAXBW-NEXT:    [[TMP13:%.*]] = sext <vscale x 16 x i8> [[WIDE_MASKED_LOAD]] to <vscale x 16 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[TMP14]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr [[TMP15]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i8> poison)
+; CHECK-MAXBW-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr [[TMP14]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i8> poison)
 ; CHECK-MAXBW-NEXT:    [[TMP16:%.*]] = sext <vscale x 16 x i8> [[WIDE_MASKED_LOAD1]] to <vscale x 16 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP17:%.*]] = mul nsw <vscale x 16 x i32> [[TMP16]], [[TMP13]]
 ; CHECK-MAXBW-NEXT:    [[TMP18:%.*]] = select <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i32> [[TMP17]], <vscale x 16 x i32> zeroinitializer
@@ -1671,12 +1610,10 @@ define i32 @not_dotp_extend_user(ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[TMP3]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP4]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP3]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP9:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP6]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP12:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP13:%.*]] = mul <vscale x 4 x i32> [[TMP12]], [[TMP9]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP14]] = add <vscale x 4 x i32> [[TMP13]], [[VEC_PHI]]
@@ -1712,20 +1649,18 @@ define i32 @not_dotp_extend_user(ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP23:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP24:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[TMP3]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP15]], 4
 ; CHECK-INTERLEAVED-NEXT:    [[TMP11:%.*]] = getelementptr i8, ptr [[TMP3]], i64 [[TMP10]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP4]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP3]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i8>, ptr [[TMP11]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP12:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP13:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD2]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[TMP8]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP16:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP17:%.*]] = mul nuw i64 [[TMP16]], 4
 ; CHECK-INTERLEAVED-NEXT:    [[TMP18:%.*]] = getelementptr i8, ptr [[TMP8]], i64 [[TMP17]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x i8>, ptr [[TMP9]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x i8>, ptr [[TMP8]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 4 x i8>, ptr [[TMP18]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP19:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD3]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP20:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD4]] to <vscale x 4 x i32>
@@ -1765,12 +1700,10 @@ define i32 @not_dotp_extend_user(ptr %a, ptr %b) #0 {
 ; CHECK-MAXBW-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 8 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP24:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[TMP7]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP8]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP7]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP13:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD2]] to <vscale x 8 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP15:%.*]] = getelementptr i8, ptr [[TMP14]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x i8>, ptr [[TMP15]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x i8>, ptr [[TMP14]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP20:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD4]] to <vscale x 8 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP22:%.*]] = mul <vscale x 8 x i32> [[TMP20]], [[TMP13]]
 ; CHECK-MAXBW-NEXT:    [[TMP24]] = add <vscale x 8 x i32> [[TMP22]], [[VEC_PHI1]]
@@ -1829,13 +1762,11 @@ define i64 @dotp_cost_disagreement(ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i8>, ptr [[TMP8]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i8>, ptr [[TMP7]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP9:%.*]] = zext <vscale x 2 x i8> [[WIDE_LOAD]] to <vscale x 2 x i64>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP10:%.*]] = add nuw nsw i64 [[INDEX]], 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[TMP10]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP11]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 2 x i8>, ptr [[TMP12]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 2 x i8>, ptr [[TMP11]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP13:%.*]] = zext <vscale x 2 x i8> [[WIDE_LOAD1]] to <vscale x 2 x i64>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP14:%.*]] = mul nuw nsw <vscale x 2 x i64> [[TMP13]], [[TMP9]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP15]] = add <vscale x 2 x i64> [[VEC_PHI]], [[TMP14]]
@@ -1868,21 +1799,19 @@ define i64 @dotp_cost_disagreement(ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP24:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP25:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 2
 ; CHECK-INTERLEAVED-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i64 [[TMP10]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i8>, ptr [[TMP8]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i8>, ptr [[TMP7]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 2 x i8>, ptr [[TMP11]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP12:%.*]] = zext <vscale x 2 x i8> [[WIDE_LOAD]] to <vscale x 2 x i64>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP13:%.*]] = zext <vscale x 2 x i8> [[WIDE_LOAD2]] to <vscale x 2 x i64>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP14:%.*]] = add nuw nsw i64 [[INDEX]], 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[TMP14]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP15]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP17:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP18:%.*]] = mul nuw i64 [[TMP17]], 2
 ; CHECK-INTERLEAVED-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP15]], i64 [[TMP18]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 2 x i8>, ptr [[TMP16]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 2 x i8>, ptr [[TMP15]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 2 x i8>, ptr [[TMP19]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP20:%.*]] = zext <vscale x 2 x i8> [[WIDE_LOAD3]] to <vscale x 2 x i64>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP21:%.*]] = zext <vscale x 2 x i8> [[WIDE_LOAD4]] to <vscale x 2 x i64>
@@ -1919,23 +1848,10 @@ define i64 @dotp_cost_disagreement(ptr %a, ptr %b) #0 {
 ; CHECK-MAXBW-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP8]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP7]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP9:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i64>
 ; CHECK-MAXBW-NEXT:    [[TMP10:%.*]] = add nuw nsw i64 [[INDEX]], 1
 ; CHECK-MAXBW-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[TMP10]]
-; CHECK-MAXBW-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP11]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP12]], align 1
-; CHECK-MAXBW-NEXT:    [[TMP17:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD1]] to <vscale x 8 x i64>
-; CHECK-MAXBW-NEXT:    [[TMP13:%.*]] = mul nuw nsw <vscale x 8 x i64> [[TMP17]], [[TMP9]]
-; CHECK-MAXBW-NEXT:    [[TMP14]] = add <vscale x 8 x i64> [[VEC_PHI]], [[TMP13]]
-; CHECK-MAXBW-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
-; CHECK-MAXBW-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-MAXBW-NEXT:    br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
-; CHECK-MAXBW:       middle.block:
-; CHECK-MAXBW-NEXT:    [[TMP16:%.*]] = call i64 @llvm.vector.reduce.add.nxv8i64(<vscale x 8 x i64> [[TMP14]])
-; CHECK-MAXBW-NEXT:    [[CMP_N:%.*]] = icmp eq i64 41, [[N_VEC]]
-; CHECK-MAXBW-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
 ; CHECK-MAXBW:       scalar.ph:
 ;
 entry:
@@ -2155,8 +2071,7 @@ define i64 @not_dotp_ext_outside_plan(ptr %a, i16 %b, i64 %n) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <8 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw i16, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP1]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP2]], align 2
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP1]], align 2
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP3:%.*]] = zext <8 x i16> [[WIDE_LOAD]] to <8 x i64>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP4:%.*]] = mul nuw nsw <8 x i64> [[TMP3]], [[BROADCAST_SPLAT]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP5]] = add <8 x i64> [[TMP4]], [[VEC_PHI]]
@@ -2189,9 +2104,8 @@ define i64 @not_dotp_ext_outside_plan(ptr %a, i16 %b, i64 %n) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI:%.*]] = phi <8 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw i16, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP1]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP1]], i32 8
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP2]], align 2
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP1]], align 2
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i16>, ptr [[TMP3]], align 2
 ; CHECK-INTERLEAVED-NEXT:    [[TMP4:%.*]] = zext <8 x i16> [[WIDE_LOAD]] to <8 x i64>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP5:%.*]] = zext <8 x i16> [[WIDE_LOAD2]] to <8 x i64>
@@ -2234,8 +2148,7 @@ define i64 @not_dotp_ext_outside_plan(ptr %a, i16 %b, i64 %n) #0 {
 ; CHECK-MAXBW-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP11:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i16, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP7]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP8]], align 2
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP7]], align 2
 ; CHECK-MAXBW-NEXT:    [[TMP9:%.*]] = zext <vscale x 4 x i16> [[WIDE_LOAD]] to <vscale x 4 x i64>
 ; CHECK-MAXBW-NEXT:    [[TMP10:%.*]] = mul nuw nsw <vscale x 4 x i64> [[TMP9]], [[BROADCAST_SPLAT]]
 ; CHECK-MAXBW-NEXT:    [[TMP11]] = add <vscale x 4 x i64> [[TMP10]], [[VEC_PHI]]
@@ -2293,8 +2206,7 @@ define i64 @not_dotp_ext_outside_plan2(ptr %a, i16 %b, i64 %n) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <8 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw i16, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP1]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP2]], align 2
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP1]], align 2
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP3:%.*]] = zext <8 x i16> [[WIDE_LOAD]] to <8 x i64>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP4:%.*]] = mul nuw nsw <8 x i64> [[BROADCAST_SPLAT]], [[TMP3]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP5]] = add <8 x i64> [[TMP4]], [[VEC_PHI]]
@@ -2327,9 +2239,8 @@ define i64 @not_dotp_ext_outside_plan2(ptr %a, i16 %b, i64 %n) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI:%.*]] = phi <8 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw i16, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP1]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP1]], i32 8
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP2]], align 2
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP1]], align 2
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i16>, ptr [[TMP3]], align 2
 ; CHECK-INTERLEAVED-NEXT:    [[TMP4:%.*]] = zext <8 x i16> [[WIDE_LOAD]] to <8 x i64>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP5:%.*]] = zext <8 x i16> [[WIDE_LOAD2]] to <8 x i64>
@@ -2372,8 +2283,7 @@ define i64 @not_dotp_ext_outside_plan2(ptr %a, i16 %b, i64 %n) #0 {
 ; CHECK-MAXBW-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP11:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i16, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP7]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP8]], align 2
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP7]], align 2
 ; CHECK-MAXBW-NEXT:    [[TMP9:%.*]] = zext <vscale x 4 x i16> [[WIDE_LOAD]] to <vscale x 4 x i64>
 ; CHECK-MAXBW-NEXT:    [[TMP10:%.*]] = mul nuw nsw <vscale x 4 x i64> [[BROADCAST_SPLAT]], [[TMP9]]
 ; CHECK-MAXBW-NEXT:    [[TMP11]] = add <vscale x 4 x i64> [[TMP10]], [[VEC_PHI]]
@@ -2440,11 +2350,9 @@ define dso_local i32 @not_dotp_vscale1(ptr %a, ptr %b, i32 %n, i64 %cost) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i64> [ [[TMP10]], [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-INTERLEAVE1-NEXT:    [[NEXT_GEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP13:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i8>, ptr [[TMP13]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i8>, ptr [[NEXT_GEP]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP14:%.*]] = zext <vscale x 2 x i8> [[WIDE_LOAD]] to <vscale x 2 x i64>
-; CHECK-INTERLEAVE1-NEXT:    [[TMP15:%.*]] = getelementptr i8, ptr [[NEXT_GEP1]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 2 x i8>, ptr [[TMP15]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 2 x i8>, ptr [[NEXT_GEP1]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP16:%.*]] = zext <vscale x 2 x i8> [[WIDE_LOAD2]] to <vscale x 2 x i64>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP17:%.*]] = mul nuw nsw <vscale x 2 x i64> [[TMP16]], [[TMP14]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP18]] = add <vscale x 2 x i64> [[TMP17]], [[VEC_PHI]]
@@ -2486,19 +2394,17 @@ define dso_local i32 @not_dotp_vscale1(ptr %a, ptr %b, i32 %n, i64 %cost) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP28:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-INTERLEAVED-NEXT:    [[NEXT_GEP2:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP13:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP14]], 2
 ; CHECK-INTERLEAVED-NEXT:    [[TMP16:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i64 [[TMP15]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i8>, ptr [[TMP13]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i8>, ptr [[NEXT_GEP]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 2 x i8>, ptr [[TMP16]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP17:%.*]] = zext <vscale x 2 x i8> [[WIDE_LOAD]] to <vscale x 2 x i64>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP18:%.*]] = zext <vscale x 2 x i8> [[WIDE_LOAD3]] to <vscale x 2 x i64>
-; CHECK-INTERLEAVED-NEXT:    [[TMP19:%.*]] = getelementptr i8, ptr [[NEXT_GEP2]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP20:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP21:%.*]] = mul nuw i64 [[TMP20]], 2
 ; CHECK-INTERLEAVED-NEXT:    [[TMP22:%.*]] = getelementptr i8, ptr [[NEXT_GEP2]], i64 [[TMP21]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 2 x i8>, ptr [[TMP19]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 2 x i8>, ptr [[NEXT_GEP2]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 2 x i8>, ptr [[TMP22]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP23:%.*]] = zext <vscale x 2 x i8> [[WIDE_LOAD4]] to <vscale x 2 x i64>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP24:%.*]] = zext <vscale x 2 x i8> [[WIDE_LOAD5]] to <vscale x 2 x i64>
@@ -2544,11 +2450,9 @@ define dso_local i32 @not_dotp_vscale1(ptr %a, ptr %b, i32 %n, i64 %cost) #0 {
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x i64> [ [[TMP10]], [[VECTOR_PH]] ], [ [[TMP20:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
 ; CHECK-MAXBW-NEXT:    [[NEXT_GEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP13:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP13]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[NEXT_GEP]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP14:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i64>
-; CHECK-MAXBW-NEXT:    [[TMP15:%.*]] = getelementptr i8, ptr [[NEXT_GEP1]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP15]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[NEXT_GEP1]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP16:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD2]] to <vscale x 8 x i64>
 ; CHECK-MAXBW-NEXT:    [[TMP17:%.*]] = mul nuw nsw <vscale x 8 x i64> [[TMP16]], [[TMP14]]
 ; CHECK-MAXBW-NEXT:    [[TMP20]] = add <vscale x 8 x i64> [[TMP17]], [[VEC_PHI]]
@@ -2637,8 +2541,7 @@ define dso_local void @not_dotp_high_register_pressure(ptr %a, ptr %b, ptr %sum,
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI6:%.*]] = phi <4 x i32> [ [[TMP6]], [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI7:%.*]] = phi <4 x i32> [ [[TMP7]], [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP9]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP8]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP10:%.*]] = sext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP11:%.*]] = shl nsw i64 [[INDEX]], 3
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[TMP11]]
@@ -2738,8 +2641,7 @@ define dso_local void @not_dotp_high_register_pressure(ptr %a, ptr %b, ptr %sum,
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI6:%.*]] = phi <4 x i32> [ [[TMP6]], [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE15:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI7:%.*]] = phi <4 x i32> [ [[TMP7]], [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP9]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP8]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP10:%.*]] = sext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP11:%.*]] = shl nsw i64 [[INDEX]], 3
 ; CHECK-INTERLEAVED-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[TMP11]]
@@ -2839,8 +2741,7 @@ define dso_local void @not_dotp_high_register_pressure(ptr %a, ptr %b, ptr %sum,
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI6:%.*]] = phi <4 x i32> [ [[TMP6]], [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE15:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI7:%.*]] = phi <4 x i32> [ [[TMP7]], [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP8]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP9]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP8]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP10:%.*]] = sext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP11:%.*]] = shl nsw i64 [[INDEX]], 3
 ; CHECK-MAXBW-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[TMP11]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-interleave.ll b/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-interleave.ll
index 3515365..e24b47d 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-interleave.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-interleave.ll
@@ -20,9 +20,8 @@ define i32 @partial_reduce_with_non_constant_start_value(ptr %src, i32 %rdx.star
 ; IC2-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ [[TMP0]], %[[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], %[[VECTOR_BODY]] ]
 ; IC2-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i32> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[PARTIAL_REDUCE3:%.*]], %[[VECTOR_BODY]] ]
 ; IC2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[INDEX]]
-; IC2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
 ; IC2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 16
-; IC2-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; IC2-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; IC2-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; IC2-NEXT:    [[TMP4:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; IC2-NEXT:    [[TMP5:%.*]] = zext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
@@ -74,11 +73,10 @@ define i32 @partial_reduce_with_non_constant_start_value(ptr %src, i32 %rdx.star
 ; IC4-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x i32> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[PARTIAL_REDUCE8:%.*]], %[[VECTOR_BODY]] ]
 ; IC4-NEXT:    [[VEC_PHI3:%.*]] = phi <4 x i32> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[PARTIAL_REDUCE9:%.*]], %[[VECTOR_BODY]] ]
 ; IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[INDEX]]
-; IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
 ; IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 16
 ; IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 32
 ; IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 48
-; IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
 ; IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-no-dotprod.ll b/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-no-dotprod.ll
index 09d09e2..ae33e46 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-no-dotprod.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-no-dotprod.ll
@@ -16,16 +16,14 @@ define i32 @not_dotp(ptr %a, ptr %b) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <16 x i32> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP13:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <16 x i32> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP14:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[TMP1]], i32 16
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-NEXT:    [[TMP5:%.*]] = zext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[TMP6]], i32 16
-; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <16 x i8>, ptr [[TMP7]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <16 x i8>, ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <16 x i8>, ptr [[TMP8]], align 1
 ; CHECK-NEXT:    [[TMP9:%.*]] = zext <16 x i8> [[WIDE_LOAD3]] to <16 x i32>
 ; CHECK-NEXT:    [[TMP10:%.*]] = zext <16 x i8> [[WIDE_LOAD4]] to <16 x i32>
@@ -39,7 +37,7 @@ define i32 @not_dotp(ptr %a, ptr %b) {
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[BIN_RDX:%.*]] = add <16 x i32> [[TMP14]], [[TMP13]]
 ; CHECK-NEXT:    [[TMP16:%.*]] = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[BIN_RDX]])
-; CHECK-NEXT:    br i1 false, [[FOR_EXIT:label %.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ;
 entry:
@@ -69,7 +67,7 @@ for.exit:                        ; preds = %for.body
 define i40 @partial_reduce_not_known_factor(i32 %a, i32 %b, i16 %N) {
 ; CHECK-LABEL: define i40 @partial_reduce_not_known_factor(
 ; CHECK-SAME: i32 [[A:%.*]], i32 [[B:%.*]], i16 [[N:%.*]]) #[[ATTR0]] {
-; CHECK-NEXT:  [[ENTRY:.*]]:
+; CHECK-NEXT:  [[ENTRY:.*:]]
 ; CHECK-NEXT:    [[SMAX:%.*]] = call i16 @llvm.smax.i16(i16 [[N]], i16 0)
 ; CHECK-NEXT:    [[TMP0:%.*]] = zext nneg i16 [[SMAX]] to i32
 ; CHECK-NEXT:    [[TMP1:%.*]] = add nuw nsw i32 [[TMP0]], 1
@@ -100,7 +98,9 @@ define i40 @partial_reduce_not_known_factor(i32 %a, i32 %b, i16 %N) {
 ; CHECK-NEXT:    [[BIN_RDX:%.*]] = or <2 x i40> [[TMP8]], [[TMP6]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = call i40 @llvm.vector.reduce.or.v2i40(<2 x i40> [[BIN_RDX]])
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP1]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br i1 [[CMP_N]], [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; CHECK:       [[SCALAR_PH]]:
+;
 entry:
   br label %for.body
 
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-sub.ll b/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-sub.ll
index a471c00..a46340c 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-sub.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-sub.ll
@@ -25,12 +25,10 @@ define i32 @dotp(ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[TMP7]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP8]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP9:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP11:%.*]] = getelementptr i8, ptr [[TMP10]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP11]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP12:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP13:%.*]] = mul <vscale x 4 x i32> [[TMP12]], [[TMP9]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP14:%.*]] = sub <vscale x 4 x i32> zeroinitializer, [[TMP13]]
@@ -63,20 +61,18 @@ define i32 @dotp(ptr %a, ptr %b) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP25:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP26:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[TMP7]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
 ; CHECK-INTERLEAVED-NEXT:    [[TMP11:%.*]] = getelementptr i8, ptr [[TMP7]], i64 [[TMP10]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP8]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i8>, ptr [[TMP11]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP12:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP13:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD2]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP15:%.*]] = getelementptr i8, ptr [[TMP14]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP16:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-INTERLEAVED-NEXT:    [[TMP17:%.*]] = mul nuw i64 [[TMP16]], 4
 ; CHECK-INTERLEAVED-NEXT:    [[TMP18:%.*]] = getelementptr i8, ptr [[TMP14]], i64 [[TMP17]]
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x i8>, ptr [[TMP15]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x i8>, ptr [[TMP14]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 4 x i8>, ptr [[TMP18]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP19:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD3]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP20:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD4]] to <vscale x 4 x i32>
@@ -114,12 +110,10 @@ define i32 @dotp(ptr %a, ptr %b) #0 {
 ; CHECK-MAXBW-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i32> [ zeroinitializer, [[ENTRY]] ], [ [[PARTIAL_REDUCE:%.*]], [[FOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[A]], i64 [[IV]]
-; CHECK-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[TMP7]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP8]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP7]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP9:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[B]], i64 [[IV]]
-; CHECK-MAXBW-NEXT:    [[TMP11:%.*]] = getelementptr i8, ptr [[TMP10]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP11]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP10]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP12:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD1]] to <vscale x 8 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP13:%.*]] = mul <vscale x 8 x i32> [[TMP12]], [[TMP9]]
 ; CHECK-MAXBW-NEXT:    [[TMP14:%.*]] = sub <vscale x 8 x i32> zeroinitializer, [[TMP13]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce.ll b/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce.ll
index 63eb97a..d2c03d1 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce.ll
@@ -19,8 +19,7 @@ define i32 @zext_add_reduc_i8_i32_sve(ptr %a) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <16 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP0]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP2:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP3]] = add <16 x i32> [[TMP2]], [[VEC_PHI]]
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
@@ -28,7 +27,7 @@ define i32 @zext_add_reduc_i8_i32_sve(ptr %a) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK-INTERLEAVE1:       middle.block:
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP5:%.*]] = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[TMP3]])
-; CHECK-INTERLEAVE1-NEXT:    br i1 false, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVE1-NEXT:    br label [[SCALAR_PH]]
 ; CHECK-INTERLEAVE1:       scalar.ph:
 ;
 ; CHECK-INTERLEAVED-LABEL: define i32 @zext_add_reduc_i8_i32_sve(
@@ -42,9 +41,8 @@ define i32 @zext_add_reduc_i8_i32_sve(ptr %a) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI:%.*]] = phi <16 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI1:%.*]] = phi <16 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP0]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP0]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP3:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP4:%.*]] = zext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
@@ -56,7 +54,7 @@ define i32 @zext_add_reduc_i8_i32_sve(ptr %a) #0 {
 ; CHECK-INTERLEAVED:       middle.block:
 ; CHECK-INTERLEAVED-NEXT:    [[BIN_RDX:%.*]] = add <16 x i32> [[TMP6]], [[TMP5]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP8:%.*]] = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[BIN_RDX]])
-; CHECK-INTERLEAVED-NEXT:    br i1 false, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVED-NEXT:    br label [[SCALAR_PH]]
 ; CHECK-INTERLEAVED:       scalar.ph:
 ;
 ; CHECK-MAXBW-LABEL: define i32 @zext_add_reduc_i8_i32_sve(
@@ -77,8 +75,7 @@ define i32 @zext_add_reduc_i8_i32_sve(ptr %a) #0 {
 ; CHECK-MAXBW-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP7]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP6]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP8:%.*]] = zext <vscale x 16 x i8> [[WIDE_LOAD]] to <vscale x 16 x i32>
 ; CHECK-MAXBW-NEXT:    [[PARTIAL_REDUCE]] = call <vscale x 4 x i32> @llvm.experimental.vector.partial.reduce.add.nxv4i32.nxv16i32(<vscale x 4 x i32> [[VEC_PHI]], <vscale x 16 x i32> [[TMP8]])
 ; CHECK-MAXBW-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -119,8 +116,7 @@ define i32 @zext_add_reduc_i8_i32_neon(ptr %a) #2 {
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP0]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP2:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[PARTIAL_REDUCE]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI]], <16 x i32> [[TMP2]])
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
@@ -128,7 +124,7 @@ define i32 @zext_add_reduc_i8_i32_neon(ptr %a) #2 {
 ; CHECK-INTERLEAVE1-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK-INTERLEAVE1:       middle.block:
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP4:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[PARTIAL_REDUCE]])
-; CHECK-INTERLEAVE1-NEXT:    br i1 false, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVE1-NEXT:    br label [[SCALAR_PH]]
 ; CHECK-INTERLEAVE1:       scalar.ph:
 ;
 ; CHECK-INTERLEAVED-LABEL: define i32 @zext_add_reduc_i8_i32_neon(
@@ -142,9 +138,8 @@ define i32 @zext_add_reduc_i8_i32_neon(ptr %a) #2 {
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP0]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP0]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP3:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP4:%.*]] = zext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
@@ -156,7 +151,7 @@ define i32 @zext_add_reduc_i8_i32_neon(ptr %a) #2 {
 ; CHECK-INTERLEAVED:       middle.block:
 ; CHECK-INTERLEAVED-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[PARTIAL_REDUCE3]], [[PARTIAL_REDUCE]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
-; CHECK-INTERLEAVED-NEXT:    br i1 false, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVED-NEXT:    br label [[SCALAR_PH]]
 ; CHECK-INTERLEAVED:       scalar.ph:
 ;
 ; CHECK-MAXBW-LABEL: define i32 @zext_add_reduc_i8_i32_neon(
@@ -169,8 +164,7 @@ define i32 @zext_add_reduc_i8_i32_neon(ptr %a) #2 {
 ; CHECK-MAXBW-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP0]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP2:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-MAXBW-NEXT:    [[PARTIAL_REDUCE]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI]], <16 x i32> [[TMP2]])
 ; CHECK-MAXBW-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
@@ -178,7 +172,7 @@ define i32 @zext_add_reduc_i8_i32_neon(ptr %a) #2 {
 ; CHECK-MAXBW-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK-MAXBW:       middle.block:
 ; CHECK-MAXBW-NEXT:    [[TMP4:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[PARTIAL_REDUCE]])
-; CHECK-MAXBW-NEXT:    br i1 false, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-MAXBW-NEXT:    br label [[SCALAR_PH]]
 ; CHECK-MAXBW:       scalar.ph:
 ;
 entry:
@@ -211,8 +205,7 @@ define i64 @zext_add_reduc_i8_i64(ptr %a) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <16 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP3:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i64>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP4]] = add <16 x i64> [[TMP3]], [[VEC_PHI]]
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
@@ -220,7 +213,7 @@ define i64 @zext_add_reduc_i8_i64(ptr %a) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK-INTERLEAVE1:       middle.block:
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP6:%.*]] = call i64 @llvm.vector.reduce.add.v16i64(<16 x i64> [[TMP4]])
-; CHECK-INTERLEAVE1-NEXT:    br i1 false, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVE1-NEXT:    br label [[SCALAR_PH]]
 ; CHECK-INTERLEAVE1:       scalar.ph:
 ;
 ; CHECK-INTERLEAVED-LABEL: define i64 @zext_add_reduc_i8_i64(
@@ -234,9 +227,8 @@ define i64 @zext_add_reduc_i8_i64(ptr %a) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI:%.*]] = phi <16 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI1:%.*]] = phi <16 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[TMP1]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP4:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i64>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP5:%.*]] = zext <16 x i8> [[WIDE_LOAD2]] to <16 x i64>
@@ -248,7 +240,7 @@ define i64 @zext_add_reduc_i8_i64(ptr %a) #0 {
 ; CHECK-INTERLEAVED:       middle.block:
 ; CHECK-INTERLEAVED-NEXT:    [[BIN_RDX:%.*]] = add <16 x i64> [[TMP7]], [[TMP6]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vector.reduce.add.v16i64(<16 x i64> [[BIN_RDX]])
-; CHECK-INTERLEAVED-NEXT:    br i1 false, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVED-NEXT:    br label [[SCALAR_PH]]
 ; CHECK-INTERLEAVED:       scalar.ph:
 ;
 ; CHECK-MAXBW-LABEL: define i64 @zext_add_reduc_i8_i64(
@@ -269,8 +261,7 @@ define i64 @zext_add_reduc_i8_i64(ptr %a) #0 {
 ; CHECK-MAXBW-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[TMP7]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP8]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP7]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP9:%.*]] = zext <vscale x 16 x i8> [[WIDE_LOAD]] to <vscale x 16 x i64>
 ; CHECK-MAXBW-NEXT:    [[PARTIAL_REDUCE]] = call <vscale x 2 x i64> @llvm.experimental.vector.partial.reduce.add.nxv2i64.nxv16i64(<vscale x 2 x i64> [[VEC_PHI]], <vscale x 16 x i64> [[TMP9]])
 ; CHECK-MAXBW-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -312,8 +303,7 @@ define i64 @zext_add_reduc_i16_i64(ptr %a) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <8 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP1:%.*]] = getelementptr i16, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP2:%.*]] = getelementptr i16, ptr [[TMP1]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP2]], align 2
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP1]], align 2
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP3:%.*]] = zext <8 x i16> [[WIDE_LOAD]] to <8 x i64>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP4]] = add <8 x i64> [[TMP3]], [[VEC_PHI]]
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
@@ -321,7 +311,7 @@ define i64 @zext_add_reduc_i16_i64(ptr %a) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK-INTERLEAVE1:       middle.block:
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP6:%.*]] = call i64 @llvm.vector.reduce.add.v8i64(<8 x i64> [[TMP4]])
-; CHECK-INTERLEAVE1-NEXT:    br i1 false, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVE1-NEXT:    br label [[SCALAR_PH]]
 ; CHECK-INTERLEAVE1:       scalar.ph:
 ;
 ; CHECK-INTERLEAVED-LABEL: define i64 @zext_add_reduc_i16_i64(
@@ -335,9 +325,8 @@ define i64 @zext_add_reduc_i16_i64(ptr %a) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI:%.*]] = phi <8 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP1:%.*]] = getelementptr i16, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP2:%.*]] = getelementptr i16, ptr [[TMP1]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP3:%.*]] = getelementptr i16, ptr [[TMP1]], i32 8
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP2]], align 2
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP1]], align 2
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i16>, ptr [[TMP3]], align 2
 ; CHECK-INTERLEAVED-NEXT:    [[TMP4:%.*]] = zext <8 x i16> [[WIDE_LOAD]] to <8 x i64>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP5:%.*]] = zext <8 x i16> [[WIDE_LOAD2]] to <8 x i64>
@@ -349,7 +338,7 @@ define i64 @zext_add_reduc_i16_i64(ptr %a) #0 {
 ; CHECK-INTERLEAVED:       middle.block:
 ; CHECK-INTERLEAVED-NEXT:    [[BIN_RDX:%.*]] = add <8 x i64> [[TMP7]], [[TMP6]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vector.reduce.add.v8i64(<8 x i64> [[BIN_RDX]])
-; CHECK-INTERLEAVED-NEXT:    br i1 false, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVED-NEXT:    br label [[SCALAR_PH]]
 ; CHECK-INTERLEAVED:       scalar.ph:
 ;
 ; CHECK-MAXBW-LABEL: define i64 @zext_add_reduc_i16_i64(
@@ -370,8 +359,7 @@ define i64 @zext_add_reduc_i16_i64(ptr %a) #0 {
 ; CHECK-MAXBW-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr i16, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr i16, ptr [[TMP7]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i16>, ptr [[TMP8]], align 2
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i16>, ptr [[TMP7]], align 2
 ; CHECK-MAXBW-NEXT:    [[TMP9:%.*]] = zext <vscale x 8 x i16> [[WIDE_LOAD]] to <vscale x 8 x i64>
 ; CHECK-MAXBW-NEXT:    [[PARTIAL_REDUCE]] = call <vscale x 2 x i64> @llvm.experimental.vector.partial.reduce.add.nxv2i64.nxv8i64(<vscale x 2 x i64> [[VEC_PHI]], <vscale x 8 x i64> [[TMP9]])
 ; CHECK-MAXBW-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -413,8 +401,7 @@ define i32 @zext_add_reduc_i8_i32_has_neon_dotprod(ptr %a) #1 {
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP3:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[PARTIAL_REDUCE]] = call <4 x i32> @llvm.experimental.vector.partial.reduce.add.v4i32.v16i32(<4 x i32> [[VEC_PHI]], <16 x i32> [[TMP3]])
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
@@ -422,7 +409,7 @@ define i32 @zext_add_reduc_i8_i32_has_neon_dotprod(ptr %a) #1 {
 ; CHECK-INTERLEAVE1-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK-INTERLEAVE1:       middle.block:
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP5:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[PARTIAL_REDUCE]])
-; CHECK-INTERLEAVE1-NEXT:    br i1 false, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVE1-NEXT:    br label [[SCALAR_PH]]
 ; CHECK-INTERLEAVE1:       scalar.ph:
 ;
 ; CHECK-INTERLEAVED-LABEL: define i32 @zext_add_reduc_i8_i32_has_neon_dotprod(
@@ -436,9 +423,8 @@ define i32 @zext_add_reduc_i8_i32_has_neon_dotprod(ptr %a) #1 {
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[TMP1]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP4:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP5:%.*]] = zext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
@@ -450,7 +436,7 @@ define i32 @zext_add_reduc_i8_i32_has_neon_dotprod(ptr %a) #1 {
 ; CHECK-INTERLEAVED:       middle.block:
 ; CHECK-INTERLEAVED-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[PARTIAL_REDUCE3]], [[PARTIAL_REDUCE]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP7:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
-; CHECK-INTERLEAVED-NEXT:    br i1 false, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVED-NEXT:    br label [[SCALAR_PH]]
 ; CHECK-INTERLEAVED:       scalar.ph:
 ;
 ; CHECK-MAXBW-LABEL: define i32 @zext_add_reduc_i8_i32_has_neon_dotprod(
@@ -471,8 +457,7 @@ define i32 @zext_add_reduc_i8_i32_has_neon_dotprod(ptr %a) #1 {
 ; CHECK-MAXBW-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[TMP7]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP8]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP7]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP9:%.*]] = zext <vscale x 16 x i8> [[WIDE_LOAD]] to <vscale x 16 x i32>
 ; CHECK-MAXBW-NEXT:    [[PARTIAL_REDUCE]] = call <vscale x 4 x i32> @llvm.experimental.vector.partial.reduce.add.nxv4i32.nxv16i32(<vscale x 4 x i32> [[VEC_PHI]], <vscale x 16 x i32> [[TMP9]])
 ; CHECK-MAXBW-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -524,8 +509,7 @@ define i32 @zext_add_reduc_i8_i32_predicated(ptr %a) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i8> @llvm.masked.load.nxv4i8.p0(ptr [[TMP7]], i32 1, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i8> poison)
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i8> @llvm.masked.load.nxv4i8.p0(ptr [[TMP6]], i32 1, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i8> poison)
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP8:%.*]] = zext <vscale x 4 x i8> [[WIDE_MASKED_LOAD]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP9:%.*]] = add <vscale x 4 x i32> [[TMP8]], [[VEC_PHI]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP10]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> [[TMP9]], <vscale x 4 x i32> [[VEC_PHI]]
@@ -559,8 +543,7 @@ define i32 @zext_add_reduc_i8_i32_predicated(ptr %a) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i8> @llvm.masked.load.nxv4i8.p0(ptr [[TMP7]], i32 1, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i8> poison)
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i8> @llvm.masked.load.nxv4i8.p0(ptr [[TMP6]], i32 1, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i8> poison)
 ; CHECK-INTERLEAVED-NEXT:    [[TMP8:%.*]] = zext <vscale x 4 x i8> [[WIDE_MASKED_LOAD]] to <vscale x 4 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP9:%.*]] = add <vscale x 4 x i32> [[TMP8]], [[VEC_PHI]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP10]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> [[TMP9]], <vscale x 4 x i32> [[VEC_PHI]]
@@ -594,8 +577,7 @@ define i32 @zext_add_reduc_i8_i32_predicated(ptr %a) #0 {
 ; CHECK-MAXBW-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr [[TMP7]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i8> poison)
+; CHECK-MAXBW-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.masked.load.nxv16i8.p0(ptr [[TMP6]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i8> poison)
 ; CHECK-MAXBW-NEXT:    [[TMP8:%.*]] = zext <vscale x 16 x i8> [[WIDE_MASKED_LOAD]] to <vscale x 16 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP9:%.*]] = select <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i32> [[TMP8]], <vscale x 16 x i32> zeroinitializer
 ; CHECK-MAXBW-NEXT:    [[PARTIAL_REDUCE]] = call <vscale x 4 x i32> @llvm.experimental.vector.partial.reduce.add.nxv4i32.nxv16i32(<vscale x 4 x i32> [[VEC_PHI]], <vscale x 16 x i32> [[TMP9]])
@@ -721,8 +703,7 @@ define i32 @zext_sub_reduc_i8_i32_has_neon_dotprod(ptr %a) #1 {
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <16 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP3:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP4]] = sub <16 x i32> [[VEC_PHI]], [[TMP3]]
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
@@ -730,7 +711,7 @@ define i32 @zext_sub_reduc_i8_i32_has_neon_dotprod(ptr %a) #1 {
 ; CHECK-INTERLEAVE1-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; CHECK-INTERLEAVE1:       middle.block:
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[TMP4]])
-; CHECK-INTERLEAVE1-NEXT:    br i1 false, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVE1-NEXT:    br label [[SCALAR_PH]]
 ; CHECK-INTERLEAVE1:       scalar.ph:
 ;
 ; CHECK-INTERLEAVED-LABEL: define i32 @zext_sub_reduc_i8_i32_has_neon_dotprod(
@@ -744,9 +725,8 @@ define i32 @zext_sub_reduc_i8_i32_has_neon_dotprod(ptr %a) #1 {
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI:%.*]] = phi <16 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI1:%.*]] = phi <16 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[TMP1]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP4:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP5:%.*]] = zext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
@@ -758,7 +738,7 @@ define i32 @zext_sub_reduc_i8_i32_has_neon_dotprod(ptr %a) #1 {
 ; CHECK-INTERLEAVED:       middle.block:
 ; CHECK-INTERLEAVED-NEXT:    [[BIN_RDX:%.*]] = add <16 x i32> [[TMP7]], [[TMP6]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[BIN_RDX]])
-; CHECK-INTERLEAVED-NEXT:    br i1 false, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVED-NEXT:    br label [[SCALAR_PH]]
 ; CHECK-INTERLEAVED:       scalar.ph:
 ;
 ; CHECK-MAXBW-LABEL: define i32 @zext_sub_reduc_i8_i32_has_neon_dotprod(
@@ -779,8 +759,7 @@ define i32 @zext_sub_reduc_i8_i32_has_neon_dotprod(ptr %a) #1 {
 ; CHECK-MAXBW-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[TMP7]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP8]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP7]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP9:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-MAXBW-NEXT:    [[TMP10]] = sub <vscale x 8 x i32> [[VEC_PHI]], [[TMP9]]
 ; CHECK-MAXBW-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -822,8 +801,7 @@ define i32 @sext_add_reduc_i8_i32(ptr %a) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <16 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-INTERLEAVE1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP3:%.*]] = sext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP4]] = add <16 x i32> [[TMP3]], [[VEC_PHI]]
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
@@ -831,7 +809,7 @@ define i32 @sext_add_reduc_i8_i32(ptr %a) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; CHECK-INTERLEAVE1:       middle.block:
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[TMP4]])
-; CHECK-INTERLEAVE1-NEXT:    br i1 false, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVE1-NEXT:    br label [[SCALAR_PH]]
 ; CHECK-INTERLEAVE1:       scalar.ph:
 ;
 ; CHECK-INTERLEAVED-LABEL: define i32 @sext_add_reduc_i8_i32(
@@ -845,9 +823,8 @@ define i32 @sext_add_reduc_i8_i32(ptr %a) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI:%.*]] = phi <16 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI1:%.*]] = phi <16 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[TMP1]], i32 16
-; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP4:%.*]] = sext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
 ; CHECK-INTERLEAVED-NEXT:    [[TMP5:%.*]] = sext <16 x i8> [[WIDE_LOAD2]] to <16 x i32>
@@ -859,7 +836,7 @@ define i32 @sext_add_reduc_i8_i32(ptr %a) #0 {
 ; CHECK-INTERLEAVED:       middle.block:
 ; CHECK-INTERLEAVED-NEXT:    [[BIN_RDX:%.*]] = add <16 x i32> [[TMP7]], [[TMP6]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[BIN_RDX]])
-; CHECK-INTERLEAVED-NEXT:    br i1 false, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-INTERLEAVED-NEXT:    br label [[SCALAR_PH]]
 ; CHECK-INTERLEAVED:       scalar.ph:
 ;
 ; CHECK-MAXBW-LABEL: define i32 @sext_add_reduc_i8_i32(
@@ -880,8 +857,7 @@ define i32 @sext_add_reduc_i8_i32(ptr %a) #0 {
 ; CHECK-MAXBW-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; CHECK-MAXBW-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[TMP7]], i32 0
-; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP8]], align 1
+; CHECK-MAXBW-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP7]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP9:%.*]] = sext <vscale x 16 x i8> [[WIDE_LOAD]] to <vscale x 16 x i32>
 ; CHECK-MAXBW-NEXT:    [[PARTIAL_REDUCE]] = call <vscale x 4 x i32> @llvm.experimental.vector.partial.reduce.add.nxv4i32.nxv16i32(<vscale x 4 x i32> [[VEC_PHI]], <vscale x 16 x i32> [[TMP9]])
 ; CHECK-MAXBW-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -933,8 +909,7 @@ define i32 @add_of_zext_outside_loop(i32 %a, ptr noalias %b, i8 %c, i32 %d) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <16 x i32> [ [[TMP2]], [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[FOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[OFFSET_IDX:%.*]] = add i32 [[D]], [[INDEX]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[B]], i32 [[OFFSET_IDX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP4]], align 1
+; CHECK-INTERLEAVE1-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP3]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP5]] = add <16 x i32> [[VEC_PHI]], [[BROADCAST_SPLAT]]
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 16
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
@@ -966,9 +941,8 @@ define i32 @add_of_zext_outside_loop(i32 %a, ptr noalias %b, i8 %c, i32 %d) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI2:%.*]] = phi <16 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[OFFSET_IDX:%.*]] = add i32 [[D]], [[VEC_PHI1]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[B]], i32 [[OFFSET_IDX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 16
-; CHECK-INTERLEAVED-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP4]], align 1
+; CHECK-INTERLEAVED-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP3]], align 1
 ; CHECK-INTERLEAVED-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP5]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP6]] = add <16 x i32> [[VEC_PHI]], [[BROADCAST_SPLAT]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP7]] = add <16 x i32> [[VEC_PHI2]], [[BROADCAST_SPLAT]]
@@ -1008,8 +982,7 @@ define i32 @add_of_zext_outside_loop(i32 %a, ptr noalias %b, i8 %c, i32 %d) #0 {
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 16 x i32> [ [[TMP8]], [[VECTOR_PH]] ], [ [[TMP11:%.*]], [[FOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[OFFSET_IDX:%.*]] = add i32 [[D]], [[INDEX]]
 ; CHECK-MAXBW-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[B]], i32 [[OFFSET_IDX]]
-; CHECK-MAXBW-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[TMP9]], i32 0
-; CHECK-MAXBW-NEXT:    store <vscale x 16 x i8> zeroinitializer, ptr [[TMP10]], align 1
+; CHECK-MAXBW-NEXT:    store <vscale x 16 x i8> zeroinitializer, ptr [[TMP9]], align 1
 ; CHECK-MAXBW-NEXT:    [[TMP11]] = add <vscale x 16 x i32> [[VEC_PHI]], [[BROADCAST_SPLAT]]
 ; CHECK-MAXBW-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP6]]
 ; CHECK-MAXBW-NEXT:    [[TMP12:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
@@ -1060,8 +1033,7 @@ define i32 @add_of_loop_invariant_zext(i32 %a, ptr %b, i8 %c, i32 %d) #0 {
 ; CHECK-INTERLEAVE1-NEXT:    [[VEC_PHI:%.*]] = phi <16 x i32> [ [[TMP2]], [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[FOR_BODY]] ]
 ; CHECK-INTERLEAVE1-NEXT:    [[OFFSET_IDX:%.*]] = add i32 [[D]], [[INDEX]]
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[B]], i32 [[OFFSET_IDX]]
-; CHECK-INTERLEAVE1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 0
-; CHECK-INTERLEAVE1-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP5]], align 1
+; CHECK-INTERLEAVE1-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP4]], align 1
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP6]] = add <16 x i32> [[VEC_PHI]], [[TMP3]]
 ; CHECK-INTERLEAVE1-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 16
 ; CHECK-INTERLEAVE1-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
@@ -1093,9 +1065,8 @@ define i32 @add_of_loop_invariant_zext(i32 %a, ptr %b, i8 %c, i32 %d) #0 {
 ; CHECK-INTERLEAVED-NEXT:    [[VEC_PHI2:%.*]] = phi <16 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-INTERLEAVED-NEXT:    [[OFFSET_IDX:%.*]] = add i32 [[D]], [[VEC_PHI1]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[B]], i32 [[OFFSET_IDX]]
-; CHECK-INTERLEAVED-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 0
 ; CHECK-INTERLEAVED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 16
-; CHECK-INTERLEAVED-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP5]], align 1
+; CHECK-INTERLEAVED-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP4]], align 1
 ; CHECK-INTERLEAVED-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP6]], align 1
 ; CHECK-INTERLEAVED-NEXT:    [[TMP7]] = add <16 x i32> [[VEC_PHI]], [[TMP3]]
 ; CHECK-INTERLEAVED-NEXT:    [[TMP8]] = add <16 x i32> [[VEC_PHI2]], [[TMP3]]
@@ -1135,8 +1106,7 @@ define i32 @add_of_loop_invariant_zext(i32 %a, ptr %b, i8 %c, i32 %d) #0 {
 ; CHECK-MAXBW-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP8]], [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[FOR_BODY]] ]
 ; CHECK-MAXBW-NEXT:    [[OFFSET_IDX:%.*]] = add i32 [[D]], [[INDEX]]
 ; CHECK-MAXBW-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[B]], i32 [[OFFSET_IDX]]
-; CHECK-MAXBW-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[TMP10]], i32 0
-; CHECK-MAXBW-NEXT:    store <vscale x 16 x i8> zeroinitializer, ptr [[TMP11]], align 1
+; CHECK-MAXBW-NEXT:    store <vscale x 16 x i8> zeroinitializer, ptr [[TMP10]], align 1
 ; CHECK-MAXBW-NEXT:    [[PARTIAL_REDUCE]] = call <vscale x 4 x i32> @llvm.experimental.vector.partial.reduce.add.nxv4i32.nxv16i32(<vscale x 4 x i32> [[VEC_PHI]], <vscale x 16 x i32> [[TMP9]])
 ; CHECK-MAXBW-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP6]]
 ; CHECK-MAXBW-NEXT:    [[TMP12:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/reduction-recurrence-costs-sve.ll b/llvm/test/Transforms/LoopVectorize/AArch64/reduction-recurrence-costs-sve.ll
index 492ab563..08d35f7 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/reduction-recurrence-costs-sve.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/reduction-recurrence-costs-sve.ll
@@ -353,11 +353,10 @@ define i16 @reduce_udiv(ptr %src, i16 %x, i64 %N) #0 {
 ; DEFAULT-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i16> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP21:%.*]], %[[VECTOR_BODY]] ]
 ; DEFAULT-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 4 x i16> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP22:%.*]], %[[VECTOR_BODY]] ]
 ; DEFAULT-NEXT:    [[TMP13:%.*]] = getelementptr i16, ptr [[SRC]], i64 [[INDEX]]
-; DEFAULT-NEXT:    [[TMP15:%.*]] = getelementptr i16, ptr [[TMP13]], i32 0
 ; DEFAULT-NEXT:    [[TMP16:%.*]] = call i64 @llvm.vscale.i64()
 ; DEFAULT-NEXT:    [[TMP17:%.*]] = mul nuw i64 [[TMP16]], 4
 ; DEFAULT-NEXT:    [[TMP18:%.*]] = getelementptr i16, ptr [[TMP13]], i64 [[TMP17]]
-; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP15]], align 2
+; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP13]], align 2
 ; DEFAULT-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i16>, ptr [[TMP18]], align 2
 ; DEFAULT-NEXT:    [[TMP19:%.*]] = udiv <vscale x 4 x i16> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
 ; DEFAULT-NEXT:    [[TMP20:%.*]] = udiv <vscale x 4 x i16> [[WIDE_LOAD2]], [[BROADCAST_SPLAT]]
@@ -412,11 +411,10 @@ define i16 @reduce_udiv(ptr %src, i16 %x, i64 %N) #0 {
 ; VSCALEFORTUNING2-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i16> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP15:%.*]], %[[VECTOR_BODY]] ]
 ; VSCALEFORTUNING2-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 4 x i16> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP16:%.*]], %[[VECTOR_BODY]] ]
 ; VSCALEFORTUNING2-NEXT:    [[TMP8:%.*]] = getelementptr i16, ptr [[SRC]], i64 [[INDEX]]
-; VSCALEFORTUNING2-NEXT:    [[TMP9:%.*]] = getelementptr i16, ptr [[TMP8]], i32 0
 ; VSCALEFORTUNING2-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
 ; VSCALEFORTUNING2-NEXT:    [[TMP11:%.*]] = mul nuw i64 [[TMP10]], 4
 ; VSCALEFORTUNING2-NEXT:    [[TMP12:%.*]] = getelementptr i16, ptr [[TMP8]], i64 [[TMP11]]
-; VSCALEFORTUNING2-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP9]], align 2
+; VSCALEFORTUNING2-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP8]], align 2
 ; VSCALEFORTUNING2-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x i16>, ptr [[TMP12]], align 2
 ; VSCALEFORTUNING2-NEXT:    [[TMP13:%.*]] = udiv <vscale x 4 x i16> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
 ; VSCALEFORTUNING2-NEXT:    [[TMP14:%.*]] = udiv <vscale x 4 x i16> [[WIDE_LOAD2]], [[BROADCAST_SPLAT]]
@@ -476,8 +474,7 @@ define i16 @reduce_udiv(ptr %src, i16 %x, i64 %N) #0 {
 ; PRED-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], %[[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; PRED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x i16> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP16:%.*]], %[[VECTOR_BODY]] ]
 ; PRED-NEXT:    [[TMP14:%.*]] = getelementptr i16, ptr [[SRC]], i64 [[INDEX]]
-; PRED-NEXT:    [[TMP15:%.*]] = getelementptr i16, ptr [[TMP14]], i32 0
-; PRED-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x i16> @llvm.masked.load.nxv8i16.p0(ptr [[TMP15]], i32 2, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x i16> poison)
+; PRED-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x i16> @llvm.masked.load.nxv8i16.p0(ptr [[TMP14]], i32 2, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x i16> poison)
 ; PRED-NEXT:    [[TMP20:%.*]] = udiv <vscale x 8 x i16> [[WIDE_MASKED_LOAD]], [[BROADCAST_SPLAT]]
 ; PRED-NEXT:    [[TMP21:%.*]] = or <vscale x 8 x i16> [[TMP20]], [[VEC_PHI]]
 ; PRED-NEXT:    [[TMP16]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x i16> [[TMP21]], <vscale x 8 x i16> [[VEC_PHI]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/scalable-avoid-scalarization.ll b/llvm/test/Transforms/LoopVectorize/AArch64/scalable-avoid-scalarization.ll
index 0f2eae1..eb3d724 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/scalable-avoid-scalarization.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/scalable-avoid-scalarization.ll
@@ -41,11 +41,9 @@ define void @test_no_scalarization(ptr %a, ptr noalias %b, i32 %idx, i32 %n) #0
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i32 [[IDX]], [[INDEX]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr i64, ptr [[A:%.*]], <vscale x 2 x i32> [[VEC_IND]]
 ; CHECK-NEXT:    [[TMP16:%.*]] = extractelement <vscale x 2 x ptr> [[TMP15]], i32 0
-; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr double, ptr [[TMP16]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x double>, ptr [[TMP17]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x double>, ptr [[TMP16]], align 8
 ; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr i64, ptr [[B:%.*]], i32 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr double, ptr [[TMP18]], i32 0
-; CHECK-NEXT:    store <vscale x 2 x double> [[WIDE_LOAD]], ptr [[TMP19]], align 8
+; CHECK-NEXT:    store <vscale x 2 x double> [[WIDE_LOAD]], ptr [[TMP18]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP7]]
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 2 x i32> [[VEC_IND]], [[DOTSPLAT2]]
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/scalable-fp-ext-trunc-illegal-type.ll b/llvm/test/Transforms/LoopVectorize/AArch64/scalable-fp-ext-trunc-illegal-type.ll
index 7294452..d751d39 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/scalable-fp-ext-trunc-illegal-type.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/scalable-fp-ext-trunc-illegal-type.ll
@@ -25,13 +25,11 @@ define void @load_ext_trunc_store(ptr readonly %in, ptr noalias %out, i64 %N) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw double, ptr [[IN]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw double, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP4]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP2]], align 8
 ; CHECK-NEXT:    [[TMP3:%.*]] = fpext <4 x double> [[WIDE_LOAD]] to <4 x fp128>
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw float, ptr [[OUT]], i64 [[INDEX]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = fptrunc <4 x fp128> [[TMP3]] to <4 x float>
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw float, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <4 x float> [[TMP5]], ptr [[TMP12]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP5]], ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/scalable-reduction-inloop-cond.ll b/llvm/test/Transforms/LoopVectorize/AArch64/scalable-reduction-inloop-cond.ll
index 1b489dd..011b823 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/scalable-reduction-inloop-cond.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/scalable-reduction-inloop-cond.ll
@@ -21,12 +21,10 @@ define float @cond_fadd(ptr noalias nocapture readonly %a, ptr noalias nocapture
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi float [ 1.000000e+00, [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[COND:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[TMP9:%.*]] = fcmp une <vscale x 4 x float> [[WIDE_LOAD]], splat (float 2.000000e+00)
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr float, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr float, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[TMP9]], <vscale x 4 x float> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP10]], i32 4, <vscale x 4 x i1> [[TMP9]], <vscale x 4 x float> poison)
 ; CHECK-NEXT:    [[TMP12:%.*]] = select fast <vscale x 4 x i1> [[TMP9]], <vscale x 4 x float> [[WIDE_MASKED_LOAD]], <vscale x 4 x float> zeroinitializer
 ; CHECK-NEXT:    [[TMP13:%.*]] = call fast float @llvm.vector.reduce.fadd.nxv4f32(float 0.000000e+00, <vscale x 4 x float> [[TMP12]])
 ; CHECK-NEXT:    [[TMP14]] = fadd fast float [[TMP13]], [[VEC_PHI]]
@@ -107,12 +105,10 @@ define float @cond_cmp_sel(ptr noalias %a, ptr noalias %cond, i64 %N) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi float [ 1.000000e+00, [[VECTOR_PH]] ], [ [[RDX_MINMAX_SELECT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[COND:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[TMP9:%.*]] = fcmp une <vscale x 4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00)
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr float, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr float, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[TMP9]], <vscale x 4 x float> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP10]], i32 4, <vscale x 4 x i1> [[TMP9]], <vscale x 4 x float> poison)
 ; CHECK-NEXT:    [[TMP12:%.*]] = select fast <vscale x 4 x i1> [[TMP9]], <vscale x 4 x float> [[WIDE_MASKED_LOAD]], <vscale x 4 x float> splat (float 0x47EFFFFFE0000000)
 ; CHECK-NEXT:    [[TMP13:%.*]] = call fast float @llvm.vector.reduce.fmin.nxv4f32(<vscale x 4 x float> [[TMP12]])
 ; CHECK-NEXT:    [[RDX_MINMAX_CMP:%.*]] = fcmp fast olt float [[TMP13]], [[VEC_PHI]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/scalable-strict-fadd.ll b/llvm/test/Transforms/LoopVectorize/AArch64/scalable-strict-fadd.ll
index d73cdc1..a60d35d 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/scalable-strict-fadd.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/scalable-strict-fadd.ll
@@ -47,33 +47,32 @@ define float @fadd_strict(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-UNORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-UNORDERED:       vector.body:
 ; CHECK-UNORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> splat (float -0.000000e+00), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> splat (float -0.000000e+00), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 0
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP7]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP8]] = fadd <vscale x 8 x float> [[WIDE_LOAD]], [[VEC_PHI]]
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP6]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP7]] = fadd <vscale x 8 x float> [[WIDE_LOAD]], [[VEC_PHI]]
 ; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
-; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-UNORDERED-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-UNORDERED-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK-UNORDERED:       middle.block:
-; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float -0.000000e+00, <vscale x 8 x float> [[TMP8]])
+; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float -0.000000e+00, <vscale x 8 x float> [[TMP7]])
 ; CHECK-UNORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-UNORDERED:       scalar.ph:
 ; CHECK-UNORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-UNORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP10]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
+; CHECK-UNORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP9]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
 ; CHECK-UNORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-UNORDERED:       for.body:
 ; CHECK-UNORDERED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[SUM_07:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = load float, ptr [[ARRAYIDX]], align 4
-; CHECK-UNORDERED-NEXT:    [[ADD]] = fadd float [[TMP11]], [[SUM_07]]
+; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-UNORDERED-NEXT:    [[ADD]] = fadd float [[TMP10]], [[SUM_07]]
 ; CHECK-UNORDERED-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-UNORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK-UNORDERED:       for.end:
-; CHECK-UNORDERED-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP10]], [[MIDDLE_BLOCK]] ]
+; CHECK-UNORDERED-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP9]], [[MIDDLE_BLOCK]] ]
 ; CHECK-UNORDERED-NEXT:    ret float [[ADD_LCSSA]]
 ;
 ; CHECK-ORDERED-LABEL: define float @fadd_strict
@@ -93,32 +92,31 @@ define float @fadd_strict(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-ORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED:       vector.body:
 ; CHECK-ORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 0
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP7]], align 4
-; CHECK-ORDERED-NEXT:    [[TMP8]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[WIDE_LOAD]])
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP6]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP7]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[WIDE_LOAD]])
 ; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
-; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-ORDERED-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-ORDERED-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK-ORDERED:       middle.block:
 ; CHECK-ORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-ORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED:       scalar.ph:
 ; CHECK-ORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP8]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
+; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
 ; CHECK-ORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED:       for.body:
 ; CHECK-ORDERED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-NEXT:    [[SUM_07:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = load float, ptr [[ARRAYIDX]], align 4
-; CHECK-ORDERED-NEXT:    [[ADD]] = fadd float [[TMP10]], [[SUM_07]]
+; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-ORDERED-NEXT:    [[ADD]] = fadd float [[TMP9]], [[SUM_07]]
 ; CHECK-ORDERED-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-ORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK-ORDERED:       for.end:
-; CHECK-ORDERED-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP7]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    ret float [[ADD_LCSSA]]
 ;
 ; CHECK-ORDERED-TF-LABEL: define float @fadd_strict
@@ -144,17 +142,16 @@ define float @fadd_strict(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-ORDERED-TF:       vector.body:
 ; CHECK-ORDERED-TF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[TMP10]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP11]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[TMP12:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> [[WIDE_MASKED_LOAD]], <vscale x 8 x float> splat (float -0.000000e+00)
-; CHECK-ORDERED-TF-NEXT:    [[TMP13]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP12]])
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP10]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[TMP11:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> [[WIDE_MASKED_LOAD]], <vscale x 8 x float> splat (float -0.000000e+00)
+; CHECK-ORDERED-TF-NEXT:    [[TMP12]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP11]])
 ; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP4]]
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP14:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
-; CHECK-ORDERED-TF-NEXT:    [[TMP15:%.*]] = extractelement <vscale x 8 x i1> [[TMP14]], i32 0
-; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP13:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
+; CHECK-ORDERED-TF-NEXT:    [[TMP14:%.*]] = extractelement <vscale x 8 x i1> [[TMP13]], i32 0
+; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK-ORDERED-TF:       middle.block:
 ; CHECK-ORDERED-TF-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK-ORDERED-TF:       scalar.ph:
@@ -165,13 +162,13 @@ define float @fadd_strict(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-ORDERED-TF-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[SUM_07:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP16:%.*]] = load float, ptr [[ARRAYIDX]], align 4
-; CHECK-ORDERED-TF-NEXT:    [[ADD]] = fadd float [[TMP16]], [[SUM_07]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP15:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-ORDERED-TF-NEXT:    [[ADD]] = fadd float [[TMP15]], [[SUM_07]]
 ; CHECK-ORDERED-TF-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-ORDERED-TF-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-TF-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK-ORDERED-TF:       for.end:
-; CHECK-ORDERED-TF-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP13]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    ret float [[ADD_LCSSA]]
 ;
 
@@ -230,54 +227,53 @@ define float @fadd_strict_unroll(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-UNORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-UNORDERED:       vector.body:
 ; CHECK-UNORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> splat (float -0.000000e+00), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP17:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 8 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI2:%.*]] = phi <vscale x 8 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP19:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI3:%.*]] = phi <vscale x 8 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP20:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> splat (float -0.000000e+00), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP16:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 8 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP17:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI2:%.*]] = phi <vscale x 8 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI3:%.*]] = phi <vscale x 8 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP19:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 0
-; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 8
-; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP9]]
-; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 16
-; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP12]]
-; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP14]], 24
-; CHECK-UNORDERED-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP15]]
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP7]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP10]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP13]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP16]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP17]] = fadd <vscale x 8 x float> [[WIDE_LOAD]], [[VEC_PHI]]
-; CHECK-UNORDERED-NEXT:    [[TMP18]] = fadd <vscale x 8 x float> [[WIDE_LOAD4]], [[VEC_PHI1]]
-; CHECK-UNORDERED-NEXT:    [[TMP19]] = fadd <vscale x 8 x float> [[WIDE_LOAD5]], [[VEC_PHI2]]
-; CHECK-UNORDERED-NEXT:    [[TMP20]] = fadd <vscale x 8 x float> [[WIDE_LOAD6]], [[VEC_PHI3]]
+; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 8
+; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP8]]
+; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = mul nuw i64 [[TMP10]], 16
+; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP11]]
+; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = mul nuw i64 [[TMP13]], 24
+; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP14]]
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP6]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP9]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP12]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP15]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP16]] = fadd <vscale x 8 x float> [[WIDE_LOAD]], [[VEC_PHI]]
+; CHECK-UNORDERED-NEXT:    [[TMP17]] = fadd <vscale x 8 x float> [[WIDE_LOAD4]], [[VEC_PHI1]]
+; CHECK-UNORDERED-NEXT:    [[TMP18]] = fadd <vscale x 8 x float> [[WIDE_LOAD5]], [[VEC_PHI2]]
+; CHECK-UNORDERED-NEXT:    [[TMP19]] = fadd <vscale x 8 x float> [[WIDE_LOAD6]], [[VEC_PHI3]]
 ; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
-; CHECK-UNORDERED-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-UNORDERED-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK-UNORDERED-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-UNORDERED-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK-UNORDERED:       middle.block:
-; CHECK-UNORDERED-NEXT:    [[BIN_RDX:%.*]] = fadd <vscale x 8 x float> [[TMP18]], [[TMP17]]
-; CHECK-UNORDERED-NEXT:    [[BIN_RDX7:%.*]] = fadd <vscale x 8 x float> [[TMP19]], [[BIN_RDX]]
-; CHECK-UNORDERED-NEXT:    [[BIN_RDX8:%.*]] = fadd <vscale x 8 x float> [[TMP20]], [[BIN_RDX7]]
-; CHECK-UNORDERED-NEXT:    [[TMP22:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float -0.000000e+00, <vscale x 8 x float> [[BIN_RDX8]])
+; CHECK-UNORDERED-NEXT:    [[BIN_RDX:%.*]] = fadd <vscale x 8 x float> [[TMP17]], [[TMP16]]
+; CHECK-UNORDERED-NEXT:    [[BIN_RDX7:%.*]] = fadd <vscale x 8 x float> [[TMP18]], [[BIN_RDX]]
+; CHECK-UNORDERED-NEXT:    [[BIN_RDX8:%.*]] = fadd <vscale x 8 x float> [[TMP19]], [[BIN_RDX7]]
+; CHECK-UNORDERED-NEXT:    [[TMP21:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float -0.000000e+00, <vscale x 8 x float> [[BIN_RDX8]])
 ; CHECK-UNORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-UNORDERED:       scalar.ph:
 ; CHECK-UNORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-UNORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP22]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
+; CHECK-UNORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP21]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
 ; CHECK-UNORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-UNORDERED:       for.body:
 ; CHECK-UNORDERED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[SUM_07:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-UNORDERED-NEXT:    [[TMP23:%.*]] = load float, ptr [[ARRAYIDX]], align 4
-; CHECK-UNORDERED-NEXT:    [[ADD]] = fadd float [[TMP23]], [[SUM_07]]
+; CHECK-UNORDERED-NEXT:    [[TMP22:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-UNORDERED-NEXT:    [[ADD]] = fadd float [[TMP22]], [[SUM_07]]
 ; CHECK-UNORDERED-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-UNORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK-UNORDERED:       for.end:
-; CHECK-UNORDERED-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP22]], [[MIDDLE_BLOCK]] ]
+; CHECK-UNORDERED-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP21]], [[MIDDLE_BLOCK]] ]
 ; CHECK-UNORDERED-NEXT:    ret float [[ADD_LCSSA]]
 ;
 ; CHECK-ORDERED-LABEL: define float @fadd_strict_unroll
@@ -297,47 +293,46 @@ define float @fadd_strict_unroll(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-ORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED:       vector.body:
 ; CHECK-ORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP20:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP19:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 0
-; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 8
-; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP9]]
-; CHECK-ORDERED-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 16
-; CHECK-ORDERED-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP12]]
-; CHECK-ORDERED-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP14]], 24
-; CHECK-ORDERED-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP15]]
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP7]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x float>, ptr [[TMP10]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x float>, ptr [[TMP13]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x float>, ptr [[TMP16]], align 4
-; CHECK-ORDERED-NEXT:    [[TMP17:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[WIDE_LOAD]])
-; CHECK-ORDERED-NEXT:    [[TMP18:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP17]], <vscale x 8 x float> [[WIDE_LOAD1]])
-; CHECK-ORDERED-NEXT:    [[TMP19:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP18]], <vscale x 8 x float> [[WIDE_LOAD2]])
-; CHECK-ORDERED-NEXT:    [[TMP20]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP19]], <vscale x 8 x float> [[WIDE_LOAD3]])
+; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 8
+; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP8]]
+; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP11:%.*]] = mul nuw i64 [[TMP10]], 16
+; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP11]]
+; CHECK-ORDERED-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP14:%.*]] = mul nuw i64 [[TMP13]], 24
+; CHECK-ORDERED-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP14]]
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP6]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x float>, ptr [[TMP9]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x float>, ptr [[TMP12]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x float>, ptr [[TMP15]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP16:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[WIDE_LOAD]])
+; CHECK-ORDERED-NEXT:    [[TMP17:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP16]], <vscale x 8 x float> [[WIDE_LOAD1]])
+; CHECK-ORDERED-NEXT:    [[TMP18:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP17]], <vscale x 8 x float> [[WIDE_LOAD2]])
+; CHECK-ORDERED-NEXT:    [[TMP19]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP18]], <vscale x 8 x float> [[WIDE_LOAD3]])
 ; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
-; CHECK-ORDERED-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-ORDERED-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK-ORDERED-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-ORDERED-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK-ORDERED:       middle.block:
 ; CHECK-ORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-ORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED:       scalar.ph:
 ; CHECK-ORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP20]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
+; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP19]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
 ; CHECK-ORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED:       for.body:
 ; CHECK-ORDERED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-NEXT:    [[SUM_07:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-ORDERED-NEXT:    [[TMP22:%.*]] = load float, ptr [[ARRAYIDX]], align 4
-; CHECK-ORDERED-NEXT:    [[ADD]] = fadd float [[TMP22]], [[SUM_07]]
+; CHECK-ORDERED-NEXT:    [[TMP21:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-ORDERED-NEXT:    [[ADD]] = fadd float [[TMP21]], [[SUM_07]]
 ; CHECK-ORDERED-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-ORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK-ORDERED:       for.end:
-; CHECK-ORDERED-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP20]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP19]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    ret float [[ADD_LCSSA]]
 ;
 ; CHECK-ORDERED-TF-LABEL: define float @fadd_strict_unroll
@@ -378,47 +373,46 @@ define float @fadd_strict_unroll(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK6:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY3]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT12:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK7:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY4]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT13:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK8:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY5]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT14:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP34:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP33:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP18:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP19:%.*]] = mul nuw i64 [[TMP18]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP19]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP21:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP22:%.*]] = mul nuw i64 [[TMP21]], 16
-; CHECK-ORDERED-TF-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP22]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP24:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP25:%.*]] = mul nuw i64 [[TMP24]], 24
-; CHECK-ORDERED-TF-NEXT:    [[TMP26:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP25]]
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP17]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP20]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP23]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP26]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[TMP27:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> [[WIDE_MASKED_LOAD]], <vscale x 8 x float> splat (float -0.000000e+00)
-; CHECK-ORDERED-TF-NEXT:    [[TMP28:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP27]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP29:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> [[WIDE_MASKED_LOAD9]], <vscale x 8 x float> splat (float -0.000000e+00)
-; CHECK-ORDERED-TF-NEXT:    [[TMP30:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP28]], <vscale x 8 x float> [[TMP29]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP31:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> [[WIDE_MASKED_LOAD10]], <vscale x 8 x float> splat (float -0.000000e+00)
-; CHECK-ORDERED-TF-NEXT:    [[TMP32:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP30]], <vscale x 8 x float> [[TMP31]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP33:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> [[WIDE_MASKED_LOAD11]], <vscale x 8 x float> splat (float -0.000000e+00)
-; CHECK-ORDERED-TF-NEXT:    [[TMP34]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP32]], <vscale x 8 x float> [[TMP33]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP17:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP18:%.*]] = mul nuw i64 [[TMP17]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP18]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP20:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP21:%.*]] = mul nuw i64 [[TMP20]], 16
+; CHECK-ORDERED-TF-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP21]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP23:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP24:%.*]] = mul nuw i64 [[TMP23]], 24
+; CHECK-ORDERED-TF-NEXT:    [[TMP25:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP24]]
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP16]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP19]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP22]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP25]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[TMP26:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> [[WIDE_MASKED_LOAD]], <vscale x 8 x float> splat (float -0.000000e+00)
+; CHECK-ORDERED-TF-NEXT:    [[TMP27:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP26]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP28:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> [[WIDE_MASKED_LOAD9]], <vscale x 8 x float> splat (float -0.000000e+00)
+; CHECK-ORDERED-TF-NEXT:    [[TMP29:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP27]], <vscale x 8 x float> [[TMP28]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP30:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> [[WIDE_MASKED_LOAD10]], <vscale x 8 x float> splat (float -0.000000e+00)
+; CHECK-ORDERED-TF-NEXT:    [[TMP31:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP29]], <vscale x 8 x float> [[TMP30]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP32:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> [[WIDE_MASKED_LOAD11]], <vscale x 8 x float> splat (float -0.000000e+00)
+; CHECK-ORDERED-TF-NEXT:    [[TMP33]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP31]], <vscale x 8 x float> [[TMP32]])
 ; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP4]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP35:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP36:%.*]] = mul nuw i64 [[TMP35]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP37:%.*]] = add i64 [[INDEX]], [[TMP36]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP38:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP39:%.*]] = mul nuw i64 [[TMP38]], 16
-; CHECK-ORDERED-TF-NEXT:    [[TMP40:%.*]] = add i64 [[INDEX]], [[TMP39]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP41:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP42:%.*]] = mul nuw i64 [[TMP41]], 24
-; CHECK-ORDERED-TF-NEXT:    [[TMP43:%.*]] = add i64 [[INDEX]], [[TMP42]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP34:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP35:%.*]] = mul nuw i64 [[TMP34]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP36:%.*]] = add i64 [[INDEX]], [[TMP35]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP37:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP38:%.*]] = mul nuw i64 [[TMP37]], 16
+; CHECK-ORDERED-TF-NEXT:    [[TMP39:%.*]] = add i64 [[INDEX]], [[TMP38]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP40:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP41:%.*]] = mul nuw i64 [[TMP40]], 24
+; CHECK-ORDERED-TF-NEXT:    [[TMP42:%.*]] = add i64 [[INDEX]], [[TMP41]]
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT12]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP37]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT13]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP40]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT14]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP43]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP44:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
-; CHECK-ORDERED-TF-NEXT:    [[TMP45:%.*]] = extractelement <vscale x 8 x i1> [[TMP44]], i32 0
-; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP45]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT12]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP36]], i64 [[TMP9]])
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT13]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP39]], i64 [[TMP9]])
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT14]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP42]], i64 [[TMP9]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP43:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
+; CHECK-ORDERED-TF-NEXT:    [[TMP44:%.*]] = extractelement <vscale x 8 x i1> [[TMP43]], i32 0
+; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP44]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK-ORDERED-TF:       middle.block:
 ; CHECK-ORDERED-TF-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK-ORDERED-TF:       scalar.ph:
@@ -429,13 +423,13 @@ define float @fadd_strict_unroll(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-ORDERED-TF-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[SUM_07:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP46:%.*]] = load float, ptr [[ARRAYIDX]], align 4
-; CHECK-ORDERED-TF-NEXT:    [[ADD]] = fadd float [[TMP46]], [[SUM_07]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP45:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-ORDERED-TF-NEXT:    [[ADD]] = fadd float [[TMP45]], [[SUM_07]]
 ; CHECK-ORDERED-TF-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-ORDERED-TF-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-TF-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK-ORDERED-TF:       for.end:
-; CHECK-ORDERED-TF-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP34]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ], [ [[TMP33]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    ret float [[ADD_LCSSA]]
 ;
 
@@ -781,40 +775,38 @@ define float @fadd_of_sum(ptr noalias nocapture readonly %a, ptr noalias nocaptu
 ; CHECK-UNORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-UNORDERED:       vector.body:
 ; CHECK-UNORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ insertelement (<vscale x 4 x float> splat (float -0.000000e+00), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ insertelement (<vscale x 4 x float> splat (float -0.000000e+00), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
-; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[TMP9]], i32 0
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP10]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = fadd <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD1]]
-; CHECK-UNORDERED-NEXT:    [[TMP12]] = fadd <vscale x 4 x float> [[VEC_PHI]], [[TMP11]]
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = fadd <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD1]]
+; CHECK-UNORDERED-NEXT:    [[TMP10]] = fadd <vscale x 4 x float> [[VEC_PHI]], [[TMP9]]
 ; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
-; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-UNORDERED-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-UNORDERED-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK-UNORDERED:       middle.block:
-; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = call float @llvm.vector.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[TMP12]])
+; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = call float @llvm.vector.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[TMP10]])
 ; CHECK-UNORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
 ; CHECK-UNORDERED:       scalar.ph:
 ; CHECK-UNORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
-; CHECK-UNORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP14]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[FOR_BODY_PREHEADER]] ]
+; CHECK-UNORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP12]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[FOR_BODY_PREHEADER]] ]
 ; CHECK-UNORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-UNORDERED:       for.body:
 ; CHECK-UNORDERED-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
 ; CHECK-UNORDERED-NEXT:    [[RES_014:%.*]] = phi float [ [[RDX:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
 ; CHECK-UNORDERED-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
 ; CHECK-UNORDERED-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; CHECK-UNORDERED-NEXT:    [[TMP16:%.*]] = load float, ptr [[ARRAYIDX4]], align 4
-; CHECK-UNORDERED-NEXT:    [[ADD:%.*]] = fadd float [[TMP15]], [[TMP16]]
+; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = load float, ptr [[ARRAYIDX4]], align 4
+; CHECK-UNORDERED-NEXT:    [[ADD:%.*]] = fadd float [[TMP13]], [[TMP14]]
 ; CHECK-UNORDERED-NEXT:    [[RDX]] = fadd float [[RES_014]], [[ADD]]
 ; CHECK-UNORDERED-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-UNORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
 ; CHECK-UNORDERED:       for.end.loopexit:
-; CHECK-UNORDERED-NEXT:    [[RDX_LCSSA:%.*]] = phi float [ [[RDX]], [[FOR_BODY]] ], [ [[TMP14]], [[MIDDLE_BLOCK]] ]
+; CHECK-UNORDERED-NEXT:    [[RDX_LCSSA:%.*]] = phi float [ [[RDX]], [[FOR_BODY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
 ; CHECK-UNORDERED-NEXT:    br label [[FOR_END]]
 ; CHECK-UNORDERED:       for.end:
 ; CHECK-UNORDERED-NEXT:    [[RES:%.*]] = phi float [ 0.000000e+00, [[ENTRY:%.*]] ], [ [[RDX_LCSSA]], [[FOR_END_LOOPEXIT]] ]
@@ -842,39 +834,37 @@ define float @fadd_of_sum(ptr noalias nocapture readonly %a, ptr noalias nocaptu
 ; CHECK-ORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED:       vector.body:
 ; CHECK-ORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
-; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
-; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[TMP9]], i32 0
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP10]], align 4
-; CHECK-ORDERED-NEXT:    [[TMP11:%.*]] = fadd <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD1]]
-; CHECK-ORDERED-NEXT:    [[TMP12]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[TMP11]])
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = fadd <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD1]]
+; CHECK-ORDERED-NEXT:    [[TMP10]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[TMP9]])
 ; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
-; CHECK-ORDERED-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-ORDERED-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK-ORDERED-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-ORDERED-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK-ORDERED:       middle.block:
 ; CHECK-ORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-ORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED:       scalar.ph:
 ; CHECK-ORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
-; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP12]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[FOR_BODY_PREHEADER]] ]
+; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP10]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[FOR_BODY_PREHEADER]] ]
 ; CHECK-ORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED:       for.body:
 ; CHECK-ORDERED-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
 ; CHECK-ORDERED-NEXT:    [[RES_014:%.*]] = phi float [ [[RDX:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
 ; CHECK-ORDERED-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-ORDERED-NEXT:    [[TMP14:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
 ; CHECK-ORDERED-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; CHECK-ORDERED-NEXT:    [[TMP15:%.*]] = load float, ptr [[ARRAYIDX4]], align 4
-; CHECK-ORDERED-NEXT:    [[ADD:%.*]] = fadd float [[TMP14]], [[TMP15]]
+; CHECK-ORDERED-NEXT:    [[TMP13:%.*]] = load float, ptr [[ARRAYIDX4]], align 4
+; CHECK-ORDERED-NEXT:    [[ADD:%.*]] = fadd float [[TMP12]], [[TMP13]]
 ; CHECK-ORDERED-NEXT:    [[RDX]] = fadd float [[RES_014]], [[ADD]]
 ; CHECK-ORDERED-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-ORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
 ; CHECK-ORDERED:       for.end.loopexit:
-; CHECK-ORDERED-NEXT:    [[RDX_LCSSA:%.*]] = phi float [ [[RDX]], [[FOR_BODY]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[RDX_LCSSA:%.*]] = phi float [ [[RDX]], [[FOR_BODY]] ], [ [[TMP10]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    br label [[FOR_END]]
 ; CHECK-ORDERED:       for.end:
 ; CHECK-ORDERED-NEXT:    [[RES:%.*]] = phi float [ 0.000000e+00, [[ENTRY:%.*]] ], [ [[RDX_LCSSA]], [[FOR_END_LOOPEXIT]] ]
@@ -908,21 +898,19 @@ define float @fadd_of_sum(ptr noalias nocapture readonly %a, ptr noalias nocaptu
 ; CHECK-ORDERED-TF:       vector.body:
 ; CHECK-ORDERED-TF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP17:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP11]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[TMP13]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP14]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[TMP15:%.*]] = fadd <vscale x 4 x float> [[WIDE_MASKED_LOAD]], [[WIDE_MASKED_LOAD1]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP16:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> [[TMP15]], <vscale x 4 x float> splat (float -0.000000e+00)
-; CHECK-ORDERED-TF-NEXT:    [[TMP17]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[TMP16]])
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[TMP13:%.*]] = fadd <vscale x 4 x float> [[WIDE_MASKED_LOAD]], [[WIDE_MASKED_LOAD1]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP14:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> [[TMP13]], <vscale x 4 x float> splat (float -0.000000e+00)
+; CHECK-ORDERED-TF-NEXT:    [[TMP15]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[TMP14]])
 ; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP5]]
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX]], i64 [[TMP10]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP18:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
-; CHECK-ORDERED-TF-NEXT:    [[TMP19:%.*]] = extractelement <vscale x 4 x i1> [[TMP18]], i32 0
-; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP16:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
+; CHECK-ORDERED-TF-NEXT:    [[TMP17:%.*]] = extractelement <vscale x 4 x i1> [[TMP16]], i32 0
+; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK-ORDERED-TF:       middle.block:
 ; CHECK-ORDERED-TF-NEXT:    br label [[FOR_END_LOOPEXIT:%.*]]
 ; CHECK-ORDERED-TF:       scalar.ph:
@@ -933,16 +921,16 @@ define float @fadd_of_sum(ptr noalias nocapture readonly %a, ptr noalias nocaptu
 ; CHECK-ORDERED-TF-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[RES_014:%.*]] = phi float [ [[RDX:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP20:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
+; CHECK-ORDERED-TF-NEXT:    [[TMP18:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
 ; CHECK-ORDERED-TF-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP21:%.*]] = load float, ptr [[ARRAYIDX4]], align 4
-; CHECK-ORDERED-TF-NEXT:    [[ADD:%.*]] = fadd float [[TMP20]], [[TMP21]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP19:%.*]] = load float, ptr [[ARRAYIDX4]], align 4
+; CHECK-ORDERED-TF-NEXT:    [[ADD:%.*]] = fadd float [[TMP18]], [[TMP19]]
 ; CHECK-ORDERED-TF-NEXT:    [[RDX]] = fadd float [[RES_014]], [[ADD]]
 ; CHECK-ORDERED-TF-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-ORDERED-TF-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-TF-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
 ; CHECK-ORDERED-TF:       for.end.loopexit:
-; CHECK-ORDERED-TF-NEXT:    [[RDX_LCSSA:%.*]] = phi float [ [[RDX]], [[FOR_BODY]] ], [ [[TMP17]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[RDX_LCSSA:%.*]] = phi float [ [[RDX]], [[FOR_BODY]] ], [ [[TMP15]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    br label [[FOR_END]]
 ; CHECK-ORDERED-TF:       for.end:
 ; CHECK-ORDERED-TF-NEXT:    [[RES:%.*]] = phi float [ 0.000000e+00, [[ENTRY:%.*]] ], [ [[RDX_LCSSA]], [[FOR_END_LOOPEXIT]] ]
@@ -1019,46 +1007,44 @@ define float @fadd_conditional(ptr noalias nocapture readonly %a, ptr noalias no
 ; CHECK-UNORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-UNORDERED:       vector.body:
 ; CHECK-UNORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ insertelement (<vscale x 4 x float> splat (float -0.000000e+00), float 1.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP11:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ insertelement (<vscale x 4 x float> splat (float -0.000000e+00), float 1.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
-; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 0
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = fcmp une <vscale x 4 x float> [[WIDE_LOAD]], zeroinitializer
-; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]]
-; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = getelementptr float, ptr [[TMP9]], i32 0
-; CHECK-UNORDERED-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP10]], i32 4, <vscale x 4 x i1> [[TMP8]], <vscale x 4 x float> poison)
-; CHECK-UNORDERED-NEXT:    [[PREDPHI:%.*]] = select <vscale x 4 x i1> [[TMP8]], <vscale x 4 x float> [[WIDE_MASKED_LOAD]], <vscale x 4 x float> splat (float 3.000000e+00)
-; CHECK-UNORDERED-NEXT:    [[TMP11]] = fadd <vscale x 4 x float> [[VEC_PHI]], [[PREDPHI]]
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP6]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = fcmp une <vscale x 4 x float> [[WIDE_LOAD]], zeroinitializer
+; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]]
+; CHECK-UNORDERED-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP8]], i32 4, <vscale x 4 x i1> [[TMP7]], <vscale x 4 x float> poison)
+; CHECK-UNORDERED-NEXT:    [[PREDPHI:%.*]] = select <vscale x 4 x i1> [[TMP7]], <vscale x 4 x float> [[WIDE_MASKED_LOAD]], <vscale x 4 x float> splat (float 3.000000e+00)
+; CHECK-UNORDERED-NEXT:    [[TMP9]] = fadd <vscale x 4 x float> [[VEC_PHI]], [[PREDPHI]]
 ; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
-; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-UNORDERED-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-UNORDERED-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK-UNORDERED:       middle.block:
-; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = call float @llvm.vector.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[TMP11]])
+; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = call float @llvm.vector.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[TMP9]])
 ; CHECK-UNORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-UNORDERED:       scalar.ph:
 ; CHECK-UNORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-UNORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP13]], [[MIDDLE_BLOCK]] ], [ 1.000000e+00, [[ENTRY]] ]
+; CHECK-UNORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP11]], [[MIDDLE_BLOCK]] ], [ 1.000000e+00, [[ENTRY]] ]
 ; CHECK-UNORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-UNORDERED:       for.body:
 ; CHECK-UNORDERED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_INC:%.*]] ]
 ; CHECK-UNORDERED-NEXT:    [[RES:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[FADD:%.*]], [[FOR_INC]] ]
 ; CHECK-UNORDERED-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = load float, ptr [[ARRAYIDX]], align 4
-; CHECK-UNORDERED-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP14]], 0.000000e+00
+; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-UNORDERED-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP12]], 0.000000e+00
 ; CHECK-UNORDERED-NEXT:    br i1 [[TOBOOL]], label [[IF_THEN:%.*]], label [[FOR_INC]]
 ; CHECK-UNORDERED:       if.then:
 ; CHECK-UNORDERED-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
 ; CHECK-UNORDERED-NEXT:    br label [[FOR_INC]]
 ; CHECK-UNORDERED:       for.inc:
-; CHECK-UNORDERED-NEXT:    [[PHI:%.*]] = phi float [ [[TMP15]], [[IF_THEN]] ], [ 3.000000e+00, [[FOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[PHI:%.*]] = phi float [ [[TMP13]], [[IF_THEN]] ], [ 3.000000e+00, [[FOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[FADD]] = fadd float [[RES]], [[PHI]]
 ; CHECK-UNORDERED-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-UNORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ; CHECK-UNORDERED:       for.end:
-; CHECK-UNORDERED-NEXT:    [[RDX:%.*]] = phi float [ [[FADD]], [[FOR_INC]] ], [ [[TMP13]], [[MIDDLE_BLOCK]] ]
+; CHECK-UNORDERED-NEXT:    [[RDX:%.*]] = phi float [ [[FADD]], [[FOR_INC]] ], [ [[TMP11]], [[MIDDLE_BLOCK]] ]
 ; CHECK-UNORDERED-NEXT:    ret float [[RDX]]
 ;
 ; CHECK-ORDERED-LABEL: define float @fadd_conditional
@@ -1078,45 +1064,43 @@ define float @fadd_conditional(ptr noalias nocapture readonly %a, ptr noalias no
 ; CHECK-ORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED:       vector.body:
 ; CHECK-ORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 1.000000e+00, [[VECTOR_PH]] ], [ [[TMP11:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 1.000000e+00, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
-; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 0
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
-; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = fcmp une <vscale x 4 x float> [[WIDE_LOAD]], zeroinitializer
-; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]]
-; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = getelementptr float, ptr [[TMP9]], i32 0
-; CHECK-ORDERED-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP10]], i32 4, <vscale x 4 x i1> [[TMP8]], <vscale x 4 x float> poison)
-; CHECK-ORDERED-NEXT:    [[PREDPHI:%.*]] = select <vscale x 4 x i1> [[TMP8]], <vscale x 4 x float> [[WIDE_MASKED_LOAD]], <vscale x 4 x float> splat (float 3.000000e+00)
-; CHECK-ORDERED-NEXT:    [[TMP11]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[PREDPHI]])
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP6]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = fcmp une <vscale x 4 x float> [[WIDE_LOAD]], zeroinitializer
+; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]]
+; CHECK-ORDERED-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP8]], i32 4, <vscale x 4 x i1> [[TMP7]], <vscale x 4 x float> poison)
+; CHECK-ORDERED-NEXT:    [[PREDPHI:%.*]] = select <vscale x 4 x i1> [[TMP7]], <vscale x 4 x float> [[WIDE_MASKED_LOAD]], <vscale x 4 x float> splat (float 3.000000e+00)
+; CHECK-ORDERED-NEXT:    [[TMP9]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[PREDPHI]])
 ; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
-; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-ORDERED-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-ORDERED-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK-ORDERED:       middle.block:
 ; CHECK-ORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-ORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED:       scalar.ph:
 ; CHECK-ORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP11]], [[MIDDLE_BLOCK]] ], [ 1.000000e+00, [[ENTRY]] ]
+; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP9]], [[MIDDLE_BLOCK]] ], [ 1.000000e+00, [[ENTRY]] ]
 ; CHECK-ORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED:       for.body:
 ; CHECK-ORDERED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_INC:%.*]] ]
 ; CHECK-ORDERED-NEXT:    [[RES:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[FADD:%.*]], [[FOR_INC]] ]
 ; CHECK-ORDERED-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; CHECK-ORDERED-NEXT:    [[TMP13:%.*]] = load float, ptr [[ARRAYIDX]], align 4
-; CHECK-ORDERED-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP13]], 0.000000e+00
+; CHECK-ORDERED-NEXT:    [[TMP11:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-ORDERED-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP11]], 0.000000e+00
 ; CHECK-ORDERED-NEXT:    br i1 [[TOBOOL]], label [[IF_THEN:%.*]], label [[FOR_INC]]
 ; CHECK-ORDERED:       if.then:
 ; CHECK-ORDERED-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-ORDERED-NEXT:    [[TMP14:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
 ; CHECK-ORDERED-NEXT:    br label [[FOR_INC]]
 ; CHECK-ORDERED:       for.inc:
-; CHECK-ORDERED-NEXT:    [[PHI:%.*]] = phi float [ [[TMP14]], [[IF_THEN]] ], [ 3.000000e+00, [[FOR_BODY]] ]
+; CHECK-ORDERED-NEXT:    [[PHI:%.*]] = phi float [ [[TMP12]], [[IF_THEN]] ], [ 3.000000e+00, [[FOR_BODY]] ]
 ; CHECK-ORDERED-NEXT:    [[FADD]] = fadd float [[RES]], [[PHI]]
 ; CHECK-ORDERED-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-ORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ; CHECK-ORDERED:       for.end:
-; CHECK-ORDERED-NEXT:    [[RDX:%.*]] = phi float [ [[FADD]], [[FOR_INC]] ], [ [[TMP11]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[RDX:%.*]] = phi float [ [[FADD]], [[FOR_INC]] ], [ [[TMP9]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    ret float [[RDX]]
 ;
 ; CHECK-ORDERED-TF-LABEL: define float @fadd_conditional
@@ -1142,23 +1126,21 @@ define float @fadd_conditional(ptr noalias nocapture readonly %a, ptr noalias no
 ; CHECK-ORDERED-TF:       vector.body:
 ; CHECK-ORDERED-TF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 1.000000e+00, [[VECTOR_PH]] ], [ [[TMP17:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 1.000000e+00, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[TMP10]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[TMP12:%.*]] = fcmp une <vscale x 4 x float> [[WIDE_MASKED_LOAD]], zeroinitializer
-; CHECK-ORDERED-TF-NEXT:    [[TMP13:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i1> [[TMP12]], <vscale x 4 x i1> zeroinitializer
-; CHECK-ORDERED-TF-NEXT:    [[TMP14:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP15:%.*]] = getelementptr float, ptr [[TMP14]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP15]], i32 4, <vscale x 4 x i1> [[TMP13]], <vscale x 4 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[PREDPHI:%.*]] = select <vscale x 4 x i1> [[TMP13]], <vscale x 4 x float> [[WIDE_MASKED_LOAD1]], <vscale x 4 x float> splat (float 3.000000e+00)
-; CHECK-ORDERED-TF-NEXT:    [[TMP16:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> [[PREDPHI]], <vscale x 4 x float> splat (float -0.000000e+00)
-; CHECK-ORDERED-TF-NEXT:    [[TMP17]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[TMP16]])
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP10]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[TMP11:%.*]] = fcmp une <vscale x 4 x float> [[WIDE_MASKED_LOAD]], zeroinitializer
+; CHECK-ORDERED-TF-NEXT:    [[TMP12:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i1> [[TMP11]], <vscale x 4 x i1> zeroinitializer
+; CHECK-ORDERED-TF-NEXT:    [[TMP13:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]]
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP13]], i32 4, <vscale x 4 x i1> [[TMP12]], <vscale x 4 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[PREDPHI:%.*]] = select <vscale x 4 x i1> [[TMP12]], <vscale x 4 x float> [[WIDE_MASKED_LOAD1]], <vscale x 4 x float> splat (float 3.000000e+00)
+; CHECK-ORDERED-TF-NEXT:    [[TMP14:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> [[PREDPHI]], <vscale x 4 x float> splat (float -0.000000e+00)
+; CHECK-ORDERED-TF-NEXT:    [[TMP15]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[TMP14]])
 ; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP4]]
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP18:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
-; CHECK-ORDERED-TF-NEXT:    [[TMP19:%.*]] = extractelement <vscale x 4 x i1> [[TMP18]], i32 0
-; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP16:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
+; CHECK-ORDERED-TF-NEXT:    [[TMP17:%.*]] = extractelement <vscale x 4 x i1> [[TMP16]], i32 0
+; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK-ORDERED-TF:       middle.block:
 ; CHECK-ORDERED-TF-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK-ORDERED-TF:       scalar.ph:
@@ -1169,21 +1151,21 @@ define float @fadd_conditional(ptr noalias nocapture readonly %a, ptr noalias no
 ; CHECK-ORDERED-TF-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_INC:%.*]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[RES:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[FADD:%.*]], [[FOR_INC]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP20:%.*]] = load float, ptr [[ARRAYIDX]], align 4
-; CHECK-ORDERED-TF-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP20]], 0.000000e+00
+; CHECK-ORDERED-TF-NEXT:    [[TMP18:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-ORDERED-TF-NEXT:    [[TOBOOL:%.*]] = fcmp une float [[TMP18]], 0.000000e+00
 ; CHECK-ORDERED-TF-NEXT:    br i1 [[TOBOOL]], label [[IF_THEN:%.*]], label [[FOR_INC]]
 ; CHECK-ORDERED-TF:       if.then:
 ; CHECK-ORDERED-TF-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP21:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
+; CHECK-ORDERED-TF-NEXT:    [[TMP19:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
 ; CHECK-ORDERED-TF-NEXT:    br label [[FOR_INC]]
 ; CHECK-ORDERED-TF:       for.inc:
-; CHECK-ORDERED-TF-NEXT:    [[PHI:%.*]] = phi float [ [[TMP21]], [[IF_THEN]] ], [ 3.000000e+00, [[FOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[PHI:%.*]] = phi float [ [[TMP19]], [[IF_THEN]] ], [ 3.000000e+00, [[FOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[FADD]] = fadd float [[RES]], [[PHI]]
 ; CHECK-ORDERED-TF-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-ORDERED-TF-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-TF-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ; CHECK-ORDERED-TF:       for.end:
-; CHECK-ORDERED-TF-NEXT:    [[RDX:%.*]] = phi float [ [[FADD]], [[FOR_INC]] ], [ [[TMP17]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[RDX:%.*]] = phi float [ [[FADD]], [[FOR_INC]] ], [ [[TMP15]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    ret float [[RDX]]
 ;
 
@@ -1257,40 +1239,38 @@ define float @fadd_multiple(ptr noalias nocapture %a, ptr noalias nocapture %b,
 ; CHECK-UNORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-UNORDERED:       vector.body:
 ; CHECK-UNORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> splat (float -0.000000e+00), float -0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP11:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> splat (float -0.000000e+00), float -0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 0
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP7]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = fadd <vscale x 8 x float> [[VEC_PHI]], [[WIDE_LOAD]]
-; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
-; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[TMP9]], i32 0
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x float>, ptr [[TMP10]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP11]] = fadd <vscale x 8 x float> [[TMP8]], [[WIDE_LOAD1]]
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP6]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = fadd <vscale x 8 x float> [[VEC_PHI]], [[WIDE_LOAD]]
+; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x float>, ptr [[TMP8]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP9]] = fadd <vscale x 8 x float> [[TMP7]], [[WIDE_LOAD1]]
 ; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
-; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-UNORDERED-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
+; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-UNORDERED-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK-UNORDERED:       middle.block:
-; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float -0.000000e+00, <vscale x 8 x float> [[TMP11]])
+; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float -0.000000e+00, <vscale x 8 x float> [[TMP9]])
 ; CHECK-UNORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-UNORDERED:       scalar.ph:
 ; CHECK-UNORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-UNORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP13]], [[MIDDLE_BLOCK]] ], [ -0.000000e+00, [[ENTRY]] ]
+; CHECK-UNORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP11]], [[MIDDLE_BLOCK]] ], [ -0.000000e+00, [[ENTRY]] ]
 ; CHECK-UNORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-UNORDERED:       for.body:
 ; CHECK-UNORDERED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[SUM:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[ADD3:%.*]], [[FOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = load float, ptr [[ARRAYIDX]], align 4
-; CHECK-UNORDERED-NEXT:    [[ADD:%.*]] = fadd float [[SUM]], [[TMP14]]
+; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-UNORDERED-NEXT:    [[ADD:%.*]] = fadd float [[SUM]], [[TMP12]]
 ; CHECK-UNORDERED-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
-; CHECK-UNORDERED-NEXT:    [[ADD3]] = fadd float [[ADD]], [[TMP15]]
+; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
+; CHECK-UNORDERED-NEXT:    [[ADD3]] = fadd float [[ADD]], [[TMP13]]
 ; CHECK-UNORDERED-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-UNORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
 ; CHECK-UNORDERED:       for.end:
-; CHECK-UNORDERED-NEXT:    [[RDX:%.*]] = phi float [ [[ADD3]], [[FOR_BODY]] ], [ [[TMP13]], [[MIDDLE_BLOCK]] ]
+; CHECK-UNORDERED-NEXT:    [[RDX:%.*]] = phi float [ [[ADD3]], [[FOR_BODY]] ], [ [[TMP11]], [[MIDDLE_BLOCK]] ]
 ; CHECK-UNORDERED-NEXT:    ret float [[RDX]]
 ;
 ; CHECK-ORDERED-LABEL: define float @fadd_multiple
@@ -1396,71 +1376,69 @@ define float @fmuladd_strict(ptr %a, ptr %b, i64 %n) #0 {
 ; CHECK-UNORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-UNORDERED:       vector.body:
 ; CHECK-UNORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> splat (float -0.000000e+00), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP28:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 8 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP29:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI2:%.*]] = phi <vscale x 8 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP30:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI3:%.*]] = phi <vscale x 8 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP31:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> splat (float -0.000000e+00), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP26:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 8 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP27:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI2:%.*]] = phi <vscale x 8 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP28:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI3:%.*]] = phi <vscale x 8 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP29:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 0
-; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 8
-; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP9]]
-; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 16
-; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP12]]
-; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP14]], 24
-; CHECK-UNORDERED-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP15]]
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP7]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP10]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP13]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP16]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
-; CHECK-UNORDERED-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i32 0
-; CHECK-UNORDERED-NEXT:    [[TMP19:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP20:%.*]] = mul nuw i64 [[TMP19]], 8
-; CHECK-UNORDERED-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i64 [[TMP20]]
-; CHECK-UNORDERED-NEXT:    [[TMP22:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP23:%.*]] = mul nuw i64 [[TMP22]], 16
-; CHECK-UNORDERED-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i64 [[TMP23]]
-; CHECK-UNORDERED-NEXT:    [[TMP25:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP26:%.*]] = mul nuw i64 [[TMP25]], 24
-; CHECK-UNORDERED-NEXT:    [[TMP27:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i64 [[TMP26]]
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD7:%.*]] = load <vscale x 8 x float>, ptr [[TMP18]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD8:%.*]] = load <vscale x 8 x float>, ptr [[TMP21]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD9:%.*]] = load <vscale x 8 x float>, ptr [[TMP24]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD10:%.*]] = load <vscale x 8 x float>, ptr [[TMP27]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP28]] = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD]], <vscale x 8 x float> [[WIDE_LOAD7]], <vscale x 8 x float> [[VEC_PHI]])
-; CHECK-UNORDERED-NEXT:    [[TMP29]] = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD4]], <vscale x 8 x float> [[WIDE_LOAD8]], <vscale x 8 x float> [[VEC_PHI1]])
-; CHECK-UNORDERED-NEXT:    [[TMP30]] = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD5]], <vscale x 8 x float> [[WIDE_LOAD9]], <vscale x 8 x float> [[VEC_PHI2]])
-; CHECK-UNORDERED-NEXT:    [[TMP31]] = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD6]], <vscale x 8 x float> [[WIDE_LOAD10]], <vscale x 8 x float> [[VEC_PHI3]])
+; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 8
+; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP8]]
+; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = mul nuw i64 [[TMP10]], 16
+; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP11]]
+; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = mul nuw i64 [[TMP13]], 24
+; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP14]]
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP6]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP9]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP12]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP15]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
+; CHECK-UNORDERED-NEXT:    [[TMP17:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP18:%.*]] = mul nuw i64 [[TMP17]], 8
+; CHECK-UNORDERED-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP18]]
+; CHECK-UNORDERED-NEXT:    [[TMP20:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP21:%.*]] = mul nuw i64 [[TMP20]], 16
+; CHECK-UNORDERED-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP21]]
+; CHECK-UNORDERED-NEXT:    [[TMP23:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP24:%.*]] = mul nuw i64 [[TMP23]], 24
+; CHECK-UNORDERED-NEXT:    [[TMP25:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP24]]
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD7:%.*]] = load <vscale x 8 x float>, ptr [[TMP16]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD8:%.*]] = load <vscale x 8 x float>, ptr [[TMP19]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD9:%.*]] = load <vscale x 8 x float>, ptr [[TMP22]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD10:%.*]] = load <vscale x 8 x float>, ptr [[TMP25]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP26]] = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD]], <vscale x 8 x float> [[WIDE_LOAD7]], <vscale x 8 x float> [[VEC_PHI]])
+; CHECK-UNORDERED-NEXT:    [[TMP27]] = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD4]], <vscale x 8 x float> [[WIDE_LOAD8]], <vscale x 8 x float> [[VEC_PHI1]])
+; CHECK-UNORDERED-NEXT:    [[TMP28]] = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD5]], <vscale x 8 x float> [[WIDE_LOAD9]], <vscale x 8 x float> [[VEC_PHI2]])
+; CHECK-UNORDERED-NEXT:    [[TMP29]] = call <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD6]], <vscale x 8 x float> [[WIDE_LOAD10]], <vscale x 8 x float> [[VEC_PHI3]])
 ; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
-; CHECK-UNORDERED-NEXT:    [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-UNORDERED-NEXT:    br i1 [[TMP32]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
+; CHECK-UNORDERED-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-UNORDERED-NEXT:    br i1 [[TMP30]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; CHECK-UNORDERED:       middle.block:
-; CHECK-UNORDERED-NEXT:    [[BIN_RDX:%.*]] = fadd <vscale x 8 x float> [[TMP29]], [[TMP28]]
-; CHECK-UNORDERED-NEXT:    [[BIN_RDX11:%.*]] = fadd <vscale x 8 x float> [[TMP30]], [[BIN_RDX]]
-; CHECK-UNORDERED-NEXT:    [[BIN_RDX12:%.*]] = fadd <vscale x 8 x float> [[TMP31]], [[BIN_RDX11]]
-; CHECK-UNORDERED-NEXT:    [[TMP33:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float -0.000000e+00, <vscale x 8 x float> [[BIN_RDX12]])
+; CHECK-UNORDERED-NEXT:    [[BIN_RDX:%.*]] = fadd <vscale x 8 x float> [[TMP27]], [[TMP26]]
+; CHECK-UNORDERED-NEXT:    [[BIN_RDX11:%.*]] = fadd <vscale x 8 x float> [[TMP28]], [[BIN_RDX]]
+; CHECK-UNORDERED-NEXT:    [[BIN_RDX12:%.*]] = fadd <vscale x 8 x float> [[TMP29]], [[BIN_RDX11]]
+; CHECK-UNORDERED-NEXT:    [[TMP31:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float -0.000000e+00, <vscale x 8 x float> [[BIN_RDX12]])
 ; CHECK-UNORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-UNORDERED:       scalar.ph:
 ; CHECK-UNORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-UNORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP33]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
+; CHECK-UNORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP31]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
 ; CHECK-UNORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-UNORDERED:       for.body:
 ; CHECK-UNORDERED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[SUM_07:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MULADD:%.*]], [[FOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-UNORDERED-NEXT:    [[TMP34:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP32:%.*]] = load float, ptr [[ARRAYIDX]], align 4
 ; CHECK-UNORDERED-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; CHECK-UNORDERED-NEXT:    [[TMP35:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
-; CHECK-UNORDERED-NEXT:    [[MULADD]] = tail call float @llvm.fmuladd.f32(float [[TMP34]], float [[TMP35]], float [[SUM_07]])
+; CHECK-UNORDERED-NEXT:    [[TMP33:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
+; CHECK-UNORDERED-NEXT:    [[MULADD]] = tail call float @llvm.fmuladd.f32(float [[TMP32]], float [[TMP33]], float [[SUM_07]])
 ; CHECK-UNORDERED-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-UNORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; CHECK-UNORDERED:       for.end:
-; CHECK-UNORDERED-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP33]], [[MIDDLE_BLOCK]] ]
+; CHECK-UNORDERED-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP31]], [[MIDDLE_BLOCK]] ]
 ; CHECK-UNORDERED-NEXT:    ret float [[MULADD_LCSSA]]
 ;
 ; CHECK-ORDERED-LABEL: define float @fmuladd_strict
@@ -1480,68 +1458,66 @@ define float @fmuladd_strict(ptr %a, ptr %b, i64 %n) #0 {
 ; CHECK-ORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED:       vector.body:
 ; CHECK-ORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP35:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP33:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 0
-; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 8
-; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP9]]
-; CHECK-ORDERED-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 16
-; CHECK-ORDERED-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP12]]
-; CHECK-ORDERED-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP14]], 24
-; CHECK-ORDERED-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP15]]
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP7]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x float>, ptr [[TMP10]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x float>, ptr [[TMP13]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x float>, ptr [[TMP16]], align 4
-; CHECK-ORDERED-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
-; CHECK-ORDERED-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i32 0
-; CHECK-ORDERED-NEXT:    [[TMP19:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP20:%.*]] = mul nuw i64 [[TMP19]], 8
-; CHECK-ORDERED-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i64 [[TMP20]]
-; CHECK-ORDERED-NEXT:    [[TMP22:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP23:%.*]] = mul nuw i64 [[TMP22]], 16
-; CHECK-ORDERED-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i64 [[TMP23]]
-; CHECK-ORDERED-NEXT:    [[TMP25:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP26:%.*]] = mul nuw i64 [[TMP25]], 24
-; CHECK-ORDERED-NEXT:    [[TMP27:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i64 [[TMP26]]
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP18]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP21]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP24]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD7:%.*]] = load <vscale x 8 x float>, ptr [[TMP27]], align 4
-; CHECK-ORDERED-NEXT:    [[TMP28:%.*]] = fmul <vscale x 8 x float> [[WIDE_LOAD]], [[WIDE_LOAD4]]
-; CHECK-ORDERED-NEXT:    [[TMP29:%.*]] = fmul <vscale x 8 x float> [[WIDE_LOAD1]], [[WIDE_LOAD5]]
-; CHECK-ORDERED-NEXT:    [[TMP30:%.*]] = fmul <vscale x 8 x float> [[WIDE_LOAD2]], [[WIDE_LOAD6]]
-; CHECK-ORDERED-NEXT:    [[TMP31:%.*]] = fmul <vscale x 8 x float> [[WIDE_LOAD3]], [[WIDE_LOAD7]]
-; CHECK-ORDERED-NEXT:    [[TMP32:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP28]])
-; CHECK-ORDERED-NEXT:    [[TMP33:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP32]], <vscale x 8 x float> [[TMP29]])
-; CHECK-ORDERED-NEXT:    [[TMP34:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP33]], <vscale x 8 x float> [[TMP30]])
-; CHECK-ORDERED-NEXT:    [[TMP35]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP34]], <vscale x 8 x float> [[TMP31]])
+; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 8
+; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP8]]
+; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP11:%.*]] = mul nuw i64 [[TMP10]], 16
+; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP11]]
+; CHECK-ORDERED-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP14:%.*]] = mul nuw i64 [[TMP13]], 24
+; CHECK-ORDERED-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP14]]
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP6]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x float>, ptr [[TMP9]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x float>, ptr [[TMP12]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x float>, ptr [[TMP15]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
+; CHECK-ORDERED-NEXT:    [[TMP17:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP18:%.*]] = mul nuw i64 [[TMP17]], 8
+; CHECK-ORDERED-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP18]]
+; CHECK-ORDERED-NEXT:    [[TMP20:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP21:%.*]] = mul nuw i64 [[TMP20]], 16
+; CHECK-ORDERED-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP21]]
+; CHECK-ORDERED-NEXT:    [[TMP23:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP24:%.*]] = mul nuw i64 [[TMP23]], 24
+; CHECK-ORDERED-NEXT:    [[TMP25:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP24]]
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP16]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP19]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP22]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD7:%.*]] = load <vscale x 8 x float>, ptr [[TMP25]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP26:%.*]] = fmul <vscale x 8 x float> [[WIDE_LOAD]], [[WIDE_LOAD4]]
+; CHECK-ORDERED-NEXT:    [[TMP27:%.*]] = fmul <vscale x 8 x float> [[WIDE_LOAD1]], [[WIDE_LOAD5]]
+; CHECK-ORDERED-NEXT:    [[TMP28:%.*]] = fmul <vscale x 8 x float> [[WIDE_LOAD2]], [[WIDE_LOAD6]]
+; CHECK-ORDERED-NEXT:    [[TMP29:%.*]] = fmul <vscale x 8 x float> [[WIDE_LOAD3]], [[WIDE_LOAD7]]
+; CHECK-ORDERED-NEXT:    [[TMP30:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP26]])
+; CHECK-ORDERED-NEXT:    [[TMP31:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP30]], <vscale x 8 x float> [[TMP27]])
+; CHECK-ORDERED-NEXT:    [[TMP32:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP31]], <vscale x 8 x float> [[TMP28]])
+; CHECK-ORDERED-NEXT:    [[TMP33]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP32]], <vscale x 8 x float> [[TMP29]])
 ; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
-; CHECK-ORDERED-NEXT:    [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-ORDERED-NEXT:    br i1 [[TMP36]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; CHECK-ORDERED-NEXT:    [[TMP34:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-ORDERED-NEXT:    br i1 [[TMP34]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; CHECK-ORDERED:       middle.block:
 ; CHECK-ORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-ORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED:       scalar.ph:
 ; CHECK-ORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP35]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
+; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP33]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
 ; CHECK-ORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED:       for.body:
 ; CHECK-ORDERED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-NEXT:    [[SUM_07:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MULADD:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-ORDERED-NEXT:    [[TMP37:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP35:%.*]] = load float, ptr [[ARRAYIDX]], align 4
 ; CHECK-ORDERED-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; CHECK-ORDERED-NEXT:    [[TMP38:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
-; CHECK-ORDERED-NEXT:    [[MULADD]] = tail call float @llvm.fmuladd.f32(float [[TMP37]], float [[TMP38]], float [[SUM_07]])
+; CHECK-ORDERED-NEXT:    [[TMP36:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
+; CHECK-ORDERED-NEXT:    [[MULADD]] = tail call float @llvm.fmuladd.f32(float [[TMP35]], float [[TMP36]], float [[SUM_07]])
 ; CHECK-ORDERED-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-ORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
 ; CHECK-ORDERED:       for.end:
-; CHECK-ORDERED-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP35]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP33]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    ret float [[MULADD_LCSSA]]
 ;
 ; CHECK-ORDERED-TF-LABEL: define float @fmuladd_strict
@@ -1582,66 +1558,64 @@ define float @fmuladd_strict(ptr %a, ptr %b, i64 %n) #0 {
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK6:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY3]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT16:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK7:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY4]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT17:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK8:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY5]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT18:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP49:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP47:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP18:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP19:%.*]] = mul nuw i64 [[TMP18]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP19]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP21:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP22:%.*]] = mul nuw i64 [[TMP21]], 16
-; CHECK-ORDERED-TF-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP22]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP24:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP25:%.*]] = mul nuw i64 [[TMP24]], 24
-; CHECK-ORDERED-TF-NEXT:    [[TMP26:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP25]]
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP17]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP20]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP23]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP26]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[TMP27:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP28:%.*]] = getelementptr inbounds float, ptr [[TMP27]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP29:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP30:%.*]] = mul nuw i64 [[TMP29]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP31:%.*]] = getelementptr inbounds float, ptr [[TMP27]], i64 [[TMP30]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP32:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP33:%.*]] = mul nuw i64 [[TMP32]], 16
-; CHECK-ORDERED-TF-NEXT:    [[TMP34:%.*]] = getelementptr inbounds float, ptr [[TMP27]], i64 [[TMP33]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP35:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP36:%.*]] = mul nuw i64 [[TMP35]], 24
-; CHECK-ORDERED-TF-NEXT:    [[TMP37:%.*]] = getelementptr inbounds float, ptr [[TMP27]], i64 [[TMP36]]
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP28]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP31]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD14:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP34]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD15:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP37]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[TMP38:%.*]] = fmul <vscale x 8 x float> [[WIDE_MASKED_LOAD]], [[WIDE_MASKED_LOAD12]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP39:%.*]] = fmul <vscale x 8 x float> [[WIDE_MASKED_LOAD9]], [[WIDE_MASKED_LOAD13]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP40:%.*]] = fmul <vscale x 8 x float> [[WIDE_MASKED_LOAD10]], [[WIDE_MASKED_LOAD14]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP41:%.*]] = fmul <vscale x 8 x float> [[WIDE_MASKED_LOAD11]], [[WIDE_MASKED_LOAD15]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP42:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> [[TMP38]], <vscale x 8 x float> splat (float -0.000000e+00)
-; CHECK-ORDERED-TF-NEXT:    [[TMP43:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP42]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP44:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> [[TMP39]], <vscale x 8 x float> splat (float -0.000000e+00)
+; CHECK-ORDERED-TF-NEXT:    [[TMP17:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP18:%.*]] = mul nuw i64 [[TMP17]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP18]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP20:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP21:%.*]] = mul nuw i64 [[TMP20]], 16
+; CHECK-ORDERED-TF-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP21]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP23:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP24:%.*]] = mul nuw i64 [[TMP23]], 24
+; CHECK-ORDERED-TF-NEXT:    [[TMP25:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP24]]
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP16]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP19]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP22]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP25]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[TMP26:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP27:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP28:%.*]] = mul nuw i64 [[TMP27]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP29:%.*]] = getelementptr inbounds float, ptr [[TMP26]], i64 [[TMP28]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP30:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP31:%.*]] = mul nuw i64 [[TMP30]], 16
+; CHECK-ORDERED-TF-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, ptr [[TMP26]], i64 [[TMP31]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP33:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP34:%.*]] = mul nuw i64 [[TMP33]], 24
+; CHECK-ORDERED-TF-NEXT:    [[TMP35:%.*]] = getelementptr inbounds float, ptr [[TMP26]], i64 [[TMP34]]
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP26]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP29]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD14:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP32]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD15:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP35]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[TMP36:%.*]] = fmul <vscale x 8 x float> [[WIDE_MASKED_LOAD]], [[WIDE_MASKED_LOAD12]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP37:%.*]] = fmul <vscale x 8 x float> [[WIDE_MASKED_LOAD9]], [[WIDE_MASKED_LOAD13]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP38:%.*]] = fmul <vscale x 8 x float> [[WIDE_MASKED_LOAD10]], [[WIDE_MASKED_LOAD14]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP39:%.*]] = fmul <vscale x 8 x float> [[WIDE_MASKED_LOAD11]], [[WIDE_MASKED_LOAD15]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP40:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> [[TMP36]], <vscale x 8 x float> splat (float -0.000000e+00)
+; CHECK-ORDERED-TF-NEXT:    [[TMP41:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP40]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP42:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> [[TMP37]], <vscale x 8 x float> splat (float -0.000000e+00)
+; CHECK-ORDERED-TF-NEXT:    [[TMP43:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP41]], <vscale x 8 x float> [[TMP42]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP44:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> [[TMP38]], <vscale x 8 x float> splat (float -0.000000e+00)
 ; CHECK-ORDERED-TF-NEXT:    [[TMP45:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP43]], <vscale x 8 x float> [[TMP44]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP46:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> [[TMP40]], <vscale x 8 x float> splat (float -0.000000e+00)
-; CHECK-ORDERED-TF-NEXT:    [[TMP47:%.*]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP45]], <vscale x 8 x float> [[TMP46]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP48:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> [[TMP41]], <vscale x 8 x float> splat (float -0.000000e+00)
-; CHECK-ORDERED-TF-NEXT:    [[TMP49]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP47]], <vscale x 8 x float> [[TMP48]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP46:%.*]] = select <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> [[TMP39]], <vscale x 8 x float> splat (float -0.000000e+00)
+; CHECK-ORDERED-TF-NEXT:    [[TMP47]] = call float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP45]], <vscale x 8 x float> [[TMP46]])
 ; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP4]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP50:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP51:%.*]] = mul nuw i64 [[TMP50]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP52:%.*]] = add i64 [[INDEX]], [[TMP51]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP53:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP54:%.*]] = mul nuw i64 [[TMP53]], 16
-; CHECK-ORDERED-TF-NEXT:    [[TMP55:%.*]] = add i64 [[INDEX]], [[TMP54]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP56:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP57:%.*]] = mul nuw i64 [[TMP56]], 24
-; CHECK-ORDERED-TF-NEXT:    [[TMP58:%.*]] = add i64 [[INDEX]], [[TMP57]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP48:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP49:%.*]] = mul nuw i64 [[TMP48]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP50:%.*]] = add i64 [[INDEX]], [[TMP49]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP51:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP52:%.*]] = mul nuw i64 [[TMP51]], 16
+; CHECK-ORDERED-TF-NEXT:    [[TMP53:%.*]] = add i64 [[INDEX]], [[TMP52]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP54:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP55:%.*]] = mul nuw i64 [[TMP54]], 24
+; CHECK-ORDERED-TF-NEXT:    [[TMP56:%.*]] = add i64 [[INDEX]], [[TMP55]]
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT16]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP52]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT17]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP55]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT18]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP58]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP59:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
-; CHECK-ORDERED-TF-NEXT:    [[TMP60:%.*]] = extractelement <vscale x 8 x i1> [[TMP59]], i32 0
-; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP60]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT16]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP50]], i64 [[TMP9]])
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT17]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP53]], i64 [[TMP9]])
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT18]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP56]], i64 [[TMP9]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP57:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
+; CHECK-ORDERED-TF-NEXT:    [[TMP58:%.*]] = extractelement <vscale x 8 x i1> [[TMP57]], i32 0
+; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP58]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; CHECK-ORDERED-TF:       middle.block:
 ; CHECK-ORDERED-TF-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK-ORDERED-TF:       scalar.ph:
@@ -1652,15 +1626,15 @@ define float @fmuladd_strict(ptr %a, ptr %b, i64 %n) #0 {
 ; CHECK-ORDERED-TF-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[SUM_07:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MULADD:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP61:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-ORDERED-TF-NEXT:    [[TMP59:%.*]] = load float, ptr [[ARRAYIDX]], align 4
 ; CHECK-ORDERED-TF-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP62:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
-; CHECK-ORDERED-TF-NEXT:    [[MULADD]] = tail call float @llvm.fmuladd.f32(float [[TMP61]], float [[TMP62]], float [[SUM_07]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP60:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
+; CHECK-ORDERED-TF-NEXT:    [[MULADD]] = tail call float @llvm.fmuladd.f32(float [[TMP59]], float [[TMP60]], float [[SUM_07]])
 ; CHECK-ORDERED-TF-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-ORDERED-TF-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-TF-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
 ; CHECK-ORDERED-TF:       for.end:
-; CHECK-ORDERED-TF-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP49]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP47]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    ret float [[MULADD_LCSSA]]
 ;
 
@@ -1724,71 +1698,69 @@ define float @fmuladd_strict_fmf(ptr %a, ptr %b, i64 %n) #0 {
 ; CHECK-UNORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-UNORDERED:       vector.body:
 ; CHECK-UNORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> splat (float -0.000000e+00), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP28:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 8 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP29:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI2:%.*]] = phi <vscale x 8 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP30:%.*]], [[VECTOR_BODY]] ]
-; CHECK-UNORDERED-NEXT:    [[VEC_PHI3:%.*]] = phi <vscale x 8 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP31:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 8 x float> [ insertelement (<vscale x 8 x float> splat (float -0.000000e+00), float 0.000000e+00, i32 0), [[VECTOR_PH]] ], [ [[TMP26:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 8 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP27:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI2:%.*]] = phi <vscale x 8 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP28:%.*]], [[VECTOR_BODY]] ]
+; CHECK-UNORDERED-NEXT:    [[VEC_PHI3:%.*]] = phi <vscale x 8 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP29:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 0
-; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 8
-; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP9]]
-; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 16
-; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP12]]
-; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP14]], 24
-; CHECK-UNORDERED-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP15]]
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP7]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP10]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP13]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP16]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
-; CHECK-UNORDERED-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i32 0
-; CHECK-UNORDERED-NEXT:    [[TMP19:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP20:%.*]] = mul nuw i64 [[TMP19]], 8
-; CHECK-UNORDERED-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i64 [[TMP20]]
-; CHECK-UNORDERED-NEXT:    [[TMP22:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP23:%.*]] = mul nuw i64 [[TMP22]], 16
-; CHECK-UNORDERED-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i64 [[TMP23]]
-; CHECK-UNORDERED-NEXT:    [[TMP25:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-UNORDERED-NEXT:    [[TMP26:%.*]] = mul nuw i64 [[TMP25]], 24
-; CHECK-UNORDERED-NEXT:    [[TMP27:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i64 [[TMP26]]
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD7:%.*]] = load <vscale x 8 x float>, ptr [[TMP18]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD8:%.*]] = load <vscale x 8 x float>, ptr [[TMP21]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD9:%.*]] = load <vscale x 8 x float>, ptr [[TMP24]], align 4
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD10:%.*]] = load <vscale x 8 x float>, ptr [[TMP27]], align 4
-; CHECK-UNORDERED-NEXT:    [[TMP28]] = call nnan <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD]], <vscale x 8 x float> [[WIDE_LOAD7]], <vscale x 8 x float> [[VEC_PHI]])
-; CHECK-UNORDERED-NEXT:    [[TMP29]] = call nnan <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD4]], <vscale x 8 x float> [[WIDE_LOAD8]], <vscale x 8 x float> [[VEC_PHI1]])
-; CHECK-UNORDERED-NEXT:    [[TMP30]] = call nnan <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD5]], <vscale x 8 x float> [[WIDE_LOAD9]], <vscale x 8 x float> [[VEC_PHI2]])
-; CHECK-UNORDERED-NEXT:    [[TMP31]] = call nnan <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD6]], <vscale x 8 x float> [[WIDE_LOAD10]], <vscale x 8 x float> [[VEC_PHI3]])
+; CHECK-UNORDERED-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 8
+; CHECK-UNORDERED-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP8]]
+; CHECK-UNORDERED-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP11:%.*]] = mul nuw i64 [[TMP10]], 16
+; CHECK-UNORDERED-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP11]]
+; CHECK-UNORDERED-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP14:%.*]] = mul nuw i64 [[TMP13]], 24
+; CHECK-UNORDERED-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP14]]
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP6]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP9]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP12]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP15]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
+; CHECK-UNORDERED-NEXT:    [[TMP17:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP18:%.*]] = mul nuw i64 [[TMP17]], 8
+; CHECK-UNORDERED-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP18]]
+; CHECK-UNORDERED-NEXT:    [[TMP20:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP21:%.*]] = mul nuw i64 [[TMP20]], 16
+; CHECK-UNORDERED-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP21]]
+; CHECK-UNORDERED-NEXT:    [[TMP23:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-UNORDERED-NEXT:    [[TMP24:%.*]] = mul nuw i64 [[TMP23]], 24
+; CHECK-UNORDERED-NEXT:    [[TMP25:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP24]]
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD7:%.*]] = load <vscale x 8 x float>, ptr [[TMP16]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD8:%.*]] = load <vscale x 8 x float>, ptr [[TMP19]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD9:%.*]] = load <vscale x 8 x float>, ptr [[TMP22]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD10:%.*]] = load <vscale x 8 x float>, ptr [[TMP25]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP26]] = call nnan <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD]], <vscale x 8 x float> [[WIDE_LOAD7]], <vscale x 8 x float> [[VEC_PHI]])
+; CHECK-UNORDERED-NEXT:    [[TMP27]] = call nnan <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD4]], <vscale x 8 x float> [[WIDE_LOAD8]], <vscale x 8 x float> [[VEC_PHI1]])
+; CHECK-UNORDERED-NEXT:    [[TMP28]] = call nnan <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD5]], <vscale x 8 x float> [[WIDE_LOAD9]], <vscale x 8 x float> [[VEC_PHI2]])
+; CHECK-UNORDERED-NEXT:    [[TMP29]] = call nnan <vscale x 8 x float> @llvm.fmuladd.nxv8f32(<vscale x 8 x float> [[WIDE_LOAD6]], <vscale x 8 x float> [[WIDE_LOAD10]], <vscale x 8 x float> [[VEC_PHI3]])
 ; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
-; CHECK-UNORDERED-NEXT:    [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-UNORDERED-NEXT:    br i1 [[TMP32]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; CHECK-UNORDERED-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-UNORDERED-NEXT:    br i1 [[TMP30]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; CHECK-UNORDERED:       middle.block:
-; CHECK-UNORDERED-NEXT:    [[BIN_RDX:%.*]] = fadd nnan <vscale x 8 x float> [[TMP29]], [[TMP28]]
-; CHECK-UNORDERED-NEXT:    [[BIN_RDX11:%.*]] = fadd nnan <vscale x 8 x float> [[TMP30]], [[BIN_RDX]]
-; CHECK-UNORDERED-NEXT:    [[BIN_RDX12:%.*]] = fadd nnan <vscale x 8 x float> [[TMP31]], [[BIN_RDX11]]
-; CHECK-UNORDERED-NEXT:    [[TMP33:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float -0.000000e+00, <vscale x 8 x float> [[BIN_RDX12]])
+; CHECK-UNORDERED-NEXT:    [[BIN_RDX:%.*]] = fadd nnan <vscale x 8 x float> [[TMP27]], [[TMP26]]
+; CHECK-UNORDERED-NEXT:    [[BIN_RDX11:%.*]] = fadd nnan <vscale x 8 x float> [[TMP28]], [[BIN_RDX]]
+; CHECK-UNORDERED-NEXT:    [[BIN_RDX12:%.*]] = fadd nnan <vscale x 8 x float> [[TMP29]], [[BIN_RDX11]]
+; CHECK-UNORDERED-NEXT:    [[TMP31:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float -0.000000e+00, <vscale x 8 x float> [[BIN_RDX12]])
 ; CHECK-UNORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-UNORDERED:       scalar.ph:
 ; CHECK-UNORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-UNORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP33]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
+; CHECK-UNORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP31]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
 ; CHECK-UNORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-UNORDERED:       for.body:
 ; CHECK-UNORDERED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[SUM_07:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MULADD:%.*]], [[FOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-UNORDERED-NEXT:    [[TMP34:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-UNORDERED-NEXT:    [[TMP32:%.*]] = load float, ptr [[ARRAYIDX]], align 4
 ; CHECK-UNORDERED-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; CHECK-UNORDERED-NEXT:    [[TMP35:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
-; CHECK-UNORDERED-NEXT:    [[MULADD]] = tail call nnan float @llvm.fmuladd.f32(float [[TMP34]], float [[TMP35]], float [[SUM_07]])
+; CHECK-UNORDERED-NEXT:    [[TMP33:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
+; CHECK-UNORDERED-NEXT:    [[MULADD]] = tail call nnan float @llvm.fmuladd.f32(float [[TMP32]], float [[TMP33]], float [[SUM_07]])
 ; CHECK-UNORDERED-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-UNORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-UNORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
 ; CHECK-UNORDERED:       for.end:
-; CHECK-UNORDERED-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP33]], [[MIDDLE_BLOCK]] ]
+; CHECK-UNORDERED-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP31]], [[MIDDLE_BLOCK]] ]
 ; CHECK-UNORDERED-NEXT:    ret float [[MULADD_LCSSA]]
 ;
 ; CHECK-ORDERED-LABEL: define float @fmuladd_strict_fmf
@@ -1808,68 +1780,66 @@ define float @fmuladd_strict_fmf(ptr %a, ptr %b, i64 %n) #0 {
 ; CHECK-ORDERED-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-ORDERED:       vector.body:
 ; CHECK-ORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP35:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP33:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 0
-; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 8
-; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP9]]
-; CHECK-ORDERED-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 16
-; CHECK-ORDERED-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP12]]
-; CHECK-ORDERED-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP14]], 24
-; CHECK-ORDERED-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP15]]
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP7]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x float>, ptr [[TMP10]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x float>, ptr [[TMP13]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x float>, ptr [[TMP16]], align 4
-; CHECK-ORDERED-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
-; CHECK-ORDERED-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i32 0
-; CHECK-ORDERED-NEXT:    [[TMP19:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP20:%.*]] = mul nuw i64 [[TMP19]], 8
-; CHECK-ORDERED-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i64 [[TMP20]]
-; CHECK-ORDERED-NEXT:    [[TMP22:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP23:%.*]] = mul nuw i64 [[TMP22]], 16
-; CHECK-ORDERED-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i64 [[TMP23]]
-; CHECK-ORDERED-NEXT:    [[TMP25:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-NEXT:    [[TMP26:%.*]] = mul nuw i64 [[TMP25]], 24
-; CHECK-ORDERED-NEXT:    [[TMP27:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i64 [[TMP26]]
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP18]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP21]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP24]], align 4
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD7:%.*]] = load <vscale x 8 x float>, ptr [[TMP27]], align 4
-; CHECK-ORDERED-NEXT:    [[TMP28:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_LOAD]], [[WIDE_LOAD4]]
-; CHECK-ORDERED-NEXT:    [[TMP29:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_LOAD1]], [[WIDE_LOAD5]]
-; CHECK-ORDERED-NEXT:    [[TMP30:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_LOAD2]], [[WIDE_LOAD6]]
-; CHECK-ORDERED-NEXT:    [[TMP31:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_LOAD3]], [[WIDE_LOAD7]]
-; CHECK-ORDERED-NEXT:    [[TMP32:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP28]])
-; CHECK-ORDERED-NEXT:    [[TMP33:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP32]], <vscale x 8 x float> [[TMP29]])
-; CHECK-ORDERED-NEXT:    [[TMP34:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP33]], <vscale x 8 x float> [[TMP30]])
-; CHECK-ORDERED-NEXT:    [[TMP35]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP34]], <vscale x 8 x float> [[TMP31]])
+; CHECK-ORDERED-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 8
+; CHECK-ORDERED-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP8]]
+; CHECK-ORDERED-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP11:%.*]] = mul nuw i64 [[TMP10]], 16
+; CHECK-ORDERED-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP11]]
+; CHECK-ORDERED-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP14:%.*]] = mul nuw i64 [[TMP13]], 24
+; CHECK-ORDERED-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i64 [[TMP14]]
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x float>, ptr [[TMP6]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x float>, ptr [[TMP9]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x float>, ptr [[TMP12]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x float>, ptr [[TMP15]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
+; CHECK-ORDERED-NEXT:    [[TMP17:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP18:%.*]] = mul nuw i64 [[TMP17]], 8
+; CHECK-ORDERED-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP18]]
+; CHECK-ORDERED-NEXT:    [[TMP20:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP21:%.*]] = mul nuw i64 [[TMP20]], 16
+; CHECK-ORDERED-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP21]]
+; CHECK-ORDERED-NEXT:    [[TMP23:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-NEXT:    [[TMP24:%.*]] = mul nuw i64 [[TMP23]], 24
+; CHECK-ORDERED-NEXT:    [[TMP25:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP24]]
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 8 x float>, ptr [[TMP16]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x float>, ptr [[TMP19]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 8 x float>, ptr [[TMP22]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD7:%.*]] = load <vscale x 8 x float>, ptr [[TMP25]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP26:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_LOAD]], [[WIDE_LOAD4]]
+; CHECK-ORDERED-NEXT:    [[TMP27:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_LOAD1]], [[WIDE_LOAD5]]
+; CHECK-ORDERED-NEXT:    [[TMP28:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_LOAD2]], [[WIDE_LOAD6]]
+; CHECK-ORDERED-NEXT:    [[TMP29:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_LOAD3]], [[WIDE_LOAD7]]
+; CHECK-ORDERED-NEXT:    [[TMP30:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP26]])
+; CHECK-ORDERED-NEXT:    [[TMP31:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP30]], <vscale x 8 x float> [[TMP27]])
+; CHECK-ORDERED-NEXT:    [[TMP32:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP31]], <vscale x 8 x float> [[TMP28]])
+; CHECK-ORDERED-NEXT:    [[TMP33]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP32]], <vscale x 8 x float> [[TMP29]])
 ; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
-; CHECK-ORDERED-NEXT:    [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-ORDERED-NEXT:    br i1 [[TMP36]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; CHECK-ORDERED-NEXT:    [[TMP34:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-ORDERED-NEXT:    br i1 [[TMP34]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; CHECK-ORDERED:       middle.block:
 ; CHECK-ORDERED-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-ORDERED-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; CHECK-ORDERED:       scalar.ph:
 ; CHECK-ORDERED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP35]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
+; CHECK-ORDERED-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP33]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
 ; CHECK-ORDERED-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-ORDERED:       for.body:
 ; CHECK-ORDERED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-NEXT:    [[SUM_07:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MULADD:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-ORDERED-NEXT:    [[TMP37:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-ORDERED-NEXT:    [[TMP35:%.*]] = load float, ptr [[ARRAYIDX]], align 4
 ; CHECK-ORDERED-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; CHECK-ORDERED-NEXT:    [[TMP38:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
-; CHECK-ORDERED-NEXT:    [[MULADD]] = tail call nnan float @llvm.fmuladd.f32(float [[TMP37]], float [[TMP38]], float [[SUM_07]])
+; CHECK-ORDERED-NEXT:    [[TMP36:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
+; CHECK-ORDERED-NEXT:    [[MULADD]] = tail call nnan float @llvm.fmuladd.f32(float [[TMP35]], float [[TMP36]], float [[SUM_07]])
 ; CHECK-ORDERED-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-ORDERED-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
 ; CHECK-ORDERED:       for.end:
-; CHECK-ORDERED-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP35]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP33]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-NEXT:    ret float [[MULADD_LCSSA]]
 ;
 ; CHECK-ORDERED-TF-LABEL: define float @fmuladd_strict_fmf
@@ -1910,66 +1880,64 @@ define float @fmuladd_strict_fmf(ptr %a, ptr %b, i64 %n) #0 {
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK6:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY3]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT16:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK7:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY4]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT17:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK8:%.*]] = phi <vscale x 8 x i1> [ [[ACTIVE_LANE_MASK_ENTRY5]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT18:%.*]], [[VECTOR_BODY]] ]
-; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP49:%.*]], [[VECTOR_BODY]] ]
+; CHECK-ORDERED-TF-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP47:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP18:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP19:%.*]] = mul nuw i64 [[TMP18]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP19]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP21:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP22:%.*]] = mul nuw i64 [[TMP21]], 16
-; CHECK-ORDERED-TF-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP22]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP24:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP25:%.*]] = mul nuw i64 [[TMP24]], 24
-; CHECK-ORDERED-TF-NEXT:    [[TMP26:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP25]]
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP17]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP20]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP23]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP26]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[TMP27:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP28:%.*]] = getelementptr inbounds float, ptr [[TMP27]], i32 0
-; CHECK-ORDERED-TF-NEXT:    [[TMP29:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP30:%.*]] = mul nuw i64 [[TMP29]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP31:%.*]] = getelementptr inbounds float, ptr [[TMP27]], i64 [[TMP30]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP32:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP33:%.*]] = mul nuw i64 [[TMP32]], 16
-; CHECK-ORDERED-TF-NEXT:    [[TMP34:%.*]] = getelementptr inbounds float, ptr [[TMP27]], i64 [[TMP33]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP35:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP36:%.*]] = mul nuw i64 [[TMP35]], 24
-; CHECK-ORDERED-TF-NEXT:    [[TMP37:%.*]] = getelementptr inbounds float, ptr [[TMP27]], i64 [[TMP36]]
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP28]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP31]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD14:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP34]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD15:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP37]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
-; CHECK-ORDERED-TF-NEXT:    [[TMP38:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_MASKED_LOAD]], [[WIDE_MASKED_LOAD12]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP39:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_MASKED_LOAD9]], [[WIDE_MASKED_LOAD13]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP40:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_MASKED_LOAD10]], [[WIDE_MASKED_LOAD14]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP41:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_MASKED_LOAD11]], [[WIDE_MASKED_LOAD15]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP42:%.*]] = select nnan <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> [[TMP38]], <vscale x 8 x float> splat (float -0.000000e+00)
-; CHECK-ORDERED-TF-NEXT:    [[TMP43:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP42]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP44:%.*]] = select nnan <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> [[TMP39]], <vscale x 8 x float> splat (float -0.000000e+00)
+; CHECK-ORDERED-TF-NEXT:    [[TMP17:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP18:%.*]] = mul nuw i64 [[TMP17]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP18]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP20:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP21:%.*]] = mul nuw i64 [[TMP20]], 16
+; CHECK-ORDERED-TF-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP21]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP23:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP24:%.*]] = mul nuw i64 [[TMP23]], 24
+; CHECK-ORDERED-TF-NEXT:    [[TMP25:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[TMP24]]
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP16]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP19]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP22]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP25]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[TMP26:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP27:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP28:%.*]] = mul nuw i64 [[TMP27]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP29:%.*]] = getelementptr inbounds float, ptr [[TMP26]], i64 [[TMP28]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP30:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP31:%.*]] = mul nuw i64 [[TMP30]], 16
+; CHECK-ORDERED-TF-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, ptr [[TMP26]], i64 [[TMP31]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP33:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP34:%.*]] = mul nuw i64 [[TMP33]], 24
+; CHECK-ORDERED-TF-NEXT:    [[TMP35:%.*]] = getelementptr inbounds float, ptr [[TMP26]], i64 [[TMP34]]
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP26]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP29]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD14:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP32]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[WIDE_MASKED_LOAD15:%.*]] = call <vscale x 8 x float> @llvm.masked.load.nxv8f32.p0(ptr [[TMP35]], i32 4, <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> poison)
+; CHECK-ORDERED-TF-NEXT:    [[TMP36:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_MASKED_LOAD]], [[WIDE_MASKED_LOAD12]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP37:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_MASKED_LOAD9]], [[WIDE_MASKED_LOAD13]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP38:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_MASKED_LOAD10]], [[WIDE_MASKED_LOAD14]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP39:%.*]] = fmul nnan <vscale x 8 x float> [[WIDE_MASKED_LOAD11]], [[WIDE_MASKED_LOAD15]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP40:%.*]] = select nnan <vscale x 8 x i1> [[ACTIVE_LANE_MASK]], <vscale x 8 x float> [[TMP36]], <vscale x 8 x float> splat (float -0.000000e+00)
+; CHECK-ORDERED-TF-NEXT:    [[TMP41:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[VEC_PHI]], <vscale x 8 x float> [[TMP40]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP42:%.*]] = select nnan <vscale x 8 x i1> [[ACTIVE_LANE_MASK6]], <vscale x 8 x float> [[TMP37]], <vscale x 8 x float> splat (float -0.000000e+00)
+; CHECK-ORDERED-TF-NEXT:    [[TMP43:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP41]], <vscale x 8 x float> [[TMP42]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP44:%.*]] = select nnan <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> [[TMP38]], <vscale x 8 x float> splat (float -0.000000e+00)
 ; CHECK-ORDERED-TF-NEXT:    [[TMP45:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP43]], <vscale x 8 x float> [[TMP44]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP46:%.*]] = select nnan <vscale x 8 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 8 x float> [[TMP40]], <vscale x 8 x float> splat (float -0.000000e+00)
-; CHECK-ORDERED-TF-NEXT:    [[TMP47:%.*]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP45]], <vscale x 8 x float> [[TMP46]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP48:%.*]] = select nnan <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> [[TMP41]], <vscale x 8 x float> splat (float -0.000000e+00)
-; CHECK-ORDERED-TF-NEXT:    [[TMP49]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP47]], <vscale x 8 x float> [[TMP48]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP46:%.*]] = select nnan <vscale x 8 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 8 x float> [[TMP39]], <vscale x 8 x float> splat (float -0.000000e+00)
+; CHECK-ORDERED-TF-NEXT:    [[TMP47]] = call nnan float @llvm.vector.reduce.fadd.nxv8f32(float [[TMP45]], <vscale x 8 x float> [[TMP46]])
 ; CHECK-ORDERED-TF-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP4]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP50:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP51:%.*]] = mul nuw i64 [[TMP50]], 8
-; CHECK-ORDERED-TF-NEXT:    [[TMP52:%.*]] = add i64 [[INDEX]], [[TMP51]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP53:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP54:%.*]] = mul nuw i64 [[TMP53]], 16
-; CHECK-ORDERED-TF-NEXT:    [[TMP55:%.*]] = add i64 [[INDEX]], [[TMP54]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP56:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-ORDERED-TF-NEXT:    [[TMP57:%.*]] = mul nuw i64 [[TMP56]], 24
-; CHECK-ORDERED-TF-NEXT:    [[TMP58:%.*]] = add i64 [[INDEX]], [[TMP57]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP48:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP49:%.*]] = mul nuw i64 [[TMP48]], 8
+; CHECK-ORDERED-TF-NEXT:    [[TMP50:%.*]] = add i64 [[INDEX]], [[TMP49]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP51:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP52:%.*]] = mul nuw i64 [[TMP51]], 16
+; CHECK-ORDERED-TF-NEXT:    [[TMP53:%.*]] = add i64 [[INDEX]], [[TMP52]]
+; CHECK-ORDERED-TF-NEXT:    [[TMP54:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-ORDERED-TF-NEXT:    [[TMP55:%.*]] = mul nuw i64 [[TMP54]], 24
+; CHECK-ORDERED-TF-NEXT:    [[TMP56:%.*]] = add i64 [[INDEX]], [[TMP55]]
 ; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[INDEX]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT16]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP52]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT17]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP55]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT18]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP58]], i64 [[TMP9]])
-; CHECK-ORDERED-TF-NEXT:    [[TMP59:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
-; CHECK-ORDERED-TF-NEXT:    [[TMP60:%.*]] = extractelement <vscale x 8 x i1> [[TMP59]], i32 0
-; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP60]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT16]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP50]], i64 [[TMP9]])
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT17]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP53]], i64 [[TMP9]])
+; CHECK-ORDERED-TF-NEXT:    [[ACTIVE_LANE_MASK_NEXT18]] = call <vscale x 8 x i1> @llvm.get.active.lane.mask.nxv8i1.i64(i64 [[TMP56]], i64 [[TMP9]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP57:%.*]] = xor <vscale x 8 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
+; CHECK-ORDERED-TF-NEXT:    [[TMP58:%.*]] = extractelement <vscale x 8 x i1> [[TMP57]], i32 0
+; CHECK-ORDERED-TF-NEXT:    br i1 [[TMP58]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; CHECK-ORDERED-TF:       middle.block:
 ; CHECK-ORDERED-TF-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK-ORDERED-TF:       scalar.ph:
@@ -1980,15 +1948,15 @@ define float @fmuladd_strict_fmf(ptr %a, ptr %b, i64 %n) #0 {
 ; CHECK-ORDERED-TF-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[SUM_07:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[MULADD:%.*]], [[FOR_BODY]] ]
 ; CHECK-ORDERED-TF-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP61:%.*]] = load float, ptr [[ARRAYIDX]], align 4
+; CHECK-ORDERED-TF-NEXT:    [[TMP59:%.*]] = load float, ptr [[ARRAYIDX]], align 4
 ; CHECK-ORDERED-TF-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; CHECK-ORDERED-TF-NEXT:    [[TMP62:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
-; CHECK-ORDERED-TF-NEXT:    [[MULADD]] = tail call nnan float @llvm.fmuladd.f32(float [[TMP61]], float [[TMP62]], float [[SUM_07]])
+; CHECK-ORDERED-TF-NEXT:    [[TMP60:%.*]] = load float, ptr [[ARRAYIDX2]], align 4
+; CHECK-ORDERED-TF-NEXT:    [[MULADD]] = tail call nnan float @llvm.fmuladd.f32(float [[TMP59]], float [[TMP60]], float [[SUM_07]])
 ; CHECK-ORDERED-TF-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-ORDERED-TF-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-ORDERED-TF-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
 ; CHECK-ORDERED-TF:       for.end:
-; CHECK-ORDERED-TF-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP49]], [[MIDDLE_BLOCK]] ]
+; CHECK-ORDERED-TF-NEXT:    [[MULADD_LCSSA:%.*]] = phi float [ [[MULADD]], [[FOR_BODY]] ], [ [[TMP47]], [[MIDDLE_BLOCK]] ]
 ; CHECK-ORDERED-TF-NEXT:    ret float [[MULADD_LCSSA]]
 ;
 
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/simple_early_exit.ll b/llvm/test/Transforms/LoopVectorize/AArch64/simple_early_exit.ll
index 3bb20e2..eaf85694 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/simple_early_exit.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/simple_early_exit.ll
@@ -30,11 +30,9 @@ define i64 @same_exit_block_pre_inc_use1() #1 {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP12]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP11]], align 1
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 16 x i8>, ptr [[TMP14]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 16 x i8>, ptr [[TMP13]], align 1
 ; CHECK-NEXT:    [[TMP16:%.*]] = icmp ne <vscale x 16 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP17:%.*]] = call i1 @llvm.vector.reduce.or.nxv16i1(<vscale x 16 x i1> [[TMP16]])
@@ -114,8 +112,7 @@ define i64 @same_exit_block_pre_inc_use4() {
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 3, i64 4>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp uge <2 x i64> [[VEC_IND]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT2]] = add nuw i64 [[INDEX1]], 2
 ; CHECK-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v2i1(<2 x i1> [[TMP4]])
@@ -126,14 +123,14 @@ define i64 @same_exit_block_pre_inc_use4() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v2i1(<2 x i1> [[TMP4]], i1 true)
 ; CHECK-NEXT:    [[TMP8:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i64 3, [[TMP8]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -189,8 +186,7 @@ define i64 @loop_contains_safe_call() #1 {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = call fast <4 x float> @llvm.sqrt.v4f32(<4 x float> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[TMP5:%.*]] = fcmp fast oge <4 x float> [[TMP3]], splat (float 3.000000e+00)
 ; CHECK-NEXT:    [[INDEX_NEXT2]] = add nuw i64 [[INDEX1]], 4
@@ -201,14 +197,14 @@ define i64 @loop_contains_safe_call() #1 {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP5]], i1 true)
 ; CHECK-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i64 3, [[TMP9]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -276,8 +272,7 @@ define i64 @loop_contains_safe_div() #1 {
 ; CHECK-NEXT:    [[INDEX2:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX1:%.*]] = add i64 3, [[INDEX2]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[P1]], i64 [[OFFSET_IDX1]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP13:%.*]] = udiv <vscale x 4 x i32> [[WIDE_LOAD]], splat (i32 20000)
 ; CHECK-NEXT:    [[TMP15:%.*]] = icmp ne <vscale x 4 x i32> [[TMP13]], splat (i32 1)
 ; CHECK-NEXT:    [[INDEX_NEXT2]] = add nuw i64 [[INDEX2]], [[TMP5]]
@@ -356,12 +351,10 @@ define i64 @loop_contains_load_after_early_exit(ptr dereferenceable(1024) align(
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp ne <4 x i32> [[WIDE_LOAD]], splat (i32 1)
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i64, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i64, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i64>, ptr [[TMP5]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]])
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT3]], 64
@@ -371,14 +364,14 @@ define i64 @loop_contains_load_after_early_exit(ptr dereferenceable(1024) align(
 ; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <4 x i64> [[WIDE_LOAD2]], i32 3
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP6]], i1 true)
 ; CHECK-NEXT:    [[TMP11:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i64 3, [[TMP11]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -457,11 +450,9 @@ define i32 @diff_exit_block_needs_scev_check(i32 %end) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[P1]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP11]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP10]], align 4
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[P2]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP13]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP12]], align 4
 ; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP15:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP14]])
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/single-early-exit-interleave.ll b/llvm/test/Transforms/LoopVectorize/AArch64/single-early-exit-interleave.ll
index 61ef3ce..070f658 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/single-early-exit-interleave.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/single-early-exit-interleave.ll
@@ -14,40 +14,88 @@ define i64 @same_exit_block_pre_inc_use1() #0 {
 ; CHECK-NEXT:    call void @init_mem(ptr [[P1]], i64 1024)
 ; CHECK-NEXT:    call void @init_mem(ptr [[P2]], i64 1024)
 ; CHECK-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 16
-; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 64
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 510, [[TMP1]]
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16
+; CHECK-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 64
 ; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 510, [[TMP3]]
 ; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 510, [[N_MOD_VF]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
-; CHECK-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 16
+; CHECK-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 64
 ; CHECK-NEXT:    [[INDEX_NEXT:%.*]] = add i64 3, [[N_VEC]]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[LOOP]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 16 x i8>, ptr [[TMP8]], align 1
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[TMP9]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <vscale x 16 x i8>, ptr [[TMP10]], align 1
-; CHECK-NEXT:    [[TMP32:%.*]] = icmp ne <vscale x 16 x i8> [[WIDE_LOAD4]], [[WIDE_LOAD8]]
+; CHECK-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 16
+; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i64 [[TMP9]]
+; CHECK-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP27:%.*]] = mul nuw i64 [[TMP11]], 32
+; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i64 [[TMP27]]
+; CHECK-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP14]], 48
+; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i64 [[TMP15]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP7]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 16 x i8>, ptr [[TMP10]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 16 x i8>, ptr [[TMP13]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 16 x i8>, ptr [[TMP28]], align 1
+; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
+; CHECK-NEXT:    [[TMP18:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP19:%.*]] = mul nuw i64 [[TMP18]], 16
+; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i8, ptr [[TMP29]], i64 [[TMP19]]
+; CHECK-NEXT:    [[TMP21:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP22:%.*]] = mul nuw i64 [[TMP21]], 32
+; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i8, ptr [[TMP29]], i64 [[TMP22]]
+; CHECK-NEXT:    [[TMP24:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP25:%.*]] = mul nuw i64 [[TMP24]], 48
+; CHECK-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i8, ptr [[TMP29]], i64 [[TMP25]]
+; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 16 x i8>, ptr [[TMP29]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <vscale x 16 x i8>, ptr [[TMP20]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <vscale x 16 x i8>, ptr [[TMP23]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <vscale x 16 x i8>, ptr [[TMP26]], align 1
+; CHECK-NEXT:    [[TMP32:%.*]] = icmp ne <vscale x 16 x i8> [[WIDE_LOAD]], [[WIDE_LOAD5]]
+; CHECK-NEXT:    [[TMP30:%.*]] = icmp ne <vscale x 16 x i8> [[WIDE_LOAD2]], [[WIDE_LOAD6]]
+; CHECK-NEXT:    [[TMP31:%.*]] = icmp ne <vscale x 16 x i8> [[WIDE_LOAD3]], [[WIDE_LOAD7]]
+; CHECK-NEXT:    [[TMP59:%.*]] = icmp ne <vscale x 16 x i8> [[WIDE_LOAD4]], [[WIDE_LOAD8]]
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], [[TMP5]]
-; CHECK-NEXT:    [[TMP12:%.*]] = call i1 @llvm.vector.reduce.or.nxv16i1(<vscale x 16 x i1> [[TMP32]])
-; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT3]], [[N_VEC]]
-; CHECK-NEXT:    [[TMP14:%.*]] = or i1 [[TMP12]], [[TMP13]]
-; CHECK-NEXT:    br i1 [[TMP14]], label [[MIDDLE_SPLIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-NEXT:    [[TMP34:%.*]] = or <vscale x 16 x i1> [[TMP32]], [[TMP30]]
+; CHECK-NEXT:    [[TMP37:%.*]] = or <vscale x 16 x i1> [[TMP34]], [[TMP31]]
+; CHECK-NEXT:    [[TMP33:%.*]] = or <vscale x 16 x i1> [[TMP37]], [[TMP59]]
+; CHECK-NEXT:    [[TMP12:%.*]] = call i1 @llvm.vector.reduce.or.nxv16i1(<vscale x 16 x i1> [[TMP33]])
+; CHECK-NEXT:    [[TMP35:%.*]] = icmp eq i64 [[INDEX_NEXT3]], [[N_VEC]]
+; CHECK-NEXT:    [[TMP36:%.*]] = or i1 [[TMP12]], [[TMP35]]
+; CHECK-NEXT:    br i1 [[TMP36]], label [[MIDDLE_SPLIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP12]], label [[VECTOR_EARLY_EXIT:%.*]], label [[LOOP_INC:%.*]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 510, [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[LOOP_END:%.*]], label [[SCALAR_PH]]
 ; CHECK:       vector.early.exit:
+; CHECK-NEXT:    [[TMP39:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP40:%.*]] = mul nuw i64 [[TMP39]], 16
+; CHECK-NEXT:    [[TMP41:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.nxv16i1(<vscale x 16 x i1> [[TMP59]], i1 true)
+; CHECK-NEXT:    [[TMP42:%.*]] = mul i64 [[TMP40]], 3
+; CHECK-NEXT:    [[TMP43:%.*]] = add i64 [[TMP42]], [[TMP41]]
+; CHECK-NEXT:    [[TMP44:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.nxv16i1(<vscale x 16 x i1> [[TMP31]], i1 true)
+; CHECK-NEXT:    [[TMP45:%.*]] = mul i64 [[TMP40]], 2
+; CHECK-NEXT:    [[TMP46:%.*]] = add i64 [[TMP45]], [[TMP44]]
+; CHECK-NEXT:    [[TMP47:%.*]] = icmp ne i64 [[TMP44]], [[TMP40]]
+; CHECK-NEXT:    [[TMP48:%.*]] = select i1 [[TMP47]], i64 [[TMP46]], i64 [[TMP43]]
+; CHECK-NEXT:    [[TMP49:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.nxv16i1(<vscale x 16 x i1> [[TMP30]], i1 true)
+; CHECK-NEXT:    [[TMP50:%.*]] = mul i64 [[TMP40]], 1
+; CHECK-NEXT:    [[TMP51:%.*]] = add i64 [[TMP50]], [[TMP49]]
+; CHECK-NEXT:    [[TMP52:%.*]] = icmp ne i64 [[TMP49]], [[TMP40]]
+; CHECK-NEXT:    [[TMP53:%.*]] = select i1 [[TMP52]], i64 [[TMP51]], i64 [[TMP48]]
 ; CHECK-NEXT:    [[TMP61:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.nxv16i1(<vscale x 16 x i1> [[TMP32]], i1 true)
-; CHECK-NEXT:    [[TMP16:%.*]] = add i64 [[INDEX1]], [[TMP61]]
+; CHECK-NEXT:    [[TMP55:%.*]] = mul i64 [[TMP40]], 0
+; CHECK-NEXT:    [[TMP56:%.*]] = add i64 [[TMP55]], [[TMP61]]
+; CHECK-NEXT:    [[TMP57:%.*]] = icmp ne i64 [[TMP61]], [[TMP40]]
+; CHECK-NEXT:    [[TMP58:%.*]] = select i1 [[TMP57]], i64 [[TMP56]], i64 [[TMP53]]
+; CHECK-NEXT:    [[TMP16:%.*]] = add i64 [[INDEX1]], [[TMP58]]
 ; CHECK-NEXT:    [[TMP17:%.*]] = add i64 3, [[TMP16]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/store-costs-sve.ll b/llvm/test/Transforms/LoopVectorize/AArch64/store-costs-sve.ll
index 77b768e..51efbe9 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/store-costs-sve.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/store-costs-sve.ll
@@ -29,11 +29,10 @@ define void @cost_store_i8(ptr %dst) #0 {
 ; DEFAULT:       vector.body:
 ; DEFAULT-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; DEFAULT-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INDEX]]
-; DEFAULT-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[TMP9]], i32 0
 ; DEFAULT-NEXT:    [[TMP22:%.*]] = call i64 @llvm.vscale.i64()
 ; DEFAULT-NEXT:    [[TMP23:%.*]] = mul nuw i64 [[TMP22]], 16
 ; DEFAULT-NEXT:    [[TMP24:%.*]] = getelementptr i8, ptr [[TMP9]], i64 [[TMP23]]
-; DEFAULT-NEXT:    store <vscale x 16 x i8> zeroinitializer, ptr [[TMP10]], align 1
+; DEFAULT-NEXT:    store <vscale x 16 x i8> zeroinitializer, ptr [[TMP9]], align 1
 ; DEFAULT-NEXT:    store <vscale x 16 x i8> zeroinitializer, ptr [[TMP24]], align 1
 ; DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP7]]
 ; DEFAULT-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -59,8 +58,7 @@ define void @cost_store_i8(ptr %dst) #0 {
 ; DEFAULT:       vec.epilog.vector.body:
 ; DEFAULT-NEXT:    [[INDEX5:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT6:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; DEFAULT-NEXT:    [[TMP19:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INDEX5]]
-; DEFAULT-NEXT:    [[TMP20:%.*]] = getelementptr i8, ptr [[TMP19]], i32 0
-; DEFAULT-NEXT:    store <vscale x 8 x i8> zeroinitializer, ptr [[TMP20]], align 1
+; DEFAULT-NEXT:    store <vscale x 8 x i8> zeroinitializer, ptr [[TMP19]], align 1
 ; DEFAULT-NEXT:    [[INDEX_NEXT6]] = add nuw i64 [[INDEX5]], [[TMP17]]
 ; DEFAULT-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT6]], [[N_VEC3]]
 ; DEFAULT-NEXT:    br i1 [[TMP21]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
@@ -104,8 +102,7 @@ define void @cost_store_i8(ptr %dst) #0 {
 ; PRED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; PRED-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 16 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; PRED-NEXT:    [[TMP13:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INDEX]]
-; PRED-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[TMP13]], i32 0
-; PRED-NEXT:    call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> zeroinitializer, ptr [[TMP14]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
+; PRED-NEXT:    call void @llvm.masked.store.nxv16i8.p0(<vscale x 16 x i8> zeroinitializer, ptr [[TMP13]], i32 1, <vscale x 16 x i1> [[ACTIVE_LANE_MASK]])
 ; PRED-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
 ; PRED-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i64(i64 [[INDEX]], i64 [[TMP11]])
 ; PRED-NEXT:    [[TMP15:%.*]] = xor <vscale x 16 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
@@ -169,9 +166,8 @@ define void @trunc_store(ptr %dst, ptr %src, i16 %x) #1 {
 ; DEFAULT-NEXT:    [[TMP8:%.*]] = and <16 x i8> [[TMP5]], [[TMP7]]
 ; DEFAULT-NEXT:    [[TMP9:%.*]] = and <16 x i8> [[TMP5]], [[TMP7]]
 ; DEFAULT-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INDEX]]
-; DEFAULT-NEXT:    [[TMP12:%.*]] = getelementptr i8, ptr [[TMP10]], i32 0
 ; DEFAULT-NEXT:    [[TMP13:%.*]] = getelementptr i8, ptr [[TMP10]], i32 16
-; DEFAULT-NEXT:    store <16 x i8> [[TMP8]], ptr [[TMP12]], align 1, !alias.scope [[META8:![0-9]+]], !noalias [[META5]]
+; DEFAULT-NEXT:    store <16 x i8> [[TMP8]], ptr [[TMP10]], align 1, !alias.scope [[META8:![0-9]+]], !noalias [[META5]]
 ; DEFAULT-NEXT:    store <16 x i8> [[TMP9]], ptr [[TMP13]], align 1, !alias.scope [[META8]], !noalias [[META5]]
 ; DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
 ; DEFAULT-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 992
@@ -194,8 +190,7 @@ define void @trunc_store(ptr %dst, ptr %src, i16 %x) #1 {
 ; DEFAULT-NEXT:    [[TMP18:%.*]] = trunc <8 x i64> [[BROADCAST_SPLAT8]] to <8 x i8>
 ; DEFAULT-NEXT:    [[TMP14:%.*]] = and <8 x i8> [[TMP18]], [[TMP15]]
 ; DEFAULT-NEXT:    [[TMP26:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INDEX5]]
-; DEFAULT-NEXT:    [[TMP27:%.*]] = getelementptr i8, ptr [[TMP26]], i32 0
-; DEFAULT-NEXT:    store <8 x i8> [[TMP14]], ptr [[TMP27]], align 1, !alias.scope [[META8]], !noalias [[META5]]
+; DEFAULT-NEXT:    store <8 x i8> [[TMP14]], ptr [[TMP26]], align 1, !alias.scope [[META8]], !noalias [[META5]]
 ; DEFAULT-NEXT:    [[INDEX_NEXT8]] = add nuw i64 [[INDEX5]], 8
 ; DEFAULT-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT8]], 1000
 ; DEFAULT-NEXT:    br i1 [[TMP17]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
@@ -252,8 +247,7 @@ define void @trunc_store(ptr %dst, ptr %src, i16 %x) #1 {
 ; PRED-NEXT:    [[TMP8:%.*]] = trunc <vscale x 2 x i64> [[BROADCAST_SPLAT3]] to <vscale x 2 x i8>
 ; PRED-NEXT:    [[TMP9:%.*]] = and <vscale x 2 x i8> [[TMP8]], [[TMP11]]
 ; PRED-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INDEX]]
-; PRED-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[TMP5]], i32 0
-; PRED-NEXT:    call void @llvm.masked.store.nxv2i8.p0(<vscale x 2 x i8> [[TMP9]], ptr [[TMP6]], i32 1, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]]), !alias.scope [[META7:![0-9]+]], !noalias [[META4]]
+; PRED-NEXT:    call void @llvm.masked.store.nxv2i8.p0(<vscale x 2 x i8> [[TMP9]], ptr [[TMP5]], i32 1, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]]), !alias.scope [[META7:![0-9]+]], !noalias [[META4]]
 ; PRED-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP4]]
 ; PRED-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[INDEX_NEXT]], i64 1000)
 ; PRED-NEXT:    [[TMP12:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/streaming-compatible-sve-no-maximize-bandwidth.ll b/llvm/test/Transforms/LoopVectorize/AArch64/streaming-compatible-sve-no-maximize-bandwidth.ll
index 045f1c4..1213d97 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/streaming-compatible-sve-no-maximize-bandwidth.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/streaming-compatible-sve-no-maximize-bandwidth.ll
@@ -27,20 +27,17 @@ define i32 @foo(i32 noundef %n, i32 noundef %lag, i32 noundef %shift) vscale_ran
 ; SC_SVE-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP17:%.*]], [[VECTOR_BODY]] ]
 ; SC_SVE-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; SC_SVE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [32 x i16], ptr @a, i64 0, i64 [[INDEX]]
-; SC_SVE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, ptr [[TMP2]], i32 0
-; SC_SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP3]], align 2
+; SC_SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP2]], align 2
 ; SC_SVE-NEXT:    [[TMP4:%.*]] = sext <4 x i16> [[WIDE_LOAD]] to <4 x i32>
 ; SC_SVE-NEXT:    [[TMP5:%.*]] = ashr <4 x i32> [[TMP4]], [[VEC_IND]]
 ; SC_SVE-NEXT:    [[TMP6:%.*]] = add nsw i64 [[INDEX]], [[TMP0]]
 ; SC_SVE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [32 x i16], ptr @b, i64 0, i64 [[TMP6]]
-; SC_SVE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[TMP7]], i32 0
-; SC_SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i16>, ptr [[TMP8]], align 2
+; SC_SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i16>, ptr [[TMP7]], align 2
 ; SC_SVE-NEXT:    [[TMP9:%.*]] = sext <4 x i16> [[WIDE_LOAD1]] to <4 x i32>
 ; SC_SVE-NEXT:    [[TMP10:%.*]] = shl <4 x i32> [[TMP9]], [[VEC_IND]]
 ; SC_SVE-NEXT:    [[TMP11:%.*]] = mul nsw <4 x i32> [[TMP10]], [[TMP5]]
 ; SC_SVE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [32 x i16], ptr @c, i64 0, i64 [[INDEX]]
-; SC_SVE-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i16, ptr [[TMP12]], i32 0
-; SC_SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i16>, ptr [[TMP13]], align 2
+; SC_SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i16>, ptr [[TMP12]], align 2
 ; SC_SVE-NEXT:    [[TMP14:%.*]] = sext <4 x i16> [[WIDE_LOAD2]] to <4 x i32>
 ; SC_SVE-NEXT:    [[TMP15:%.*]] = add nsw <4 x i32> [[TMP11]], [[TMP14]]
 ; SC_SVE-NEXT:    [[TMP16:%.*]] = shl <4 x i32> [[TMP15]], [[BROADCAST_SPLAT]]
@@ -101,20 +98,17 @@ define i32 @foo(i32 noundef %n, i32 noundef %lag, i32 noundef %shift) vscale_ran
 ; NO_SC_SVE-NEXT:    [[VEC_PHI:%.*]] = phi <8 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP17:%.*]], [[VECTOR_BODY]] ]
 ; NO_SC_SVE-NEXT:    [[VEC_IND:%.*]] = phi <8 x i32> [ <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO_SC_SVE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [32 x i16], ptr @a, i64 0, i64 [[INDEX]]
-; NO_SC_SVE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, ptr [[TMP2]], i32 0
-; NO_SC_SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP3]], align 2
+; NO_SC_SVE-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP2]], align 2
 ; NO_SC_SVE-NEXT:    [[TMP4:%.*]] = sext <8 x i16> [[WIDE_LOAD]] to <8 x i32>
 ; NO_SC_SVE-NEXT:    [[TMP5:%.*]] = ashr <8 x i32> [[TMP4]], [[VEC_IND]]
 ; NO_SC_SVE-NEXT:    [[TMP6:%.*]] = add nsw i64 [[INDEX]], [[TMP0]]
 ; NO_SC_SVE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [32 x i16], ptr @b, i64 0, i64 [[TMP6]]
-; NO_SC_SVE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[TMP7]], i32 0
-; NO_SC_SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i16>, ptr [[TMP8]], align 2
+; NO_SC_SVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i16>, ptr [[TMP7]], align 2
 ; NO_SC_SVE-NEXT:    [[TMP9:%.*]] = sext <8 x i16> [[WIDE_LOAD1]] to <8 x i32>
 ; NO_SC_SVE-NEXT:    [[TMP10:%.*]] = shl <8 x i32> [[TMP9]], [[VEC_IND]]
 ; NO_SC_SVE-NEXT:    [[TMP11:%.*]] = mul nsw <8 x i32> [[TMP10]], [[TMP5]]
 ; NO_SC_SVE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [32 x i16], ptr @c, i64 0, i64 [[INDEX]]
-; NO_SC_SVE-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i16, ptr [[TMP12]], i32 0
-; NO_SC_SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i16>, ptr [[TMP13]], align 2
+; NO_SC_SVE-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i16>, ptr [[TMP12]], align 2
 ; NO_SC_SVE-NEXT:    [[TMP14:%.*]] = sext <8 x i16> [[WIDE_LOAD2]] to <8 x i32>
 ; NO_SC_SVE-NEXT:    [[TMP15:%.*]] = add nsw <8 x i32> [[TMP11]], [[TMP14]]
 ; NO_SC_SVE-NEXT:    [[TMP16:%.*]] = shl <8 x i32> [[TMP15]], [[BROADCAST_SPLAT]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/strict-fadd.ll b/llvm/test/Transforms/LoopVectorize/AArch64/strict-fadd.ll
index 8399614..f4eebca 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/strict-fadd.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/strict-fadd.ll
@@ -970,7 +970,7 @@ loop:
   %red.next = fadd double %for, %red
   %for.next = sitofp i32 %iv to double
   %iv.next = add nsw i32 %iv, 1
-  %ec = icmp eq i32 %iv.next, 1024
+  %ec = icmp eq i32 %iv.next, 1025
   br i1 %ec, label %exit, label %loop, !llvm.loop !13
 
 exit:
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect-inloop-reductions.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect-inloop-reductions.ll
index 0fddadd..3d81541 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect-inloop-reductions.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect-inloop-reductions.ll
@@ -27,11 +27,10 @@ define i64 @int_reduction_and(ptr noalias nocapture %a, i64 %N) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi i64 [ 1, [[VECTOR_PH]] ], [ [[TMP17:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI2:%.*]] = phi i64 [ -1, [[VECTOR_PH]] ], [ [[TMP19:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i32 0
 ; CHECK-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP14:%.*]] = mul nuw i64 [[TMP13]], 2
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i64 [[TMP14]]
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP12]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP10]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 2 x i64>, ptr [[TMP15]], align 8
 ; CHECK-NEXT:    [[TMP16:%.*]] = call i64 @llvm.vector.reduce.and.nxv2i64(<vscale x 2 x i64> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[TMP17]] = and i64 [[TMP16]], [[VEC_PHI]]
@@ -58,8 +57,7 @@ define i64 @int_reduction_and(ptr noalias nocapture %a, i64 %N) {
 ; CHECK-NEXT:    [[INDEX7:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT10:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI8:%.*]] = phi i64 [ [[BC_MERGE_RDX]], [[VEC_EPILOG_PH]] ], [ [[TMP27:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX7]]
-; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i64, ptr [[TMP24]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <2 x i64>, ptr [[TMP25]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <2 x i64>, ptr [[TMP24]], align 8
 ; CHECK-NEXT:    [[TMP26:%.*]] = call i64 @llvm.vector.reduce.and.v2i64(<2 x i64> [[WIDE_LOAD9]])
 ; CHECK-NEXT:    [[TMP27]] = and i64 [[TMP26]], [[VEC_PHI8]]
 ; CHECK-NEXT:    [[INDEX_NEXT10]] = add nuw i64 [[INDEX7]], 2
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect-reductions.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect-reductions.ll
index de8fcb0..af9c39e 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect-reductions.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect-reductions.ll
@@ -27,11 +27,10 @@ define i64 @int_reduction_add(ptr %a, i64 %N) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i64> [ insertelement (<vscale x 2 x i64> zeroinitializer, i64 5, i32 0), [[VECTOR_PH]] ], [ [[TMP16:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI2:%.*]] = phi <vscale x 2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP17:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i32 0
 ; CHECK-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP14:%.*]] = mul nuw i64 [[TMP13]], 2
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i64 [[TMP14]]
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP12]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP10]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 2 x i64>, ptr [[TMP15]], align 8
 ; CHECK-NEXT:    [[TMP16]] = add <vscale x 2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP17]] = add <vscale x 2 x i64> [[WIDE_LOAD3]], [[VEC_PHI2]]
@@ -58,8 +57,7 @@ define i64 @int_reduction_add(ptr %a, i64 %N) {
 ; CHECK-NEXT:    [[INDEX7:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT10:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI8:%.*]] = phi <2 x i64> [ [[TMP22]], [[VEC_EPILOG_PH]] ], [ [[TMP26:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX7]]
-; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i64, ptr [[TMP24]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <2 x i64>, ptr [[TMP25]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <2 x i64>, ptr [[TMP24]], align 8
 ; CHECK-NEXT:    [[TMP26]] = add <2 x i64> [[WIDE_LOAD9]], [[VEC_PHI8]]
 ; CHECK-NEXT:    [[INDEX_NEXT10]] = add nuw i64 [[INDEX7]], 2
 ; CHECK-NEXT:    [[TMP27:%.*]] = icmp eq i64 [[INDEX_NEXT10]], [[N_VEC5]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect-strict-reductions.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect-strict-reductions.ll
index 83f2b2a..fca29cd 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect-strict-reductions.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect-strict-reductions.ll
@@ -26,11 +26,10 @@ define float @fadd_strict(ptr noalias nocapture readonly %a, i64 %n) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0xFFFFFFFFE0000000, [[VECTOR_PH]] ], [ [[TMP19:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[TMP12]], i32 0
 ; CHECK-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP16:%.*]] = mul nuw i64 [[TMP15]], 4
 ; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[TMP12]], i64 [[TMP16]]
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP14]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP12]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP17]], align 4
 ; CHECK-NEXT:    [[TMP18:%.*]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[TMP19]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[TMP18]], <vscale x 4 x float> [[WIDE_LOAD2]])
@@ -54,8 +53,7 @@ define float @fadd_strict(ptr noalias nocapture readonly %a, i64 %n) {
 ; CHECK-NEXT:    [[INDEX6:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT9:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI7:%.*]] = phi float [ [[BC_MERGE_RDX]], [[VEC_EPILOG_PH]] ], [ [[TMP24:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX6]]
-; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, ptr [[TMP22]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <2 x float>, ptr [[TMP23]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <2 x float>, ptr [[TMP22]], align 4
 ; CHECK-NEXT:    [[TMP24]] = call float @llvm.vector.reduce.fadd.v2f32(float [[VEC_PHI7]], <2 x float> [[WIDE_LOAD8]])
 ; CHECK-NEXT:    [[INDEX_NEXT9]] = add nuw i64 [[INDEX6]], 2
 ; CHECK-NEXT:    [[TMP25:%.*]] = icmp eq i64 [[INDEX_NEXT9]], [[N_VEC4]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect.ll
index 5209374..0f407cd 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-epilog-vect.ll
@@ -41,11 +41,10 @@ define void @main_vf_vscale_x_16(ptr %A) #0 {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[TMP14]], i32 0
 ; CHECK-NEXT:    [[TMP17:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP18:%.*]] = mul nuw i64 [[TMP17]], 16
 ; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[TMP14]], i64 [[TMP18]]
-; CHECK-NEXT:    store <vscale x 16 x i8> splat (i8 1), ptr [[TMP16]], align 1
+; CHECK-NEXT:    store <vscale x 16 x i8> splat (i8 1), ptr [[TMP14]], align 1
 ; CHECK-NEXT:    store <vscale x 16 x i8> splat (i8 1), ptr [[TMP19]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP7]]
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -71,8 +70,7 @@ define void @main_vf_vscale_x_16(ptr %A) #0 {
 ; CHECK:       vec.epilog.vector.body:
 ; CHECK-NEXT:    [[INDEX5:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT6:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX5]]
-; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i8, ptr [[TMP28]], i32 0
-; CHECK-NEXT:    store <vscale x 8 x i8> splat (i8 1), ptr [[TMP29]], align 1
+; CHECK-NEXT:    store <vscale x 8 x i8> splat (i8 1), ptr [[TMP28]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT6]] = add nuw i64 [[INDEX5]], [[TMP26]]
 ; CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT6]], [[N_VEC3]]
 ; CHECK-NEXT:    br i1 [[TMP30]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
@@ -103,11 +101,10 @@ define void @main_vf_vscale_x_16(ptr %A) #0 {
 ; CHECK-VF8:       vector.body:
 ; CHECK-VF8-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-VF8-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-VF8-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP12]], i32 0
 ; CHECK-VF8-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-VF8-NEXT:    [[TMP16:%.*]] = mul nuw i64 [[TMP15]], 16
 ; CHECK-VF8-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP12]], i64 [[TMP16]]
-; CHECK-VF8-NEXT:    store <vscale x 16 x i8> splat (i8 1), ptr [[TMP14]], align 1
+; CHECK-VF8-NEXT:    store <vscale x 16 x i8> splat (i8 1), ptr [[TMP12]], align 1
 ; CHECK-VF8-NEXT:    store <vscale x 16 x i8> splat (i8 1), ptr [[TMP17]], align 1
 ; CHECK-VF8-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-VF8-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -125,8 +122,7 @@ define void @main_vf_vscale_x_16(ptr %A) #0 {
 ; CHECK-VF8:       vec.epilog.vector.body:
 ; CHECK-VF8-NEXT:    [[INDEX1:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-VF8-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX1]]
-; CHECK-VF8-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i8, ptr [[TMP20]], i32 0
-; CHECK-VF8-NEXT:    store <8 x i8> splat (i8 1), ptr [[TMP21]], align 1
+; CHECK-VF8-NEXT:    store <8 x i8> splat (i8 1), ptr [[TMP20]], align 1
 ; CHECK-VF8-NEXT:    [[INDEX_NEXT2]] = add nuw i64 [[INDEX1]], 8
 ; CHECK-VF8-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT2]], 1024
 ; CHECK-VF8-NEXT:    br i1 [[TMP22]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
@@ -152,6 +148,104 @@ exit:
   ret void
 }
 
+define void @main_vf_vscale_x_2_no_epi_iteration(ptr %A) #0 vscale_range(8, 8) {
+; CHECK-LABEL: @main_vf_vscale_x_2_no_epi_iteration(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 4
+; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]]
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]]
+; CHECK-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
+; CHECK-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 2
+; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i64 [[TMP8]]
+; CHECK-NEXT:    store <vscale x 2 x i64> splat (i64 1), ptr [[TMP6]], align 1
+; CHECK-NEXT:    store <vscale x 2 x i64> splat (i64 1), ptr [[TMP9]], align 1
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+;
+; CHECK-VF8-LABEL: @main_vf_vscale_x_2_no_epi_iteration(
+; CHECK-VF8-NEXT:  iter.check:
+; CHECK-VF8-NEXT:    br i1 false, label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
+; CHECK-VF8:       vector.main.loop.iter.check:
+; CHECK-VF8-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-VF8-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 4
+; CHECK-VF8-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP1]]
+; CHECK-VF8-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK-VF8:       vector.ph:
+; CHECK-VF8-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-VF8-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4
+; CHECK-VF8-NEXT:    [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]]
+; CHECK-VF8-NEXT:    [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]]
+; CHECK-VF8-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-VF8-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4
+; CHECK-VF8-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK-VF8:       vector.body:
+; CHECK-VF8-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-VF8-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
+; CHECK-VF8-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-VF8-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 2
+; CHECK-VF8-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i64 [[TMP8]]
+; CHECK-VF8-NEXT:    store <vscale x 2 x i64> splat (i64 1), ptr [[TMP6]], align 1
+; CHECK-VF8-NEXT:    store <vscale x 2 x i64> splat (i64 1), ptr [[TMP9]], align 1
+; CHECK-VF8-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; CHECK-VF8-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-VF8-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-VF8:       middle.block:
+; CHECK-VF8-NEXT:    [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]]
+; CHECK-VF8-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
+; CHECK-VF8:       vec.epilog.iter.check:
+; CHECK-VF8-NEXT:    [[N_VEC_REMAINING:%.*]] = sub i64 1024, [[N_VEC]]
+; CHECK-VF8-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 8
+; CHECK-VF8-NEXT:    br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
+; CHECK-VF8:       vec.epilog.ph:
+; CHECK-VF8-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
+; CHECK-VF8-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
+; CHECK-VF8:       vec.epilog.vector.body:
+; CHECK-VF8-NEXT:    [[INDEX1:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
+; CHECK-VF8-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX1]]
+; CHECK-VF8-NEXT:    store <8 x i64> splat (i64 1), ptr [[TMP11]], align 1
+; CHECK-VF8-NEXT:    [[INDEX_NEXT2]] = add nuw i64 [[INDEX1]], 8
+; CHECK-VF8-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT2]], 1024
+; CHECK-VF8-NEXT:    br i1 [[TMP14]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK-VF8:       vec.epilog.middle.block:
+; CHECK-VF8-NEXT:    br i1 true, label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
+; CHECK-VF8:       vec.epilog.scalar.ph:
+; CHECK-VF8-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ]
+; CHECK-VF8-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK-VF8:       for.body:
+;
+entry:
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %arrayidx = getelementptr inbounds i64, ptr %A, i64 %iv
+  store i64 1, ptr %arrayidx, align 1
+  %iv.next = add nuw nsw i64 %iv, 1
+  %exitcond = icmp ne i64 %iv.next, 1024
+  br i1 %exitcond, label %for.body, label %exit
+
+exit:
+  ret void
+}
 
 ; DEBUG: LV: Checking a loop in 'main_vf_vscale_x_2'
 ; DEBUG: Create Skeleton for epilogue vectorized loop (first pass)
@@ -167,110 +261,114 @@ exit:
 ; fixed-width VF=8 for the epilogue if the vectors are known to be
 ; sufficiently wide. This information can be deduced from vscale_range or
 ; VScaleForTuning (set by mcpu/mtune).
-define void @main_vf_vscale_x_2(ptr %A) #0 vscale_range(8, 8) {
+define void @main_vf_vscale_x_2(ptr %A, i64 %n) #0 vscale_range(8, 8) {
 ; CHECK-LABEL: @main_vf_vscale_x_2(
 ; CHECK-NEXT:  iter.check:
-; CHECK-NEXT:    br i1 false, label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], 8
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
 ; CHECK:       vector.main.loop.iter.check:
 ; CHECK-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 4
-; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP1]]
-; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK-NEXT:    [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[N]], [[TMP1]]
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4
-; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]]
-; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]]
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i64, ptr [[TMP12]], i32 0
 ; CHECK-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP16:%.*]] = mul nuw i64 [[TMP15]], 2
 ; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i64, ptr [[TMP12]], i64 [[TMP16]]
-; CHECK-NEXT:    store <vscale x 2 x i64> splat (i64 1), ptr [[TMP14]], align 1
+; CHECK-NEXT:    store <vscale x 2 x i64> splat (i64 1), ptr [[TMP12]], align 1
 ; CHECK-NEXT:    store <vscale x 2 x i64> splat (i64 1), ptr [[TMP17]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]]
-; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH:%.*]]
 ; CHECK:       vec.epilog.iter.check:
-; CHECK-NEXT:    [[N_VEC_REMAINING:%.*]] = sub i64 1024, [[N_VEC]]
+; CHECK-NEXT:    [[N_VEC_REMAINING:%.*]] = sub i64 [[N]], [[N_VEC]]
 ; CHECK-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 8
 ; CHECK-NEXT:    br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
 ; CHECK:       vec.epilog.ph:
-; CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
-; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
+; CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[SCALAR_PH]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
+; CHECK-NEXT:    [[N_MOD_VF2:%.*]] = urem i64 [[N]], 8
+; CHECK-NEXT:    [[N_VEC3:%.*]] = sub i64 [[N]], [[N_MOD_VF2]]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       vec.epilog.vector.body:
-; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i64, ptr [[TMP20]], i32 0
-; CHECK-NEXT:    store <8 x i64> splat (i64 1), ptr [[TMP21]], align 1
-; CHECK-NEXT:    [[INDEX_NEXT2]] = add nuw i64 [[INDEX1]], 8
-; CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT2]], 1024
-; CHECK-NEXT:    br i1 [[TMP22]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK-NEXT:    [[INDEX4:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT5:%.*]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX4]]
+; CHECK-NEXT:    store <8 x i64> splat (i64 1), ptr [[TMP19]], align 1
+; CHECK-NEXT:    [[INDEX_NEXT5]] = add nuw i64 [[INDEX4]], 8
+; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT5]], [[N_VEC3]]
+; CHECK-NEXT:    br i1 [[TMP20]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       vec.epilog.middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
+; CHECK-NEXT:    [[CMP_N6:%.*]] = icmp eq i64 [[N]], [[N_VEC3]]
+; CHECK-NEXT:    br i1 [[CMP_N6]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
 ; CHECK:       vec.epilog.scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ]
-; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC3]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[SCALAR_PH]] ], [ 0, [[ITER_CHECK:%.*]] ]
+; CHECK-NEXT:    br label [[FOR_BODY1:%.*]]
 ; CHECK:       for.body:
 ;
 ; CHECK-VF8-LABEL: @main_vf_vscale_x_2(
 ; CHECK-VF8-NEXT:  iter.check:
-; CHECK-VF8-NEXT:    br i1 false, label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
+; CHECK-VF8-NEXT:    [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[N:%.*]], 8
+; CHECK-VF8-NEXT:    br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
 ; CHECK-VF8:       vector.main.loop.iter.check:
 ; CHECK-VF8-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-VF8-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 4
-; CHECK-VF8-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP1]]
+; CHECK-VF8-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP1]]
 ; CHECK-VF8-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK-VF8:       vector.ph:
 ; CHECK-VF8-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-VF8-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4
-; CHECK-VF8-NEXT:    [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]]
-; CHECK-VF8-NEXT:    [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]]
+; CHECK-VF8-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
+; CHECK-VF8-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; CHECK-VF8-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-VF8-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4
 ; CHECK-VF8-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-VF8:       vector.body:
 ; CHECK-VF8-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-VF8-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-VF8-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i64, ptr [[TMP12]], i32 0
 ; CHECK-VF8-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-VF8-NEXT:    [[TMP16:%.*]] = mul nuw i64 [[TMP15]], 2
 ; CHECK-VF8-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i64, ptr [[TMP12]], i64 [[TMP16]]
-; CHECK-VF8-NEXT:    store <vscale x 2 x i64> splat (i64 1), ptr [[TMP14]], align 1
+; CHECK-VF8-NEXT:    store <vscale x 2 x i64> splat (i64 1), ptr [[TMP12]], align 1
 ; CHECK-VF8-NEXT:    store <vscale x 2 x i64> splat (i64 1), ptr [[TMP17]], align 1
 ; CHECK-VF8-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-VF8-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-VF8-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-VF8-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK-VF8:       middle.block:
-; CHECK-VF8-NEXT:    [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]]
+; CHECK-VF8-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-VF8-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
 ; CHECK-VF8:       vec.epilog.iter.check:
-; CHECK-VF8-NEXT:    [[N_VEC_REMAINING:%.*]] = sub i64 1024, [[N_VEC]]
+; CHECK-VF8-NEXT:    [[N_VEC_REMAINING:%.*]] = sub i64 [[N]], [[N_VEC]]
 ; CHECK-VF8-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], 8
 ; CHECK-VF8-NEXT:    br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
 ; CHECK-VF8:       vec.epilog.ph:
 ; CHECK-VF8-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
+; CHECK-VF8-NEXT:    [[N_MOD_VF2:%.*]] = urem i64 [[N]], 8
+; CHECK-VF8-NEXT:    [[N_VEC3:%.*]] = sub i64 [[N]], [[N_MOD_VF2]]
 ; CHECK-VF8-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
 ; CHECK-VF8:       vec.epilog.vector.body:
 ; CHECK-VF8-NEXT:    [[INDEX1:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-VF8-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX1]]
-; CHECK-VF8-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i64, ptr [[TMP20]], i32 0
-; CHECK-VF8-NEXT:    store <8 x i64> splat (i64 1), ptr [[TMP21]], align 1
+; CHECK-VF8-NEXT:    store <8 x i64> splat (i64 1), ptr [[TMP20]], align 1
 ; CHECK-VF8-NEXT:    [[INDEX_NEXT2]] = add nuw i64 [[INDEX1]], 8
-; CHECK-VF8-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT2]], 1024
-; CHECK-VF8-NEXT:    br i1 [[TMP22]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK-VF8-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT2]], [[N_VEC3]]
+; CHECK-VF8-NEXT:    br i1 [[TMP19]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
 ; CHECK-VF8:       vec.epilog.middle.block:
-; CHECK-VF8-NEXT:    br i1 true, label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
+; CHECK-VF8-NEXT:    [[CMP_N6:%.*]] = icmp eq i64 [[N]], [[N_VEC3]]
+; CHECK-VF8-NEXT:    br i1 [[CMP_N6]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
 ; CHECK-VF8:       vec.epilog.scalar.ph:
-; CHECK-VF8-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ]
+; CHECK-VF8-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC3]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ]
 ; CHECK-VF8-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK-VF8:       for.body:
 ;
@@ -282,7 +380,7 @@ for.body:
   %arrayidx = getelementptr inbounds i64, ptr %A, i64 %iv
   store i64 1, ptr %arrayidx, align 1
   %iv.next = add nuw nsw i64 %iv, 1
-  %exitcond = icmp ne i64 %iv.next, 1024
+  %exitcond = icmp ne i64 %iv.next, %n
   br i1 %exitcond, label %for.body, label %exit
 
 exit:
@@ -314,15 +412,14 @@ define void @test_pr57912_pointer_induction(ptr %start) #0 {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[START:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr i8, ptr [[TMP14]], i32 0
 ; CHECK-NEXT:    [[TMP17:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP18:%.*]] = mul nuw i64 [[TMP17]], 16
 ; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr i8, ptr [[TMP14]], i64 [[TMP18]]
-; CHECK-NEXT:    store <vscale x 16 x i8> zeroinitializer, ptr [[TMP16]], align 1
+; CHECK-NEXT:    store <vscale x 16 x i8> zeroinitializer, ptr [[TMP14]], align 1
 ; CHECK-NEXT:    store <vscale x 16 x i8> zeroinitializer, ptr [[TMP19]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP7]]
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 10000, [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
@@ -346,11 +443,10 @@ define void @test_pr57912_pointer_induction(ptr %start) #0 {
 ; CHECK:       vec.epilog.vector.body:
 ; CHECK-NEXT:    [[INDEX7:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT8:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr i8, ptr [[START]], i64 [[INDEX7]]
-; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr i8, ptr [[TMP28]], i32 0
-; CHECK-NEXT:    store <vscale x 8 x i8> zeroinitializer, ptr [[TMP29]], align 1
+; CHECK-NEXT:    store <vscale x 8 x i8> zeroinitializer, ptr [[TMP28]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT8]] = add nuw i64 [[INDEX7]], [[TMP26]]
 ; CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT8]], [[N_VEC3]]
-; CHECK-NEXT:    br i1 [[TMP30]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; CHECK-NEXT:    br i1 [[TMP30]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ; CHECK:       vec.epilog.middle.block:
 ; CHECK-NEXT:    [[CMP_N6:%.*]] = icmp eq i64 10000, [[N_VEC3]]
 ; CHECK-NEXT:    br i1 [[CMP_N6]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
@@ -379,15 +475,14 @@ define void @test_pr57912_pointer_induction(ptr %start) #0 {
 ; CHECK-VF8:       vector.body:
 ; CHECK-VF8-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-VF8-NEXT:    [[TMP12:%.*]] = getelementptr i8, ptr [[START:%.*]], i64 [[INDEX]]
-; CHECK-VF8-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[TMP12]], i32 0
 ; CHECK-VF8-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-VF8-NEXT:    [[TMP16:%.*]] = mul nuw i64 [[TMP15]], 16
 ; CHECK-VF8-NEXT:    [[TMP17:%.*]] = getelementptr i8, ptr [[TMP12]], i64 [[TMP16]]
-; CHECK-VF8-NEXT:    store <vscale x 16 x i8> zeroinitializer, ptr [[TMP14]], align 1
+; CHECK-VF8-NEXT:    store <vscale x 16 x i8> zeroinitializer, ptr [[TMP12]], align 1
 ; CHECK-VF8-NEXT:    store <vscale x 16 x i8> zeroinitializer, ptr [[TMP17]], align 1
 ; CHECK-VF8-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-VF8-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-VF8-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK-VF8-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ; CHECK-VF8:       middle.block:
 ; CHECK-VF8-NEXT:    [[CMP_N:%.*]] = icmp eq i64 10000, [[N_VEC]]
 ; CHECK-VF8-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
@@ -403,11 +498,10 @@ define void @test_pr57912_pointer_induction(ptr %start) #0 {
 ; CHECK-VF8:       vec.epilog.vector.body:
 ; CHECK-VF8-NEXT:    [[INDEX3:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT4:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-VF8-NEXT:    [[TMP20:%.*]] = getelementptr i8, ptr [[START]], i64 [[INDEX3]]
-; CHECK-VF8-NEXT:    [[TMP21:%.*]] = getelementptr i8, ptr [[TMP20]], i32 0
-; CHECK-VF8-NEXT:    store <8 x i8> zeroinitializer, ptr [[TMP21]], align 1
+; CHECK-VF8-NEXT:    store <8 x i8> zeroinitializer, ptr [[TMP20]], align 1
 ; CHECK-VF8-NEXT:    [[INDEX_NEXT4]] = add nuw i64 [[INDEX3]], 8
 ; CHECK-VF8-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT4]], 10000
-; CHECK-VF8-NEXT:    br i1 [[TMP22]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; CHECK-VF8-NEXT:    br i1 [[TMP22]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK-VF8:       vec.epilog.middle.block:
 ; CHECK-VF8-NEXT:    br i1 true, label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
 ; CHECK-VF8:       vec.epilog.scalar.ph:
@@ -432,4 +526,307 @@ exit:
   ret void
 }
 
+; Loop with vscale-based trip count vscale x 1033.
+define void @trip_count_vscale(ptr noalias %a, ptr noalias %b) vscale_range(1, 16) #0 {
+; CHECK-LABEL: @trip_count_vscale(
+; CHECK-NEXT:  iter.check:
+; CHECK-NEXT:    [[V:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[N:%.*]] = mul nuw nsw i64 [[V]], 1033
+; CHECK-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 2
+; CHECK-NEXT:    br i1 false, label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
+; CHECK:       vector.main.loop.iter.check:
+; CHECK-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 8
+; CHECK-NEXT:    [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[N]], [[TMP3]]
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 8
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP5]]
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[TMP6:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP7:%.*]] = mul nuw i64 [[TMP6]], 8
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw float, ptr [[A:%.*]], i64 [[INDEX]]
+; CHECK-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP11:%.*]] = mul nuw i64 [[TMP10]], 4
+; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw float, ptr [[TMP8]], i64 [[TMP11]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP12]], align 4
+; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw float, ptr [[B:%.*]], i64 [[INDEX]]
+; CHECK-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP16:%.*]] = mul nuw i64 [[TMP15]], 4
+; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw float, ptr [[TMP13]], i64 [[TMP16]]
+; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x float>, ptr [[TMP13]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 4 x float>, ptr [[TMP17]], align 4
+; CHECK-NEXT:    [[TMP18:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD3]]
+; CHECK-NEXT:    [[TMP19:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD2]], [[WIDE_LOAD4]]
+; CHECK-NEXT:    [[TMP20:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP21:%.*]] = mul nuw i64 [[TMP20]], 4
+; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw float, ptr [[TMP13]], i64 [[TMP21]]
+; CHECK-NEXT:    store <vscale x 4 x float> [[TMP18]], ptr [[TMP13]], align 4
+; CHECK-NEXT:    store <vscale x 4 x float> [[TMP19]], ptr [[TMP22]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP7]]
+; CHECK-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP23]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
+; CHECK:       vec.epilog.iter.check:
+; CHECK-NEXT:    [[N_VEC_REMAINING:%.*]] = sub i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    [[TMP24:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP25:%.*]] = mul nuw i64 [[TMP24]], 2
+; CHECK-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], [[TMP25]]
+; CHECK-NEXT:    br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
+; CHECK:       vec.epilog.ph:
+; CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
+; CHECK-NEXT:    [[TMP26:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP27:%.*]] = mul nuw i64 [[TMP26]], 2
+; CHECK-NEXT:    [[N_MOD_VF5:%.*]] = urem i64 [[N]], [[TMP27]]
+; CHECK-NEXT:    [[N_VEC6:%.*]] = sub i64 [[N]], [[N_MOD_VF5]]
+; CHECK-NEXT:    [[TMP28:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP29:%.*]] = mul nuw i64 [[TMP28]], 2
+; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
+; CHECK:       vec.epilog.vector.body:
+; CHECK-NEXT:    [[INDEX7:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT10:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw float, ptr [[A]], i64 [[INDEX7]]
+; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <vscale x 2 x float>, ptr [[TMP30]], align 4
+; CHECK-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw float, ptr [[B]], i64 [[INDEX7]]
+; CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <vscale x 2 x float>, ptr [[TMP32]], align 4
+; CHECK-NEXT:    [[TMP34:%.*]] = fmul <vscale x 2 x float> [[WIDE_LOAD8]], [[WIDE_LOAD9]]
+; CHECK-NEXT:    store <vscale x 2 x float> [[TMP34]], ptr [[TMP32]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT10]] = add nuw i64 [[INDEX7]], [[TMP29]]
+; CHECK-NEXT:    [[TMP35:%.*]] = icmp eq i64 [[INDEX_NEXT10]], [[N_VEC6]]
+; CHECK-NEXT:    br i1 [[TMP35]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
+; CHECK:       vec.epilog.middle.block:
+; CHECK-NEXT:    [[CMP_N11:%.*]] = icmp eq i64 [[N]], [[N_VEC6]]
+; CHECK-NEXT:    br i1 [[CMP_N11]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
+; CHECK:       vec.epilog.scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC6]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+;
+; CHECK-VF8-LABEL: @trip_count_vscale(
+; CHECK-VF8-NEXT:  entry:
+; CHECK-VF8-NEXT:    [[V:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-VF8-NEXT:    [[N:%.*]] = mul nuw nsw i64 [[V]], 1033
+; CHECK-VF8-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-VF8-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 8
+; CHECK-VF8-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK-VF8:       vector.ph:
+; CHECK-VF8-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-VF8-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 8
+; CHECK-VF8-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
+; CHECK-VF8-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; CHECK-VF8-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-VF8-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 8
+; CHECK-VF8-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK-VF8:       vector.body:
+; CHECK-VF8-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-VF8-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw float, ptr [[A:%.*]], i64 [[INDEX]]
+; CHECK-VF8-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-VF8-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 4
+; CHECK-VF8-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw float, ptr [[TMP6]], i64 [[TMP9]]
+; CHECK-VF8-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP6]], align 4
+; CHECK-VF8-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP10]], align 4
+; CHECK-VF8-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw float, ptr [[B:%.*]], i64 [[INDEX]]
+; CHECK-VF8-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-VF8-NEXT:    [[TMP14:%.*]] = mul nuw i64 [[TMP13]], 4
+; CHECK-VF8-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw float, ptr [[TMP11]], i64 [[TMP14]]
+; CHECK-VF8-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4
+; CHECK-VF8-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x float>, ptr [[TMP15]], align 4
+; CHECK-VF8-NEXT:    [[TMP16:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD2]]
+; CHECK-VF8-NEXT:    [[TMP17:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD1]], [[WIDE_LOAD3]]
+; CHECK-VF8-NEXT:    [[TMP18:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-VF8-NEXT:    [[TMP19:%.*]] = mul nuw i64 [[TMP18]], 4
+; CHECK-VF8-NEXT:    [[TMP20:%.*]] = getelementptr inbounds nuw float, ptr [[TMP11]], i64 [[TMP19]]
+; CHECK-VF8-NEXT:    store <vscale x 4 x float> [[TMP16]], ptr [[TMP11]], align 4
+; CHECK-VF8-NEXT:    store <vscale x 4 x float> [[TMP17]], ptr [[TMP20]], align 4
+; CHECK-VF8-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; CHECK-VF8-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-VF8-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
+; CHECK-VF8:       middle.block:
+; CHECK-VF8-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-VF8-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-VF8:       scalar.ph:
+; CHECK-VF8-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-VF8-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK-VF8:       for.body:
+;
+entry:
+  %v = tail call i64 @llvm.vscale.i64()
+  %n = mul nuw nsw i64 %v, 1033
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %arrayidx = getelementptr inbounds nuw float, ptr %a, i64 %iv
+  %l.a = load float, ptr %arrayidx, align 4
+  %arrayidx3 = getelementptr inbounds nuw float, ptr %b, i64 %iv
+  %l.b = load float, ptr %arrayidx3, align 4
+  %mul4 = fmul float %l.a, %l.b
+  store float %mul4, ptr %arrayidx3, align 4
+  %iv.next = add nuw nsw i64 %iv, 1
+  %ec = icmp eq i64 %iv.next, %n
+  br i1 %ec, label %exit, label %for.body
+
+exit:
+  ret void
+}
+
+; Loop with vscale-based trip count vscale x 1024.
+; TODO: No epilogue vectorizations should remain when choosing VF = vscale x 4.
+define void @trip_count_vscale_no_epilogue_iterations(ptr noalias %a, ptr noalias %b) vscale_range(1, 16) #0 {
+; CHECK-LABEL: @trip_count_vscale_no_epilogue_iterations(
+; CHECK-NEXT:  iter.check:
+; CHECK-NEXT:    [[V:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[N:%.*]] = mul nuw nsw i64 [[V]], 1024
+; CHECK-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 2
+; CHECK-NEXT:    br i1 false, label [[VEC_EPILOG_SCALAR_PH:%.*]], label [[VECTOR_MAIN_LOOP_ITER_CHECK:%.*]]
+; CHECK:       vector.main.loop.iter.check:
+; CHECK-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 8
+; CHECK-NEXT:    [[MIN_ITERS_CHECK1:%.*]] = icmp ult i64 [[N]], [[TMP3]]
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK1]], label [[VEC_EPILOG_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 8
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP5]]
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[TMP6:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP7:%.*]] = mul nuw i64 [[TMP6]], 8
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw float, ptr [[A:%.*]], i64 [[INDEX]]
+; CHECK-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP11:%.*]] = mul nuw i64 [[TMP10]], 4
+; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw float, ptr [[TMP8]], i64 [[TMP11]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP12]], align 4
+; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw float, ptr [[B:%.*]], i64 [[INDEX]]
+; CHECK-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP16:%.*]] = mul nuw i64 [[TMP15]], 4
+; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw float, ptr [[TMP13]], i64 [[TMP16]]
+; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x float>, ptr [[TMP13]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 4 x float>, ptr [[TMP17]], align 4
+; CHECK-NEXT:    [[TMP18:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD3]]
+; CHECK-NEXT:    [[TMP19:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD2]], [[WIDE_LOAD4]]
+; CHECK-NEXT:    [[TMP20:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP21:%.*]] = mul nuw i64 [[TMP20]], 4
+; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw float, ptr [[TMP13]], i64 [[TMP21]]
+; CHECK-NEXT:    store <vscale x 4 x float> [[TMP18]], ptr [[TMP13]], align 4
+; CHECK-NEXT:    store <vscale x 4 x float> [[TMP19]], ptr [[TMP22]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP7]]
+; CHECK-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP23]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
+; CHECK:       vec.epilog.iter.check:
+; CHECK-NEXT:    [[N_VEC_REMAINING:%.*]] = sub i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    [[TMP24:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP25:%.*]] = mul nuw i64 [[TMP24]], 2
+; CHECK-NEXT:    [[MIN_EPILOG_ITERS_CHECK:%.*]] = icmp ult i64 [[N_VEC_REMAINING]], [[TMP25]]
+; CHECK-NEXT:    br i1 [[MIN_EPILOG_ITERS_CHECK]], label [[VEC_EPILOG_SCALAR_PH]], label [[VEC_EPILOG_PH]]
+; CHECK:       vec.epilog.ph:
+; CHECK-NEXT:    [[VEC_EPILOG_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[VECTOR_MAIN_LOOP_ITER_CHECK]] ]
+; CHECK-NEXT:    [[TMP26:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP27:%.*]] = mul nuw i64 [[TMP26]], 2
+; CHECK-NEXT:    [[N_MOD_VF5:%.*]] = urem i64 [[N]], [[TMP27]]
+; CHECK-NEXT:    [[N_VEC6:%.*]] = sub i64 [[N]], [[N_MOD_VF5]]
+; CHECK-NEXT:    [[TMP28:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP29:%.*]] = mul nuw i64 [[TMP28]], 2
+; CHECK-NEXT:    br label [[VEC_EPILOG_VECTOR_BODY:%.*]]
+; CHECK:       vec.epilog.vector.body:
+; CHECK-NEXT:    [[INDEX7:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT10:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr inbounds nuw float, ptr [[A]], i64 [[INDEX7]]
+; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <vscale x 2 x float>, ptr [[TMP30]], align 4
+; CHECK-NEXT:    [[TMP32:%.*]] = getelementptr inbounds nuw float, ptr [[B]], i64 [[INDEX7]]
+; CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <vscale x 2 x float>, ptr [[TMP32]], align 4
+; CHECK-NEXT:    [[TMP34:%.*]] = fmul <vscale x 2 x float> [[WIDE_LOAD8]], [[WIDE_LOAD9]]
+; CHECK-NEXT:    store <vscale x 2 x float> [[TMP34]], ptr [[TMP32]], align 4
+; CHECK-NEXT:    [[INDEX_NEXT10]] = add nuw i64 [[INDEX7]], [[TMP29]]
+; CHECK-NEXT:    [[TMP35:%.*]] = icmp eq i64 [[INDEX_NEXT10]], [[N_VEC6]]
+; CHECK-NEXT:    br i1 [[TMP35]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
+; CHECK:       vec.epilog.middle.block:
+; CHECK-NEXT:    [[CMP_N11:%.*]] = icmp eq i64 [[N]], [[N_VEC6]]
+; CHECK-NEXT:    br i1 [[CMP_N11]], label [[EXIT]], label [[VEC_EPILOG_SCALAR_PH]]
+; CHECK:       vec.epilog.scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC6]], [[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], [[VEC_EPILOG_ITER_CHECK]] ], [ 0, [[ITER_CHECK:%.*]] ]
+; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK:       for.body:
+;
+; CHECK-VF8-LABEL: @trip_count_vscale_no_epilogue_iterations(
+; CHECK-VF8-NEXT:  entry:
+; CHECK-VF8-NEXT:    [[V:%.*]] = tail call i64 @llvm.vscale.i64()
+; CHECK-VF8-NEXT:    [[N:%.*]] = mul nuw nsw i64 [[V]], 1024
+; CHECK-VF8-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-VF8-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 8
+; CHECK-VF8-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK-VF8:       vector.ph:
+; CHECK-VF8-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-VF8-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 8
+; CHECK-VF8-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
+; CHECK-VF8-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; CHECK-VF8-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-VF8-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 8
+; CHECK-VF8-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK-VF8:       vector.body:
+; CHECK-VF8-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-VF8-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw float, ptr [[A:%.*]], i64 [[INDEX]]
+; CHECK-VF8-NEXT:    [[TMP8:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-VF8-NEXT:    [[TMP9:%.*]] = mul nuw i64 [[TMP8]], 4
+; CHECK-VF8-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw float, ptr [[TMP6]], i64 [[TMP9]]
+; CHECK-VF8-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP6]], align 4
+; CHECK-VF8-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP10]], align 4
+; CHECK-VF8-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw float, ptr [[B:%.*]], i64 [[INDEX]]
+; CHECK-VF8-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-VF8-NEXT:    [[TMP14:%.*]] = mul nuw i64 [[TMP13]], 4
+; CHECK-VF8-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw float, ptr [[TMP11]], i64 [[TMP14]]
+; CHECK-VF8-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4
+; CHECK-VF8-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x float>, ptr [[TMP15]], align 4
+; CHECK-VF8-NEXT:    [[TMP16:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD2]]
+; CHECK-VF8-NEXT:    [[TMP17:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD1]], [[WIDE_LOAD3]]
+; CHECK-VF8-NEXT:    [[TMP18:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-VF8-NEXT:    [[TMP19:%.*]] = mul nuw i64 [[TMP18]], 4
+; CHECK-VF8-NEXT:    [[TMP20:%.*]] = getelementptr inbounds nuw float, ptr [[TMP11]], i64 [[TMP19]]
+; CHECK-VF8-NEXT:    store <vscale x 4 x float> [[TMP16]], ptr [[TMP11]], align 4
+; CHECK-VF8-NEXT:    store <vscale x 4 x float> [[TMP17]], ptr [[TMP20]], align 4
+; CHECK-VF8-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; CHECK-VF8-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-VF8-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; CHECK-VF8:       middle.block:
+; CHECK-VF8-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-VF8-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-VF8:       scalar.ph:
+; CHECK-VF8-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-VF8-NEXT:    br label [[FOR_BODY:%.*]]
+; CHECK-VF8:       for.body:
+;
+entry:
+  %v = tail call i64 @llvm.vscale.i64()
+  %n = mul nuw nsw i64 %v, 1024
+  br label %for.body
+
+for.body:
+  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
+  %arrayidx = getelementptr inbounds nuw float, ptr %a, i64 %iv
+  %l.a = load float, ptr %arrayidx, align 4
+  %arrayidx3 = getelementptr inbounds nuw float, ptr %b, i64 %iv
+  %l.b = load float, ptr %arrayidx3, align 4
+  %mul4 = fmul float %l.a, %l.b
+  store float %mul4, ptr %arrayidx3, align 4
+  %iv.next = add nuw nsw i64 %iv, 1
+  %ec = icmp eq i64 %iv.next, %n
+  br i1 %ec, label %exit, label %for.body
+
+exit:
+  ret void
+}
+
 attributes #0 = { "target-features"="+sve" }
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-fixed-width-inorder-core.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-fixed-width-inorder-core.ll
index 51e2492..20bc0af 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-fixed-width-inorder-core.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-fixed-width-inorder-core.ll
@@ -28,21 +28,18 @@ define void @sve_add(ptr  %dst, ptr  %a, ptr  %b, i64 %n) {
 ; CHECK-CA510:       [[VECTOR_BODY]]:
 ; CHECK-CA510-NEXT:    [[TMP2:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-CA510-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw float, ptr [[A]], i64 [[TMP2]]
-; CHECK-CA510-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw float, ptr [[TMP3]], i32 0
 ; CHECK-CA510-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw float, ptr [[TMP3]], i32 4
-; CHECK-CA510-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
+; CHECK-CA510-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
 ; CHECK-CA510-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
 ; CHECK-CA510-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw float, ptr [[B]], i64 [[TMP2]]
-; CHECK-CA510-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw float, ptr [[TMP6]], i32 0
 ; CHECK-CA510-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw float, ptr [[TMP6]], i32 4
-; CHECK-CA510-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP7]], align 4
+; CHECK-CA510-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP6]], align 4
 ; CHECK-CA510-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x float>, ptr [[TMP8]], align 4
 ; CHECK-CA510-NEXT:    [[TMP9:%.*]] = fadd fast <4 x float> [[WIDE_LOAD6]], [[WIDE_LOAD]]
 ; CHECK-CA510-NEXT:    [[TMP10:%.*]] = fadd fast <4 x float> [[WIDE_LOAD7]], [[WIDE_LOAD5]]
 ; CHECK-CA510-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw float, ptr [[DST]], i64 [[TMP2]]
-; CHECK-CA510-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw float, ptr [[TMP11]], i32 0
 ; CHECK-CA510-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw float, ptr [[TMP11]], i32 4
-; CHECK-CA510-NEXT:    store <4 x float> [[TMP9]], ptr [[TMP12]], align 4
+; CHECK-CA510-NEXT:    store <4 x float> [[TMP9]], ptr [[TMP11]], align 4
 ; CHECK-CA510-NEXT:    store <4 x float> [[TMP10]], ptr [[TMP13]], align 4
 ; CHECK-CA510-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP2]], 8
 ; CHECK-CA510-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -95,21 +92,18 @@ define void @sve_add(ptr  %dst, ptr  %a, ptr  %b, i64 %n) {
 ; CHECK-CA520:       [[VECTOR_BODY]]:
 ; CHECK-CA520-NEXT:    [[TMP2:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-CA520-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw float, ptr [[A]], i64 [[TMP2]]
-; CHECK-CA520-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw float, ptr [[TMP3]], i32 0
 ; CHECK-CA520-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw float, ptr [[TMP3]], i32 4
-; CHECK-CA520-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
+; CHECK-CA520-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
 ; CHECK-CA520-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
 ; CHECK-CA520-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw float, ptr [[B]], i64 [[TMP2]]
-; CHECK-CA520-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw float, ptr [[TMP6]], i32 0
 ; CHECK-CA520-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw float, ptr [[TMP6]], i32 4
-; CHECK-CA520-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP7]], align 4
+; CHECK-CA520-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP6]], align 4
 ; CHECK-CA520-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x float>, ptr [[TMP8]], align 4
 ; CHECK-CA520-NEXT:    [[TMP9:%.*]] = fadd fast <4 x float> [[WIDE_LOAD6]], [[WIDE_LOAD]]
 ; CHECK-CA520-NEXT:    [[TMP10:%.*]] = fadd fast <4 x float> [[WIDE_LOAD7]], [[WIDE_LOAD5]]
 ; CHECK-CA520-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw float, ptr [[DST]], i64 [[TMP2]]
-; CHECK-CA520-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw float, ptr [[TMP11]], i32 0
 ; CHECK-CA520-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw float, ptr [[TMP11]], i32 4
-; CHECK-CA520-NEXT:    store <4 x float> [[TMP9]], ptr [[TMP12]], align 4
+; CHECK-CA520-NEXT:    store <4 x float> [[TMP9]], ptr [[TMP11]], align 4
 ; CHECK-CA520-NEXT:    store <4 x float> [[TMP10]], ptr [[TMP13]], align 4
 ; CHECK-CA520-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP2]], 8
 ; CHECK-CA520-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-fneg.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-fneg.ll
index 0322f74..24f93f0 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-fneg.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-fneg.ll
@@ -37,20 +37,18 @@ define void @fneg(ptr nocapture noundef writeonly %d, ptr nocapture noundef read
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds half, ptr [[S]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds half, ptr [[TMP11]], i32 0
 ; CHECK-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP14:%.*]] = mul nuw i64 [[TMP13]], 8
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds half, ptr [[TMP11]], i64 [[TMP14]]
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x half>, ptr [[TMP12]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x half>, ptr [[TMP11]], align 2
 ; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 8 x half>, ptr [[TMP15]], align 2
 ; CHECK-NEXT:    [[TMP16:%.*]] = fneg <vscale x 8 x half> [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP17:%.*]] = fneg <vscale x 8 x half> [[WIDE_LOAD3]]
 ; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds half, ptr [[D]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds half, ptr [[TMP18]], i32 0
 ; CHECK-NEXT:    [[TMP20:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP21:%.*]] = mul nuw i64 [[TMP20]], 8
 ; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds half, ptr [[TMP18]], i64 [[TMP21]]
-; CHECK-NEXT:    store <vscale x 8 x half> [[TMP16]], ptr [[TMP19]], align 2
+; CHECK-NEXT:    store <vscale x 8 x half> [[TMP16]], ptr [[TMP18]], align 2
 ; CHECK-NEXT:    store <vscale x 8 x half> [[TMP17]], ptr [[TMP22]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
 ; CHECK-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-inductions-unusual-types.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-inductions-unusual-types.ll
index 70042ca..fefb5af 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-inductions-unusual-types.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-inductions-unusual-types.ll
@@ -37,11 +37,10 @@ define void @induction_i7(ptr %dst) #0 {
 ; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i64, ptr [[DST]], i64 [[INDEX]]
 ; CHECK-NEXT:    [[TMP23:%.*]] = zext <vscale x 2 x i7> [[TMP19]] to <vscale x 2 x i64>
 ; CHECK-NEXT:    [[TMP24:%.*]] = zext <vscale x 2 x i7> [[TMP20]] to <vscale x 2 x i64>
-; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i64, ptr [[TMP21]], i32 0
 ; CHECK-NEXT:    [[TMP26:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP27:%.*]] = mul nuw i64 [[TMP26]], 2
 ; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i64, ptr [[TMP21]], i64 [[TMP27]]
-; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP23]], ptr [[TMP25]], align 8
+; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP23]], ptr [[TMP21]], align 8
 ; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP24]], ptr [[TMP28]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 2 x i7> [[STEP_ADD]], [[DOTSPLAT]]
@@ -100,11 +99,10 @@ define void @induction_i3_zext(ptr %dst) #0 {
 ; CHECK-NEXT:    [[TMP19:%.*]] = zext <vscale x 2 x i3> [[VEC_IND]] to <vscale x 2 x i64>
 ; CHECK-NEXT:    [[TMP20:%.*]] = zext <vscale x 2 x i3> [[STEP_ADD]] to <vscale x 2 x i64>
 ; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i64, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i64, ptr [[TMP21]], i32 0
 ; CHECK-NEXT:    [[TMP24:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP25:%.*]] = mul nuw i64 [[TMP24]], 2
 ; CHECK-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i64, ptr [[TMP21]], i64 [[TMP25]]
-; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP19]], ptr [[TMP23]], align 8
+; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP19]], ptr [[TMP21]], align 8
 ; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP20]], ptr [[TMP26]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 2 x i3> [[STEP_ADD]], [[DOTSPLAT]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-inv-store.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-inv-store.ll
index f8c635b..f0675a4 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-inv-store.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-inv-store.ll
@@ -21,8 +21,7 @@ define void @inv_store_i16(ptr noalias %dst, ptr noalias readonly %src, i64 %N)
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i16, ptr [[SRC:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP8]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP7]], align 2
 ; CHECK-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vscale.i32()
 ; CHECK-NEXT:    [[TMP10:%.*]] = mul nuw i32 [[TMP9]], 4
 ; CHECK-NEXT:    [[TMP11:%.*]] = sub i32 [[TMP10]], 1
@@ -73,8 +72,7 @@ define void @cond_inv_store_i32(ptr noalias %dst, ptr noalias readonly %src, i64
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp sgt <vscale x 4 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    call void @llvm.masked.scatter.nxv4i32.nxv4p0(<vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x ptr> [[BROADCAST_SPLAT]], i32 4, <vscale x 4 x i1> [[TMP9]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-live-out-pointer-induction.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-live-out-pointer-induction.ll
index cac526f..2b4aad1 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-live-out-pointer-induction.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-live-out-pointer-induction.ll
@@ -30,11 +30,10 @@ define ptr @test(ptr %start.1, ptr %start.2, ptr %end) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr i8, ptr [[START_2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP32:%.*]] = getelementptr i64, ptr [[TMP30]], i32 0
 ; CHECK-NEXT:    [[TMP33:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP34:%.*]] = mul nuw i64 [[TMP33]], 2
 ; CHECK-NEXT:    [[TMP35:%.*]] = getelementptr i64, ptr [[TMP30]], i64 [[TMP34]]
-; CHECK-NEXT:    store <vscale x 2 x i64> zeroinitializer, ptr [[TMP32]], align 8
+; CHECK-NEXT:    store <vscale x 2 x i64> zeroinitializer, ptr [[TMP30]], align 8
 ; CHECK-NEXT:    store <vscale x 2 x i64> zeroinitializer, ptr [[TMP35]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP11]]
 ; CHECK-NEXT:    [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-multiexit.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-multiexit.ll
index 246beb2..993c048 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-multiexit.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-multiexit.ll
@@ -39,18 +39,16 @@ define void @multiple_exits_unique_exit_block(ptr %A, ptr %B, i32 %N) #0 {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, ptr [[TMP19]], i32 0
 ; CHECK-NEXT:    [[TMP22:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP23:%.*]] = mul nuw i64 [[TMP22]], 4
 ; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, ptr [[TMP19]], i64 [[TMP23]]
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP21]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP19]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x i32>, ptr [[TMP24]], align 4
 ; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i32, ptr [[B]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i32, ptr [[TMP25]], i32 0
 ; CHECK-NEXT:    [[TMP28:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP29:%.*]] = mul nuw i64 [[TMP28]], 4
 ; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i32, ptr [[TMP25]], i64 [[TMP29]]
-; CHECK-NEXT:    store <vscale x 4 x i32> [[WIDE_LOAD]], ptr [[TMP27]], align 4
+; CHECK-NEXT:    store <vscale x 4 x i32> [[WIDE_LOAD]], ptr [[TMP25]], align 4
 ; CHECK-NEXT:    store <vscale x 4 x i32> [[WIDE_LOAD3]], ptr [[TMP30]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP12]]
 ; CHECK-NEXT:    [[TMP31:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
@@ -113,18 +111,16 @@ define i32 @multiple_exits_multiple_exit_blocks(ptr %A, ptr %B, i32 %N) #0 {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, ptr [[TMP19]], i32 0
 ; CHECK-NEXT:    [[TMP22:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP23:%.*]] = mul nuw i64 [[TMP22]], 4
 ; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, ptr [[TMP19]], i64 [[TMP23]]
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP21]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP19]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x i32>, ptr [[TMP24]], align 4
 ; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i32, ptr [[B]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i32, ptr [[TMP25]], i32 0
 ; CHECK-NEXT:    [[TMP28:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP29:%.*]] = mul nuw i64 [[TMP28]], 4
 ; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i32, ptr [[TMP25]], i64 [[TMP29]]
-; CHECK-NEXT:    store <vscale x 4 x i32> [[WIDE_LOAD]], ptr [[TMP27]], align 4
+; CHECK-NEXT:    store <vscale x 4 x i32> [[WIDE_LOAD]], ptr [[TMP25]], align 4
 ; CHECK-NEXT:    store <vscale x 4 x i32> [[WIDE_LOAD3]], ptr [[TMP30]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP12]]
 ; CHECK-NEXT:    [[TMP31:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-runtime-check-size-based-threshold.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-runtime-check-size-based-threshold.ll
index b6f723e..893ebef 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-runtime-check-size-based-threshold.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-runtime-check-size-based-threshold.ll
@@ -51,34 +51,30 @@ define void @min_trip_count_due_to_runtime_checks_1(ptr %dst.1, ptr %dst.2, ptr
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr i64, ptr [[SRC_1]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr i64, ptr [[SRC_2]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP27:%.*]] = getelementptr i64, ptr [[TMP23]], i32 0
+; CHECK-NEXT:    [[TMP27:%.*]] = getelementptr i64, ptr [[SRC_1]], i64 [[INDEX]]
+; CHECK-NEXT:    [[TMP31:%.*]] = getelementptr i64, ptr [[SRC_2]], i64 [[INDEX]]
 ; CHECK-NEXT:    [[TMP28:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP29:%.*]] = mul nuw i64 [[TMP28]], 2
-; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr i64, ptr [[TMP23]], i64 [[TMP29]]
+; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr i64, ptr [[TMP27]], i64 [[TMP29]]
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP27]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD12:%.*]] = load <vscale x 2 x i64>, ptr [[TMP30]], align 8
-; CHECK-NEXT:    [[TMP31:%.*]] = getelementptr i64, ptr [[TMP25]], i32 0
 ; CHECK-NEXT:    [[TMP32:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP33:%.*]] = mul nuw i64 [[TMP32]], 2
-; CHECK-NEXT:    [[TMP34:%.*]] = getelementptr i64, ptr [[TMP25]], i64 [[TMP33]]
+; CHECK-NEXT:    [[TMP34:%.*]] = getelementptr i64, ptr [[TMP31]], i64 [[TMP33]]
 ; CHECK-NEXT:    [[WIDE_LOAD13:%.*]] = load <vscale x 2 x i64>, ptr [[TMP31]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD14:%.*]] = load <vscale x 2 x i64>, ptr [[TMP34]], align 8
 ; CHECK-NEXT:    [[TMP35:%.*]] = add <vscale x 2 x i64> [[WIDE_LOAD]], [[WIDE_LOAD13]]
 ; CHECK-NEXT:    [[TMP36:%.*]] = add <vscale x 2 x i64> [[WIDE_LOAD12]], [[WIDE_LOAD14]]
-; CHECK-NEXT:    [[TMP37:%.*]] = getelementptr i64, ptr [[DST_1]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP39:%.*]] = getelementptr i64, ptr [[DST_2]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP41:%.*]] = getelementptr i64, ptr [[TMP37]], i32 0
+; CHECK-NEXT:    [[TMP41:%.*]] = getelementptr i64, ptr [[DST_1]], i64 [[INDEX]]
+; CHECK-NEXT:    [[TMP45:%.*]] = getelementptr i64, ptr [[DST_2]], i64 [[INDEX]]
 ; CHECK-NEXT:    [[TMP42:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP43:%.*]] = mul nuw i64 [[TMP42]], 2
-; CHECK-NEXT:    [[TMP44:%.*]] = getelementptr i64, ptr [[TMP37]], i64 [[TMP43]]
+; CHECK-NEXT:    [[TMP44:%.*]] = getelementptr i64, ptr [[TMP41]], i64 [[TMP43]]
 ; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP35]], ptr [[TMP41]], align 8
 ; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP36]], ptr [[TMP44]], align 8
-; CHECK-NEXT:    [[TMP45:%.*]] = getelementptr i64, ptr [[TMP39]], i32 0
 ; CHECK-NEXT:    [[TMP46:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP47:%.*]] = mul nuw i64 [[TMP46]], 2
-; CHECK-NEXT:    [[TMP48:%.*]] = getelementptr i64, ptr [[TMP39]], i64 [[TMP47]]
+; CHECK-NEXT:    [[TMP48:%.*]] = getelementptr i64, ptr [[TMP45]], i64 [[TMP47]]
 ; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP35]], ptr [[TMP45]], align 8
 ; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP36]], ptr [[TMP48]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP50]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-forced.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-forced.ll
index eb8f218..1cda568 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-forced.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-forced.ll
@@ -67,8 +67,7 @@ define void @simple_memset(i32 %val, ptr %ptr, i64 %n) #0 {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT2]] = add i64 [[INDEX1]], [[TMP14]]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX1]], i64 [[TMP9]])
 ; CHECK-NEXT:    [[TMP15:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-optsize.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-optsize.ll
index 90b4901..fb0447b 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-optsize.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-optsize.ll
@@ -22,14 +22,12 @@ define void @trip1025_i64(ptr noalias nocapture noundef %dst, ptr noalias nocapt
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 2 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[SRC:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP9]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i64> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP8]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i64> poison)
 ; CHECK-NEXT:    [[TMP10:%.*]] = shl nsw <vscale x 2 x i64> [[WIDE_MASKED_LOAD]], splat (i64 1)
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[DST:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP12]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i64> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 2 x i64> @llvm.masked.load.nxv2i64.p0(ptr [[TMP11]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]], <vscale x 2 x i64> poison)
 ; CHECK-NEXT:    [[TMP13:%.*]] = add nsw <vscale x 2 x i64> [[WIDE_MASKED_LOAD1]], [[TMP10]]
-; CHECK-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP13]], ptr [[TMP12]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP13]], ptr [[TMP11]], i32 8, <vscale x 2 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP6]]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 2 x i1> @llvm.get.active.lane.mask.nxv2i1.i64(i64 [[INDEX_NEXT]], i64 1025)
 ; CHECK-NEXT:    [[TMP14:%.*]] = xor <vscale x 2 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-reductions.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-reductions.ll
index 2de24b0..fc86e3a 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-reductions.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-reductions.ll
@@ -32,8 +32,7 @@ define i32 @add_reduction_i32(ptr %ptr, i64 %n) #0 {
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
 ; CHECK-NEXT:    [[TMP13:%.*]] = add <vscale x 4 x i32> [[VEC_PHI]], [[WIDE_MASKED_LOAD]]
 ; CHECK-NEXT:    [[TMP14]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> [[TMP13]], <vscale x 4 x i32> [[VEC_PHI]]
 ; CHECK-NEXT:    [[INDEX_NEXT2]] = add i64 [[INDEX1]], [[TMP16]]
@@ -86,8 +85,7 @@ define i32 @add_reduction_i32(ptr %ptr, i64 %n) #0 {
 ; CHECK-IN-LOOP-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-IN-LOOP-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-IN-LOOP-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[INDEX1]]
-; CHECK-IN-LOOP-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr [[TMP11]], i32 0
-; CHECK-IN-LOOP-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
+; CHECK-IN-LOOP-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
 ; CHECK-IN-LOOP-NEXT:    [[TMP13:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> [[WIDE_MASKED_LOAD]], <vscale x 4 x i32> zeroinitializer
 ; CHECK-IN-LOOP-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vector.reduce.add.nxv4i32(<vscale x 4 x i32> [[TMP13]])
 ; CHECK-IN-LOOP-NEXT:    [[TMP15]] = add i32 [[TMP14]], [[VEC_PHI]]
@@ -158,8 +156,7 @@ define float @add_reduction_f32(ptr %ptr, i64 %n) #0 {
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr float, ptr [[PTR:%.*]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr float, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
 ; CHECK-NEXT:    [[TMP13:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> [[WIDE_MASKED_LOAD]], <vscale x 4 x float> splat (float -0.000000e+00)
 ; CHECK-NEXT:    [[TMP14]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[TMP13]])
 ; CHECK-NEXT:    [[INDEX_NEXT2]] = add i64 [[INDEX1]], [[TMP16]]
@@ -211,8 +208,7 @@ define float @add_reduction_f32(ptr %ptr, i64 %n) #0 {
 ; CHECK-IN-LOOP-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-IN-LOOP-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-IN-LOOP-NEXT:    [[TMP11:%.*]] = getelementptr float, ptr [[PTR:%.*]], i64 [[INDEX1]]
-; CHECK-IN-LOOP-NEXT:    [[TMP12:%.*]] = getelementptr float, ptr [[TMP11]], i32 0
-; CHECK-IN-LOOP-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
+; CHECK-IN-LOOP-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
 ; CHECK-IN-LOOP-NEXT:    [[TMP13:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> [[WIDE_MASKED_LOAD]], <vscale x 4 x float> splat (float -0.000000e+00)
 ; CHECK-IN-LOOP-NEXT:    [[TMP14]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[TMP13]])
 ; CHECK-IN-LOOP-NEXT:    [[INDEX_NEXT2]] = add i64 [[INDEX1]], [[TMP16]]
@@ -281,13 +277,11 @@ define i32 @cond_xor_reduction(ptr noalias %a, ptr noalias %cond, i64 %N) #0 {
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ insertelement (<vscale x 4 x i32> zeroinitializer, i32 7, i32 0), [[VECTOR_PH]] ], [ [[TMP20:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[COND:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq <vscale x 4 x i32> [[WIDE_MASKED_LOAD]], splat (i32 5)
 ; CHECK-NEXT:    [[TMP15:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i1> [[TMP13]], <vscale x 4 x i1> zeroinitializer
 ; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr i32, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr i32, ptr [[TMP14]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP16]], i32 4, <vscale x 4 x i1> [[TMP15]], <vscale x 4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP14]], i32 4, <vscale x 4 x i1> [[TMP15]], <vscale x 4 x i32> poison)
 ; CHECK-NEXT:    [[TMP17:%.*]] = xor <vscale x 4 x i32> [[VEC_PHI]], [[WIDE_MASKED_LOAD1]]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <vscale x 4 x i1> [[TMP15]], <vscale x 4 x i32> [[TMP17]], <vscale x 4 x i32> [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP20]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> [[PREDPHI]], <vscale x 4 x i32> [[VEC_PHI]]
@@ -348,13 +342,11 @@ define i32 @cond_xor_reduction(ptr noalias %a, ptr noalias %cond, i64 %N) #0 {
 ; CHECK-IN-LOOP-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-IN-LOOP-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ 7, [[VECTOR_PH]] ], [ [[TMP19:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-IN-LOOP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[COND:%.*]], i64 [[INDEX]]
-; CHECK-IN-LOOP-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; CHECK-IN-LOOP-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
+; CHECK-IN-LOOP-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
 ; CHECK-IN-LOOP-NEXT:    [[TMP13:%.*]] = icmp eq <vscale x 4 x i32> [[WIDE_MASKED_LOAD]], splat (i32 5)
 ; CHECK-IN-LOOP-NEXT:    [[TMP15:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i1> [[TMP13]], <vscale x 4 x i1> zeroinitializer
 ; CHECK-IN-LOOP-NEXT:    [[TMP14:%.*]] = getelementptr i32, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-IN-LOOP-NEXT:    [[TMP16:%.*]] = getelementptr i32, ptr [[TMP14]], i32 0
-; CHECK-IN-LOOP-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP16]], i32 4, <vscale x 4 x i1> [[TMP15]], <vscale x 4 x i32> poison)
+; CHECK-IN-LOOP-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP14]], i32 4, <vscale x 4 x i1> [[TMP15]], <vscale x 4 x i32> poison)
 ; CHECK-IN-LOOP-NEXT:    [[TMP17:%.*]] = select <vscale x 4 x i1> [[TMP15]], <vscale x 4 x i32> [[WIDE_MASKED_LOAD1]], <vscale x 4 x i32> zeroinitializer
 ; CHECK-IN-LOOP-NEXT:    [[TMP18:%.*]] = call i32 @llvm.vector.reduce.xor.nxv4i32(<vscale x 4 x i32> [[TMP17]])
 ; CHECK-IN-LOOP-NEXT:    [[TMP19]] = xor i32 [[TMP18]], [[VEC_PHI]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-unroll.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-unroll.ll
index ea9cd3f..4ec7d4d 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-unroll.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding-unroll.ll
@@ -46,7 +46,6 @@ define void @simple_memset(i32 %val, ptr %ptr, i64 %n) #0 {
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK8:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY4]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT12:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK9:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY5]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT13:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP47:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[INDEX6]]
-; CHECK-NEXT:    [[TMP51:%.*]] = getelementptr i32, ptr [[TMP47]], i32 0
 ; CHECK-NEXT:    [[TMP52:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP53:%.*]] = mul nuw i64 [[TMP52]], 4
 ; CHECK-NEXT:    [[TMP54:%.*]] = getelementptr i32, ptr [[TMP47]], i64 [[TMP53]]
@@ -56,7 +55,7 @@ define void @simple_memset(i32 %val, ptr %ptr, i64 %n) #0 {
 ; CHECK-NEXT:    [[TMP58:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP59:%.*]] = mul nuw i64 [[TMP58]], 12
 ; CHECK-NEXT:    [[TMP60:%.*]] = getelementptr i32, ptr [[TMP47]], i64 [[TMP59]]
-; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP51]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP47]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP54]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK7]])
 ; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP57]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK8]])
 ; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP60]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK9]])
@@ -138,7 +137,6 @@ define void @cond_memset(i32 %val, ptr noalias readonly %cond_ptr, ptr noalias %
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK8:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY4]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT15:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK9:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY5]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT16:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP47:%.*]] = getelementptr i32, ptr [[COND_PTR:%.*]], i64 [[INDEX6]]
-; CHECK-NEXT:    [[TMP51:%.*]] = getelementptr i32, ptr [[TMP47]], i32 0
 ; CHECK-NEXT:    [[TMP52:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP53:%.*]] = mul nuw i64 [[TMP52]], 4
 ; CHECK-NEXT:    [[TMP54:%.*]] = getelementptr i32, ptr [[TMP47]], i64 [[TMP53]]
@@ -148,7 +146,7 @@ define void @cond_memset(i32 %val, ptr noalias readonly %cond_ptr, ptr noalias %
 ; CHECK-NEXT:    [[TMP58:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP59:%.*]] = mul nuw i64 [[TMP58]], 12
 ; CHECK-NEXT:    [[TMP60:%.*]] = getelementptr i32, ptr [[TMP47]], i64 [[TMP59]]
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP51]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP47]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP54]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK7]], <vscale x 4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP57]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP60]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK9]], <vscale x 4 x i32> poison)
@@ -161,7 +159,6 @@ define void @cond_memset(i32 %val, ptr noalias readonly %cond_ptr, ptr noalias %
 ; CHECK-NEXT:    [[TMP71:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK8]], <vscale x 4 x i1> [[TMP63]], <vscale x 4 x i1> zeroinitializer
 ; CHECK-NEXT:    [[TMP72:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK9]], <vscale x 4 x i1> [[TMP64]], <vscale x 4 x i1> zeroinitializer
 ; CHECK-NEXT:    [[TMP65:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[INDEX6]]
-; CHECK-NEXT:    [[TMP73:%.*]] = getelementptr i32, ptr [[TMP65]], i32 0
 ; CHECK-NEXT:    [[TMP74:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP75:%.*]] = mul nuw i64 [[TMP74]], 4
 ; CHECK-NEXT:    [[TMP76:%.*]] = getelementptr i32, ptr [[TMP65]], i64 [[TMP75]]
@@ -171,7 +168,7 @@ define void @cond_memset(i32 %val, ptr noalias readonly %cond_ptr, ptr noalias %
 ; CHECK-NEXT:    [[TMP80:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP81:%.*]] = mul nuw i64 [[TMP80]], 12
 ; CHECK-NEXT:    [[TMP82:%.*]] = getelementptr i32, ptr [[TMP65]], i64 [[TMP81]]
-; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP73]], i32 4, <vscale x 4 x i1> [[TMP69]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP65]], i32 4, <vscale x 4 x i1> [[TMP69]])
 ; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP76]], i32 4, <vscale x 4 x i1> [[TMP70]])
 ; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP79]], i32 4, <vscale x 4 x i1> [[TMP71]])
 ; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP82]], i32 4, <vscale x 4 x i1> [[TMP72]])
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding.ll
index f6f8895..672523e 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-tail-folding.ll
@@ -31,8 +31,7 @@ define void @simple_memset(i32 %val, ptr %ptr, i64 %n) #0 {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT2]] = add i64 [[INDEX1]], [[TMP4]]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX1]], i64 [[TMP9]])
 ; CHECK-NEXT:    [[TMP13:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
@@ -78,8 +77,7 @@ define void @simple_memset_v4i32(i32 %val, ptr %ptr, i64 %n) #0 {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP5]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP4]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT2]] = add i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i64(i64 [[INDEX1]], i64 [[TMP2]])
 ; CHECK-NEXT:    [[TMP6:%.*]] = xor <4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
@@ -130,11 +128,9 @@ define void @simple_memcpy(ptr noalias %dst, ptr noalias %src, i64 %n) #0 {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[SRC:%.*]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr i32, ptr [[DST:%.*]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr i32, ptr [[TMP13]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[WIDE_MASKED_LOAD]], ptr [[TMP14]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[WIDE_MASKED_LOAD]], ptr [[TMP13]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT2]] = add i64 [[INDEX1]], [[TMP4]]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX1]], i64 [[TMP9]])
 ; CHECK-NEXT:    [[TMP15:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
@@ -253,8 +249,7 @@ define void @simple_gather_scatter(ptr noalias %dst, ptr noalias %src, ptr noali
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[IND:%.*]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr i32, ptr [[SRC:%.*]], <vscale x 4 x i32> [[WIDE_MASKED_LOAD]]
 ; CHECK-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 4 x i32> @llvm.masked.gather.nxv4i32.nxv4p0(<vscale x 4 x ptr> [[TMP13]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
 ; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr i32, ptr [[DST:%.*]], <vscale x 4 x i32> [[WIDE_MASKED_LOAD]]
@@ -317,8 +312,7 @@ define void @uniform_load(ptr noalias %dst, ptr noalias readonly %src, i64 %n) #
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[TMP11]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP13]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP4]]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX]], i64 [[TMP9]])
 ; CHECK-NEXT:    [[TMP14:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
@@ -376,15 +370,13 @@ define void @cond_uniform_load(ptr noalias %dst, ptr noalias readonly %src, ptr
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[COND:%.*]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
 ; CHECK-NEXT:    [[TMP14:%.*]] = icmp ne <vscale x 4 x i32> [[WIDE_MASKED_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP15:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i1> [[TMP14]], <vscale x 4 x i1> zeroinitializer
 ; CHECK-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 4 x i32> @llvm.masked.gather.nxv4i32.nxv4p0(<vscale x 4 x ptr> [[BROADCAST_SPLAT]], i32 4, <vscale x 4 x i1> [[TMP15]], <vscale x 4 x i32> poison)
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <vscale x 4 x i1> [[TMP15]], <vscale x 4 x i32> [[WIDE_MASKED_GATHER]], <vscale x 4 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[PREDPHI]], ptr [[TMP17]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[PREDPHI]], ptr [[TMP16]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT2]] = add i64 [[INDEX1]], [[TMP4]]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX1]], i64 [[TMP9]])
 ; CHECK-NEXT:    [[TMP18:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
@@ -450,8 +442,7 @@ define void @uniform_store(ptr noalias %dst, ptr noalias readonly %src, i64 %n)
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
 ; CHECK-NEXT:    call void @llvm.masked.scatter.nxv4i32.nxv4p0(<vscale x 4 x i32> [[WIDE_MASKED_LOAD]], <vscale x 4 x ptr> [[BROADCAST_SPLAT]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP4]]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX]], i64 [[TMP9]])
@@ -506,12 +497,10 @@ define void @simple_fdiv(ptr noalias %dst, ptr noalias %src, i64 %n) #0 {
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr float, ptr [[SRC:%.*]], i64 [[INDEX1]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr float, ptr [[DST:%.*]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr float, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP13]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr float, ptr [[TMP12]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD2:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP14]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD2:%.*]] = call <vscale x 4 x float> @llvm.masked.load.nxv4f32.p0(ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x float> poison)
 ; CHECK-NEXT:    [[TMP15:%.*]] = fdiv <vscale x 4 x float> [[WIDE_MASKED_LOAD]], [[WIDE_MASKED_LOAD2]]
-; CHECK-NEXT:    call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[TMP15]], ptr [[TMP14]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv4f32.p0(<vscale x 4 x float> [[TMP15]], ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add i64 [[INDEX1]], [[TMP4]]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX1]], i64 [[TMP9]])
 ; CHECK-NEXT:    [[TMP16:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
@@ -568,13 +557,11 @@ define void @simple_idiv(ptr noalias %dst, ptr noalias %src, i64 %n) #0 {
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[SRC:%.*]], i64 [[INDEX1]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr [[DST:%.*]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr i32, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP13]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr i32, ptr [[TMP12]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD2:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP14]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD2:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> poison)
 ; CHECK-NEXT:    [[TMP15:%.*]] = select <vscale x 4 x i1> [[ACTIVE_LANE_MASK]], <vscale x 4 x i32> [[WIDE_MASKED_LOAD2]], <vscale x 4 x i32> splat (i32 1)
 ; CHECK-NEXT:    [[TMP16:%.*]] = udiv <vscale x 4 x i32> [[WIDE_MASKED_LOAD]], [[TMP15]]
-; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP16]], ptr [[TMP14]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP16]], ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add i64 [[INDEX1]], [[TMP4]]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX1]], i64 [[TMP9]])
 ; CHECK-NEXT:    [[TMP17:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
@@ -622,8 +609,7 @@ define void @simple_memset_trip1024(i32 %val, ptr %ptr, i64 %n) #0 {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP8]], align 4
+; CHECK-NEXT:    store <vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT2]] = add nuw i64 [[INDEX1]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT2]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-vscale-based-trip-counts.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-vscale-based-trip-counts.ll
index e2c7469..3e9f6fa 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-vscale-based-trip-counts.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-vscale-based-trip-counts.ll
@@ -24,25 +24,23 @@ define void @vscale_mul_4(ptr noalias noundef readonly captures(none) %a, ptr no
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw float, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw float, ptr [[TMP13]], i32 0
 ; CHECK-NEXT:    [[TMP18:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP11:%.*]] = mul nuw i64 [[TMP18]], 4
 ; CHECK-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw float, ptr [[TMP13]], i64 [[TMP11]]
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP14]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP13]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP26]], align 4
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw float, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw float, ptr [[TMP12]], i32 0
 ; CHECK-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP16:%.*]] = mul nuw i64 [[TMP15]], 4
 ; CHECK-NEXT:    [[TMP27:%.*]] = getelementptr inbounds nuw float, ptr [[TMP12]], i64 [[TMP16]]
-; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x float>, ptr [[TMP17]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x float>, ptr [[TMP12]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 4 x float>, ptr [[TMP27]], align 4
 ; CHECK-NEXT:    [[TMP19:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD2]], [[WIDE_LOAD3]]
 ; CHECK-NEXT:    [[TMP28:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD1]], [[WIDE_LOAD4]]
 ; CHECK-NEXT:    [[TMP20:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP21:%.*]] = mul nuw i64 [[TMP20]], 4
 ; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw float, ptr [[TMP12]], i64 [[TMP21]]
-; CHECK-NEXT:    store <vscale x 4 x float> [[TMP19]], ptr [[TMP17]], align 4
+; CHECK-NEXT:    store <vscale x 4 x float> [[TMP19]], ptr [[TMP12]], align 4
 ; CHECK-NEXT:    store <vscale x 4 x float> [[TMP28]], ptr [[TMP22]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP7]]
 ; CHECK-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -102,25 +100,22 @@ define  void @vscale_mul_8(ptr noalias noundef readonly captures(none) %a, ptr n
 ; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[MUL1]], [[N_MOD_VF]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 8
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw float, ptr [[A]], i32 0
 ; CHECK-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw float, ptr [[A]], i64 [[TMP10]]
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[A]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw float, ptr [[B]], i32 0
 ; CHECK-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP14]], 4
 ; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw float, ptr [[B]], i64 [[TMP15]]
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP13]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[B]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x float>, ptr [[TMP16]], align 4
 ; CHECK-NEXT:    [[TMP17:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[TMP18:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD1]], [[WIDE_LOAD3]]
-; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw float, ptr [[B]], i32 0
 ; CHECK-NEXT:    [[TMP20:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP21:%.*]] = mul nuw i64 [[TMP20]], 4
 ; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw float, ptr [[B]], i64 [[TMP21]]
-; CHECK-NEXT:    store <vscale x 4 x float> [[TMP17]], ptr [[TMP19]], align 4
+; CHECK-NEXT:    store <vscale x 4 x float> [[TMP17]], ptr [[B]], align 4
 ; CHECK-NEXT:    store <vscale x 4 x float> [[TMP18]], ptr [[TMP22]], align 4
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[MUL1]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label %[[FOR_COND_CLEANUP:.*]], label %[[FOR_BODY:.*]]
@@ -180,25 +175,23 @@ define void @vscale_mul_12(ptr noalias noundef readonly captures(none) %a, ptr n
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw float, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw float, ptr [[TMP7]], i32 0
 ; CHECK-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw float, ptr [[TMP7]], i64 [[TMP10]]
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw float, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw float, ptr [[TMP12]], i32 0
 ; CHECK-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP14]], 4
 ; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw float, ptr [[TMP12]], i64 [[TMP15]]
-; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x float>, ptr [[TMP13]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x float>, ptr [[TMP12]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <vscale x 4 x float>, ptr [[TMP16]], align 4
 ; CHECK-NEXT:    [[TMP18:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD1]], [[WIDE_LOAD3]]
 ; CHECK-NEXT:    [[TMP25:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD2]], [[WIDE_LOAD4]]
 ; CHECK-NEXT:    [[TMP19:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP20:%.*]] = mul nuw i64 [[TMP19]], 4
 ; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw float, ptr [[TMP12]], i64 [[TMP20]]
-; CHECK-NEXT:    store <vscale x 4 x float> [[TMP18]], ptr [[TMP13]], align 4
+; CHECK-NEXT:    store <vscale x 4 x float> [[TMP18]], ptr [[TMP12]], align 4
 ; CHECK-NEXT:    store <vscale x 4 x float> [[TMP25]], ptr [[TMP21]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
 ; CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -265,25 +258,23 @@ define void @vscale_mul_31(ptr noalias noundef readonly captures(none) %a, ptr n
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw float, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw float, ptr [[TMP7]], i32 0
 ; CHECK-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw float, ptr [[TMP7]], i64 [[TMP10]]
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw float, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw float, ptr [[TMP12]], i32 0
 ; CHECK-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP14]], 4
 ; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw float, ptr [[TMP12]], i64 [[TMP15]]
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP13]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP12]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x float>, ptr [[TMP16]], align 4
 ; CHECK-NEXT:    [[TMP17:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[TMP18:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD1]], [[WIDE_LOAD3]]
 ; CHECK-NEXT:    [[TMP19:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP20:%.*]] = mul nuw i64 [[TMP19]], 4
 ; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw float, ptr [[TMP12]], i64 [[TMP20]]
-; CHECK-NEXT:    store <vscale x 4 x float> [[TMP17]], ptr [[TMP13]], align 4
+; CHECK-NEXT:    store <vscale x 4 x float> [[TMP17]], ptr [[TMP12]], align 4
 ; CHECK-NEXT:    store <vscale x 4 x float> [[TMP18]], ptr [[TMP21]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
 ; CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -350,25 +341,23 @@ define void @vscale_mul_64(ptr noalias noundef readonly captures(none) %a, ptr n
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw float, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw float, ptr [[TMP7]], i32 0
 ; CHECK-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw float, ptr [[TMP7]], i64 [[TMP10]]
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw float, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw float, ptr [[TMP12]], i32 0
 ; CHECK-NEXT:    [[TMP14:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP14]], 4
 ; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw float, ptr [[TMP12]], i64 [[TMP15]]
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP13]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP12]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <vscale x 4 x float>, ptr [[TMP16]], align 4
 ; CHECK-NEXT:    [[TMP17:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[TMP18:%.*]] = fmul <vscale x 4 x float> [[WIDE_LOAD1]], [[WIDE_LOAD3]]
 ; CHECK-NEXT:    [[TMP19:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP20:%.*]] = mul nuw i64 [[TMP19]], 4
 ; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw float, ptr [[TMP12]], i64 [[TMP20]]
-; CHECK-NEXT:    store <vscale x 4 x float> [[TMP17]], ptr [[TMP13]], align 4
+; CHECK-NEXT:    store <vscale x 4 x float> [[TMP17]], ptr [[TMP12]], align 4
 ; CHECK-NEXT:    store <vscale x 4 x float> [[TMP18]], ptr [[TMP21]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
 ; CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-gep.ll b/llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-gep.ll
index e58ea65..0754a38 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-gep.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/sve-widen-gep.ll
@@ -9,38 +9,6 @@ target triple = "aarch64-unknown-linux-gnu"
 ; CHECK-NOT:    LV: Found {{.*}} scalar instruction:   %ptr.iv.2.next = getelementptr inbounds i8, ptr %ptr.iv.2, i64 1
 ;
 ; CHECK:        VPlan 'Initial VPlan for VF={vscale x 2},UF>=1' {
-; CHECK-NEXT:   Live-in vp<[[VF:%.+]]> = VF
-; CHECK-NEXT:   Live-in vp<[[VFxUF:%.+]]> = VF * UF
-; CHECK-NEXT:   Live-in vp<[[VEC_TC:%.+]]> = vector-trip-count
-; CHECK-NEXT:   Live-in ir<%N> = original trip-count
-; CHECK-EMPTY:
-; CHECK-NEXT: ir-bb<entry>:
-; CHECK-NEXT: Successor(s): scalar.ph, vector.ph
-; CHECK-EMPTY:
-; CHECK-NEXT: vector.ph:
-; CHECK-NEXT:  vp<[[END1:%.+]]> = DERIVED-IV ir<%start.1> + vp<[[VEC_TC]]> * ir<8>
-; CHECK-NEXT:  vp<[[END2:%.+]]> = DERIVED-IV ir<%start.2> + vp<[[VEC_TC]]> * ir<1>
-; CHECK-NEXT: Successor(s): vector loop
-; CHECK-EMPTY:
-; CHECK-NEXT:   <x1> vector loop: {
-; CHECK-NEXT:   vector.body:
-; CHECK-NEXT:     EMIT vp<[[CAN_IV:%.+]]> = CANONICAL-INDUCTION
-; CHECK-NEXT:     EMIT ir<%ptr.iv.2> = WIDEN-POINTER-INDUCTION ir<%start.2>, ir<1>
-; CHECK-NEXT:     vp<[[PTR_IDX:%.+]]> = DERIVED-IV ir<0> + vp<[[CAN_IV]]> * ir<8>
-; CHECK-NEXT:     vp<[[PTR_IDX_STEPS:%.+]]> = SCALAR-STEPS vp<[[PTR_IDX]]>, ir<8>, vp<[[VF]]>
-; CHECK-NEXT:     EMIT vp<[[PTR_IV_1:%.+]]> = ptradd ir<%start.1>, vp<[[PTR_IDX_STEPS]]>
-; CHECK-NEXT:     WIDEN-GEP Var[Inv] ir<%ptr.iv.2.next> = getelementptr inbounds ir<%ptr.iv.2>, ir<1>
-; CHECK-NEXT:     vp<[[VEC_PTR:%.+]]> = vector-pointer vp<[[PTR_IV_1]]>
-; CHECK-NEXT:     WIDEN store vp<[[VEC_PTR]]>, ir<%ptr.iv.2.next>
-; CHECK-NEXT:     vp<[[VEC_PTR2:%.+]]> = vector-pointer ir<%ptr.iv.2>
-; CHECK-NEXT:     WIDEN ir<%lv> = load vp<[[VEC_PTR2]]>
-; CHECK-NEXT:     WIDEN ir<%add> = add ir<%lv>, ir<1>
-; CHECK-NEXT:     vp<[[VEC_PTR3:%.+]]> = vector-pointer ir<%ptr.iv.2>
-; CHECK-NEXT:     WIDEN store vp<[[VEC_PTR3]]>, ir<%add>
-; CHECK-NEXT:     EMIT vp<[[CAN_IV_NEXT:%.+]]> = add nuw vp<[[CAN_IV]]>, vp<[[VFxUF]]>
-; CHECK-NEXT:     EMIT branch-on-count vp<[[CAN_IV_NEXT]]>, vp<[[VEC_TC]]>
-; CHECK-NEXT:   No successors
-; CHECK-NEXT:   }
 
 ; In the test below the pointer phi %ptr.iv.2 is used as
 ;  1. As a uniform address for the load, and
@@ -80,13 +48,11 @@ define void @pointer_induction_used_as_vector(ptr noalias %start.1, ptr noalias
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[START_1]], i64 [[OFFSET_IDX]]
 ; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, <vscale x 2 x ptr> [[VECTOR_GEP]], i64 1
-; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr ptr, ptr [[NEXT_GEP]], i32 0
-; CHECK-NEXT:    store <vscale x 2 x ptr> [[TMP16]], ptr [[TMP17]], align 8
+; CHECK-NEXT:    store <vscale x 2 x ptr> [[TMP16]], ptr [[NEXT_GEP]], align 8
 ; CHECK-NEXT:    [[TMP18:%.*]] = extractelement <vscale x 2 x ptr> [[VECTOR_GEP]], i32 0
-; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr i8, ptr [[TMP18]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i8>, ptr [[TMP19]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i8>, ptr [[TMP18]], align 1
 ; CHECK-NEXT:    [[TMP20:%.*]] = add <vscale x 2 x i8> [[WIDE_LOAD]], splat (i8 1)
-; CHECK-NEXT:    store <vscale x 2 x i8> [[TMP20]], ptr [[TMP19]], align 1
+; CHECK-NEXT:    store <vscale x 2 x i8> [[TMP20]], ptr [[TMP18]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
 ; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i64 [[TMP11]]
 ; CHECK-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -170,10 +136,9 @@ define void @pointer_induction(ptr noalias %start, i64 %N) {
 ; CHECK-NEXT:    [[TMP14:%.*]] = mul <vscale x 2 x i64> [[TMP13]], splat (i64 1)
 ; CHECK-NEXT:    [[VECTOR_GEP:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <vscale x 2 x i64> [[TMP14]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = extractelement <vscale x 2 x ptr> [[VECTOR_GEP]], i32 0
-; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr i8, ptr [[TMP15]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i8>, ptr [[TMP16]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i8>, ptr [[TMP15]], align 1
 ; CHECK-NEXT:    [[TMP17:%.*]] = add <vscale x 2 x i8> [[WIDE_LOAD]], splat (i8 1)
-; CHECK-NEXT:    store <vscale x 2 x i8> [[TMP17]], ptr [[TMP16]], align 1
+; CHECK-NEXT:    store <vscale x 2 x i8> [[TMP17]], ptr [[TMP15]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX2]], [[TMP6]]
 ; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i64 [[TMP10]]
 ; CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/synthesize-mask-for-call.ll b/llvm/test/Transforms/LoopVectorize/AArch64/synthesize-mask-for-call.ll
index 46b4762..3b04df3 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/synthesize-mask-for-call.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/synthesize-mask-for-call.ll
@@ -309,19 +309,17 @@ define void @test_v4_v4m(ptr noalias %a, ptr readonly %b) #3 {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i64, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; CHECK-NEXT:    [[TMP3:%.*]] = call <4 x i64> @foo_vector_fixed4_nomask(<4 x i64> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <4 x i64> [[TMP3]], ptr [[TMP5]], align 8
+; CHECK-NEXT:    store <4 x i64> [[TMP3]], ptr [[TMP4]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
@@ -364,19 +362,17 @@ define void @test_v2_v4m(ptr noalias %a, ptr readonly %b) #3 {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i64, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; CHECK-NEXT:    [[TMP3:%.*]] = call <4 x i64> @foo_vector_fixed4_mask(<4 x i64> [[WIDE_LOAD]], <4 x i1> splat (i1 true))
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <4 x i64> [[TMP3]], ptr [[TMP5]], align 8
+; CHECK-NEXT:    store <4 x i64> [[TMP3]], ptr [[TMP4]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
@@ -419,19 +415,17 @@ define void @test_v2_v4(ptr noalias %a, ptr readonly %b) #3 {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i64, ptr [[B:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i64, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; CHECK-NEXT:    [[TMP3:%.*]] = call <4 x i64> @foo_vector_fixed4_nomask(<4 x i64> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <4 x i64> [[TMP3]], ptr [[TMP5]], align 8
+; CHECK-NEXT:    store <4 x i64> [[TMP3]], ptr [[TMP4]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/tail-fold-uniform-memops.ll b/llvm/test/Transforms/LoopVectorize/AArch64/tail-fold-uniform-memops.ll
index f6443b1e..e55e322 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/tail-fold-uniform-memops.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/tail-fold-uniform-memops.ll
@@ -22,8 +22,7 @@ define void @uniform_load(ptr noalias %dst, ptr noalias readonly %src, i64 %n) #
 ; CHECK-NEXT:    [[TMP4:%.*]] = insertelement <4 x i32> poison, i32 [[LOAD_VAL]], i64 0
 ; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <4 x i32> [[TMP4]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr %dst, i64 [[IDX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[TMP5]], ptr [[TMP7]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[TMP5]], ptr [[TMP6]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[IDX_NEXT]] = add i64 [[IDX]], 4
 ; CHECK-NEXT:    [[NEXT_ACTIVE_LANE_MASK]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i64(i64 [[IDX]], i64 [[N2]])
 ; CHECK-NEXT:    [[NOT_ACTIVE_LANE_MASK:%.*]] = xor <4 x i1> [[NEXT_ACTIVE_LANE_MASK]], splat (i1 true)
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/tail-folding-styles.ll b/llvm/test/Transforms/LoopVectorize/AArch64/tail-folding-styles.ll
index 0326fe5..7308129 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/tail-folding-styles.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/tail-folding-styles.ll
@@ -30,8 +30,7 @@ define void @simple_memset_tailfold(i32 %val, ptr %ptr, i64 %n) "target-features
 ; NONE:       vector.body:
 ; NONE-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
 ; NONE-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[INDEX1]]
-; NONE-NEXT:    [[TMP6:%.*]] = getelementptr i32, ptr [[TMP5]], i32 0
-; NONE-NEXT:    store <vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP6]], align 4
+; NONE-NEXT:    store <vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP5]], align 4
 ; NONE-NEXT:    [[INDEX_NEXT2]] = add nuw i64 [[INDEX1]], [[TMP8]]
 ; NONE-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT2]], [[N_VEC]]
 ; NONE-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -71,8 +70,7 @@ define void @simple_memset_tailfold(i32 %val, ptr %ptr, i64 %n) "target-features
 ; DATA-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
 ; DATA-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX1]], i64 [[UMAX]])
 ; DATA-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[INDEX1]]
-; DATA-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[TMP10]], i32 0
-; DATA-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
+; DATA-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP10]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
 ; DATA-NEXT:    [[INDEX_NEXT2]] = add i64 [[INDEX1]], [[TMP13]]
 ; DATA-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT2]], [[N_VEC]]
 ; DATA-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -119,8 +117,7 @@ define void @simple_memset_tailfold(i32 %val, ptr %ptr, i64 %n) "target-features
 ; DATA_NO_LANEMASK-NEXT:    [[VEC_IV:%.*]] = add <vscale x 4 x i64> [[BROADCAST_SPLAT3]], [[TMP11]]
 ; DATA_NO_LANEMASK-NEXT:    [[TMP12:%.*]] = icmp ule <vscale x 4 x i64> [[VEC_IV]], [[BROADCAST_SPLAT]]
 ; DATA_NO_LANEMASK-NEXT:    [[TMP13:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[INDEX1]]
-; DATA_NO_LANEMASK-NEXT:    [[TMP14:%.*]] = getelementptr i32, ptr [[TMP13]], i32 0
-; DATA_NO_LANEMASK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT5]], ptr [[TMP14]], i32 4, <vscale x 4 x i1> [[TMP12]])
+; DATA_NO_LANEMASK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT5]], ptr [[TMP13]], i32 4, <vscale x 4 x i1> [[TMP12]])
 ; DATA_NO_LANEMASK-NEXT:    [[INDEX_NEXT6]] = add i64 [[INDEX1]], [[TMP16]]
 ; DATA_NO_LANEMASK-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT6]], [[N_VEC]]
 ; DATA_NO_LANEMASK-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -160,8 +157,7 @@ define void @simple_memset_tailfold(i32 %val, ptr %ptr, i64 %n) "target-features
 ; DATA_AND_CONTROL-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
 ; DATA_AND_CONTROL-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; DATA_AND_CONTROL-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[INDEX1]]
-; DATA_AND_CONTROL-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[TMP10]], i32 0
-; DATA_AND_CONTROL-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
+; DATA_AND_CONTROL-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP10]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
 ; DATA_AND_CONTROL-NEXT:    [[INDEX_NEXT2]] = add i64 [[INDEX1]], [[TMP13]]
 ; DATA_AND_CONTROL-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX_NEXT2]], i64 [[UMAX]])
 ; DATA_AND_CONTROL-NEXT:    [[TMP14:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
@@ -208,8 +204,7 @@ define void @simple_memset_tailfold(i32 %val, ptr %ptr, i64 %n) "target-features
 ; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
 ; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = phi <vscale x 4 x i1> [ [[ACTIVE_LANE_MASK_ENTRY]], [[VECTOR_PH]] ], [ [[ACTIVE_LANE_MASK_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[INDEX1]]
-; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr [[TMP11]], i32 0
-; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP12]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
+; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[BROADCAST_SPLAT]], ptr [[TMP11]], i32 4, <vscale x 4 x i1> [[ACTIVE_LANE_MASK]])
 ; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[INDEX_NEXT2]] = add i64 [[INDEX1]], [[TMP14]]
 ; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[ACTIVE_LANE_MASK_NEXT]] = call <vscale x 4 x i1> @llvm.get.active.lane.mask.nxv4i1.i64(i64 [[INDEX1]], i64 [[TMP9]])
 ; DATA_AND_CONTROL_NO_RT_CHECK-NEXT:    [[TMP15:%.*]] = xor <vscale x 4 x i1> [[ACTIVE_LANE_MASK_NEXT]], splat (i1 true)
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-constant-ops.ll b/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-constant-ops.ll
index 94f46bf..3215356b 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-constant-ops.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-constant-ops.ll
@@ -28,7 +28,7 @@ define void @test_add_double_same_const_args_1(ptr %res, ptr noalias %A, ptr noa
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br [[EXIT:label %.*]]
 ; CHECK:       [[SCALAR_PH]]:
 ;
 entry:
@@ -78,7 +78,7 @@ define void @test_add_double_same_const_args_2(ptr %res, ptr noalias %A, ptr noa
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br [[EXIT:label %.*]]
 ; CHECK:       [[SCALAR_PH]]:
 ;
 entry:
@@ -138,7 +138,7 @@ define void @test_add_double_mixed_const_args(ptr %res, ptr noalias %A, ptr noal
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br [[EXIT:label %.*]]
 ; CHECK:       [[SCALAR_PH]]:
 ;
 entry:
@@ -200,7 +200,7 @@ define void @test_add_double_same_var_args_1(ptr %res, ptr noalias %A, ptr noali
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br [[EXIT:label %.*]]
 ; CHECK:       [[SCALAR_PH]]:
 ;
 entry:
@@ -262,7 +262,7 @@ define void @test_add_double_same_var_args_2(ptr %res, ptr noalias %A, ptr noali
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br [[EXIT:label %.*]]
 ; CHECK:       [[SCALAR_PH]]:
 ;
 entry:
@@ -324,7 +324,7 @@ define void @test_add_double_same_var_args_at_different_positions(ptr %res, ptr
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br [[EXIT:label %.*]]
 ; CHECK:       [[SCALAR_PH]]:
 ;
 entry:
@@ -388,7 +388,7 @@ define void @test_add_double_different_var_args_1(ptr %res, ptr noalias %A, ptr
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br [[EXIT:label %.*]]
 ; CHECK:       [[SCALAR_PH]]:
 ;
 entry:
@@ -452,7 +452,7 @@ define void @test_add_double_different_var_args_2(ptr %res, ptr noalias %A, ptr
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br [[EXIT:label %.*]]
 ; CHECK:       [[SCALAR_PH]]:
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-cost.ll b/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-cost.ll
index ccfa725..ac39ecff 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-cost.ll
@@ -180,11 +180,10 @@ define void @test_interleave_store_one_constant(ptr noalias %src, ptr noalias %d
 ; CHECK-NEXT:    [[TMP11:%.*]] = add i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP12:%.*]] = add i64 [[INDEX]], 6
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr double, ptr [[SRC]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr double, ptr [[TMP13]], i32 0
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr double, ptr [[TMP13]], i32 2
 ; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr double, ptr [[TMP13]], i32 4
 ; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr double, ptr [[TMP13]], i32 6
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP14]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP13]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <2 x double>, ptr [[TMP15]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <2 x double>, ptr [[TMP16]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <2 x double>, ptr [[TMP17]], align 8
@@ -226,8 +225,7 @@ define void @test_interleave_store_one_constant(ptr noalias %src, ptr noalias %d
 ; CHECK:       [[VEC_EPILOG_VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX14:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], %[[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT17:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP31:%.*]] = getelementptr double, ptr [[SRC]], i64 [[INDEX14]]
-; CHECK-NEXT:    [[TMP32:%.*]] = getelementptr double, ptr [[TMP31]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD15:%.*]] = load <2 x double>, ptr [[TMP32]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD15:%.*]] = load <2 x double>, ptr [[TMP31]], align 8
 ; CHECK-NEXT:    [[TMP33:%.*]] = fmul <2 x double> [[WIDE_LOAD15]], splat (double 5.000000e+00)
 ; CHECK-NEXT:    [[TMP34:%.*]] = getelementptr [2 x double], ptr [[DST]], i64 [[INDEX14]]
 ; CHECK-NEXT:    [[TMP35:%.*]] = shufflevector <2 x double> [[TMP33]], <2 x double> zeroinitializer, <4 x i32> <i32 0, i32 1, i32 2, i32 3>
@@ -332,11 +330,10 @@ define void @single_fmul_used_by_each_member(ptr noalias %A, ptr noalias %B, ptr
 ; CHECK-NEXT:    [[TMP21:%.*]] = add i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP22:%.*]] = add i64 [[INDEX]], 6
 ; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr double, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr double, ptr [[TMP23]], i32 0
 ; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr double, ptr [[TMP23]], i32 2
 ; CHECK-NEXT:    [[TMP26:%.*]] = getelementptr double, ptr [[TMP23]], i32 4
 ; CHECK-NEXT:    [[TMP27:%.*]] = getelementptr double, ptr [[TMP23]], i32 6
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP24]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP23]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD12:%.*]] = load <2 x double>, ptr [[TMP25]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD13:%.*]] = load <2 x double>, ptr [[TMP26]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD14:%.*]] = load <2 x double>, ptr [[TMP27]], align 8
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-derived-ivs.ll b/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-derived-ivs.ll
index 3cde3f3..8f241c9 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-derived-ivs.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-derived-ivs.ll
@@ -325,7 +325,7 @@ define void @narrow_with_uniform_add_and_gep(ptr noalias %p) {
 ; VF2-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 512
 ; VF2-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF2IC2-LABEL: define void @narrow_with_uniform_add_and_gep(
@@ -352,7 +352,7 @@ define void @narrow_with_uniform_add_and_gep(ptr noalias %p) {
 ; VF2IC2-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 512
 ; VF2IC2-NEXT:    br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; VF2IC2:       [[MIDDLE_BLOCK]]:
-; VF2IC2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2IC2-NEXT:    br [[EXIT:label %.*]]
 ; VF2IC2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @narrow_with_uniform_add_and_gep(
@@ -378,7 +378,7 @@ define void @narrow_with_uniform_add_and_gep(ptr noalias %p) {
 ; VF4-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 512
 ; VF4-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-metadata.ll b/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-metadata.ll
index 0a83ff6..0244414 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-metadata.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-metadata.ll
@@ -23,7 +23,7 @@ define void @load_store_interleave_group_with_metadata(ptr noalias %data) {
 ; VF2-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP2]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 entry:
@@ -54,11 +54,6 @@ exit:
 !4 = !{ i64 0, i64 2 }
 !5 = !{ i64 0, i64 2 }
 
-;.
-; VF2: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
-; VF2: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
-; VF2: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
-;.
 ; VF4: [[TBAA0]] = !{[[META1:![0-9]+]], [[META1]], i64 0}
 ; VF4: [[META1]] = !{!"omnipotent char", [[META2:![0-9]+]], i64 0}
 ; VF4: [[META2]] = !{!"Simple C/C++ TBAA"}
@@ -66,3 +61,7 @@ exit:
 ; VF4: [[META4]] = !{!"llvm.loop.isvectorized", i32 1}
 ; VF4: [[META5]] = !{!"llvm.loop.unroll.runtime.disable"}
 ;.
+; VF2: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; VF2: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; VF2: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+;.
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-remove-loop-region.ll b/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-remove-loop-region.ll
index fea57fa..d0ea828 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-remove-loop-region.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-remove-loop-region.ll
@@ -21,9 +21,9 @@ define void @load_store_interleave_group_tc_2(ptr noalias %data) {
 ; VF2-NEXT:    store <4 x i64> [[INTERLEAVED_VEC]], ptr [[DATA]], align 8
 ; VF2-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br label %[[EXIT:.*]]
 ; VF2:       [[SCALAR_PH]]:
-; VF2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 2, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; VF2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; VF2-NEXT:    br label %[[LOOP:.*]]
 ; VF2:       [[LOOP]]:
 ; VF2-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -251,9 +251,9 @@ define void @test_complex_add_float_tc_4(ptr %res, ptr noalias %A, ptr noalias %
 ; VF2-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4
 ; VF2-NEXT:    br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br label %[[EXIT:.*]]
 ; VF2:       [[SCALAR_PH]]:
-; VF2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; VF2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; VF2-NEXT:    br label %[[LOOP:.*]]
 ; VF2:       [[LOOP]]:
 ; VF2-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -297,9 +297,9 @@ define void @test_complex_add_float_tc_4(ptr %res, ptr noalias %A, ptr noalias %
 ; VF4-NEXT:    store <8 x float> [[INTERLEAVED_VEC]], ptr [[RES]], align 4
 ; VF4-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br label %[[EXIT:.*]]
 ; VF4:       [[SCALAR_PH]]:
-; VF4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; VF4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; VF4-NEXT:    br label %[[LOOP:.*]]
 ; VF4:       [[LOOP]]:
 ; VF4-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-unroll.ll b/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-unroll.ll
index 3a7b448..4df02a7 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-unroll.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-unroll.ll
@@ -26,9 +26,9 @@ define void @load_store_interleave_group(ptr noalias %data) {
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -78,9 +78,8 @@ define void @test_2xi64_with_wide_load(ptr noalias %data, ptr noalias %factor) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[FACTOR]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 2
-; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = load <2 x i64>, ptr [[TMP1]], align 8
 ; CHECK-NEXT:    [[BROADCAST_SPLAT3:%.*]] = load <2 x i64>, ptr [[TMP3]], align 8
 ; CHECK-NEXT:    [[TMP6:%.*]] = shl nsw i64 [[INDEX]], 1
 ; CHECK-NEXT:    [[TMP7:%.*]] = shl nsw i64 [[TMP0]], 1
@@ -106,9 +105,9 @@ define void @test_2xi64_with_wide_load(ptr noalias %data, ptr noalias %factor) {
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-with-wide-ops.ll b/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-with-wide-ops.ll
index 38e224f..e04b550 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-with-wide-ops.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory-with-wide-ops.ll
@@ -23,7 +23,7 @@ define void @test_2xi64_unary_op_load_interleave_group(ptr noalias %data, ptr no
 ; VF2-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @test_2xi64_unary_op_load_interleave_group(
@@ -48,7 +48,7 @@ define void @test_2xi64_unary_op_load_interleave_group(ptr noalias %data, ptr no
 ; VF4-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
@@ -85,13 +85,11 @@ define void @test_2xi64_unary_op_wide_load(ptr noalias %data, ptr noalias %A, pt
 ; VF2-NEXT:    [[TMP14:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF2-NEXT:    [[TMP0:%.*]] = shl nsw i64 [[TMP14]], 1
 ; VF2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[TMP14]]
-; VF2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds double, ptr [[TMP6]], i32 0
-; VF2-NEXT:    [[WIDE_LOAD3:%.*]] = load <2 x double>, ptr [[TMP7]], align 8
+; VF2-NEXT:    [[WIDE_LOAD3:%.*]] = load <2 x double>, ptr [[TMP6]], align 8
 ; VF2-NEXT:    [[TMP15:%.*]] = fneg <2 x double> [[WIDE_LOAD3]]
 ; VF2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds double, ptr [[DATA]], i64 [[TMP0]]
 ; VF2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[TMP14]]
-; VF2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds double, ptr [[TMP5]], i32 0
-; VF2-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x double>, ptr [[TMP8]], align 8
+; VF2-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x double>, ptr [[TMP5]], align 8
 ; VF2-NEXT:    [[TMP9:%.*]] = fneg <2 x double> [[WIDE_LOAD1]]
 ; VF2-NEXT:    [[TMP17:%.*]] = shufflevector <2 x double> [[TMP15]], <2 x double> [[TMP9]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; VF2-NEXT:    [[INTERLEAVED_VEC4:%.*]] = shufflevector <4 x double> [[TMP17]], <4 x double> poison, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
@@ -100,7 +98,7 @@ define void @test_2xi64_unary_op_wide_load(ptr noalias %data, ptr noalias %A, pt
 ; VF2-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @test_2xi64_unary_op_wide_load(
@@ -113,13 +111,11 @@ define void @test_2xi64_unary_op_wide_load(ptr noalias %data, ptr noalias %A, pt
 ; VF4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF4-NEXT:    [[TMP1:%.*]] = shl nsw i64 [[INDEX]], 1
 ; VF4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds double, ptr [[TMP2]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP3]], align 8
+; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP2]], align 8
 ; VF4-NEXT:    [[TMP4:%.*]] = fneg <4 x double> [[WIDE_LOAD]]
 ; VF4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds double, ptr [[DATA]], i64 [[TMP1]]
 ; VF4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds double, ptr [[TMP6]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x double>, ptr [[TMP7]], align 8
+; VF4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x double>, ptr [[TMP6]], align 8
 ; VF4-NEXT:    [[TMP8:%.*]] = fneg <4 x double> [[WIDE_LOAD1]]
 ; VF4-NEXT:    [[TMP9:%.*]] = shufflevector <4 x double> [[TMP4]], <4 x double> [[TMP8]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; VF4-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <8 x double> [[TMP9]], <8 x double> poison, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
@@ -128,7 +124,7 @@ define void @test_2xi64_unary_op_wide_load(ptr noalias %data, ptr noalias %A, pt
 ; VF4-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
@@ -178,7 +174,7 @@ define void @test_2xi64(ptr noalias %data, ptr noalias %factor) {
 ; VF2-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @test_2xi64(
@@ -190,8 +186,7 @@ define void @test_2xi64(ptr noalias %data, ptr noalias %factor) {
 ; VF4:       [[VECTOR_BODY]]:
 ; VF4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[FACTOR]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; VF4-NEXT:    [[TMP10:%.*]] = shl nsw i64 [[INDEX]], 1
 ; VF4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP10]]
 ; VF4-NEXT:    [[WIDE_VEC:%.*]] = load <8 x i64>, ptr [[TMP11]], align 8
@@ -206,7 +201,7 @@ define void @test_2xi64(ptr noalias %data, ptr noalias %factor) {
 ; VF4-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
@@ -244,8 +239,7 @@ define void @test_2xi64_different_opcodes(ptr noalias %data, ptr noalias %factor
 ; VF2:       [[VECTOR_BODY]]:
 ; VF2-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[FACTOR]], i64 [[TMP0]]
-; VF2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8
+; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 8
 ; VF2-NEXT:    [[TMP3:%.*]] = shl nsw i64 [[TMP0]], 1
 ; VF2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP3]]
 ; VF2-NEXT:    [[WIDE_VEC:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
@@ -260,7 +254,7 @@ define void @test_2xi64_different_opcodes(ptr noalias %data, ptr noalias %factor
 ; VF2-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @test_2xi64_different_opcodes(
@@ -272,8 +266,7 @@ define void @test_2xi64_different_opcodes(ptr noalias %data, ptr noalias %factor
 ; VF4:       [[VECTOR_BODY]]:
 ; VF4-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[FACTOR]], i64 [[TMP0]]
-; VF4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; VF4-NEXT:    [[TMP3:%.*]] = shl nsw i64 [[TMP0]], 1
 ; VF4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP3]]
 ; VF4-NEXT:    [[WIDE_VEC:%.*]] = load <8 x i64>, ptr [[TMP4]], align 8
@@ -288,7 +281,7 @@ define void @test_2xi64_different_opcodes(ptr noalias %data, ptr noalias %factor
 ; VF4-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
@@ -326,8 +319,7 @@ define void @test_2xi64_interleave_loads_order_flipped(ptr noalias %data, ptr no
 ; VF2:       [[VECTOR_BODY]]:
 ; VF2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[FACTOR]], i64 [[INDEX]]
-; VF2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8
+; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 8
 ; VF2-NEXT:    [[TMP15:%.*]] = shl nsw i64 [[INDEX]], 1
 ; VF2-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP15]]
 ; VF2-NEXT:    [[WIDE_VEC:%.*]] = load <4 x i64>, ptr [[TMP16]], align 8
@@ -342,7 +334,7 @@ define void @test_2xi64_interleave_loads_order_flipped(ptr noalias %data, ptr no
 ; VF2-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @test_2xi64_interleave_loads_order_flipped(
@@ -354,8 +346,7 @@ define void @test_2xi64_interleave_loads_order_flipped(ptr noalias %data, ptr no
 ; VF4:       [[VECTOR_BODY]]:
 ; VF4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[FACTOR]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; VF4-NEXT:    [[TMP27:%.*]] = shl nsw i64 [[INDEX]], 1
 ; VF4-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP27]]
 ; VF4-NEXT:    [[WIDE_VEC:%.*]] = load <8 x i64>, ptr [[TMP28]], align 8
@@ -370,7 +361,7 @@ define void @test_2xi64_interleave_loads_order_flipped(ptr noalias %data, ptr no
 ; VF4-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
@@ -408,8 +399,7 @@ define void @test_2xi64_store_order_flipped_1(ptr noalias %data, ptr noalias %fa
 ; VF2:       [[VECTOR_BODY]]:
 ; VF2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[FACTOR]], i64 [[INDEX]]
-; VF2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8
+; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 8
 ; VF2-NEXT:    [[TMP6:%.*]] = shl nsw i64 [[INDEX]], 1
 ; VF2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP6]]
 ; VF2-NEXT:    [[WIDE_VEC:%.*]] = load <4 x i64>, ptr [[TMP7]], align 8
@@ -424,7 +414,7 @@ define void @test_2xi64_store_order_flipped_1(ptr noalias %data, ptr noalias %fa
 ; VF2-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @test_2xi64_store_order_flipped_1(
@@ -436,8 +426,7 @@ define void @test_2xi64_store_order_flipped_1(ptr noalias %data, ptr noalias %fa
 ; VF4:       [[VECTOR_BODY]]:
 ; VF4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[FACTOR]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; VF4-NEXT:    [[TMP10:%.*]] = shl nsw i64 [[INDEX]], 1
 ; VF4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP10]]
 ; VF4-NEXT:    [[WIDE_VEC:%.*]] = load <8 x i64>, ptr [[TMP11]], align 8
@@ -452,7 +441,7 @@ define void @test_2xi64_store_order_flipped_1(ptr noalias %data, ptr noalias %fa
 ; VF4-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
@@ -490,8 +479,7 @@ define void @test_2xi64_store_order_flipped_2(ptr noalias %data, ptr noalias %fa
 ; VF2:       [[VECTOR_BODY]]:
 ; VF2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[FACTOR]], i64 [[INDEX]]
-; VF2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8
+; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 8
 ; VF2-NEXT:    [[TMP6:%.*]] = shl nsw i64 [[INDEX]], 1
 ; VF2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP6]]
 ; VF2-NEXT:    [[WIDE_VEC:%.*]] = load <4 x i64>, ptr [[TMP7]], align 8
@@ -506,7 +494,7 @@ define void @test_2xi64_store_order_flipped_2(ptr noalias %data, ptr noalias %fa
 ; VF2-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @test_2xi64_store_order_flipped_2(
@@ -518,8 +506,7 @@ define void @test_2xi64_store_order_flipped_2(ptr noalias %data, ptr noalias %fa
 ; VF4:       [[VECTOR_BODY]]:
 ; VF4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[FACTOR]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; VF4-NEXT:    [[TMP10:%.*]] = shl nsw i64 [[INDEX]], 1
 ; VF4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP10]]
 ; VF4-NEXT:    [[WIDE_VEC:%.*]] = load <8 x i64>, ptr [[TMP11]], align 8
@@ -534,7 +521,7 @@ define void @test_2xi64_store_order_flipped_2(ptr noalias %data, ptr noalias %fa
 ; VF4-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
@@ -572,26 +559,24 @@ define void @test_2xi64_different_loads_feeding_fmul(ptr noalias %data, ptr noal
 ; VF2:       [[VECTOR_BODY]]:
 ; VF2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[SRC_0]], i64 [[INDEX]]
-; VF2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8
+; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 8
 ; VF2-NEXT:    [[TMP6:%.*]] = shl nsw i64 [[INDEX]], 1
 ; VF2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP6]]
-; VF2-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8
+; VF2-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i64>, ptr [[TMP1]], align 8
 ; VF2-NEXT:    [[TMP8:%.*]] = mul <2 x i64> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; VF2-NEXT:    [[TMP14:%.*]] = or disjoint i64 [[TMP6]], 1
 ; VF2-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP14]]
 ; VF2-NEXT:    [[WIDE_VEC:%.*]] = load <4 x i64>, ptr [[TMP15]], align 8
 ; VF2-NEXT:    [[TMP19:%.*]] = shufflevector <4 x i64> [[WIDE_VEC]], <4 x i64> poison, <2 x i32> <i32 0, i32 2>
 ; VF2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i64, ptr [[SRC_1]], i64 [[INDEX]]
-; VF2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i64, ptr [[TMP20]], i32 0
-; VF2-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i64>, ptr [[TMP21]], align 8
+; VF2-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i64>, ptr [[TMP20]], align 8
 ; VF2-NEXT:    [[TMP22:%.*]] = mul <2 x i64> [[WIDE_LOAD2]], [[TMP19]]
 ; VF2-NEXT:    [[TMP13:%.*]] = shufflevector <2 x i64> [[TMP8]], <2 x i64> [[TMP22]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; VF2-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <4 x i64> [[TMP13]], <4 x i64> poison, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
 ; VF2-NEXT:    store <4 x i64> [[INTERLEAVED_VEC]], ptr [[TMP7]], align 8
 ; VF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; VF2-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 98
-; VF2-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; VF2-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 98
+; VF2-NEXT:    br i1 [[TMP11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
 ; VF2-NEXT:    br label %[[SCALAR_PH]]
 ; VF2:       [[SCALAR_PH]]:
@@ -605,19 +590,17 @@ define void @test_2xi64_different_loads_feeding_fmul(ptr noalias %data, ptr noal
 ; VF4:       [[VECTOR_BODY]]:
 ; VF4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[SRC_0]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; VF4-NEXT:    [[TMP10:%.*]] = shl nsw i64 [[INDEX]], 1
 ; VF4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP10]]
-; VF4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; VF4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; VF4-NEXT:    [[TMP12:%.*]] = mul <4 x i64> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; VF4-NEXT:    [[TMP24:%.*]] = or disjoint i64 [[TMP10]], 1
 ; VF4-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP24]]
 ; VF4-NEXT:    [[WIDE_VEC:%.*]] = load <8 x i64>, ptr [[TMP25]], align 8
 ; VF4-NEXT:    [[TMP33:%.*]] = shufflevector <8 x i64> [[WIDE_VEC]], <8 x i64> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
 ; VF4-NEXT:    [[TMP34:%.*]] = getelementptr inbounds i64, ptr [[SRC_1]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP35:%.*]] = getelementptr inbounds i64, ptr [[TMP34]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i64>, ptr [[TMP35]], align 8
+; VF4-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i64>, ptr [[TMP34]], align 8
 ; VF4-NEXT:    [[TMP36:%.*]] = mul <4 x i64> [[WIDE_LOAD2]], [[TMP33]]
 ; VF4-NEXT:    [[TMP13:%.*]] = shufflevector <4 x i64> [[TMP12]], <4 x i64> [[TMP36]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; VF4-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <8 x i64> [[TMP13]], <8 x i64> poison, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
@@ -666,8 +649,7 @@ define void @test_3xi64(ptr noalias %data, ptr noalias %factor) {
 ; VF2:       [[VECTOR_BODY]]:
 ; VF2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[FACTOR]], i64 [[INDEX]]
-; VF2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
-; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP3]], align 8
+; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8
 ; VF2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds { i64, i64, i64 }, ptr [[DATA]], i64 [[INDEX]], i32 0
 ; VF2-NEXT:    [[WIDE_VEC:%.*]] = load <6 x i64>, ptr [[TMP4]], align 8
 ; VF2-NEXT:    [[TMP9:%.*]] = shufflevector <6 x i64> [[WIDE_VEC]], <6 x i64> poison, <2 x i32> <i32 0, i32 3>
@@ -685,7 +667,7 @@ define void @test_3xi64(ptr noalias %data, ptr noalias %factor) {
 ; VF2-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @test_3xi64(
@@ -697,8 +679,7 @@ define void @test_3xi64(ptr noalias %data, ptr noalias %factor) {
 ; VF4:       [[VECTOR_BODY]]:
 ; VF4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[FACTOR]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP5]], align 8
+; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
 ; VF4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds { i64, i64, i64 }, ptr [[DATA]], i64 [[INDEX]], i32 0
 ; VF4-NEXT:    [[WIDE_VEC:%.*]] = load <12 x i64>, ptr [[TMP6]], align 8
 ; VF4-NEXT:    [[TMP17:%.*]] = shufflevector <12 x i64> [[WIDE_VEC]], <12 x i64> poison, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
@@ -716,7 +697,7 @@ define void @test_3xi64(ptr noalias %data, ptr noalias %factor) {
 ; VF4-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
@@ -973,7 +954,7 @@ define void @test_2xi64_sub_of_wide_loads(ptr noalias %data, ptr noalias %A, ptr
 ; VF2-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], 100
 ; VF2-NEXT:    br i1 [[EC]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @test_2xi64_sub_of_wide_loads(
@@ -985,11 +966,9 @@ define void @test_2xi64_sub_of_wide_loads(ptr noalias %data, ptr noalias %A, ptr
 ; VF4:       [[VECTOR_BODY]]:
 ; VF4-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP0]]
-; VF4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; VF4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP0]]
-; VF4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP3]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
+; VF4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; VF4-NEXT:    [[TMP5:%.*]] = sub <4 x i64> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; VF4-NEXT:    [[TMP6:%.*]] = shl nsw i64 [[TMP0]], 1
 ; VF4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP6]]
@@ -1001,7 +980,7 @@ define void @test_2xi64_sub_of_wide_loads(ptr noalias %data, ptr noalias %A, ptr
 ; VF4-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
@@ -1039,11 +1018,9 @@ define void @test_2xi64_sub_of_wide_loads_ops_swapped(ptr noalias %data, ptr noa
 ; VF2:       [[VECTOR_BODY]]:
 ; VF2-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[IV_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP0]]
-; VF2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP7]], i32 0
-; VF2-NEXT:    [[BROADCAST_SPLAT6:%.*]] = load <2 x i64>, ptr [[TMP8]], align 8
+; VF2-NEXT:    [[BROADCAST_SPLAT6:%.*]] = load <2 x i64>, ptr [[TMP7]], align 8
 ; VF2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP0]]
-; VF2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
-; VF2-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i64>, ptr [[TMP3]], align 8
+; VF2-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8
 ; VF2-NEXT:    [[TMP13:%.*]] = sub <2 x i64> [[BROADCAST_SPLAT6]], [[WIDE_LOAD1]]
 ; VF2-NEXT:    [[TMP20:%.*]] = shl nsw i64 [[TMP0]], 1
 ; VF2-NEXT:    [[DATA_1:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP20]]
@@ -1055,7 +1032,7 @@ define void @test_2xi64_sub_of_wide_loads_ops_swapped(ptr noalias %data, ptr noa
 ; VF2-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], 100
 ; VF2-NEXT:    br i1 [[EC]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @test_2xi64_sub_of_wide_loads_ops_swapped(
@@ -1067,11 +1044,9 @@ define void @test_2xi64_sub_of_wide_loads_ops_swapped(ptr noalias %data, ptr noa
 ; VF4:       [[VECTOR_BODY]]:
 ; VF4-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP0]]
-; VF4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; VF4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP0]]
-; VF4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP3]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
+; VF4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; VF4-NEXT:    [[TMP5:%.*]] = sub <4 x i64> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; VF4-NEXT:    [[TMP6:%.*]] = shl nsw i64 [[TMP0]], 1
 ; VF4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP6]]
@@ -1083,7 +1058,7 @@ define void @test_2xi64_sub_of_wide_loads_ops_swapped(ptr noalias %data, ptr noa
 ; VF4-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
@@ -1121,17 +1096,14 @@ define void @test_2xi64_sub_of_wide_loads_with_different_base_ptrs(ptr noalias %
 ; VF2:       [[VECTOR_BODY]]:
 ; VF2-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP0]]
-; VF2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8
+; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 8
 ; VF2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP0]]
-; VF2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP3]], i32 0
-; VF2-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i64>, ptr [[TMP4]], align 8
+; VF2-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i64>, ptr [[TMP3]], align 8
 ; VF2-NEXT:    [[TMP5:%.*]] = sub <2 x i64> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; VF2-NEXT:    [[TMP6:%.*]] = shl nsw i64 [[TMP0]], 1
 ; VF2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP6]]
 ; VF2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[C]], i64 [[TMP0]]
-; VF2-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP8]], i32 0
-; VF2-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i64>, ptr [[TMP9]], align 8
+; VF2-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i64>, ptr [[TMP8]], align 8
 ; VF2-NEXT:    [[TMP10:%.*]] = sub <2 x i64> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; VF2-NEXT:    [[TMP11:%.*]] = shufflevector <2 x i64> [[TMP5]], <2 x i64> [[TMP10]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; VF2-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <4 x i64> [[TMP11]], <4 x i64> poison, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
@@ -1140,7 +1112,7 @@ define void @test_2xi64_sub_of_wide_loads_with_different_base_ptrs(ptr noalias %
 ; VF2-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @test_2xi64_sub_of_wide_loads_with_different_base_ptrs(
@@ -1152,17 +1124,14 @@ define void @test_2xi64_sub_of_wide_loads_with_different_base_ptrs(ptr noalias %
 ; VF4:       [[VECTOR_BODY]]:
 ; VF4-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP0]]
-; VF4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; VF4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP0]]
-; VF4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP3]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
+; VF4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; VF4-NEXT:    [[TMP5:%.*]] = sub <4 x i64> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; VF4-NEXT:    [[TMP6:%.*]] = shl nsw i64 [[TMP0]], 1
 ; VF4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP6]]
 ; VF4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[C]], i64 [[TMP0]]
-; VF4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP8]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i64>, ptr [[TMP9]], align 8
+; VF4-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i64>, ptr [[TMP8]], align 8
 ; VF4-NEXT:    [[TMP10:%.*]] = sub <4 x i64> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; VF4-NEXT:    [[TMP11:%.*]] = shufflevector <4 x i64> [[TMP5]], <4 x i64> [[TMP10]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; VF4-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <8 x i64> [[TMP11]], <8 x i64> poison, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
@@ -1171,7 +1140,7 @@ define void @test_2xi64_sub_of_wide_loads_with_different_base_ptrs(ptr noalias %
 ; VF4-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory.ll b/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory.ll
index 6acd798..41b89dc 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/transform-narrow-interleave-to-widen-memory.ll
@@ -22,7 +22,7 @@ define void @load_store_interleave_group(ptr noalias %data) {
 ; VF2-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @load_store_interleave_group(
@@ -45,7 +45,7 @@ define void @load_store_interleave_group(ptr noalias %data) {
 ; VF4-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
@@ -87,7 +87,7 @@ define void @load_store_interleave_group_different_objecs(ptr noalias %src, ptr
 ; VF2-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @load_store_interleave_group_different_objecs(
@@ -111,7 +111,7 @@ define void @load_store_interleave_group_different_objecs(ptr noalias %src, ptr
 ; VF4-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
@@ -148,8 +148,7 @@ define void @single_wide_load_store_interleave_group(ptr noalias %src, ptr noali
 ; VF2-NEXT:    [[TMP7:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF2-NEXT:    [[TMP9:%.*]] = shl nsw i64 [[TMP7]], 1
 ; VF2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[SRC]], i64 [[TMP7]]
-; VF2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
-; VF2-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i64>, ptr [[TMP3]], align 8
+; VF2-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8
 ; VF2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[DST]], i64 [[TMP9]]
 ; VF2-NEXT:    [[TMP10:%.*]] = shufflevector <2 x i64> [[WIDE_LOAD1]], <2 x i64> [[WIDE_LOAD1]], <4 x i32> <i32 0, i32 1, i32 2, i32 3>
 ; VF2-NEXT:    [[INTERLEAVED_VEC2:%.*]] = shufflevector <4 x i64> [[TMP10]], <4 x i64> poison, <4 x i32> <i32 0, i32 2, i32 1, i32 3>
@@ -158,7 +157,7 @@ define void @single_wide_load_store_interleave_group(ptr noalias %src, ptr noali
 ; VF2-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @single_wide_load_store_interleave_group(
@@ -171,8 +170,7 @@ define void @single_wide_load_store_interleave_group(ptr noalias %src, ptr noali
 ; VF4-NEXT:    [[TMP7:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF4-NEXT:    [[TMP9:%.*]] = shl nsw i64 [[TMP7]], 1
 ; VF4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[SRC]], i64 [[TMP7]]
-; VF4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
+; VF4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
 ; VF4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[DST]], i64 [[TMP9]]
 ; VF4-NEXT:    [[TMP10:%.*]] = shufflevector <4 x i64> [[WIDE_LOAD1]], <4 x i64> [[WIDE_LOAD1]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; VF4-NEXT:    [[INTERLEAVED_VEC2:%.*]] = shufflevector <8 x i64> [[TMP10]], <8 x i64> poison, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
@@ -181,7 +179,7 @@ define void @single_wide_load_store_interleave_group(ptr noalias %src, ptr noali
 ; VF4-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
@@ -221,7 +219,7 @@ define void @same_constant_store_interleave_group(i64 %x, ptr noalias %dst) {
 ; VF2-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP2]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @same_constant_store_interleave_group(
@@ -239,7 +237,7 @@ define void @same_constant_store_interleave_group(i64 %x, ptr noalias %dst) {
 ; VF4-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP2]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
@@ -277,7 +275,7 @@ define void @different_constants_store_interleave_group(i64 %x, i64 %y, ptr noal
 ; VF2-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP2]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @different_constants_store_interleave_group(
@@ -295,7 +293,7 @@ define void @different_constants_store_interleave_group(i64 %x, i64 %y, ptr noal
 ; VF4-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP2]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
@@ -337,7 +335,7 @@ define void @same_live_in_store_interleave_group(i64 %x, ptr noalias %dst) {
 ; VF2-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @same_live_in_store_interleave_group(
@@ -359,7 +357,7 @@ define void @same_live_in_store_interleave_group(i64 %x, ptr noalias %dst) {
 ; VF4-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
@@ -403,7 +401,7 @@ define void @different_live_ins_store_interleave_group(i64 %x, i64 %y, ptr noali
 ; VF2-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @different_live_ins_store_interleave_group(
@@ -427,7 +425,7 @@ define void @different_live_ins_store_interleave_group(i64 %x, i64 %y, ptr noali
 ; VF4-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
@@ -468,7 +466,7 @@ define void @single_uniform_load_store_interleave_group(ptr noalias %src, ptr no
 ; VF2-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @single_uniform_load_store_interleave_group(
@@ -491,7 +489,7 @@ define void @single_uniform_load_store_interleave_group(ptr noalias %src, ptr no
 ; VF4-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
@@ -538,7 +536,7 @@ define void @multiple_uniform_load_store_interleave_group(ptr noalias %src.0, pt
 ; VF2-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF2-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; VF2:       [[MIDDLE_BLOCK]]:
-; VF2-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF2-NEXT:    br [[EXIT:label %.*]]
 ; VF2:       [[SCALAR_PH]]:
 ;
 ; VF4-LABEL: define void @multiple_uniform_load_store_interleave_group(
@@ -564,7 +562,7 @@ define void @multiple_uniform_load_store_interleave_group(ptr noalias %src.0, pt
 ; VF4-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; VF4-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; VF4:       [[MIDDLE_BLOCK]]:
-; VF4-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; VF4-NEXT:    br [[EXIT:label %.*]]
 ; VF4:       [[SCALAR_PH]]:
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/type-shrinkage-insertelt.ll b/llvm/test/Transforms/LoopVectorize/AArch64/type-shrinkage-insertelt.ll
index c017c1f..a2cbf6f 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/type-shrinkage-insertelt.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/type-shrinkage-insertelt.ll
@@ -18,40 +18,39 @@ define void @test0(ptr noalias %M3, ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 3
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP5]], align 2
-; CHECK-NEXT:    [[TMP6:%.*]] = add <4 x i16> [[WIDE_LOAD]], splat (i16 10)
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP2]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP3]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP4]], align 2
+; CHECK-NEXT:    [[TMP5:%.*]] = add <4 x i16> [[WIDE_LOAD]], splat (i16 10)
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP0]]
+; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
+; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP2]]
+; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP3]]
+; CHECK-NEXT:    [[TMP10:%.*]] = load i64, ptr [[TMP6]], align 8
 ; CHECK-NEXT:    [[TMP11:%.*]] = load i64, ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[TMP12:%.*]] = load i64, ptr [[TMP8]], align 8
 ; CHECK-NEXT:    [[TMP13:%.*]] = load i64, ptr [[TMP9]], align 8
-; CHECK-NEXT:    [[TMP14:%.*]] = load i64, ptr [[TMP10]], align 8
+; CHECK-NEXT:    [[TMP14:%.*]] = ashr exact i64 [[TMP10]], 32
 ; CHECK-NEXT:    [[TMP15:%.*]] = ashr exact i64 [[TMP11]], 32
 ; CHECK-NEXT:    [[TMP16:%.*]] = ashr exact i64 [[TMP12]], 32
 ; CHECK-NEXT:    [[TMP17:%.*]] = ashr exact i64 [[TMP13]], 32
-; CHECK-NEXT:    [[TMP18:%.*]] = ashr exact i64 [[TMP14]], 32
+; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i16, ptr [[M3]], i64 [[TMP14]]
 ; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i16, ptr [[M3]], i64 [[TMP15]]
 ; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i16, ptr [[M3]], i64 [[TMP16]]
 ; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i16, ptr [[M3]], i64 [[TMP17]]
-; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i16, ptr [[M3]], i64 [[TMP18]]
-; CHECK-NEXT:    [[TMP23:%.*]] = extractelement <4 x i16> [[TMP6]], i32 0
+; CHECK-NEXT:    [[TMP22:%.*]] = extractelement <4 x i16> [[TMP5]], i32 0
+; CHECK-NEXT:    store i16 [[TMP22]], ptr [[TMP18]], align 2
+; CHECK-NEXT:    [[TMP23:%.*]] = extractelement <4 x i16> [[TMP5]], i32 1
 ; CHECK-NEXT:    store i16 [[TMP23]], ptr [[TMP19]], align 2
-; CHECK-NEXT:    [[TMP24:%.*]] = extractelement <4 x i16> [[TMP6]], i32 1
+; CHECK-NEXT:    [[TMP24:%.*]] = extractelement <4 x i16> [[TMP5]], i32 2
 ; CHECK-NEXT:    store i16 [[TMP24]], ptr [[TMP20]], align 2
-; CHECK-NEXT:    [[TMP25:%.*]] = extractelement <4 x i16> [[TMP6]], i32 2
+; CHECK-NEXT:    [[TMP25:%.*]] = extractelement <4 x i16> [[TMP5]], i32 3
 ; CHECK-NEXT:    store i16 [[TMP25]], ptr [[TMP21]], align 2
-; CHECK-NEXT:    [[TMP26:%.*]] = extractelement <4 x i16> [[TMP6]], i32 3
-; CHECK-NEXT:    store i16 [[TMP26]], ptr [[TMP22]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP27:%.*]] = icmp eq i64 [[INDEX_NEXT]], 16
-; CHECK-NEXT:    br i1 [[TMP27]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], 16
+; CHECK-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_INC1286_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_INC1286_LOOPEXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 16, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[IF_THEN1165_US:%.*]]
 ; CHECK:       if.then1165.us:
 ; CHECK-NEXT:    [[INDVARS_IV1783:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT1784:%.*]], [[IF_THEN1165_US]] ]
@@ -111,41 +110,40 @@ define void @test1(ptr noalias %M3, ptr noalias %A, ptr noalias %B, ptr noalias
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TMP4]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP6]], align 2
-; CHECK-NEXT:    [[TMP7:%.*]] = trunc <4 x i32> [[BROADCAST_SPLAT]] to <4 x i16>
-; CHECK-NEXT:    [[TMP8:%.*]] = add <4 x i16> [[WIDE_LOAD]], [[TMP7]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP2]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP3]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP5]], align 2
+; CHECK-NEXT:    [[TMP6:%.*]] = trunc <4 x i32> [[BROADCAST_SPLAT]] to <4 x i16>
+; CHECK-NEXT:    [[TMP7:%.*]] = add <4 x i16> [[WIDE_LOAD]], [[TMP6]]
+; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP0]]
+; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
+; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP2]]
+; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP3]]
+; CHECK-NEXT:    [[TMP12:%.*]] = load i64, ptr [[TMP8]], align 8
 ; CHECK-NEXT:    [[TMP13:%.*]] = load i64, ptr [[TMP9]], align 8
 ; CHECK-NEXT:    [[TMP14:%.*]] = load i64, ptr [[TMP10]], align 8
 ; CHECK-NEXT:    [[TMP15:%.*]] = load i64, ptr [[TMP11]], align 8
-; CHECK-NEXT:    [[TMP16:%.*]] = load i64, ptr [[TMP12]], align 8
+; CHECK-NEXT:    [[TMP16:%.*]] = ashr exact i64 [[TMP12]], 32
 ; CHECK-NEXT:    [[TMP17:%.*]] = ashr exact i64 [[TMP13]], 32
 ; CHECK-NEXT:    [[TMP18:%.*]] = ashr exact i64 [[TMP14]], 32
 ; CHECK-NEXT:    [[TMP19:%.*]] = ashr exact i64 [[TMP15]], 32
-; CHECK-NEXT:    [[TMP20:%.*]] = ashr exact i64 [[TMP16]], 32
+; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i16, ptr [[M3]], i64 [[TMP16]]
 ; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i16, ptr [[M3]], i64 [[TMP17]]
 ; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i16, ptr [[M3]], i64 [[TMP18]]
 ; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i16, ptr [[M3]], i64 [[TMP19]]
-; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i16, ptr [[M3]], i64 [[TMP20]]
-; CHECK-NEXT:    [[TMP25:%.*]] = extractelement <4 x i16> [[TMP8]], i32 0
+; CHECK-NEXT:    [[TMP24:%.*]] = extractelement <4 x i16> [[TMP7]], i32 0
+; CHECK-NEXT:    store i16 [[TMP24]], ptr [[TMP20]], align 2
+; CHECK-NEXT:    [[TMP25:%.*]] = extractelement <4 x i16> [[TMP7]], i32 1
 ; CHECK-NEXT:    store i16 [[TMP25]], ptr [[TMP21]], align 2
-; CHECK-NEXT:    [[TMP26:%.*]] = extractelement <4 x i16> [[TMP8]], i32 1
+; CHECK-NEXT:    [[TMP26:%.*]] = extractelement <4 x i16> [[TMP7]], i32 2
 ; CHECK-NEXT:    store i16 [[TMP26]], ptr [[TMP22]], align 2
-; CHECK-NEXT:    [[TMP27:%.*]] = extractelement <4 x i16> [[TMP8]], i32 2
+; CHECK-NEXT:    [[TMP27:%.*]] = extractelement <4 x i16> [[TMP7]], i32 3
 ; CHECK-NEXT:    store i16 [[TMP27]], ptr [[TMP23]], align 2
-; CHECK-NEXT:    [[TMP28:%.*]] = extractelement <4 x i16> [[TMP8]], i32 3
-; CHECK-NEXT:    store i16 [[TMP28]], ptr [[TMP24]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP29:%.*]] = icmp eq i64 [[INDEX_NEXT]], 16
-; CHECK-NEXT:    br i1 [[TMP29]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK-NEXT:    [[TMP28:%.*]] = icmp eq i64 [[INDEX_NEXT]], 16
+; CHECK-NEXT:    br i1 [[TMP28]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_INC1286_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_INC1286_LOOPEXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 16, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[IF_THEN1165_US:%.*]]
 ; CHECK:       if.then1165.us:
 ; CHECK-NEXT:    [[INDVARS_IV1783:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT1784:%.*]], [[IF_THEN1165_US]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/vector-loop-backedge-elimination-epilogue.ll b/llvm/test/Transforms/LoopVectorize/AArch64/vector-loop-backedge-elimination-epilogue.ll
index 2705d69..a431fdd 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/vector-loop-backedge-elimination-epilogue.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/vector-loop-backedge-elimination-epilogue.ll
@@ -17,11 +17,10 @@ define void @test_remove_vector_loop_region_epilogue(ptr %dst, i1 %c)  {
 ; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TC]], [[N_MOD_VF]]
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[DST]], i32 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[DST]], i32 16
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[DST]], i32 32
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[DST]], i32 48
-; CHECK-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP1]], align 4
+; CHECK-NEXT:    store <16 x i8> zeroinitializer, ptr [[DST]], align 4
 ; CHECK-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP3]], align 4
 ; CHECK-NEXT:    store <16 x i8> zeroinitializer, ptr [[TMP4]], align 4
@@ -40,8 +39,7 @@ define void @test_remove_vector_loop_region_epilogue(ptr %dst, i1 %c)  {
 ; CHECK-NEXT:    br label %[[VEC_EPILOG_VECTOR_BODY:.*]]
 ; CHECK:       [[VEC_EPILOG_VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[DST]], i64 [[VEC_EPILOG_RESUME_VAL]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <8 x i8> zeroinitializer, ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <8 x i8> zeroinitializer, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    br label %[[VEC_EPILOG_MIDDLE_BLOCK:.*]]
 ; CHECK:       [[VEC_EPILOG_MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[CMP_N4:%.*]] = icmp eq i64 [[TC]], [[N_VEC3]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/vplan-printing.ll b/llvm/test/Transforms/LoopVectorize/AArch64/vplan-printing.ll
index 399d676..bba9293 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/vplan-printing.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/vplan-printing.ll
@@ -90,12 +90,10 @@ define i32 @print_partial_reduction(ptr %a, ptr %b) {
 ; CHECK-NEXT:   EMIT-SCALAR vp<[[EP_IV:%.+]]> = phi [ ir<0>, ir-bb<vector.ph> ], [ vp<%index.next>, vector.body ]
 ; CHECK-NEXT:   WIDEN-REDUCTION-PHI ir<%accum> = phi vp<[[RDX_START]]>, ir<%add> (VF scaled by 1/4)
 ; CHECK-NEXT:   CLONE ir<%gep.a> = getelementptr ir<%a>, vp<[[EP_IV]]>
-; CHECK-NEXT:   vp<[[PTR_A:%.+]]> = vector-pointer ir<%gep.a>
-; CHECK-NEXT:   WIDEN ir<%load.a> = load vp<[[PTR_A]]>
+; CHECK-NEXT:   WIDEN ir<%load.a> = load ir<%gep.a>
 ; CHECK-NEXT:   WIDEN-CAST ir<%ext.a> = zext ir<%load.a> to i32
 ; CHECK-NEXT:   CLONE ir<%gep.b> = getelementptr ir<%b>, vp<[[EP_IV]]>
-; CHECK-NEXT:   vp<[[PTR_B:%.+]]> = vector-pointer ir<%gep.b>
-; CHECK-NEXT:   WIDEN ir<%load.b> = load vp<[[PTR_B]]>
+; CHECK-NEXT:   WIDEN ir<%load.b> = load ir<%gep.b>
 ; CHECK-NEXT:   WIDEN-CAST ir<%ext.b> = zext ir<%load.b> to i32
 ; CHECK-NEXT:   WIDEN ir<%mul> = mul ir<%ext.b>, ir<%ext.a>
 ; CHECK-NEXT:   PARTIAL-REDUCE ir<%add> = add ir<%accum>, ir<%mul>
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/widen-call-with-intrinsic-or-libfunc.ll b/llvm/test/Transforms/LoopVectorize/AArch64/widen-call-with-intrinsic-or-libfunc.ll
index 1f2e918..d014468 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/widen-call-with-intrinsic-or-libfunc.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/widen-call-with-intrinsic-or-libfunc.ll
@@ -109,8 +109,7 @@ define void @test(ptr noalias %src, ptr noalias %dst) {
 ; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
 ; CHECK-NEXT:    [[TMP1:%.*]] = add i64 [[INDEX]], 1
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[SRC:%.*]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP3]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = fpext <2 x float> [[WIDE_LOAD]] to <2 x double>
 ; CHECK-NEXT:    [[TMP5:%.*]] = call fast <2 x double> @__simd_sin_v2f64(<2 x double> [[TMP4]])
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[DST:%.*]], i64 [[TMP0]]
diff --git a/llvm/test/Transforms/LoopVectorize/AArch64/wider-VF-for-callinst.ll b/llvm/test/Transforms/LoopVectorize/AArch64/wider-VF-for-callinst.ll
index 3545c6b..c1d4317 100644
--- a/llvm/test/Transforms/LoopVectorize/AArch64/wider-VF-for-callinst.ll
+++ b/llvm/test/Transforms/LoopVectorize/AArch64/wider-VF-for-callinst.ll
@@ -68,7 +68,7 @@ define void @test_widen(ptr noalias %a, ptr readnone %b) #1 {
 ; NARROW-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; NARROW-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; NARROW:       middle.block:
-; NARROW-NEXT:    br i1 false, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; NARROW-NEXT:    br label [[SCALAR_PH]]
 ; NARROW:       scalar.ph:
 ; NARROW-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; NARROW-NEXT:    br label [[FOR_BODY:%.*]]
@@ -82,7 +82,7 @@ define void @test_widen(ptr noalias %a, ptr readnone %b) #1 {
 ; NARROW-NEXT:    store float [[CALL]], ptr [[ARRAYIDX]], align 4
 ; NARROW-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; NARROW-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 1025
-; NARROW-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
+; NARROW-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; NARROW:       for.cond.cleanup:
 ; NARROW-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/AMDGPU/packed-math.ll b/llvm/test/Transforms/LoopVectorize/AMDGPU/packed-math.ll
index ab7bb66..8aab77e 100644
--- a/llvm/test/Transforms/LoopVectorize/AMDGPU/packed-math.ll
+++ b/llvm/test/Transforms/LoopVectorize/AMDGPU/packed-math.ll
@@ -25,7 +25,7 @@ define half @vectorize_v2f16_loop(ptr addrspace(1) noalias %s) {
 ; GFX9:       middle.block:
 ; GFX9-NEXT:    [[BIN_RDX:%.*]] = fadd fast <2 x half> [[TMP3]], [[TMP2]]
 ; GFX9-NEXT:    [[TMP5:%.*]] = call fast half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> [[BIN_RDX]])
-; GFX9-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; GFX9-NEXT:    br label [[FOR_END:%.*]]
 ; GFX9:       scalar.ph:
 ; GFX9-NEXT:    br label [[FOR_BODY:%.*]]
 ; GFX9:       for.body:
@@ -55,7 +55,7 @@ define half @vectorize_v2f16_loop(ptr addrspace(1) noalias %s) {
 ; VI:       middle.block:
 ; VI-NEXT:    [[BIN_RDX:%.*]] = fadd fast <2 x half> [[TMP3]], [[TMP2]]
 ; VI-NEXT:    [[TMP5:%.*]] = call fast half @llvm.vector.reduce.fadd.v2f16(half 0xH0000, <2 x half> [[BIN_RDX]])
-; VI-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; VI-NEXT:    br label [[FOR_END:%.*]]
 ; VI:       scalar.ph:
 ; VI-NEXT:    br label [[FOR_BODY:%.*]]
 ; VI:       for.body:
diff --git a/llvm/test/Transforms/LoopVectorize/ARM/mve-gather-scatter-tailpred.ll b/llvm/test/Transforms/LoopVectorize/ARM/mve-gather-scatter-tailpred.ll
index 9522c7e..66bb80b 100644
--- a/llvm/test/Transforms/LoopVectorize/ARM/mve-gather-scatter-tailpred.ll
+++ b/llvm/test/Transforms/LoopVectorize/ARM/mve-gather-scatter-tailpred.ll
@@ -17,12 +17,10 @@ define void @test_stride1_4i32(ptr readonly %data, ptr noalias nocapture %dst, i
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 [[INDEX]], i32 [[N]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = add nuw nsw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[DATA:%.*]], i32 [[TMP1]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP3]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP2]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> poison)
 ; CHECK-NEXT:    [[TMP4:%.*]] = add nsw <4 x i32> splat (i32 5), [[WIDE_MASKED_LOAD]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[TMP4]], ptr [[TMP6]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[TMP4]], ptr [[TMP5]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -84,8 +82,7 @@ define void @test_stride-1_4i32(ptr readonly %data, ptr noalias nocapture %dst,
 ; CHECK-NEXT:    [[REVERSE:%.*]] = shufflevector <4 x i32> [[WIDE_LOAD]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; CHECK-NEXT:    [[TMP6:%.*]] = add nsw <4 x i32> splat (i32 5), [[REVERSE]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP6]], ptr [[TMP8]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP6]], ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -149,8 +146,7 @@ define void @test_stride2_4i32(ptr readonly %data, ptr noalias nocapture %dst, i
 ; CHECK-NEXT:    [[STRIDED_VEC:%.*]] = shufflevector <8 x i32> [[WIDE_VEC]], <8 x i32> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
 ; CHECK-NEXT:    [[TMP7:%.*]] = add nsw <4 x i32> splat (i32 5), [[STRIDED_VEC]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP7]], ptr [[TMP9]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP7]], ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
@@ -211,8 +207,7 @@ define void @test_stride3_4i32(ptr readonly %data, ptr noalias nocapture %dst, i
 ; CHECK-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0(<4 x ptr> [[TMP3]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> poison)
 ; CHECK-NEXT:    [[TMP4:%.*]] = add nsw <4 x i32> splat (i32 5), [[WIDE_MASKED_GATHER]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[TMP4]], ptr [[TMP6]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[TMP4]], ptr [[TMP5]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
@@ -274,8 +269,7 @@ define void @test_stride4_4i32(ptr readonly %data, ptr noalias nocapture %dst, i
 ; CHECK-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0(<4 x ptr> [[TMP3]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> poison)
 ; CHECK-NEXT:    [[TMP4:%.*]] = add nsw <4 x i32> splat (i32 5), [[WIDE_MASKED_GATHER]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[TMP4]], ptr [[TMP6]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[TMP4]], ptr [[TMP5]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
@@ -335,12 +329,10 @@ define void @test_stride_loopinvar_4i32(ptr readonly %data, ptr noalias nocaptur
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 [[INDEX]], i32 [[N]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = add nuw nsw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[DATA:%.*]], i32 [[TMP1]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP3]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP2]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> poison)
 ; CHECK-NEXT:    [[TMP4:%.*]] = add nsw <4 x i32> splat (i32 5), [[WIDE_MASKED_LOAD]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[TMP4]], ptr [[TMP6]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]])
+; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[TMP4]], ptr [[TMP5]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
@@ -403,8 +395,7 @@ define void @test_stride_noninvar_4i32(ptr readonly %data, ptr noalias nocapture
 ; CHECK-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0(<4 x ptr> [[TMP4]], i32 4, <4 x i1> splat (i1 true), <4 x i32> poison)
 ; CHECK-NEXT:    [[TMP5:%.*]] = add nsw <4 x i32> splat (i32 5), [[WIDE_MASKED_GATHER]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP7]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
 ; CHECK-NEXT:    [[VEC_IND_NEXT3]] = add <4 x i32> [[VEC_IND2]], splat (i32 32)
@@ -523,8 +514,7 @@ define void @test_stride_noninvar3_4i32(ptr readonly %data, ptr noalias nocaptur
 ; CHECK-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0(<4 x ptr> [[TMP6]], i32 4, <4 x i1> splat (i1 true), <4 x i32> poison)
 ; CHECK-NEXT:    [[TMP7:%.*]] = add nsw <4 x i32> splat (i32 5), [[WIDE_MASKED_GATHER]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP7]], ptr [[TMP9]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP7]], ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
 ; CHECK-NEXT:    [[VEC_IND_NEXT5]] = add <4 x i32> [[VEC_IND4]], [[DOTSPLAT3]]
diff --git a/llvm/test/Transforms/LoopVectorize/ARM/mve-hoist-runtime-checks.ll b/llvm/test/Transforms/LoopVectorize/ARM/mve-hoist-runtime-checks.ll
index 68f7e53..029bffd 100644
--- a/llvm/test/Transforms/LoopVectorize/ARM/mve-hoist-runtime-checks.ll
+++ b/llvm/test/Transforms/LoopVectorize/ARM/mve-hoist-runtime-checks.ll
@@ -70,14 +70,12 @@ define void @diff_checks(ptr nocapture noundef writeonly %dst, ptr nocapture nou
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 [[INDEX]], i32 [[N]])
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[TMP7]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP12]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> poison), !alias.scope [[META0:![0-9]+]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr i32, ptr [[TMP8]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr i32, ptr [[TMP13]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[WIDE_MASKED_LOAD]], ptr [[TMP14]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]]), !alias.scope [[META3:![0-9]+]], !noalias [[META0]]
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP11]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> poison), !alias.scope [[META0:![0-9]+]]
+; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr [[TMP8]], i32 [[INDEX]]
+; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[WIDE_MASKED_LOAD]], ptr [[TMP12]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]]), !alias.scope [[META3:![0-9]+]], !noalias [[META0]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i32 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    br label [[INNER_LOOP_EXIT]]
 ; CHECK:       scalar.ph:
@@ -86,9 +84,9 @@ define void @diff_checks(ptr nocapture noundef writeonly %dst, ptr nocapture nou
 ; CHECK:       inner.loop:
 ; CHECK-NEXT:    [[J_021_US:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC_US:%.*]], [[INNER_LOOP]] ]
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr i32, ptr [[TMP7]], i32 [[J_021_US]]
-; CHECK-NEXT:    [[TMP16:%.*]] = load i32, ptr [[ARRAYIDX_US]], align 4
+; CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[ARRAYIDX_US]], align 4
 ; CHECK-NEXT:    [[ARRAYIDX8_US:%.*]] = getelementptr i32, ptr [[TMP8]], i32 [[J_021_US]]
-; CHECK-NEXT:    store i32 [[TMP16]], ptr [[ARRAYIDX8_US]], align 4
+; CHECK-NEXT:    store i32 [[TMP14]], ptr [[ARRAYIDX8_US]], align 4
 ; CHECK-NEXT:    [[INC_US]] = add nuw nsw i32 [[J_021_US]], 1
 ; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i32 [[INC_US]], [[N]]
 ; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[INNER_LOOP_EXIT]], label [[INNER_LOOP]], !llvm.loop [[LOOP8:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/ARM/mve-multiexit.ll b/llvm/test/Transforms/LoopVectorize/ARM/mve-multiexit.ll
index 951e3c1..4fc2778 100644
--- a/llvm/test/Transforms/LoopVectorize/ARM/mve-multiexit.ll
+++ b/llvm/test/Transforms/LoopVectorize/ARM/mve-multiexit.ll
@@ -30,11 +30,9 @@ define void @multiple_exits_unique_exit_block(ptr %A, ptr %B, i32 %N) #0 {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[B]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[WIDE_LOAD]], ptr [[TMP8]], align 4
+; CHECK-NEXT:    store <4 x i32> [[WIDE_LOAD]], ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -103,11 +101,9 @@ define i32 @multiple_exits_multiple_exit_blocks(ptr %A, ptr %B, i32 %N) #0 {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[B]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[WIDE_LOAD]], ptr [[TMP8]], align 4
+; CHECK-NEXT:    store <4 x i32> [[WIDE_LOAD]], ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/ARM/mve-reduction-predselect.ll b/llvm/test/Transforms/LoopVectorize/ARM/mve-reduction-predselect.ll
index 6d42dee..bc02595 100644
--- a/llvm/test/Transforms/LoopVectorize/ARM/mve-reduction-predselect.ll
+++ b/llvm/test/Transforms/LoopVectorize/ARM/mve-reduction-predselect.ll
@@ -436,7 +436,7 @@ define i32 @reduction_min(ptr nocapture %A, ptr nocapture %B) {
 ; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.smin.v4i32(<4 x i32> [[TMP1]])
-; CHECK-NEXT:    br i1 false, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP3]], [[MIDDLE_BLOCK]] ], [ poison, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
@@ -448,10 +448,9 @@ define i32 @reduction_min(ptr nocapture %A, ptr nocapture %B) {
 ; CHECK-NEXT:    [[V0]] = call i32 @llvm.smin.i32(i32 [[RESULT_08]], i32 [[L0]])
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add i32 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], 257
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
 ; CHECK:       for.end:
-; CHECK-NEXT:    [[RESULT_0_LCSSA:%.*]] = phi i32 [ [[V0]], [[FOR_BODY]] ], [ poison, [[MIDDLE_BLOCK]] ]
-; CHECK-NEXT:    ret i32 [[RESULT_0_LCSSA]]
+; CHECK-NEXT:    ret i32 [[V0]]
 ;
 entry:
   br label %for.body
@@ -489,7 +488,7 @@ define i32 @reduction_max(ptr nocapture %A, ptr nocapture %B) {
 ; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.umax.v4i32(<4 x i32> [[TMP1]])
-; CHECK-NEXT:    br i1 false, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP3]], [[MIDDLE_BLOCK]] ], [ poison, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
@@ -501,10 +500,9 @@ define i32 @reduction_max(ptr nocapture %A, ptr nocapture %B) {
 ; CHECK-NEXT:    [[V0]] = call i32 @llvm.umax.i32(i32 [[RESULT_08]], i32 [[L0]])
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add i32 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INDVARS_IV_NEXT]], 257
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_END:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
 ; CHECK:       for.end:
-; CHECK-NEXT:    [[RESULT_0_LCSSA:%.*]] = phi i32 [ [[V0]], [[FOR_BODY]] ], [ poison, [[MIDDLE_BLOCK]] ]
-; CHECK-NEXT:    ret i32 [[RESULT_0_LCSSA]]
+; CHECK-NEXT:    ret i32 [[V0]]
 ;
 entry:
   br label %for.body
diff --git a/llvm/test/Transforms/LoopVectorize/ARM/mve-reduction-types.ll b/llvm/test/Transforms/LoopVectorize/ARM/mve-reduction-types.ll
index 18d607f..f1bee3b 100644
--- a/llvm/test/Transforms/LoopVectorize/ARM/mve-reduction-types.ll
+++ b/llvm/test/Transforms/LoopVectorize/ARM/mve-reduction-types.ll
@@ -21,11 +21,9 @@ define i32 @mla_i32(ptr noalias nocapture readonly %A, ptr noalias nocapture rea
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <16 x i1> @llvm.get.active.lane.mask.v16i1.i32(i32 [[INDEX]], i32 [[N]])
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <16 x i8> @llvm.masked.load.v16i8.p0(ptr [[TMP1]], i32 1, <16 x i1> [[ACTIVE_LANE_MASK]], <16 x i8> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <16 x i8> @llvm.masked.load.v16i8.p0(ptr [[TMP0]], i32 1, <16 x i1> [[ACTIVE_LANE_MASK]], <16 x i8> poison)
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[B:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <16 x i8> @llvm.masked.load.v16i8.p0(ptr [[TMP3]], i32 1, <16 x i1> [[ACTIVE_LANE_MASK]], <16 x i8> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <16 x i8> @llvm.masked.load.v16i8.p0(ptr [[TMP2]], i32 1, <16 x i1> [[ACTIVE_LANE_MASK]], <16 x i8> poison)
 ; CHECK-NEXT:    [[TMP4:%.*]] = sext <16 x i8> [[WIDE_MASKED_LOAD1]] to <16 x i32>
 ; CHECK-NEXT:    [[TMP5:%.*]] = sext <16 x i8> [[WIDE_MASKED_LOAD]] to <16 x i32>
 ; CHECK-NEXT:    [[TMP6:%.*]] = mul nsw <16 x i32> [[TMP4]], [[TMP5]]
@@ -103,11 +101,9 @@ define i32 @mla_i8(ptr noalias nocapture readonly %A, ptr noalias nocapture read
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <16 x i1> @llvm.get.active.lane.mask.v16i1.i32(i32 [[INDEX]], i32 [[N]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <16 x i8> @llvm.masked.load.v16i8.p0(ptr [[TMP2]], i32 1, <16 x i1> [[ACTIVE_LANE_MASK]], <16 x i8> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <16 x i8> @llvm.masked.load.v16i8.p0(ptr [[TMP1]], i32 1, <16 x i1> [[ACTIVE_LANE_MASK]], <16 x i8> poison)
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[B:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <16 x i8> @llvm.masked.load.v16i8.p0(ptr [[TMP5]], i32 1, <16 x i1> [[ACTIVE_LANE_MASK]], <16 x i8> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <16 x i8> @llvm.masked.load.v16i8.p0(ptr [[TMP4]], i32 1, <16 x i1> [[ACTIVE_LANE_MASK]], <16 x i8> poison)
 ; CHECK-NEXT:    [[TMP6:%.*]] = sext <16 x i8> [[WIDE_MASKED_LOAD1]] to <16 x i32>
 ; CHECK-NEXT:    [[TMP3:%.*]] = sext <16 x i8> [[WIDE_MASKED_LOAD]] to <16 x i32>
 ; CHECK-NEXT:    [[TMP7:%.*]] = mul nsw <16 x i32> [[TMP6]], [[TMP3]]
@@ -185,8 +181,7 @@ define i32 @add_i32(ptr nocapture readonly %x, i32 %n) #0 {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 [[INDEX]], i32 [[N]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[X:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP2]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP1]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> poison)
 ; CHECK-NEXT:    [[TMP3:%.*]] = select <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> [[WIDE_MASKED_LOAD]], <4 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP3]])
 ; CHECK-NEXT:    [[TMP5]] = add i32 [[TMP4]], [[VEC_PHI]]
@@ -251,8 +246,7 @@ define i32 @mul_i32(ptr nocapture readonly %x, i32 %n) #0 {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 1), [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 [[INDEX]], i32 [[N]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[X:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP2]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP1]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> poison)
 ; CHECK-NEXT:    [[TMP3:%.*]] = mul <4 x i32> [[WIDE_MASKED_LOAD]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP4]] = select <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> [[TMP3]], <4 x i32> [[VEC_PHI]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
@@ -317,8 +311,7 @@ define i32 @and_i32(ptr nocapture readonly %x, i32 %n) #0 {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 -1), [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 [[INDEX]], i32 [[N]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[X:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP2]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP1]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> poison)
 ; CHECK-NEXT:    [[TMP3:%.*]] = and <4 x i32> [[WIDE_MASKED_LOAD]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP4]] = select <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> [[TMP3]], <4 x i32> [[VEC_PHI]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
@@ -383,8 +376,7 @@ define i32 @or_i32(ptr nocapture readonly %x, i32 %n) #0 {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 [[INDEX]], i32 [[N]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[X:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP2]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP1]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> poison)
 ; CHECK-NEXT:    [[TMP3:%.*]] = or <4 x i32> [[WIDE_MASKED_LOAD]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP4]] = select <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> [[TMP3]], <4 x i32> [[VEC_PHI]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
@@ -449,8 +441,7 @@ define i32 @xor_i32(ptr nocapture readonly %x, i32 %n) #0 {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 [[INDEX]], i32 [[N]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[X:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP2]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP1]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> poison)
 ; CHECK-NEXT:    [[TMP3:%.*]] = xor <4 x i32> [[WIDE_MASKED_LOAD]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP4]] = select <4 x i1> [[ACTIVE_LANE_MASK]], <4 x i32> [[TMP3]], <4 x i32> [[VEC_PHI]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
@@ -515,8 +506,7 @@ define float @fadd_f32(ptr nocapture readonly %x, i32 %n) #0 {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 [[INDEX]], i32 [[N]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[X:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP2]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x float> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP1]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x float> poison)
 ; CHECK-NEXT:    [[TMP3:%.*]] = fadd fast <4 x float> [[WIDE_MASKED_LOAD]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP4]] = select fast <4 x i1> [[ACTIVE_LANE_MASK]], <4 x float> [[TMP3]], <4 x float> [[VEC_PHI]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
@@ -581,8 +571,7 @@ define float @fmul_f32(ptr nocapture readonly %x, i32 %n) #0 {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ splat (float 1.000000e+00), [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <4 x i1> @llvm.get.active.lane.mask.v4i1.i32(i32 [[INDEX]], i32 [[N]])
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[X:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP2]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x float> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP1]], i32 4, <4 x i1> [[ACTIVE_LANE_MASK]], <4 x float> poison)
 ; CHECK-NEXT:    [[TMP3:%.*]] = fmul fast <4 x float> [[WIDE_MASKED_LOAD]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP4]] = select fast <4 x i1> [[ACTIVE_LANE_MASK]], <4 x float> [[TMP3]], <4 x float> [[VEC_PHI]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
@@ -646,8 +635,7 @@ define i32 @smin_i32(ptr nocapture readonly %x, i32 %n) #0 {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 2147483647), [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[X:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp slt <4 x i32> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i32> [[VEC_PHI]], <4 x i32> [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
@@ -714,8 +702,7 @@ define i32 @smax_i32(ptr nocapture readonly %x, i32 %n) #0 {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 -2147483648), [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[X:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp sgt <4 x i32> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i32> [[VEC_PHI]], <4 x i32> [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
@@ -782,8 +769,7 @@ define i32 @umin_i32(ptr nocapture readonly %x, i32 %n) #0 {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 -1), [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[X:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp ult <4 x i32> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i32> [[VEC_PHI]], <4 x i32> [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
@@ -850,8 +836,7 @@ define i32 @umax_i32(ptr nocapture readonly %x, i32 %n) #0 {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[X:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp ugt <4 x i32> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i32> [[VEC_PHI]], <4 x i32> [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
diff --git a/llvm/test/Transforms/LoopVectorize/ARM/optsize_minsize.ll b/llvm/test/Transforms/LoopVectorize/ARM/optsize_minsize.ll
index 1d898fb..0f4d40f 100644
--- a/llvm/test/Transforms/LoopVectorize/ARM/optsize_minsize.ll
+++ b/llvm/test/Transforms/LoopVectorize/ARM/optsize_minsize.ll
@@ -24,16 +24,14 @@ define void @always_vectorize(ptr %p, i32 %x) {
 ; DEFAULT-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; DEFAULT-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; DEFAULT:       [[VECTOR_BODY]]:
-; DEFAULT-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[P]], i32 0
-; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[P]], align 4
 ; DEFAULT-NEXT:    [[TMP3:%.*]] = add nsw <4 x i32> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
-; DEFAULT-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[P]], i32 0
-; DEFAULT-NEXT:    store <4 x i32> [[TMP3]], ptr [[TMP5]], align 4
+; DEFAULT-NEXT:    store <4 x i32> [[TMP3]], ptr [[P]], align 4
 ; DEFAULT-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; DEFAULT:       [[MIDDLE_BLOCK]]:
-; DEFAULT-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
+; DEFAULT-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
 ; DEFAULT:       [[SCALAR_PH]]:
-; DEFAULT-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; DEFAULT-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; DEFAULT-NEXT:    br label %[[FOR_BODY:.*]]
 ; DEFAULT:       [[FOR_BODY]]:
 ; DEFAULT-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -56,16 +54,14 @@ define void @always_vectorize(ptr %p, i32 %x) {
 ; OPTSIZE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; OPTSIZE-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; OPTSIZE:       [[VECTOR_BODY]]:
-; OPTSIZE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[P]], i32 0
-; OPTSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; OPTSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[P]], align 4
 ; OPTSIZE-NEXT:    [[TMP3:%.*]] = add nsw <4 x i32> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
-; OPTSIZE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[P]], i32 0
-; OPTSIZE-NEXT:    store <4 x i32> [[TMP3]], ptr [[TMP5]], align 4
+; OPTSIZE-NEXT:    store <4 x i32> [[TMP3]], ptr [[P]], align 4
 ; OPTSIZE-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; OPTSIZE:       [[MIDDLE_BLOCK]]:
-; OPTSIZE-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
+; OPTSIZE-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
 ; OPTSIZE:       [[SCALAR_PH]]:
-; OPTSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; OPTSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; OPTSIZE-NEXT:    br label %[[FOR_BODY:.*]]
 ; OPTSIZE:       [[FOR_BODY]]:
 ; OPTSIZE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -88,16 +84,14 @@ define void @always_vectorize(ptr %p, i32 %x) {
 ; MINSIZE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; MINSIZE-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; MINSIZE:       [[VECTOR_BODY]]:
-; MINSIZE-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[P]], i32 0
-; MINSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
+; MINSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[P]], align 4
 ; MINSIZE-NEXT:    [[TMP2:%.*]] = add nsw <4 x i32> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
-; MINSIZE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[P]], i32 0
-; MINSIZE-NEXT:    store <4 x i32> [[TMP2]], ptr [[TMP3]], align 4
+; MINSIZE-NEXT:    store <4 x i32> [[TMP2]], ptr [[P]], align 4
 ; MINSIZE-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; MINSIZE:       [[MIDDLE_BLOCK]]:
-; MINSIZE-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
+; MINSIZE-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
 ; MINSIZE:       [[SCALAR_PH]]:
-; MINSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; MINSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; MINSIZE-NEXT:    br label %[[FOR_BODY:.*]]
 ; MINSIZE:       [[FOR_BODY]]:
 ; MINSIZE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -145,10 +139,9 @@ define void @vectorize_without_optsize(ptr %p, i32 %x, i64 %n) {
 ; DEFAULT:       [[VECTOR_BODY]]:
 ; DEFAULT-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; DEFAULT-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[P]], i64 [[TMP0]]
-; DEFAULT-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; DEFAULT-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; DEFAULT-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; DEFAULT-NEXT:    [[TMP5:%.*]] = add nsw <4 x i32> [[WIDE_LOAD1]], [[BROADCAST_SPLAT]]
-; DEFAULT-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP2]], align 4
+; DEFAULT-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP1]], align 4
 ; DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 4
 ; DEFAULT-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; DEFAULT-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
@@ -235,9 +228,9 @@ define void @tail_predicate_without_optsize(ptr %p, i8 %a, i8 %b, i8 %c, i32 %n)
 ; DEFAULT-NEXT:    [[BROADCAST_SPLAT6:%.*]] = shufflevector <16 x i8> [[BROADCAST_SPLATINSERT5]], <16 x i8> poison, <16 x i32> zeroinitializer
 ; DEFAULT-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; DEFAULT:       [[VECTOR_BODY]]:
-; DEFAULT-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE36:.*]] ]
-; DEFAULT-NEXT:    [[VEC_IND:%.*]] = phi <16 x i8> [ <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[PRED_STORE_CONTINUE36]] ]
-; DEFAULT-NEXT:    [[VEC_IND1:%.*]] = phi <16 x i8> [ <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT2:%.*]], %[[PRED_STORE_CONTINUE36]] ]
+; DEFAULT-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE35:.*]] ]
+; DEFAULT-NEXT:    [[VEC_IND:%.*]] = phi <16 x i8> [ <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[PRED_STORE_CONTINUE35]] ]
+; DEFAULT-NEXT:    [[VEC_IND1:%.*]] = phi <16 x i8> [ <i8 0, i8 1, i8 2, i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT2:%.*]], %[[PRED_STORE_CONTINUE35]] ]
 ; DEFAULT-NEXT:    [[TMP0:%.*]] = icmp ule <16 x i8> [[VEC_IND]], splat (i8 14)
 ; DEFAULT-NEXT:    [[TMP1:%.*]] = mul <16 x i8> [[BROADCAST_SPLAT]], [[VEC_IND1]]
 ; DEFAULT-NEXT:    [[TMP2:%.*]] = lshr <16 x i8> [[VEC_IND1]], splat (i8 1)
@@ -256,140 +249,140 @@ define void @tail_predicate_without_optsize(ptr %p, i8 %a, i8 %b, i8 %c, i32 %n)
 ; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE]]
 ; DEFAULT:       [[PRED_STORE_CONTINUE]]:
 ; DEFAULT-NEXT:    [[TMP12:%.*]] = extractelement <16 x i1> [[TMP0]], i32 1
-; DEFAULT-NEXT:    br i1 [[TMP12]], label %[[PRED_STORE_IF7:.*]], label %[[PRED_STORE_CONTINUE8:.*]]
-; DEFAULT:       [[PRED_STORE_IF7]]:
+; DEFAULT-NEXT:    br i1 [[TMP12]], label %[[PRED_STORE_IF6:.*]], label %[[PRED_STORE_CONTINUE7:.*]]
+; DEFAULT:       [[PRED_STORE_IF6]]:
 ; DEFAULT-NEXT:    [[TMP13:%.*]] = add i64 [[INDEX]], 1
 ; DEFAULT-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP13]]
 ; DEFAULT-NEXT:    [[TMP15:%.*]] = extractelement <16 x i8> [[TMP7]], i32 1
 ; DEFAULT-NEXT:    store i8 [[TMP15]], ptr [[TMP14]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE8]]
-; DEFAULT:       [[PRED_STORE_CONTINUE8]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE7]]
+; DEFAULT:       [[PRED_STORE_CONTINUE7]]:
 ; DEFAULT-NEXT:    [[TMP16:%.*]] = extractelement <16 x i1> [[TMP0]], i32 2
-; DEFAULT-NEXT:    br i1 [[TMP16]], label %[[PRED_STORE_IF9:.*]], label %[[PRED_STORE_CONTINUE10:.*]]
-; DEFAULT:       [[PRED_STORE_IF9]]:
+; DEFAULT-NEXT:    br i1 [[TMP16]], label %[[PRED_STORE_IF8:.*]], label %[[PRED_STORE_CONTINUE9:.*]]
+; DEFAULT:       [[PRED_STORE_IF8]]:
 ; DEFAULT-NEXT:    [[TMP17:%.*]] = add i64 [[INDEX]], 2
 ; DEFAULT-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP17]]
 ; DEFAULT-NEXT:    [[TMP19:%.*]] = extractelement <16 x i8> [[TMP7]], i32 2
 ; DEFAULT-NEXT:    store i8 [[TMP19]], ptr [[TMP18]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE10]]
-; DEFAULT:       [[PRED_STORE_CONTINUE10]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE9]]
+; DEFAULT:       [[PRED_STORE_CONTINUE9]]:
 ; DEFAULT-NEXT:    [[TMP20:%.*]] = extractelement <16 x i1> [[TMP0]], i32 3
-; DEFAULT-NEXT:    br i1 [[TMP20]], label %[[PRED_STORE_IF11:.*]], label %[[PRED_STORE_CONTINUE12:.*]]
-; DEFAULT:       [[PRED_STORE_IF11]]:
+; DEFAULT-NEXT:    br i1 [[TMP20]], label %[[PRED_STORE_IF10:.*]], label %[[PRED_STORE_CONTINUE11:.*]]
+; DEFAULT:       [[PRED_STORE_IF10]]:
 ; DEFAULT-NEXT:    [[TMP21:%.*]] = add i64 [[INDEX]], 3
 ; DEFAULT-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP21]]
 ; DEFAULT-NEXT:    [[TMP23:%.*]] = extractelement <16 x i8> [[TMP7]], i32 3
 ; DEFAULT-NEXT:    store i8 [[TMP23]], ptr [[TMP22]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE12]]
-; DEFAULT:       [[PRED_STORE_CONTINUE12]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE11]]
+; DEFAULT:       [[PRED_STORE_CONTINUE11]]:
 ; DEFAULT-NEXT:    [[TMP24:%.*]] = extractelement <16 x i1> [[TMP0]], i32 4
-; DEFAULT-NEXT:    br i1 [[TMP24]], label %[[PRED_STORE_IF13:.*]], label %[[PRED_STORE_CONTINUE14:.*]]
-; DEFAULT:       [[PRED_STORE_IF13]]:
+; DEFAULT-NEXT:    br i1 [[TMP24]], label %[[PRED_STORE_IF12:.*]], label %[[PRED_STORE_CONTINUE13:.*]]
+; DEFAULT:       [[PRED_STORE_IF12]]:
 ; DEFAULT-NEXT:    [[TMP25:%.*]] = add i64 [[INDEX]], 4
 ; DEFAULT-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP25]]
 ; DEFAULT-NEXT:    [[TMP27:%.*]] = extractelement <16 x i8> [[TMP7]], i32 4
 ; DEFAULT-NEXT:    store i8 [[TMP27]], ptr [[TMP26]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE14]]
-; DEFAULT:       [[PRED_STORE_CONTINUE14]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE13]]
+; DEFAULT:       [[PRED_STORE_CONTINUE13]]:
 ; DEFAULT-NEXT:    [[TMP28:%.*]] = extractelement <16 x i1> [[TMP0]], i32 5
-; DEFAULT-NEXT:    br i1 [[TMP28]], label %[[PRED_STORE_IF15:.*]], label %[[PRED_STORE_CONTINUE16:.*]]
-; DEFAULT:       [[PRED_STORE_IF15]]:
+; DEFAULT-NEXT:    br i1 [[TMP28]], label %[[PRED_STORE_IF14:.*]], label %[[PRED_STORE_CONTINUE15:.*]]
+; DEFAULT:       [[PRED_STORE_IF14]]:
 ; DEFAULT-NEXT:    [[TMP29:%.*]] = add i64 [[INDEX]], 5
 ; DEFAULT-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP29]]
 ; DEFAULT-NEXT:    [[TMP31:%.*]] = extractelement <16 x i8> [[TMP7]], i32 5
 ; DEFAULT-NEXT:    store i8 [[TMP31]], ptr [[TMP30]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE16]]
-; DEFAULT:       [[PRED_STORE_CONTINUE16]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE15]]
+; DEFAULT:       [[PRED_STORE_CONTINUE15]]:
 ; DEFAULT-NEXT:    [[TMP32:%.*]] = extractelement <16 x i1> [[TMP0]], i32 6
-; DEFAULT-NEXT:    br i1 [[TMP32]], label %[[PRED_STORE_IF17:.*]], label %[[PRED_STORE_CONTINUE18:.*]]
-; DEFAULT:       [[PRED_STORE_IF17]]:
+; DEFAULT-NEXT:    br i1 [[TMP32]], label %[[PRED_STORE_IF16:.*]], label %[[PRED_STORE_CONTINUE17:.*]]
+; DEFAULT:       [[PRED_STORE_IF16]]:
 ; DEFAULT-NEXT:    [[TMP33:%.*]] = add i64 [[INDEX]], 6
 ; DEFAULT-NEXT:    [[TMP34:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP33]]
 ; DEFAULT-NEXT:    [[TMP35:%.*]] = extractelement <16 x i8> [[TMP7]], i32 6
 ; DEFAULT-NEXT:    store i8 [[TMP35]], ptr [[TMP34]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE18]]
-; DEFAULT:       [[PRED_STORE_CONTINUE18]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE17]]
+; DEFAULT:       [[PRED_STORE_CONTINUE17]]:
 ; DEFAULT-NEXT:    [[TMP36:%.*]] = extractelement <16 x i1> [[TMP0]], i32 7
-; DEFAULT-NEXT:    br i1 [[TMP36]], label %[[PRED_STORE_IF19:.*]], label %[[PRED_STORE_CONTINUE20:.*]]
-; DEFAULT:       [[PRED_STORE_IF19]]:
+; DEFAULT-NEXT:    br i1 [[TMP36]], label %[[PRED_STORE_IF18:.*]], label %[[PRED_STORE_CONTINUE19:.*]]
+; DEFAULT:       [[PRED_STORE_IF18]]:
 ; DEFAULT-NEXT:    [[TMP37:%.*]] = add i64 [[INDEX]], 7
 ; DEFAULT-NEXT:    [[TMP38:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP37]]
 ; DEFAULT-NEXT:    [[TMP39:%.*]] = extractelement <16 x i8> [[TMP7]], i32 7
 ; DEFAULT-NEXT:    store i8 [[TMP39]], ptr [[TMP38]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE20]]
-; DEFAULT:       [[PRED_STORE_CONTINUE20]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE19]]
+; DEFAULT:       [[PRED_STORE_CONTINUE19]]:
 ; DEFAULT-NEXT:    [[TMP40:%.*]] = extractelement <16 x i1> [[TMP0]], i32 8
-; DEFAULT-NEXT:    br i1 [[TMP40]], label %[[PRED_STORE_IF21:.*]], label %[[PRED_STORE_CONTINUE22:.*]]
-; DEFAULT:       [[PRED_STORE_IF21]]:
+; DEFAULT-NEXT:    br i1 [[TMP40]], label %[[PRED_STORE_IF20:.*]], label %[[PRED_STORE_CONTINUE21:.*]]
+; DEFAULT:       [[PRED_STORE_IF20]]:
 ; DEFAULT-NEXT:    [[TMP41:%.*]] = add i64 [[INDEX]], 8
 ; DEFAULT-NEXT:    [[TMP42:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP41]]
 ; DEFAULT-NEXT:    [[TMP43:%.*]] = extractelement <16 x i8> [[TMP7]], i32 8
 ; DEFAULT-NEXT:    store i8 [[TMP43]], ptr [[TMP42]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE22]]
-; DEFAULT:       [[PRED_STORE_CONTINUE22]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE21]]
+; DEFAULT:       [[PRED_STORE_CONTINUE21]]:
 ; DEFAULT-NEXT:    [[TMP44:%.*]] = extractelement <16 x i1> [[TMP0]], i32 9
-; DEFAULT-NEXT:    br i1 [[TMP44]], label %[[PRED_STORE_IF23:.*]], label %[[PRED_STORE_CONTINUE24:.*]]
-; DEFAULT:       [[PRED_STORE_IF23]]:
+; DEFAULT-NEXT:    br i1 [[TMP44]], label %[[PRED_STORE_IF22:.*]], label %[[PRED_STORE_CONTINUE23:.*]]
+; DEFAULT:       [[PRED_STORE_IF22]]:
 ; DEFAULT-NEXT:    [[TMP45:%.*]] = add i64 [[INDEX]], 9
 ; DEFAULT-NEXT:    [[TMP46:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP45]]
 ; DEFAULT-NEXT:    [[TMP47:%.*]] = extractelement <16 x i8> [[TMP7]], i32 9
 ; DEFAULT-NEXT:    store i8 [[TMP47]], ptr [[TMP46]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE24]]
-; DEFAULT:       [[PRED_STORE_CONTINUE24]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE23]]
+; DEFAULT:       [[PRED_STORE_CONTINUE23]]:
 ; DEFAULT-NEXT:    [[TMP48:%.*]] = extractelement <16 x i1> [[TMP0]], i32 10
-; DEFAULT-NEXT:    br i1 [[TMP48]], label %[[PRED_STORE_IF25:.*]], label %[[PRED_STORE_CONTINUE26:.*]]
-; DEFAULT:       [[PRED_STORE_IF25]]:
+; DEFAULT-NEXT:    br i1 [[TMP48]], label %[[PRED_STORE_IF24:.*]], label %[[PRED_STORE_CONTINUE25:.*]]
+; DEFAULT:       [[PRED_STORE_IF24]]:
 ; DEFAULT-NEXT:    [[TMP49:%.*]] = add i64 [[INDEX]], 10
 ; DEFAULT-NEXT:    [[TMP50:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP49]]
 ; DEFAULT-NEXT:    [[TMP51:%.*]] = extractelement <16 x i8> [[TMP7]], i32 10
 ; DEFAULT-NEXT:    store i8 [[TMP51]], ptr [[TMP50]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE26]]
-; DEFAULT:       [[PRED_STORE_CONTINUE26]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE25]]
+; DEFAULT:       [[PRED_STORE_CONTINUE25]]:
 ; DEFAULT-NEXT:    [[TMP52:%.*]] = extractelement <16 x i1> [[TMP0]], i32 11
-; DEFAULT-NEXT:    br i1 [[TMP52]], label %[[PRED_STORE_IF27:.*]], label %[[PRED_STORE_CONTINUE28:.*]]
-; DEFAULT:       [[PRED_STORE_IF27]]:
+; DEFAULT-NEXT:    br i1 [[TMP52]], label %[[PRED_STORE_IF26:.*]], label %[[PRED_STORE_CONTINUE27:.*]]
+; DEFAULT:       [[PRED_STORE_IF26]]:
 ; DEFAULT-NEXT:    [[TMP53:%.*]] = add i64 [[INDEX]], 11
 ; DEFAULT-NEXT:    [[TMP54:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP53]]
 ; DEFAULT-NEXT:    [[TMP55:%.*]] = extractelement <16 x i8> [[TMP7]], i32 11
 ; DEFAULT-NEXT:    store i8 [[TMP55]], ptr [[TMP54]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE28]]
-; DEFAULT:       [[PRED_STORE_CONTINUE28]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE27]]
+; DEFAULT:       [[PRED_STORE_CONTINUE27]]:
 ; DEFAULT-NEXT:    [[TMP56:%.*]] = extractelement <16 x i1> [[TMP0]], i32 12
-; DEFAULT-NEXT:    br i1 [[TMP56]], label %[[PRED_STORE_IF29:.*]], label %[[PRED_STORE_CONTINUE30:.*]]
-; DEFAULT:       [[PRED_STORE_IF29]]:
+; DEFAULT-NEXT:    br i1 [[TMP56]], label %[[PRED_STORE_IF28:.*]], label %[[PRED_STORE_CONTINUE29:.*]]
+; DEFAULT:       [[PRED_STORE_IF28]]:
 ; DEFAULT-NEXT:    [[TMP57:%.*]] = add i64 [[INDEX]], 12
 ; DEFAULT-NEXT:    [[TMP58:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP57]]
 ; DEFAULT-NEXT:    [[TMP59:%.*]] = extractelement <16 x i8> [[TMP7]], i32 12
 ; DEFAULT-NEXT:    store i8 [[TMP59]], ptr [[TMP58]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE30]]
-; DEFAULT:       [[PRED_STORE_CONTINUE30]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE29]]
+; DEFAULT:       [[PRED_STORE_CONTINUE29]]:
 ; DEFAULT-NEXT:    [[TMP60:%.*]] = extractelement <16 x i1> [[TMP0]], i32 13
-; DEFAULT-NEXT:    br i1 [[TMP60]], label %[[PRED_STORE_IF31:.*]], label %[[PRED_STORE_CONTINUE32:.*]]
-; DEFAULT:       [[PRED_STORE_IF31]]:
+; DEFAULT-NEXT:    br i1 [[TMP60]], label %[[PRED_STORE_IF30:.*]], label %[[PRED_STORE_CONTINUE31:.*]]
+; DEFAULT:       [[PRED_STORE_IF30]]:
 ; DEFAULT-NEXT:    [[TMP61:%.*]] = add i64 [[INDEX]], 13
 ; DEFAULT-NEXT:    [[TMP62:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP61]]
 ; DEFAULT-NEXT:    [[TMP63:%.*]] = extractelement <16 x i8> [[TMP7]], i32 13
 ; DEFAULT-NEXT:    store i8 [[TMP63]], ptr [[TMP62]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE32]]
-; DEFAULT:       [[PRED_STORE_CONTINUE32]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE31]]
+; DEFAULT:       [[PRED_STORE_CONTINUE31]]:
 ; DEFAULT-NEXT:    [[TMP64:%.*]] = extractelement <16 x i1> [[TMP0]], i32 14
-; DEFAULT-NEXT:    br i1 [[TMP64]], label %[[PRED_STORE_IF33:.*]], label %[[PRED_STORE_CONTINUE34:.*]]
-; DEFAULT:       [[PRED_STORE_IF33]]:
+; DEFAULT-NEXT:    br i1 [[TMP64]], label %[[PRED_STORE_IF32:.*]], label %[[PRED_STORE_CONTINUE33:.*]]
+; DEFAULT:       [[PRED_STORE_IF32]]:
 ; DEFAULT-NEXT:    [[TMP65:%.*]] = add i64 [[INDEX]], 14
 ; DEFAULT-NEXT:    [[TMP66:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP65]]
 ; DEFAULT-NEXT:    [[TMP67:%.*]] = extractelement <16 x i8> [[TMP7]], i32 14
 ; DEFAULT-NEXT:    store i8 [[TMP67]], ptr [[TMP66]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE34]]
-; DEFAULT:       [[PRED_STORE_CONTINUE34]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE33]]
+; DEFAULT:       [[PRED_STORE_CONTINUE33]]:
 ; DEFAULT-NEXT:    [[TMP68:%.*]] = extractelement <16 x i1> [[TMP0]], i32 15
-; DEFAULT-NEXT:    br i1 [[TMP68]], label %[[PRED_STORE_IF35:.*]], label %[[PRED_STORE_CONTINUE36]]
-; DEFAULT:       [[PRED_STORE_IF35]]:
+; DEFAULT-NEXT:    br i1 [[TMP68]], label %[[PRED_STORE_IF34:.*]], label %[[PRED_STORE_CONTINUE35]]
+; DEFAULT:       [[PRED_STORE_IF34]]:
 ; DEFAULT-NEXT:    [[TMP69:%.*]] = add i64 [[INDEX]], 15
 ; DEFAULT-NEXT:    [[TMP70:%.*]] = getelementptr inbounds i8, ptr [[P]], i64 [[TMP69]]
 ; DEFAULT-NEXT:    [[TMP71:%.*]] = extractelement <16 x i8> [[TMP7]], i32 15
 ; DEFAULT-NEXT:    store i8 [[TMP71]], ptr [[TMP70]], align 1
-; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE36]]
-; DEFAULT:       [[PRED_STORE_CONTINUE36]]:
+; DEFAULT-NEXT:    br label %[[PRED_STORE_CONTINUE35]]
+; DEFAULT:       [[PRED_STORE_CONTINUE35]]:
 ; DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; DEFAULT-NEXT:    [[VEC_IND_NEXT]] = add <16 x i8> [[VEC_IND]], splat (i8 16)
 ; DEFAULT-NEXT:    [[VEC_IND_NEXT2]] = add <16 x i8> [[VEC_IND1]], splat (i8 16)
@@ -499,25 +492,22 @@ define void @dont_vectorize_with_minsize() {
 ; DEFAULT:       [[VECTOR_BODY]]:
 ; DEFAULT-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; DEFAULT-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @B, i64 0, i64 [[TMP0]]
-; DEFAULT-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i32 0
-; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; DEFAULT-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @C, i64 0, i64 [[TMP0]]
-; DEFAULT-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP4]], i32 0
-; DEFAULT-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4
+; DEFAULT-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
 ; DEFAULT-NEXT:    [[TMP6:%.*]] = mul nsw <4 x i32> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; DEFAULT-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [1000 x i16], ptr @A, i64 0, i64 [[TMP0]]
-; DEFAULT-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP9]], i32 0
-; DEFAULT-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i16>, ptr [[TMP10]], align 2
+; DEFAULT-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i16>, ptr [[TMP9]], align 2
 ; DEFAULT-NEXT:    [[TMP8:%.*]] = trunc <4 x i32> [[TMP6]] to <4 x i16>
 ; DEFAULT-NEXT:    [[TMP11:%.*]] = add <4 x i16> [[TMP8]], [[WIDE_LOAD2]]
-; DEFAULT-NEXT:    store <4 x i16> [[TMP11]], ptr [[TMP10]], align 2
+; DEFAULT-NEXT:    store <4 x i16> [[TMP11]], ptr [[TMP9]], align 2
 ; DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 4
 ; DEFAULT-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; DEFAULT-NEXT:    br i1 [[TMP16]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
 ; DEFAULT:       [[MIDDLE_BLOCK]]:
-; DEFAULT-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
+; DEFAULT-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
 ; DEFAULT:       [[SCALAR_PH]]:
-; DEFAULT-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 64, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; DEFAULT-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; DEFAULT-NEXT:    br label %[[FOR_BODY:.*]]
 ; DEFAULT:       [[FOR_BODY]]:
 ; DEFAULT-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -546,25 +536,22 @@ define void @dont_vectorize_with_minsize() {
 ; OPTSIZE:       [[VECTOR_BODY]]:
 ; OPTSIZE-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; OPTSIZE-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @B, i64 0, i64 [[TMP0]]
-; OPTSIZE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i32 0
-; OPTSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; OPTSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; OPTSIZE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @C, i64 0, i64 [[TMP0]]
-; OPTSIZE-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP3]], i32 0
-; OPTSIZE-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
+; OPTSIZE-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
 ; OPTSIZE-NEXT:    [[TMP5:%.*]] = mul nsw <4 x i32> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; OPTSIZE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [1000 x i16], ptr @A, i64 0, i64 [[TMP0]]
-; OPTSIZE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP6]], i32 0
-; OPTSIZE-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i16>, ptr [[TMP7]], align 2
+; OPTSIZE-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i16>, ptr [[TMP6]], align 2
 ; OPTSIZE-NEXT:    [[TMP8:%.*]] = trunc <4 x i32> [[TMP5]] to <4 x i16>
 ; OPTSIZE-NEXT:    [[TMP9:%.*]] = add <4 x i16> [[TMP8]], [[WIDE_LOAD2]]
-; OPTSIZE-NEXT:    store <4 x i16> [[TMP9]], ptr [[TMP7]], align 2
+; OPTSIZE-NEXT:    store <4 x i16> [[TMP9]], ptr [[TMP6]], align 2
 ; OPTSIZE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 4
 ; OPTSIZE-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; OPTSIZE-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; OPTSIZE:       [[MIDDLE_BLOCK]]:
-; OPTSIZE-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
+; OPTSIZE-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
 ; OPTSIZE:       [[SCALAR_PH]]:
-; OPTSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 64, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; OPTSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; OPTSIZE-NEXT:    br label %[[FOR_BODY:.*]]
 ; OPTSIZE:       [[FOR_BODY]]:
 ; OPTSIZE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -593,25 +580,22 @@ define void @dont_vectorize_with_minsize() {
 ; MINSIZE:       [[VECTOR_BODY]]:
 ; MINSIZE-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; MINSIZE-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @B, i64 0, i64 [[TMP0]]
-; MINSIZE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i32 0
-; MINSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; MINSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; MINSIZE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @C, i64 0, i64 [[TMP0]]
-; MINSIZE-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP3]], i32 0
-; MINSIZE-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i32>, ptr [[TMP4]], align 4
+; MINSIZE-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i32>, ptr [[TMP3]], align 4
 ; MINSIZE-NEXT:    [[TMP5:%.*]] = mul nsw <2 x i32> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; MINSIZE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [1000 x i16], ptr @A, i64 0, i64 [[TMP0]]
-; MINSIZE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP6]], i32 0
-; MINSIZE-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i16>, ptr [[TMP7]], align 2
+; MINSIZE-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i16>, ptr [[TMP6]], align 2
 ; MINSIZE-NEXT:    [[TMP8:%.*]] = trunc <2 x i32> [[TMP5]] to <2 x i16>
 ; MINSIZE-NEXT:    [[TMP9:%.*]] = add <2 x i16> [[TMP8]], [[WIDE_LOAD2]]
-; MINSIZE-NEXT:    store <2 x i16> [[TMP9]], ptr [[TMP7]], align 2
+; MINSIZE-NEXT:    store <2 x i16> [[TMP9]], ptr [[TMP6]], align 2
 ; MINSIZE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; MINSIZE-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; MINSIZE-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; MINSIZE:       [[MIDDLE_BLOCK]]:
-; MINSIZE-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
+; MINSIZE-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
 ; MINSIZE:       [[SCALAR_PH]]:
-; MINSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 64, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; MINSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; MINSIZE-NEXT:    br label %[[FOR_BODY:.*]]
 ; MINSIZE:       [[FOR_BODY]]:
 ; MINSIZE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -668,25 +652,22 @@ define void @vectorization_forced() {
 ; DEFAULT:       [[VECTOR_BODY]]:
 ; DEFAULT-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; DEFAULT-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @B, i64 0, i64 [[TMP0]]
-; DEFAULT-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i32 0
-; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; DEFAULT-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; DEFAULT-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @C, i64 0, i64 [[TMP0]]
-; DEFAULT-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP4]], i32 0
-; DEFAULT-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4
+; DEFAULT-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
 ; DEFAULT-NEXT:    [[TMP6:%.*]] = mul nsw <4 x i32> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; DEFAULT-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw [1000 x i16], ptr @A, i64 0, i64 [[TMP0]]
-; DEFAULT-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP9]], i32 0
-; DEFAULT-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i16>, ptr [[TMP10]], align 2
+; DEFAULT-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i16>, ptr [[TMP9]], align 2
 ; DEFAULT-NEXT:    [[TMP8:%.*]] = trunc <4 x i32> [[TMP6]] to <4 x i16>
 ; DEFAULT-NEXT:    [[TMP11:%.*]] = add <4 x i16> [[TMP8]], [[WIDE_LOAD2]]
-; DEFAULT-NEXT:    store <4 x i16> [[TMP11]], ptr [[TMP10]], align 2
+; DEFAULT-NEXT:    store <4 x i16> [[TMP11]], ptr [[TMP9]], align 2
 ; DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 4
 ; DEFAULT-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; DEFAULT-NEXT:    br i1 [[TMP16]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
 ; DEFAULT:       [[MIDDLE_BLOCK]]:
-; DEFAULT-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
+; DEFAULT-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
 ; DEFAULT:       [[SCALAR_PH]]:
-; DEFAULT-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 64, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; DEFAULT-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; DEFAULT-NEXT:    br label %[[FOR_BODY:.*]]
 ; DEFAULT:       [[FOR_BODY]]:
 ; DEFAULT-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -715,25 +696,22 @@ define void @vectorization_forced() {
 ; OPTSIZE:       [[VECTOR_BODY]]:
 ; OPTSIZE-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; OPTSIZE-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @B, i64 0, i64 [[TMP0]]
-; OPTSIZE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i32 0
-; OPTSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; OPTSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; OPTSIZE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @C, i64 0, i64 [[TMP0]]
-; OPTSIZE-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP3]], i32 0
-; OPTSIZE-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
+; OPTSIZE-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
 ; OPTSIZE-NEXT:    [[TMP5:%.*]] = mul nsw <4 x i32> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; OPTSIZE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [1000 x i16], ptr @A, i64 0, i64 [[TMP0]]
-; OPTSIZE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP6]], i32 0
-; OPTSIZE-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i16>, ptr [[TMP7]], align 2
+; OPTSIZE-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i16>, ptr [[TMP6]], align 2
 ; OPTSIZE-NEXT:    [[TMP8:%.*]] = trunc <4 x i32> [[TMP5]] to <4 x i16>
 ; OPTSIZE-NEXT:    [[TMP9:%.*]] = add <4 x i16> [[TMP8]], [[WIDE_LOAD2]]
-; OPTSIZE-NEXT:    store <4 x i16> [[TMP9]], ptr [[TMP7]], align 2
+; OPTSIZE-NEXT:    store <4 x i16> [[TMP9]], ptr [[TMP6]], align 2
 ; OPTSIZE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 4
 ; OPTSIZE-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; OPTSIZE-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; OPTSIZE:       [[MIDDLE_BLOCK]]:
-; OPTSIZE-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
+; OPTSIZE-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
 ; OPTSIZE:       [[SCALAR_PH]]:
-; OPTSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 64, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; OPTSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; OPTSIZE-NEXT:    br label %[[FOR_BODY:.*]]
 ; OPTSIZE:       [[FOR_BODY]]:
 ; OPTSIZE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -762,25 +740,22 @@ define void @vectorization_forced() {
 ; MINSIZE:       [[VECTOR_BODY]]:
 ; MINSIZE-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; MINSIZE-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @B, i64 0, i64 [[TMP0]]
-; MINSIZE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i32 0
-; MINSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; MINSIZE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; MINSIZE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw [1000 x i32], ptr @C, i64 0, i64 [[TMP0]]
-; MINSIZE-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP3]], i32 0
-; MINSIZE-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i32>, ptr [[TMP4]], align 4
+; MINSIZE-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i32>, ptr [[TMP3]], align 4
 ; MINSIZE-NEXT:    [[TMP5:%.*]] = mul nsw <2 x i32> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; MINSIZE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw [1000 x i16], ptr @A, i64 0, i64 [[TMP0]]
-; MINSIZE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i16, ptr [[TMP6]], i32 0
-; MINSIZE-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i16>, ptr [[TMP7]], align 2
+; MINSIZE-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i16>, ptr [[TMP6]], align 2
 ; MINSIZE-NEXT:    [[TMP8:%.*]] = trunc <2 x i32> [[TMP5]] to <2 x i16>
 ; MINSIZE-NEXT:    [[TMP9:%.*]] = add <2 x i16> [[TMP8]], [[WIDE_LOAD2]]
-; MINSIZE-NEXT:    store <2 x i16> [[TMP9]], ptr [[TMP7]], align 2
+; MINSIZE-NEXT:    store <2 x i16> [[TMP9]], ptr [[TMP6]], align 2
 ; MINSIZE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; MINSIZE-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; MINSIZE-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; MINSIZE:       [[MIDDLE_BLOCK]]:
-; MINSIZE-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP:.*]], label %[[SCALAR_PH]]
+; MINSIZE-NEXT:    br label %[[FOR_COND_CLEANUP:.*]]
 ; MINSIZE:       [[SCALAR_PH]]:
-; MINSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 64, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; MINSIZE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; MINSIZE-NEXT:    br label %[[FOR_BODY:.*]]
 ; MINSIZE:       [[FOR_BODY]]:
 ; MINSIZE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/ARM/pointer_iv.ll b/llvm/test/Transforms/LoopVectorize/ARM/pointer_iv.ll
index fa03e29..7c6e705 100644
--- a/llvm/test/Transforms/LoopVectorize/ARM/pointer_iv.ll
+++ b/llvm/test/Transforms/LoopVectorize/ARM/pointer_iv.ll
@@ -62,8 +62,8 @@ define hidden void @pointer_phi_v4i32_add2(ptr noalias nocapture readonly %A, pt
 ; CHECK-NEXT:    [[TMP0:%.*]] = add nsw <4 x i32> [[STRIDED_VEC]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    store <4 x i32> [[TMP0]], ptr [[NEXT_GEP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], 996
-; CHECK-NEXT:    br i1 [[TMP1]], label [[FOR_BODY:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 996
+; CHECK-NEXT:    br i1 [[TMP3]], label [[FOR_BODY:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[A_ADDR_09:%.*]] = phi ptr [ [[ADD_PTR:%.*]], [[FOR_BODY]] ], [ [[IND_END]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[I_08:%.*]] = phi i32 [ [[INC:%.*]], [[FOR_BODY]] ], [ 996, [[VECTOR_BODY]] ]
@@ -106,8 +106,8 @@ define hidden void @pointer_phi_v4i32_add3(ptr noalias nocapture readonly %A, pt
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[POINTER_PHI:%.*]] = phi ptr [ [[A]], [[ENTRY:%.*]] ], [ [[PTR_IND:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[POINTER_PHI:%.*]] = phi ptr [ [[A]], [[ENTRY]] ], [ [[PTR_IND:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <4 x i32> <i32 0, i32 12, i32 24, i32 36>
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = shl i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[B]], i32 [[OFFSET_IDX]]
@@ -116,8 +116,8 @@ define hidden void @pointer_phi_v4i32_add3(ptr noalias nocapture readonly %A, pt
 ; CHECK-NEXT:    store <4 x i32> [[TMP1]], ptr [[NEXT_GEP]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i32 48
-; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 996
-; CHECK-NEXT:    br i1 [[TMP2]], label [[FOR_BODY:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], 996
+; CHECK-NEXT:    br i1 [[TMP4]], label [[FOR_BODY:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[A_ADDR_09:%.*]] = phi ptr [ [[ADD_PTR:%.*]], [[FOR_BODY]] ], [ [[IND_END]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[I_08:%.*]] = phi i32 [ [[INC:%.*]], [[FOR_BODY]] ], [ 996, [[VECTOR_BODY]] ]
@@ -212,8 +212,8 @@ define hidden void @pointer_phi_v8i16_add2(ptr noalias nocapture readonly %A, pt
 ; CHECK-NEXT:    [[TMP1:%.*]] = add <8 x i16> [[STRIDED_VEC]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    store <8 x i16> [[TMP1]], ptr [[NEXT_GEP5]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
-; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 992
-; CHECK-NEXT:    br i1 [[TMP2]], label [[FOR_BODY:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], 992
+; CHECK-NEXT:    br i1 [[TMP4]], label [[FOR_BODY:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[A_ADDR_011:%.*]] = phi ptr [ [[ADD_PTR:%.*]], [[FOR_BODY]] ], [ [[IND_END]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[I_010:%.*]] = phi i32 [ [[INC:%.*]], [[FOR_BODY]] ], [ 992, [[VECTOR_BODY]] ]
@@ -495,8 +495,8 @@ define hidden void @pointer_phi_v4f32_add2(ptr noalias nocapture readonly %A, pt
 ; CHECK-NEXT:    [[TMP0:%.*]] = fadd fast <4 x float> [[STRIDED_VEC]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    store <4 x float> [[TMP0]], ptr [[NEXT_GEP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], 996
-; CHECK-NEXT:    br i1 [[TMP1]], label [[FOR_BODY:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 996
+; CHECK-NEXT:    br i1 [[TMP3]], label [[FOR_BODY:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[A_ADDR_09:%.*]] = phi ptr [ [[ADD_PTR:%.*]], [[FOR_BODY]] ], [ [[IND_END]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[I_08:%.*]] = phi i32 [ [[INC:%.*]], [[FOR_BODY]] ], [ 996, [[VECTOR_BODY]] ]
@@ -539,8 +539,8 @@ define hidden void @pointer_phi_v4f32_add3(ptr noalias nocapture readonly %A, pt
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x float> [[BROADCAST_SPLATINSERT]], <4 x float> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[POINTER_PHI:%.*]] = phi ptr [ [[A]], [[ENTRY:%.*]] ], [ [[PTR_IND:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[POINTER_PHI:%.*]] = phi ptr [ [[A]], [[ENTRY]] ], [ [[PTR_IND:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <4 x i32> <i32 0, i32 12, i32 24, i32 36>
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = shl i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[B]], i32 [[OFFSET_IDX]]
@@ -549,8 +549,8 @@ define hidden void @pointer_phi_v4f32_add3(ptr noalias nocapture readonly %A, pt
 ; CHECK-NEXT:    store <4 x float> [[TMP1]], ptr [[NEXT_GEP]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i32 48
-; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 996
-; CHECK-NEXT:    br i1 [[TMP2]], label [[FOR_BODY:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], 996
+; CHECK-NEXT:    br i1 [[TMP4]], label [[FOR_BODY:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[A_ADDR_09:%.*]] = phi ptr [ [[ADD_PTR:%.*]], [[FOR_BODY]] ], [ [[IND_END]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[I_08:%.*]] = phi i32 [ [[INC:%.*]], [[FOR_BODY]] ], [ 996, [[VECTOR_BODY]] ]
@@ -642,8 +642,8 @@ define hidden void @pointer_phi_v4half_add2(ptr noalias nocapture readonly %A, p
 ; CHECK-NEXT:    [[TMP0:%.*]] = fadd fast <8 x half> [[STRIDED_VEC]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    store <8 x half> [[TMP0]], ptr [[NEXT_GEP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
-; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], 992
-; CHECK-NEXT:    br i1 [[TMP1]], label [[FOR_BODY:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 992
+; CHECK-NEXT:    br i1 [[TMP3]], label [[FOR_BODY:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[A_ADDR_09:%.*]] = phi ptr [ [[ADD_PTR:%.*]], [[FOR_BODY]] ], [ [[IND_END]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[I_08:%.*]] = phi i32 [ [[INC:%.*]], [[FOR_BODY]] ], [ 992, [[VECTOR_BODY]] ]
@@ -696,8 +696,8 @@ define hidden void @pointer_phi_v4half_add3(ptr noalias nocapture readonly %A, p
 ; CHECK-NEXT:    [[TMP0:%.*]] = fadd fast <8 x half> [[STRIDED_VEC]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    store <8 x half> [[TMP0]], ptr [[NEXT_GEP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
-; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], 992
-; CHECK-NEXT:    br i1 [[TMP1]], label [[FOR_BODY:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 992
+; CHECK-NEXT:    br i1 [[TMP3]], label [[FOR_BODY:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[A_ADDR_09:%.*]] = phi ptr [ [[ADD_PTR:%.*]], [[FOR_BODY]] ], [ [[IND_END]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[I_08:%.*]] = phi i32 [ [[INC:%.*]], [[FOR_BODY]] ], [ 992, [[VECTOR_BODY]] ]
@@ -743,8 +743,8 @@ define hidden void @pointer_phi_v4i32_uf2(ptr noalias nocapture readonly %A, ptr
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[POINTER_PHI:%.*]] = phi ptr [ [[A]], [[ENTRY:%.*]] ], [ [[PTR_IND:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[POINTER_PHI:%.*]] = phi ptr [ [[A]], [[ENTRY]] ], [ [[PTR_IND:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <4 x i32> <i32 0, i32 24, i32 48, i32 72>
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <4 x i32> <i32 96, i32 120, i32 144, i32 168>
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = shl i32 [[INDEX]], 2
@@ -808,8 +808,8 @@ define hidden void @pointer_phi_v4i32_uf4(ptr noalias nocapture readonly %A, ptr
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[POINTER_PHI:%.*]] = phi ptr [ [[A]], [[ENTRY:%.*]] ], [ [[PTR_IND:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[POINTER_PHI:%.*]] = phi ptr [ [[A]], [[ENTRY]] ], [ [[PTR_IND:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <4 x i32> <i32 0, i32 24, i32 48, i32 72>
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <4 x i32> <i32 96, i32 120, i32 144, i32 168>
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <4 x i32> <i32 192, i32 216, i32 240, i32 264>
@@ -880,8 +880,8 @@ define hidden void @mult_ptr_iv(ptr noalias nocapture readonly %x, ptr noalias n
 ; CHECK-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
 ; CHECK-NEXT:    br i1 [[FOUND_CONFLICT]], label [[FOR_BODY:%.*]], label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ], [ 0, [[ENTRY]] ]
-; CHECK-NEXT:    [[POINTER_PHI:%.*]] = phi ptr [ [[PTR_IND:%.*]], [[VECTOR_BODY]] ], [ [[X]], [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[POINTER_PHI:%.*]] = phi ptr [ [[PTR_IND:%.*]], [[VECTOR_BODY]] ], [ [[X]], [[ENTRY]] ]
 ; CHECK-NEXT:    [[POINTER_PHI5:%.*]] = phi ptr [ [[PTR_IND6:%.*]], [[VECTOR_BODY]] ], [ [[Z]], [[ENTRY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <4 x i32> <i32 0, i32 3, i32 6, i32 9>
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[POINTER_PHI5]], <4 x i32> <i32 0, i32 3, i32 6, i32 9>
diff --git a/llvm/test/Transforms/LoopVectorize/ARM/sphinx.ll b/llvm/test/Transforms/LoopVectorize/ARM/sphinx.ll
index ce714f6..91b0508 100644
--- a/llvm/test/Transforms/LoopVectorize/ARM/sphinx.ll
+++ b/llvm/test/Transforms/LoopVectorize/ARM/sphinx.ll
@@ -45,17 +45,14 @@ define i32 @test(ptr nocapture readonly %x) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x double> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[X:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP3]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[T4]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x float>, ptr [[TMP5]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x float>, ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = fsub fast <2 x float> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = fpext <2 x float> [[TMP6]] to <2 x double>
 ; CHECK-NEXT:    [[TMP8:%.*]] = fmul fast <2 x double> [[TMP7]], [[TMP7]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[T6]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[TMP9]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x float>, ptr [[TMP10]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x float>, ptr [[TMP9]], align 4
 ; CHECK-NEXT:    [[TMP11:%.*]] = fpext <2 x float> [[WIDE_LOAD2]] to <2 x double>
 ; CHECK-NEXT:    [[TMP12:%.*]] = fmul fast <2 x double> [[TMP8]], [[TMP11]]
 ; CHECK-NEXT:    [[TMP13]] = fsub fast <2 x double> [[VEC_PHI]], [[TMP12]]
diff --git a/llvm/test/Transforms/LoopVectorize/ARM/tail-folding-not-allowed.ll b/llvm/test/Transforms/LoopVectorize/ARM/tail-folding-not-allowed.ll
index 2c659d1..957f8ee 100644
--- a/llvm/test/Transforms/LoopVectorize/ARM/tail-folding-not-allowed.ll
+++ b/llvm/test/Transforms/LoopVectorize/ARM/tail-folding-not-allowed.ll
@@ -14,25 +14,21 @@ define void @trunc_not_allowed_different_vec_elemns(ptr noalias nocapture %A, pt
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = add nsw <4 x i32> [[WIDE_LOAD1]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP7]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = trunc <4 x i32> [[TMP5]] to <4 x i16>
 ; CHECK-NEXT:    [[TMP9:%.*]] = shl <4 x i16> [[TMP8]], splat (i16 1)
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i16, ptr [[D:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i16, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <4 x i16> [[TMP9]], ptr [[TMP11]], align 2
+; CHECK-NEXT:    store <4 x i16> [[TMP9]], ptr [[TMP10]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i32 [[INDEX_NEXT]], 428
 ; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 428, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
@@ -53,7 +49,7 @@ define void @trunc_not_allowed_different_vec_elemns(ptr noalias nocapture %A, pt
 ; CHECK-NEXT:    store i16 [[CONV7]], ptr [[ARRAYIDX8]], align 2
 ; CHECK-NEXT:    [[ADD9]] = add nuw nsw i32 [[I_021]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[ADD9]], 431
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ;
 entry:
   br label %for.body
@@ -127,21 +123,18 @@ define void @narrowing_load_not_allowed(ptr noalias nocapture %A, ptr noalias no
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[C:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP2]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i16>, ptr [[TMP1]], align 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[B:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i8>, ptr [[TMP4]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[TMP5:%.*]] = trunc <8 x i16> [[WIDE_LOAD]] to <8 x i8>
 ; CHECK-NEXT:    [[TMP6:%.*]] = add <8 x i8> [[WIDE_LOAD1]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <8 x i8> [[TMP6]], ptr [[TMP8]], align 1
+; CHECK-NEXT:    store <8 x i8> [[TMP6]], ptr [[TMP7]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i32 [[INDEX_NEXT]], 424
 ; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 424, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
@@ -159,7 +152,7 @@ define void @narrowing_load_not_allowed(ptr noalias nocapture %A, ptr noalias no
 ; CHECK-NEXT:    store i8 [[ADD]], ptr [[ARRAYIDX5]], align 1
 ; CHECK-NEXT:    [[ADD6]] = add nuw nsw i32 [[I_012]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[ADD6]], 431
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ;
 entry:
   br label %for.body
@@ -196,20 +189,17 @@ define void @trunc_not_allowed(ptr noalias nocapture %A, ptr noalias nocapture r
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = add nsw <4 x i32> [[WIDE_LOAD1]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP7]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], 428
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 428, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
@@ -227,7 +217,7 @@ define void @trunc_not_allowed(ptr noalias nocapture %A, ptr noalias nocapture r
 ; CHECK-NEXT:    [[ADD3]] = add nuw nsw i32 [[I_09]], 1
 ; CHECK-NEXT:    [[ADD_IV:%.*]] = trunc i32 [[ADD3]] to i16
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i16 [[ADD_IV]], 431
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
 ;
 entry:
   br label %for.body
@@ -264,8 +254,7 @@ define void @strides_different_direction(ptr noalias nocapture %A, ptr noalias n
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = sub nsw i32 [[N:%.*]], [[INDEX]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i32 [[TMP3]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
@@ -274,13 +263,12 @@ define void @strides_different_direction(ptr noalias nocapture %A, ptr noalias n
 ; CHECK-NEXT:    [[REVERSE:%.*]] = shufflevector <4 x i32> [[WIDE_LOAD1]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; CHECK-NEXT:    [[TMP7:%.*]] = add nsw <4 x i32> [[REVERSE]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP7]], ptr [[TMP9]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP7]], ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[INDEX_NEXT]], 428
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 428, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
@@ -298,7 +286,7 @@ define void @strides_different_direction(ptr noalias nocapture %A, ptr noalias n
 ; CHECK-NEXT:    store i32 [[ADD]], ptr [[ARRAYIDX2]], align 4
 ; CHECK-NEXT:    [[ADD3]] = add nuw nsw i32 [[I_09]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[ADD3]], 431
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
 ;
 entry:
   br label %for.body
@@ -330,20 +318,17 @@ define void @too_many_loop_blocks(ptr noalias nocapture %A, ptr noalias nocaptur
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = add nsw <4 x i32> [[WIDE_LOAD1]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP7]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], 428
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 428, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
@@ -362,7 +347,7 @@ define void @too_many_loop_blocks(ptr noalias nocapture %A, ptr noalias nocaptur
 ; CHECK:       loopincr:
 ; CHECK-NEXT:    [[ADD3]] = add nuw nsw i32 [[I_09]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[ADD3]], 431
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ;
 entry:
   br label %for.body
@@ -435,25 +420,21 @@ define void @fptrunc_not_allowed(ptr noalias nocapture %A, ptr noalias nocapture
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[C:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = fadd fast <4 x float> [[WIDE_LOAD1]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <4 x float> [[TMP5]], ptr [[TMP7]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP5]], ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = fptrunc <4 x float> [[TMP5]] to <4 x half>
 ; CHECK-NEXT:    [[TMP9:%.*]] = fmul fast <4 x half> [[TMP8]], splat (half 0xH4000)
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds half, ptr [[D:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds half, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <4 x half> [[TMP9]], ptr [[TMP11]], align 2
+; CHECK-NEXT:    store <4 x half> [[TMP9]], ptr [[TMP10]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i32 [[INDEX_NEXT]], 428
 ; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 428, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
@@ -474,7 +455,7 @@ define void @fptrunc_not_allowed(ptr noalias nocapture %A, ptr noalias nocapture
 ; CHECK-NEXT:    store half [[FACTOR]], ptr [[ARRAYIDX5]], align 2
 ; CHECK-NEXT:    [[ADD6]] = add nuw nsw i32 [[I_017]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[ADD6]], 431
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
 ;
 entry:
   br label %for.body
@@ -525,15 +506,13 @@ define dso_local void @select_not_allowed(ptr noalias nocapture %A, ptr noalias
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[COND:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = select <4 x i1> [[TMP3]], <4 x ptr> [[BROADCAST_SPLAT]], <4 x ptr> [[BROADCAST_SPLAT2]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, <4 x ptr> [[TMP4]], <4 x i32> [[VEC_IND]]
 ; CHECK-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <4 x i32> @llvm.masked.gather.v4i32.v4p0(<4 x ptr> [[TMP5]], i32 4, <4 x i1> splat (i1 true), <4 x i32> poison)
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[WIDE_MASKED_GATHER]], ptr [[TMP7]], align 4
+; CHECK-NEXT:    store <4 x i32> [[WIDE_MASKED_GATHER]], ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
@@ -606,8 +585,7 @@ define i32 @i32_smin_reduction(ptr nocapture readonly %x, i32 %n) #0 {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 2147483647), [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[X:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp slt <4 x i32> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i32> [[VEC_PHI]], <4 x i32> [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
@@ -674,8 +652,7 @@ define i32 @i32_smax_reduction(ptr nocapture readonly %x, i32 %n) #0 {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 -2147483648), [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[X:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp sgt <4 x i32> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i32> [[VEC_PHI]], <4 x i32> [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
@@ -742,8 +719,7 @@ define i32 @i32_umin_reduction(ptr nocapture readonly %x, i32 %n) #0 {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 -1), [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[X:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp ult <4 x i32> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i32> [[VEC_PHI]], <4 x i32> [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
@@ -810,8 +786,7 @@ define i32 @i32_umax_reduction(ptr nocapture readonly %x, i32 %n) #0 {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[X:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp ugt <4 x i32> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i32> [[VEC_PHI]], <4 x i32> [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
diff --git a/llvm/test/Transforms/LoopVectorize/ARM/tail-folding-scalar-epilogue-fallback.ll b/llvm/test/Transforms/LoopVectorize/ARM/tail-folding-scalar-epilogue-fallback.ll
index 06e345f..e08614b 100644
--- a/llvm/test/Transforms/LoopVectorize/ARM/tail-folding-scalar-epilogue-fallback.ll
+++ b/llvm/test/Transforms/LoopVectorize/ARM/tail-folding-scalar-epilogue-fallback.ll
@@ -25,10 +25,8 @@ define void @outside_user_blocks_tail_folding(ptr nocapture readonly %ptr, i32 %
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[PTR]], i32 [[INDEX]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[NEXT_GEP]], i32 1
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
-; CHECK-NEXT:    store <16 x i8> [[WIDE_LOAD]], ptr [[TMP3]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEXT:    store <16 x i8> [[WIDE_LOAD]], ptr [[NEXT_GEP]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 16
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/LoongArch/defaults.ll b/llvm/test/Transforms/LoopVectorize/LoongArch/defaults.ll
index 631f3c7..17eeafa 100644
--- a/llvm/test/Transforms/LoopVectorize/LoongArch/defaults.ll
+++ b/llvm/test/Transforms/LoopVectorize/LoongArch/defaults.ll
@@ -22,21 +22,20 @@ define void @vector_add(ptr noalias nocapture %a, i64 %v) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 4
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP5]], align 8
 ; CHECK-NEXT:    [[TMP6:%.*]] = add <4 x i64> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = add <4 x i64> [[WIDE_LOAD1]], [[BROADCAST_SPLAT]]
-; CHECK-NEXT:    store <4 x i64> [[TMP6]], ptr [[TMP4]], align 8
+; CHECK-NEXT:    store <4 x i64> [[TMP6]], ptr [[TMP2]], align 8
 ; CHECK-NEXT:    store <4 x i64> [[TMP7]], ptr [[TMP5]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/PowerPC/exit-branch-cost.ll b/llvm/test/Transforms/LoopVectorize/PowerPC/exit-branch-cost.ll
index b498712..1d0751a 100644
--- a/llvm/test/Transforms/LoopVectorize/PowerPC/exit-branch-cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/PowerPC/exit-branch-cost.ll
@@ -51,7 +51,6 @@ define i1 @select_exit_cond(ptr %start, ptr %end, i64 %N) {
 ; CHECK-NEXT:    [[STEP_ADD_10:%.*]] = add <2 x i64> [[STEP_ADD_9]], splat (i64 2)
 ; CHECK-NEXT:    [[STEP_ADD_11:%.*]] = add <2 x i64> [[STEP_ADD_10]], splat (i64 2)
 ; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[START]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 2
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 4
 ; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 6
@@ -63,7 +62,7 @@ define i1 @select_exit_cond(ptr %start, ptr %end, i64 %N) {
 ; CHECK-NEXT:    [[TMP69:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 18
 ; CHECK-NEXT:    [[TMP70:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 20
 ; CHECK-NEXT:    [[TMP71:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 22
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[TMP11]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[NEXT_GEP]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD25:%.*]] = load <2 x i8>, ptr [[TMP12]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD26:%.*]] = load <2 x i8>, ptr [[TMP13]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD27:%.*]] = load <2 x i8>, ptr [[TMP14]], align 1
@@ -163,8 +162,7 @@ define i1 @select_exit_cond(ptr %start, ptr %end, i64 %N) {
 ; CHECK-NEXT:    [[VEC_IND27:%.*]] = phi <2 x i64> [ [[INDUCTION]], %[[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT28:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI29:%.*]] = phi <2 x i64> [ [[TMP57]], %[[VEC_EPILOG_PH]] ], [ [[TMP58:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[NEXT_GEP30:%.*]] = getelementptr i8, ptr [[START]], i64 [[INDEX38]]
-; CHECK-NEXT:    [[TMP60:%.*]] = getelementptr i8, ptr [[NEXT_GEP30]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD32:%.*]] = load <2 x i8>, ptr [[TMP60]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD32:%.*]] = load <2 x i8>, ptr [[NEXT_GEP30]], align 1
 ; CHECK-NEXT:    [[TMP61:%.*]] = zext <2 x i8> [[WIDE_LOAD32]] to <2 x i64>
 ; CHECK-NEXT:    [[TMP62:%.*]] = shl <2 x i64> [[VEC_IND27]], splat (i64 1)
 ; CHECK-NEXT:    [[TMP63:%.*]] = shl <2 x i64> [[TMP61]], [[TMP62]]
diff --git a/llvm/test/Transforms/LoopVectorize/PowerPC/optimal-epilog-vectorization.ll b/llvm/test/Transforms/LoopVectorize/PowerPC/optimal-epilog-vectorization.ll
index ed83af6..4d14624 100644
--- a/llvm/test/Transforms/LoopVectorize/PowerPC/optimal-epilog-vectorization.ll
+++ b/llvm/test/Transforms/LoopVectorize/PowerPC/optimal-epilog-vectorization.ll
@@ -26,7 +26,6 @@ define dso_local void @f1(ptr noalias %aa, ptr noalias %bb, ptr noalias %cc, i32
 ; VF-TWO-CHECK:       vector.body:
 ; VF-TWO-CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; VF-TWO-CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[BB]], i64 [[INDEX]]
-; VF-TWO-CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 0
 ; VF-TWO-CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 4
 ; VF-TWO-CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 8
 ; VF-TWO-CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 12
@@ -34,7 +33,7 @@ define dso_local void @f1(ptr noalias %aa, ptr noalias %bb, ptr noalias %cc, i32
 ; VF-TWO-CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 20
 ; VF-TWO-CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 24
 ; VF-TWO-CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 28
-; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP16]], align 4
+; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP8]], align 4
 ; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP17]], align 4
 ; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x float>, ptr [[TMP18]], align 4
 ; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x float>, ptr [[TMP19]], align 4
@@ -43,7 +42,6 @@ define dso_local void @f1(ptr noalias %aa, ptr noalias %bb, ptr noalias %cc, i32
 ; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x float>, ptr [[TMP22]], align 4
 ; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x float>, ptr [[TMP23]], align 4
 ; VF-TWO-CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, ptr [[CC]], i64 [[INDEX]]
-; VF-TWO-CHECK-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i32 0
 ; VF-TWO-CHECK-NEXT:    [[TMP33:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i32 4
 ; VF-TWO-CHECK-NEXT:    [[TMP34:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i32 8
 ; VF-TWO-CHECK-NEXT:    [[TMP35:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i32 12
@@ -51,7 +49,7 @@ define dso_local void @f1(ptr noalias %aa, ptr noalias %bb, ptr noalias %cc, i32
 ; VF-TWO-CHECK-NEXT:    [[TMP37:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i32 20
 ; VF-TWO-CHECK-NEXT:    [[TMP38:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i32 24
 ; VF-TWO-CHECK-NEXT:    [[TMP39:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i32 28
-; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x float>, ptr [[TMP32]], align 4
+; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x float>, ptr [[TMP24]], align 4
 ; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD10:%.*]] = load <4 x float>, ptr [[TMP33]], align 4
 ; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x float>, ptr [[TMP34]], align 4
 ; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD12:%.*]] = load <4 x float>, ptr [[TMP35]], align 4
@@ -68,7 +66,6 @@ define dso_local void @f1(ptr noalias %aa, ptr noalias %bb, ptr noalias %cc, i32
 ; VF-TWO-CHECK-NEXT:    [[TMP46:%.*]] = fadd fast <4 x float> [[WIDE_LOAD7]], [[WIDE_LOAD15]]
 ; VF-TWO-CHECK-NEXT:    [[TMP47:%.*]] = fadd fast <4 x float> [[WIDE_LOAD8]], [[WIDE_LOAD16]]
 ; VF-TWO-CHECK-NEXT:    [[TMP48:%.*]] = getelementptr inbounds float, ptr [[AA]], i64 [[INDEX]]
-; VF-TWO-CHECK-NEXT:    [[TMP56:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i32 0
 ; VF-TWO-CHECK-NEXT:    [[TMP57:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i32 4
 ; VF-TWO-CHECK-NEXT:    [[TMP58:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i32 8
 ; VF-TWO-CHECK-NEXT:    [[TMP59:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i32 12
@@ -76,7 +73,7 @@ define dso_local void @f1(ptr noalias %aa, ptr noalias %bb, ptr noalias %cc, i32
 ; VF-TWO-CHECK-NEXT:    [[TMP61:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i32 20
 ; VF-TWO-CHECK-NEXT:    [[TMP62:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i32 24
 ; VF-TWO-CHECK-NEXT:    [[TMP63:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i32 28
-; VF-TWO-CHECK-NEXT:    store <4 x float> [[TMP40]], ptr [[TMP56]], align 4
+; VF-TWO-CHECK-NEXT:    store <4 x float> [[TMP40]], ptr [[TMP48]], align 4
 ; VF-TWO-CHECK-NEXT:    store <4 x float> [[TMP41]], ptr [[TMP57]], align 4
 ; VF-TWO-CHECK-NEXT:    store <4 x float> [[TMP42]], ptr [[TMP58]], align 4
 ; VF-TWO-CHECK-NEXT:    store <4 x float> [[TMP43]], ptr [[TMP59]], align 4
@@ -102,15 +99,12 @@ define dso_local void @f1(ptr noalias %aa, ptr noalias %bb, ptr noalias %cc, i32
 ; VF-TWO-CHECK:       vec.epilog.vector.body:
 ; VF-TWO-CHECK-NEXT:    [[INDEX20:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT23:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; VF-TWO-CHECK-NEXT:    [[TMP66:%.*]] = getelementptr inbounds float, ptr [[BB]], i64 [[INDEX20]]
-; VF-TWO-CHECK-NEXT:    [[TMP67:%.*]] = getelementptr inbounds float, ptr [[TMP66]], i32 0
-; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD21:%.*]] = load <2 x float>, ptr [[TMP67]], align 4
+; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD21:%.*]] = load <2 x float>, ptr [[TMP66]], align 4
 ; VF-TWO-CHECK-NEXT:    [[TMP68:%.*]] = getelementptr inbounds float, ptr [[CC]], i64 [[INDEX20]]
-; VF-TWO-CHECK-NEXT:    [[TMP69:%.*]] = getelementptr inbounds float, ptr [[TMP68]], i32 0
-; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD22:%.*]] = load <2 x float>, ptr [[TMP69]], align 4
+; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD22:%.*]] = load <2 x float>, ptr [[TMP68]], align 4
 ; VF-TWO-CHECK-NEXT:    [[TMP70:%.*]] = fadd fast <2 x float> [[WIDE_LOAD21]], [[WIDE_LOAD22]]
 ; VF-TWO-CHECK-NEXT:    [[TMP71:%.*]] = getelementptr inbounds float, ptr [[AA]], i64 [[INDEX20]]
-; VF-TWO-CHECK-NEXT:    [[TMP72:%.*]] = getelementptr inbounds float, ptr [[TMP71]], i32 0
-; VF-TWO-CHECK-NEXT:    store <2 x float> [[TMP70]], ptr [[TMP72]], align 4
+; VF-TWO-CHECK-NEXT:    store <2 x float> [[TMP70]], ptr [[TMP71]], align 4
 ; VF-TWO-CHECK-NEXT:    [[INDEX_NEXT23]] = add nuw i64 [[INDEX20]], 2
 ; VF-TWO-CHECK-NEXT:    [[TMP73:%.*]] = icmp eq i64 [[INDEX_NEXT23]], [[N_VEC18]]
 ; VF-TWO-CHECK-NEXT:    br i1 [[TMP73]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
@@ -156,7 +150,6 @@ define dso_local void @f1(ptr noalias %aa, ptr noalias %bb, ptr noalias %cc, i32
 ; VF-FOUR-CHECK:       vector.body:
 ; VF-FOUR-CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; VF-FOUR-CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[BB]], i64 [[INDEX]]
-; VF-FOUR-CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 0
 ; VF-FOUR-CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 4
 ; VF-FOUR-CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 8
 ; VF-FOUR-CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 12
@@ -164,7 +157,7 @@ define dso_local void @f1(ptr noalias %aa, ptr noalias %bb, ptr noalias %cc, i32
 ; VF-FOUR-CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 20
 ; VF-FOUR-CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 24
 ; VF-FOUR-CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 28
-; VF-FOUR-CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP16]], align 4
+; VF-FOUR-CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP8]], align 4
 ; VF-FOUR-CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP17]], align 4
 ; VF-FOUR-CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x float>, ptr [[TMP18]], align 4
 ; VF-FOUR-CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x float>, ptr [[TMP19]], align 4
@@ -173,7 +166,6 @@ define dso_local void @f1(ptr noalias %aa, ptr noalias %bb, ptr noalias %cc, i32
 ; VF-FOUR-CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x float>, ptr [[TMP22]], align 4
 ; VF-FOUR-CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x float>, ptr [[TMP23]], align 4
 ; VF-FOUR-CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds float, ptr [[CC]], i64 [[INDEX]]
-; VF-FOUR-CHECK-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i32 0
 ; VF-FOUR-CHECK-NEXT:    [[TMP33:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i32 4
 ; VF-FOUR-CHECK-NEXT:    [[TMP34:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i32 8
 ; VF-FOUR-CHECK-NEXT:    [[TMP35:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i32 12
@@ -181,7 +173,7 @@ define dso_local void @f1(ptr noalias %aa, ptr noalias %bb, ptr noalias %cc, i32
 ; VF-FOUR-CHECK-NEXT:    [[TMP37:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i32 20
 ; VF-FOUR-CHECK-NEXT:    [[TMP38:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i32 24
 ; VF-FOUR-CHECK-NEXT:    [[TMP39:%.*]] = getelementptr inbounds float, ptr [[TMP24]], i32 28
-; VF-FOUR-CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x float>, ptr [[TMP32]], align 4
+; VF-FOUR-CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x float>, ptr [[TMP24]], align 4
 ; VF-FOUR-CHECK-NEXT:    [[WIDE_LOAD10:%.*]] = load <4 x float>, ptr [[TMP33]], align 4
 ; VF-FOUR-CHECK-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x float>, ptr [[TMP34]], align 4
 ; VF-FOUR-CHECK-NEXT:    [[WIDE_LOAD12:%.*]] = load <4 x float>, ptr [[TMP35]], align 4
@@ -198,7 +190,6 @@ define dso_local void @f1(ptr noalias %aa, ptr noalias %bb, ptr noalias %cc, i32
 ; VF-FOUR-CHECK-NEXT:    [[TMP46:%.*]] = fadd fast <4 x float> [[WIDE_LOAD7]], [[WIDE_LOAD15]]
 ; VF-FOUR-CHECK-NEXT:    [[TMP47:%.*]] = fadd fast <4 x float> [[WIDE_LOAD8]], [[WIDE_LOAD16]]
 ; VF-FOUR-CHECK-NEXT:    [[TMP48:%.*]] = getelementptr inbounds float, ptr [[AA]], i64 [[INDEX]]
-; VF-FOUR-CHECK-NEXT:    [[TMP56:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i32 0
 ; VF-FOUR-CHECK-NEXT:    [[TMP57:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i32 4
 ; VF-FOUR-CHECK-NEXT:    [[TMP58:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i32 8
 ; VF-FOUR-CHECK-NEXT:    [[TMP59:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i32 12
@@ -206,7 +197,7 @@ define dso_local void @f1(ptr noalias %aa, ptr noalias %bb, ptr noalias %cc, i32
 ; VF-FOUR-CHECK-NEXT:    [[TMP61:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i32 20
 ; VF-FOUR-CHECK-NEXT:    [[TMP62:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i32 24
 ; VF-FOUR-CHECK-NEXT:    [[TMP63:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i32 28
-; VF-FOUR-CHECK-NEXT:    store <4 x float> [[TMP40]], ptr [[TMP56]], align 4
+; VF-FOUR-CHECK-NEXT:    store <4 x float> [[TMP40]], ptr [[TMP48]], align 4
 ; VF-FOUR-CHECK-NEXT:    store <4 x float> [[TMP41]], ptr [[TMP57]], align 4
 ; VF-FOUR-CHECK-NEXT:    store <4 x float> [[TMP42]], ptr [[TMP58]], align 4
 ; VF-FOUR-CHECK-NEXT:    store <4 x float> [[TMP43]], ptr [[TMP59]], align 4
@@ -232,15 +223,12 @@ define dso_local void @f1(ptr noalias %aa, ptr noalias %bb, ptr noalias %cc, i32
 ; VF-FOUR-CHECK:       vec.epilog.vector.body:
 ; VF-FOUR-CHECK-NEXT:    [[INDEX20:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT23:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; VF-FOUR-CHECK-NEXT:    [[TMP66:%.*]] = getelementptr inbounds float, ptr [[BB]], i64 [[INDEX20]]
-; VF-FOUR-CHECK-NEXT:    [[TMP67:%.*]] = getelementptr inbounds float, ptr [[TMP66]], i32 0
-; VF-FOUR-CHECK-NEXT:    [[WIDE_LOAD21:%.*]] = load <4 x float>, ptr [[TMP67]], align 4
+; VF-FOUR-CHECK-NEXT:    [[WIDE_LOAD21:%.*]] = load <4 x float>, ptr [[TMP66]], align 4
 ; VF-FOUR-CHECK-NEXT:    [[TMP68:%.*]] = getelementptr inbounds float, ptr [[CC]], i64 [[INDEX20]]
-; VF-FOUR-CHECK-NEXT:    [[TMP69:%.*]] = getelementptr inbounds float, ptr [[TMP68]], i32 0
-; VF-FOUR-CHECK-NEXT:    [[WIDE_LOAD22:%.*]] = load <4 x float>, ptr [[TMP69]], align 4
+; VF-FOUR-CHECK-NEXT:    [[WIDE_LOAD22:%.*]] = load <4 x float>, ptr [[TMP68]], align 4
 ; VF-FOUR-CHECK-NEXT:    [[TMP70:%.*]] = fadd fast <4 x float> [[WIDE_LOAD21]], [[WIDE_LOAD22]]
 ; VF-FOUR-CHECK-NEXT:    [[TMP71:%.*]] = getelementptr inbounds float, ptr [[AA]], i64 [[INDEX20]]
-; VF-FOUR-CHECK-NEXT:    [[TMP72:%.*]] = getelementptr inbounds float, ptr [[TMP71]], i32 0
-; VF-FOUR-CHECK-NEXT:    store <4 x float> [[TMP70]], ptr [[TMP72]], align 4
+; VF-FOUR-CHECK-NEXT:    store <4 x float> [[TMP70]], ptr [[TMP71]], align 4
 ; VF-FOUR-CHECK-NEXT:    [[INDEX_NEXT23]] = add nuw i64 [[INDEX20]], 4
 ; VF-FOUR-CHECK-NEXT:    [[TMP73:%.*]] = icmp eq i64 [[INDEX_NEXT23]], [[N_VEC18]]
 ; VF-FOUR-CHECK-NEXT:    br i1 [[TMP73]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
@@ -375,7 +363,6 @@ define dso_local signext i32 @f2(ptr noalias %A, ptr noalias %B, i32 signext %n)
 ; VF-TWO-CHECK-NEXT:    [[TMP78:%.*]] = fadd fast <4 x float> [[REVERSE13]], splat (float 1.000000e+00)
 ; VF-TWO-CHECK-NEXT:    [[TMP79:%.*]] = fadd fast <4 x float> [[REVERSE15]], splat (float 1.000000e+00)
 ; VF-TWO-CHECK-NEXT:    [[TMP80:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; VF-TWO-CHECK-NEXT:    [[TMP88:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i32 0
 ; VF-TWO-CHECK-NEXT:    [[TMP89:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i32 4
 ; VF-TWO-CHECK-NEXT:    [[TMP90:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i32 8
 ; VF-TWO-CHECK-NEXT:    [[TMP91:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i32 12
@@ -383,7 +370,7 @@ define dso_local signext i32 @f2(ptr noalias %A, ptr noalias %B, i32 signext %n)
 ; VF-TWO-CHECK-NEXT:    [[TMP93:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i32 20
 ; VF-TWO-CHECK-NEXT:    [[TMP94:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i32 24
 ; VF-TWO-CHECK-NEXT:    [[TMP95:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i32 28
-; VF-TWO-CHECK-NEXT:    store <4 x float> [[TMP72]], ptr [[TMP88]], align 4
+; VF-TWO-CHECK-NEXT:    store <4 x float> [[TMP72]], ptr [[TMP80]], align 4
 ; VF-TWO-CHECK-NEXT:    store <4 x float> [[TMP73]], ptr [[TMP89]], align 4
 ; VF-TWO-CHECK-NEXT:    store <4 x float> [[TMP74]], ptr [[TMP90]], align 4
 ; VF-TWO-CHECK-NEXT:    store <4 x float> [[TMP75]], ptr [[TMP91]], align 4
@@ -421,8 +408,7 @@ define dso_local signext i32 @f2(ptr noalias %A, ptr noalias %B, i32 signext %n)
 ; VF-TWO-CHECK-NEXT:    [[REVERSE24:%.*]] = shufflevector <2 x float> [[WIDE_LOAD23]], <2 x float> poison, <2 x i32> <i32 1, i32 0>
 ; VF-TWO-CHECK-NEXT:    [[TMP105:%.*]] = fadd fast <2 x float> [[REVERSE24]], splat (float 1.000000e+00)
 ; VF-TWO-CHECK-NEXT:    [[TMP106:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX21]]
-; VF-TWO-CHECK-NEXT:    [[TMP107:%.*]] = getelementptr inbounds float, ptr [[TMP106]], i32 0
-; VF-TWO-CHECK-NEXT:    store <2 x float> [[TMP105]], ptr [[TMP107]], align 4
+; VF-TWO-CHECK-NEXT:    store <2 x float> [[TMP105]], ptr [[TMP106]], align 4
 ; VF-TWO-CHECK-NEXT:    [[INDEX_NEXT25]] = add nuw i64 [[INDEX21]], 2
 ; VF-TWO-CHECK-NEXT:    [[TMP108:%.*]] = icmp eq i64 [[INDEX_NEXT25]], [[N_VEC17]]
 ; VF-TWO-CHECK-NEXT:    br i1 [[TMP108]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
@@ -530,7 +516,6 @@ define dso_local signext i32 @f2(ptr noalias %A, ptr noalias %B, i32 signext %n)
 ; VF-FOUR-CHECK-NEXT:    [[TMP78:%.*]] = fadd fast <4 x float> [[REVERSE13]], splat (float 1.000000e+00)
 ; VF-FOUR-CHECK-NEXT:    [[TMP79:%.*]] = fadd fast <4 x float> [[REVERSE15]], splat (float 1.000000e+00)
 ; VF-FOUR-CHECK-NEXT:    [[TMP80:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; VF-FOUR-CHECK-NEXT:    [[TMP88:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i32 0
 ; VF-FOUR-CHECK-NEXT:    [[TMP89:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i32 4
 ; VF-FOUR-CHECK-NEXT:    [[TMP90:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i32 8
 ; VF-FOUR-CHECK-NEXT:    [[TMP91:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i32 12
@@ -538,7 +523,7 @@ define dso_local signext i32 @f2(ptr noalias %A, ptr noalias %B, i32 signext %n)
 ; VF-FOUR-CHECK-NEXT:    [[TMP93:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i32 20
 ; VF-FOUR-CHECK-NEXT:    [[TMP94:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i32 24
 ; VF-FOUR-CHECK-NEXT:    [[TMP95:%.*]] = getelementptr inbounds float, ptr [[TMP80]], i32 28
-; VF-FOUR-CHECK-NEXT:    store <4 x float> [[TMP72]], ptr [[TMP88]], align 4
+; VF-FOUR-CHECK-NEXT:    store <4 x float> [[TMP72]], ptr [[TMP80]], align 4
 ; VF-FOUR-CHECK-NEXT:    store <4 x float> [[TMP73]], ptr [[TMP89]], align 4
 ; VF-FOUR-CHECK-NEXT:    store <4 x float> [[TMP74]], ptr [[TMP90]], align 4
 ; VF-FOUR-CHECK-NEXT:    store <4 x float> [[TMP75]], ptr [[TMP91]], align 4
@@ -576,8 +561,7 @@ define dso_local signext i32 @f2(ptr noalias %A, ptr noalias %B, i32 signext %n)
 ; VF-FOUR-CHECK-NEXT:    [[REVERSE24:%.*]] = shufflevector <4 x float> [[WIDE_LOAD23]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; VF-FOUR-CHECK-NEXT:    [[TMP105:%.*]] = fadd fast <4 x float> [[REVERSE24]], splat (float 1.000000e+00)
 ; VF-FOUR-CHECK-NEXT:    [[TMP106:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX21]]
-; VF-FOUR-CHECK-NEXT:    [[TMP107:%.*]] = getelementptr inbounds float, ptr [[TMP106]], i32 0
-; VF-FOUR-CHECK-NEXT:    store <4 x float> [[TMP105]], ptr [[TMP107]], align 4
+; VF-FOUR-CHECK-NEXT:    store <4 x float> [[TMP105]], ptr [[TMP106]], align 4
 ; VF-FOUR-CHECK-NEXT:    [[INDEX_NEXT25]] = add nuw i64 [[INDEX21]], 4
 ; VF-FOUR-CHECK-NEXT:    [[TMP108:%.*]] = icmp eq i64 [[INDEX_NEXT25]], [[N_VEC17]]
 ; VF-FOUR-CHECK-NEXT:    br i1 [[TMP108]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/PowerPC/small-loop-rdx.ll b/llvm/test/Transforms/LoopVectorize/PowerPC/small-loop-rdx.ll
index 1e64a20..4dd5403 100644
--- a/llvm/test/Transforms/LoopVectorize/PowerPC/small-loop-rdx.ll
+++ b/llvm/test/Transforms/LoopVectorize/PowerPC/small-loop-rdx.ll
@@ -34,7 +34,6 @@ define void @test(ptr %arr, i32 %len) {
 ; CHECK-NEXT:    [[VEC_PHI7:%.*]] = phi <2 x double> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP18:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI8:%.*]] = phi <2 x double> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP19:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds double, ptr [[ARR]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds double, ptr [[TMP3]], i32 0
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds double, ptr [[TMP3]], i32 2
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds double, ptr [[TMP3]], i32 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds double, ptr [[TMP3]], i32 6
@@ -42,7 +41,7 @@ define void @test(ptr %arr, i32 %len) {
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds double, ptr [[TMP3]], i32 10
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds double, ptr [[TMP3]], i32 12
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds double, ptr [[TMP3]], i32 14
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP4]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP3]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <2 x double>, ptr [[TMP5]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD10:%.*]] = load <2 x double>, ptr [[TMP6]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD11:%.*]] = load <2 x double>, ptr [[TMP7]], align 8
@@ -87,8 +86,7 @@ define void @test(ptr %arr, i32 %len) {
 ; CHECK-NEXT:    [[TMP23:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], %[[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT27:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI25:%.*]] = phi <2 x double> [ [[TMP22]], %[[VEC_EPILOG_PH]] ], [ [[TMP26:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds double, ptr [[ARR]], i64 [[TMP23]]
-; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds double, ptr [[TMP24]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD26:%.*]] = load <2 x double>, ptr [[TMP25]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD26:%.*]] = load <2 x double>, ptr [[TMP24]], align 8
 ; CHECK-NEXT:    [[TMP26]] = fadd fast <2 x double> [[WIDE_LOAD26]], [[VEC_PHI25]]
 ; CHECK-NEXT:    [[INDEX_NEXT27]] = add nuw i64 [[TMP23]], 2
 ; CHECK-NEXT:    [[TMP27:%.*]] = icmp eq i64 [[INDEX_NEXT27]], [[N_VEC23]]
diff --git a/llvm/test/Transforms/LoopVectorize/PowerPC/vectorize-bswap.ll b/llvm/test/Transforms/LoopVectorize/PowerPC/vectorize-bswap.ll
index a515b10..2c85b75 100644
--- a/llvm/test/Transforms/LoopVectorize/PowerPC/vectorize-bswap.ll
+++ b/llvm/test/Transforms/LoopVectorize/PowerPC/vectorize-bswap.ll
@@ -17,13 +17,11 @@ define dso_local void @test(ptr %Arr, i32 signext %Len) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = sext i32 [[INDEX]] to i64
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[ARR:%.*]], i64 [[TMP1]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = call <4 x i32> @llvm.bswap.v4i32(<4 x i32> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[TMP5:%.*]] = sext i32 [[INDEX]] to i64
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[ARR]], i64 [[TMP5]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP4]], ptr [[TMP7]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP4]], ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/bf16.ll b/llvm/test/Transforms/LoopVectorize/RISCV/bf16.ll
index 2b4d8b9..d5b25bf 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/bf16.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/bf16.ll
@@ -40,12 +40,10 @@ define void @fadd(ptr noalias %a, ptr noalias %b, i64 %n) {
 ; ZVFBFMIN-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; ZVFBFMIN-NEXT:    [[TMP1:%.*]] = getelementptr bfloat, ptr [[A]], i64 [[TMP0]]
 ; ZVFBFMIN-NEXT:    [[TMP2:%.*]] = getelementptr bfloat, ptr [[B]], i64 [[TMP0]]
-; ZVFBFMIN-NEXT:    [[TMP3:%.*]] = getelementptr bfloat, ptr [[TMP1]], i32 0
-; ZVFBFMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x bfloat>, ptr [[TMP3]], align 2
-; ZVFBFMIN-NEXT:    [[TMP4:%.*]] = getelementptr bfloat, ptr [[TMP2]], i32 0
-; ZVFBFMIN-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x bfloat>, ptr [[TMP4]], align 2
+; ZVFBFMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x bfloat>, ptr [[TMP1]], align 2
+; ZVFBFMIN-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x bfloat>, ptr [[TMP2]], align 2
 ; ZVFBFMIN-NEXT:    [[TMP11:%.*]] = fadd <vscale x 8 x bfloat> [[WIDE_LOAD]], [[WIDE_LOAD1]]
-; ZVFBFMIN-NEXT:    store <vscale x 8 x bfloat> [[TMP11]], ptr [[TMP3]], align 2
+; ZVFBFMIN-NEXT:    store <vscale x 8 x bfloat> [[TMP11]], ptr [[TMP1]], align 2
 ; ZVFBFMIN-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], [[TMP5]]
 ; ZVFBFMIN-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; ZVFBFMIN-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -101,16 +99,13 @@ define void @vfwmaccbf16.vv(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64
 ; NO-ZVFBFMIN-NEXT:    [[TMP1:%.*]] = getelementptr bfloat, ptr [[A]], i64 [[TMP0]]
 ; NO-ZVFBFMIN-NEXT:    [[TMP2:%.*]] = getelementptr bfloat, ptr [[B]], i64 [[TMP0]]
 ; NO-ZVFBFMIN-NEXT:    [[TMP3:%.*]] = getelementptr float, ptr [[C]], i64 [[TMP0]]
-; NO-ZVFBFMIN-NEXT:    [[TMP4:%.*]] = getelementptr bfloat, ptr [[TMP1]], i32 0
-; NO-ZVFBFMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x bfloat>, ptr [[TMP4]], align 2
-; NO-ZVFBFMIN-NEXT:    [[TMP5:%.*]] = getelementptr bfloat, ptr [[TMP2]], i32 0
-; NO-ZVFBFMIN-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x bfloat>, ptr [[TMP5]], align 2
-; NO-ZVFBFMIN-NEXT:    [[TMP6:%.*]] = getelementptr float, ptr [[TMP3]], i32 0
-; NO-ZVFBFMIN-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP6]], align 4
+; NO-ZVFBFMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x bfloat>, ptr [[TMP1]], align 2
+; NO-ZVFBFMIN-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x bfloat>, ptr [[TMP2]], align 2
+; NO-ZVFBFMIN-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
 ; NO-ZVFBFMIN-NEXT:    [[TMP7:%.*]] = fpext <4 x bfloat> [[WIDE_LOAD]] to <4 x float>
 ; NO-ZVFBFMIN-NEXT:    [[TMP8:%.*]] = fpext <4 x bfloat> [[WIDE_LOAD1]] to <4 x float>
 ; NO-ZVFBFMIN-NEXT:    [[TMP9:%.*]] = call <4 x float> @llvm.fmuladd.v4f32(<4 x float> [[TMP7]], <4 x float> [[TMP8]], <4 x float> [[WIDE_LOAD2]])
-; NO-ZVFBFMIN-NEXT:    store <4 x float> [[TMP9]], ptr [[TMP6]], align 4
+; NO-ZVFBFMIN-NEXT:    store <4 x float> [[TMP9]], ptr [[TMP3]], align 4
 ; NO-ZVFBFMIN-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 4
 ; NO-ZVFBFMIN-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; NO-ZVFBFMIN-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -158,16 +153,13 @@ define void @vfwmaccbf16.vv(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64
 ; ZVFBFMIN-NEXT:    [[TMP7:%.*]] = getelementptr bfloat, ptr [[A]], i64 [[TMP6]]
 ; ZVFBFMIN-NEXT:    [[TMP8:%.*]] = getelementptr bfloat, ptr [[B]], i64 [[TMP6]]
 ; ZVFBFMIN-NEXT:    [[TMP9:%.*]] = getelementptr float, ptr [[C]], i64 [[TMP6]]
-; ZVFBFMIN-NEXT:    [[TMP10:%.*]] = getelementptr bfloat, ptr [[TMP7]], i32 0
-; ZVFBFMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x bfloat>, ptr [[TMP10]], align 2
-; ZVFBFMIN-NEXT:    [[TMP11:%.*]] = getelementptr bfloat, ptr [[TMP8]], i32 0
-; ZVFBFMIN-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x bfloat>, ptr [[TMP11]], align 2
-; ZVFBFMIN-NEXT:    [[TMP12:%.*]] = getelementptr float, ptr [[TMP9]], i32 0
-; ZVFBFMIN-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP12]], align 4
+; ZVFBFMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x bfloat>, ptr [[TMP7]], align 2
+; ZVFBFMIN-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x bfloat>, ptr [[TMP8]], align 2
+; ZVFBFMIN-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 4 x float>, ptr [[TMP9]], align 4
 ; ZVFBFMIN-NEXT:    [[TMP13:%.*]] = fpext <vscale x 4 x bfloat> [[WIDE_LOAD]] to <vscale x 4 x float>
 ; ZVFBFMIN-NEXT:    [[TMP14:%.*]] = fpext <vscale x 4 x bfloat> [[WIDE_LOAD1]] to <vscale x 4 x float>
 ; ZVFBFMIN-NEXT:    [[TMP15:%.*]] = call <vscale x 4 x float> @llvm.fmuladd.nxv4f32(<vscale x 4 x float> [[TMP13]], <vscale x 4 x float> [[TMP14]], <vscale x 4 x float> [[WIDE_LOAD2]])
-; ZVFBFMIN-NEXT:    store <vscale x 4 x float> [[TMP15]], ptr [[TMP12]], align 4
+; ZVFBFMIN-NEXT:    store <vscale x 4 x float> [[TMP15]], ptr [[TMP9]], align 4
 ; ZVFBFMIN-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP6]], [[TMP5]]
 ; ZVFBFMIN-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; ZVFBFMIN-NEXT:    br i1 [[TMP16]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/blocks-with-dead-instructions.ll b/llvm/test/Transforms/LoopVectorize/RISCV/blocks-with-dead-instructions.ll
index f8b83ff..aad9128 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/blocks-with-dead-instructions.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/blocks-with-dead-instructions.ll
@@ -595,11 +595,10 @@ define void @empty_block_with_phi_1(ptr %src, i64 %N) #0 {
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP9:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i16, ptr [[SRC]], i64 [[TMP9]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i16, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i16>, ptr [[TMP11]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i16>, ptr [[TMP10]], align 2
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq <vscale x 8 x i16> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <vscale x 8 x i1> [[TMP12]], <vscale x 8 x i16> splat (i16 99), <vscale x 8 x i16> [[WIDE_LOAD]]
-; CHECK-NEXT:    store <vscale x 8 x i16> [[PREDPHI]], ptr [[TMP11]], align 2
+; CHECK-NEXT:    store <vscale x 8 x i16> [[PREDPHI]], ptr [[TMP10]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP9]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
@@ -674,11 +673,10 @@ define void @empty_block_with_phi_2(ptr %src, i64 %N) #0 {
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP9:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i16, ptr [[SRC]], i64 [[TMP9]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i16, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i16>, ptr [[TMP11]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i16>, ptr [[TMP10]], align 2
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq <vscale x 8 x i16> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <vscale x 8 x i1> [[TMP12]], <vscale x 8 x i16> [[WIDE_LOAD]], <vscale x 8 x i16> splat (i16 99)
-; CHECK-NEXT:    store <vscale x 8 x i16> [[PREDPHI]], ptr [[TMP11]], align 2
+; CHECK-NEXT:    store <vscale x 8 x i16> [[PREDPHI]], ptr [[TMP10]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP9]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/dead-ops-cost.ll b/llvm/test/Transforms/LoopVectorize/RISCV/dead-ops-cost.ll
index 22c56c8..ab8875b 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/dead-ops-cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/dead-ops-cost.ll
@@ -248,8 +248,7 @@ define i32 @cost_of_exit_branch_and_cond_insts(ptr %a, ptr %b, i1 %c, i16 %x) #0
 ; CHECK-NEXT:    store i1 false, ptr [[A]], align 1, !alias.scope [[META11]], !noalias [[META14]]
 ; CHECK-NEXT:    br label %[[PRED_STORE_CONTINUE18]]
 ; CHECK:       [[PRED_STORE_CONTINUE18]]:
-; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr i32, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> zeroinitializer, ptr [[TMP20]], i32 4, <8 x i1> [[BROADCAST_SPLAT]]), !alias.scope [[META14]]
+; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> zeroinitializer, ptr [[TMP11]], i32 4, <8 x i1> [[BROADCAST_SPLAT]]), !alias.scope [[META14]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
 ; CHECK-NEXT:    [[TMP21:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP21]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/defaults.ll b/llvm/test/Transforms/LoopVectorize/RISCV/defaults.ll
index a2faaaa..db3215a6 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/defaults.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/defaults.ll
@@ -30,10 +30,9 @@ define void @vector_add(ptr noalias nocapture %a, i64 %v) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP8]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[TMP9:%.*]] = add <vscale x 2 x i64> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
-; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP9]], ptr [[TMP8]], align 8
+; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP9]], ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -91,8 +90,7 @@ define i64 @vector_add_reduce(ptr noalias nocapture %a) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP8]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[TMP9]] = add <vscale x 2 x i64> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/divrem.ll b/llvm/test/Transforms/LoopVectorize/RISCV/divrem.ll
index c6661cb..f02e5de 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/divrem.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/divrem.ll
@@ -27,10 +27,9 @@ define void @vector_udiv(ptr noalias nocapture %a, i64 %v, i64 %n) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP8]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[TMP9:%.*]] = udiv <vscale x 2 x i64> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
-; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP9]], ptr [[TMP8]], align 8
+; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP9]], ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -62,21 +61,20 @@ define void @vector_udiv(ptr noalias nocapture %a, i64 %v, i64 %n) {
 ; FIXED:       vector.body:
 ; FIXED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; FIXED-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
 ; FIXED-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 4
-; FIXED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; FIXED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; FIXED-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; FIXED-NEXT:    [[TMP4:%.*]] = udiv <4 x i64> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
 ; FIXED-NEXT:    [[TMP5:%.*]] = udiv <4 x i64> [[WIDE_LOAD1]], [[BROADCAST_SPLAT]]
-; FIXED-NEXT:    store <4 x i64> [[TMP4]], ptr [[TMP2]], align 8
+; FIXED-NEXT:    store <4 x i64> [[TMP4]], ptr [[TMP1]], align 8
 ; FIXED-NEXT:    store <4 x i64> [[TMP5]], ptr [[TMP3]], align 8
 ; FIXED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FIXED-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[FOR_END:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[FOR_BODY:%.*]]
 ; FIXED:       for.body:
 ; FIXED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
@@ -127,10 +125,9 @@ define void @vector_sdiv(ptr noalias nocapture %a, i64 %v, i64 %n) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP8]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[TMP9:%.*]] = sdiv <vscale x 2 x i64> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
-; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP9]], ptr [[TMP8]], align 8
+; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP9]], ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -162,21 +159,20 @@ define void @vector_sdiv(ptr noalias nocapture %a, i64 %v, i64 %n) {
 ; FIXED:       vector.body:
 ; FIXED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; FIXED-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
 ; FIXED-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 4
-; FIXED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; FIXED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; FIXED-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; FIXED-NEXT:    [[TMP4:%.*]] = sdiv <4 x i64> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
 ; FIXED-NEXT:    [[TMP5:%.*]] = sdiv <4 x i64> [[WIDE_LOAD1]], [[BROADCAST_SPLAT]]
-; FIXED-NEXT:    store <4 x i64> [[TMP4]], ptr [[TMP2]], align 8
+; FIXED-NEXT:    store <4 x i64> [[TMP4]], ptr [[TMP1]], align 8
 ; FIXED-NEXT:    store <4 x i64> [[TMP5]], ptr [[TMP3]], align 8
 ; FIXED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FIXED-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[FOR_END:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[FOR_BODY:%.*]]
 ; FIXED:       for.body:
 ; FIXED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
@@ -227,10 +223,9 @@ define void @vector_urem(ptr noalias nocapture %a, i64 %v, i64 %n) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP8]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[TMP9:%.*]] = urem <vscale x 2 x i64> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
-; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP9]], ptr [[TMP8]], align 8
+; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP9]], ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
@@ -262,21 +257,20 @@ define void @vector_urem(ptr noalias nocapture %a, i64 %v, i64 %n) {
 ; FIXED:       vector.body:
 ; FIXED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; FIXED-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
 ; FIXED-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 4
-; FIXED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; FIXED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; FIXED-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; FIXED-NEXT:    [[TMP4:%.*]] = urem <4 x i64> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
 ; FIXED-NEXT:    [[TMP5:%.*]] = urem <4 x i64> [[WIDE_LOAD1]], [[BROADCAST_SPLAT]]
-; FIXED-NEXT:    store <4 x i64> [[TMP4]], ptr [[TMP2]], align 8
+; FIXED-NEXT:    store <4 x i64> [[TMP4]], ptr [[TMP1]], align 8
 ; FIXED-NEXT:    store <4 x i64> [[TMP5]], ptr [[TMP3]], align 8
 ; FIXED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FIXED-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[FOR_END:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[FOR_BODY:%.*]]
 ; FIXED:       for.body:
 ; FIXED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
@@ -327,10 +321,9 @@ define void @vector_srem(ptr noalias nocapture %a, i64 %v, i64 %n) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP8]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[TMP9:%.*]] = srem <vscale x 2 x i64> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
-; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP9]], ptr [[TMP8]], align 8
+; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP9]], ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
@@ -362,21 +355,20 @@ define void @vector_srem(ptr noalias nocapture %a, i64 %v, i64 %n) {
 ; FIXED:       vector.body:
 ; FIXED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; FIXED-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
 ; FIXED-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 4
-; FIXED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; FIXED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; FIXED-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; FIXED-NEXT:    [[TMP4:%.*]] = srem <4 x i64> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
 ; FIXED-NEXT:    [[TMP5:%.*]] = srem <4 x i64> [[WIDE_LOAD1]], [[BROADCAST_SPLAT]]
-; FIXED-NEXT:    store <4 x i64> [[TMP4]], ptr [[TMP2]], align 8
+; FIXED-NEXT:    store <4 x i64> [[TMP4]], ptr [[TMP1]], align 8
 ; FIXED-NEXT:    store <4 x i64> [[TMP5]], ptr [[TMP3]], align 8
 ; FIXED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FIXED-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[FOR_END:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[FOR_BODY:%.*]]
 ; FIXED:       for.body:
 ; FIXED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
@@ -429,11 +421,10 @@ define void @predicated_udiv(ptr noalias nocapture %a, i64 %v, i64 %n) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP9]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP8]], align 8
 ; CHECK-NEXT:    [[TMP11:%.*]] = udiv <vscale x 2 x i64> [[WIDE_LOAD]], [[TMP10]]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <vscale x 2 x i1> [[TMP6]], <vscale x 2 x i64> [[TMP11]], <vscale x 2 x i64> [[WIDE_LOAD]]
-; CHECK-NEXT:    store <vscale x 2 x i64> [[PREDPHI]], ptr [[TMP9]], align 8
+; CHECK-NEXT:    store <vscale x 2 x i64> [[PREDPHI]], ptr [[TMP8]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
@@ -473,23 +464,22 @@ define void @predicated_udiv(ptr noalias nocapture %a, i64 %v, i64 %n) {
 ; FIXED:       vector.body:
 ; FIXED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; FIXED-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
 ; FIXED-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 4
-; FIXED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
+; FIXED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
 ; FIXED-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
 ; FIXED-NEXT:    [[TMP7:%.*]] = udiv <4 x i64> [[WIDE_LOAD]], [[TMP5]]
 ; FIXED-NEXT:    [[TMP8:%.*]] = udiv <4 x i64> [[WIDE_LOAD1]], [[TMP5]]
 ; FIXED-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP0]], <4 x i64> [[TMP7]], <4 x i64> [[WIDE_LOAD]]
 ; FIXED-NEXT:    [[PREDPHI2:%.*]] = select <4 x i1> [[TMP0]], <4 x i64> [[TMP8]], <4 x i64> [[WIDE_LOAD1]]
-; FIXED-NEXT:    store <4 x i64> [[PREDPHI]], ptr [[TMP3]], align 8
+; FIXED-NEXT:    store <4 x i64> [[PREDPHI]], ptr [[TMP2]], align 8
 ; FIXED-NEXT:    store <4 x i64> [[PREDPHI2]], ptr [[TMP4]], align 8
 ; FIXED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FIXED-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[FOR_END:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[FOR_BODY:%.*]]
 ; FIXED:       for.body:
 ; FIXED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
@@ -554,11 +544,10 @@ define void @predicated_sdiv(ptr noalias nocapture %a, i64 %v, i64 %n) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP9]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP8]], align 8
 ; CHECK-NEXT:    [[TMP11:%.*]] = sdiv <vscale x 2 x i64> [[WIDE_LOAD]], [[TMP10]]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <vscale x 2 x i1> [[TMP6]], <vscale x 2 x i64> [[TMP11]], <vscale x 2 x i64> [[WIDE_LOAD]]
-; CHECK-NEXT:    store <vscale x 2 x i64> [[PREDPHI]], ptr [[TMP9]], align 8
+; CHECK-NEXT:    store <vscale x 2 x i64> [[PREDPHI]], ptr [[TMP8]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
@@ -598,23 +587,22 @@ define void @predicated_sdiv(ptr noalias nocapture %a, i64 %v, i64 %n) {
 ; FIXED:       vector.body:
 ; FIXED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; FIXED-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
 ; FIXED-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 4
-; FIXED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
+; FIXED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
 ; FIXED-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
 ; FIXED-NEXT:    [[TMP7:%.*]] = sdiv <4 x i64> [[WIDE_LOAD]], [[TMP5]]
 ; FIXED-NEXT:    [[TMP8:%.*]] = sdiv <4 x i64> [[WIDE_LOAD1]], [[TMP5]]
 ; FIXED-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP0]], <4 x i64> [[TMP7]], <4 x i64> [[WIDE_LOAD]]
 ; FIXED-NEXT:    [[PREDPHI2:%.*]] = select <4 x i1> [[TMP0]], <4 x i64> [[TMP8]], <4 x i64> [[WIDE_LOAD1]]
-; FIXED-NEXT:    store <4 x i64> [[PREDPHI]], ptr [[TMP3]], align 8
+; FIXED-NEXT:    store <4 x i64> [[PREDPHI]], ptr [[TMP2]], align 8
 ; FIXED-NEXT:    store <4 x i64> [[PREDPHI2]], ptr [[TMP4]], align 8
 ; FIXED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FIXED-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[FOR_END:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[FOR_BODY:%.*]]
 ; FIXED:       for.body:
 ; FIXED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
@@ -675,12 +663,11 @@ define void @predicated_udiv_by_constant(ptr noalias nocapture %a, i64 %n) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP8]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp ne <vscale x 2 x i64> [[WIDE_LOAD]], splat (i64 42)
 ; CHECK-NEXT:    [[TMP10:%.*]] = udiv <vscale x 2 x i64> [[WIDE_LOAD]], splat (i64 27)
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <vscale x 2 x i1> [[TMP9]], <vscale x 2 x i64> [[TMP10]], <vscale x 2 x i64> [[WIDE_LOAD]]
-; CHECK-NEXT:    store <vscale x 2 x i64> [[PREDPHI]], ptr [[TMP8]], align 8
+; CHECK-NEXT:    store <vscale x 2 x i64> [[PREDPHI]], ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
@@ -716,9 +703,8 @@ define void @predicated_udiv_by_constant(ptr noalias nocapture %a, i64 %n) {
 ; FIXED:       vector.body:
 ; FIXED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; FIXED-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
 ; FIXED-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 4
-; FIXED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; FIXED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; FIXED-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; FIXED-NEXT:    [[TMP4:%.*]] = icmp ne <4 x i64> [[WIDE_LOAD]], splat (i64 42)
 ; FIXED-NEXT:    [[TMP5:%.*]] = icmp ne <4 x i64> [[WIDE_LOAD1]], splat (i64 42)
@@ -726,15 +712,15 @@ define void @predicated_udiv_by_constant(ptr noalias nocapture %a, i64 %n) {
 ; FIXED-NEXT:    [[TMP7:%.*]] = udiv <4 x i64> [[WIDE_LOAD1]], splat (i64 27)
 ; FIXED-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP4]], <4 x i64> [[TMP6]], <4 x i64> [[WIDE_LOAD]]
 ; FIXED-NEXT:    [[PREDPHI2:%.*]] = select <4 x i1> [[TMP5]], <4 x i64> [[TMP7]], <4 x i64> [[WIDE_LOAD1]]
-; FIXED-NEXT:    store <4 x i64> [[PREDPHI]], ptr [[TMP2]], align 8
+; FIXED-NEXT:    store <4 x i64> [[PREDPHI]], ptr [[TMP1]], align 8
 ; FIXED-NEXT:    store <4 x i64> [[PREDPHI2]], ptr [[TMP3]], align 8
 ; FIXED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FIXED-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[FOR_END:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[FOR_BODY:%.*]]
 ; FIXED:       for.body:
 ; FIXED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
@@ -795,12 +781,11 @@ define void @predicated_sdiv_by_constant(ptr noalias nocapture %a, i64 %n) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP8]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp ne <vscale x 2 x i64> [[WIDE_LOAD]], splat (i64 42)
 ; CHECK-NEXT:    [[TMP10:%.*]] = sdiv <vscale x 2 x i64> [[WIDE_LOAD]], splat (i64 27)
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <vscale x 2 x i1> [[TMP9]], <vscale x 2 x i64> [[TMP10]], <vscale x 2 x i64> [[WIDE_LOAD]]
-; CHECK-NEXT:    store <vscale x 2 x i64> [[PREDPHI]], ptr [[TMP8]], align 8
+; CHECK-NEXT:    store <vscale x 2 x i64> [[PREDPHI]], ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
@@ -836,9 +821,8 @@ define void @predicated_sdiv_by_constant(ptr noalias nocapture %a, i64 %n) {
 ; FIXED:       vector.body:
 ; FIXED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; FIXED-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
 ; FIXED-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 4
-; FIXED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; FIXED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; FIXED-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; FIXED-NEXT:    [[TMP4:%.*]] = icmp ne <4 x i64> [[WIDE_LOAD]], splat (i64 42)
 ; FIXED-NEXT:    [[TMP5:%.*]] = icmp ne <4 x i64> [[WIDE_LOAD1]], splat (i64 42)
@@ -846,15 +830,15 @@ define void @predicated_sdiv_by_constant(ptr noalias nocapture %a, i64 %n) {
 ; FIXED-NEXT:    [[TMP7:%.*]] = sdiv <4 x i64> [[WIDE_LOAD1]], splat (i64 27)
 ; FIXED-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP4]], <4 x i64> [[TMP6]], <4 x i64> [[WIDE_LOAD]]
 ; FIXED-NEXT:    [[PREDPHI2:%.*]] = select <4 x i1> [[TMP5]], <4 x i64> [[TMP7]], <4 x i64> [[WIDE_LOAD1]]
-; FIXED-NEXT:    store <4 x i64> [[PREDPHI]], ptr [[TMP2]], align 8
+; FIXED-NEXT:    store <4 x i64> [[PREDPHI]], ptr [[TMP1]], align 8
 ; FIXED-NEXT:    store <4 x i64> [[PREDPHI2]], ptr [[TMP3]], align 8
 ; FIXED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FIXED-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[FOR_END:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[FOR_BODY:%.*]]
 ; FIXED:       for.body:
 ; FIXED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
@@ -915,13 +899,12 @@ define void @predicated_sdiv_by_minus_one(ptr noalias nocapture %a, i64 %n) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP8]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP7]], align 1
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp ne <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 -128)
 ; CHECK-NEXT:    [[TMP10:%.*]] = select <vscale x 16 x i1> [[TMP9]], <vscale x 16 x i8> splat (i8 -1), <vscale x 16 x i8> splat (i8 1)
 ; CHECK-NEXT:    [[TMP11:%.*]] = sdiv <vscale x 16 x i8> [[WIDE_LOAD]], [[TMP10]]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <vscale x 16 x i1> [[TMP9]], <vscale x 16 x i8> [[TMP11]], <vscale x 16 x i8> [[WIDE_LOAD]]
-; CHECK-NEXT:    store <vscale x 16 x i8> [[PREDPHI]], ptr [[TMP8]], align 1
+; CHECK-NEXT:    store <vscale x 16 x i8> [[PREDPHI]], ptr [[TMP7]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
@@ -957,9 +940,8 @@ define void @predicated_sdiv_by_minus_one(ptr noalias nocapture %a, i64 %n) {
 ; FIXED:       vector.body:
 ; FIXED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[INDEX]]
-; FIXED-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
 ; FIXED-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 32
-; FIXED-NEXT:    [[WIDE_LOAD:%.*]] = load <32 x i8>, ptr [[TMP2]], align 1
+; FIXED-NEXT:    [[WIDE_LOAD:%.*]] = load <32 x i8>, ptr [[TMP1]], align 1
 ; FIXED-NEXT:    [[WIDE_LOAD1:%.*]] = load <32 x i8>, ptr [[TMP3]], align 1
 ; FIXED-NEXT:    [[TMP4:%.*]] = icmp ne <32 x i8> [[WIDE_LOAD]], splat (i8 -128)
 ; FIXED-NEXT:    [[TMP5:%.*]] = icmp ne <32 x i8> [[WIDE_LOAD1]], splat (i8 -128)
@@ -969,15 +951,15 @@ define void @predicated_sdiv_by_minus_one(ptr noalias nocapture %a, i64 %n) {
 ; FIXED-NEXT:    [[TMP9:%.*]] = sdiv <32 x i8> [[WIDE_LOAD1]], [[TMP7]]
 ; FIXED-NEXT:    [[PREDPHI:%.*]] = select <32 x i1> [[TMP4]], <32 x i8> [[TMP8]], <32 x i8> [[WIDE_LOAD]]
 ; FIXED-NEXT:    [[PREDPHI2:%.*]] = select <32 x i1> [[TMP5]], <32 x i8> [[TMP9]], <32 x i8> [[WIDE_LOAD1]]
-; FIXED-NEXT:    store <32 x i8> [[PREDPHI]], ptr [[TMP2]], align 1
+; FIXED-NEXT:    store <32 x i8> [[PREDPHI]], ptr [[TMP1]], align 1
 ; FIXED-NEXT:    store <32 x i8> [[PREDPHI2]], ptr [[TMP3]], align 1
 ; FIXED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 64
 ; FIXED-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[FOR_END:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[FOR_BODY:%.*]]
 ; FIXED:       for.body:
 ; FIXED-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/evl-compatible-loops.ll b/llvm/test/Transforms/LoopVectorize/RISCV/evl-compatible-loops.ll
index 25f52b2..5d032e6 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/evl-compatible-loops.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/evl-compatible-loops.ll
@@ -1,6 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
-; RUN: opt -passes=loop-vectorize -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: opt -passes=loop-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s
 
 ; Make sure we do not vectorize a loop with a widened int induction.
@@ -33,8 +33,7 @@ define void @test_wide_integer_induction(ptr noalias %a, i64 %N) {
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP13]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[EVL_BASED_IV]]
-; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i64, ptr [[TMP14]], i32 0
-; CHECK-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VEC_IND]], ptr align 8 [[TMP15]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP11]])
+; CHECK-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VEC_IND]], ptr align 8 [[TMP14]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP11]])
 ; CHECK-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP11]] to i64
 ; CHECK-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP16]], [[EVL_BASED_IV]]
 ; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], [[TMP8]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/f16.ll b/llvm/test/Transforms/LoopVectorize/RISCV/f16.ll
index 0b307c2..5b56552 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/f16.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/f16.ll
@@ -40,12 +40,10 @@ define void @fadd(ptr noalias %a, ptr noalias %b, i64 %n) {
 ; ZVFHMIN-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; ZVFHMIN-NEXT:    [[TMP1:%.*]] = getelementptr half, ptr [[A]], i64 [[INDEX]]
 ; ZVFHMIN-NEXT:    [[TMP2:%.*]] = getelementptr half, ptr [[B]], i64 [[INDEX]]
-; ZVFHMIN-NEXT:    [[TMP3:%.*]] = getelementptr half, ptr [[TMP1]], i32 0
-; ZVFHMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x half>, ptr [[TMP3]], align 2
-; ZVFHMIN-NEXT:    [[TMP4:%.*]] = getelementptr half, ptr [[TMP2]], i32 0
-; ZVFHMIN-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x half>, ptr [[TMP4]], align 2
+; ZVFHMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x half>, ptr [[TMP1]], align 2
+; ZVFHMIN-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 8 x half>, ptr [[TMP2]], align 2
 ; ZVFHMIN-NEXT:    [[TMP11:%.*]] = fadd <vscale x 8 x half> [[WIDE_LOAD]], [[WIDE_LOAD1]]
-; ZVFHMIN-NEXT:    store <vscale x 8 x half> [[TMP11]], ptr [[TMP3]], align 2
+; ZVFHMIN-NEXT:    store <vscale x 8 x half> [[TMP11]], ptr [[TMP1]], align 2
 ; ZVFHMIN-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; ZVFHMIN-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; ZVFHMIN-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/first-order-recurrence-scalable-vf1.ll b/llvm/test/Transforms/LoopVectorize/RISCV/first-order-recurrence-scalable-vf1.ll
index 1f6d71e..ce58ae1 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/first-order-recurrence-scalable-vf1.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/first-order-recurrence-scalable-vf1.ll
@@ -24,12 +24,10 @@ define i64 @pr97452_scalable_vf1_for(ptr %src, ptr noalias %dst) #0 {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 1 x i64> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[SRC]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD]] = load <vscale x 1 x i64>, ptr [[TMP6]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <vscale x 1 x i64>, ptr [[TMP5]], align 8
 ; CHECK-NEXT:    [[TMP7:%.*]] = call <vscale x 1 x i64> @llvm.vector.splice.nxv1i64(<vscale x 1 x i64> [[VECTOR_RECUR]], <vscale x 1 x i64> [[WIDE_LOAD]], i32 -1)
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <vscale x 1 x i64> [[TMP7]], ptr [[TMP9]], align 8
+; CHECK-NEXT:    store <vscale x 1 x i64> [[TMP7]], ptr [[TMP8]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP2]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/fminimumnum.ll b/llvm/test/Transforms/LoopVectorize/RISCV/fminimumnum.ll
index 283688c..c9ba2af 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/fminimumnum.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/fminimumnum.ll
@@ -37,15 +37,12 @@ define void @fmin32(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw float, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP3]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw float, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 4 x float>, ptr [[TMP6]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 4 x float>, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[TMP17:%.*]] = call <vscale x 4 x float> @llvm.minimumnum.nxv4f32(<vscale x 4 x float> [[WIDE_LOAD]], <vscale x 4 x float> [[WIDE_LOAD5]])
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw float, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <vscale x 4 x float> [[TMP17]], ptr [[TMP11]], align 4
+; CHECK-NEXT:    store <vscale x 4 x float> [[TMP17]], ptr [[TMP10]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -103,15 +100,12 @@ define void @fmin32(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; ZVFHMIN:       [[VECTOR_BODY]]:
 ; ZVFHMIN-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; ZVFHMIN-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; ZVFHMIN-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw float, ptr [[TMP13]], i32 0
-; ZVFHMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP14]], align 4
+; ZVFHMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP13]], align 4
 ; ZVFHMIN-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; ZVFHMIN-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw float, ptr [[TMP15]], i32 0
-; ZVFHMIN-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 4 x float>, ptr [[TMP16]], align 4
+; ZVFHMIN-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 4 x float>, ptr [[TMP15]], align 4
 ; ZVFHMIN-NEXT:    [[TMP17:%.*]] = call <vscale x 4 x float> @llvm.minimumnum.nxv4f32(<vscale x 4 x float> [[WIDE_LOAD]], <vscale x 4 x float> [[WIDE_LOAD5]])
 ; ZVFHMIN-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; ZVFHMIN-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw float, ptr [[TMP18]], i32 0
-; ZVFHMIN-NEXT:    store <vscale x 4 x float> [[TMP17]], ptr [[TMP19]], align 4
+; ZVFHMIN-NEXT:    store <vscale x 4 x float> [[TMP17]], ptr [[TMP18]], align 4
 ; ZVFHMIN-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
 ; ZVFHMIN-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; ZVFHMIN-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -192,15 +186,12 @@ define void @fmax32(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw float, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP3]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw float, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 4 x float>, ptr [[TMP6]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 4 x float>, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[TMP17:%.*]] = call <vscale x 4 x float> @llvm.maximumnum.nxv4f32(<vscale x 4 x float> [[WIDE_LOAD]], <vscale x 4 x float> [[WIDE_LOAD5]])
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw float, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <vscale x 4 x float> [[TMP17]], ptr [[TMP11]], align 4
+; CHECK-NEXT:    store <vscale x 4 x float> [[TMP17]], ptr [[TMP10]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -258,15 +249,12 @@ define void @fmax32(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; ZVFHMIN:       [[VECTOR_BODY]]:
 ; ZVFHMIN-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; ZVFHMIN-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; ZVFHMIN-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw float, ptr [[TMP13]], i32 0
-; ZVFHMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP14]], align 4
+; ZVFHMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP13]], align 4
 ; ZVFHMIN-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; ZVFHMIN-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw float, ptr [[TMP15]], i32 0
-; ZVFHMIN-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 4 x float>, ptr [[TMP16]], align 4
+; ZVFHMIN-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 4 x float>, ptr [[TMP15]], align 4
 ; ZVFHMIN-NEXT:    [[TMP17:%.*]] = call <vscale x 4 x float> @llvm.maximumnum.nxv4f32(<vscale x 4 x float> [[WIDE_LOAD]], <vscale x 4 x float> [[WIDE_LOAD5]])
 ; ZVFHMIN-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; ZVFHMIN-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw float, ptr [[TMP18]], i32 0
-; ZVFHMIN-NEXT:    store <vscale x 4 x float> [[TMP17]], ptr [[TMP19]], align 4
+; ZVFHMIN-NEXT:    store <vscale x 4 x float> [[TMP17]], ptr [[TMP18]], align 4
 ; ZVFHMIN-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
 ; ZVFHMIN-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; ZVFHMIN-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -347,15 +335,12 @@ define void @fmin64(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw double, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x double>, ptr [[TMP3]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x double>, ptr [[TMP2]], align 8
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw double, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 2 x double>, ptr [[TMP6]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 2 x double>, ptr [[TMP5]], align 8
 ; CHECK-NEXT:    [[TMP17:%.*]] = call <vscale x 2 x double> @llvm.minimumnum.nxv2f64(<vscale x 2 x double> [[WIDE_LOAD]], <vscale x 2 x double> [[WIDE_LOAD5]])
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw double, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <vscale x 2 x double> [[TMP17]], ptr [[TMP11]], align 8
+; CHECK-NEXT:    store <vscale x 2 x double> [[TMP17]], ptr [[TMP10]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
@@ -413,15 +398,12 @@ define void @fmin64(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; ZVFHMIN:       [[VECTOR_BODY]]:
 ; ZVFHMIN-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; ZVFHMIN-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; ZVFHMIN-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw double, ptr [[TMP13]], i32 0
-; ZVFHMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x double>, ptr [[TMP14]], align 8
+; ZVFHMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x double>, ptr [[TMP13]], align 8
 ; ZVFHMIN-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; ZVFHMIN-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw double, ptr [[TMP15]], i32 0
-; ZVFHMIN-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 2 x double>, ptr [[TMP16]], align 8
+; ZVFHMIN-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 2 x double>, ptr [[TMP15]], align 8
 ; ZVFHMIN-NEXT:    [[TMP17:%.*]] = call <vscale x 2 x double> @llvm.minimumnum.nxv2f64(<vscale x 2 x double> [[WIDE_LOAD]], <vscale x 2 x double> [[WIDE_LOAD5]])
 ; ZVFHMIN-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; ZVFHMIN-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw double, ptr [[TMP18]], i32 0
-; ZVFHMIN-NEXT:    store <vscale x 2 x double> [[TMP17]], ptr [[TMP19]], align 8
+; ZVFHMIN-NEXT:    store <vscale x 2 x double> [[TMP17]], ptr [[TMP18]], align 8
 ; ZVFHMIN-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
 ; ZVFHMIN-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; ZVFHMIN-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
@@ -502,15 +484,12 @@ define void @fmax64(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw double, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x double>, ptr [[TMP3]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x double>, ptr [[TMP2]], align 8
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw double, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 2 x double>, ptr [[TMP6]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 2 x double>, ptr [[TMP5]], align 8
 ; CHECK-NEXT:    [[TMP17:%.*]] = call <vscale x 2 x double> @llvm.maximumnum.nxv2f64(<vscale x 2 x double> [[WIDE_LOAD]], <vscale x 2 x double> [[WIDE_LOAD5]])
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw double, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <vscale x 2 x double> [[TMP17]], ptr [[TMP11]], align 8
+; CHECK-NEXT:    store <vscale x 2 x double> [[TMP17]], ptr [[TMP10]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
@@ -568,15 +547,12 @@ define void @fmax64(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; ZVFHMIN:       [[VECTOR_BODY]]:
 ; ZVFHMIN-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; ZVFHMIN-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; ZVFHMIN-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw double, ptr [[TMP13]], i32 0
-; ZVFHMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x double>, ptr [[TMP14]], align 8
+; ZVFHMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x double>, ptr [[TMP13]], align 8
 ; ZVFHMIN-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; ZVFHMIN-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw double, ptr [[TMP15]], i32 0
-; ZVFHMIN-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 2 x double>, ptr [[TMP16]], align 8
+; ZVFHMIN-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 2 x double>, ptr [[TMP15]], align 8
 ; ZVFHMIN-NEXT:    [[TMP17:%.*]] = call <vscale x 2 x double> @llvm.maximumnum.nxv2f64(<vscale x 2 x double> [[WIDE_LOAD]], <vscale x 2 x double> [[WIDE_LOAD5]])
 ; ZVFHMIN-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; ZVFHMIN-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw double, ptr [[TMP18]], i32 0
-; ZVFHMIN-NEXT:    store <vscale x 2 x double> [[TMP17]], ptr [[TMP19]], align 8
+; ZVFHMIN-NEXT:    store <vscale x 2 x double> [[TMP17]], ptr [[TMP18]], align 8
 ; ZVFHMIN-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
 ; ZVFHMIN-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; ZVFHMIN-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
@@ -657,15 +633,12 @@ define void @fmin16(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw half, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x half>, ptr [[TMP3]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x half>, ptr [[TMP2]], align 2
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw half, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x half>, ptr [[TMP5]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x half>, ptr [[TMP4]], align 2
 ; CHECK-NEXT:    [[TMP17:%.*]] = call <vscale x 8 x half> @llvm.minimumnum.nxv8f16(<vscale x 8 x half> [[WIDE_LOAD]], <vscale x 8 x half> [[WIDE_LOAD5]])
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw half, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <vscale x 8 x half> [[TMP17]], ptr [[TMP8]], align 2
+; CHECK-NEXT:    store <vscale x 8 x half> [[TMP17]], ptr [[TMP7]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
@@ -723,15 +696,12 @@ define void @fmin16(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; ZVFHMIN:       [[VECTOR_BODY]]:
 ; ZVFHMIN-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; ZVFHMIN-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; ZVFHMIN-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw half, ptr [[TMP13]], i32 0
-; ZVFHMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x half>, ptr [[TMP14]], align 2
+; ZVFHMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x half>, ptr [[TMP13]], align 2
 ; ZVFHMIN-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; ZVFHMIN-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw half, ptr [[TMP15]], i32 0
-; ZVFHMIN-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x half>, ptr [[TMP16]], align 2
+; ZVFHMIN-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x half>, ptr [[TMP15]], align 2
 ; ZVFHMIN-NEXT:    [[TMP17:%.*]] = call <vscale x 8 x half> @llvm.minimumnum.nxv8f16(<vscale x 8 x half> [[WIDE_LOAD]], <vscale x 8 x half> [[WIDE_LOAD5]])
 ; ZVFHMIN-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; ZVFHMIN-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw half, ptr [[TMP18]], i32 0
-; ZVFHMIN-NEXT:    store <vscale x 8 x half> [[TMP17]], ptr [[TMP19]], align 2
+; ZVFHMIN-NEXT:    store <vscale x 8 x half> [[TMP17]], ptr [[TMP18]], align 2
 ; ZVFHMIN-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
 ; ZVFHMIN-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; ZVFHMIN-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
@@ -812,15 +782,12 @@ define void @fmax16(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw half, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x half>, ptr [[TMP3]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x half>, ptr [[TMP2]], align 2
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw half, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x half>, ptr [[TMP5]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x half>, ptr [[TMP4]], align 2
 ; CHECK-NEXT:    [[TMP17:%.*]] = call <vscale x 8 x half> @llvm.maximumnum.nxv8f16(<vscale x 8 x half> [[WIDE_LOAD]], <vscale x 8 x half> [[WIDE_LOAD5]])
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw half, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <vscale x 8 x half> [[TMP17]], ptr [[TMP8]], align 2
+; CHECK-NEXT:    store <vscale x 8 x half> [[TMP17]], ptr [[TMP7]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
@@ -878,15 +845,12 @@ define void @fmax16(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; ZVFHMIN:       [[VECTOR_BODY]]:
 ; ZVFHMIN-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; ZVFHMIN-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; ZVFHMIN-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw half, ptr [[TMP13]], i32 0
-; ZVFHMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x half>, ptr [[TMP14]], align 2
+; ZVFHMIN-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x half>, ptr [[TMP13]], align 2
 ; ZVFHMIN-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; ZVFHMIN-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw half, ptr [[TMP15]], i32 0
-; ZVFHMIN-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x half>, ptr [[TMP16]], align 2
+; ZVFHMIN-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 8 x half>, ptr [[TMP15]], align 2
 ; ZVFHMIN-NEXT:    [[TMP17:%.*]] = call <vscale x 8 x half> @llvm.maximumnum.nxv8f16(<vscale x 8 x half> [[WIDE_LOAD]], <vscale x 8 x half> [[WIDE_LOAD5]])
 ; ZVFHMIN-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; ZVFHMIN-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw half, ptr [[TMP18]], i32 0
-; ZVFHMIN-NEXT:    store <vscale x 8 x half> [[TMP17]], ptr [[TMP19]], align 2
+; ZVFHMIN-NEXT:    store <vscale x 8 x half> [[TMP17]], ptr [[TMP18]], align 2
 ; ZVFHMIN-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
 ; ZVFHMIN-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; ZVFHMIN-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
index d485a743..5b2d308 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll
@@ -1,8 +1,8 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -mtriple riscv64-linux-gnu -mattr=+v,+d -passes=loop-vectorize < %s -S -o - | FileCheck %s -check-prefix=OUTLOOP
 ; RUN: opt -mtriple riscv64-linux-gnu -mattr=+v,+d -passes=loop-vectorize -prefer-inloop-reductions < %s -S -o - | FileCheck %s -check-prefix=INLOOP
-; RUN: opt -passes=loop-vectorize -force-tail-folding-style=data-with-evl -prefer-predicate-over-epilogue=predicate-dont-vectorize -mtriple=riscv64 -mattr=+v -S < %s 2>&1 | FileCheck --check-prefix=IF-EVL-OUTLOOP %s
-; RUN: opt -passes=loop-vectorize -prefer-inloop-reductions -force-tail-folding-style=data-with-evl -prefer-predicate-over-epilogue=predicate-dont-vectorize -mtriple=riscv64 -mattr=+v -S < %s 2>&1 | FileCheck --check-prefix=IF-EVL-INLOOP %s
+; RUN: opt -passes=loop-vectorize -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue -mtriple=riscv64 -mattr=+v -S < %s 2>&1 | FileCheck --check-prefix=IF-EVL-OUTLOOP %s
+; RUN: opt -passes=loop-vectorize -prefer-inloop-reductions -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue -mtriple=riscv64 -mattr=+v -S < %s 2>&1 | FileCheck --check-prefix=IF-EVL-INLOOP %s
 
 
 target datalayout = "e-m:e-p:64:64-i64:64-i128:128-n64-S128"
@@ -30,8 +30,7 @@ define i32 @add_i16_i32(ptr nocapture readonly %x, i32 %n) {
 ; OUTLOOP-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; OUTLOOP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; OUTLOOP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i16, ptr [[X:%.*]], i32 [[INDEX]]
-; OUTLOOP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[TMP7]], i32 0
-; OUTLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP8]], align 2
+; OUTLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP7]], align 2
 ; OUTLOOP-NEXT:    [[TMP9:%.*]] = sext <vscale x 4 x i16> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; OUTLOOP-NEXT:    [[TMP10]] = add <vscale x 4 x i32> [[VEC_PHI]], [[TMP9]]
 ; OUTLOOP-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP5]]
@@ -83,8 +82,7 @@ define i32 @add_i16_i32(ptr nocapture readonly %x, i32 %n) {
 ; INLOOP-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; INLOOP-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP11:%.*]], [[VECTOR_BODY]] ]
 ; INLOOP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i16, ptr [[X:%.*]], i32 [[INDEX]]
-; INLOOP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[TMP7]], i32 0
-; INLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i16>, ptr [[TMP8]], align 2
+; INLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i16>, ptr [[TMP7]], align 2
 ; INLOOP-NEXT:    [[TMP9:%.*]] = sext <vscale x 8 x i16> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; INLOOP-NEXT:    [[TMP10:%.*]] = call i32 @llvm.vector.reduce.add.nxv8i32(<vscale x 8 x i32> [[TMP9]])
 ; INLOOP-NEXT:    [[TMP11]] = add i32 [[TMP10]], [[VEC_PHI]]
@@ -138,8 +136,7 @@ define i32 @add_i16_i32(ptr nocapture readonly %x, i32 %n) {
 ; IF-EVL-OUTLOOP-NEXT:    [[AVL:%.*]] = sub i32 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP5:%.*]] = call i32 @llvm.experimental.get.vector.length.i32(i32 [[AVL]], i32 4, i1 true)
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i16, ptr [[X:%.*]], i32 [[EVL_BASED_IV]]
-; IF-EVL-OUTLOOP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[TMP7]], i32 0
-; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i16> @llvm.vp.load.nxv4i16.p0(ptr align 2 [[TMP8]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP5]])
+; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i16> @llvm.vp.load.nxv4i16.p0(ptr align 2 [[TMP7]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP5]])
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP9:%.*]] = sext <vscale x 4 x i16> [[VP_OP_LOAD]] to <vscale x 4 x i32>
 ; IF-EVL-OUTLOOP-NEXT:    [[VP_OP:%.*]] = add <vscale x 4 x i32> [[VEC_PHI]], [[TMP9]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP10]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[VP_OP]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP5]])
@@ -194,8 +191,7 @@ define i32 @add_i16_i32(ptr nocapture readonly %x, i32 %n) {
 ; IF-EVL-INLOOP-NEXT:    [[TMP5:%.*]] = sub i32 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-INLOOP-NEXT:    [[TMP6:%.*]] = call i32 @llvm.experimental.get.vector.length.i32(i32 [[TMP5]], i32 8, i1 true)
 ; IF-EVL-INLOOP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[X:%.*]], i32 [[EVL_BASED_IV]]
-; IF-EVL-INLOOP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i16, ptr [[TMP8]], i32 0
-; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 8 x i16> @llvm.vp.load.nxv8i16.p0(ptr align 2 [[TMP9]], <vscale x 8 x i1> splat (i1 true), i32 [[TMP6]])
+; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 8 x i16> @llvm.vp.load.nxv8i16.p0(ptr align 2 [[TMP8]], <vscale x 8 x i1> splat (i1 true), i32 [[TMP6]])
 ; IF-EVL-INLOOP-NEXT:    [[TMP14:%.*]] = sext <vscale x 8 x i16> [[VP_OP_LOAD]] to <vscale x 8 x i32>
 ; IF-EVL-INLOOP-NEXT:    [[TMP10:%.*]] = call i32 @llvm.vp.reduce.add.nxv8i32(i32 0, <vscale x 8 x i32> [[TMP14]], <vscale x 8 x i1> splat (i1 true), i32 [[TMP6]])
 ; IF-EVL-INLOOP-NEXT:    [[TMP11]] = add i32 [[TMP10]], [[VEC_PHI]]
@@ -267,8 +263,7 @@ define i32 @smin(ptr %a, i64 %n, i32 %start) {
 ; OUTLOOP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; OUTLOOP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; OUTLOOP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; OUTLOOP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; OUTLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; OUTLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; OUTLOOP-NEXT:    [[TMP9:%.*]] = icmp slt <vscale x 4 x i32> [[WIDE_LOAD]], [[VEC_PHI]]
 ; OUTLOOP-NEXT:    [[TMP10]] = select <vscale x 4 x i1> [[TMP9]], <vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[VEC_PHI]]
 ; OUTLOOP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -314,8 +309,7 @@ define i32 @smin(ptr %a, i64 %n, i32 %start) {
 ; INLOOP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; INLOOP-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ [[START:%.*]], [[VECTOR_PH]] ], [ [[RDX_MINMAX:%.*]], [[VECTOR_BODY]] ]
 ; INLOOP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; INLOOP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; INLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; INLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; INLOOP-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.smin.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]])
 ; INLOOP-NEXT:    [[RDX_MINMAX]] = call i32 @llvm.smin.i32(i32 [[TMP9]], i32 [[VEC_PHI]])
 ; INLOOP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -364,8 +358,7 @@ define i32 @smin(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-OUTLOOP-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-OUTLOOP-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP13:%.*]] = icmp slt <vscale x 4 x i32> [[VP_OP_LOAD]], [[VEC_PHI]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP14:%.*]] = select <vscale x 4 x i1> [[TMP13]], <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i32> [[VEC_PHI]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP15]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[TMP14]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP9]])
@@ -415,8 +408,7 @@ define i32 @smin(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-INLOOP-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-INLOOP-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-INLOOP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-INLOOP-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-INLOOP-NEXT:    [[TMP13:%.*]] = call i32 @llvm.vp.reduce.smin.nxv4i32(i32 2147483647, <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-INLOOP-NEXT:    [[RDX_MINMAX]] = call i32 @llvm.smin.i32(i32 [[TMP13]], i32 [[VEC_PHI]])
 ; IF-EVL-INLOOP-NEXT:    [[TMP14:%.*]] = zext i32 [[TMP9]] to i64
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/interleaved-accesses.ll b/llvm/test/Transforms/LoopVectorize/RISCV/interleaved-accesses.ll
index 85ccbab..e226eea 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/interleaved-accesses.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/interleaved-accesses.ll
@@ -78,9 +78,9 @@ define void @load_store_factor2_i32(ptr %p) {
 ; FIXED-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[EXIT:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[LOOP:%.*]]
 ; FIXED:       loop:
 ; FIXED-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ]
@@ -252,9 +252,9 @@ define void @load_store_factor2_i64(ptr %p) {
 ; FIXED-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[EXIT:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[LOOP:%.*]]
 ; FIXED:       loop:
 ; FIXED-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ]
@@ -437,9 +437,9 @@ define void @load_store_factor3_i32(ptr %p) {
 ; FIXED-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[EXIT:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[LOOP:%.*]]
 ; FIXED:       loop:
 ; FIXED-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ]
@@ -640,9 +640,9 @@ define void @load_store_factor3_i64(ptr %p) {
 ; FIXED-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[EXIT:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[LOOP:%.*]]
 ; FIXED:       loop:
 ; FIXED-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ]
@@ -852,9 +852,9 @@ define void @load_store_factor4(ptr %p) {
 ; FIXED-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[EXIT:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[LOOP:%.*]]
 ; FIXED:       loop:
 ; FIXED-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ]
@@ -1090,9 +1090,9 @@ define void @load_store_factor5(ptr %p) {
 ; FIXED-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[EXIT:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[LOOP:%.*]]
 ; FIXED:       loop:
 ; FIXED-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ]
@@ -1353,9 +1353,9 @@ define void @load_store_factor6(ptr %p) {
 ; FIXED-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[EXIT:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[LOOP:%.*]]
 ; FIXED:       loop:
 ; FIXED-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ]
@@ -1645,9 +1645,9 @@ define void @load_store_factor7(ptr %p) {
 ; FIXED-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[EXIT:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[LOOP:%.*]]
 ; FIXED:       loop:
 ; FIXED-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ]
@@ -1963,9 +1963,9 @@ define void @load_store_factor8(ptr %p) {
 ; FIXED-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[EXIT:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[LOOP:%.*]]
 ; FIXED:       loop:
 ; FIXED-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ]
@@ -2192,8 +2192,7 @@ define void @combine_load_factor2_i32(ptr noalias %p, ptr noalias %q) {
 ; CHECK-NEXT:    [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[STRIDED_VEC]], 1
 ; CHECK-NEXT:    [[TMP10:%.*]] = add <vscale x 4 x i32> [[TMP8]], [[TMP9]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[Q:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    store <vscale x 4 x i32> [[TMP10]], ptr [[TMP12]], align 4
+; CHECK-NEXT:    store <vscale x 4 x i32> [[TMP10]], ptr [[TMP11]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
@@ -2241,17 +2240,16 @@ define void @combine_load_factor2_i32(ptr noalias %p, ptr noalias %q) {
 ; FIXED-NEXT:    [[TMP5:%.*]] = add <8 x i32> [[STRIDED_VEC]], [[STRIDED_VEC1]]
 ; FIXED-NEXT:    [[TMP6:%.*]] = add <8 x i32> [[STRIDED_VEC3]], [[STRIDED_VEC4]]
 ; FIXED-NEXT:    [[TMP7:%.*]] = getelementptr i32, ptr [[Q:%.*]], i64 [[INDEX]]
-; FIXED-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[TMP7]], i32 0
 ; FIXED-NEXT:    [[TMP9:%.*]] = getelementptr i32, ptr [[TMP7]], i32 8
-; FIXED-NEXT:    store <8 x i32> [[TMP5]], ptr [[TMP8]], align 4
+; FIXED-NEXT:    store <8 x i32> [[TMP5]], ptr [[TMP7]], align 4
 ; FIXED-NEXT:    store <8 x i32> [[TMP6]], ptr [[TMP9]], align 4
 ; FIXED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; FIXED-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[EXIT:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[LOOP:%.*]]
 ; FIXED:       loop:
 ; FIXED-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ]
@@ -2294,8 +2292,7 @@ define void @combine_load_factor2_i32(ptr noalias %p, ptr noalias %q) {
 ; SCALABLE-NEXT:    [[TMP9:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[STRIDED_VEC]], 1
 ; SCALABLE-NEXT:    [[TMP10:%.*]] = add <vscale x 4 x i32> [[TMP8]], [[TMP9]]
 ; SCALABLE-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[Q:%.*]], i64 [[INDEX]]
-; SCALABLE-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr [[TMP11]], i32 0
-; SCALABLE-NEXT:    store <vscale x 4 x i32> [[TMP10]], ptr [[TMP12]], align 4
+; SCALABLE-NEXT:    store <vscale x 4 x i32> [[TMP10]], ptr [[TMP11]], align 4
 ; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; SCALABLE-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; SCALABLE-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
@@ -2372,8 +2369,7 @@ define void @combine_load_factor2_i64(ptr noalias %p, ptr noalias %q) {
 ; CHECK-NEXT:    [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[STRIDED_VEC]], 1
 ; CHECK-NEXT:    [[TMP10:%.*]] = add <vscale x 2 x i64> [[TMP8]], [[TMP9]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i64, ptr [[Q:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i64, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP10]], ptr [[TMP12]], align 8
+; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP10]], ptr [[TMP11]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
@@ -2421,17 +2417,16 @@ define void @combine_load_factor2_i64(ptr noalias %p, ptr noalias %q) {
 ; FIXED-NEXT:    [[TMP5:%.*]] = add <4 x i64> [[STRIDED_VEC]], [[STRIDED_VEC1]]
 ; FIXED-NEXT:    [[TMP6:%.*]] = add <4 x i64> [[STRIDED_VEC3]], [[STRIDED_VEC4]]
 ; FIXED-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[Q:%.*]], i64 [[INDEX]]
-; FIXED-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[TMP7]], i32 0
 ; FIXED-NEXT:    [[TMP9:%.*]] = getelementptr i64, ptr [[TMP7]], i32 4
-; FIXED-NEXT:    store <4 x i64> [[TMP5]], ptr [[TMP8]], align 8
+; FIXED-NEXT:    store <4 x i64> [[TMP5]], ptr [[TMP7]], align 8
 ; FIXED-NEXT:    store <4 x i64> [[TMP6]], ptr [[TMP9]], align 8
 ; FIXED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FIXED-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXED-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
 ; FIXED:       middle.block:
-; FIXED-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; FIXED-NEXT:    br label [[EXIT:%.*]]
 ; FIXED:       scalar.ph:
-; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; FIXED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; FIXED-NEXT:    br label [[LOOP:%.*]]
 ; FIXED:       loop:
 ; FIXED-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[NEXTI:%.*]], [[LOOP]] ]
@@ -2474,8 +2469,7 @@ define void @combine_load_factor2_i64(ptr noalias %p, ptr noalias %q) {
 ; SCALABLE-NEXT:    [[TMP9:%.*]] = extractvalue { <vscale x 2 x i64>, <vscale x 2 x i64> } [[STRIDED_VEC]], 1
 ; SCALABLE-NEXT:    [[TMP10:%.*]] = add <vscale x 2 x i64> [[TMP8]], [[TMP9]]
 ; SCALABLE-NEXT:    [[TMP11:%.*]] = getelementptr i64, ptr [[Q:%.*]], i64 [[INDEX]]
-; SCALABLE-NEXT:    [[TMP12:%.*]] = getelementptr i64, ptr [[TMP11]], i32 0
-; SCALABLE-NEXT:    store <vscale x 2 x i64> [[TMP10]], ptr [[TMP12]], align 8
+; SCALABLE-NEXT:    store <vscale x 2 x i64> [[TMP10]], ptr [[TMP11]], align 8
 ; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; SCALABLE-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; SCALABLE-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/interleaved-masked-access.ll b/llvm/test/Transforms/LoopVectorize/RISCV/interleaved-masked-access.ll
index 2a4cd28..c35a6ab 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/interleaved-masked-access.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/interleaved-masked-access.ll
@@ -1,7 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --filter-out-after "^scalar.ph:" --version 2
 ; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize,instcombine -mattr=+v -prefer-predicate-over-epilogue=scalar-epilogue %s 2>&1 | FileCheck %s -check-prefix=SCALAR_EPILOGUE
-; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize,instcombine -mattr=+v -prefer-predicate-over-epilogue=predicate-dont-vectorize -force-tail-folding-style=data %s 2>&1 | FileCheck %s -check-prefix=PREDICATED_DATA
-; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize,instcombine -mattr=+v -prefer-predicate-over-epilogue=predicate-dont-vectorize -force-tail-folding-style=data-with-evl %s 2>&1 | FileCheck %s -check-prefix=PREDICATED_DATA-WITH-EVL
+; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize,instcombine -mattr=+v -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue -force-tail-folding-style=data %s 2>&1 | FileCheck %s -check-prefix=PREDICATED_DATA
+; RUN: opt -mtriple=riscv64-none-linux-gnu -S -passes=loop-vectorize,instcombine -mattr=+v -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue %s 2>&1 | FileCheck %s -check-prefix=PREDICATED_DATA-WITH-EVL
 
 target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"
 
@@ -30,25 +30,26 @@ define void @masked_strided_factor2(ptr noalias nocapture readonly %p, ptr noali
 ; SCALAR_EPILOGUE-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; SCALAR_EPILOGUE-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP5]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; SCALAR_EPILOGUE-NEXT:    [[TMP6:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
-; SCALAR_EPILOGUE-NEXT:    [[TMP7:%.*]] = shl i32 [[INDEX]], 1
-; SCALAR_EPILOGUE-NEXT:    [[TMP8:%.*]] = sext i32 [[TMP7]] to i64
-; SCALAR_EPILOGUE-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP8]]
-; SCALAR_EPILOGUE-NEXT:    [[INTERLEAVED_MASK:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP6]], <vscale x 16 x i1> [[TMP6]])
-; SCALAR_EPILOGUE-NEXT:    [[WIDE_MASKED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.masked.load.nxv32i8.p0(ptr [[TMP9]], i32 1, <vscale x 32 x i1> [[INTERLEAVED_MASK]], <vscale x 32 x i8> poison)
-; SCALAR_EPILOGUE-NEXT:    [[STRIDED_VEC:%.*]] = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.vector.deinterleave2.nxv32i8(<vscale x 32 x i8> [[WIDE_MASKED_VEC]])
-; SCALAR_EPILOGUE-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
-; SCALAR_EPILOGUE-NEXT:    [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
-; SCALAR_EPILOGUE-NEXT:    [[TMP12:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]])
-; SCALAR_EPILOGUE-NEXT:    [[TMP13:%.*]] = sext i32 [[TMP7]] to i64
-; SCALAR_EPILOGUE-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[Q]], i64 [[TMP13]]
-; SCALAR_EPILOGUE-NEXT:    [[TMP15:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP12]]
-; SCALAR_EPILOGUE-NEXT:    [[INTERLEAVED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.vector.interleave2.nxv32i8(<vscale x 16 x i8> [[TMP12]], <vscale x 16 x i8> [[TMP15]])
-; SCALAR_EPILOGUE-NEXT:    [[INTERLEAVED_MASK3:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP6]], <vscale x 16 x i1> [[TMP6]])
-; SCALAR_EPILOGUE-NEXT:    call void @llvm.masked.store.nxv32i8.p0(<vscale x 32 x i8> [[INTERLEAVED_VEC]], ptr [[TMP14]], i32 1, <vscale x 32 x i1> [[INTERLEAVED_MASK3]])
+; SCALAR_EPILOGUE-NEXT:    [[TMP7:%.*]] = shl nuw nsw <vscale x 16 x i32> [[VEC_IND]], splat (i32 1)
+; SCALAR_EPILOGUE-NEXT:    [[TMP8:%.*]] = zext nneg <vscale x 16 x i32> [[TMP7]] to <vscale x 16 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 16 x i64> [[TMP8]]
+; SCALAR_EPILOGUE-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 16 x i8> @llvm.masked.gather.nxv16i8.nxv16p0(<vscale x 16 x ptr> [[TMP9]], i32 1, <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i8> poison)
+; SCALAR_EPILOGUE-NEXT:    [[TMP10:%.*]] = or disjoint <vscale x 16 x i32> [[TMP7]], splat (i32 1)
+; SCALAR_EPILOGUE-NEXT:    [[TMP11:%.*]] = zext nneg <vscale x 16 x i32> [[TMP10]] to <vscale x 16 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 16 x i64> [[TMP11]]
+; SCALAR_EPILOGUE-NEXT:    [[WIDE_MASKED_GATHER3:%.*]] = call <vscale x 16 x i8> @llvm.masked.gather.nxv16i8.nxv16p0(<vscale x 16 x ptr> [[TMP12]], i32 1, <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i8> poison)
+; SCALAR_EPILOGUE-NEXT:    [[TMP13:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[WIDE_MASKED_GATHER]], <vscale x 16 x i8> [[WIDE_MASKED_GATHER3]])
+; SCALAR_EPILOGUE-NEXT:    [[TMP14:%.*]] = zext nneg <vscale x 16 x i32> [[TMP7]] to <vscale x 16 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP14]]
+; SCALAR_EPILOGUE-NEXT:    call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[TMP13]], <vscale x 16 x ptr> [[TMP15]], i32 1, <vscale x 16 x i1> [[TMP6]])
+; SCALAR_EPILOGUE-NEXT:    [[TMP16:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP13]]
+; SCALAR_EPILOGUE-NEXT:    [[TMP17:%.*]] = zext nneg <vscale x 16 x i32> [[TMP10]] to <vscale x 16 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP17]]
+; SCALAR_EPILOGUE-NEXT:    call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[TMP16]], <vscale x 16 x ptr> [[TMP18]], i32 1, <vscale x 16 x i1> [[TMP6]])
 ; SCALAR_EPILOGUE-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP4]]
 ; SCALAR_EPILOGUE-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
-; SCALAR_EPILOGUE-NEXT:    [[TMP16:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
-; SCALAR_EPILOGUE-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; SCALAR_EPILOGUE-NEXT:    [[TMP19:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; SCALAR_EPILOGUE-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; SCALAR_EPILOGUE:       middle.block:
 ; SCALAR_EPILOGUE-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[N_MOD_VF]], 0
 ; SCALAR_EPILOGUE-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
@@ -79,25 +80,26 @@ define void @masked_strided_factor2(ptr noalias nocapture readonly %p, ptr noali
 ; PREDICATED_DATA-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i32(i32 [[INDEX]], i32 1024)
 ; PREDICATED_DATA-NEXT:    [[TMP5:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
 ; PREDICATED_DATA-NEXT:    [[TMP6:%.*]] = select <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> zeroinitializer
-; PREDICATED_DATA-NEXT:    [[TMP7:%.*]] = shl i32 [[INDEX]], 1
-; PREDICATED_DATA-NEXT:    [[TMP8:%.*]] = sext i32 [[TMP7]] to i64
-; PREDICATED_DATA-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP8]]
-; PREDICATED_DATA-NEXT:    [[INTERLEAVED_MASK:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP6]], <vscale x 16 x i1> [[TMP6]])
-; PREDICATED_DATA-NEXT:    [[WIDE_MASKED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.masked.load.nxv32i8.p0(ptr [[TMP9]], i32 1, <vscale x 32 x i1> [[INTERLEAVED_MASK]], <vscale x 32 x i8> poison)
-; PREDICATED_DATA-NEXT:    [[STRIDED_VEC:%.*]] = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.vector.deinterleave2.nxv32i8(<vscale x 32 x i8> [[WIDE_MASKED_VEC]])
-; PREDICATED_DATA-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
-; PREDICATED_DATA-NEXT:    [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
-; PREDICATED_DATA-NEXT:    [[TMP12:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]])
-; PREDICATED_DATA-NEXT:    [[TMP13:%.*]] = sext i32 [[TMP7]] to i64
-; PREDICATED_DATA-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[Q]], i64 [[TMP13]]
-; PREDICATED_DATA-NEXT:    [[TMP15:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP12]]
-; PREDICATED_DATA-NEXT:    [[INTERLEAVED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.vector.interleave2.nxv32i8(<vscale x 16 x i8> [[TMP12]], <vscale x 16 x i8> [[TMP15]])
-; PREDICATED_DATA-NEXT:    [[INTERLEAVED_MASK3:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP6]], <vscale x 16 x i1> [[TMP6]])
-; PREDICATED_DATA-NEXT:    call void @llvm.masked.store.nxv32i8.p0(<vscale x 32 x i8> [[INTERLEAVED_VEC]], ptr [[TMP14]], i32 1, <vscale x 32 x i1> [[INTERLEAVED_MASK3]])
+; PREDICATED_DATA-NEXT:    [[TMP7:%.*]] = shl nuw nsw <vscale x 16 x i32> [[VEC_IND]], splat (i32 1)
+; PREDICATED_DATA-NEXT:    [[TMP8:%.*]] = zext nneg <vscale x 16 x i32> [[TMP7]] to <vscale x 16 x i64>
+; PREDICATED_DATA-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 16 x i64> [[TMP8]]
+; PREDICATED_DATA-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 16 x i8> @llvm.masked.gather.nxv16i8.nxv16p0(<vscale x 16 x ptr> [[TMP9]], i32 1, <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i8> poison)
+; PREDICATED_DATA-NEXT:    [[TMP10:%.*]] = or disjoint <vscale x 16 x i32> [[TMP7]], splat (i32 1)
+; PREDICATED_DATA-NEXT:    [[TMP11:%.*]] = zext nneg <vscale x 16 x i32> [[TMP10]] to <vscale x 16 x i64>
+; PREDICATED_DATA-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 16 x i64> [[TMP11]]
+; PREDICATED_DATA-NEXT:    [[WIDE_MASKED_GATHER3:%.*]] = call <vscale x 16 x i8> @llvm.masked.gather.nxv16i8.nxv16p0(<vscale x 16 x ptr> [[TMP12]], i32 1, <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i8> poison)
+; PREDICATED_DATA-NEXT:    [[TMP13:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[WIDE_MASKED_GATHER]], <vscale x 16 x i8> [[WIDE_MASKED_GATHER3]])
+; PREDICATED_DATA-NEXT:    [[TMP14:%.*]] = zext nneg <vscale x 16 x i32> [[TMP7]] to <vscale x 16 x i64>
+; PREDICATED_DATA-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP14]]
+; PREDICATED_DATA-NEXT:    call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[TMP13]], <vscale x 16 x ptr> [[TMP15]], i32 1, <vscale x 16 x i1> [[TMP6]])
+; PREDICATED_DATA-NEXT:    [[TMP16:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP13]]
+; PREDICATED_DATA-NEXT:    [[TMP17:%.*]] = zext nneg <vscale x 16 x i32> [[TMP10]] to <vscale x 16 x i64>
+; PREDICATED_DATA-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP17]]
+; PREDICATED_DATA-NEXT:    call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[TMP16]], <vscale x 16 x ptr> [[TMP18]], i32 1, <vscale x 16 x i1> [[TMP6]])
 ; PREDICATED_DATA-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP3]]
 ; PREDICATED_DATA-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
-; PREDICATED_DATA-NEXT:    [[TMP16:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
-; PREDICATED_DATA-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; PREDICATED_DATA-NEXT:    [[TMP19:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; PREDICATED_DATA-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; PREDICATED_DATA:       middle.block:
 ; PREDICATED_DATA-NEXT:    br label [[FOR_END:%.*]]
 ; PREDICATED_DATA:       scalar.ph:
@@ -118,34 +120,37 @@ define void @masked_strided_factor2(ptr noalias nocapture readonly %p, ptr noali
 ; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
 ; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
 ; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP4:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP3]], i64 0
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
 ; PREDICATED_DATA-WITH-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; PREDICATED_DATA-WITH-EVL:       vector.body:
 ; PREDICATED_DATA-WITH-EVL-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; PREDICATED_DATA-WITH-EVL-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP4]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP5:%.*]] = icmp ult <vscale x 16 x i32> [[VEC_IND]], splat (i32 1024)
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[AVL:%.*]] = sub i32 1024, [[EVL_BASED_IV]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP5:%.*]] = call i32 @llvm.experimental.get.vector.length.i32(i32 [[AVL]], i32 16, i1 true)
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP5]], i64 0
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
 ; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP6:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP7:%.*]] = select <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i1> zeroinitializer
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP8:%.*]] = shl i32 [[INDEX]], 1
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP9]]
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[INTERLEAVED_MASK:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP7]], <vscale x 16 x i1> [[TMP7]])
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[WIDE_MASKED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.masked.load.nxv32i8.p0(ptr [[TMP10]], i32 1, <vscale x 32 x i1> [[INTERLEAVED_MASK]], <vscale x 32 x i8> poison)
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[STRIDED_VEC:%.*]] = call { <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.vector.deinterleave2.nxv32i8(<vscale x 32 x i8> [[WIDE_MASKED_VEC]])
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP12:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP13:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP11]], <vscale x 16 x i8> [[TMP12]])
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP14:%.*]] = sext i32 [[TMP8]] to i64
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP15:%.*]] = getelementptr i8, ptr [[Q]], i64 [[TMP14]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP7:%.*]] = shl nuw nsw <vscale x 16 x i32> [[VEC_IND]], splat (i32 1)
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP8:%.*]] = zext nneg <vscale x 16 x i32> [[TMP7]] to <vscale x 16 x i64>
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 16 x i64> [[TMP8]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 16 x i8> @llvm.vp.gather.nxv16i8.nxv16p0(<vscale x 16 x ptr> align 1 [[TMP9]], <vscale x 16 x i1> [[TMP6]], i32 [[TMP5]])
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP10:%.*]] = or disjoint <vscale x 16 x i32> [[TMP7]], splat (i32 1)
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP11:%.*]] = zext nneg <vscale x 16 x i32> [[TMP10]] to <vscale x 16 x i64>
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 16 x i64> [[TMP11]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[WIDE_MASKED_GATHER3:%.*]] = call <vscale x 16 x i8> @llvm.vp.gather.nxv16i8.nxv16p0(<vscale x 16 x ptr> align 1 [[TMP12]], <vscale x 16 x i1> [[TMP6]], i32 [[TMP5]])
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP13:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[WIDE_MASKED_GATHER]], <vscale x 16 x i8> [[WIDE_MASKED_GATHER3]])
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP14:%.*]] = zext nneg <vscale x 16 x i32> [[TMP7]] to <vscale x 16 x i64>
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP14]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    call void @llvm.vp.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[TMP13]], <vscale x 16 x ptr> align 1 [[TMP15]], <vscale x 16 x i1> [[TMP6]], i32 [[TMP5]])
 ; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP16:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP13]]
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[INTERLEAVED_VEC:%.*]] = call <vscale x 32 x i8> @llvm.vector.interleave2.nxv32i8(<vscale x 16 x i8> [[TMP13]], <vscale x 16 x i8> [[TMP16]])
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[INTERLEAVED_MASK3:%.*]] = call <vscale x 32 x i1> @llvm.vector.interleave2.nxv32i1(<vscale x 16 x i1> [[TMP7]], <vscale x 16 x i1> [[TMP7]])
-; PREDICATED_DATA-WITH-EVL-NEXT:    call void @llvm.masked.store.nxv32i8.p0(<vscale x 32 x i8> [[INTERLEAVED_VEC]], ptr [[TMP15]], i32 1, <vscale x 32 x i1> [[INTERLEAVED_MASK3]])
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP17:%.*]] = zext nneg <vscale x 16 x i32> [[TMP10]] to <vscale x 16 x i64>
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP17]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    call void @llvm.vp.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[TMP16]], <vscale x 16 x ptr> align 1 [[TMP18]], <vscale x 16 x i1> [[TMP6]], i32 [[TMP5]])
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i32 [[TMP5]], [[EVL_BASED_IV]]
 ; PREDICATED_DATA-WITH-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP3]]
 ; PREDICATED_DATA-WITH-EVL-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP17:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
-; PREDICATED_DATA-WITH-EVL-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP19:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; PREDICATED_DATA-WITH-EVL:       middle.block:
 ; PREDICATED_DATA-WITH-EVL-NEXT:    br label [[FOR_END:%.*]]
 ; PREDICATED_DATA-WITH-EVL:       scalar.ph:
@@ -210,29 +215,42 @@ define void @masked_strided_factor4(ptr noalias nocapture readonly %p, ptr noali
 ; SCALAR_EPILOGUE-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; SCALAR_EPILOGUE-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP5]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; SCALAR_EPILOGUE-NEXT:    [[TMP6:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
-; SCALAR_EPILOGUE-NEXT:    [[TMP7:%.*]] = shl i32 [[INDEX]], 2
-; SCALAR_EPILOGUE-NEXT:    [[TMP8:%.*]] = sext i32 [[TMP7]] to i64
-; SCALAR_EPILOGUE-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP8]]
-; SCALAR_EPILOGUE-NEXT:    [[INTERLEAVED_MASK:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP6]], <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i1> [[TMP6]])
-; SCALAR_EPILOGUE-NEXT:    [[WIDE_MASKED_VEC:%.*]] = call <vscale x 64 x i8> @llvm.masked.load.nxv64i8.p0(ptr [[TMP9]], i32 1, <vscale x 64 x i1> [[INTERLEAVED_MASK]], <vscale x 64 x i8> poison)
-; SCALAR_EPILOGUE-NEXT:    [[STRIDED_VEC:%.*]] = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.vector.deinterleave4.nxv64i8(<vscale x 64 x i8> [[WIDE_MASKED_VEC]])
-; SCALAR_EPILOGUE-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
-; SCALAR_EPILOGUE-NEXT:    [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
-; SCALAR_EPILOGUE-NEXT:    [[TMP12:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 2
-; SCALAR_EPILOGUE-NEXT:    [[TMP13:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 3
-; SCALAR_EPILOGUE-NEXT:    [[TMP14:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]])
-; SCALAR_EPILOGUE-NEXT:    [[TMP15:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP14]]
-; SCALAR_EPILOGUE-NEXT:    [[TMP16:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP12]], <vscale x 16 x i8> [[TMP13]])
-; SCALAR_EPILOGUE-NEXT:    [[TMP17:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP16]]
-; SCALAR_EPILOGUE-NEXT:    [[TMP18:%.*]] = sext i32 [[TMP7]] to i64
-; SCALAR_EPILOGUE-NEXT:    [[TMP19:%.*]] = getelementptr i8, ptr [[Q]], i64 [[TMP18]]
-; SCALAR_EPILOGUE-NEXT:    [[INTERLEAVED_VEC:%.*]] = call <vscale x 64 x i8> @llvm.vector.interleave4.nxv64i8(<vscale x 16 x i8> [[TMP14]], <vscale x 16 x i8> [[TMP15]], <vscale x 16 x i8> [[TMP16]], <vscale x 16 x i8> [[TMP17]])
-; SCALAR_EPILOGUE-NEXT:    [[INTERLEAVED_MASK3:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP6]], <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i1> [[TMP6]])
-; SCALAR_EPILOGUE-NEXT:    call void @llvm.masked.store.nxv64i8.p0(<vscale x 64 x i8> [[INTERLEAVED_VEC]], ptr [[TMP19]], i32 1, <vscale x 64 x i1> [[INTERLEAVED_MASK3]])
+; SCALAR_EPILOGUE-NEXT:    [[TMP7:%.*]] = shl nuw nsw <vscale x 16 x i32> [[VEC_IND]], splat (i32 2)
+; SCALAR_EPILOGUE-NEXT:    [[TMP8:%.*]] = or disjoint <vscale x 16 x i32> [[TMP7]], splat (i32 1)
+; SCALAR_EPILOGUE-NEXT:    [[TMP9:%.*]] = or disjoint <vscale x 16 x i32> [[TMP7]], splat (i32 2)
+; SCALAR_EPILOGUE-NEXT:    [[TMP10:%.*]] = or disjoint <vscale x 16 x i32> [[TMP7]], splat (i32 3)
+; SCALAR_EPILOGUE-NEXT:    [[TMP11:%.*]] = zext nneg <vscale x 16 x i32> [[TMP7]] to <vscale x 16 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 16 x i64> [[TMP11]]
+; SCALAR_EPILOGUE-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 16 x i8> @llvm.masked.gather.nxv16i8.nxv16p0(<vscale x 16 x ptr> [[TMP12]], i32 1, <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i8> poison)
+; SCALAR_EPILOGUE-NEXT:    [[TMP13:%.*]] = zext nneg <vscale x 16 x i32> [[TMP8]] to <vscale x 16 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 16 x i64> [[TMP13]]
+; SCALAR_EPILOGUE-NEXT:    [[WIDE_MASKED_GATHER3:%.*]] = call <vscale x 16 x i8> @llvm.masked.gather.nxv16i8.nxv16p0(<vscale x 16 x ptr> [[TMP14]], i32 1, <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i8> poison)
+; SCALAR_EPILOGUE-NEXT:    [[TMP15:%.*]] = zext nneg <vscale x 16 x i32> [[TMP9]] to <vscale x 16 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 16 x i64> [[TMP15]]
+; SCALAR_EPILOGUE-NEXT:    [[WIDE_MASKED_GATHER4:%.*]] = call <vscale x 16 x i8> @llvm.masked.gather.nxv16i8.nxv16p0(<vscale x 16 x ptr> [[TMP16]], i32 1, <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i8> poison)
+; SCALAR_EPILOGUE-NEXT:    [[TMP17:%.*]] = zext nneg <vscale x 16 x i32> [[TMP10]] to <vscale x 16 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 16 x i64> [[TMP17]]
+; SCALAR_EPILOGUE-NEXT:    [[WIDE_MASKED_GATHER5:%.*]] = call <vscale x 16 x i8> @llvm.masked.gather.nxv16i8.nxv16p0(<vscale x 16 x ptr> [[TMP18]], i32 1, <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i8> poison)
+; SCALAR_EPILOGUE-NEXT:    [[TMP19:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[WIDE_MASKED_GATHER]], <vscale x 16 x i8> [[WIDE_MASKED_GATHER3]])
+; SCALAR_EPILOGUE-NEXT:    [[TMP20:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP19]]
+; SCALAR_EPILOGUE-NEXT:    [[TMP21:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[WIDE_MASKED_GATHER4]], <vscale x 16 x i8> [[WIDE_MASKED_GATHER5]])
+; SCALAR_EPILOGUE-NEXT:    [[TMP22:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP21]]
+; SCALAR_EPILOGUE-NEXT:    [[TMP23:%.*]] = zext nneg <vscale x 16 x i32> [[TMP7]] to <vscale x 16 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP23]]
+; SCALAR_EPILOGUE-NEXT:    call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[TMP19]], <vscale x 16 x ptr> [[TMP24]], i32 1, <vscale x 16 x i1> [[TMP6]])
+; SCALAR_EPILOGUE-NEXT:    [[TMP25:%.*]] = zext nneg <vscale x 16 x i32> [[TMP8]] to <vscale x 16 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP25]]
+; SCALAR_EPILOGUE-NEXT:    call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[TMP20]], <vscale x 16 x ptr> [[TMP26]], i32 1, <vscale x 16 x i1> [[TMP6]])
+; SCALAR_EPILOGUE-NEXT:    [[TMP27:%.*]] = zext nneg <vscale x 16 x i32> [[TMP9]] to <vscale x 16 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP27]]
+; SCALAR_EPILOGUE-NEXT:    call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[TMP21]], <vscale x 16 x ptr> [[TMP28]], i32 1, <vscale x 16 x i1> [[TMP6]])
+; SCALAR_EPILOGUE-NEXT:    [[TMP29:%.*]] = zext nneg <vscale x 16 x i32> [[TMP10]] to <vscale x 16 x i64>
+; SCALAR_EPILOGUE-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP29]]
+; SCALAR_EPILOGUE-NEXT:    call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[TMP22]], <vscale x 16 x ptr> [[TMP30]], i32 1, <vscale x 16 x i1> [[TMP6]])
 ; SCALAR_EPILOGUE-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP4]]
 ; SCALAR_EPILOGUE-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
-; SCALAR_EPILOGUE-NEXT:    [[TMP20:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
-; SCALAR_EPILOGUE-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; SCALAR_EPILOGUE-NEXT:    [[TMP31:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; SCALAR_EPILOGUE-NEXT:    br i1 [[TMP31]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; SCALAR_EPILOGUE:       middle.block:
 ; SCALAR_EPILOGUE-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[N_MOD_VF]], 0
 ; SCALAR_EPILOGUE-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
@@ -263,29 +281,42 @@ define void @masked_strided_factor4(ptr noalias nocapture readonly %p, ptr noali
 ; PREDICATED_DATA-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = call <vscale x 16 x i1> @llvm.get.active.lane.mask.nxv16i1.i32(i32 [[INDEX]], i32 1024)
 ; PREDICATED_DATA-NEXT:    [[TMP5:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
 ; PREDICATED_DATA-NEXT:    [[TMP6:%.*]] = select <vscale x 16 x i1> [[ACTIVE_LANE_MASK]], <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> zeroinitializer
-; PREDICATED_DATA-NEXT:    [[TMP7:%.*]] = shl i32 [[INDEX]], 2
-; PREDICATED_DATA-NEXT:    [[TMP8:%.*]] = sext i32 [[TMP7]] to i64
-; PREDICATED_DATA-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP8]]
-; PREDICATED_DATA-NEXT:    [[INTERLEAVED_MASK:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP6]], <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i1> [[TMP6]])
-; PREDICATED_DATA-NEXT:    [[WIDE_MASKED_VEC:%.*]] = call <vscale x 64 x i8> @llvm.masked.load.nxv64i8.p0(ptr [[TMP9]], i32 1, <vscale x 64 x i1> [[INTERLEAVED_MASK]], <vscale x 64 x i8> poison)
-; PREDICATED_DATA-NEXT:    [[STRIDED_VEC:%.*]] = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.vector.deinterleave4.nxv64i8(<vscale x 64 x i8> [[WIDE_MASKED_VEC]])
-; PREDICATED_DATA-NEXT:    [[TMP10:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
-; PREDICATED_DATA-NEXT:    [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
-; PREDICATED_DATA-NEXT:    [[TMP12:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 2
-; PREDICATED_DATA-NEXT:    [[TMP13:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 3
-; PREDICATED_DATA-NEXT:    [[TMP14:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP10]], <vscale x 16 x i8> [[TMP11]])
-; PREDICATED_DATA-NEXT:    [[TMP15:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP14]]
-; PREDICATED_DATA-NEXT:    [[TMP16:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP12]], <vscale x 16 x i8> [[TMP13]])
-; PREDICATED_DATA-NEXT:    [[TMP17:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP16]]
-; PREDICATED_DATA-NEXT:    [[TMP18:%.*]] = sext i32 [[TMP7]] to i64
-; PREDICATED_DATA-NEXT:    [[TMP19:%.*]] = getelementptr i8, ptr [[Q]], i64 [[TMP18]]
-; PREDICATED_DATA-NEXT:    [[INTERLEAVED_VEC:%.*]] = call <vscale x 64 x i8> @llvm.vector.interleave4.nxv64i8(<vscale x 16 x i8> [[TMP14]], <vscale x 16 x i8> [[TMP15]], <vscale x 16 x i8> [[TMP16]], <vscale x 16 x i8> [[TMP17]])
-; PREDICATED_DATA-NEXT:    [[INTERLEAVED_MASK3:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP6]], <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i1> [[TMP6]])
-; PREDICATED_DATA-NEXT:    call void @llvm.masked.store.nxv64i8.p0(<vscale x 64 x i8> [[INTERLEAVED_VEC]], ptr [[TMP19]], i32 1, <vscale x 64 x i1> [[INTERLEAVED_MASK3]])
+; PREDICATED_DATA-NEXT:    [[TMP7:%.*]] = shl nuw nsw <vscale x 16 x i32> [[VEC_IND]], splat (i32 2)
+; PREDICATED_DATA-NEXT:    [[TMP8:%.*]] = or disjoint <vscale x 16 x i32> [[TMP7]], splat (i32 1)
+; PREDICATED_DATA-NEXT:    [[TMP9:%.*]] = or disjoint <vscale x 16 x i32> [[TMP7]], splat (i32 2)
+; PREDICATED_DATA-NEXT:    [[TMP10:%.*]] = or disjoint <vscale x 16 x i32> [[TMP7]], splat (i32 3)
+; PREDICATED_DATA-NEXT:    [[TMP11:%.*]] = zext nneg <vscale x 16 x i32> [[TMP7]] to <vscale x 16 x i64>
+; PREDICATED_DATA-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 16 x i64> [[TMP11]]
+; PREDICATED_DATA-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 16 x i8> @llvm.masked.gather.nxv16i8.nxv16p0(<vscale x 16 x ptr> [[TMP12]], i32 1, <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i8> poison)
+; PREDICATED_DATA-NEXT:    [[TMP13:%.*]] = zext nneg <vscale x 16 x i32> [[TMP8]] to <vscale x 16 x i64>
+; PREDICATED_DATA-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 16 x i64> [[TMP13]]
+; PREDICATED_DATA-NEXT:    [[WIDE_MASKED_GATHER3:%.*]] = call <vscale x 16 x i8> @llvm.masked.gather.nxv16i8.nxv16p0(<vscale x 16 x ptr> [[TMP14]], i32 1, <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i8> poison)
+; PREDICATED_DATA-NEXT:    [[TMP15:%.*]] = zext nneg <vscale x 16 x i32> [[TMP9]] to <vscale x 16 x i64>
+; PREDICATED_DATA-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 16 x i64> [[TMP15]]
+; PREDICATED_DATA-NEXT:    [[WIDE_MASKED_GATHER4:%.*]] = call <vscale x 16 x i8> @llvm.masked.gather.nxv16i8.nxv16p0(<vscale x 16 x ptr> [[TMP16]], i32 1, <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i8> poison)
+; PREDICATED_DATA-NEXT:    [[TMP17:%.*]] = zext nneg <vscale x 16 x i32> [[TMP10]] to <vscale x 16 x i64>
+; PREDICATED_DATA-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 16 x i64> [[TMP17]]
+; PREDICATED_DATA-NEXT:    [[WIDE_MASKED_GATHER5:%.*]] = call <vscale x 16 x i8> @llvm.masked.gather.nxv16i8.nxv16p0(<vscale x 16 x ptr> [[TMP18]], i32 1, <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i8> poison)
+; PREDICATED_DATA-NEXT:    [[TMP19:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[WIDE_MASKED_GATHER]], <vscale x 16 x i8> [[WIDE_MASKED_GATHER3]])
+; PREDICATED_DATA-NEXT:    [[TMP20:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP19]]
+; PREDICATED_DATA-NEXT:    [[TMP21:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[WIDE_MASKED_GATHER4]], <vscale x 16 x i8> [[WIDE_MASKED_GATHER5]])
+; PREDICATED_DATA-NEXT:    [[TMP22:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP21]]
+; PREDICATED_DATA-NEXT:    [[TMP23:%.*]] = zext nneg <vscale x 16 x i32> [[TMP7]] to <vscale x 16 x i64>
+; PREDICATED_DATA-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP23]]
+; PREDICATED_DATA-NEXT:    call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[TMP19]], <vscale x 16 x ptr> [[TMP24]], i32 1, <vscale x 16 x i1> [[TMP6]])
+; PREDICATED_DATA-NEXT:    [[TMP25:%.*]] = zext nneg <vscale x 16 x i32> [[TMP8]] to <vscale x 16 x i64>
+; PREDICATED_DATA-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP25]]
+; PREDICATED_DATA-NEXT:    call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[TMP20]], <vscale x 16 x ptr> [[TMP26]], i32 1, <vscale x 16 x i1> [[TMP6]])
+; PREDICATED_DATA-NEXT:    [[TMP27:%.*]] = zext nneg <vscale x 16 x i32> [[TMP9]] to <vscale x 16 x i64>
+; PREDICATED_DATA-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP27]]
+; PREDICATED_DATA-NEXT:    call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[TMP21]], <vscale x 16 x ptr> [[TMP28]], i32 1, <vscale x 16 x i1> [[TMP6]])
+; PREDICATED_DATA-NEXT:    [[TMP29:%.*]] = zext nneg <vscale x 16 x i32> [[TMP10]] to <vscale x 16 x i64>
+; PREDICATED_DATA-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP29]]
+; PREDICATED_DATA-NEXT:    call void @llvm.masked.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[TMP22]], <vscale x 16 x ptr> [[TMP30]], i32 1, <vscale x 16 x i1> [[TMP6]])
 ; PREDICATED_DATA-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP3]]
 ; PREDICATED_DATA-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
-; PREDICATED_DATA-NEXT:    [[TMP20:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
-; PREDICATED_DATA-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; PREDICATED_DATA-NEXT:    [[TMP31:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; PREDICATED_DATA-NEXT:    br i1 [[TMP31]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; PREDICATED_DATA:       middle.block:
 ; PREDICATED_DATA-NEXT:    br label [[FOR_END:%.*]]
 ; PREDICATED_DATA:       scalar.ph:
@@ -306,38 +337,53 @@ define void @masked_strided_factor4(ptr noalias nocapture readonly %p, ptr noali
 ; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[CONV]], i64 0
 ; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
 ; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP4:%.*]] = call <vscale x 16 x i32> @llvm.stepvector.nxv16i32()
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP3]], i64 0
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
 ; PREDICATED_DATA-WITH-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; PREDICATED_DATA-WITH-EVL:       vector.body:
 ; PREDICATED_DATA-WITH-EVL-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; PREDICATED_DATA-WITH-EVL-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 16 x i32> [ [[TMP4]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP5:%.*]] = icmp ult <vscale x 16 x i32> [[VEC_IND]], splat (i32 1024)
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[AVL:%.*]] = sub i32 1024, [[EVL_BASED_IV]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP5:%.*]] = call i32 @llvm.experimental.get.vector.length.i32(i32 [[AVL]], i32 16, i1 true)
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 16 x i32> poison, i32 [[TMP5]], i64 0
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 16 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 16 x i32> poison, <vscale x 16 x i32> zeroinitializer
 ; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP6:%.*]] = icmp ugt <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP7:%.*]] = select <vscale x 16 x i1> [[TMP5]], <vscale x 16 x i1> [[TMP6]], <vscale x 16 x i1> zeroinitializer
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP8:%.*]] = shl i32 [[INDEX]], 2
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP9:%.*]] = sext i32 [[TMP8]] to i64
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[P]], i64 [[TMP9]]
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[INTERLEAVED_MASK:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP7]], <vscale x 16 x i1> [[TMP7]], <vscale x 16 x i1> [[TMP7]], <vscale x 16 x i1> [[TMP7]])
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[WIDE_MASKED_VEC:%.*]] = call <vscale x 64 x i8> @llvm.masked.load.nxv64i8.p0(ptr [[TMP10]], i32 1, <vscale x 64 x i1> [[INTERLEAVED_MASK]], <vscale x 64 x i8> poison)
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[STRIDED_VEC:%.*]] = call { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } @llvm.vector.deinterleave4.nxv64i8(<vscale x 64 x i8> [[WIDE_MASKED_VEC]])
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP11:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 0
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP12:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 1
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP13:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 2
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP14:%.*]] = extractvalue { <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8>, <vscale x 16 x i8> } [[STRIDED_VEC]], 3
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP15:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP11]], <vscale x 16 x i8> [[TMP12]])
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP16:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP15]]
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP17:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[TMP13]], <vscale x 16 x i8> [[TMP14]])
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP18:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP17]]
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP19:%.*]] = sext i32 [[TMP8]] to i64
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP20:%.*]] = getelementptr i8, ptr [[Q]], i64 [[TMP19]]
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[INTERLEAVED_VEC:%.*]] = call <vscale x 64 x i8> @llvm.vector.interleave4.nxv64i8(<vscale x 16 x i8> [[TMP15]], <vscale x 16 x i8> [[TMP16]], <vscale x 16 x i8> [[TMP17]], <vscale x 16 x i8> [[TMP18]])
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[INTERLEAVED_MASK3:%.*]] = call <vscale x 64 x i1> @llvm.vector.interleave4.nxv64i1(<vscale x 16 x i1> [[TMP7]], <vscale x 16 x i1> [[TMP7]], <vscale x 16 x i1> [[TMP7]], <vscale x 16 x i1> [[TMP7]])
-; PREDICATED_DATA-WITH-EVL-NEXT:    call void @llvm.masked.store.nxv64i8.p0(<vscale x 64 x i8> [[INTERLEAVED_VEC]], ptr [[TMP20]], i32 1, <vscale x 64 x i1> [[INTERLEAVED_MASK3]])
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP7:%.*]] = shl nuw nsw <vscale x 16 x i32> [[VEC_IND]], splat (i32 2)
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP8:%.*]] = or disjoint <vscale x 16 x i32> [[TMP7]], splat (i32 1)
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP9:%.*]] = or disjoint <vscale x 16 x i32> [[TMP7]], splat (i32 2)
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP10:%.*]] = or disjoint <vscale x 16 x i32> [[TMP7]], splat (i32 3)
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP11:%.*]] = zext nneg <vscale x 16 x i32> [[TMP7]] to <vscale x 16 x i64>
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 16 x i64> [[TMP11]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 16 x i8> @llvm.vp.gather.nxv16i8.nxv16p0(<vscale x 16 x ptr> align 1 [[TMP12]], <vscale x 16 x i1> [[TMP6]], i32 [[TMP5]])
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP13:%.*]] = zext nneg <vscale x 16 x i32> [[TMP8]] to <vscale x 16 x i64>
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 16 x i64> [[TMP13]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[WIDE_MASKED_GATHER3:%.*]] = call <vscale x 16 x i8> @llvm.vp.gather.nxv16i8.nxv16p0(<vscale x 16 x ptr> align 1 [[TMP14]], <vscale x 16 x i1> [[TMP6]], i32 [[TMP5]])
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP15:%.*]] = zext nneg <vscale x 16 x i32> [[TMP9]] to <vscale x 16 x i64>
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 16 x i64> [[TMP15]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[WIDE_MASKED_GATHER4:%.*]] = call <vscale x 16 x i8> @llvm.vp.gather.nxv16i8.nxv16p0(<vscale x 16 x ptr> align 1 [[TMP16]], <vscale x 16 x i1> [[TMP6]], i32 [[TMP5]])
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP17:%.*]] = zext nneg <vscale x 16 x i32> [[TMP10]] to <vscale x 16 x i64>
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[P]], <vscale x 16 x i64> [[TMP17]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[WIDE_MASKED_GATHER5:%.*]] = call <vscale x 16 x i8> @llvm.vp.gather.nxv16i8.nxv16p0(<vscale x 16 x ptr> align 1 [[TMP18]], <vscale x 16 x i1> [[TMP6]], i32 [[TMP5]])
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP19:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[WIDE_MASKED_GATHER]], <vscale x 16 x i8> [[WIDE_MASKED_GATHER3]])
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP20:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP19]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP21:%.*]] = call <vscale x 16 x i8> @llvm.smax.nxv16i8(<vscale x 16 x i8> [[WIDE_MASKED_GATHER4]], <vscale x 16 x i8> [[WIDE_MASKED_GATHER5]])
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP22:%.*]] = sub <vscale x 16 x i8> zeroinitializer, [[TMP21]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP23:%.*]] = zext nneg <vscale x 16 x i32> [[TMP7]] to <vscale x 16 x i64>
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP23]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    call void @llvm.vp.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[TMP19]], <vscale x 16 x ptr> align 1 [[TMP24]], <vscale x 16 x i1> [[TMP6]], i32 [[TMP5]])
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP25:%.*]] = zext nneg <vscale x 16 x i32> [[TMP8]] to <vscale x 16 x i64>
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP25]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    call void @llvm.vp.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[TMP20]], <vscale x 16 x ptr> align 1 [[TMP26]], <vscale x 16 x i1> [[TMP6]], i32 [[TMP5]])
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP27:%.*]] = zext nneg <vscale x 16 x i32> [[TMP9]] to <vscale x 16 x i64>
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP27]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    call void @llvm.vp.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[TMP21]], <vscale x 16 x ptr> align 1 [[TMP28]], <vscale x 16 x i1> [[TMP6]], i32 [[TMP5]])
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP29:%.*]] = zext nneg <vscale x 16 x i32> [[TMP10]] to <vscale x 16 x i64>
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i8, ptr [[Q]], <vscale x 16 x i64> [[TMP29]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    call void @llvm.vp.scatter.nxv16i8.nxv16p0(<vscale x 16 x i8> [[TMP22]], <vscale x 16 x ptr> align 1 [[TMP30]], <vscale x 16 x i1> [[TMP6]], i32 [[TMP5]])
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i32 [[TMP5]], [[EVL_BASED_IV]]
 ; PREDICATED_DATA-WITH-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP3]]
 ; PREDICATED_DATA-WITH-EVL-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 16 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
-; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP21:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
-; PREDICATED_DATA-WITH-EVL-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    [[TMP31:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; PREDICATED_DATA-WITH-EVL-NEXT:    br i1 [[TMP31]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; PREDICATED_DATA-WITH-EVL:       middle.block:
 ; PREDICATED_DATA-WITH-EVL-NEXT:    br label [[FOR_END:%.*]]
 ; PREDICATED_DATA-WITH-EVL:       scalar.ph:
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/lmul.ll b/llvm/test/Transforms/LoopVectorize/RISCV/lmul.ll
index 4d59f27..93e0f90 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/lmul.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/lmul.ll
@@ -20,10 +20,9 @@ define void @load_store(ptr %p) {
 ; LMUL1:       vector.body:
 ; LMUL1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; LMUL1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[P:%.*]], i64 [[INDEX]]
-; LMUL1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP3]], i32 0
-; LMUL1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 1 x i64>, ptr [[TMP4]], align 8
+; LMUL1-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 1 x i64>, ptr [[TMP3]], align 8
 ; LMUL1-NEXT:    [[TMP5:%.*]] = add <vscale x 1 x i64> [[WIDE_LOAD]], splat (i64 1)
-; LMUL1-NEXT:    store <vscale x 1 x i64> [[TMP5]], ptr [[TMP4]], align 8
+; LMUL1-NEXT:    store <vscale x 1 x i64> [[TMP5]], ptr [[TMP3]], align 8
 ; LMUL1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
 ; LMUL1-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; LMUL1-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -62,10 +61,9 @@ define void @load_store(ptr %p) {
 ; LMUL2:       vector.body:
 ; LMUL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; LMUL2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[P:%.*]], i64 [[INDEX]]
-; LMUL2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 0
-; LMUL2-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP6]], align 8
+; LMUL2-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP5]], align 8
 ; LMUL2-NEXT:    [[TMP7:%.*]] = add <vscale x 2 x i64> [[WIDE_LOAD]], splat (i64 1)
-; LMUL2-NEXT:    store <vscale x 2 x i64> [[TMP7]], ptr [[TMP6]], align 8
+; LMUL2-NEXT:    store <vscale x 2 x i64> [[TMP7]], ptr [[TMP5]], align 8
 ; LMUL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
 ; LMUL2-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; LMUL2-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -104,10 +102,9 @@ define void @load_store(ptr %p) {
 ; LMUL4:       vector.body:
 ; LMUL4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; LMUL4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[P:%.*]], i64 [[INDEX]]
-; LMUL4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 0
-; LMUL4-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i64>, ptr [[TMP6]], align 8
+; LMUL4-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i64>, ptr [[TMP5]], align 8
 ; LMUL4-NEXT:    [[TMP7:%.*]] = add <vscale x 4 x i64> [[WIDE_LOAD]], splat (i64 1)
-; LMUL4-NEXT:    store <vscale x 4 x i64> [[TMP7]], ptr [[TMP6]], align 8
+; LMUL4-NEXT:    store <vscale x 4 x i64> [[TMP7]], ptr [[TMP5]], align 8
 ; LMUL4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
 ; LMUL4-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; LMUL4-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -146,10 +143,9 @@ define void @load_store(ptr %p) {
 ; LMUL8:       vector.body:
 ; LMUL8-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; LMUL8-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[P:%.*]], i64 [[INDEX]]
-; LMUL8-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 0
-; LMUL8-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i64>, ptr [[TMP6]], align 8
+; LMUL8-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i64>, ptr [[TMP5]], align 8
 ; LMUL8-NEXT:    [[TMP7:%.*]] = add <vscale x 8 x i64> [[WIDE_LOAD]], splat (i64 1)
-; LMUL8-NEXT:    store <vscale x 8 x i64> [[TMP7]], ptr [[TMP6]], align 8
+; LMUL8-NEXT:    store <vscale x 8 x i64> [[TMP7]], ptr [[TMP5]], align 8
 ; LMUL8-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
 ; LMUL8-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; LMUL8-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/low-trip-count.ll b/llvm/test/Transforms/LoopVectorize/RISCV/low-trip-count.ll
index e2641ab..0a87257 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/low-trip-count.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/low-trip-count.ll
@@ -134,14 +134,11 @@ define void @trip8_i8(ptr noalias nocapture noundef %dst, ptr noalias nocapture
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[TMP5:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 8, i32 4, i1 true)
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[TMP8:%.*]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i8> @llvm.vp.load.nxv4i8.p0(ptr align 1 [[TMP9]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP5]])
-; CHECK-NEXT:    [[TMP10:%.*]] = shl <vscale x 4 x i8> [[WIDE_MASKED_LOAD]], splat (i8 1)
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP11:%.*]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <vscale x 4 x i8> @llvm.vp.load.nxv4i8.p0(ptr align 1 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP5]])
-; CHECK-NEXT:    [[TMP13:%.*]] = add <vscale x 4 x i8> [[TMP10]], [[WIDE_MASKED_LOAD1]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    call void @llvm.vp.store.nxv4i8.p0(<vscale x 4 x i8> [[TMP13]], ptr align 1 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP5]])
+; CHECK-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i8> @llvm.vp.load.nxv4i8.p0(ptr align 1 [[TMP9:%.*]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP5]])
+; CHECK-NEXT:    [[TMP6:%.*]] = shl <vscale x 4 x i8> [[VP_OP_LOAD]], splat (i8 1)
+; CHECK-NEXT:    [[VP_OP_LOAD1:%.*]] = call <vscale x 4 x i8> @llvm.vp.load.nxv4i8.p0(ptr align 1 [[TMP12:%.*]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP5]])
+; CHECK-NEXT:    [[TMP7:%.*]] = add <vscale x 4 x i8> [[TMP6]], [[VP_OP_LOAD1]]
+; CHECK-NEXT:    call void @llvm.vp.store.nxv4i8.p0(<vscale x 4 x i8> [[TMP7]], ptr align 1 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP5]])
 ; CHECK-NEXT:    br label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    br label [[FOR_END:%.*]]
@@ -150,10 +147,10 @@ define void @trip8_i8(ptr noalias nocapture noundef %dst, ptr noalias nocapture
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[I_08:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i64 [[I_08]]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[TMP9]], i64 [[I_08]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
 ; CHECK-NEXT:    [[MUL:%.*]] = shl i8 [[TMP15]], 1
-; CHECK-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i64 [[I_08]]
+; CHECK-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[TMP12]], i64 [[I_08]]
 ; CHECK-NEXT:    [[TMP16:%.*]] = load i8, ptr [[ARRAYIDX1]], align 1
 ; CHECK-NEXT:    [[ADD:%.*]] = add i8 [[MUL]], [[TMP16]]
 ; CHECK-NEXT:    store i8 [[ADD]], ptr [[ARRAYIDX1]], align 1
@@ -190,26 +187,23 @@ define void @trip16_i8(ptr noalias nocapture noundef %dst, ptr noalias nocapture
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TMP0:%.*]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1:%.*]], align 1
 ; CHECK-NEXT:    [[TMP2:%.*]] = shl <16 x i8> [[WIDE_LOAD]], splat (i8 1)
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3:%.*]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP4:%.*]], align 1
 ; CHECK-NEXT:    [[TMP5:%.*]] = add <16 x i8> [[TMP2]], [[WIDE_LOAD1]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    store <16 x i8> [[TMP5]], ptr [[TMP6]], align 1
+; CHECK-NEXT:    store <16 x i8> [[TMP5]], ptr [[TMP4]], align 1
 ; CHECK-NEXT:    br label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 16, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[I_08:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i64 [[I_08]]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i64 [[I_08]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
 ; CHECK-NEXT:    [[MUL:%.*]] = shl i8 [[TMP7]], 1
-; CHECK-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i64 [[I_08]]
+; CHECK-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i64 [[I_08]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = load i8, ptr [[ARRAYIDX1]], align 1
 ; CHECK-NEXT:    [[ADD:%.*]] = add i8 [[MUL]], [[TMP8]]
 ; CHECK-NEXT:    store i8 [[ADD]], ptr [[ARRAYIDX1]], align 1
@@ -247,26 +241,23 @@ define void @trip32_i8(ptr noalias nocapture noundef %dst, ptr noalias nocapture
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TMP0:%.*]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <32 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <32 x i8>, ptr [[TMP1:%.*]], align 1
 ; CHECK-NEXT:    [[TMP2:%.*]] = shl <32 x i8> [[WIDE_LOAD]], splat (i8 1)
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3:%.*]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <32 x i8>, ptr [[TMP4]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <32 x i8>, ptr [[TMP4:%.*]], align 1
 ; CHECK-NEXT:    [[TMP5:%.*]] = add <32 x i8> [[TMP2]], [[WIDE_LOAD1]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    store <32 x i8> [[TMP5]], ptr [[TMP6]], align 1
+; CHECK-NEXT:    store <32 x i8> [[TMP5]], ptr [[TMP4]], align 1
 ; CHECK-NEXT:    br label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 32, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[I_08:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i64 [[I_08]]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i64 [[I_08]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
 ; CHECK-NEXT:    [[MUL:%.*]] = shl i8 [[TMP7]], 1
-; CHECK-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i64 [[I_08]]
+; CHECK-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i64 [[I_08]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = load i8, ptr [[ARRAYIDX1]], align 1
 ; CHECK-NEXT:    [[ADD:%.*]] = add i8 [[MUL]], [[TMP8]]
 ; CHECK-NEXT:    store i8 [[ADD]], ptr [[ARRAYIDX1]], align 1
@@ -305,21 +296,19 @@ define void @trip24_i8(ptr noalias nocapture noundef %dst, ptr noalias nocapture
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[SRC:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = shl <8 x i8> [[WIDE_LOAD]], splat (i8 1)
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[DST:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i8>, ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i8>, ptr [[TMP4]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = add <8 x i8> [[TMP3]], [[WIDE_LOAD1]]
-; CHECK-NEXT:    store <8 x i8> [[TMP6]], ptr [[TMP5]], align 1
+; CHECK-NEXT:    store <8 x i8> [[TMP6]], ptr [[TMP4]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 24
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 24, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[I_08:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
@@ -377,8 +366,7 @@ define i8 @mul_non_pow_2_low_trip_count(ptr noalias %a) {
 ; CHECK-NEXT:    [[VEC_IV:%.*]] = add <16 x i64> [[BROADCAST_SPLAT]], <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11, i64 12, i64 13, i64 14, i64 15>
 ; CHECK-NEXT:    [[ACTIVE_LANE_MASK:%.*]] = icmp ule <16 x i64> [[VEC_IV]], splat (i64 9)
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <16 x i8> @llvm.masked.load.v16i8.p0(ptr [[TMP1]], i32 1, <16 x i1> [[ACTIVE_LANE_MASK]], <16 x i8> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <16 x i8> @llvm.masked.load.v16i8.p0(ptr [[TMP0]], i32 1, <16 x i1> [[ACTIVE_LANE_MASK]], <16 x i8> poison)
 ; CHECK-NEXT:    [[TMP2]] = mul <16 x i8> [[WIDE_MASKED_LOAD]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = select <16 x i1> [[ACTIVE_LANE_MASK]], <16 x i8> [[TMP2]], <16 x i8> [[VEC_PHI]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/mask-index-type.ll b/llvm/test/Transforms/LoopVectorize/RISCV/mask-index-type.ll
index 4ccc45d4..10ba208 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/mask-index-type.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/mask-index-type.ll
@@ -36,13 +36,11 @@ define void @test(ptr noalias nocapture %a, ptr noalias nocapture %b, i32 %v) {
 ; VLENUNK-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 4 x i64> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; VLENUNK-NEXT:    [[TMP13:%.*]] = icmp ult <vscale x 4 x i64> [[VEC_IND]], splat (i64 512)
 ; VLENUNK-NEXT:    [[TMP14:%.*]] = getelementptr i32, ptr [[A:%.*]], i64 [[INDEX]]
-; VLENUNK-NEXT:    [[TMP15:%.*]] = getelementptr i32, ptr [[TMP14]], i32 0
-; VLENUNK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP15]], i32 4, <vscale x 4 x i1> [[TMP13]], <vscale x 4 x i32> poison)
+; VLENUNK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP14]], i32 4, <vscale x 4 x i1> [[TMP13]], <vscale x 4 x i32> poison)
 ; VLENUNK-NEXT:    [[PREDPHI:%.*]] = select <vscale x 4 x i1> [[TMP13]], <vscale x 4 x i32> [[WIDE_MASKED_LOAD]], <vscale x 4 x i32> zeroinitializer
 ; VLENUNK-NEXT:    [[TMP17:%.*]] = add <vscale x 4 x i32> [[PREDPHI]], [[BROADCAST_SPLAT]]
 ; VLENUNK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
-; VLENUNK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[TMP18]], i32 0
-; VLENUNK-NEXT:    store <vscale x 4 x i32> [[TMP17]], ptr [[TMP19]], align 4
+; VLENUNK-NEXT:    store <vscale x 4 x i32> [[TMP17]], ptr [[TMP18]], align 4
 ; VLENUNK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; VLENUNK-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 4 x i64> [[VEC_IND]], [[DOTSPLAT]]
 ; VLENUNK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/only-compute-cost-for-vplan-vfs.ll b/llvm/test/Transforms/LoopVectorize/RISCV/only-compute-cost-for-vplan-vfs.ll
index 4f3b8d8..0afe04e 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/only-compute-cost-for-vplan-vfs.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/only-compute-cost-for-vplan-vfs.ll
@@ -1,5 +1,5 @@
-; RUN: opt -passes=loop-vectorize -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: opt -passes=loop-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S -debug %s 2>&1 | FileCheck %s
 
 ; REQUIRES: asserts
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/ordered-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/ordered-reduction.ll
index 6668cd6..9ec8ba4 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/ordered-reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/ordered-reduction.ll
@@ -20,8 +20,7 @@ define float @fadd(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-ORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-NEXT:    [[VEC_PHI:%.*]] = phi float [ 0.000000e+00, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-ORDERED-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-ORDERED-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-ORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-ORDERED-NEXT:    [[TMP3]] = call float @llvm.vector.reduce.fadd.v4f32(float [[VEC_PHI]], <4 x float> [[WIDE_LOAD]])
 ; CHECK-ORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-ORDERED-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -58,8 +57,7 @@ define float @fadd(ptr noalias nocapture readonly %a, i64 %n) #0 {
 ; CHECK-UNORDERED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ <float 0.000000e+00, float -0.000000e+00, float -0.000000e+00, float -0.000000e+00>, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-UNORDERED-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-UNORDERED-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-UNORDERED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-UNORDERED-NEXT:    [[TMP3]] = fadd <4 x float> [[WIDE_LOAD]], [[VEC_PHI]]
 ; CHECK-UNORDERED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-UNORDERED-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/partial-reduce-dot-product.ll b/llvm/test/Transforms/LoopVectorize/RISCV/partial-reduce-dot-product.ll
index b8d5cbd..ee6b950 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/partial-reduce-dot-product.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/partial-reduce-dot-product.ll
@@ -26,12 +26,10 @@ define i32 @vqdot(ptr %a, ptr %b) #0 {
 ; V-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; V-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
 ; V-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; V-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
-; V-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
+; V-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP6]], align 1
 ; V-NEXT:    [[TMP8:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; V-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; V-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[TMP9]], i32 0
-; V-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
+; V-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP9]], align 1
 ; V-NEXT:    [[TMP11:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; V-NEXT:    [[TMP12:%.*]] = mul <vscale x 4 x i32> [[TMP11]], [[TMP8]]
 ; V-NEXT:    [[TMP13]] = add <vscale x 4 x i32> [[TMP12]], [[VEC_PHI]]
@@ -63,12 +61,10 @@ define i32 @vqdot(ptr %a, ptr %b) #0 {
 ; ZVQDOTQ-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; ZVQDOTQ-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 1 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; ZVQDOTQ-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; ZVQDOTQ-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
-; ZVQDOTQ-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
+; ZVQDOTQ-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP6]], align 1
 ; ZVQDOTQ-NEXT:    [[TMP8:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; ZVQDOTQ-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; ZVQDOTQ-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[TMP9]], i32 0
-; ZVQDOTQ-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
+; ZVQDOTQ-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP9]], align 1
 ; ZVQDOTQ-NEXT:    [[TMP11:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; ZVQDOTQ-NEXT:    [[TMP12:%.*]] = mul <vscale x 4 x i32> [[TMP11]], [[TMP8]]
 ; ZVQDOTQ-NEXT:    [[PARTIAL_REDUCE]] = call <vscale x 1 x i32> @llvm.experimental.vector.partial.reduce.add.nxv1i32.nxv4i32(<vscale x 1 x i32> [[VEC_PHI]], <vscale x 4 x i32> [[TMP12]])
@@ -92,16 +88,14 @@ define i32 @vqdot(ptr %a, ptr %b) #0 {
 ; FIXED-V-NEXT:    [[VEC_PHI:%.*]] = phi <8 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-V-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-V-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; FIXED-V-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
 ; FIXED-V-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP0]], i32 8
-; FIXED-V-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP1]], align 1
+; FIXED-V-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP0]], align 1
 ; FIXED-V-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i8>, ptr [[TMP2]], align 1
 ; FIXED-V-NEXT:    [[TMP3:%.*]] = sext <8 x i8> [[WIDE_LOAD]] to <8 x i32>
 ; FIXED-V-NEXT:    [[TMP4:%.*]] = sext <8 x i8> [[WIDE_LOAD2]] to <8 x i32>
 ; FIXED-V-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; FIXED-V-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[TMP5]], i32 0
 ; FIXED-V-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP5]], i32 8
-; FIXED-V-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i8>, ptr [[TMP6]], align 1
+; FIXED-V-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i8>, ptr [[TMP5]], align 1
 ; FIXED-V-NEXT:    [[WIDE_LOAD4:%.*]] = load <8 x i8>, ptr [[TMP7]], align 1
 ; FIXED-V-NEXT:    [[TMP8:%.*]] = sext <8 x i8> [[WIDE_LOAD3]] to <8 x i32>
 ; FIXED-V-NEXT:    [[TMP9:%.*]] = sext <8 x i8> [[WIDE_LOAD4]] to <8 x i32>
@@ -115,7 +109,7 @@ define i32 @vqdot(ptr %a, ptr %b) #0 {
 ; FIXED-V:       middle.block:
 ; FIXED-V-NEXT:    [[BIN_RDX:%.*]] = add <8 x i32> [[TMP13]], [[TMP12]]
 ; FIXED-V-NEXT:    [[TMP15:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[BIN_RDX]])
-; FIXED-V-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; FIXED-V-NEXT:    br label [[FOR_EXIT:%.*]]
 ; FIXED-V:       scalar.ph:
 ;
 ; FIXED-ZVQDOTQ-LABEL: define i32 @vqdot(
@@ -129,16 +123,14 @@ define i32 @vqdot(ptr %a, ptr %b) #0 {
 ; FIXED-ZVQDOTQ-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-ZVQDOTQ-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE5:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; FIXED-ZVQDOTQ-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP0]], i32 8
-; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP1]], align 1
+; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP0]], align 1
 ; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i8>, ptr [[TMP2]], align 1
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP3:%.*]] = sext <8 x i8> [[WIDE_LOAD]] to <8 x i32>
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP4:%.*]] = sext <8 x i8> [[WIDE_LOAD2]] to <8 x i32>
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; FIXED-ZVQDOTQ-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[TMP5]], i32 0
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP5]], i32 8
-; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i8>, ptr [[TMP6]], align 1
+; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i8>, ptr [[TMP5]], align 1
 ; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD4:%.*]] = load <8 x i8>, ptr [[TMP7]], align 1
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP8:%.*]] = sext <8 x i8> [[WIDE_LOAD3]] to <8 x i32>
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP9:%.*]] = sext <8 x i8> [[WIDE_LOAD4]] to <8 x i32>
@@ -152,7 +144,7 @@ define i32 @vqdot(ptr %a, ptr %b) #0 {
 ; FIXED-ZVQDOTQ:       middle.block:
 ; FIXED-ZVQDOTQ-NEXT:    [[BIN_RDX:%.*]] = add <2 x i32> [[PARTIAL_REDUCE5]], [[PARTIAL_REDUCE]]
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP13:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[BIN_RDX]])
-; FIXED-ZVQDOTQ-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; FIXED-ZVQDOTQ-NEXT:    br label [[FOR_EXIT:%.*]]
 ; FIXED-ZVQDOTQ:       scalar.ph:
 ;
 entry:
@@ -198,12 +190,10 @@ define i32 @vqdotu(ptr %a, ptr %b) #0 {
 ; V-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; V-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
 ; V-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; V-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
-; V-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
+; V-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP6]], align 1
 ; V-NEXT:    [[TMP8:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; V-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; V-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[TMP9]], i32 0
-; V-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
+; V-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP9]], align 1
 ; V-NEXT:    [[TMP11:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; V-NEXT:    [[TMP12:%.*]] = mul <vscale x 4 x i32> [[TMP11]], [[TMP8]]
 ; V-NEXT:    [[TMP13]] = add <vscale x 4 x i32> [[TMP12]], [[VEC_PHI]]
@@ -235,12 +225,10 @@ define i32 @vqdotu(ptr %a, ptr %b) #0 {
 ; ZVQDOTQ-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; ZVQDOTQ-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 1 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; ZVQDOTQ-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; ZVQDOTQ-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
-; ZVQDOTQ-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
+; ZVQDOTQ-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP6]], align 1
 ; ZVQDOTQ-NEXT:    [[TMP8:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; ZVQDOTQ-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; ZVQDOTQ-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[TMP9]], i32 0
-; ZVQDOTQ-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
+; ZVQDOTQ-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP9]], align 1
 ; ZVQDOTQ-NEXT:    [[TMP11:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; ZVQDOTQ-NEXT:    [[TMP12:%.*]] = mul <vscale x 4 x i32> [[TMP11]], [[TMP8]]
 ; ZVQDOTQ-NEXT:    [[PARTIAL_REDUCE]] = call <vscale x 1 x i32> @llvm.experimental.vector.partial.reduce.add.nxv1i32.nxv4i32(<vscale x 1 x i32> [[VEC_PHI]], <vscale x 4 x i32> [[TMP12]])
@@ -264,16 +252,14 @@ define i32 @vqdotu(ptr %a, ptr %b) #0 {
 ; FIXED-V-NEXT:    [[VEC_PHI:%.*]] = phi <8 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-V-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-V-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; FIXED-V-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
 ; FIXED-V-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP0]], i32 8
-; FIXED-V-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP1]], align 1
+; FIXED-V-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP0]], align 1
 ; FIXED-V-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i8>, ptr [[TMP2]], align 1
 ; FIXED-V-NEXT:    [[TMP3:%.*]] = zext <8 x i8> [[WIDE_LOAD]] to <8 x i32>
 ; FIXED-V-NEXT:    [[TMP4:%.*]] = zext <8 x i8> [[WIDE_LOAD2]] to <8 x i32>
 ; FIXED-V-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; FIXED-V-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[TMP5]], i32 0
 ; FIXED-V-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP5]], i32 8
-; FIXED-V-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i8>, ptr [[TMP6]], align 1
+; FIXED-V-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i8>, ptr [[TMP5]], align 1
 ; FIXED-V-NEXT:    [[WIDE_LOAD4:%.*]] = load <8 x i8>, ptr [[TMP7]], align 1
 ; FIXED-V-NEXT:    [[TMP8:%.*]] = zext <8 x i8> [[WIDE_LOAD3]] to <8 x i32>
 ; FIXED-V-NEXT:    [[TMP9:%.*]] = zext <8 x i8> [[WIDE_LOAD4]] to <8 x i32>
@@ -287,7 +273,7 @@ define i32 @vqdotu(ptr %a, ptr %b) #0 {
 ; FIXED-V:       middle.block:
 ; FIXED-V-NEXT:    [[BIN_RDX:%.*]] = add <8 x i32> [[TMP13]], [[TMP12]]
 ; FIXED-V-NEXT:    [[TMP15:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[BIN_RDX]])
-; FIXED-V-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; FIXED-V-NEXT:    br label [[FOR_EXIT:%.*]]
 ; FIXED-V:       scalar.ph:
 ;
 ; FIXED-ZVQDOTQ-LABEL: define i32 @vqdotu(
@@ -301,16 +287,14 @@ define i32 @vqdotu(ptr %a, ptr %b) #0 {
 ; FIXED-ZVQDOTQ-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-ZVQDOTQ-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE5:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; FIXED-ZVQDOTQ-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP0]], i32 8
-; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP1]], align 1
+; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP0]], align 1
 ; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i8>, ptr [[TMP2]], align 1
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP3:%.*]] = zext <8 x i8> [[WIDE_LOAD]] to <8 x i32>
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP4:%.*]] = zext <8 x i8> [[WIDE_LOAD2]] to <8 x i32>
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; FIXED-ZVQDOTQ-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[TMP5]], i32 0
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP5]], i32 8
-; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i8>, ptr [[TMP6]], align 1
+; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i8>, ptr [[TMP5]], align 1
 ; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD4:%.*]] = load <8 x i8>, ptr [[TMP7]], align 1
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP8:%.*]] = zext <8 x i8> [[WIDE_LOAD3]] to <8 x i32>
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP9:%.*]] = zext <8 x i8> [[WIDE_LOAD4]] to <8 x i32>
@@ -324,7 +308,7 @@ define i32 @vqdotu(ptr %a, ptr %b) #0 {
 ; FIXED-ZVQDOTQ:       middle.block:
 ; FIXED-ZVQDOTQ-NEXT:    [[BIN_RDX:%.*]] = add <2 x i32> [[PARTIAL_REDUCE5]], [[PARTIAL_REDUCE]]
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP13:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[BIN_RDX]])
-; FIXED-ZVQDOTQ-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; FIXED-ZVQDOTQ-NEXT:    br label [[FOR_EXIT:%.*]]
 ; FIXED-ZVQDOTQ:       scalar.ph:
 ;
 entry:
@@ -370,12 +354,10 @@ define i32 @vqdotsu(ptr %a, ptr %b) #0 {
 ; V-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; V-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
 ; V-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; V-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
-; V-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
+; V-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP6]], align 1
 ; V-NEXT:    [[TMP8:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; V-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; V-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[TMP9]], i32 0
-; V-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
+; V-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP9]], align 1
 ; V-NEXT:    [[TMP11:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; V-NEXT:    [[TMP12:%.*]] = mul <vscale x 4 x i32> [[TMP11]], [[TMP8]]
 ; V-NEXT:    [[TMP13]] = add <vscale x 4 x i32> [[TMP12]], [[VEC_PHI]]
@@ -407,12 +389,10 @@ define i32 @vqdotsu(ptr %a, ptr %b) #0 {
 ; ZVQDOTQ-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; ZVQDOTQ-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 1 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; ZVQDOTQ-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; ZVQDOTQ-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
-; ZVQDOTQ-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
+; ZVQDOTQ-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP6]], align 1
 ; ZVQDOTQ-NEXT:    [[TMP8:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; ZVQDOTQ-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; ZVQDOTQ-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[TMP9]], i32 0
-; ZVQDOTQ-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
+; ZVQDOTQ-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP9]], align 1
 ; ZVQDOTQ-NEXT:    [[TMP11:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; ZVQDOTQ-NEXT:    [[TMP12:%.*]] = mul <vscale x 4 x i32> [[TMP11]], [[TMP8]]
 ; ZVQDOTQ-NEXT:    [[PARTIAL_REDUCE]] = call <vscale x 1 x i32> @llvm.experimental.vector.partial.reduce.add.nxv1i32.nxv4i32(<vscale x 1 x i32> [[VEC_PHI]], <vscale x 4 x i32> [[TMP12]])
@@ -436,16 +416,14 @@ define i32 @vqdotsu(ptr %a, ptr %b) #0 {
 ; FIXED-V-NEXT:    [[VEC_PHI:%.*]] = phi <8 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-V-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-V-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; FIXED-V-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
 ; FIXED-V-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP0]], i32 8
-; FIXED-V-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP1]], align 1
+; FIXED-V-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP0]], align 1
 ; FIXED-V-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i8>, ptr [[TMP2]], align 1
 ; FIXED-V-NEXT:    [[TMP3:%.*]] = zext <8 x i8> [[WIDE_LOAD]] to <8 x i32>
 ; FIXED-V-NEXT:    [[TMP4:%.*]] = zext <8 x i8> [[WIDE_LOAD2]] to <8 x i32>
 ; FIXED-V-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; FIXED-V-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[TMP5]], i32 0
 ; FIXED-V-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP5]], i32 8
-; FIXED-V-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i8>, ptr [[TMP6]], align 1
+; FIXED-V-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i8>, ptr [[TMP5]], align 1
 ; FIXED-V-NEXT:    [[WIDE_LOAD4:%.*]] = load <8 x i8>, ptr [[TMP7]], align 1
 ; FIXED-V-NEXT:    [[TMP8:%.*]] = sext <8 x i8> [[WIDE_LOAD3]] to <8 x i32>
 ; FIXED-V-NEXT:    [[TMP9:%.*]] = sext <8 x i8> [[WIDE_LOAD4]] to <8 x i32>
@@ -459,7 +437,7 @@ define i32 @vqdotsu(ptr %a, ptr %b) #0 {
 ; FIXED-V:       middle.block:
 ; FIXED-V-NEXT:    [[BIN_RDX:%.*]] = add <8 x i32> [[TMP13]], [[TMP12]]
 ; FIXED-V-NEXT:    [[TMP15:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[BIN_RDX]])
-; FIXED-V-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; FIXED-V-NEXT:    br label [[FOR_EXIT:%.*]]
 ; FIXED-V:       scalar.ph:
 ;
 ; FIXED-ZVQDOTQ-LABEL: define i32 @vqdotsu(
@@ -473,16 +451,14 @@ define i32 @vqdotsu(ptr %a, ptr %b) #0 {
 ; FIXED-ZVQDOTQ-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-ZVQDOTQ-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE5:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; FIXED-ZVQDOTQ-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP0]], i32 8
-; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP1]], align 1
+; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP0]], align 1
 ; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i8>, ptr [[TMP2]], align 1
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP3:%.*]] = zext <8 x i8> [[WIDE_LOAD]] to <8 x i32>
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP4:%.*]] = zext <8 x i8> [[WIDE_LOAD2]] to <8 x i32>
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; FIXED-ZVQDOTQ-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[TMP5]], i32 0
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP5]], i32 8
-; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i8>, ptr [[TMP6]], align 1
+; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i8>, ptr [[TMP5]], align 1
 ; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD4:%.*]] = load <8 x i8>, ptr [[TMP7]], align 1
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP8:%.*]] = sext <8 x i8> [[WIDE_LOAD3]] to <8 x i32>
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP9:%.*]] = sext <8 x i8> [[WIDE_LOAD4]] to <8 x i32>
@@ -496,7 +472,7 @@ define i32 @vqdotsu(ptr %a, ptr %b) #0 {
 ; FIXED-ZVQDOTQ:       middle.block:
 ; FIXED-ZVQDOTQ-NEXT:    [[BIN_RDX:%.*]] = add <2 x i32> [[PARTIAL_REDUCE5]], [[PARTIAL_REDUCE]]
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP13:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[BIN_RDX]])
-; FIXED-ZVQDOTQ-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; FIXED-ZVQDOTQ-NEXT:    br label [[FOR_EXIT:%.*]]
 ; FIXED-ZVQDOTQ:       scalar.ph:
 ;
 entry:
@@ -541,12 +517,10 @@ define i32 @vqdotsu2(ptr %a, ptr %b) #0 {
 ; V-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; V-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
 ; V-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; V-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
-; V-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
+; V-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP6]], align 1
 ; V-NEXT:    [[TMP8:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; V-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; V-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[TMP9]], i32 0
-; V-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
+; V-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP9]], align 1
 ; V-NEXT:    [[TMP11:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; V-NEXT:    [[TMP12:%.*]] = mul <vscale x 4 x i32> [[TMP11]], [[TMP8]]
 ; V-NEXT:    [[TMP13]] = add <vscale x 4 x i32> [[TMP12]], [[VEC_PHI]]
@@ -578,12 +552,10 @@ define i32 @vqdotsu2(ptr %a, ptr %b) #0 {
 ; ZVQDOTQ-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; ZVQDOTQ-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 1 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; ZVQDOTQ-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; ZVQDOTQ-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP6]], i32 0
-; ZVQDOTQ-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP7]], align 1
+; ZVQDOTQ-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i8>, ptr [[TMP6]], align 1
 ; ZVQDOTQ-NEXT:    [[TMP8:%.*]] = sext <vscale x 4 x i8> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; ZVQDOTQ-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; ZVQDOTQ-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[TMP9]], i32 0
-; ZVQDOTQ-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP10]], align 1
+; ZVQDOTQ-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i8>, ptr [[TMP9]], align 1
 ; ZVQDOTQ-NEXT:    [[TMP11:%.*]] = zext <vscale x 4 x i8> [[WIDE_LOAD1]] to <vscale x 4 x i32>
 ; ZVQDOTQ-NEXT:    [[TMP12:%.*]] = mul <vscale x 4 x i32> [[TMP11]], [[TMP8]]
 ; ZVQDOTQ-NEXT:    [[PARTIAL_REDUCE]] = call <vscale x 1 x i32> @llvm.experimental.vector.partial.reduce.add.nxv1i32.nxv4i32(<vscale x 1 x i32> [[VEC_PHI]], <vscale x 4 x i32> [[TMP12]])
@@ -607,16 +579,14 @@ define i32 @vqdotsu2(ptr %a, ptr %b) #0 {
 ; FIXED-V-NEXT:    [[VEC_PHI:%.*]] = phi <8 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-V-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-V-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; FIXED-V-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
 ; FIXED-V-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP0]], i32 8
-; FIXED-V-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP1]], align 1
+; FIXED-V-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP0]], align 1
 ; FIXED-V-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i8>, ptr [[TMP2]], align 1
 ; FIXED-V-NEXT:    [[TMP3:%.*]] = sext <8 x i8> [[WIDE_LOAD]] to <8 x i32>
 ; FIXED-V-NEXT:    [[TMP4:%.*]] = sext <8 x i8> [[WIDE_LOAD2]] to <8 x i32>
 ; FIXED-V-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; FIXED-V-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[TMP5]], i32 0
 ; FIXED-V-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP5]], i32 8
-; FIXED-V-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i8>, ptr [[TMP6]], align 1
+; FIXED-V-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i8>, ptr [[TMP5]], align 1
 ; FIXED-V-NEXT:    [[WIDE_LOAD4:%.*]] = load <8 x i8>, ptr [[TMP7]], align 1
 ; FIXED-V-NEXT:    [[TMP8:%.*]] = zext <8 x i8> [[WIDE_LOAD3]] to <8 x i32>
 ; FIXED-V-NEXT:    [[TMP9:%.*]] = zext <8 x i8> [[WIDE_LOAD4]] to <8 x i32>
@@ -630,7 +600,7 @@ define i32 @vqdotsu2(ptr %a, ptr %b) #0 {
 ; FIXED-V:       middle.block:
 ; FIXED-V-NEXT:    [[BIN_RDX:%.*]] = add <8 x i32> [[TMP13]], [[TMP12]]
 ; FIXED-V-NEXT:    [[TMP15:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[BIN_RDX]])
-; FIXED-V-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; FIXED-V-NEXT:    br label [[FOR_EXIT:%.*]]
 ; FIXED-V:       scalar.ph:
 ;
 ; FIXED-ZVQDOTQ-LABEL: define i32 @vqdotsu2(
@@ -644,16 +614,14 @@ define i32 @vqdotsu2(ptr %a, ptr %b) #0 {
 ; FIXED-ZVQDOTQ-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-ZVQDOTQ-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PARTIAL_REDUCE5:%.*]], [[VECTOR_BODY]] ]
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; FIXED-ZVQDOTQ-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP0]], i32 8
-; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP1]], align 1
+; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP0]], align 1
 ; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i8>, ptr [[TMP2]], align 1
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP3:%.*]] = sext <8 x i8> [[WIDE_LOAD]] to <8 x i32>
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP4:%.*]] = sext <8 x i8> [[WIDE_LOAD2]] to <8 x i32>
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; FIXED-ZVQDOTQ-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[TMP5]], i32 0
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[TMP5]], i32 8
-; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i8>, ptr [[TMP6]], align 1
+; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i8>, ptr [[TMP5]], align 1
 ; FIXED-ZVQDOTQ-NEXT:    [[WIDE_LOAD4:%.*]] = load <8 x i8>, ptr [[TMP7]], align 1
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP8:%.*]] = zext <8 x i8> [[WIDE_LOAD3]] to <8 x i32>
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP9:%.*]] = zext <8 x i8> [[WIDE_LOAD4]] to <8 x i32>
@@ -667,7 +635,7 @@ define i32 @vqdotsu2(ptr %a, ptr %b) #0 {
 ; FIXED-ZVQDOTQ:       middle.block:
 ; FIXED-ZVQDOTQ-NEXT:    [[BIN_RDX:%.*]] = add <2 x i32> [[PARTIAL_REDUCE5]], [[PARTIAL_REDUCE]]
 ; FIXED-ZVQDOTQ-NEXT:    [[TMP13:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[BIN_RDX]])
-; FIXED-ZVQDOTQ-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; FIXED-ZVQDOTQ-NEXT:    br label [[FOR_EXIT:%.*]]
 ; FIXED-ZVQDOTQ:       scalar.ph:
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/pr87378-vpinstruction-or-drop-poison-generating-flags.ll b/llvm/test/Transforms/LoopVectorize/RISCV/pr87378-vpinstruction-or-drop-poison-generating-flags.ll
index 8088a65..b5b62d0 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/pr87378-vpinstruction-or-drop-poison-generating-flags.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/pr87378-vpinstruction-or-drop-poison-generating-flags.ll
@@ -49,8 +49,7 @@ define void @pr87378_vpinstruction_or_drop_poison_generating_flags(ptr %arg, i64
 ; CHECK-NEXT:    [[TMP23:%.*]] = extractelement <vscale x 8 x i1> [[TMP21]], i32 0
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select i1 [[TMP23]], i64 poison, i64 [[INDEX]]
 ; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr i16, ptr [[ARG]], i64 [[PREDPHI]]
-; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr i16, ptr [[TMP24]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.nxv8i16.p0(<vscale x 8 x i16> zeroinitializer, ptr [[TMP25]], i32 2, <vscale x 8 x i1> [[TMP22]])
+; CHECK-NEXT:    call void @llvm.masked.store.nxv8i16.p0(<vscale x 8 x i16> zeroinitializer, ptr [[TMP24]], i32 2, <vscale x 8 x i1> [[TMP22]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 8 x i64> [[VEC_IND]], [[DOTSPLAT]]
 ; CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/preserve-dbg-loc.ll b/llvm/test/Transforms/LoopVectorize/RISCV/preserve-dbg-loc.ll
index e4892db..f4817bb 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/preserve-dbg-loc.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/preserve-dbg-loc.ll
@@ -1,6 +1,5 @@
 ; RUN: opt -passes=debugify,loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -S < %s 2>&1 | FileCheck --check-prefix=DEBUGLOC %s
 
 ; Testing the debug locations of the generated vector intrinsic is same as
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/riscv-unroll.ll b/llvm/test/Transforms/LoopVectorize/RISCV/riscv-unroll.ll
index 0a367c0..642eef5 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/riscv-unroll.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/riscv-unroll.ll
@@ -21,15 +21,12 @@ define ptr @array_add(ptr noalias nocapture readonly %a, ptr noalias nocapture r
 ; LMUL1:       vector.body:
 ; LMUL1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; LMUL1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; LMUL1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
-; LMUL1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP3]], align 4
+; LMUL1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP2]], align 4
 ; LMUL1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
-; LMUL1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
-; LMUL1-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i32>, ptr [[TMP5]], align 4
+; LMUL1-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i32>, ptr [[TMP4]], align 4
 ; LMUL1-NEXT:    [[TMP6:%.*]] = add nsw <8 x i32> [[WIDE_LOAD1]], [[WIDE_LOAD]]
 ; LMUL1-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[INDEX]]
-; LMUL1-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; LMUL1-NEXT:    store <8 x i32> [[TMP6]], ptr [[TMP8]], align 4
+; LMUL1-NEXT:    store <8 x i32> [[TMP6]], ptr [[TMP7]], align 4
 ; LMUL1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; LMUL1-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; LMUL1-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -72,15 +69,12 @@ define ptr @array_add(ptr noalias nocapture readonly %a, ptr noalias nocapture r
 ; LMUL2:       vector.body:
 ; LMUL2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; LMUL2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; LMUL2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
-; LMUL2-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP3]], align 4
+; LMUL2-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP2]], align 4
 ; LMUL2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
-; LMUL2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
-; LMUL2-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i32>, ptr [[TMP5]], align 4
+; LMUL2-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i32>, ptr [[TMP4]], align 4
 ; LMUL2-NEXT:    [[TMP6:%.*]] = add nsw <8 x i32> [[WIDE_LOAD1]], [[WIDE_LOAD]]
 ; LMUL2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[INDEX]]
-; LMUL2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; LMUL2-NEXT:    store <8 x i32> [[TMP6]], ptr [[TMP8]], align 4
+; LMUL2-NEXT:    store <8 x i32> [[TMP6]], ptr [[TMP7]], align 4
 ; LMUL2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; LMUL2-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; LMUL2-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/riscv-vector-reverse.ll b/llvm/test/Transforms/LoopVectorize/RISCV/riscv-vector-reverse.ll
index f59ab56..0b3dcf8 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/riscv-vector-reverse.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/riscv-vector-reverse.ll
@@ -986,7 +986,7 @@ define void @vector_reverse_irregular_type(ptr noalias %A, ptr noalias %B) {
 ; RV64-NEXT:    [[TMP29:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1020
 ; RV64-NEXT:    br i1 [[TMP29]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; RV64:       [[MIDDLE_BLOCK]]:
-; RV64-NEXT:    br i1 false, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; RV64-NEXT:    br label %[[SCALAR_PH]]
 ; RV64:       [[SCALAR_PH]]:
 ; RV64-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, %[[MIDDLE_BLOCK]] ], [ 1023, %[[ENTRY]] ]
 ; RV64-NEXT:    br label %[[FOR_BODY:.*]]
@@ -1038,7 +1038,7 @@ define void @vector_reverse_irregular_type(ptr noalias %A, ptr noalias %B) {
 ; RV32-NEXT:    [[TMP29:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1020
 ; RV32-NEXT:    br i1 [[TMP29]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; RV32:       [[MIDDLE_BLOCK]]:
-; RV32-NEXT:    br i1 false, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; RV32-NEXT:    br label %[[SCALAR_PH]]
 ; RV32:       [[SCALAR_PH]]:
 ; RV32-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, %[[MIDDLE_BLOCK]] ], [ 1023, %[[ENTRY]] ]
 ; RV32-NEXT:    br label %[[FOR_BODY:.*]]
@@ -1123,7 +1123,7 @@ define void @vector_reverse_irregular_type(ptr noalias %A, ptr noalias %B) {
 ; RV64-UF2-NEXT:    [[TMP58:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1016
 ; RV64-UF2-NEXT:    br i1 [[TMP58]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; RV64-UF2:       [[MIDDLE_BLOCK]]:
-; RV64-UF2-NEXT:    br i1 false, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; RV64-UF2-NEXT:    br label %[[SCALAR_PH]]
 ; RV64-UF2:       [[SCALAR_PH]]:
 ; RV64-UF2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 7, %[[MIDDLE_BLOCK]] ], [ 1023, %[[ENTRY]] ]
 ; RV64-UF2-NEXT:    br label %[[FOR_BODY:.*]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/safe-dep-distance.ll b/llvm/test/Transforms/LoopVectorize/RISCV/safe-dep-distance.ll
index a901fe7..3370e92 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/safe-dep-distance.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/safe-dep-distance.ll
@@ -25,12 +25,10 @@ define void @test(ptr %p) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP8]], align 32
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 32
 ; CHECK-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX]], 200
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i64, ptr [[P]], i64 [[TMP9]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i64, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <vscale x 2 x i64> [[WIDE_LOAD]], ptr [[TMP11]], align 32
+; CHECK-NEXT:    store <vscale x 2 x i64> [[WIDE_LOAD]], ptr [[TMP10]], align 32
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -82,19 +80,17 @@ define void @test_may_clobber(ptr %p) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i64, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 32
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 32
 ; CHECK-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 100
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i64, ptr [[P]], i64 [[TMP3]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i64, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <4 x i64> [[WIDE_LOAD]], ptr [[TMP5]], align 32
+; CHECK-NEXT:    store <4 x i64> [[WIDE_LOAD]], ptr [[TMP4]], align 32
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 200
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 200, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -146,12 +142,10 @@ define void @trivial_due_max_vscale(ptr %p) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP8]], align 32
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 32
 ; CHECK-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX]], 8192
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i64, ptr [[P]], i64 [[TMP9]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i64, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <vscale x 2 x i64> [[WIDE_LOAD]], ptr [[TMP11]], align 32
+; CHECK-NEXT:    store <vscale x 2 x i64> [[WIDE_LOAD]], ptr [[TMP10]], align 32
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
@@ -211,12 +205,10 @@ define void @no_high_lmul_or_interleave(ptr %p) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP8]], align 32
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 32
 ; CHECK-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX]], 1024
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i64, ptr [[P]], i64 [[TMP9]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i64, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <vscale x 2 x i64> [[WIDE_LOAD]], ptr [[TMP11]], align 32
+; CHECK-NEXT:    store <vscale x 2 x i64> [[WIDE_LOAD]], ptr [[TMP10]], align 32
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/scalable-basics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/scalable-basics.ll
index c3c4abb..e51f6fa 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/scalable-basics.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/scalable-basics.ll
@@ -25,10 +25,9 @@ define void @vector_add(ptr noalias nocapture %a, i64 %v, i64 %n) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP6]], align 8
 ; CHECK-NEXT:    [[TMP8:%.*]] = add <vscale x 2 x i64> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
-; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP8]], ptr [[TMP7]], align 8
+; CHECK-NEXT:    store <vscale x 2 x i64> [[TMP8]], ptr [[TMP6]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -90,10 +89,9 @@ define void @vector_add_i32(ptr noalias nocapture %a, i32 %v, i64 %n) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = add <vscale x 4 x i32> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
-; CHECK-NEXT:    store <vscale x 4 x i32> [[TMP8]], ptr [[TMP7]], align 4
+; CHECK-NEXT:    store <vscale x 4 x i32> [[TMP8]], ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -193,8 +191,7 @@ define void @indexed_store(ptr noalias nocapture %a, ptr noalias nocapture %b, i
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[B:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP6]], align 8
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], <vscale x 2 x i64> [[WIDE_LOAD]]
 ; CHECK-NEXT:    call void @llvm.masked.scatter.nxv2i64.nxv2p0(<vscale x 2 x i64> [[BROADCAST_SPLAT]], <vscale x 2 x ptr> [[TMP8]], i32 8, <vscale x 2 x i1> splat (i1 true))
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -254,8 +251,7 @@ define i64 @indexed_load(ptr noalias nocapture %a, ptr noalias nocapture %b, i64
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[B:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP6]], align 8
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], <vscale x 2 x i64> [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 2 x i64> @llvm.masked.gather.nxv2i64.nxv2p0(<vscale x 2 x ptr> [[TMP8]], i32 8, <vscale x 2 x i1> splat (i1 true), <vscale x 2 x i64> poison)
 ; CHECK-NEXT:    [[TMP9]] = add <vscale x 2 x i64> [[VEC_PHI]], [[WIDE_MASKED_GATHER]]
@@ -324,8 +320,7 @@ define void @splat_int(ptr noalias nocapture %a, i64 %v, i64 %n) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP7]], align 8
+; CHECK-NEXT:    store <vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP6]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
@@ -380,8 +375,7 @@ define void @splat_ptr(ptr noalias nocapture %a, ptr %v, i64 %n) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds ptr, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <vscale x 2 x ptr> [[BROADCAST_SPLAT]], ptr [[TMP7]], align 8
+; CHECK-NEXT:    store <vscale x 2 x ptr> [[BROADCAST_SPLAT]], ptr [[TMP6]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/scalable-tailfold.ll b/llvm/test/Transforms/LoopVectorize/RISCV/scalable-tailfold.ll
index b046f61..f6f8f6a 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/scalable-tailfold.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/scalable-tailfold.ll
@@ -1,5 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
-; RUN: opt < %s -passes=loop-vectorize -scalable-vectorization=on -prefer-predicate-over-epilogue=predicate-dont-vectorize -mtriple riscv64-linux-gnu -mattr=+v,+f -S 2>%t | FileCheck %s -check-prefix=CHECK
+; RUN: opt < %s -passes=loop-vectorize -scalable-vectorization=on -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue -mtriple riscv64-linux-gnu -mattr=+v,+f -S 2>%t | FileCheck %s -check-prefix=CHECK
 
 ; Exercise tail folding on RISCV w/scalable vectors.
 
@@ -26,14 +26,13 @@ define void @vector_add(ptr noalias nocapture %a, i64 %v, i64 %n) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[AVL:%.*]] = sub i64 1025, [[EVL_BASED_IV]]
-; CHECK-NEXT:    [[TMP7:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP9]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP7]])
-; CHECK-NEXT:    [[TMP10:%.*]] = add <vscale x 2 x i64> [[WIDE_MASKED_LOAD]], [[BROADCAST_SPLAT]]
-; CHECK-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP10]], ptr align 8 [[TMP9]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP7]])
-; CHECK-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP7]] to i64
-; CHECK-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP12]], [[EVL_BASED_IV]]
+; CHECK-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
+; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
+; CHECK-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP10]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP9]])
+; CHECK-NEXT:    [[TMP7:%.*]] = add <vscale x 2 x i64> [[VP_OP_LOAD]], [[BROADCAST_SPLAT]]
+; CHECK-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP7]], ptr align 8 [[TMP10]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP9]])
+; CHECK-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP9]] to i64
+; CHECK-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP8]], [[EVL_BASED_IV]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -95,8 +94,7 @@ define void @indexed_store(ptr noalias nocapture %a, ptr noalias nocapture %b, i
 ; CHECK-NEXT:    [[AVL:%.*]] = sub i64 1025, [[EVL_BASED_IV]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[B:%.*]], i64 [[EVL_BASED_IV]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP9]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP7]])
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP8]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP7]])
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], <vscale x 2 x i64> [[WIDE_MASKED_LOAD]]
 ; CHECK-NEXT:    call void @llvm.vp.scatter.nxv2i64.nxv2p0(<vscale x 2 x i64> [[BROADCAST_SPLAT]], <vscale x 2 x ptr> align 8 [[TMP10]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP7]])
 ; CHECK-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP7]] to i64
@@ -159,8 +157,7 @@ define i64 @indexed_load(ptr noalias nocapture %a, ptr noalias nocapture %b, i64
 ; CHECK-NEXT:    [[AVL:%.*]] = sub i64 1025, [[EVL_BASED_IV]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[B:%.*]], i64 [[EVL_BASED_IV]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP9]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP7]])
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP8]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP7]])
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], <vscale x 2 x i64> [[WIDE_MASKED_LOAD]]
 ; CHECK-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 2 x i64> @llvm.vp.gather.nxv2i64.nxv2p0(<vscale x 2 x ptr> align 8 [[TMP10]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP7]])
 ; CHECK-NEXT:    [[TMP12:%.*]] = add <vscale x 2 x i64> [[VEC_PHI]], [[WIDE_MASKED_GATHER]]
@@ -231,12 +228,11 @@ define void @splat_int(ptr noalias nocapture %a, i64 %v, i64 %n) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[AVL:%.*]] = sub i64 1025, [[EVL_BASED_IV]]
-; CHECK-NEXT:    [[TMP7:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr align 8 [[TMP9]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP7]])
-; CHECK-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP7]] to i64
-; CHECK-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP11]], [[EVL_BASED_IV]]
+; CHECK-NEXT:    [[TMP8:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
+; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
+; CHECK-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr align 8 [[TMP9]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP8]])
+; CHECK-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP8]] to i64
+; CHECK-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP7]], [[EVL_BASED_IV]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
@@ -293,10 +289,9 @@ define void @uniform_store(ptr noalias nocapture %a, ptr noalias nocapture %b, i
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
 ; CHECK-NEXT:    store i64 [[V]], ptr [[B:%.*]], align 8
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr align 8 [[TMP9]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP7]])
-; CHECK-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP7]] to i64
-; CHECK-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP11]], [[EVL_BASED_IV]]
+; CHECK-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr align 8 [[TMP8]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP7]])
+; CHECK-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP7]] to i64
+; CHECK-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP9]], [[EVL_BASED_IV]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
@@ -335,17 +330,44 @@ for.end:
 define i64 @uniform_load(ptr noalias nocapture %a, ptr noalias nocapture %b, i64 %n) {
 ; CHECK-LABEL: @uniform_load(
 ; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 2
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1025, [[TMP1]]
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 2
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 1025, [[TMP3]]
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 1025, [[N_MOD_VF]]
+; CHECK-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; CHECK-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
-; CHECK:       for.body:
-; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-NEXT:    [[V:%.*]] = load i64, ptr [[B:%.*]], align 8
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[V]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[IV]]
-; CHECK-NEXT:    store i64 [[V]], ptr [[ARRAYIDX]], align 8
-; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
-; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025
-; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END:%.*]], label [[FOR_BODY]]
+; CHECK-NEXT:    store <vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[ARRAYIDX]], align 8
+; CHECK-NEXT:    [[IV_NEXT]] = add nuw i64 [[IV]], [[TMP5]]
+; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 1025, [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
+; CHECK-NEXT:    br label [[FOR_BODY1:%.*]]
+; CHECK:       for.body:
+; CHECK-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ]
+; CHECK-NEXT:    [[V1:%.*]] = load i64, ptr [[B]], align 8
+; CHECK-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV1]]
+; CHECK-NEXT:    store i64 [[V1]], ptr [[ARRAYIDX1]], align 8
+; CHECK-NEXT:    [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT1]], 1025
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP14:![0-9]+]]
 ; CHECK:       for.end:
-; CHECK-NEXT:    [[V_LCSSA:%.*]] = phi i64 [ [[V]], [[FOR_BODY]] ]
+; CHECK-NEXT:    [[V_LCSSA:%.*]] = phi i64 [ [[V1]], [[FOR_BODY1]] ], [ [[V]], [[MIDDLE_BLOCK]] ]
 ; CHECK-NEXT:    ret i64 [[V_LCSSA]]
 ;
 entry:
@@ -385,17 +407,16 @@ define void @vector_add_trip1024(ptr noalias nocapture %a, i64 %v, i64 %n) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[AVL:%.*]] = sub i64 1024, [[EVL_BASED_IV]]
-; CHECK-NEXT:    [[TMP7:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP9]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP7]])
-; CHECK-NEXT:    [[TMP10:%.*]] = add <vscale x 2 x i64> [[WIDE_MASKED_LOAD]], [[BROADCAST_SPLAT]]
-; CHECK-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP10]], ptr align 8 [[TMP9]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP7]])
-; CHECK-NEXT:    [[TMP12:%.*]] = zext i32 [[TMP7]] to i64
-; CHECK-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP12]], [[EVL_BASED_IV]]
+; CHECK-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
+; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
+; CHECK-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP10]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP9]])
+; CHECK-NEXT:    [[TMP7:%.*]] = add <vscale x 2 x i64> [[VP_OP_LOAD]], [[BROADCAST_SPLAT]]
+; CHECK-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP7]], ptr align 8 [[TMP10]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP9]])
+; CHECK-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP9]] to i64
+; CHECK-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP8]], [[EVL_BASED_IV]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
+; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
@@ -409,7 +430,7 @@ define void @vector_add_trip1024(ptr noalias nocapture %a, i64 %v, i64 %n) {
 ; CHECK-NEXT:    store i64 [[ADD]], ptr [[ARRAYIDX]], align 8
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1024
-; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; CHECK:       for.end:
 ; CHECK-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/select-cmp-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/select-cmp-reduction.ll
index 01b4ad2..745b8ba 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/select-cmp-reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/select-cmp-reduction.ll
@@ -17,8 +17,7 @@ define i32 @select_icmp(i32 %x, i32 %y, ptr nocapture readonly %c, i64 %n) #0 {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp sge <4 x i32> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI]], [[TMP4]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -45,8 +44,7 @@ define i32 @select_icmp(i32 %x, i32 %y, ptr nocapture readonly %c, i64 %n) #0 {
 ; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; SCALABLE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; SCALABLE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[INDEX]]
-; SCALABLE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP6]], align 4
+; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP5]], align 4
 ; SCALABLE-NEXT:    [[TMP8:%.*]] = icmp sge <vscale x 4 x i32> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
 ; SCALABLE-NEXT:    [[TMP9]] = or <vscale x 4 x i1> [[VEC_PHI]], [[TMP8]]
 ; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP11]]
@@ -88,8 +86,7 @@ define i32 @select_fcmp(float %x, i32 %y, ptr nocapture readonly %c, i64 %n) #0
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[C:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = fcmp fast uge <4 x float> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI]], [[TMP4]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -116,8 +113,7 @@ define i32 @select_fcmp(float %x, i32 %y, ptr nocapture readonly %c, i64 %n) #0
 ; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; SCALABLE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; SCALABLE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[C:%.*]], i64 [[INDEX]]
-; SCALABLE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i32 0
-; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP6]], align 4
+; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP5]], align 4
 ; SCALABLE-NEXT:    [[TMP8:%.*]] = fcmp fast uge <vscale x 4 x float> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
 ; SCALABLE-NEXT:    [[TMP9]] = or <vscale x 4 x i1> [[VEC_PHI]], [[TMP8]]
 ; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP11]]
@@ -157,8 +153,7 @@ define i32 @select_const_i32_from_icmp(ptr nocapture readonly %v, i64 %n) #0 {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[V:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp ne <4 x i32> [[WIDE_LOAD]], splat (i32 3)
 ; CHECK-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI]], [[TMP4]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -183,8 +178,7 @@ define i32 @select_const_i32_from_icmp(ptr nocapture readonly %v, i64 %n) #0 {
 ; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; SCALABLE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; SCALABLE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[V:%.*]], i64 [[INDEX]]
-; SCALABLE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP6]], align 4
+; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP5]], align 4
 ; SCALABLE-NEXT:    [[TMP8:%.*]] = icmp ne <vscale x 4 x i32> [[WIDE_LOAD]], splat (i32 3)
 ; SCALABLE-NEXT:    [[TMP9]] = or <vscale x 4 x i1> [[VEC_PHI]], [[TMP8]]
 ; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP11]]
@@ -224,8 +218,7 @@ define i32 @select_i32_from_icmp(ptr nocapture readonly %v, i32 %a, i32 %b, i64
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[V:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp ne <4 x i32> [[WIDE_LOAD]], splat (i32 3)
 ; CHECK-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI]], [[TMP4]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -250,8 +243,7 @@ define i32 @select_i32_from_icmp(ptr nocapture readonly %v, i32 %a, i32 %b, i64
 ; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; SCALABLE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; SCALABLE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[V:%.*]], i64 [[INDEX]]
-; SCALABLE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP6]], align 4
+; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP5]], align 4
 ; SCALABLE-NEXT:    [[TMP8:%.*]] = icmp ne <vscale x 4 x i32> [[WIDE_LOAD]], splat (i32 3)
 ; SCALABLE-NEXT:    [[TMP9]] = or <vscale x 4 x i1> [[VEC_PHI]], [[TMP8]]
 ; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP11]]
@@ -291,8 +283,7 @@ define i32 @select_const_i32_from_fcmp(ptr nocapture readonly %v, i64 %n) #0 {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[V:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = fcmp fast one <4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00)
 ; CHECK-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI]], [[TMP4]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -317,8 +308,7 @@ define i32 @select_const_i32_from_fcmp(ptr nocapture readonly %v, i64 %n) #0 {
 ; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; SCALABLE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; SCALABLE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[V:%.*]], i64 [[INDEX]]
-; SCALABLE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i32 0
-; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP6]], align 4
+; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP5]], align 4
 ; SCALABLE-NEXT:    [[TMP8:%.*]] = fcmp fast one <vscale x 4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00)
 ; SCALABLE-NEXT:    [[TMP9]] = or <vscale x 4 x i1> [[VEC_PHI]], [[TMP8]]
 ; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP11]]
@@ -383,12 +373,10 @@ define i32 @pred_select_const_i32_from_icmp(ptr noalias nocapture readonly %src1
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PREDPHI:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[SRC1:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp sgt <4 x i32> [[WIDE_LOAD]], splat (i32 35)
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[SRC2:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP6]], i32 4, <4 x i1> [[TMP4]], <4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP5]], i32 4, <4 x i1> [[TMP4]], <4 x i32> poison)
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq <4 x i32> [[WIDE_MASKED_LOAD]], splat (i32 2)
 ; CHECK-NEXT:    [[TMP9:%.*]] = or <4 x i1> [[VEC_PHI]], [[TMP8]]
 ; CHECK-NEXT:    [[PREDPHI]] = select <4 x i1> [[TMP4]], <4 x i1> [[TMP9]], <4 x i1> [[VEC_PHI]]
@@ -414,12 +402,10 @@ define i32 @pred_select_const_i32_from_icmp(ptr noalias nocapture readonly %src1
 ; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; SCALABLE-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PREDPHI:%.*]], [[VECTOR_BODY]] ]
 ; SCALABLE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[SRC1:%.*]], i64 [[INDEX]]
-; SCALABLE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP6]], align 4
+; SCALABLE-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP5]], align 4
 ; SCALABLE-NEXT:    [[TMP8:%.*]] = icmp sgt <vscale x 4 x i32> [[WIDE_LOAD]], splat (i32 35)
 ; SCALABLE-NEXT:    [[TMP9:%.*]] = getelementptr i32, ptr [[SRC2:%.*]], i64 [[INDEX]]
-; SCALABLE-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[TMP9]], i32 0
-; SCALABLE-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP10]], i32 4, <vscale x 4 x i1> [[TMP8]], <vscale x 4 x i32> poison)
+; SCALABLE-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP9]], i32 4, <vscale x 4 x i1> [[TMP8]], <vscale x 4 x i32> poison)
 ; SCALABLE-NEXT:    [[TMP12:%.*]] = icmp eq <vscale x 4 x i32> [[WIDE_MASKED_LOAD]], splat (i32 2)
 ; SCALABLE-NEXT:    [[TMP13:%.*]] = or <vscale x 4 x i1> [[VEC_PHI]], [[TMP12]]
 ; SCALABLE-NEXT:    [[PREDPHI]] = select <vscale x 4 x i1> [[TMP8]], <vscale x 4 x i1> [[TMP13]], <vscale x 4 x i1> [[VEC_PHI]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/short-trip-count.ll b/llvm/test/Transforms/LoopVectorize/RISCV/short-trip-count.ll
index d9598bc..13a4b16 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/short-trip-count.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/short-trip-count.ll
@@ -8,20 +8,18 @@ define void @small_trip_count_min_vlen_128(ptr nocapture %a) nounwind vscale_ran
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TMP0:%.*]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1:%.*]], align 4
 ; CHECK-NEXT:    [[TMP2:%.*]] = add nsw <4 x i32> [[WIDE_LOAD]], splat (i32 1)
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP2]], ptr [[TMP3]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP2]], ptr [[TMP1]], align 4
 ; CHECK-NEXT:    br label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 4, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP1:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[LOOP1]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
-; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 [[IV]]
+; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 [[IV]]
 ; CHECK-NEXT:    [[V:%.*]] = load i32, ptr [[GEP]], align 4
 ; CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 [[V]], 1
 ; CHECK-NEXT:    store i32 [[ADD]], ptr [[GEP]], align 4
@@ -55,20 +53,18 @@ define void @small_trip_count_min_vlen_32(ptr nocapture %a) nounwind vscale_rang
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TMP0:%.*]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1:%.*]], align 4
 ; CHECK-NEXT:    [[TMP2:%.*]] = add nsw <4 x i32> [[WIDE_LOAD]], splat (i32 1)
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP2]], ptr [[TMP3]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP2]], ptr [[TMP1]], align 4
 ; CHECK-NEXT:    br label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 4, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP1:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[LOOP1]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
-; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 [[IV]]
+; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 [[IV]]
 ; CHECK-NEXT:    [[V:%.*]] = load i32, ptr [[GEP]], align 4
 ; CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 [[V]], 1
 ; CHECK-NEXT:    store i32 [[ADD]], ptr [[GEP]], align 4
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/strided-accesses.ll b/llvm/test/Transforms/LoopVectorize/RISCV/strided-accesses.ll
index 25dac36..80f0274 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/strided-accesses.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/strided-accesses.ll
@@ -250,10 +250,9 @@ define void @single_stride_int_scaled(ptr %p, i64 %stride) {
 ; NOSTRIDED:       vector.body:
 ; NOSTRIDED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NOSTRIDED-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[P:%.*]], i64 [[INDEX]]
-; NOSTRIDED-NEXT:    [[TMP9:%.*]] = getelementptr i32, ptr [[TMP8]], i32 0
-; NOSTRIDED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP9]], align 4
+; NOSTRIDED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
 ; NOSTRIDED-NEXT:    [[TMP10:%.*]] = add <vscale x 4 x i32> [[WIDE_LOAD]], splat (i32 1)
-; NOSTRIDED-NEXT:    store <vscale x 4 x i32> [[TMP10]], ptr [[TMP9]], align 4
+; NOSTRIDED-NEXT:    store <vscale x 4 x i32> [[TMP10]], ptr [[TMP8]], align 4
 ; NOSTRIDED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
 ; NOSTRIDED-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; NOSTRIDED-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
@@ -332,10 +331,9 @@ define void @single_stride_int_iv(ptr %p, i64 %stride) {
 ; NOSTRIDED:       vector.body:
 ; NOSTRIDED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NOSTRIDED-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[P:%.*]], i64 [[INDEX]]
-; NOSTRIDED-NEXT:    [[TMP9:%.*]] = getelementptr i32, ptr [[TMP8]], i32 0
-; NOSTRIDED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP9]], align 4
+; NOSTRIDED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
 ; NOSTRIDED-NEXT:    [[TMP10:%.*]] = add <vscale x 4 x i32> [[WIDE_LOAD]], splat (i32 1)
-; NOSTRIDED-NEXT:    store <vscale x 4 x i32> [[TMP10]], ptr [[TMP9]], align 4
+; NOSTRIDED-NEXT:    store <vscale x 4 x i32> [[TMP10]], ptr [[TMP8]], align 4
 ; NOSTRIDED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
 ; NOSTRIDED-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; NOSTRIDED-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
@@ -463,12 +461,10 @@ define void @double_stride_int_scaled(ptr %p, ptr %p2, i64 %stride) {
 ; NOSTRIDED:       vector.body:
 ; NOSTRIDED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NOSTRIDED-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr [[P]], i64 [[INDEX]]
-; NOSTRIDED-NEXT:    [[TMP13:%.*]] = getelementptr i32, ptr [[TMP12]], i32 0
-; NOSTRIDED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP13]], align 4
+; NOSTRIDED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP12]], align 4
 ; NOSTRIDED-NEXT:    [[TMP14:%.*]] = add <vscale x 4 x i32> [[WIDE_LOAD]], splat (i32 1)
 ; NOSTRIDED-NEXT:    [[TMP15:%.*]] = getelementptr i32, ptr [[P2]], i64 [[INDEX]]
-; NOSTRIDED-NEXT:    [[TMP16:%.*]] = getelementptr i32, ptr [[TMP15]], i32 0
-; NOSTRIDED-NEXT:    store <vscale x 4 x i32> [[TMP14]], ptr [[TMP16]], align 4
+; NOSTRIDED-NEXT:    store <vscale x 4 x i32> [[TMP14]], ptr [[TMP15]], align 4
 ; NOSTRIDED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP10]]
 ; NOSTRIDED-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; NOSTRIDED-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
@@ -636,10 +632,9 @@ define void @double_stride_int_iv(ptr %p, ptr %p2, i64 %stride) {
 ; NOSTRIDED:       vector.body:
 ; NOSTRIDED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NOSTRIDED-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[P:%.*]], i64 [[INDEX]]
-; NOSTRIDED-NEXT:    [[TMP9:%.*]] = getelementptr i32, ptr [[TMP8]], i32 0
-; NOSTRIDED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP9]], align 4
+; NOSTRIDED-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
 ; NOSTRIDED-NEXT:    [[TMP10:%.*]] = add <vscale x 4 x i32> [[WIDE_LOAD]], splat (i32 1)
-; NOSTRIDED-NEXT:    store <vscale x 4 x i32> [[TMP10]], ptr [[TMP9]], align 4
+; NOSTRIDED-NEXT:    store <vscale x 4 x i32> [[TMP10]], ptr [[TMP8]], align 4
 ; NOSTRIDED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP6]]
 ; NOSTRIDED-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; NOSTRIDED-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-bin-unary-ops-args.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-bin-unary-ops-args.ll
index 16c575f..f9395aa 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-bin-unary-ops-args.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-bin-unary-ops-args.ll
@@ -1,12 +1,10 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s --check-prefix=IF-EVL
 
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s --check-prefix=NO-VP
 
 
@@ -39,12 +37,10 @@ define void @test_and(ptr nocapture %a, ptr nocapture readonly %b) {
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = sub i64 100, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP10]], i32 16, i1 true)
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP14]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP13]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = and <vscale x 16 x i8> [[VP_OP_LOAD]], splat (i8 1)
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP17]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP16]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP11]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
@@ -71,9 +67,45 @@ define void @test_and(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-LABEL: define void @test_and(
 ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0:[0-9]+]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16
+; NO-VP-NEXT:    [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16
+; NO-VP-NEXT:    [[TMP4:%.*]] = sub i64 [[B1]], [[A2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1
+; NO-VP-NEXT:    [[TMP10:%.*]] = and <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 1)
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[DEC]] = add nsw i64 [[LEN]], 1
 ; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
@@ -81,7 +113,7 @@ define void @test_and(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]]
 ; NO-VP-NEXT:    store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1
 ; NO-VP-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100
-; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
 ; NO-VP:       [[FINISH_LOOPEXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -132,12 +164,10 @@ define void @test_or(ptr nocapture %a, ptr nocapture readonly %b) {
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = sub i64 100, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP10]], i32 16, i1 true)
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP14]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP13]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = or <vscale x 16 x i8> [[VP_OP_LOAD]], splat (i8 1)
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP17]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP16]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP11]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
@@ -164,9 +194,45 @@ define void @test_or(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-LABEL: define void @test_or(
 ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16
+; NO-VP-NEXT:    [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16
+; NO-VP-NEXT:    [[TMP4:%.*]] = sub i64 [[B1]], [[A2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1
+; NO-VP-NEXT:    [[TMP10:%.*]] = or <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 1)
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[DEC]] = add nsw i64 [[LEN]], 1
 ; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
@@ -174,7 +240,7 @@ define void @test_or(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]]
 ; NO-VP-NEXT:    store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1
 ; NO-VP-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100
-; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
 ; NO-VP:       [[FINISH_LOOPEXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -225,12 +291,10 @@ define void @test_xor(ptr nocapture %a, ptr nocapture readonly %b) {
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = sub i64 100, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP10]], i32 16, i1 true)
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP14]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP13]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = xor <vscale x 16 x i8> [[VP_OP_LOAD]], splat (i8 1)
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP17]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP16]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP11]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
@@ -257,9 +321,45 @@ define void @test_xor(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-LABEL: define void @test_xor(
 ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16
+; NO-VP-NEXT:    [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16
+; NO-VP-NEXT:    [[TMP4:%.*]] = sub i64 [[B1]], [[A2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1
+; NO-VP-NEXT:    [[TMP10:%.*]] = xor <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 1)
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[DEC]] = add nsw i64 [[LEN]], 1
 ; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
@@ -267,7 +367,7 @@ define void @test_xor(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]]
 ; NO-VP-NEXT:    store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1
 ; NO-VP-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100
-; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP7:![0-9]+]]
 ; NO-VP:       [[FINISH_LOOPEXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -318,12 +418,10 @@ define void @test_shl(ptr nocapture %a, ptr nocapture readonly %b) {
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = sub i64 100, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP10]], i32 16, i1 true)
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP14]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP13]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = shl <vscale x 16 x i8> [[VP_OP_LOAD]], splat (i8 1)
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP17]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP16]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP11]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
@@ -350,9 +448,45 @@ define void @test_shl(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-LABEL: define void @test_shl(
 ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16
+; NO-VP-NEXT:    [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16
+; NO-VP-NEXT:    [[TMP4:%.*]] = sub i64 [[B1]], [[A2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1
+; NO-VP-NEXT:    [[TMP10:%.*]] = shl <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 1)
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[DEC]] = add nsw i64 [[LEN]], 1
 ; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
@@ -360,7 +494,7 @@ define void @test_shl(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]]
 ; NO-VP-NEXT:    store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1
 ; NO-VP-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100
-; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP9:![0-9]+]]
 ; NO-VP:       [[FINISH_LOOPEXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -411,12 +545,10 @@ define void @test_lshr(ptr nocapture %a, ptr nocapture readonly %b) {
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = sub i64 100, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP10]], i32 16, i1 true)
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP14]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP13]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = lshr <vscale x 16 x i8> [[VP_OP_LOAD]], splat (i8 1)
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP17]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP16]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP11]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
@@ -443,9 +575,45 @@ define void @test_lshr(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-LABEL: define void @test_lshr(
 ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16
+; NO-VP-NEXT:    [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16
+; NO-VP-NEXT:    [[TMP4:%.*]] = sub i64 [[B1]], [[A2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1
+; NO-VP-NEXT:    [[TMP10:%.*]] = lshr <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 1)
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[DEC]] = add nsw i64 [[LEN]], 1
 ; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
@@ -453,7 +621,7 @@ define void @test_lshr(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]]
 ; NO-VP-NEXT:    store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1
 ; NO-VP-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100
-; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP11:![0-9]+]]
 ; NO-VP:       [[FINISH_LOOPEXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -504,12 +672,10 @@ define void @test_ashr(ptr nocapture %a, ptr nocapture readonly %b) {
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = sub i64 100, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP10]], i32 16, i1 true)
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP14]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP13]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = ashr <vscale x 16 x i8> [[VP_OP_LOAD]], splat (i8 1)
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP17]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP16]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP11]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
@@ -536,9 +702,45 @@ define void @test_ashr(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-LABEL: define void @test_ashr(
 ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16
+; NO-VP-NEXT:    [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16
+; NO-VP-NEXT:    [[TMP4:%.*]] = sub i64 [[B1]], [[A2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1
+; NO-VP-NEXT:    [[TMP10:%.*]] = ashr <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 1)
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[DEC]] = add nsw i64 [[LEN]], 1
 ; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
@@ -546,7 +748,7 @@ define void @test_ashr(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]]
 ; NO-VP-NEXT:    store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1
 ; NO-VP-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100
-; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP13:![0-9]+]]
 ; NO-VP:       [[FINISH_LOOPEXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -597,12 +799,10 @@ define void @test_add(ptr nocapture %a, ptr nocapture readonly %b) {
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = sub i64 100, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP10]], i32 16, i1 true)
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP14]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP13]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = add <vscale x 16 x i8> [[VP_OP_LOAD]], splat (i8 1)
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP17]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP16]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP11]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
@@ -629,9 +829,45 @@ define void @test_add(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-LABEL: define void @test_add(
 ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16
+; NO-VP-NEXT:    [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16
+; NO-VP-NEXT:    [[TMP4:%.*]] = sub i64 [[B1]], [[A2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1
+; NO-VP-NEXT:    [[TMP10:%.*]] = add <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 1)
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[DEC]] = add nsw i64 [[LEN]], 1
 ; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
@@ -639,7 +875,7 @@ define void @test_add(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]]
 ; NO-VP-NEXT:    store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1
 ; NO-VP-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100
-; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP15:![0-9]+]]
 ; NO-VP:       [[FINISH_LOOPEXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -690,12 +926,10 @@ define void @test_sub(ptr nocapture %a, ptr nocapture readonly %b) {
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = sub i64 100, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP10]], i32 16, i1 true)
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP14]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP13]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = sub <vscale x 16 x i8> [[VP_OP_LOAD]], splat (i8 1)
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP17]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP16]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP11]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
@@ -722,9 +956,45 @@ define void @test_sub(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-LABEL: define void @test_sub(
 ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16
+; NO-VP-NEXT:    [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16
+; NO-VP-NEXT:    [[TMP4:%.*]] = sub i64 [[B1]], [[A2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1
+; NO-VP-NEXT:    [[TMP10:%.*]] = sub <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 1)
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[DEC]] = add nsw i64 [[LEN]], 1
 ; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
@@ -732,7 +1002,7 @@ define void @test_sub(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]]
 ; NO-VP-NEXT:    store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1
 ; NO-VP-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100
-; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP17:![0-9]+]]
 ; NO-VP:       [[FINISH_LOOPEXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -783,12 +1053,10 @@ define void @test_mul(ptr nocapture %a, ptr nocapture readonly %b) {
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = sub i64 100, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP10]], i32 16, i1 true)
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP14]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP13]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = mul <vscale x 16 x i8> [[VP_OP_LOAD]], splat (i8 3)
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP17]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP16]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP11]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
@@ -815,9 +1083,45 @@ define void @test_mul(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-LABEL: define void @test_mul(
 ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16
+; NO-VP-NEXT:    [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16
+; NO-VP-NEXT:    [[TMP4:%.*]] = sub i64 [[B1]], [[A2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 3)
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[DEC]] = add nsw i64 [[LEN]], 1
 ; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
@@ -825,7 +1129,7 @@ define void @test_mul(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]]
 ; NO-VP-NEXT:    store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1
 ; NO-VP-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100
-; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP19:![0-9]+]]
 ; NO-VP:       [[FINISH_LOOPEXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -876,12 +1180,10 @@ define void @test_sdiv(ptr nocapture %a, ptr nocapture readonly %b) {
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = sub i64 100, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP10]], i32 16, i1 true)
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP14]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP13]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = sdiv <vscale x 16 x i8> [[VP_OP_LOAD]], splat (i8 3)
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP17]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP16]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP11]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
@@ -908,9 +1210,45 @@ define void @test_sdiv(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-LABEL: define void @test_sdiv(
 ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16
+; NO-VP-NEXT:    [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16
+; NO-VP-NEXT:    [[TMP4:%.*]] = sub i64 [[B1]], [[A2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1
+; NO-VP-NEXT:    [[TMP10:%.*]] = sdiv <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 3)
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[DEC]] = add nsw i64 [[LEN]], 1
 ; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
@@ -918,7 +1256,7 @@ define void @test_sdiv(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]]
 ; NO-VP-NEXT:    store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1
 ; NO-VP-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100
-; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP21:![0-9]+]]
 ; NO-VP:       [[FINISH_LOOPEXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -969,12 +1307,10 @@ define void @test_udiv(ptr nocapture %a, ptr nocapture readonly %b) {
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = sub i64 100, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP10]], i32 16, i1 true)
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP14]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP13]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = udiv <vscale x 16 x i8> [[VP_OP_LOAD]], splat (i8 3)
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP17]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP16]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP11]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
@@ -1001,9 +1337,45 @@ define void @test_udiv(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-LABEL: define void @test_udiv(
 ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16
+; NO-VP-NEXT:    [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16
+; NO-VP-NEXT:    [[TMP4:%.*]] = sub i64 [[B1]], [[A2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1
+; NO-VP-NEXT:    [[TMP10:%.*]] = udiv <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 3)
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[DEC]] = add nsw i64 [[LEN]], 1
 ; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
@@ -1011,7 +1383,7 @@ define void @test_udiv(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]]
 ; NO-VP-NEXT:    store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1
 ; NO-VP-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100
-; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP23:![0-9]+]]
 ; NO-VP:       [[FINISH_LOOPEXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -1062,12 +1434,10 @@ define void @test_srem(ptr nocapture %a, ptr nocapture readonly %b) {
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = sub i64 100, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP10]], i32 16, i1 true)
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP14]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP13]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = srem <vscale x 16 x i8> [[VP_OP_LOAD]], splat (i8 3)
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP17]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP16]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP11]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
@@ -1094,9 +1464,45 @@ define void @test_srem(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-LABEL: define void @test_srem(
 ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16
+; NO-VP-NEXT:    [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16
+; NO-VP-NEXT:    [[TMP4:%.*]] = sub i64 [[B1]], [[A2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1
+; NO-VP-NEXT:    [[TMP10:%.*]] = srem <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 3)
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[DEC]] = add nsw i64 [[LEN]], 1
 ; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
@@ -1104,7 +1510,7 @@ define void @test_srem(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]]
 ; NO-VP-NEXT:    store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1
 ; NO-VP-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100
-; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP25:![0-9]+]]
 ; NO-VP:       [[FINISH_LOOPEXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -1155,12 +1561,10 @@ define void @test_urem(ptr nocapture %a, ptr nocapture readonly %b) {
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = sub i64 100, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP10]], i32 16, i1 true)
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP14]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 16 x i8> @llvm.vp.load.nxv16i8.p0(ptr align 1 [[TMP13]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = urem <vscale x 16 x i8> [[VP_OP_LOAD]], splat (i8 3)
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP17]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv16i8.p0(<vscale x 16 x i8> [[VP_OP]], ptr align 1 [[TMP16]], <vscale x 16 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP11]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
@@ -1187,9 +1591,45 @@ define void @test_urem(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-LABEL: define void @test_urem(
 ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 16
+; NO-VP-NEXT:    [[TMP14:%.*]] = call i64 @llvm.umax.i64(i64 32, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP14]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 16
+; NO-VP-NEXT:    [[TMP4:%.*]] = sub i64 [[B1]], [[A2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP4]], [[TMP3]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 16
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 100, [[TMP6]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 16
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 16 x i8>, ptr [[TMP9]], align 1
+; NO-VP-NEXT:    [[TMP10:%.*]] = urem <vscale x 16 x i8> [[WIDE_LOAD]], splat (i8 3)
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 16 x i8> [[TMP10]], ptr [[TMP11]], align 1
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[DEC]] = add nsw i64 [[LEN]], 1
 ; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[LEN]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i8, ptr [[ARRAYIDX]], align 1
@@ -1197,7 +1637,7 @@ define void @test_urem(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[LEN]]
 ; NO-VP-NEXT:    store i8 [[TMP]], ptr [[ARRAYIDX1]], align 1
 ; NO-VP-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100
-; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP27:![0-9]+]]
 ; NO-VP:       [[FINISH_LOOPEXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -1251,12 +1691,10 @@ define void @test_fadd(ptr nocapture %a, ptr nocapture readonly %b) {
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = sub i64 100, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP11]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP14]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP15]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = fadd fast <vscale x 4 x float> [[VP_OP_LOAD]], splat (float 3.000000e+00)
 ; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4f32.p0(<vscale x 4 x float> [[VP_OP]], ptr align 4 [[TMP18]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4f32.p0(<vscale x 4 x float> [[VP_OP]], ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP12]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP19]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP10]]
@@ -1283,9 +1721,46 @@ define void @test_fadd(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-LABEL: define void @test_fadd(
 ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP15]], 4
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[B1]], [[A2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 100, [[TMP8]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4
+; NO-VP-NEXT:    [[TMP12:%.*]] = fadd fast <vscale x 4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00)
+; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 4 x float> [[TMP12]], ptr [[TMP13]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP10]]
+; NO-VP-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[DEC]] = add nsw i64 [[LEN]], 1
 ; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[LEN]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4
@@ -1293,7 +1768,7 @@ define void @test_fadd(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[LEN]]
 ; NO-VP-NEXT:    store float [[TMP]], ptr [[ARRAYIDX1]], align 4
 ; NO-VP-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100
-; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP29:![0-9]+]]
 ; NO-VP:       [[FINISH_LOOPEXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -1345,12 +1820,10 @@ define void @test_fsub(ptr nocapture %a, ptr nocapture readonly %b) {
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = sub i64 100, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP11]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP14]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP15]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = fsub fast <vscale x 4 x float> [[VP_OP_LOAD]], splat (float 3.000000e+00)
 ; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4f32.p0(<vscale x 4 x float> [[VP_OP]], ptr align 4 [[TMP18]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4f32.p0(<vscale x 4 x float> [[VP_OP]], ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP12]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP19]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP10]]
@@ -1377,9 +1850,46 @@ define void @test_fsub(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-LABEL: define void @test_fsub(
 ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP15]], 4
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[B1]], [[A2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 100, [[TMP8]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4
+; NO-VP-NEXT:    [[TMP12:%.*]] = fsub fast <vscale x 4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00)
+; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 4 x float> [[TMP12]], ptr [[TMP13]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP10]]
+; NO-VP-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[DEC]] = add nsw i64 [[LEN]], 1
 ; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[LEN]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4
@@ -1387,7 +1897,7 @@ define void @test_fsub(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[LEN]]
 ; NO-VP-NEXT:    store float [[TMP]], ptr [[ARRAYIDX1]], align 4
 ; NO-VP-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100
-; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP31:![0-9]+]]
 ; NO-VP:       [[FINISH_LOOPEXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -1439,12 +1949,10 @@ define void @test_fmul(ptr nocapture %a, ptr nocapture readonly %b) {
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = sub i64 100, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP11]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP14]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP15]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = fmul fast <vscale x 4 x float> [[VP_OP_LOAD]], splat (float 3.000000e+00)
 ; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4f32.p0(<vscale x 4 x float> [[VP_OP]], ptr align 4 [[TMP18]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4f32.p0(<vscale x 4 x float> [[VP_OP]], ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP12]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP19]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP10]]
@@ -1471,9 +1979,46 @@ define void @test_fmul(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-LABEL: define void @test_fmul(
 ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP15]], 4
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[B1]], [[A2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 100, [[TMP8]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4
+; NO-VP-NEXT:    [[TMP12:%.*]] = fmul fast <vscale x 4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00)
+; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 4 x float> [[TMP12]], ptr [[TMP13]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP10]]
+; NO-VP-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[DEC]] = add nsw i64 [[LEN]], 1
 ; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[LEN]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4
@@ -1481,7 +2026,7 @@ define void @test_fmul(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[LEN]]
 ; NO-VP-NEXT:    store float [[TMP]], ptr [[ARRAYIDX1]], align 4
 ; NO-VP-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100
-; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP33:![0-9]+]]
 ; NO-VP:       [[FINISH_LOOPEXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -1533,12 +2078,10 @@ define void @test_fdiv(ptr nocapture %a, ptr nocapture readonly %b) {
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = sub i64 100, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP11]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP14]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP15]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = fdiv fast <vscale x 4 x float> [[VP_OP_LOAD]], splat (float 3.000000e+00)
 ; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4f32.p0(<vscale x 4 x float> [[VP_OP]], ptr align 4 [[TMP18]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4f32.p0(<vscale x 4 x float> [[VP_OP]], ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP12]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP19]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP10]]
@@ -1565,9 +2108,46 @@ define void @test_fdiv(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-LABEL: define void @test_fdiv(
 ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP15]], 4
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[B1]], [[A2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 100, [[TMP8]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4
+; NO-VP-NEXT:    [[TMP12:%.*]] = fdiv fast <vscale x 4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00)
+; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 4 x float> [[TMP12]], ptr [[TMP13]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP10]]
+; NO-VP-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[DEC]] = add nsw i64 [[LEN]], 1
 ; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[LEN]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4
@@ -1575,7 +2155,7 @@ define void @test_fdiv(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[LEN]]
 ; NO-VP-NEXT:    store float [[TMP]], ptr [[ARRAYIDX1]], align 4
 ; NO-VP-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100
-; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP35:![0-9]+]]
 ; NO-VP:       [[FINISH_LOOPEXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -1680,12 +2260,10 @@ define void @test_fneg(ptr nocapture %a, ptr nocapture readonly %b) {
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = sub i64 100, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP11]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP14]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP15]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = fneg fast <vscale x 4 x float> [[VP_OP_LOAD]]
 ; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4f32.p0(<vscale x 4 x float> [[VP_OP]], ptr align 4 [[TMP18]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4f32.p0(<vscale x 4 x float> [[VP_OP]], ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP12]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP19]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP10]]
@@ -1712,9 +2290,46 @@ define void @test_fneg(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-LABEL: define void @test_fneg(
 ; NO-VP-SAME: ptr captures(none) [[A:%.*]], ptr readonly captures(none) [[B:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[A2:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[B1:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP15]], 4
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 100, [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[B1]], [[A2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 100, [[TMP8]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 100, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4
+; NO-VP-NEXT:    [[TMP12:%.*]] = fneg fast <vscale x 4 x float> [[WIDE_LOAD]]
+; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 4 x float> [[TMP12]], ptr [[TMP13]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP10]]
+; NO-VP-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 100, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[FINISH_LOOPEXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[LEN:%.*]] = phi i64 [ [[DEC:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[DEC]] = add nsw i64 [[LEN]], 1
 ; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[LEN]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4
@@ -1722,7 +2337,7 @@ define void @test_fneg(ptr nocapture %a, ptr nocapture readonly %b) {
 ; NO-VP-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[LEN]]
 ; NO-VP-NEXT:    store float [[TMP]], ptr [[ARRAYIDX1]], align 4
 ; NO-VP-NEXT:    [[DOTNOT:%.*]] = icmp eq i64 [[DEC]], 100
-; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DOTNOT]], label %[[FINISH_LOOPEXIT]], label %[[LOOP]], !llvm.loop [[LOOP37:![0-9]+]]
 ; NO-VP:       [[FINISH_LOOPEXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -1784,3 +2399,42 @@ finish.loopexit:
 ; IF-EVL: [[LOOP37]] = distinct !{[[LOOP37]], [[META1]], [[META2]], [[META3]]}
 ; IF-EVL: [[LOOP38]] = distinct !{[[LOOP38]], [[META1]]}
 ;.
+; NO-VP: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; NO-VP: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; NO-VP: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+; NO-VP: [[LOOP3]] = distinct !{[[LOOP3]], [[META1]]}
+; NO-VP: [[LOOP4]] = distinct !{[[LOOP4]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP5]] = distinct !{[[LOOP5]], [[META1]]}
+; NO-VP: [[LOOP6]] = distinct !{[[LOOP6]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP7]] = distinct !{[[LOOP7]], [[META1]]}
+; NO-VP: [[LOOP8]] = distinct !{[[LOOP8]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP9]] = distinct !{[[LOOP9]], [[META1]]}
+; NO-VP: [[LOOP10]] = distinct !{[[LOOP10]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP11]] = distinct !{[[LOOP11]], [[META1]]}
+; NO-VP: [[LOOP12]] = distinct !{[[LOOP12]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP13]] = distinct !{[[LOOP13]], [[META1]]}
+; NO-VP: [[LOOP14]] = distinct !{[[LOOP14]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP15]] = distinct !{[[LOOP15]], [[META1]]}
+; NO-VP: [[LOOP16]] = distinct !{[[LOOP16]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP17]] = distinct !{[[LOOP17]], [[META1]]}
+; NO-VP: [[LOOP18]] = distinct !{[[LOOP18]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP19]] = distinct !{[[LOOP19]], [[META1]]}
+; NO-VP: [[LOOP20]] = distinct !{[[LOOP20]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP21]] = distinct !{[[LOOP21]], [[META1]]}
+; NO-VP: [[LOOP22]] = distinct !{[[LOOP22]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP23]] = distinct !{[[LOOP23]], [[META1]]}
+; NO-VP: [[LOOP24]] = distinct !{[[LOOP24]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP25]] = distinct !{[[LOOP25]], [[META1]]}
+; NO-VP: [[LOOP26]] = distinct !{[[LOOP26]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP27]] = distinct !{[[LOOP27]], [[META1]]}
+; NO-VP: [[LOOP28]] = distinct !{[[LOOP28]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP29]] = distinct !{[[LOOP29]], [[META1]]}
+; NO-VP: [[LOOP30]] = distinct !{[[LOOP30]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP31]] = distinct !{[[LOOP31]], [[META1]]}
+; NO-VP: [[LOOP32]] = distinct !{[[LOOP32]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP33]] = distinct !{[[LOOP33]], [[META1]]}
+; NO-VP: [[LOOP34]] = distinct !{[[LOOP34]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP35]] = distinct !{[[LOOP35]], [[META1]]}
+; NO-VP: [[LOOP36]] = distinct !{[[LOOP36]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP37]] = distinct !{[[LOOP37]], [[META1]]}
+;.
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-call-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-call-intrinsics.ll
index 7a5415a..7d21c4d 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-call-intrinsics.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-call-intrinsics.ll
@@ -1,12 +1,10 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s --check-prefix=IF-EVL
 
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s --check-prefix=NO-VP
 
 define void @vp_smax(ptr %a, ptr %b, ptr %c, i64 %N) {
@@ -44,15 +42,12 @@ define void @vp_smax(ptr %a, ptr %b, ptr %c, i64 %N) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD5:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD5:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP29:%.*]] = call <vscale x 4 x i32> @llvm.smax.nxv4i32(<vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i32> [[VP_OP_LOAD5]])
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP29]], ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP29]], ptr align 4 [[TMP16]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP9]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP8]]
@@ -81,9 +76,53 @@ define void @vp_smax(ptr %a, ptr %b, ptr %c, i64 %N) {
 ; NO-VP-LABEL: define void @vp_smax(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], ptr [[C:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[C3:%.*]] = ptrtoint ptr [[C]] to i64
+; NO-VP-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[TMP18:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP19:%.*]] = mul nuw i64 [[TMP18]], 4
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP19]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[A1]], [[B2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP8:%.*]] = sub i64 [[A1]], [[C3]]
+; NO-VP-NEXT:    [[DIFF_CHECK4:%.*]] = icmp ult i64 [[TMP8]], [[TMP7]]
+; NO-VP-NEXT:    [[CONFLICT_RDX:%.*]] = or i1 [[DIFF_CHECK]], [[DIFF_CHECK4]]
+; NO-VP-NEXT:    br i1 [[CONFLICT_RDX]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP10]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 4
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP13]], align 4
+; NO-VP-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 4 x i32>, ptr [[TMP14]], align 4
+; NO-VP-NEXT:    [[TMP15:%.*]] = call <vscale x 4 x i32> @llvm.smax.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[WIDE_LOAD5]])
+; NO-VP-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP15]], ptr [[TMP16]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
+; NO-VP-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP17]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ]
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4
 ; NO-VP-NEXT:    [[GEP3:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[IV]]
@@ -93,7 +132,7 @@ define void @vp_smax(ptr %a, ptr %b, ptr %c, i64 %N) {
 ; NO-VP-NEXT:    store i32 [[DOT]], ptr [[GEP11]], align 4
 ; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -153,15 +192,12 @@ define void @vp_smin(ptr %a, ptr %b, ptr %c, i64 %N) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD5:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD5:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP29:%.*]] = call <vscale x 4 x i32> @llvm.smin.nxv4i32(<vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i32> [[VP_OP_LOAD5]])
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP29]], ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP29]], ptr align 4 [[TMP16]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP9]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP8]]
@@ -190,9 +226,53 @@ define void @vp_smin(ptr %a, ptr %b, ptr %c, i64 %N) {
 ; NO-VP-LABEL: define void @vp_smin(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], ptr [[C:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[C3:%.*]] = ptrtoint ptr [[C]] to i64
+; NO-VP-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[TMP18:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP19:%.*]] = mul nuw i64 [[TMP18]], 4
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP19]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[A1]], [[B2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP8:%.*]] = sub i64 [[A1]], [[C3]]
+; NO-VP-NEXT:    [[DIFF_CHECK4:%.*]] = icmp ult i64 [[TMP8]], [[TMP7]]
+; NO-VP-NEXT:    [[CONFLICT_RDX:%.*]] = or i1 [[DIFF_CHECK]], [[DIFF_CHECK4]]
+; NO-VP-NEXT:    br i1 [[CONFLICT_RDX]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP10]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 4
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP13]], align 4
+; NO-VP-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 4 x i32>, ptr [[TMP14]], align 4
+; NO-VP-NEXT:    [[TMP15:%.*]] = call <vscale x 4 x i32> @llvm.smin.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[WIDE_LOAD5]])
+; NO-VP-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP15]], ptr [[TMP16]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
+; NO-VP-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP17]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ]
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4
 ; NO-VP-NEXT:    [[GEP3:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[IV]]
@@ -202,7 +282,7 @@ define void @vp_smin(ptr %a, ptr %b, ptr %c, i64 %N) {
 ; NO-VP-NEXT:    store i32 [[DOT]], ptr [[GEP11]], align 4
 ; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -262,15 +342,12 @@ define void @vp_umax(ptr %a, ptr %b, ptr %c, i64 %N) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD5:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD5:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP29:%.*]] = call <vscale x 4 x i32> @llvm.umax.nxv4i32(<vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i32> [[VP_OP_LOAD5]])
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP29]], ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP29]], ptr align 4 [[TMP16]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP9]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP8]]
@@ -299,9 +376,53 @@ define void @vp_umax(ptr %a, ptr %b, ptr %c, i64 %N) {
 ; NO-VP-LABEL: define void @vp_umax(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], ptr [[C:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[C3:%.*]] = ptrtoint ptr [[C]] to i64
+; NO-VP-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[TMP18:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP19:%.*]] = mul nuw i64 [[TMP18]], 4
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP19]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[A1]], [[B2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP8:%.*]] = sub i64 [[A1]], [[C3]]
+; NO-VP-NEXT:    [[DIFF_CHECK4:%.*]] = icmp ult i64 [[TMP8]], [[TMP7]]
+; NO-VP-NEXT:    [[CONFLICT_RDX:%.*]] = or i1 [[DIFF_CHECK]], [[DIFF_CHECK4]]
+; NO-VP-NEXT:    br i1 [[CONFLICT_RDX]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP10]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 4
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP13]], align 4
+; NO-VP-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 4 x i32>, ptr [[TMP14]], align 4
+; NO-VP-NEXT:    [[TMP15:%.*]] = call <vscale x 4 x i32> @llvm.umax.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[WIDE_LOAD5]])
+; NO-VP-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP15]], ptr [[TMP16]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
+; NO-VP-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP17]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ]
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4
 ; NO-VP-NEXT:    [[GEP3:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[IV]]
@@ -311,7 +432,7 @@ define void @vp_umax(ptr %a, ptr %b, ptr %c, i64 %N) {
 ; NO-VP-NEXT:    store i32 [[DOT]], ptr [[GEP11]], align 4
 ; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP7:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -371,15 +492,12 @@ define void @vp_umin(ptr %a, ptr %b, ptr %c, i64 %N) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD5:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD5:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP29:%.*]] = call <vscale x 4 x i32> @llvm.umin.nxv4i32(<vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i32> [[VP_OP_LOAD5]])
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP29]], ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP29]], ptr align 4 [[TMP16]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP9]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP8]]
@@ -408,9 +526,53 @@ define void @vp_umin(ptr %a, ptr %b, ptr %c, i64 %N) {
 ; NO-VP-LABEL: define void @vp_umin(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], ptr [[C:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[C3:%.*]] = ptrtoint ptr [[C]] to i64
+; NO-VP-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[TMP18:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP19:%.*]] = mul nuw i64 [[TMP18]], 4
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP19]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[A1]], [[B2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP8:%.*]] = sub i64 [[A1]], [[C3]]
+; NO-VP-NEXT:    [[DIFF_CHECK4:%.*]] = icmp ult i64 [[TMP8]], [[TMP7]]
+; NO-VP-NEXT:    [[CONFLICT_RDX:%.*]] = or i1 [[DIFF_CHECK]], [[DIFF_CHECK4]]
+; NO-VP-NEXT:    br i1 [[CONFLICT_RDX]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP10]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 4
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP13]], align 4
+; NO-VP-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD5:%.*]] = load <vscale x 4 x i32>, ptr [[TMP14]], align 4
+; NO-VP-NEXT:    [[TMP15:%.*]] = call <vscale x 4 x i32> @llvm.umin.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[WIDE_LOAD5]])
+; NO-VP-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP15]], ptr [[TMP16]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP12]]
+; NO-VP-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP17]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ]
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4
 ; NO-VP-NEXT:    [[GEP3:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[IV]]
@@ -420,7 +582,7 @@ define void @vp_umin(ptr %a, ptr %b, ptr %c, i64 %N) {
 ; NO-VP-NEXT:    store i32 [[DOT]], ptr [[GEP11]], align 4
 ; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP9:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -476,12 +638,10 @@ define void @vp_ctlz(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP24:%.*]] = call <vscale x 4 x i32> @llvm.ctlz.nxv4i32(<vscale x 4 x i32> [[VP_OP_LOAD]], i1 true)
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP24]], ptr align 4 [[TMP15]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP24]], ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP9]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP16]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP8]]
@@ -508,17 +668,54 @@ define void @vp_ctlz(ptr %a, ptr %b, i64 %N) {
 ; NO-VP-LABEL: define void @vp_ctlz(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP11]], 4
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 8, i64 [[TMP15]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[A1]], [[B2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[TMP12:%.*]] = call <vscale x 4 x i32> @llvm.ctlz.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], i1 true)
+; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP12]], ptr [[TMP13]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]]
+; NO-VP-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ]
-; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
-; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
+; NO-VP-NEXT:    [[GEP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV1]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP1]], align 4
 ; NO-VP-NEXT:    [[TMP1:%.*]] = tail call range(i32 0, 33) i32 @llvm.ctlz.i32(i32 [[TMP0]], i1 true)
-; NO-VP-NEXT:    [[GEP3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    [[GEP3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]]
 ; NO-VP-NEXT:    store i32 [[TMP1]], ptr [[GEP3]], align 4
-; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP11:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -571,12 +768,10 @@ define void @vp_cttz(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[TMP15]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP16]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP13]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP15]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP13]])
 ; IF-EVL-NEXT:    [[TMP17:%.*]] = call <vscale x 4 x i32> @llvm.cttz.nxv4i32(<vscale x 4 x i32> [[VP_OP_LOAD]], i1 true)
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[TMP18]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP17]], ptr align 4 [[TMP19]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP13]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP17]], ptr align 4 [[TMP18]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP13]])
 ; IF-EVL-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP13]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP20]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP12]]
@@ -603,17 +798,54 @@ define void @vp_cttz(ptr %a, ptr %b, i64 %N) {
 ; NO-VP-LABEL: define void @vp_cttz(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP11]], 4
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 8, i64 [[TMP15]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[A1]], [[B2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[TMP12:%.*]] = call <vscale x 4 x i32> @llvm.cttz.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], i1 true)
+; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP12]], ptr [[TMP13]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]]
+; NO-VP-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ]
-; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
-; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
+; NO-VP-NEXT:    [[GEP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV1]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP1]], align 4
 ; NO-VP-NEXT:    [[TMP1:%.*]] = tail call range(i32 0, 33) i32 @llvm.cttz.i32(i32 [[TMP0]], i1 true)
-; NO-VP-NEXT:    [[GEP3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    [[GEP3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]]
 ; NO-VP-NEXT:    store i32 [[TMP1]], ptr [[GEP3]], align 4
-; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP13:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -666,14 +898,12 @@ define void @vp_lrint(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP11]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP27:%.*]] = fpext <vscale x 4 x float> [[VP_OP_LOAD]] to <vscale x 4 x double>
 ; IF-EVL-NEXT:    [[TMP28:%.*]] = call <vscale x 4 x i64> @llvm.lrint.nxv4i64.nxv4f64(<vscale x 4 x double> [[TMP27]])
 ; IF-EVL-NEXT:    [[TMP15:%.*]] = trunc <vscale x 4 x i64> [[TMP28]] to <vscale x 4 x i32>
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP15]], ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP15]], ptr align 4 [[TMP16]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP9]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP8]]
@@ -702,19 +932,58 @@ define void @vp_lrint(ptr %a, ptr %b, i64 %N) {
 ; NO-VP-LABEL: define void @vp_lrint(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP17:%.*]] = mul nuw i64 [[TMP11]], 4
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP17]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[A1]], [[B2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[TMP12:%.*]] = fpext <vscale x 4 x float> [[WIDE_LOAD]] to <vscale x 4 x double>
+; NO-VP-NEXT:    [[TMP13:%.*]] = call <vscale x 4 x i64> @llvm.lrint.nxv4i64.nxv4f64(<vscale x 4 x double> [[TMP12]])
+; NO-VP-NEXT:    [[TMP14:%.*]] = trunc <vscale x 4 x i64> [[TMP13]] to <vscale x 4 x i32>
+; NO-VP-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP14]], ptr [[TMP15]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]]
+; NO-VP-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP16]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ]
-; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
+; NO-VP-NEXT:    [[GEP1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV1]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[GEP1]], align 4
 ; NO-VP-NEXT:    [[CONV2:%.*]] = fpext float [[TMP0]] to double
 ; NO-VP-NEXT:    [[TMP1:%.*]] = tail call i64 @llvm.lrint.i64.f64(double [[CONV2]])
 ; NO-VP-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP1]] to i32
-; NO-VP-NEXT:    [[GEP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    [[GEP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]]
 ; NO-VP-NEXT:    store i32 [[CONV3]], ptr [[GEP5]], align 4
-; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP15:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -769,14 +1038,12 @@ define void @vp_llrint(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP11]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP27:%.*]] = fpext <vscale x 4 x float> [[VP_OP_LOAD]] to <vscale x 4 x double>
 ; IF-EVL-NEXT:    [[TMP28:%.*]] = call <vscale x 4 x i64> @llvm.llrint.nxv4i64.nxv4f64(<vscale x 4 x double> [[TMP27]])
 ; IF-EVL-NEXT:    [[TMP15:%.*]] = trunc <vscale x 4 x i64> [[TMP28]] to <vscale x 4 x i32>
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP15]], ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP15]], ptr align 4 [[TMP16]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP9]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP8]]
@@ -805,19 +1072,58 @@ define void @vp_llrint(ptr %a, ptr %b, i64 %N) {
 ; NO-VP-LABEL: define void @vp_llrint(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP17:%.*]] = mul nuw i64 [[TMP11]], 4
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP17]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[A1]], [[B2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[TMP12:%.*]] = fpext <vscale x 4 x float> [[WIDE_LOAD]] to <vscale x 4 x double>
+; NO-VP-NEXT:    [[TMP13:%.*]] = call <vscale x 4 x i64> @llvm.llrint.nxv4i64.nxv4f64(<vscale x 4 x double> [[TMP12]])
+; NO-VP-NEXT:    [[TMP14:%.*]] = trunc <vscale x 4 x i64> [[TMP13]] to <vscale x 4 x i32>
+; NO-VP-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP14]], ptr [[TMP15]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]]
+; NO-VP-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP16]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ]
-; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
+; NO-VP-NEXT:    [[GEP1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV1]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[GEP1]], align 4
 ; NO-VP-NEXT:    [[CONV2:%.*]] = fpext float [[TMP0]] to double
 ; NO-VP-NEXT:    [[TMP1:%.*]] = tail call i64 @llvm.llrint.i64.f64(double [[CONV2]])
 ; NO-VP-NEXT:    [[CONV3:%.*]] = trunc i64 [[TMP1]] to i32
-; NO-VP-NEXT:    [[GEP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    [[GEP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]]
 ; NO-VP-NEXT:    store i32 [[CONV3]], ptr [[GEP5]], align 4
-; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP17:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -872,12 +1178,10 @@ define void @vp_abs(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP24:%.*]] = call <vscale x 4 x i32> @llvm.abs.nxv4i32(<vscale x 4 x i32> [[VP_OP_LOAD]], i1 true)
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP24]], ptr align 4 [[TMP15]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP24]], ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP9]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP16]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP8]]
@@ -904,17 +1208,54 @@ define void @vp_abs(ptr %a, ptr %b, i64 %N) {
 ; NO-VP-LABEL: define void @vp_abs(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP11]], 4
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 8, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[A1]], [[B2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[TMP12:%.*]] = call <vscale x 4 x i32> @llvm.abs.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], i1 true)
+; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP12]], ptr [[TMP13]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]]
+; NO-VP-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ]
-; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
-; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
+; NO-VP-NEXT:    [[GEP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV1]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP1]], align 4
 ; NO-VP-NEXT:    [[COND:%.*]] = tail call i32 @llvm.abs.i32(i32 [[TMP0]], i1 true)
-; NO-VP-NEXT:    [[GEP9:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    [[GEP9:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]]
 ; NO-VP-NEXT:    store i32 [[COND]], ptr [[GEP9]], align 4
-; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP19:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -959,17 +1300,44 @@ define void @log10(ptr %a, ptr %b, i64 %N) {
 ; NO-VP-LABEL: define void @log10(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 8
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP1:%.*]] = sub i64 [[A1]], [[B2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP1]], 16
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[TMP2:%.*]] = call <4 x float> @llvm.log10.v4f32(<4 x float> [[WIDE_LOAD]])
+; NO-VP-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store <4 x float> [[TMP2]], ptr [[TMP3]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], 4
+; NO-VP-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ]
-; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
+; NO-VP-NEXT:    [[GEP1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV1]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[GEP1]], align 4
 ; NO-VP-NEXT:    [[COND:%.*]] = tail call float @llvm.log10.f32(float [[TMP0]])
-; NO-VP-NEXT:    [[GEP9:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    [[GEP9:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV1]]
 ; NO-VP-NEXT:    store float [[COND]], ptr [[GEP9]], align 4
-; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP21:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -1025,3 +1393,26 @@ declare i32 @llvm.abs.i32(i32, i1 immarg)
 ; IF-EVL: [[LOOP19]] = distinct !{[[LOOP19]], [[META1]], [[META2]], [[META3]]}
 ; IF-EVL: [[LOOP20]] = distinct !{[[LOOP20]], [[META1]]}
 ;.
+; NO-VP: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; NO-VP: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; NO-VP: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+; NO-VP: [[LOOP3]] = distinct !{[[LOOP3]], [[META1]]}
+; NO-VP: [[LOOP4]] = distinct !{[[LOOP4]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP5]] = distinct !{[[LOOP5]], [[META1]]}
+; NO-VP: [[LOOP6]] = distinct !{[[LOOP6]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP7]] = distinct !{[[LOOP7]], [[META1]]}
+; NO-VP: [[LOOP8]] = distinct !{[[LOOP8]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP9]] = distinct !{[[LOOP9]], [[META1]]}
+; NO-VP: [[LOOP10]] = distinct !{[[LOOP10]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP11]] = distinct !{[[LOOP11]], [[META1]]}
+; NO-VP: [[LOOP12]] = distinct !{[[LOOP12]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP13]] = distinct !{[[LOOP13]], [[META1]]}
+; NO-VP: [[LOOP14]] = distinct !{[[LOOP14]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP15]] = distinct !{[[LOOP15]], [[META1]]}
+; NO-VP: [[LOOP16]] = distinct !{[[LOOP16]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP17]] = distinct !{[[LOOP17]], [[META1]]}
+; NO-VP: [[LOOP18]] = distinct !{[[LOOP18]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP19]] = distinct !{[[LOOP19]], [[META1]]}
+; NO-VP: [[LOOP20]] = distinct !{[[LOOP20]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP21]] = distinct !{[[LOOP21]], [[META1]]}
+;.
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-cast-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-cast-intrinsics.ll
index 091eb87..4b718dd 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-cast-intrinsics.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-cast-intrinsics.ll
@@ -1,12 +1,10 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s --check-prefix=IF-EVL
 
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s --check-prefix=NO-VP
 
 define void @vp_sext(ptr %a, ptr %b, i64 %N) {
@@ -39,12 +37,10 @@ define void @vp_sext(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i32> @llvm.vp.load.nxv2i32.p0(ptr align 4 [[TMP15]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META0:![0-9]+]]
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i32> @llvm.vp.load.nxv2i32.p0(ptr align 4 [[TMP14]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META0:![0-9]+]]
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = sext <vscale x 2 x i32> [[VP_OP_LOAD]] to <vscale x 2 x i64>
 ; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i64, ptr [[TMP17]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP16]], ptr align 8 [[TMP18]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META3:![0-9]+]], !noalias [[META0]]
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP16]], ptr align 8 [[TMP17]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META3:![0-9]+]], !noalias [[META0]]
 ; IF-EVL-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP12]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP19]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP11]]
@@ -71,17 +67,54 @@ define void @vp_sext(ptr %a, ptr %b, i64 %N) {
 ; NO-VP-LABEL: define void @vp_sext(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0:[0-9]+]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP9]], 2
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 20, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = shl i64 [[N]], 3
+; NO-VP-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]]
+; NO-VP-NEXT:    [[TMP4:%.*]] = shl i64 [[N]], 2
+; NO-VP-NEXT:    [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP4]]
+; NO-VP-NEXT:    [[BOUND0:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]]
+; NO-VP-NEXT:    [[BOUND1:%.*]] = icmp ult ptr [[B]], [[SCEVGEP]]
+; NO-VP-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; NO-VP-NEXT:    br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 2
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP6]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 2
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i32>, ptr [[GEP]], align 4, !alias.scope [[META0:![0-9]+]]
+; NO-VP-NEXT:    [[TMP10:%.*]] = sext <vscale x 2 x i32> [[WIDE_LOAD]] to <vscale x 2 x i64>
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store <vscale x 2 x i64> [[TMP10]], ptr [[TMP11]], align 8, !alias.scope [[META3:![0-9]+]], !noalias [[META0]]
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP8]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[ENTRY]] ]
-; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
-; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
+; NO-VP-NEXT:    [[GEP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV1]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP1]], align 4
 ; NO-VP-NEXT:    [[CONV2:%.*]] = sext i32 [[TMP0]] to i64
-; NO-VP-NEXT:    [[GEP4:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    [[GEP4:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV1]]
 ; NO-VP-NEXT:    store i64 [[CONV2]], ptr [[GEP4]], align 8
-; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP8:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -133,12 +166,10 @@ define void @vp_zext(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i32> @llvm.vp.load.nxv2i32.p0(ptr align 4 [[TMP15]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META10:![0-9]+]]
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i32> @llvm.vp.load.nxv2i32.p0(ptr align 4 [[TMP14]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META10:![0-9]+]]
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = zext <vscale x 2 x i32> [[VP_OP_LOAD]] to <vscale x 2 x i64>
 ; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i64, ptr [[TMP17]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP16]], ptr align 8 [[TMP18]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META13:![0-9]+]], !noalias [[META10]]
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP16]], ptr align 8 [[TMP17]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META13:![0-9]+]], !noalias [[META10]]
 ; IF-EVL-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP12]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP19]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP11]]
@@ -165,17 +196,54 @@ define void @vp_zext(ptr %a, ptr %b, i64 %N) {
 ; NO-VP-LABEL: define void @vp_zext(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP9]], 2
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 20, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = shl i64 [[N]], 3
+; NO-VP-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]]
+; NO-VP-NEXT:    [[TMP4:%.*]] = shl i64 [[N]], 2
+; NO-VP-NEXT:    [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP4]]
+; NO-VP-NEXT:    [[BOUND0:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]]
+; NO-VP-NEXT:    [[BOUND1:%.*]] = icmp ult ptr [[B]], [[SCEVGEP]]
+; NO-VP-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; NO-VP-NEXT:    br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 2
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP6]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 2
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i32>, ptr [[GEP]], align 4, !alias.scope [[META9:![0-9]+]]
+; NO-VP-NEXT:    [[TMP10:%.*]] = zext <vscale x 2 x i32> [[WIDE_LOAD]] to <vscale x 2 x i64>
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store <vscale x 2 x i64> [[TMP10]], ptr [[TMP11]], align 8, !alias.scope [[META12:![0-9]+]], !noalias [[META9]]
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP8]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
-; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
-; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[GEP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV1]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP1]], align 4
 ; NO-VP-NEXT:    [[CONV:%.*]] = zext i32 [[TMP0]] to i64
-; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV1]]
 ; NO-VP-NEXT:    store i64 [[CONV]], ptr [[GEP2]], align 8
-; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP15:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -227,12 +295,10 @@ define void @vp_trunc(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i64, ptr [[TMP14]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP15]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META17:![0-9]+]]
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP14]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META17:![0-9]+]]
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = trunc <vscale x 2 x i64> [[VP_OP_LOAD]] to <vscale x 2 x i32>
 ; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[TMP17]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i32.p0(<vscale x 2 x i32> [[TMP16]], ptr align 4 [[TMP18]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META20:![0-9]+]], !noalias [[META17]]
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i32.p0(<vscale x 2 x i32> [[TMP16]], ptr align 4 [[TMP17]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META20:![0-9]+]], !noalias [[META17]]
 ; IF-EVL-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP12]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP19]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP11]]
@@ -259,17 +325,54 @@ define void @vp_trunc(ptr %a, ptr %b, i64 %N) {
 ; NO-VP-LABEL: define void @vp_trunc(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP9]], 2
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 20, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = shl i64 [[N]], 2
+; NO-VP-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]]
+; NO-VP-NEXT:    [[TMP4:%.*]] = shl i64 [[N]], 3
+; NO-VP-NEXT:    [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP4]]
+; NO-VP-NEXT:    [[BOUND0:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]]
+; NO-VP-NEXT:    [[BOUND1:%.*]] = icmp ult ptr [[B]], [[SCEVGEP]]
+; NO-VP-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; NO-VP-NEXT:    br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 2
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP6]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 2
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[GEP]], align 8, !alias.scope [[META16:![0-9]+]]
+; NO-VP-NEXT:    [[TMP10:%.*]] = trunc <vscale x 2 x i64> [[WIDE_LOAD]] to <vscale x 2 x i32>
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store <vscale x 2 x i32> [[TMP10]], ptr [[TMP11]], align 4, !alias.scope [[META19:![0-9]+]], !noalias [[META16]]
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP8]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
-; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[IV]]
-; NO-VP-NEXT:    [[TMP0:%.*]] = load i64, ptr [[GEP]], align 8
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[GEP1:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[IV1]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load i64, ptr [[GEP1]], align 8
 ; NO-VP-NEXT:    [[CONV:%.*]] = trunc i64 [[TMP0]] to i32
-; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]]
 ; NO-VP-NEXT:    store i32 [[CONV]], ptr [[GEP2]], align 4
-; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP22:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -321,12 +424,10 @@ define void @vp_fpext(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP14]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x float> @llvm.vp.load.nxv2f32.p0(ptr align 4 [[TMP15]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META24:![0-9]+]]
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x float> @llvm.vp.load.nxv2f32.p0(ptr align 4 [[TMP14]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META24:![0-9]+]]
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = fpext <vscale x 2 x float> [[VP_OP_LOAD]] to <vscale x 2 x double>
 ; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds double, ptr [[TMP17]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2f64.p0(<vscale x 2 x double> [[TMP16]], ptr align 8 [[TMP18]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META27:![0-9]+]], !noalias [[META24]]
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2f64.p0(<vscale x 2 x double> [[TMP16]], ptr align 8 [[TMP17]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META27:![0-9]+]], !noalias [[META24]]
 ; IF-EVL-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP12]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP19]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP11]]
@@ -353,17 +454,54 @@ define void @vp_fpext(ptr %a, ptr %b, i64 %N) {
 ; NO-VP-LABEL: define void @vp_fpext(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP9]], 2
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = shl i64 [[N]], 3
+; NO-VP-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]]
+; NO-VP-NEXT:    [[TMP4:%.*]] = shl i64 [[N]], 2
+; NO-VP-NEXT:    [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP4]]
+; NO-VP-NEXT:    [[BOUND0:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]]
+; NO-VP-NEXT:    [[BOUND1:%.*]] = icmp ult ptr [[B]], [[SCEVGEP]]
+; NO-VP-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; NO-VP-NEXT:    br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 2
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP6]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 2
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x float>, ptr [[GEP]], align 4, !alias.scope [[META23:![0-9]+]]
+; NO-VP-NEXT:    [[TMP10:%.*]] = fpext <vscale x 2 x float> [[WIDE_LOAD]] to <vscale x 2 x double>
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store <vscale x 2 x double> [[TMP10]], ptr [[TMP11]], align 8, !alias.scope [[META26:![0-9]+]], !noalias [[META23]]
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP8]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
-; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[GEP1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV1]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[GEP1]], align 4
 ; NO-VP-NEXT:    [[CONV:%.*]] = fpext float [[TMP0]] to double
-; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[IV1]]
 ; NO-VP-NEXT:    store double [[CONV]], ptr [[GEP2]], align 8
-; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP29:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -415,12 +553,10 @@ define void @vp_fptrunc(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds double, ptr [[TMP14]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x double> @llvm.vp.load.nxv2f64.p0(ptr align 8 [[TMP15]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META31:![0-9]+]]
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x double> @llvm.vp.load.nxv2f64.p0(ptr align 8 [[TMP14]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META31:![0-9]+]]
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = fptrunc <vscale x 2 x double> [[VP_OP_LOAD]] to <vscale x 2 x float>
 ; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2f32.p0(<vscale x 2 x float> [[TMP16]], ptr align 4 [[TMP18]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META34:![0-9]+]], !noalias [[META31]]
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2f32.p0(<vscale x 2 x float> [[TMP16]], ptr align 4 [[TMP17]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META34:![0-9]+]], !noalias [[META31]]
 ; IF-EVL-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP12]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP19]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP11]]
@@ -447,17 +583,54 @@ define void @vp_fptrunc(ptr %a, ptr %b, i64 %N) {
 ; NO-VP-LABEL: define void @vp_fptrunc(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP9]], 2
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = shl i64 [[N]], 2
+; NO-VP-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP3]]
+; NO-VP-NEXT:    [[TMP4:%.*]] = shl i64 [[N]], 3
+; NO-VP-NEXT:    [[SCEVGEP1:%.*]] = getelementptr i8, ptr [[B]], i64 [[TMP4]]
+; NO-VP-NEXT:    [[BOUND0:%.*]] = icmp ult ptr [[A]], [[SCEVGEP1]]
+; NO-VP-NEXT:    [[BOUND1:%.*]] = icmp ult ptr [[B]], [[SCEVGEP]]
+; NO-VP-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
+; NO-VP-NEXT:    br i1 [[FOUND_CONFLICT]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP5:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP6:%.*]] = mul nuw i64 [[TMP5]], 2
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP6]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 2
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x double>, ptr [[GEP]], align 8, !alias.scope [[META30:![0-9]+]]
+; NO-VP-NEXT:    [[TMP10:%.*]] = fptrunc <vscale x 2 x double> [[WIDE_LOAD]] to <vscale x 2 x float>
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store <vscale x 2 x float> [[TMP10]], ptr [[TMP11]], align 4, !alias.scope [[META33:![0-9]+]], !noalias [[META30]]
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP8]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP35:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
-; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[IV]]
-; NO-VP-NEXT:    [[TMP0:%.*]] = load double, ptr [[GEP]], align 8
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[GEP1:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[IV1]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load double, ptr [[GEP1]], align 8
 ; NO-VP-NEXT:    [[CONV:%.*]] = fptrunc double [[TMP0]] to float
-; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV1]]
 ; NO-VP-NEXT:    store float [[CONV]], ptr [[GEP2]], align 4
-; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP36:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -509,12 +682,10 @@ define void @vp_sitofp(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP16]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = sitofp <vscale x 4 x i32> [[VP_OP_LOAD]] to <vscale x 4 x float>
 ; IF-EVL-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, ptr [[TMP19]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4f32.p0(<vscale x 4 x float> [[TMP18]], ptr align 4 [[TMP20]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4f32.p0(<vscale x 4 x float> [[TMP18]], ptr align 4 [[TMP19]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
 ; IF-EVL-NEXT:    [[TMP21:%.*]] = zext i32 [[TMP14]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP21]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP13]]
@@ -541,17 +712,54 @@ define void @vp_sitofp(ptr %a, ptr %b, i64 %N) {
 ; NO-VP-LABEL: define void @vp_sitofp(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP11]], 4
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[A1]], [[B2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[TMP12:%.*]] = sitofp <vscale x 4 x i32> [[WIDE_LOAD]] to <vscale x 4 x float>
+; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store <vscale x 4 x float> [[TMP12]], ptr [[TMP13]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]]
+; NO-VP-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP37:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
-; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
-; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[GEP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV1]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP1]], align 4
 ; NO-VP-NEXT:    [[CONV:%.*]] = sitofp i32 [[TMP0]] to float
-; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV1]]
 ; NO-VP-NEXT:    store float [[CONV]], ptr [[GEP2]], align 4
-; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP38:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -603,12 +811,10 @@ define void @vp_uitofp(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP16]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = uitofp <vscale x 4 x i32> [[VP_OP_LOAD]] to <vscale x 4 x float>
 ; IF-EVL-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, ptr [[TMP19]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4f32.p0(<vscale x 4 x float> [[TMP18]], ptr align 4 [[TMP20]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4f32.p0(<vscale x 4 x float> [[TMP18]], ptr align 4 [[TMP19]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
 ; IF-EVL-NEXT:    [[TMP21:%.*]] = zext i32 [[TMP14]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP21]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP13]]
@@ -635,17 +841,54 @@ define void @vp_uitofp(ptr %a, ptr %b, i64 %N) {
 ; NO-VP-LABEL: define void @vp_uitofp(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP11]], 4
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[A1]], [[B2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[TMP12:%.*]] = uitofp <vscale x 4 x i32> [[WIDE_LOAD]] to <vscale x 4 x float>
+; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store <vscale x 4 x float> [[TMP12]], ptr [[TMP13]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]]
+; NO-VP-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP39:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
-; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
-; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[GEP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV1]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[GEP1]], align 4
 ; NO-VP-NEXT:    [[CONV:%.*]] = uitofp i32 [[TMP0]] to float
-; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV1]]
 ; NO-VP-NEXT:    store float [[CONV]], ptr [[GEP2]], align 4
-; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP40:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -697,12 +940,10 @@ define void @vp_fptosi(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP16]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = fptosi <vscale x 4 x float> [[VP_OP_LOAD]] to <vscale x 4 x i32>
 ; IF-EVL-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[TMP19]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP18]], ptr align 4 [[TMP20]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP18]], ptr align 4 [[TMP19]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
 ; IF-EVL-NEXT:    [[TMP21:%.*]] = zext i32 [[TMP14]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP21]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP13]]
@@ -729,17 +970,54 @@ define void @vp_fptosi(ptr %a, ptr %b, i64 %N) {
 ; NO-VP-LABEL: define void @vp_fptosi(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP11]], 4
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[A1]], [[B2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[TMP12:%.*]] = fptosi <vscale x 4 x float> [[WIDE_LOAD]] to <vscale x 4 x i32>
+; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP12]], ptr [[TMP13]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]]
+; NO-VP-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
-; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[GEP1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV1]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[GEP1]], align 4
 ; NO-VP-NEXT:    [[CONV:%.*]] = fptosi float [[TMP0]] to i32
-; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]]
 ; NO-VP-NEXT:    store i32 [[CONV]], ptr [[GEP2]], align 4
-; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP42:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -791,12 +1069,10 @@ define void @vp_fptoui(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP16]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = fptoui <vscale x 4 x float> [[VP_OP_LOAD]] to <vscale x 4 x i32>
 ; IF-EVL-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[TMP19]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP18]], ptr align 4 [[TMP20]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP18]], ptr align 4 [[TMP19]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
 ; IF-EVL-NEXT:    [[TMP21:%.*]] = zext i32 [[TMP14]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP21]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP13]]
@@ -823,17 +1099,54 @@ define void @vp_fptoui(ptr %a, ptr %b, i64 %N) {
 ; NO-VP-LABEL: define void @vp_fptoui(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP11]], 4
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP1]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 4
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[A1]], [[B2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 4
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[TMP12:%.*]] = fptoui <vscale x 4 x float> [[WIDE_LOAD]] to <vscale x 4 x i32>
+; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP12]], ptr [[TMP13]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]]
+; NO-VP-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP43:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
-; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV]]
-; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[GEP]], align 4
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[GEP1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[IV1]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[GEP1]], align 4
 ; NO-VP-NEXT:    [[CONV:%.*]] = fptoui float [[TMP0]] to i32
-; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV1]]
 ; NO-VP-NEXT:    store i32 [[CONV]], ptr [[GEP2]], align 4
-; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP44:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -885,12 +1198,10 @@ define void @vp_inttoptr(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i64, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP17]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP16]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP14]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = inttoptr <vscale x 2 x i64> [[VP_OP_LOAD]] to <vscale x 2 x ptr>
 ; IF-EVL-NEXT:    [[TMP19:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP20:%.*]] = getelementptr inbounds ptr, ptr [[TMP19]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2p0.p0(<vscale x 2 x ptr> [[TMP18]], ptr align 8 [[TMP20]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2p0.p0(<vscale x 2 x ptr> [[TMP18]], ptr align 8 [[TMP19]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP14]])
 ; IF-EVL-NEXT:    [[TMP21:%.*]] = zext i32 [[TMP14]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP21]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP13]]
@@ -917,17 +1228,54 @@ define void @vp_inttoptr(ptr %a, ptr %b, i64 %N) {
 ; NO-VP-LABEL: define void @vp_inttoptr(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[B2:%.*]] = ptrtoint ptr [[B]] to i64
+; NO-VP-NEXT:    [[A1:%.*]] = ptrtoint ptr [[A]] to i64
+; NO-VP-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP15:%.*]] = mul nuw i64 [[TMP11]], 2
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.umax.i64(i64 16, i64 [[TMP15]])
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP2]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; NO-VP:       [[VECTOR_MEMCHECK]]:
+; NO-VP-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP4:%.*]] = mul nuw i64 [[TMP3]], 2
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul i64 [[TMP4]], 8
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[A1]], [[B2]]
+; NO-VP-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP6]], [[TMP5]]
+; NO-VP-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 2
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP8]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul nuw i64 [[TMP9]], 2
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[IV]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[GEP]], align 8
+; NO-VP-NEXT:    [[TMP12:%.*]] = inttoptr <vscale x 2 x i64> [[WIDE_LOAD]] to <vscale x 2 x ptr>
+; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    store <vscale x 2 x ptr> [[TMP12]], ptr [[TMP13]], align 8
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP10]]
+; NO-VP-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
-; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[IV]]
-; NO-VP-NEXT:    [[TMP0:%.*]] = load i64, ptr [[GEP]], align 8
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[GEP1:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[IV1]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = load i64, ptr [[GEP1]], align 8
 ; NO-VP-NEXT:    [[TMP1:%.*]] = inttoptr i64 [[TMP0]] to ptr
-; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[IV]]
+; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[IV1]]
 ; NO-VP-NEXT:    store ptr [[TMP1]], ptr [[GEP2]], align 8
-; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP46:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -980,8 +1328,7 @@ define void @vp_ptrtoint(ptr %a, ptr %b, i64 %N) {
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[B]], <vscale x 2 x i64> [[VEC_IND]]
 ; IF-EVL-NEXT:    [[TMP15:%.*]] = ptrtoint <vscale x 2 x ptr> [[TMP14]] to <vscale x 2 x i64>
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i64, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP15]], ptr align 8 [[TMP17]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP15]], ptr align 8 [[TMP16]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP11]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP18]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP8]]
@@ -1008,16 +1355,50 @@ define void @vp_ptrtoint(ptr %a, ptr %b, i64 %N) {
 ; NO-VP-LABEL: define void @vp_ptrtoint(
 ; NO-VP-SAME: ptr [[A:%.*]], ptr [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[ENTRY:.*]]:
+; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP13]], 2
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], [[TMP1]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 2
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2
+; NO-VP-NEXT:    [[TMP6:%.*]] = call <vscale x 2 x i64> @llvm.stepvector.nxv2i64()
+; NO-VP-NEXT:    [[TMP7:%.*]] = mul <vscale x 2 x i64> [[TMP6]], splat (i64 1)
+; NO-VP-NEXT:    [[INDUCTION:%.*]] = add <vscale x 2 x i64> zeroinitializer, [[TMP7]]
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul i64 1, [[TMP5]]
+; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP8]], i64 0
+; NO-VP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 2 x i64> [ [[INDUCTION]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[B]], <vscale x 2 x i64> [[VEC_IND]]
+; NO-VP-NEXT:    [[TMP10:%.*]] = ptrtoint <vscale x 2 x ptr> [[TMP9]] to <vscale x 2 x i64>
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 2 x i64> [[TMP10]], ptr [[TMP11]], align 8
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; NO-VP-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP47:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
 ; NO-VP-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[IV]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = ptrtoint ptr [[GEP]] to i64
 ; NO-VP-NEXT:    [[GEP2:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
 ; NO-VP-NEXT:    store i64 [[TMP0]], ptr [[GEP2]], align 8
 ; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP48:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -1089,3 +1470,53 @@ exit:
 ; IF-EVL: [[LOOP48]] = distinct !{[[LOOP48]], [[META6]], [[META7]], [[META8]]}
 ; IF-EVL: [[LOOP49]] = distinct !{[[LOOP49]], [[META8]], [[META6]]}
 ;.
+; NO-VP: [[META0]] = !{[[META1:![0-9]+]]}
+; NO-VP: [[META1]] = distinct !{[[META1]], [[META2:![0-9]+]]}
+; NO-VP: [[META2]] = distinct !{[[META2]], !"LVerDomain"}
+; NO-VP: [[META3]] = !{[[META4:![0-9]+]]}
+; NO-VP: [[META4]] = distinct !{[[META4]], [[META2]]}
+; NO-VP: [[LOOP5]] = distinct !{[[LOOP5]], [[META6:![0-9]+]], [[META7:![0-9]+]]}
+; NO-VP: [[META6]] = !{!"llvm.loop.isvectorized", i32 1}
+; NO-VP: [[META7]] = !{!"llvm.loop.unroll.runtime.disable"}
+; NO-VP: [[LOOP8]] = distinct !{[[LOOP8]], [[META6]]}
+; NO-VP: [[META9]] = !{[[META10:![0-9]+]]}
+; NO-VP: [[META10]] = distinct !{[[META10]], [[META11:![0-9]+]]}
+; NO-VP: [[META11]] = distinct !{[[META11]], !"LVerDomain"}
+; NO-VP: [[META12]] = !{[[META13:![0-9]+]]}
+; NO-VP: [[META13]] = distinct !{[[META13]], [[META11]]}
+; NO-VP: [[LOOP14]] = distinct !{[[LOOP14]], [[META6]], [[META7]]}
+; NO-VP: [[LOOP15]] = distinct !{[[LOOP15]], [[META6]]}
+; NO-VP: [[META16]] = !{[[META17:![0-9]+]]}
+; NO-VP: [[META17]] = distinct !{[[META17]], [[META18:![0-9]+]]}
+; NO-VP: [[META18]] = distinct !{[[META18]], !"LVerDomain"}
+; NO-VP: [[META19]] = !{[[META20:![0-9]+]]}
+; NO-VP: [[META20]] = distinct !{[[META20]], [[META18]]}
+; NO-VP: [[LOOP21]] = distinct !{[[LOOP21]], [[META6]], [[META7]]}
+; NO-VP: [[LOOP22]] = distinct !{[[LOOP22]], [[META6]]}
+; NO-VP: [[META23]] = !{[[META24:![0-9]+]]}
+; NO-VP: [[META24]] = distinct !{[[META24]], [[META25:![0-9]+]]}
+; NO-VP: [[META25]] = distinct !{[[META25]], !"LVerDomain"}
+; NO-VP: [[META26]] = !{[[META27:![0-9]+]]}
+; NO-VP: [[META27]] = distinct !{[[META27]], [[META25]]}
+; NO-VP: [[LOOP28]] = distinct !{[[LOOP28]], [[META6]], [[META7]]}
+; NO-VP: [[LOOP29]] = distinct !{[[LOOP29]], [[META6]]}
+; NO-VP: [[META30]] = !{[[META31:![0-9]+]]}
+; NO-VP: [[META31]] = distinct !{[[META31]], [[META32:![0-9]+]]}
+; NO-VP: [[META32]] = distinct !{[[META32]], !"LVerDomain"}
+; NO-VP: [[META33]] = !{[[META34:![0-9]+]]}
+; NO-VP: [[META34]] = distinct !{[[META34]], [[META32]]}
+; NO-VP: [[LOOP35]] = distinct !{[[LOOP35]], [[META6]], [[META7]]}
+; NO-VP: [[LOOP36]] = distinct !{[[LOOP36]], [[META6]]}
+; NO-VP: [[LOOP37]] = distinct !{[[LOOP37]], [[META6]], [[META7]]}
+; NO-VP: [[LOOP38]] = distinct !{[[LOOP38]], [[META6]]}
+; NO-VP: [[LOOP39]] = distinct !{[[LOOP39]], [[META6]], [[META7]]}
+; NO-VP: [[LOOP40]] = distinct !{[[LOOP40]], [[META6]]}
+; NO-VP: [[LOOP41]] = distinct !{[[LOOP41]], [[META6]], [[META7]]}
+; NO-VP: [[LOOP42]] = distinct !{[[LOOP42]], [[META6]]}
+; NO-VP: [[LOOP43]] = distinct !{[[LOOP43]], [[META6]], [[META7]]}
+; NO-VP: [[LOOP44]] = distinct !{[[LOOP44]], [[META6]]}
+; NO-VP: [[LOOP45]] = distinct !{[[LOOP45]], [[META6]], [[META7]]}
+; NO-VP: [[LOOP46]] = distinct !{[[LOOP46]], [[META6]]}
+; NO-VP: [[LOOP47]] = distinct !{[[LOOP47]], [[META6]], [[META7]]}
+; NO-VP: [[LOOP48]] = distinct !{[[LOOP48]], [[META7]], [[META6]]}
+;.
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-cond-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-cond-reduction.ll
index 2926011..5f53cb6 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-cond-reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-cond-reduction.ll
@@ -1,24 +1,20 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefixes=IF-EVL-OUTLOOP
 
 ; RUN: opt -passes=loop-vectorize \
 ; RUN: -prefer-inloop-reductions \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefixes=IF-EVL-INLOOP
 
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefixes=NO-VP-OUTLOOP
 
 ; RUN: opt -passes=loop-vectorize \
 ; RUN: -prefer-inloop-reductions \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefixes=NO-VP-INLOOP
 
 define i32 @cond_add(ptr %a, i64 %n, i32 %start) {
@@ -44,8 +40,7 @@ define i32 @cond_add(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP10:%.*]] = sub i64 [[N]], [[EVL_BASED_IV1]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP11:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP10]], i32 4, i1 true)
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV1]]
-; IF-EVL-OUTLOOP-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP16]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP18:%.*]] = icmp sgt <vscale x 4 x i32> [[VP_OP_LOAD]], splat (i32 3)
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP19:%.*]] = select <vscale x 4 x i1> [[TMP18]], <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i32> zeroinitializer
 ; IF-EVL-OUTLOOP-NEXT:    [[VP_OP:%.*]] = add <vscale x 4 x i32> [[TMP19]], [[VEC_PHI]]
@@ -98,8 +93,7 @@ define i32 @cond_add(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-INLOOP-NEXT:    [[TMP11:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-INLOOP-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP11]], i32 4, i1 true)
 ; IF-EVL-INLOOP-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-INLOOP-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[TMP17]], i32 0
-; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP18]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-INLOOP-NEXT:    [[TMP19:%.*]] = icmp sgt <vscale x 4 x i32> [[VP_OP_LOAD]], splat (i32 3)
 ; IF-EVL-INLOOP-NEXT:    [[TMP20:%.*]] = select <vscale x 4 x i1> [[TMP19]], <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i32> zeroinitializer
 ; IF-EVL-INLOOP-NEXT:    [[TMP21:%.*]] = call i32 @llvm.vp.reduce.add.nxv4i32(i32 0, <vscale x 4 x i32> [[TMP20]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
@@ -150,8 +144,7 @@ define i32 @cond_add(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-OUTLOOP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-OUTLOOP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP11]], [[VECTOR_PH]] ], [ [[TMP17:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-OUTLOOP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
-; NO-VP-OUTLOOP-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 0
-; NO-VP-OUTLOOP-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP14]], align 4
+; NO-VP-OUTLOOP-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP13]], align 4
 ; NO-VP-OUTLOOP-NEXT:    [[TMP21:%.*]] = icmp sgt <vscale x 4 x i32> [[WIDE_MASKED_LOAD]], splat (i32 3)
 ; NO-VP-OUTLOOP-NEXT:    [[TMP16:%.*]] = select <vscale x 4 x i1> [[TMP21]], <vscale x 4 x i32> [[WIDE_MASKED_LOAD]], <vscale x 4 x i32> zeroinitializer
 ; NO-VP-OUTLOOP-NEXT:    [[TMP17]] = add <vscale x 4 x i32> [[TMP16]], [[VEC_PHI]]
@@ -200,8 +193,7 @@ define i32 @cond_add(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-INLOOP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-INLOOP-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ [[START]], [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-INLOOP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
-; NO-VP-INLOOP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; NO-VP-INLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; NO-VP-INLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; NO-VP-INLOOP-NEXT:    [[TMP9:%.*]] = icmp sgt <vscale x 4 x i32> [[WIDE_LOAD]], splat (i32 3)
 ; NO-VP-INLOOP-NEXT:    [[TMP10:%.*]] = select <vscale x 4 x i1> [[TMP9]], <vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> zeroinitializer
 ; NO-VP-INLOOP-NEXT:    [[TMP11:%.*]] = call i32 @llvm.vector.reduce.add.nxv4i32(<vscale x 4 x i32> [[TMP10]])
@@ -282,8 +274,7 @@ define i32 @cond_add_pred(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-OUTLOOP-NEXT:    [[VEC_IV:%.*]] = add <vscale x 4 x i64> [[BROADCAST_SPLAT]], [[TMP14]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP15:%.*]] = icmp ule <vscale x 4 x i64> [[VEC_IV]], [[BROADCAST_SPLAT2]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-OUTLOOP-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP16]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP18:%.*]] = icmp sle <vscale x 4 x i32> [[VP_OP_LOAD]], splat (i32 3)
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP19:%.*]] = add <vscale x 4 x i32> [[VEC_PHI]], [[VP_OP_LOAD]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP21:%.*]] = select <vscale x 4 x i1> [[TMP15]], <vscale x 4 x i1> [[TMP18]], <vscale x 4 x i1> zeroinitializer
@@ -341,8 +332,7 @@ define i32 @cond_add_pred(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-INLOOP-NEXT:    [[TMP11:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-INLOOP-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP11]], i32 4, i1 true)
 ; IF-EVL-INLOOP-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-INLOOP-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[TMP17]], i32 0
-; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP18]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-INLOOP-NEXT:    [[TMP19:%.*]] = icmp sgt <vscale x 4 x i32> [[VP_OP_LOAD]], splat (i32 3)
 ; IF-EVL-INLOOP-NEXT:    [[TMP21:%.*]] = call i32 @llvm.vp.reduce.add.nxv4i32(i32 0, <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i1> [[TMP19]], i32 [[TMP12]])
 ; IF-EVL-INLOOP-NEXT:    [[TMP22]] = add i32 [[TMP21]], [[VEC_PHI]]
@@ -396,8 +386,7 @@ define i32 @cond_add_pred(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-OUTLOOP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-OUTLOOP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP11]], [[VECTOR_PH]] ], [ [[PREDPHI:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-OUTLOOP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
-; NO-VP-OUTLOOP-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 0
-; NO-VP-OUTLOOP-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP21]], align 4
+; NO-VP-OUTLOOP-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP13]], align 4
 ; NO-VP-OUTLOOP-NEXT:    [[TMP18:%.*]] = icmp sgt <vscale x 4 x i32> [[WIDE_MASKED_LOAD]], splat (i32 3)
 ; NO-VP-OUTLOOP-NEXT:    [[TMP16:%.*]] = add <vscale x 4 x i32> [[VEC_PHI]], [[WIDE_MASKED_LOAD]]
 ; NO-VP-OUTLOOP-NEXT:    [[PREDPHI]] = select <vscale x 4 x i1> [[TMP18]], <vscale x 4 x i32> [[TMP16]], <vscale x 4 x i32> [[VEC_PHI]]
@@ -450,8 +439,7 @@ define i32 @cond_add_pred(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-INLOOP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-INLOOP-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ [[START]], [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-INLOOP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
-; NO-VP-INLOOP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; NO-VP-INLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; NO-VP-INLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; NO-VP-INLOOP-NEXT:    [[TMP9:%.*]] = icmp sgt <vscale x 4 x i32> [[WIDE_LOAD]], splat (i32 3)
 ; NO-VP-INLOOP-NEXT:    [[TMP10:%.*]] = select <vscale x 4 x i1> [[TMP9]], <vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> zeroinitializer
 ; NO-VP-INLOOP-NEXT:    [[TMP11:%.*]] = call i32 @llvm.vector.reduce.add.nxv4i32(<vscale x 4 x i32> [[TMP10]])
@@ -540,8 +528,7 @@ define i32 @step_cond_add(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-OUTLOOP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[TMP13]], i64 0
 ; IF-EVL-OUTLOOP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV1]]
-; IF-EVL-OUTLOOP-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i32 0
-; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP15]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP16:%.*]] = icmp sgt <vscale x 4 x i32> [[VP_OP_LOAD]], [[VEC_IND]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP17:%.*]] = select <vscale x 4 x i1> [[TMP16]], <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i32> zeroinitializer
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP18:%.*]] = add <vscale x 4 x i32> [[TMP17]], [[VEC_PHI]]
@@ -603,8 +590,7 @@ define i32 @step_cond_add(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-INLOOP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[TMP12]], i64 0
 ; IF-EVL-INLOOP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
 ; IF-EVL-INLOOP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV1]]
-; IF-EVL-INLOOP-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 0
-; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-INLOOP-NEXT:    [[TMP15:%.*]] = icmp sgt <vscale x 4 x i32> [[VP_OP_LOAD]], [[VEC_IND]]
 ; IF-EVL-INLOOP-NEXT:    [[TMP16:%.*]] = select <vscale x 4 x i1> [[TMP15]], <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i32> zeroinitializer
 ; IF-EVL-INLOOP-NEXT:    [[TMP17:%.*]] = call i32 @llvm.vp.reduce.add.nxv4i32(i32 0, <vscale x 4 x i32> [[TMP16]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP11]])
@@ -665,8 +651,7 @@ define i32 @step_cond_add(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-OUTLOOP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP11]], [[VECTOR_PH]] ], [ [[TMP23:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-OUTLOOP-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-OUTLOOP-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
-; NO-VP-OUTLOOP-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[TMP19]], i32 0
-; NO-VP-OUTLOOP-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP20]], align 4
+; NO-VP-OUTLOOP-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP19]], align 4
 ; NO-VP-OUTLOOP-NEXT:    [[TMP27:%.*]] = icmp sgt <vscale x 4 x i32> [[WIDE_MASKED_LOAD]], [[VEC_IND]]
 ; NO-VP-OUTLOOP-NEXT:    [[TMP22:%.*]] = select <vscale x 4 x i1> [[TMP27]], <vscale x 4 x i32> [[WIDE_MASKED_LOAD]], <vscale x 4 x i32> zeroinitializer
 ; NO-VP-OUTLOOP-NEXT:    [[TMP23]] = add <vscale x 4 x i32> [[TMP22]], [[VEC_PHI]]
@@ -725,8 +710,7 @@ define i32 @step_cond_add(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-INLOOP-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ [[START]], [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-INLOOP-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-INLOOP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
-; NO-VP-INLOOP-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 0
-; NO-VP-INLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP14]], align 4
+; NO-VP-INLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP13]], align 4
 ; NO-VP-INLOOP-NEXT:    [[TMP15:%.*]] = icmp sgt <vscale x 4 x i32> [[WIDE_LOAD]], [[VEC_IND]]
 ; NO-VP-INLOOP-NEXT:    [[TMP16:%.*]] = select <vscale x 4 x i1> [[TMP15]], <vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> zeroinitializer
 ; NO-VP-INLOOP-NEXT:    [[TMP17:%.*]] = call i32 @llvm.vector.reduce.add.nxv4i32(<vscale x 4 x i32> [[TMP16]])
@@ -820,8 +804,7 @@ define i32 @step_cond_add_pred(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-OUTLOOP-NEXT:    [[BROADCAST_SPLAT6:%.*]] = shufflevector <vscale x 4 x i64> [[BROADCAST_SPLATINSERT5]], <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP18:%.*]] = icmp ule <vscale x 4 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]]
 ; IF-EVL-OUTLOOP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
-; IF-EVL-OUTLOOP-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[ARRAYIDX]], i32 0
-; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP20]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
+; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[ARRAYIDX]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP21:%.*]] = icmp sle <vscale x 4 x i32> [[VP_OP_LOAD]], [[VEC_IND2]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP22:%.*]] = add <vscale x 4 x i32> [[VEC_PHI]], [[VP_OP_LOAD]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP23:%.*]] = select <vscale x 4 x i1> [[TMP18]], <vscale x 4 x i1> [[TMP21]], <vscale x 4 x i1> zeroinitializer
@@ -889,8 +872,7 @@ define i32 @step_cond_add_pred(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-INLOOP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[TMP12]], i64 0
 ; IF-EVL-INLOOP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
 ; IF-EVL-INLOOP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
-; IF-EVL-INLOOP-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[ARRAYIDX]], i32 0
-; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[ARRAYIDX]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-INLOOP-NEXT:    [[TMP15:%.*]] = icmp sgt <vscale x 4 x i32> [[VP_OP_LOAD]], [[VEC_IND]]
 ; IF-EVL-INLOOP-NEXT:    [[TMP16:%.*]] = call i32 @llvm.vp.reduce.add.nxv4i32(i32 0, <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i1> [[TMP15]], i32 [[TMP11]])
 ; IF-EVL-INLOOP-NEXT:    [[TMP17]] = add i32 [[TMP16]], [[VEC_PHI]]
@@ -954,8 +936,7 @@ define i32 @step_cond_add_pred(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-OUTLOOP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP11]], [[VECTOR_PH]] ], [ [[PREDPHI:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-OUTLOOP-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-OUTLOOP-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
-; NO-VP-OUTLOOP-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i32, ptr [[TMP19]], i32 0
-; NO-VP-OUTLOOP-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP27]], align 4
+; NO-VP-OUTLOOP-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP19]], align 4
 ; NO-VP-OUTLOOP-NEXT:    [[TMP28:%.*]] = icmp sgt <vscale x 4 x i32> [[WIDE_MASKED_LOAD]], [[VEC_IND]]
 ; NO-VP-OUTLOOP-NEXT:    [[TMP22:%.*]] = add <vscale x 4 x i32> [[VEC_PHI]], [[WIDE_MASKED_LOAD]]
 ; NO-VP-OUTLOOP-NEXT:    [[PREDPHI]] = select <vscale x 4 x i1> [[TMP28]], <vscale x 4 x i32> [[TMP22]], <vscale x 4 x i32> [[VEC_PHI]]
@@ -1018,8 +999,7 @@ define i32 @step_cond_add_pred(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-INLOOP-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ [[START]], [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-INLOOP-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-INLOOP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
-; NO-VP-INLOOP-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 0
-; NO-VP-INLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP14]], align 4
+; NO-VP-INLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP13]], align 4
 ; NO-VP-INLOOP-NEXT:    [[TMP15:%.*]] = icmp sgt <vscale x 4 x i32> [[WIDE_LOAD]], [[VEC_IND]]
 ; NO-VP-INLOOP-NEXT:    [[TMP16:%.*]] = select <vscale x 4 x i1> [[TMP15]], <vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> zeroinitializer
 ; NO-VP-INLOOP-NEXT:    [[TMP17:%.*]] = call i32 @llvm.vector.reduce.add.nxv4i32(<vscale x 4 x i32> [[TMP16]])
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-cost.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-cost.ll
index cd53ea0..238aed1 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-cost.ll
@@ -3,7 +3,7 @@
 ; RUN:   -mtriple riscv64-linux-gnu -mattr=+v,+f -S -disable-output -debug-only=loop-vectorize 2>&1 | FileCheck %s --check-prefix=DATA
 
 ; RUN: opt < %s -passes=loop-vectorize -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
-; RUN:   -mtriple riscv64-linux-gnu -force-tail-folding-style=data-with-evl -mattr=+v,+f -S \
+; RUN:   -mtriple riscv64-linux-gnu -mattr=+v,+f -S \
 ; RUN:   -disable-output -debug-only=loop-vectorize 2>&1 | FileCheck %s --check-prefix=EVL
 
 ; DATA: Cost of 2 for VF 2: EMIT{{.*}} = active lane mask
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-div.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-div.ll
index 3e83d8a..f27f49e 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-div.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-div.ll
@@ -1,12 +1,10 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s --check-prefix=IF-EVL
 
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s --check-prefix=NO-VP
 
 define void @test_sdiv(ptr noalias %a, ptr noalias %b, ptr noalias %c) {
@@ -30,16 +28,13 @@ define void @test_sdiv(ptr noalias %a, ptr noalias %b, ptr noalias %c) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 1024, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP5:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
 ; IF-EVL-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[TMP7]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP8]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP7]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = getelementptr i64, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP10:%.*]] = getelementptr i64, ptr [[TMP9]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD1:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP10]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD1:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP9]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = call <vscale x 2 x i64> @llvm.vp.merge.nxv2i64(<vscale x 2 x i1> splat (i1 true), <vscale x 2 x i64> [[VP_OP_LOAD1]], <vscale x 2 x i64> splat (i64 1), i32 [[TMP5]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = sdiv <vscale x 2 x i64> [[VP_OP_LOAD]], [[TMP11]]
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr i64, ptr [[C]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr i64, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VP_OP]], ptr align 8 [[TMP13]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VP_OP]], ptr align 8 [[TMP12]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = zext i32 [[TMP5]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP14]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP4]]
@@ -68,9 +63,38 @@ define void @test_sdiv(ptr noalias %a, ptr noalias %b, ptr noalias %c) {
 ; NO-VP-LABEL: define void @test_sdiv(
 ; NO-VP-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR0:[0-9]+]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 2
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP12]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP13]], 2
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP6:%.*]] = getelementptr i64, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP6]], align 8
+; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 8
+; NO-VP-NEXT:    [[TMP8:%.*]] = sdiv <vscale x 2 x i64> [[WIDE_LOAD]], [[WIDE_LOAD1]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr i64, ptr [[C]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 2 x i64> [[TMP8]], ptr [[TMP9]], align 8
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; NO-VP-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[A_GEP:%.*]] = getelementptr i64, ptr [[A]], i64 [[IV]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_GEP]], align 8
 ; NO-VP-NEXT:    [[B_GEP:%.*]] = getelementptr i64, ptr [[B]], i64 [[IV]]
@@ -80,7 +104,7 @@ define void @test_sdiv(ptr noalias %a, ptr noalias %b, ptr noalias %c) {
 ; NO-VP-NEXT:    store i64 [[TMP2]], ptr [[C_GEP]], align 8
 ; NO-VP-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; NO-VP-NEXT:    [[DONE:%.*]] = icmp eq i64 [[IV_NEXT]], 1024
-; NO-VP-NEXT:    br i1 [[DONE]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DONE]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -126,16 +150,13 @@ define void @test_udiv(ptr noalias %a, ptr noalias %b, ptr noalias %c) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 1024, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP5:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
 ; IF-EVL-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[TMP7]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP8]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP7]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = getelementptr i64, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP10:%.*]] = getelementptr i64, ptr [[TMP9]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD1:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP10]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD1:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP9]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = call <vscale x 2 x i64> @llvm.vp.merge.nxv2i64(<vscale x 2 x i1> splat (i1 true), <vscale x 2 x i64> [[VP_OP_LOAD1]], <vscale x 2 x i64> splat (i64 1), i32 [[TMP5]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = udiv <vscale x 2 x i64> [[VP_OP_LOAD]], [[TMP11]]
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr i64, ptr [[C]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr i64, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VP_OP]], ptr align 8 [[TMP13]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VP_OP]], ptr align 8 [[TMP12]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = zext i32 [[TMP5]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP14]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP4]]
@@ -164,9 +185,38 @@ define void @test_udiv(ptr noalias %a, ptr noalias %b, ptr noalias %c) {
 ; NO-VP-LABEL: define void @test_udiv(
 ; NO-VP-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 2
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP12]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP13]], 2
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP6:%.*]] = getelementptr i64, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP6]], align 8
+; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 8
+; NO-VP-NEXT:    [[TMP8:%.*]] = udiv <vscale x 2 x i64> [[WIDE_LOAD]], [[WIDE_LOAD1]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr i64, ptr [[C]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 2 x i64> [[TMP8]], ptr [[TMP9]], align 8
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; NO-VP-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[A_GEP:%.*]] = getelementptr i64, ptr [[A]], i64 [[IV]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_GEP]], align 8
 ; NO-VP-NEXT:    [[B_GEP:%.*]] = getelementptr i64, ptr [[B]], i64 [[IV]]
@@ -176,7 +226,7 @@ define void @test_udiv(ptr noalias %a, ptr noalias %b, ptr noalias %c) {
 ; NO-VP-NEXT:    store i64 [[TMP2]], ptr [[C_GEP]], align 8
 ; NO-VP-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; NO-VP-NEXT:    [[DONE:%.*]] = icmp eq i64 [[IV_NEXT]], 1024
-; NO-VP-NEXT:    br i1 [[DONE]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DONE]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -221,16 +271,13 @@ define void @test_srem(ptr noalias %a, ptr noalias %b, ptr noalias %c) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 1024, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP5:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
 ; IF-EVL-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[TMP7]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP8]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP7]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = getelementptr i64, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP10:%.*]] = getelementptr i64, ptr [[TMP9]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD1:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP10]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD1:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP9]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = call <vscale x 2 x i64> @llvm.vp.merge.nxv2i64(<vscale x 2 x i1> splat (i1 true), <vscale x 2 x i64> [[VP_OP_LOAD1]], <vscale x 2 x i64> splat (i64 1), i32 [[TMP5]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = srem <vscale x 2 x i64> [[VP_OP_LOAD]], [[TMP11]]
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr i64, ptr [[C]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr i64, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VP_OP]], ptr align 8 [[TMP13]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VP_OP]], ptr align 8 [[TMP12]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = zext i32 [[TMP5]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP14]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP4]]
@@ -259,9 +306,38 @@ define void @test_srem(ptr noalias %a, ptr noalias %b, ptr noalias %c) {
 ; NO-VP-LABEL: define void @test_srem(
 ; NO-VP-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 2
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP12]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP13]], 2
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP6:%.*]] = getelementptr i64, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP6]], align 8
+; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 8
+; NO-VP-NEXT:    [[TMP8:%.*]] = srem <vscale x 2 x i64> [[WIDE_LOAD]], [[WIDE_LOAD1]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr i64, ptr [[C]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 2 x i64> [[TMP8]], ptr [[TMP9]], align 8
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; NO-VP-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[A_GEP:%.*]] = getelementptr i64, ptr [[A]], i64 [[IV]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_GEP]], align 8
 ; NO-VP-NEXT:    [[B_GEP:%.*]] = getelementptr i64, ptr [[B]], i64 [[IV]]
@@ -271,7 +347,7 @@ define void @test_srem(ptr noalias %a, ptr noalias %b, ptr noalias %c) {
 ; NO-VP-NEXT:    store i64 [[TMP2]], ptr [[C_GEP]], align 8
 ; NO-VP-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; NO-VP-NEXT:    [[DONE:%.*]] = icmp eq i64 [[IV_NEXT]], 1024
-; NO-VP-NEXT:    br i1 [[DONE]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DONE]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP7:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -316,16 +392,13 @@ define void @test_urem(ptr noalias %a, ptr noalias %b, ptr noalias %c) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 1024, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP5:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
 ; IF-EVL-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[TMP7]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP8]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP7]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = getelementptr i64, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP10:%.*]] = getelementptr i64, ptr [[TMP9]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD1:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP10]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD1:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP9]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = call <vscale x 2 x i64> @llvm.vp.merge.nxv2i64(<vscale x 2 x i1> splat (i1 true), <vscale x 2 x i64> [[VP_OP_LOAD1]], <vscale x 2 x i64> splat (i64 1), i32 [[TMP5]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = urem <vscale x 2 x i64> [[VP_OP_LOAD]], [[TMP11]]
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr i64, ptr [[C]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr i64, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VP_OP]], ptr align 8 [[TMP13]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VP_OP]], ptr align 8 [[TMP12]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = zext i32 [[TMP5]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP14]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP4]]
@@ -354,9 +427,38 @@ define void @test_urem(ptr noalias %a, ptr noalias %b, ptr noalias %c) {
 ; NO-VP-LABEL: define void @test_urem(
 ; NO-VP-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]]) #[[ATTR0]] {
 ; NO-VP-NEXT:  [[LOOP_PREHEADER:.*]]:
+; NO-VP-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 2
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP12]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; NO-VP:       [[VECTOR_PH]]:
+; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP13]], 2
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2
+; NO-VP-NEXT:    br label %[[VECTOR_BODY:.*]]
+; NO-VP:       [[VECTOR_BODY]]:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP6:%.*]] = getelementptr i64, ptr [[A]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[TMP6]], align 8
+; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[B]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 2 x i64>, ptr [[TMP7]], align 8
+; NO-VP-NEXT:    [[TMP8:%.*]] = urem <vscale x 2 x i64> [[WIDE_LOAD]], [[WIDE_LOAD1]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr i64, ptr [[C]], i64 [[INDEX]]
+; NO-VP-NEXT:    store <vscale x 2 x i64> [[TMP8]], ptr [[TMP9]], align 8
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; NO-VP-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; NO-VP:       [[MIDDLE_BLOCK]]:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; NO-VP:       [[SCALAR_PH]]:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ]
 ; NO-VP-NEXT:    br label %[[LOOP:.*]]
 ; NO-VP:       [[LOOP]]:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
 ; NO-VP-NEXT:    [[A_GEP:%.*]] = getelementptr i64, ptr [[A]], i64 [[IV]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i64, ptr [[A_GEP]], align 8
 ; NO-VP-NEXT:    [[B_GEP:%.*]] = getelementptr i64, ptr [[B]], i64 [[IV]]
@@ -366,7 +468,7 @@ define void @test_urem(ptr noalias %a, ptr noalias %b, ptr noalias %c) {
 ; NO-VP-NEXT:    store i64 [[TMP2]], ptr [[C_GEP]], align 8
 ; NO-VP-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; NO-VP-NEXT:    [[DONE:%.*]] = icmp eq i64 [[IV_NEXT]], 1024
-; NO-VP-NEXT:    br i1 [[DONE]], label %[[EXIT:.*]], label %[[LOOP]]
+; NO-VP-NEXT:    br i1 [[DONE]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP9:![0-9]+]]
 ; NO-VP:       [[EXIT]]:
 ; NO-VP-NEXT:    ret void
 ;
@@ -402,3 +504,14 @@ exit:
 ; IF-EVL: [[LOOP9]] = distinct !{[[LOOP9]], [[META1]], [[META2]], [[META3]]}
 ; IF-EVL: [[LOOP10]] = distinct !{[[LOOP10]], [[META3]], [[META1]]}
 ;.
+; NO-VP: [[LOOP0]] = distinct !{[[LOOP0]], [[META1:![0-9]+]], [[META2:![0-9]+]]}
+; NO-VP: [[META1]] = !{!"llvm.loop.isvectorized", i32 1}
+; NO-VP: [[META2]] = !{!"llvm.loop.unroll.runtime.disable"}
+; NO-VP: [[LOOP3]] = distinct !{[[LOOP3]], [[META2]], [[META1]]}
+; NO-VP: [[LOOP4]] = distinct !{[[LOOP4]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP5]] = distinct !{[[LOOP5]], [[META2]], [[META1]]}
+; NO-VP: [[LOOP6]] = distinct !{[[LOOP6]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP7]] = distinct !{[[LOOP7]], [[META2]], [[META1]]}
+; NO-VP: [[LOOP8]] = distinct !{[[LOOP8]], [[META1]], [[META2]]}
+; NO-VP: [[LOOP9]] = distinct !{[[LOOP9]], [[META2]], [[META1]]}
+;.
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-fixed-order-recurrence.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-fixed-order-recurrence.ll
index e31d7ff..6a6bca2d 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-fixed-order-recurrence.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-fixed-order-recurrence.ll
@@ -1,14 +1,12 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
 ; RUN: opt -passes=loop-vectorize \
 ; RUN: -prefer-inloop-reductions \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v,+f -S < %s| FileCheck %s --check-prefix=IF-EVL
 
 ; RUN: opt -passes=loop-vectorize \
 ; RUN: -prefer-inloop-reductions \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v,+f -S < %s| FileCheck %s --check-prefix=NO-VP
 
 define void @first_order_recurrence(ptr noalias %A, ptr noalias %B, i64 %TC) {
@@ -39,13 +37,11 @@ define void @first_order_recurrence(ptr noalias %A, ptr noalias %B, i64 %TC) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[TC]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP12]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP14]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP15]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = call <vscale x 4 x i32> @llvm.experimental.vp.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR]], <vscale x 4 x i32> [[VP_OP_LOAD]], i32 -1, <vscale x 4 x i1> splat (i1 true), i32 [[PREV_EVL]], i32 [[TMP12]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = add nsw <vscale x 4 x i32> [[TMP16]], [[VP_OP_LOAD]]
 ; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw i32, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP17]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[VP_OP]], ptr align 4 [[TMP18]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[VP_OP]], ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-NEXT:    [[TMP19:%.*]] = zext i32 [[TMP12]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP19]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP8]]
@@ -94,13 +90,11 @@ define void @first_order_recurrence(ptr noalias %A, ptr noalias %B, i64 %TC) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i32> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP10]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i32>, ptr [[TMP11]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i32>, ptr [[TMP10]], align 4
 ; NO-VP-NEXT:    [[TMP12:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR]], <vscale x 4 x i32> [[WIDE_LOAD]], i32 -1)
 ; NO-VP-NEXT:    [[TMP13:%.*]] = add nsw <vscale x 4 x i32> [[TMP12]], [[WIDE_LOAD]]
 ; NO-VP-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw i32, ptr [[B]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP14]], i32 0
-; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP13]], ptr [[TMP15]], align 4
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP13]], ptr [[TMP14]], align 4
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; NO-VP-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; NO-VP-NEXT:    br i1 [[TMP16]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -181,14 +175,12 @@ define void @second_order_recurrence(ptr noalias %A, ptr noalias %B, i64 %TC) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[TC]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP15]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP17]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP18]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP15]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP15]])
 ; IF-EVL-NEXT:    [[TMP19]] = call <vscale x 4 x i32> @llvm.experimental.vp.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR]], <vscale x 4 x i32> [[VP_OP_LOAD]], i32 -1, <vscale x 4 x i1> splat (i1 true), i32 [[PREV_EVL]], i32 [[TMP15]])
 ; IF-EVL-NEXT:    [[TMP20:%.*]] = call <vscale x 4 x i32> @llvm.experimental.vp.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR2]], <vscale x 4 x i32> [[TMP19]], i32 -1, <vscale x 4 x i1> splat (i1 true), i32 [[PREV_EVL]], i32 [[TMP15]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = add nsw <vscale x 4 x i32> [[TMP19]], [[TMP20]]
 ; IF-EVL-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw i32, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP22:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP21]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[VP_OP]], ptr align 4 [[TMP22]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP15]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[VP_OP]], ptr align 4 [[TMP21]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP15]])
 ; IF-EVL-NEXT:    [[TMP23:%.*]] = zext i32 [[TMP15]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP23]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP8]]
@@ -244,14 +236,12 @@ define void @second_order_recurrence(ptr noalias %A, ptr noalias %B, i64 %TC) {
 ; NO-VP-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i32> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VECTOR_RECUR2:%.*]] = phi <vscale x 4 x i32> [ [[VECTOR_RECUR_INIT1]], %[[VECTOR_PH]] ], [ [[TMP15:%.*]], %[[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP13]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i32>, ptr [[TMP14]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i32>, ptr [[TMP13]], align 4
 ; NO-VP-NEXT:    [[TMP15]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR]], <vscale x 4 x i32> [[WIDE_LOAD]], i32 -1)
 ; NO-VP-NEXT:    [[TMP16:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR2]], <vscale x 4 x i32> [[TMP15]], i32 -1)
 ; NO-VP-NEXT:    [[TMP17:%.*]] = add nsw <vscale x 4 x i32> [[TMP15]], [[TMP16]]
 ; NO-VP-NEXT:    [[TMP18:%.*]] = getelementptr inbounds nuw i32, ptr [[B]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP19:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP18]], i32 0
-; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP17]], ptr [[TMP19]], align 4
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP17]], ptr [[TMP18]], align 4
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; NO-VP-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; NO-VP-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -344,16 +334,14 @@ define void @third_order_recurrence(ptr noalias %A, ptr noalias %B, i64 %TC) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[TC]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP18]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP20:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP21:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP20]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP21]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP18]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP20]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP18]])
 ; IF-EVL-NEXT:    [[TMP22]] = call <vscale x 4 x i32> @llvm.experimental.vp.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR]], <vscale x 4 x i32> [[VP_OP_LOAD]], i32 -1, <vscale x 4 x i1> splat (i1 true), i32 [[PREV_EVL]], i32 [[TMP18]])
 ; IF-EVL-NEXT:    [[TMP23]] = call <vscale x 4 x i32> @llvm.experimental.vp.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR2]], <vscale x 4 x i32> [[TMP22]], i32 -1, <vscale x 4 x i1> splat (i1 true), i32 [[PREV_EVL]], i32 [[TMP18]])
 ; IF-EVL-NEXT:    [[TMP24:%.*]] = call <vscale x 4 x i32> @llvm.experimental.vp.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR4]], <vscale x 4 x i32> [[TMP23]], i32 -1, <vscale x 4 x i1> splat (i1 true), i32 [[PREV_EVL]], i32 [[TMP18]])
 ; IF-EVL-NEXT:    [[TMP40:%.*]] = add nsw <vscale x 4 x i32> [[TMP23]], [[TMP24]]
 ; IF-EVL-NEXT:    [[VP_OP5:%.*]] = add <vscale x 4 x i32> [[TMP40]], [[TMP22]]
 ; IF-EVL-NEXT:    [[TMP25:%.*]] = getelementptr inbounds nuw i32, ptr [[B]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP26:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP25]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[VP_OP5]], ptr align 4 [[TMP26]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP18]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[VP_OP5]], ptr align 4 [[TMP25]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP18]])
 ; IF-EVL-NEXT:    [[TMP27:%.*]] = zext i32 [[TMP18]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP27]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP8]]
@@ -417,16 +405,14 @@ define void @third_order_recurrence(ptr noalias %A, ptr noalias %B, i64 %TC) {
 ; NO-VP-NEXT:    [[VECTOR_RECUR2:%.*]] = phi <vscale x 4 x i32> [ [[VECTOR_RECUR_INIT1]], %[[VECTOR_PH]] ], [ [[TMP18:%.*]], %[[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VECTOR_RECUR4:%.*]] = phi <vscale x 4 x i32> [ [[VECTOR_RECUR_INIT3]], %[[VECTOR_PH]] ], [ [[TMP19:%.*]], %[[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP16:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP16]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i32>, ptr [[TMP17]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i32>, ptr [[TMP16]], align 4
 ; NO-VP-NEXT:    [[TMP18]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR]], <vscale x 4 x i32> [[WIDE_LOAD]], i32 -1)
 ; NO-VP-NEXT:    [[TMP19]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR2]], <vscale x 4 x i32> [[TMP18]], i32 -1)
 ; NO-VP-NEXT:    [[TMP20:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR4]], <vscale x 4 x i32> [[TMP19]], i32 -1)
 ; NO-VP-NEXT:    [[TMP21:%.*]] = add nsw <vscale x 4 x i32> [[TMP19]], [[TMP20]]
 ; NO-VP-NEXT:    [[TMP22:%.*]] = add <vscale x 4 x i32> [[TMP21]], [[TMP18]]
 ; NO-VP-NEXT:    [[TMP23:%.*]] = getelementptr inbounds nuw i32, ptr [[B]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP24:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP23]], i32 0
-; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP22]], ptr [[TMP24]], align 4
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP22]], ptr [[TMP23]], align 4
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; NO-VP-NEXT:    [[TMP25:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; NO-VP-NEXT:    br i1 [[TMP25]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
@@ -494,20 +480,62 @@ define i32 @FOR_reduction(ptr noalias %A, ptr noalias %B, i64 %TC) {
 ; IF-EVL-LABEL: define i32 @FOR_reduction(
 ; IF-EVL-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], i64 [[TC:%.*]]) #[[ATTR0]] {
 ; IF-EVL-NEXT:  [[ENTRY:.*]]:
+; IF-EVL-NEXT:    [[TMP9:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP9]], 4
+; IF-EVL-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TC]], [[TMP1]]
+; IF-EVL-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; IF-EVL:       [[VECTOR_PH]]:
+; IF-EVL-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4
+; IF-EVL-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TC]], [[TMP3]]
+; IF-EVL-NEXT:    [[N_VEC:%.*]] = sub i64 [[TC]], [[N_MOD_VF]]
+; IF-EVL-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; IF-EVL-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4
+; IF-EVL-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vscale.i32()
+; IF-EVL-NEXT:    [[TMP7:%.*]] = mul nuw i32 [[TMP6]], 4
+; IF-EVL-NEXT:    [[TMP8:%.*]] = sub i32 [[TMP7]], 1
+; IF-EVL-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 33, i32 [[TMP8]]
+; IF-EVL-NEXT:    br label %[[VECTOR_BODY:.*]]
+; IF-EVL:       [[VECTOR_BODY]]:
+; IF-EVL-NEXT:    [[INDVARS:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i32> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[INDVARS]]
+; IF-EVL-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i32>, ptr [[ARRAYIDX]], align 4
+; IF-EVL-NEXT:    [[TMP10:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR]], <vscale x 4 x i32> [[WIDE_LOAD]], i32 -1)
+; IF-EVL-NEXT:    [[TMP11:%.*]] = add nsw <vscale x 4 x i32> [[TMP10]], [[WIDE_LOAD]]
+; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw i32, ptr [[B]], i64 [[INDVARS]]
+; IF-EVL-NEXT:    store <vscale x 4 x i32> [[TMP11]], ptr [[TMP12]], align 4
+; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDVARS]], [[TMP5]]
+; IF-EVL-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IF-EVL-NEXT:    br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; IF-EVL:       [[MIDDLE_BLOCK]]:
+; IF-EVL-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vscale.i32()
+; IF-EVL-NEXT:    [[TMP15:%.*]] = mul nuw i32 [[TMP14]], 4
+; IF-EVL-NEXT:    [[TMP16:%.*]] = sub i32 [[TMP15]], 1
+; IF-EVL-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <vscale x 4 x i32> [[WIDE_LOAD]], i32 [[TMP16]]
+; IF-EVL-NEXT:    [[TMP17:%.*]] = call i32 @llvm.vscale.i32()
+; IF-EVL-NEXT:    [[TMP18:%.*]] = mul nuw i32 [[TMP17]], 4
+; IF-EVL-NEXT:    [[TMP19:%.*]] = sub i32 [[TMP18]], 2
+; IF-EVL-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <vscale x 4 x i32> [[WIDE_LOAD]], i32 [[TMP19]]
+; IF-EVL-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TC]], [[N_VEC]]
+; IF-EVL-NEXT:    br i1 [[CMP_N]], label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; IF-EVL:       [[SCALAR_PH]]:
+; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; IF-EVL-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
 ; IF-EVL-NEXT:    br label %[[FOR_BODY:.*]]
 ; IF-EVL:       [[FOR_BODY]]:
-; IF-EVL-NEXT:    [[INDVARS:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[INDVARS_NEXT:%.*]], %[[FOR_BODY]] ]
-; IF-EVL-NEXT:    [[FOR1:%.*]] = phi i32 [ 33, %[[ENTRY]] ], [ [[TMP0:%.*]], %[[FOR_BODY]] ]
-; IF-EVL-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[INDVARS]]
-; IF-EVL-NEXT:    [[TMP0]] = load i32, ptr [[ARRAYIDX]], align 4
+; IF-EVL-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_NEXT:%.*]], %[[FOR_BODY]] ]
+; IF-EVL-NEXT:    [[FOR1:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], %[[SCALAR_PH]] ], [ [[TMP0:%.*]], %[[FOR_BODY]] ]
+; IF-EVL-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[IV]]
+; IF-EVL-NEXT:    [[TMP0]] = load i32, ptr [[ARRAYIDX1]], align 4
 ; IF-EVL-NEXT:    [[ADD:%.*]] = add nsw i32 [[FOR1]], [[TMP0]]
-; IF-EVL-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds nuw i32, ptr [[B]], i64 [[INDVARS]]
+; IF-EVL-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds nuw i32, ptr [[B]], i64 [[IV]]
 ; IF-EVL-NEXT:    store i32 [[ADD]], ptr [[ARRAYIDX2]], align 4
-; IF-EVL-NEXT:    [[INDVARS_NEXT]] = add nuw nsw i64 [[INDVARS]], 1
+; IF-EVL-NEXT:    [[INDVARS_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_NEXT]], [[TC]]
-; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; IF-EVL:       [[FOR_END]]:
-; IF-EVL-NEXT:    [[FOR1_LCSSA:%.*]] = phi i32 [ [[FOR1]], %[[FOR_BODY]] ]
+; IF-EVL-NEXT:    [[FOR1_LCSSA:%.*]] = phi i32 [ [[FOR1]], %[[FOR_BODY]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ]
 ; IF-EVL-NEXT:    ret i32 [[FOR1_LCSSA]]
 ;
 ; NO-VP-LABEL: define i32 @FOR_reduction(
@@ -533,13 +561,11 @@ define i32 @FOR_reduction(ptr noalias %A, ptr noalias %B, i64 %TC) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i32> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw i32, ptr [[A]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP10]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i32>, ptr [[TMP11]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i32>, ptr [[TMP10]], align 4
 ; NO-VP-NEXT:    [[TMP12:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR]], <vscale x 4 x i32> [[WIDE_LOAD]], i32 -1)
 ; NO-VP-NEXT:    [[TMP13:%.*]] = add nsw <vscale x 4 x i32> [[TMP12]], [[WIDE_LOAD]]
 ; NO-VP-NEXT:    [[TMP14:%.*]] = getelementptr inbounds nuw i32, ptr [[B]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP14]], i32 0
-; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP13]], ptr [[TMP15]], align 4
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP13]], ptr [[TMP14]], align 4
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; NO-VP-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; NO-VP-NEXT:    br i1 [[TMP16]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
@@ -630,8 +656,7 @@ define void @first_order_recurrence_indvar(ptr noalias %A, i64 %TC) {
 ; IF-EVL-NEXT:    [[TMP20]] = add <vscale x 2 x i64> [[VEC_IND]], splat (i64 42)
 ; IF-EVL-NEXT:    [[TMP15:%.*]] = call <vscale x 2 x i64> @llvm.experimental.vp.splice.nxv2i64(<vscale x 2 x i64> [[VECTOR_RECUR]], <vscale x 2 x i64> [[TMP20]], i32 -1, <vscale x 2 x i1> splat (i1 true), i32 [[PREV_EVL]], i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw i64, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds nuw i64, ptr [[TMP9]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP15]], ptr align 8 [[TMP17]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[TMP15]], ptr align 8 [[TMP9]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[TMP21:%.*]] = zext i32 [[TMP11]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP21]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP4]]
@@ -688,8 +713,7 @@ define void @first_order_recurrence_indvar(ptr noalias %A, i64 %TC) {
 ; NO-VP-NEXT:    [[TMP12]] = add <vscale x 2 x i64> [[VEC_IND]], splat (i64 42)
 ; NO-VP-NEXT:    [[TMP13:%.*]] = call <vscale x 2 x i64> @llvm.vector.splice.nxv2i64(<vscale x 2 x i64> [[VECTOR_RECUR]], <vscale x 2 x i64> [[TMP12]], i32 -1)
 ; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw i64, ptr [[A]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP15:%.*]] = getelementptr inbounds nuw i64, ptr [[TMP11]], i32 0
-; NO-VP-NEXT:    store <vscale x 2 x i64> [[TMP13]], ptr [[TMP15]], align 8
+; NO-VP-NEXT:    store <vscale x 2 x i64> [[TMP13]], ptr [[TMP11]], align 8
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP9]]
 ; NO-VP-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]]
 ; NO-VP-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -749,8 +773,8 @@ for.end:
 ; IF-EVL: [[LOOP6]] = distinct !{[[LOOP6]], [[META3]], [[META1]]}
 ; IF-EVL: [[LOOP7]] = distinct !{[[LOOP7]], [[META1]], [[META2]], [[META3]]}
 ; IF-EVL: [[LOOP8]] = distinct !{[[LOOP8]], [[META3]], [[META1]]}
-; IF-EVL: [[LOOP9]] = distinct !{[[LOOP9]], [[META10:![0-9]+]]}
-; IF-EVL: [[META10]] = !{!"llvm.loop.vectorize.enable", i1 true}
+; IF-EVL: [[LOOP9]] = distinct !{[[LOOP9]], [[META1]], [[META3]]}
+; IF-EVL: [[LOOP10]] = distinct !{[[LOOP10]], [[META3]], [[META1]]}
 ; IF-EVL: [[LOOP11]] = distinct !{[[LOOP11]], [[META1]], [[META2]], [[META3]]}
 ; IF-EVL: [[LOOP12]] = distinct !{[[LOOP12]], [[META3]], [[META1]]}
 ;.
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-gather-scatter.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-gather-scatter.ll
index 2b7a9fb..892293b 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-gather-scatter.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-gather-scatter.ll
@@ -1,12 +1,10 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefix=IF-EVL
 
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefix=NO-VP
 
 define void @gather_scatter(ptr noalias %in, ptr noalias %out, ptr noalias %index, i64 %n) {
@@ -69,18 +67,54 @@ define void @gather_scatter(ptr noalias %in, ptr noalias %out, ptr noalias %inde
 ;
 ; NO-VP-LABEL: @gather_scatter(
 ; NO-VP-NEXT:  entry:
+; NO-VP-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP14:%.*]] = mul nuw i64 [[TMP13]], 2
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP14]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
+; NO-VP:       vector.ph:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 2
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2
+; NO-VP-NEXT:    [[TMP6:%.*]] = call <vscale x 2 x i64> @llvm.stepvector.nxv2i64()
+; NO-VP-NEXT:    [[TMP7:%.*]] = mul <vscale x 2 x i64> [[TMP6]], splat (i64 1)
+; NO-VP-NEXT:    [[INDUCTION:%.*]] = add <vscale x 2 x i64> zeroinitializer, [[TMP7]]
+; NO-VP-NEXT:    [[TMP8:%.*]] = mul i64 1, [[TMP5]]
+; NO-VP-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP8]], i64 0
+; NO-VP-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
 ; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
+; NO-VP:       vector.body:
+; NO-VP-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
+; NO-VP-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 2 x i64> [ [[INDUCTION]], [[ENTRY]] ], [ [[VEC_IND_NEXT:%.*]], [[FOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[INDEX:%.*]], <vscale x 2 x i64> [[VEC_IND]]
+; NO-VP-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 2 x i64> @llvm.masked.gather.nxv2i64.nxv2p0(<vscale x 2 x ptr> [[TMP9]], i32 8, <vscale x 2 x i1> splat (i1 true), <vscale x 2 x i64> poison)
+; NO-VP-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[IN:%.*]], <vscale x 2 x i64> [[WIDE_MASKED_GATHER]]
+; NO-VP-NEXT:    [[WIDE_MASKED_GATHER2:%.*]] = call <vscale x 2 x float> @llvm.masked.gather.nxv2f32.nxv2p0(<vscale x 2 x ptr> [[TMP10]], i32 4, <vscale x 2 x i1> splat (i1 true), <vscale x 2 x float> poison)
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[OUT:%.*]], <vscale x 2 x i64> [[WIDE_MASKED_GATHER]]
+; NO-VP-NEXT:    call void @llvm.masked.scatter.nxv2f32.nxv2p0(<vscale x 2 x float> [[WIDE_MASKED_GATHER2]], <vscale x 2 x ptr> [[TMP11]], i32 4, <vscale x 2 x i1> splat (i1 true))
+; NO-VP-NEXT:    [[INDVARS_IV_NEXT]] = add nuw i64 [[INDVARS_IV]], [[TMP5]]
+; NO-VP-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; NO-VP:       middle.block:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; NO-VP:       scalar.ph:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
+; NO-VP-NEXT:    br label [[FOR_BODY1:%.*]]
 ; NO-VP:       for.body:
-; NO-VP-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
-; NO-VP-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, ptr [[INDEX:%.*]], i64 [[INDVARS_IV]]
+; NO-VP-NEXT:    [[INDVARS_IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT1:%.*]], [[FOR_BODY1]] ]
+; NO-VP-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, ptr [[INDEX]], i64 [[INDVARS_IV1]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i64, ptr [[ARRAYIDX3]], align 8
-; NO-VP-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds float, ptr [[IN:%.*]], i64 [[TMP0]]
+; NO-VP-NEXT:    [[ARRAYIDX5:%.*]] = getelementptr inbounds float, ptr [[IN]], i64 [[TMP0]]
 ; NO-VP-NEXT:    [[TMP1:%.*]] = load float, ptr [[ARRAYIDX5]], align 4
-; NO-VP-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, ptr [[OUT:%.*]], i64 [[TMP0]]
+; NO-VP-NEXT:    [[ARRAYIDX7:%.*]] = getelementptr inbounds float, ptr [[OUT]], i64 [[TMP0]]
 ; NO-VP-NEXT:    store float [[TMP1]], ptr [[ARRAYIDX7]], align 4
-; NO-VP-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
-; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[N:%.*]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END:%.*]], label [[FOR_BODY]]
+; NO-VP-NEXT:    [[INDVARS_IV_NEXT1]] = add nuw nsw i64 [[INDVARS_IV1]], 1
+; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT1]], [[N]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP3:![0-9]+]]
 ; NO-VP:       for.end:
 ; NO-VP-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-inloop-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-inloop-reduction.ll
index 87ac697..b7d7fc1 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-inloop-reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-inloop-reduction.ll
@@ -1,14 +1,12 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -passes=loop-vectorize \
 ; RUN: -prefer-inloop-reductions \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v,+f -S < %s| FileCheck %s --check-prefix=IF-EVL
 
 ; RUN: opt -passes=loop-vectorize \
 ; RUN: -prefer-inloop-reductions \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v,+f -S < %s| FileCheck %s --check-prefix=NO-VP
 
 define i32 @add(ptr %a, i64 %n, i32 %start) {
@@ -32,8 +30,7 @@ define i32 @add(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP9]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vp.reduce.add.nxv4i32(i32 0, <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP15]] = add i32 [[TMP14]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP10]] to i64
@@ -78,8 +75,7 @@ define i32 @add(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ [[START:%.*]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.add.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]])
 ; NO-VP-NEXT:    [[TMP10]] = add i32 [[TMP9]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -143,8 +139,7 @@ define i32 @mul(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-NEXT:    [[VEC_IV:%.*]] = add <4 x i64> [[BROADCAST_SPLAT]], <i64 0, i64 1, i64 2, i64 3>
 ; IF-EVL-NEXT:    [[TMP1:%.*]] = icmp ule <4 x i64> [[VEC_IV]], [[BROADCAST_SPLAT2]]
 ; IF-EVL-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IV]]
-; IF-EVL-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
-; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP3]], i32 4, <4 x i1> [[TMP1]], <4 x i32> poison)
+; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP2]], i32 4, <4 x i1> [[TMP1]], <4 x i32> poison)
 ; IF-EVL-NEXT:    [[TMP4:%.*]] = select <4 x i1> [[TMP1]], <4 x i32> [[WIDE_MASKED_LOAD]], <4 x i32> splat (i32 1)
 ; IF-EVL-NEXT:    [[TMP5:%.*]] = call i32 @llvm.vector.reduce.mul.v4i32(<4 x i32> [[TMP4]])
 ; IF-EVL-NEXT:    [[MUL]] = mul i32 [[TMP5]], [[RDX]]
@@ -183,9 +178,8 @@ define i32 @mul(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ [[START:%.*]], [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI1:%.*]] = phi i32 [ 1, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
 ; NO-VP-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 4
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
 ; NO-VP-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4
 ; NO-VP-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.mul.v4i32(<4 x i32> [[WIDE_LOAD]])
 ; NO-VP-NEXT:    [[TMP7]] = mul i32 [[TMP6]], [[VEC_PHI]]
@@ -253,8 +247,7 @@ define i32 @or(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP9]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vp.reduce.or.nxv4i32(i32 0, <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP15]] = or i32 [[TMP14]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP10]] to i64
@@ -299,8 +292,7 @@ define i32 @or(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ [[START:%.*]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.or.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]])
 ; NO-VP-NEXT:    [[TMP10]] = or i32 [[TMP9]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -364,8 +356,7 @@ define i32 @and(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP9]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vp.reduce.and.nxv4i32(i32 -1, <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP15]] = and i32 [[TMP14]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP10]] to i64
@@ -410,8 +401,7 @@ define i32 @and(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ [[START:%.*]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.and.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]])
 ; NO-VP-NEXT:    [[TMP10]] = and i32 [[TMP9]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -475,8 +465,7 @@ define i32 @xor(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP9]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vp.reduce.xor.nxv4i32(i32 0, <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP15]] = xor i32 [[TMP14]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP10]] to i64
@@ -521,8 +510,7 @@ define i32 @xor(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ [[START:%.*]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.xor.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]])
 ; NO-VP-NEXT:    [[TMP10]] = xor i32 [[TMP9]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -586,8 +574,7 @@ define i32 @smin(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP9]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vp.reduce.smin.nxv4i32(i32 2147483647, <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[RDX_MINMAX]] = call i32 @llvm.smin.i32(i32 [[TMP14]], i32 [[VEC_PHI]])
 ; IF-EVL-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP10]] to i64
@@ -633,8 +620,7 @@ define i32 @smin(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ [[START:%.*]], [[VECTOR_PH]] ], [ [[RDX_MINMAX:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.smin.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]])
 ; NO-VP-NEXT:    [[RDX_MINMAX]] = call i32 @llvm.smin.i32(i32 [[TMP9]], i32 [[VEC_PHI]])
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -700,8 +686,7 @@ define i32 @smax(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP9]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vp.reduce.smax.nxv4i32(i32 -2147483648, <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[RDX_MINMAX]] = call i32 @llvm.smax.i32(i32 [[TMP14]], i32 [[VEC_PHI]])
 ; IF-EVL-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP10]] to i64
@@ -747,8 +732,7 @@ define i32 @smax(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ [[START:%.*]], [[VECTOR_PH]] ], [ [[RDX_MINMAX:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.smax.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]])
 ; NO-VP-NEXT:    [[RDX_MINMAX]] = call i32 @llvm.smax.i32(i32 [[TMP9]], i32 [[VEC_PHI]])
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -814,8 +798,7 @@ define i32 @umin(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP9]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vp.reduce.umin.nxv4i32(i32 -1, <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[RDX_MINMAX]] = call i32 @llvm.umin.i32(i32 [[TMP14]], i32 [[VEC_PHI]])
 ; IF-EVL-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP10]] to i64
@@ -861,8 +844,7 @@ define i32 @umin(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ [[START:%.*]], [[VECTOR_PH]] ], [ [[RDX_MINMAX:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.umin.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]])
 ; NO-VP-NEXT:    [[RDX_MINMAX]] = call i32 @llvm.umin.i32(i32 [[TMP9]], i32 [[VEC_PHI]])
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -928,8 +910,7 @@ define i32 @umax(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP9]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vp.reduce.umax.nxv4i32(i32 0, <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[RDX_MINMAX]] = call i32 @llvm.umax.i32(i32 [[TMP14]], i32 [[VEC_PHI]])
 ; IF-EVL-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP10]] to i64
@@ -975,8 +956,7 @@ define i32 @umax(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ [[START:%.*]], [[VECTOR_PH]] ], [ [[RDX_MINMAX:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = call i32 @llvm.vector.reduce.umax.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]])
 ; NO-VP-NEXT:    [[RDX_MINMAX]] = call i32 @llvm.umax.i32(i32 [[TMP9]], i32 [[VEC_PHI]])
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -1042,8 +1022,7 @@ define float @fadd(ptr %a, i64 %n, float %start) {
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP9]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = call reassoc float @llvm.vp.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP15]] = fadd reassoc float [[TMP14]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP10]] to i64
@@ -1088,8 +1067,7 @@ define float @fadd(ptr %a, i64 %n, float %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi float [ [[START:%.*]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = call reassoc float @llvm.vector.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[WIDE_LOAD]])
 ; NO-VP-NEXT:    [[TMP10]] = fadd reassoc float [[TMP9]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -1153,8 +1131,7 @@ define float @fmul(ptr %a, i64 %n, float %start) {
 ; IF-EVL-NEXT:    [[VEC_IV:%.*]] = add <4 x i64> [[BROADCAST_SPLAT]], <i64 0, i64 1, i64 2, i64 3>
 ; IF-EVL-NEXT:    [[TMP1:%.*]] = icmp ule <4 x i64> [[VEC_IV]], [[BROADCAST_SPLAT2]]
 ; IF-EVL-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[IV]]
-; IF-EVL-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
-; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP3]], i32 4, <4 x i1> [[TMP1]], <4 x float> poison)
+; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP2]], i32 4, <4 x i1> [[TMP1]], <4 x float> poison)
 ; IF-EVL-NEXT:    [[TMP4:%.*]] = select reassoc <4 x i1> [[TMP1]], <4 x float> [[WIDE_MASKED_LOAD]], <4 x float> splat (float 1.000000e+00)
 ; IF-EVL-NEXT:    [[TMP5:%.*]] = call reassoc float @llvm.vector.reduce.fmul.v4f32(float 1.000000e+00, <4 x float> [[TMP4]])
 ; IF-EVL-NEXT:    [[MUL]] = fmul reassoc float [[TMP5]], [[RDX]]
@@ -1193,9 +1170,8 @@ define float @fmul(ptr %a, i64 %n, float %start) {
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi float [ [[START:%.*]], [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI1:%.*]] = phi float [ 1.000000e+00, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
 ; NO-VP-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 4
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
 ; NO-VP-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
 ; NO-VP-NEXT:    [[TMP6:%.*]] = call reassoc float @llvm.vector.reduce.fmul.v4f32(float 1.000000e+00, <4 x float> [[WIDE_LOAD]])
 ; NO-VP-NEXT:    [[TMP7]] = fmul reassoc float [[TMP6]], [[VEC_PHI]]
@@ -1263,8 +1239,7 @@ define float @fmin(ptr %a, i64 %n, float %start) #0 {
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP9]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = call fast float @llvm.vp.reduce.fmin.nxv4f32(float 0x47EFFFFFE0000000, <vscale x 4 x float> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[RDX_MINMAX_CMP:%.*]] = fcmp fast olt float [[TMP14]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[RDX_MINMAX_SELECT]] = select fast i1 [[RDX_MINMAX_CMP]], float [[TMP14]], float [[VEC_PHI]]
@@ -1311,8 +1286,7 @@ define float @fmin(ptr %a, i64 %n, float %start) #0 {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi float [ [[START:%.*]], [[VECTOR_PH]] ], [ [[RDX_MINMAX_SELECT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = call fast float @llvm.vector.reduce.fmin.nxv4f32(<vscale x 4 x float> [[WIDE_LOAD]])
 ; NO-VP-NEXT:    [[RDX_MINMAX_CMP:%.*]] = fcmp fast olt float [[TMP9]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[RDX_MINMAX_SELECT]] = select fast i1 [[RDX_MINMAX_CMP]], float [[TMP9]], float [[VEC_PHI]]
@@ -1379,8 +1353,7 @@ define float @fmax(ptr %a, i64 %n, float %start) #0 {
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP9]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = call fast float @llvm.vp.reduce.fmax.nxv4f32(float 0xC7EFFFFFE0000000, <vscale x 4 x float> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[RDX_MINMAX_CMP:%.*]] = fcmp fast ogt float [[TMP14]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[RDX_MINMAX_SELECT]] = select fast i1 [[RDX_MINMAX_CMP]], float [[TMP14]], float [[VEC_PHI]]
@@ -1427,8 +1400,7 @@ define float @fmax(ptr %a, i64 %n, float %start) #0 {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi float [ [[START:%.*]], [[VECTOR_PH]] ], [ [[RDX_MINMAX_SELECT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = call fast float @llvm.vector.reduce.fmax.nxv4f32(<vscale x 4 x float> [[WIDE_LOAD]])
 ; NO-VP-NEXT:    [[RDX_MINMAX_CMP:%.*]] = fcmp fast ogt float [[TMP9]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[RDX_MINMAX_SELECT]] = select fast i1 [[RDX_MINMAX_CMP]], float [[TMP9]], float [[VEC_PHI]]
@@ -1496,8 +1468,7 @@ define float @fminimum(ptr %a, i64 %n, float %start) {
 ; IF-EVL-NEXT:    [[VEC_IV:%.*]] = add <8 x i64> [[BROADCAST_SPLAT2]], <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>
 ; IF-EVL-NEXT:    [[TMP1:%.*]] = icmp ule <8 x i64> [[VEC_IV]], [[BROADCAST_SPLAT4]]
 ; IF-EVL-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[IV]]
-; IF-EVL-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
-; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP3]], i32 4, <8 x i1> [[TMP1]], <8 x float> poison)
+; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP2]], i32 4, <8 x i1> [[TMP1]], <8 x float> poison)
 ; IF-EVL-NEXT:    [[TMP4]] = call <8 x float> @llvm.minimum.v8f32(<8 x float> [[VEC_PHI]], <8 x float> [[WIDE_MASKED_LOAD]])
 ; IF-EVL-NEXT:    [[TMP5:%.*]] = select <8 x i1> [[TMP1]], <8 x float> [[TMP4]], <8 x float> [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 8
@@ -1538,9 +1509,8 @@ define float @fminimum(ptr %a, i64 %n, float %start) {
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <8 x float> [ [[MINMAX_IDENT_SPLAT]], [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x float> [ [[MINMAX_IDENT_SPLAT]], [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
 ; NO-VP-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 8
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x float>, ptr [[TMP4]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x float>, ptr [[TMP2]], align 4
 ; NO-VP-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x float>, ptr [[TMP5]], align 4
 ; NO-VP-NEXT:    [[TMP6]] = call <8 x float> @llvm.minimum.v8f32(<8 x float> [[VEC_PHI]], <8 x float> [[WIDE_LOAD]])
 ; NO-VP-NEXT:    [[TMP7]] = call <8 x float> @llvm.minimum.v8f32(<8 x float> [[VEC_PHI1]], <8 x float> [[WIDE_LOAD2]])
@@ -1608,8 +1578,7 @@ define float @fmaximum(ptr %a, i64 %n, float %start) {
 ; IF-EVL-NEXT:    [[VEC_IV:%.*]] = add <8 x i64> [[BROADCAST_SPLAT2]], <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>
 ; IF-EVL-NEXT:    [[TMP1:%.*]] = icmp ule <8 x i64> [[VEC_IV]], [[BROADCAST_SPLAT4]]
 ; IF-EVL-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[IV]]
-; IF-EVL-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
-; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP3]], i32 4, <8 x i1> [[TMP1]], <8 x float> poison)
+; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP2]], i32 4, <8 x i1> [[TMP1]], <8 x float> poison)
 ; IF-EVL-NEXT:    [[TMP4]] = call <8 x float> @llvm.maximum.v8f32(<8 x float> [[VEC_PHI]], <8 x float> [[WIDE_MASKED_LOAD]])
 ; IF-EVL-NEXT:    [[TMP5:%.*]] = select <8 x i1> [[TMP1]], <8 x float> [[TMP4]], <8 x float> [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 8
@@ -1650,9 +1619,8 @@ define float @fmaximum(ptr %a, i64 %n, float %start) {
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <8 x float> [ [[MINMAX_IDENT_SPLAT]], [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x float> [ [[MINMAX_IDENT_SPLAT]], [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
 ; NO-VP-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 8
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x float>, ptr [[TMP4]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x float>, ptr [[TMP2]], align 4
 ; NO-VP-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x float>, ptr [[TMP5]], align 4
 ; NO-VP-NEXT:    [[TMP6]] = call <8 x float> @llvm.maximum.v8f32(<8 x float> [[VEC_PHI]], <8 x float> [[WIDE_LOAD]])
 ; NO-VP-NEXT:    [[TMP7]] = call <8 x float> @llvm.maximum.v8f32(<8 x float> [[VEC_PHI1]], <8 x float> [[WIDE_LOAD2]])
@@ -1719,11 +1687,9 @@ define float @fmuladd(ptr %a, ptr %b, i64 %n, float %start) {
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP9]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP14]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD1:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP15]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD1:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = fmul reassoc <vscale x 4 x float> [[VP_OP_LOAD]], [[VP_OP_LOAD1]]
 ; IF-EVL-NEXT:    [[TMP17:%.*]] = call reassoc float @llvm.vp.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[TMP16]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP18]] = fadd reassoc float [[TMP17]], [[VEC_PHI]]
@@ -1771,11 +1737,9 @@ define float @fmuladd(ptr %a, ptr %b, i64 %n, float %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi float [ [[START:%.*]], [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[TMP9]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP10]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP9]], align 4
 ; NO-VP-NEXT:    [[TMP11:%.*]] = fmul reassoc <vscale x 4 x float> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; NO-VP-NEXT:    [[TMP12:%.*]] = call reassoc float @llvm.vector.reduce.fadd.nxv4f32(float -0.000000e+00, <vscale x 4 x float> [[TMP11]])
 ; NO-VP-NEXT:    [[TMP13]] = fadd reassoc float [[TMP12]], [[VEC_PHI]]
@@ -1844,8 +1808,7 @@ define i32 @anyof_icmp(ptr %a, i64 %n, i32 %start, i32 %inv) {
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP9]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = icmp slt <vscale x 4 x i32> [[VP_OP_LOAD]], splat (i32 3)
 ; IF-EVL-NEXT:    [[TMP16]] = call <vscale x 4 x i1> @llvm.vp.merge.nxv4i1(<vscale x 4 x i1> [[TMP14]], <vscale x 4 x i1> splat (i1 true), <vscale x 4 x i1> [[VEC_PHI]], i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP10]] to i64
@@ -1894,8 +1857,7 @@ define i32 @anyof_icmp(ptr %a, i64 %n, i32 %start, i32 %inv) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = icmp slt <vscale x 4 x i32> [[WIDE_LOAD]], splat (i32 3)
 ; NO-VP-NEXT:    [[TMP10]] = or <vscale x 4 x i1> [[VEC_PHI]], [[TMP9]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -1964,8 +1926,7 @@ define i32 @anyof_fcmp(ptr %a, i64 %n, i32 %start, i32 %inv) {
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP9]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = fcmp fast olt <vscale x 4 x float> [[VP_OP_LOAD]], splat (float 3.000000e+00)
 ; IF-EVL-NEXT:    [[TMP16]] = call <vscale x 4 x i1> @llvm.vp.merge.nxv4i1(<vscale x 4 x i1> [[TMP14]], <vscale x 4 x i1> splat (i1 true), <vscale x 4 x i1> [[VEC_PHI]], i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP17:%.*]] = zext i32 [[TMP10]] to i64
@@ -2014,8 +1975,7 @@ define i32 @anyof_fcmp(ptr %a, i64 %n, i32 %start, i32 %inv) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = fcmp fast olt <vscale x 4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00)
 ; NO-VP-NEXT:    [[TMP10]] = or <vscale x 4 x i1> [[VEC_PHI]], [[TMP9]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-interleave.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-interleave.ll
index 556446c3..80b0714 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-interleave.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-interleave.ll
@@ -1,12 +1,10 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck --check-prefix=IF-EVL %s
 
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck --check-prefix=NO-VP %s
 
 ; FIXME: interleaved accesses are not supported yet with predicated vectorization.
@@ -21,33 +19,35 @@ define void @interleave(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL-NEXT:    [[N_RND_UP:%.*]] = add i64 [[N:%.*]], [[TMP6]]
 ; IF-EVL-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N_RND_UP]], [[TMP5]]
 ; IF-EVL-NEXT:    [[N_VEC:%.*]] = sub i64 [[N_RND_UP]], [[N_MOD_VF]]
-; IF-EVL-NEXT:    [[TRIP_COUNT_MINUS_1:%.*]] = sub i64 [[N]], 1
 ; IF-EVL-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vscale.i64()
 ; IF-EVL-NEXT:    [[TMP8:%.*]] = mul nuw i64 [[TMP7]], 4
-; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i64> poison, i64 [[TRIP_COUNT_MINUS_1]], i64 0
-; IF-EVL-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
+; IF-EVL-NEXT:    [[TMP10:%.*]] = call <vscale x 4 x i64> @llvm.stepvector.nxv4i64()
+; IF-EVL-NEXT:    [[TMP12:%.*]] = mul <vscale x 4 x i64> [[TMP10]], splat (i64 1)
+; IF-EVL-NEXT:    [[INDUCTION:%.*]] = add <vscale x 4 x i64> zeroinitializer, [[TMP12]]
 ; IF-EVL-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; IF-EVL:       vector.body:
+; IF-EVL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; IF-EVL-NEXT:    [[EVL_BASED_IV:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
-; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 4 x i64> poison, i64 [[EVL_BASED_IV]], i64 0
+; IF-EVL-NEXT:    [[VEC_IND:%.*]] = phi <vscale x 4 x i64> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
+; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[TMP11:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
+; IF-EVL-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP11]] to i64
+; IF-EVL-NEXT:    [[TMP9:%.*]] = mul i64 1, [[TMP13]]
+; IF-EVL-NEXT:    [[BROADCAST_SPLATINSERT1:%.*]] = insertelement <vscale x 4 x i64> poison, i64 [[TMP9]], i64 0
 ; IF-EVL-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 4 x i64> [[BROADCAST_SPLATINSERT1]], <vscale x 4 x i64> poison, <vscale x 4 x i32> zeroinitializer
-; IF-EVL-NEXT:    [[TMP9:%.*]] = call <vscale x 4 x i64> @llvm.stepvector.nxv4i64()
-; IF-EVL-NEXT:    [[TMP10:%.*]] = add <vscale x 4 x i64> zeroinitializer, [[TMP9]]
-; IF-EVL-NEXT:    [[VEC_IV:%.*]] = add <vscale x 4 x i64> [[BROADCAST_SPLAT2]], [[TMP10]]
-; IF-EVL-NEXT:    [[TMP11:%.*]] = icmp ule <vscale x 4 x i64> [[VEC_IV]], [[BROADCAST_SPLAT]]
-; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [2 x i32], ptr [[B:%.*]], i64 [[EVL_BASED_IV]], i32 0
-; IF-EVL-NEXT:    [[INTERLEAVED_MASK:%.*]] = call <vscale x 8 x i1> @llvm.vector.interleave2.nxv8i1(<vscale x 4 x i1> [[TMP11]], <vscale x 4 x i1> [[TMP11]])
-; IF-EVL-NEXT:    [[WIDE_MASKED_VEC:%.*]] = call <vscale x 8 x i32> @llvm.masked.load.nxv8i32.p0(ptr [[TMP12]], i32 4, <vscale x 8 x i1> [[INTERLEAVED_MASK]], <vscale x 8 x i32> poison)
-; IF-EVL-NEXT:    [[STRIDED_VEC:%.*]] = call { <vscale x 4 x i32>, <vscale x 4 x i32> } @llvm.vector.deinterleave2.nxv8i32(<vscale x 8 x i32> [[WIDE_MASKED_VEC]])
-; IF-EVL-NEXT:    [[WIDE_MASKED_GATHER3:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[STRIDED_VEC]], 0
-; IF-EVL-NEXT:    [[WIDE_MASKED_GATHER5:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[STRIDED_VEC]], 1
+; IF-EVL-NEXT:    [[TMP21:%.*]] = getelementptr inbounds [2 x i32], ptr [[B:%.*]], <vscale x 4 x i64> [[VEC_IND]], i32 0
+; IF-EVL-NEXT:    [[WIDE_MASKED_GATHER3:%.*]] = call <vscale x 4 x i32> @llvm.vp.gather.nxv4i32.nxv4p0(<vscale x 4 x ptr> align 4 [[TMP21]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [2 x i32], ptr [[B]], <vscale x 4 x i64> [[VEC_IND]], i32 1
+; IF-EVL-NEXT:    [[WIDE_MASKED_GATHER5:%.*]] = call <vscale x 4 x i32> @llvm.vp.gather.nxv4i32.nxv4p0(<vscale x 4 x ptr> align 4 [[TMP23]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP11]])
 ; IF-EVL-NEXT:    [[TMP26:%.*]] = add nsw <vscale x 4 x i32> [[WIDE_MASKED_GATHER5]], [[WIDE_MASKED_GATHER3]]
-; IF-EVL-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i32, ptr [[TMP27]], i32 0
-; IF-EVL-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP26]], ptr [[TMP29]], i32 4, <vscale x 4 x i1> [[TMP11]])
-; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[EVL_BASED_IV]], [[TMP8]]
-; IF-EVL-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_EVL_NEXT]], [[N_VEC]]
-; IF-EVL-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP26]], ptr align 4 [[TMP18]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP11]])
+; IF-EVL-NEXT:    [[TMP14:%.*]] = zext i32 [[TMP11]] to i64
+; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP14]], [[EVL_BASED_IV]]
+; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP8]]
+; IF-EVL-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 4 x i64> [[VEC_IND]], [[BROADCAST_SPLAT2]]
+; IF-EVL-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; IF-EVL-NEXT:    br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; IF-EVL:       middle.block:
 ; IF-EVL-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; IF-EVL:       scalar.ph:
@@ -56,15 +56,15 @@ define void @interleave(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; IF-EVL:       for.body:
 ; IF-EVL-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; IF-EVL-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds [2 x i32], ptr [[B]], i64 [[IV]], i32 0
-; IF-EVL-NEXT:    [[TMP34:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+; IF-EVL-NEXT:    [[TMP16:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
 ; IF-EVL-NEXT:    [[ARRAYIDX2:%.*]] = getelementptr inbounds [2 x i32], ptr [[B]], i64 [[IV]], i32 1
-; IF-EVL-NEXT:    [[TMP35:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4
-; IF-EVL-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP35]], [[TMP34]]
+; IF-EVL-NEXT:    [[TMP17:%.*]] = load i32, ptr [[ARRAYIDX2]], align 4
+; IF-EVL-NEXT:    [[ADD:%.*]] = add nsw i32 [[TMP17]], [[TMP16]]
 ; IF-EVL-NEXT:    [[ARRAYIDX4:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
 ; IF-EVL-NEXT:    store i32 [[ADD]], ptr [[ARRAYIDX4]], align 4
 ; IF-EVL-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; IF-EVL-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
+; IF-EVL-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; IF-EVL:       for.cond.cleanup:
 ; IF-EVL-NEXT:    ret void
 ;
@@ -91,8 +91,7 @@ define void @interleave(ptr noalias %a, ptr noalias %b, i64 %N) {
 ; NO-VP-NEXT:    [[TMP19:%.*]] = extractvalue { <vscale x 4 x i32>, <vscale x 4 x i32> } [[STRIDED_VEC2]], 1
 ; NO-VP-NEXT:    [[TMP21:%.*]] = add nsw <vscale x 4 x i32> [[TMP19]], [[TMP18]]
 ; NO-VP-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, ptr [[TMP22]], i32 0
-; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP21]], ptr [[TMP24]], align 4
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP21]], ptr [[TMP22]], align 4
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
 ; NO-VP-NEXT:    [[TMP28:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; NO-VP-NEXT:    br i1 [[TMP28]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-intermediate-store.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-intermediate-store.ll
index 3e804c0..c7808b1 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-intermediate-store.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-intermediate-store.ll
@@ -1,24 +1,20 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefixes=IF-EVL-OUTLOOP
 
 ; RUN: opt -passes=loop-vectorize \
 ; RUN: -prefer-inloop-reductions \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefixes=IF-EVL-INLOOP
 
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefixes=NO-VP-OUTLOOP
 
 ; RUN: opt -passes=loop-vectorize \
 ; RUN: -prefer-inloop-reductions \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefixes=NO-VP-INLOOP
 
 define void @reduction_intermediate_store(ptr %a, i64 %n, i32 %start, ptr %addr) {
@@ -52,8 +48,7 @@ define void @reduction_intermediate_store(ptr %a, i64 %n, i32 %start, ptr %addr)
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP11:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP11]], i32 4, i1 true)
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-OUTLOOP-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[TMP17]], i32 0
-; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP18]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META0:![0-9]+]]
+; IF-EVL-OUTLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]]), !alias.scope [[META0:![0-9]+]]
 ; IF-EVL-OUTLOOP-NEXT:    [[VP_OP:%.*]] = add <vscale x 4 x i32> [[VP_OP_LOAD]], [[VEC_PHI]]
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP19]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[VP_OP]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP12]])
 ; IF-EVL-OUTLOOP-NEXT:    [[TMP21:%.*]] = zext i32 [[TMP12]] to i64
@@ -111,8 +106,7 @@ define void @reduction_intermediate_store(ptr %a, i64 %n, i32 %start, ptr %addr)
 ; IF-EVL-INLOOP-NEXT:    [[TMP13:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-INLOOP-NEXT:    [[TMP14:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP13]], i32 4, i1 true)
 ; IF-EVL-INLOOP-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[EVL_BASED_IV]]
-; IF-EVL-INLOOP-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[TMP19]], i32 0
-; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP20]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]]), !alias.scope [[META0:![0-9]+]]
+; IF-EVL-INLOOP-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP19]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]]), !alias.scope [[META0:![0-9]+]]
 ; IF-EVL-INLOOP-NEXT:    [[TMP21:%.*]] = call i32 @llvm.vp.reduce.add.nxv4i32(i32 0, <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP14]])
 ; IF-EVL-INLOOP-NEXT:    [[TMP22]] = add i32 [[TMP21]], [[VEC_PHI]]
 ; IF-EVL-INLOOP-NEXT:    [[TMP23:%.*]] = zext i32 [[TMP14]] to i64
@@ -168,8 +162,7 @@ define void @reduction_intermediate_store(ptr %a, i64 %n, i32 %start, ptr %addr)
 ; NO-VP-OUTLOOP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-OUTLOOP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP8]], [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-OUTLOOP-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
-; NO-VP-OUTLOOP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[TMP10]], i32 0
-; NO-VP-OUTLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP11]], align 4, !alias.scope [[META0:![0-9]+]]
+; NO-VP-OUTLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP10]], align 4, !alias.scope [[META0:![0-9]+]]
 ; NO-VP-OUTLOOP-NEXT:    [[TMP12]] = add <vscale x 4 x i32> [[WIDE_LOAD]], [[VEC_PHI]]
 ; NO-VP-OUTLOOP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP7]]
 ; NO-VP-OUTLOOP-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -223,8 +216,7 @@ define void @reduction_intermediate_store(ptr %a, i64 %n, i32 %start, ptr %addr)
 ; NO-VP-INLOOP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-INLOOP-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ [[START]], [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-INLOOP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
-; NO-VP-INLOOP-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i32 0
-; NO-VP-INLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP10]], align 4, !alias.scope [[META0:![0-9]+]]
+; NO-VP-INLOOP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP9]], align 4, !alias.scope [[META0:![0-9]+]]
 ; NO-VP-INLOOP-NEXT:    [[TMP11:%.*]] = call i32 @llvm.vector.reduce.add.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]])
 ; NO-VP-INLOOP-NEXT:    [[TMP12]] = add i32 [[TMP11]], [[VEC_PHI]]
 ; NO-VP-INLOOP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP7]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-iv32.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-iv32.ll
index 3d91738..bda9145 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-iv32.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-iv32.ll
@@ -1,12 +1,10 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck --check-prefix=IF-EVL %s
 
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck --check-prefix=NO-VP %s
 
 define void @iv32(ptr noalias %a, ptr noalias %b, i32 %N) {
@@ -29,11 +27,9 @@ define void @iv32(ptr noalias %a, ptr noalias %b, i32 %N) {
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = sub i32 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i32(i32 [[TMP11]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i32 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP15]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[VP_OP_LOAD]], ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[VP_OP_LOAD]], ptr align 4 [[TMP16]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i32 [[TMP12]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[IV_NEXT]] = add i32 [[IV]], [[TMP10]]
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = icmp eq i32 [[IV_NEXT]], [[N_VEC]]
@@ -72,11 +68,9 @@ define void @iv32(ptr noalias %a, ptr noalias %b, i32 %N) {
 ; NO-VP:       vector.body:
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i32 [[INDEX]]
-; NO-VP-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP5]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP4]], align 4
 ; NO-VP-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]
-; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; NO-VP-NEXT:    store <vscale x 4 x i32> [[WIDE_LOAD]], ptr [[TMP7]], align 4
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[WIDE_LOAD]], ptr [[TMP6]], align 4
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP12]]
 ; NO-VP-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; NO-VP-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-known-no-overflow.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-known-no-overflow.ll
index 258b7ce..2601bfd 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-known-no-overflow.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-known-no-overflow.ll
@@ -1,7 +1,6 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s
 
 ; TODO: We know the IV will never overflow here so we can skip the overflow
@@ -32,10 +31,9 @@ define void @trip_count_max_1024(ptr %p, i64 %tc) vscale_range(2, 1024) {
 ; CHECK-NEXT:    [[AVL:%.*]] = sub i64 [[UMAX]], [[EVL_BASED_IV]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i64, ptr [[P]], i64 [[EVL_BASED_IV]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i64, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP12]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP9]])
+; CHECK-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP11]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP9]])
 ; CHECK-NEXT:    [[VP_OP:%.*]] = add <vscale x 2 x i64> [[VP_OP_LOAD]], splat (i64 1)
-; CHECK-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VP_OP]], ptr align 8 [[TMP12]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP9]])
+; CHECK-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VP_OP]], ptr align 8 [[TMP11]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP9]])
 ; CHECK-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP9]] to i64
 ; CHECK-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP13]], [[EVL_BASED_IV]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP8]]
@@ -103,10 +101,9 @@ define void @overflow_at_0(ptr %p, i64 %tc) vscale_range(2, 1024) {
 ; CHECK-NEXT:    [[AVL:%.*]] = sub i64 [[TC]], [[EVL_BASED_IV]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i64, ptr [[P]], i64 [[EVL_BASED_IV]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i64, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP12]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP9]])
+; CHECK-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP11]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP9]])
 ; CHECK-NEXT:    [[VP_OP:%.*]] = add <vscale x 2 x i64> [[VP_OP_LOAD]], splat (i64 1)
-; CHECK-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VP_OP]], ptr align 8 [[TMP12]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP9]])
+; CHECK-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VP_OP]], ptr align 8 [[TMP11]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP9]])
 ; CHECK-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP9]] to i64
 ; CHECK-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP13]], [[EVL_BASED_IV]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP8]]
@@ -174,10 +171,9 @@ define void @no_overflow_at_0(ptr %p, i64 %tc) vscale_range(2, 1024) {
 ; CHECK-NEXT:    [[AVL:%.*]] = sub i64 [[TC_ADD]], [[EVL_BASED_IV]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[P]], i64 [[EVL_BASED_IV]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP8]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
+; CHECK-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP7]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
 ; CHECK-NEXT:    [[VP_OP:%.*]] = add <vscale x 2 x i64> [[VP_OP_LOAD]], splat (i64 1)
-; CHECK-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VP_OP]], ptr align 8 [[TMP8]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
+; CHECK-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VP_OP]], ptr align 8 [[TMP7]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
 ; CHECK-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP5]] to i64
 ; CHECK-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP9]], [[EVL_BASED_IV]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP4]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-masked-loadstore.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-masked-loadstore.ll
index 63acbdf..226372f 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-masked-loadstore.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-masked-loadstore.ll
@@ -1,12 +1,10 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefix=IF-EVL
 
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefix=NO-VP
 
 define void @masked_loadstore(ptr noalias %a, ptr noalias %b, i64 %n) {
@@ -29,14 +27,12 @@ define void @masked_loadstore(ptr noalias %a, ptr noalias %b, i64 %n) {
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP9]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[TMP15]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP16]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP15]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP17:%.*]] = icmp ne <vscale x 4 x i32> [[VP_OP_LOAD]], zeroinitializer
 ; IF-EVL-NEXT:    [[TMP19:%.*]] = getelementptr i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP20:%.*]] = getelementptr i32, ptr [[TMP19]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD3:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP20]], <vscale x 4 x i1> [[TMP17]], i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD3:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP19]], <vscale x 4 x i1> [[TMP17]], i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = add <vscale x 4 x i32> [[VP_OP_LOAD]], [[VP_OP_LOAD3]]
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[VP_OP]], ptr align 4 [[TMP20]], <vscale x 4 x i1> [[TMP17]], i32 [[TMP10]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[VP_OP]], ptr align 4 [[TMP19]], <vscale x 4 x i1> [[TMP17]], i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP21:%.*]] = zext i32 [[TMP10]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP21]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP8]]
@@ -68,23 +64,52 @@ define void @masked_loadstore(ptr noalias %a, ptr noalias %b, i64 %n) {
 ;
 ; NO-VP-LABEL: @masked_loadstore(
 ; NO-VP-NEXT:  entry:
+; NO-VP-NEXT:    [[TMP11:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP12:%.*]] = mul nuw i64 [[TMP11]], 4
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N:%.*]], [[TMP12]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; NO-VP:       vector.ph:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], [[TMP3]]
+; NO-VP-NEXT:    [[INC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4
 ; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
+; NO-VP:       vector.body:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[FOR_BODY]] ]
+; NO-VP-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP6]], align 4
+; NO-VP-NEXT:    [[TMP7:%.*]] = icmp ne <vscale x 4 x i32> [[WIDE_LOAD]], zeroinitializer
+; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[A:%.*]], i64 [[INDEX]]
+; NO-VP-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP8]], i32 4, <vscale x 4 x i1> [[TMP7]], <vscale x 4 x i32> poison)
+; NO-VP-NEXT:    [[TMP9:%.*]] = add <vscale x 4 x i32> [[WIDE_LOAD]], [[WIDE_MASKED_LOAD]]
+; NO-VP-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[TMP9]], ptr [[TMP8]], i32 4, <vscale x 4 x i1> [[TMP7]])
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; NO-VP-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[INC]]
+; NO-VP-NEXT:    br i1 [[TMP10]], label [[FOR_INC:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; NO-VP:       middle.block:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[INC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; NO-VP:       scalar.ph:
+; NO-VP-NEXT:    [[I_011:%.*]] = phi i64 [ [[INC]], [[FOR_INC]] ], [ 0, [[ENTRY:%.*]] ]
+; NO-VP-NEXT:    br label [[FOR_BODY1:%.*]]
 ; NO-VP:       for.body:
-; NO-VP-NEXT:    [[I_011:%.*]] = phi i64 [ [[INC:%.*]], [[FOR_INC:%.*]] ], [ 0, [[ENTRY:%.*]] ]
-; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[I_011]]
+; NO-VP-NEXT:    [[I_11:%.*]] = phi i64 [ [[INC1:%.*]], [[FOR_INC1:%.*]] ], [ [[I_011]], [[SCALAR_PH]] ]
+; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[I_11]]
 ; NO-VP-NEXT:    [[TMP0:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
 ; NO-VP-NEXT:    [[CMP1:%.*]] = icmp ne i32 [[TMP0]], 0
-; NO-VP-NEXT:    br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[FOR_INC]]
+; NO-VP-NEXT:    br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[FOR_INC1]]
 ; NO-VP:       if.then:
-; NO-VP-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[I_011]]
+; NO-VP-NEXT:    [[ARRAYIDX3:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[I_11]]
 ; NO-VP-NEXT:    [[TMP1:%.*]] = load i32, ptr [[ARRAYIDX3]], align 4
 ; NO-VP-NEXT:    [[ADD:%.*]] = add i32 [[TMP0]], [[TMP1]]
 ; NO-VP-NEXT:    store i32 [[ADD]], ptr [[ARRAYIDX3]], align 4
-; NO-VP-NEXT:    br label [[FOR_INC]]
+; NO-VP-NEXT:    br label [[FOR_INC1]]
 ; NO-VP:       for.inc:
-; NO-VP-NEXT:    [[INC]] = add nuw nsw i64 [[I_011]], 1
-; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INC]], [[N:%.*]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[EXIT:%.*]], label [[FOR_BODY]]
+; NO-VP-NEXT:    [[INC1]] = add nuw nsw i64 [[I_11]], 1
+; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INC1]], [[N]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[EXIT]], label [[FOR_BODY1]], !llvm.loop [[LOOP3:![0-9]+]]
 ; NO-VP:       exit:
 ; NO-VP-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-no-masking.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-no-masking.ll
index 1c49fba..6285a6f 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-no-masking.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-no-masking.ll
@@ -1,12 +1,10 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s
 
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s
 
 ; No need to emit predicated vector code if the vector instructions with masking are not required.
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-ordered-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-ordered-reduction.ll
index a97c4b3..41bddaa 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-ordered-reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-ordered-reduction.ll
@@ -1,14 +1,12 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -passes=loop-vectorize \
 ; RUN: -force-ordered-reductions=true -hints-allow-reordering=false \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v,+f -S < %s| FileCheck %s --check-prefix=IF-EVL
 
 ; RUN: opt -passes=loop-vectorize \
 ; RUN: -force-ordered-reductions=true -hints-allow-reordering=false \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v,+f -S < %s| FileCheck %s --check-prefix=NO-VP
 
 define float @fadd(ptr noalias nocapture readonly %a, i64 %n) {
@@ -32,8 +30,7 @@ define float @fadd(ptr noalias nocapture readonly %a, i64 %n) {
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP9]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP14]] = call float @llvm.vp.reduce.fadd.nxv4f32(float [[VEC_PHI]], <vscale x 4 x float> [[VP_OP_LOAD]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP10]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP15]], [[EVL_BASED_IV]]
@@ -61,18 +58,45 @@ define float @fadd(ptr noalias nocapture readonly %a, i64 %n) {
 ;
 ; NO-VP-LABEL: @fadd(
 ; NO-VP-NEXT:  entry:
+; NO-VP-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 4
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N1:%.*]], [[TMP1]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
+; NO-VP:       vector.ph:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N1]], [[TMP3]]
+; NO-VP-NEXT:    [[N:%.*]] = sub i64 [[N1]], [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4
 ; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
-; NO-VP:       for.body:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
+; NO-VP:       vector.body:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; NO-VP-NEXT:    [[SUM_07:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
 ; NO-VP-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[IV]]
-; NO-VP-NEXT:    [[TMP0:%.*]] = load float, ptr [[ARRAYIDX]], align 4
-; NO-VP-NEXT:    [[ADD]] = fadd float [[TMP0]], [[SUM_07]]
-; NO-VP-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
-; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N:%.*]]
-; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[FOR_END:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[ARRAYIDX]], align 4
+; NO-VP-NEXT:    [[ADD]] = call float @llvm.vector.reduce.fadd.nxv4f32(float [[SUM_07]], <vscale x 4 x float> [[WIDE_LOAD]])
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw i64 [[IV]], [[TMP5]]
+; NO-VP-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; NO-VP:       middle.block:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N1]], [[N]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; NO-VP:       scalar.ph:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
+; NO-VP-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[ADD]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY1]] ]
+; NO-VP-NEXT:    br label [[FOR_BODY1:%.*]]
+; NO-VP:       for.body:
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[FOR_BODY1]] ]
+; NO-VP-NEXT:    [[SUM_7:%.*]] = phi float [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[ADD1:%.*]], [[FOR_BODY1]] ]
+; NO-VP-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV1]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = load float, ptr [[ARRAYIDX1]], align 4
+; NO-VP-NEXT:    [[ADD1]] = fadd float [[TMP9]], [[SUM_7]]
+; NO-VP-NEXT:    [[IV_NEXT1]] = add nuw nsw i64 [[IV1]], 1
+; NO-VP-NEXT:    [[EXITCOND_NOT1:%.*]] = icmp eq i64 [[IV_NEXT1]], [[N1]]
+; NO-VP-NEXT:    br i1 [[EXITCOND_NOT1]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP3:![0-9]+]]
 ; NO-VP:       for.end:
-; NO-VP-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD]], [[FOR_BODY]] ]
+; NO-VP-NEXT:    [[ADD_LCSSA:%.*]] = phi float [ [[ADD1]], [[FOR_BODY1]] ], [ [[ADD]], [[MIDDLE_BLOCK]] ]
 ; NO-VP-NEXT:    ret float [[ADD_LCSSA]]
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-reduction-cost.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-reduction-cost.ll
index 15f9993..10d83f4 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-reduction-cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-reduction-cost.ll
@@ -1,7 +1,6 @@
 ; REQUIRES: asserts
 ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize --disable-output \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s 2>&1 | FileCheck %s
 
 ; CHECK: Cost of 2 for VF vscale x 4: WIDEN-INTRINSIC vp<%{{.+}}> = call llvm.vp.merge(ir<true>, ir<%add>, ir<%rdx>, vp<%{{.+}}>)
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-reduction.ll
index 8b14414..d2dc482 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-reduction.ll
@@ -1,12 +1,10 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v,+f -S < %s| FileCheck %s --check-prefix=IF-EVL
 
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v,+f -S < %s| FileCheck %s --check-prefix=NO-VP
 
 define i32 @add(ptr %a, i64 %n, i32 %start) {
@@ -31,8 +29,7 @@ define i32 @add(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = add <vscale x 4 x i32> [[VP_OP_LOAD]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP14]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[VP_OP]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP10]] to i64
@@ -79,8 +76,7 @@ define i32 @add(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP6]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP9]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
 ; NO-VP-NEXT:    [[TMP10]] = add <vscale x 4 x i32> [[WIDE_LOAD]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -145,8 +141,7 @@ define i32 @mul(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-NEXT:    [[VEC_IV:%.*]] = add <8 x i64> [[BROADCAST_SPLAT]], <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>
 ; IF-EVL-NEXT:    [[TMP2:%.*]] = icmp ule <8 x i64> [[VEC_IV]], [[BROADCAST_SPLAT2]]
 ; IF-EVL-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[IV]]
-; IF-EVL-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
-; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP4]], i32 4, <8 x i1> [[TMP2]], <8 x i32> poison)
+; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP3]], i32 4, <8 x i1> [[TMP2]], <8 x i32> poison)
 ; IF-EVL-NEXT:    [[TMP5]] = mul <8 x i32> [[WIDE_MASKED_LOAD]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP6:%.*]] = select <8 x i1> [[TMP2]], <8 x i32> [[TMP5]], <8 x i32> [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 8
@@ -186,9 +181,8 @@ define i32 @mul(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <8 x i32> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x i32> [ splat (i32 1), [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
 ; NO-VP-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 8
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP3]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP2]], align 4
 ; NO-VP-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i32>, ptr [[TMP4]], align 4
 ; NO-VP-NEXT:    [[TMP5]] = mul <8 x i32> [[WIDE_LOAD]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[TMP6]] = mul <8 x i32> [[WIDE_LOAD2]], [[VEC_PHI1]]
@@ -256,8 +250,7 @@ define i32 @or(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = or <vscale x 4 x i32> [[VP_OP_LOAD]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP14]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[VP_OP]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP10]] to i64
@@ -304,8 +297,7 @@ define i32 @or(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP6]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP9]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
 ; NO-VP-NEXT:    [[TMP10]] = or <vscale x 4 x i32> [[WIDE_LOAD]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -370,8 +362,7 @@ define i32 @and(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = and <vscale x 4 x i32> [[VP_OP_LOAD]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP14]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[VP_OP]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP10]] to i64
@@ -418,8 +409,7 @@ define i32 @and(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP6]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP9]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
 ; NO-VP-NEXT:    [[TMP10]] = and <vscale x 4 x i32> [[WIDE_LOAD]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -484,8 +474,7 @@ define i32 @xor(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = xor <vscale x 4 x i32> [[VP_OP_LOAD]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP14]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[VP_OP]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP10]] to i64
@@ -532,8 +521,7 @@ define i32 @xor(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[TMP6]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP9]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
 ; NO-VP-NEXT:    [[TMP10]] = xor <vscale x 4 x i32> [[WIDE_LOAD]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -599,8 +587,7 @@ define i32 @smin(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = icmp slt <vscale x 4 x i32> [[VP_OP_LOAD]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = select <vscale x 4 x i1> [[TMP13]], <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i32> [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP15]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[TMP14]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP9]])
@@ -650,8 +637,7 @@ define i32 @smin(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = icmp slt <vscale x 4 x i32> [[WIDE_LOAD]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[TMP10]] = select <vscale x 4 x i1> [[TMP9]], <vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[VEC_PHI]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -720,8 +706,7 @@ define i32 @smax(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = icmp sgt <vscale x 4 x i32> [[VP_OP_LOAD]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = select <vscale x 4 x i1> [[TMP13]], <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i32> [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP15]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[TMP14]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP9]])
@@ -771,8 +756,7 @@ define i32 @smax(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = icmp sgt <vscale x 4 x i32> [[WIDE_LOAD]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[TMP10]] = select <vscale x 4 x i1> [[TMP9]], <vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[VEC_PHI]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -841,8 +825,7 @@ define i32 @umin(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = icmp ult <vscale x 4 x i32> [[VP_OP_LOAD]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = select <vscale x 4 x i1> [[TMP13]], <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i32> [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP15]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[TMP14]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP9]])
@@ -892,8 +875,7 @@ define i32 @umin(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = icmp ult <vscale x 4 x i32> [[WIDE_LOAD]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[TMP10]] = select <vscale x 4 x i1> [[TMP9]], <vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[VEC_PHI]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -962,8 +944,7 @@ define i32 @umax(ptr %a, i64 %n, i32 %start) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = icmp ugt <vscale x 4 x i32> [[VP_OP_LOAD]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = select <vscale x 4 x i1> [[TMP13]], <vscale x 4 x i32> [[VP_OP_LOAD]], <vscale x 4 x i32> [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP15]] = call <vscale x 4 x i32> @llvm.vp.merge.nxv4i32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x i32> [[TMP14]], <vscale x 4 x i32> [[VEC_PHI]], i32 [[TMP9]])
@@ -1013,8 +994,7 @@ define i32 @umax(ptr %a, i64 %n, i32 %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = icmp ugt <vscale x 4 x i32> [[WIDE_LOAD]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[TMP10]] = select <vscale x 4 x i1> [[TMP9]], <vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[VEC_PHI]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -1082,8 +1062,7 @@ define float @fadd(ptr %a, i64 %n, float %start) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = fadd reassoc <vscale x 4 x float> [[VP_OP_LOAD]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP14]] = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x float> [[VP_OP]], <vscale x 4 x float> [[VEC_PHI]], i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP10]] to i64
@@ -1130,8 +1109,7 @@ define float @fadd(ptr %a, i64 %n, float %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ [[TMP6]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP9]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
 ; NO-VP-NEXT:    [[TMP10]] = fadd reassoc <vscale x 4 x float> [[WIDE_LOAD]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -1196,8 +1174,7 @@ define float @fmul(ptr %a, i64 %n, float %start) {
 ; IF-EVL-NEXT:    [[VEC_IV:%.*]] = add <8 x i64> [[BROADCAST_SPLAT]], <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>
 ; IF-EVL-NEXT:    [[TMP2:%.*]] = icmp ule <8 x i64> [[VEC_IV]], [[BROADCAST_SPLAT2]]
 ; IF-EVL-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[IV]]
-; IF-EVL-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP3]], i32 0
-; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP4]], i32 4, <8 x i1> [[TMP2]], <8 x float> poison)
+; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP3]], i32 4, <8 x i1> [[TMP2]], <8 x float> poison)
 ; IF-EVL-NEXT:    [[TMP5]] = fmul reassoc <8 x float> [[WIDE_MASKED_LOAD]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP6:%.*]] = select reassoc <8 x i1> [[TMP2]], <8 x float> [[TMP5]], <8 x float> [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 8
@@ -1237,9 +1214,8 @@ define float @fmul(ptr %a, i64 %n, float %start) {
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <8 x float> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x float> [ splat (float 1.000000e+00), [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
 ; NO-VP-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 8
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x float>, ptr [[TMP3]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x float>, ptr [[TMP2]], align 4
 ; NO-VP-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x float>, ptr [[TMP4]], align 4
 ; NO-VP-NEXT:    [[TMP5]] = fmul reassoc <8 x float> [[WIDE_LOAD]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[TMP6]] = fmul reassoc <8 x float> [[WIDE_LOAD2]], [[VEC_PHI1]]
@@ -1308,8 +1284,7 @@ define float @fmin(ptr %a, i64 %n, float %start) #0 {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP11]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = fcmp fast olt <vscale x 4 x float> [[VP_OP_LOAD]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = select <vscale x 4 x i1> [[TMP13]], <vscale x 4 x float> [[VP_OP_LOAD]], <vscale x 4 x float> [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP15]] = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x float> [[TMP14]], <vscale x 4 x float> [[VEC_PHI]], i32 [[TMP9]])
@@ -1359,8 +1334,7 @@ define float @fmin(ptr %a, i64 %n, float %start) #0 {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = fcmp fast olt <vscale x 4 x float> [[WIDE_LOAD]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[TMP10]] = select <vscale x 4 x i1> [[TMP9]], <vscale x 4 x float> [[WIDE_LOAD]], <vscale x 4 x float> [[VEC_PHI]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -1429,8 +1403,7 @@ define float @fmax(ptr %a, i64 %n, float %start) #0 {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP11]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = fcmp fast ogt <vscale x 4 x float> [[VP_OP_LOAD]], [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = select <vscale x 4 x i1> [[TMP13]], <vscale x 4 x float> [[VP_OP_LOAD]], <vscale x 4 x float> [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[TMP15]] = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x float> [[TMP14]], <vscale x 4 x float> [[VEC_PHI]], i32 [[TMP9]])
@@ -1480,8 +1453,7 @@ define float @fmax(ptr %a, i64 %n, float %start) #0 {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = fcmp fast ogt <vscale x 4 x float> [[WIDE_LOAD]], [[VEC_PHI]]
 ; NO-VP-NEXT:    [[TMP10]] = select <vscale x 4 x i1> [[TMP9]], <vscale x 4 x float> [[WIDE_LOAD]], <vscale x 4 x float> [[VEC_PHI]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -1549,8 +1521,7 @@ define float @fminimum(ptr %a, i64 %n, float %start) {
 ; IF-EVL-NEXT:    [[VEC_IV:%.*]] = add <8 x i64> [[BROADCAST_SPLAT2]], <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>
 ; IF-EVL-NEXT:    [[TMP1:%.*]] = icmp ule <8 x i64> [[VEC_IV]], [[BROADCAST_SPLAT4]]
 ; IF-EVL-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[IV]]
-; IF-EVL-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
-; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP3]], i32 4, <8 x i1> [[TMP1]], <8 x float> poison)
+; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP2]], i32 4, <8 x i1> [[TMP1]], <8 x float> poison)
 ; IF-EVL-NEXT:    [[TMP4]] = call <8 x float> @llvm.minimum.v8f32(<8 x float> [[VEC_PHI]], <8 x float> [[WIDE_MASKED_LOAD]])
 ; IF-EVL-NEXT:    [[TMP5:%.*]] = select <8 x i1> [[TMP1]], <8 x float> [[TMP4]], <8 x float> [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 8
@@ -1591,9 +1562,8 @@ define float @fminimum(ptr %a, i64 %n, float %start) {
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
 ; NO-VP-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 8
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x float>, ptr [[TMP2]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x float>, ptr [[TMP1]], align 4
 ; NO-VP-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x float>, ptr [[TMP3]], align 4
 ; NO-VP-NEXT:    [[TMP4]] = call <8 x float> @llvm.minimum.v8f32(<8 x float> [[VEC_PHI]], <8 x float> [[WIDE_LOAD]])
 ; NO-VP-NEXT:    [[TMP5]] = call <8 x float> @llvm.minimum.v8f32(<8 x float> [[VEC_PHI1]], <8 x float> [[WIDE_LOAD2]])
@@ -1661,8 +1631,7 @@ define float @fmaximum(ptr %a, i64 %n, float %start) {
 ; IF-EVL-NEXT:    [[VEC_IV:%.*]] = add <8 x i64> [[BROADCAST_SPLAT2]], <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>
 ; IF-EVL-NEXT:    [[TMP1:%.*]] = icmp ule <8 x i64> [[VEC_IV]], [[BROADCAST_SPLAT4]]
 ; IF-EVL-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[IV]]
-; IF-EVL-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
-; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP3]], i32 4, <8 x i1> [[TMP1]], <8 x float> poison)
+; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP2]], i32 4, <8 x i1> [[TMP1]], <8 x float> poison)
 ; IF-EVL-NEXT:    [[TMP4]] = call <8 x float> @llvm.maximum.v8f32(<8 x float> [[VEC_PHI]], <8 x float> [[WIDE_MASKED_LOAD]])
 ; IF-EVL-NEXT:    [[TMP5:%.*]] = select <8 x i1> [[TMP1]], <8 x float> [[TMP4]], <8 x float> [[VEC_PHI]]
 ; IF-EVL-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 8
@@ -1703,9 +1672,8 @@ define float @fmaximum(ptr %a, i64 %n, float %start) {
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI1:%.*]] = phi <8 x float> [ [[BROADCAST_SPLAT]], [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
 ; NO-VP-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 8
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x float>, ptr [[TMP2]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x float>, ptr [[TMP1]], align 4
 ; NO-VP-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x float>, ptr [[TMP3]], align 4
 ; NO-VP-NEXT:    [[TMP4]] = call <8 x float> @llvm.maximum.v8f32(<8 x float> [[VEC_PHI]], <8 x float> [[WIDE_LOAD]])
 ; NO-VP-NEXT:    [[TMP5]] = call <8 x float> @llvm.maximum.v8f32(<8 x float> [[VEC_PHI1]], <8 x float> [[WIDE_LOAD2]])
@@ -1773,11 +1741,9 @@ define float @fmuladd(ptr %a, ptr %b, i64 %n, float %start) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP14]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD1:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP15]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD1:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = call reassoc <vscale x 4 x float> @llvm.fmuladd.nxv4f32(<vscale x 4 x float> [[VP_OP_LOAD]], <vscale x 4 x float> [[VP_OP_LOAD1]], <vscale x 4 x float> [[VEC_PHI]])
 ; IF-EVL-NEXT:    [[TMP17]] = call <vscale x 4 x float> @llvm.vp.merge.nxv4f32(<vscale x 4 x i1> splat (i1 true), <vscale x 4 x float> [[TMP16]], <vscale x 4 x float> [[VEC_PHI]], i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = zext i32 [[TMP10]] to i64
@@ -1826,11 +1792,9 @@ define float @fmuladd(ptr %a, ptr %b, i64 %n, float %start) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x float> [ [[TMP6]], [[VECTOR_PH]] ], [ [[TMP12:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP9]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
 ; NO-VP-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[TMP10]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP11]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x float>, ptr [[TMP10]], align 4
 ; NO-VP-NEXT:    [[TMP12]] = call reassoc <vscale x 4 x float> @llvm.fmuladd.nxv4f32(<vscale x 4 x float> [[WIDE_LOAD]], <vscale x 4 x float> [[WIDE_LOAD1]], <vscale x 4 x float> [[VEC_PHI]])
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; NO-VP-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -1898,8 +1862,7 @@ define i32 @anyof_icmp(ptr %a, i64 %n, i32 %start, i32 %inv) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = icmp slt <vscale x 4 x i32> [[VP_OP_LOAD]], splat (i32 3)
 ; IF-EVL-NEXT:    [[TMP15]] = call <vscale x 4 x i1> @llvm.vp.merge.nxv4i1(<vscale x 4 x i1> [[TMP13]], <vscale x 4 x i1> splat (i1 true), <vscale x 4 x i1> [[VEC_PHI]], i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP9]] to i64
@@ -1948,8 +1911,7 @@ define i32 @anyof_icmp(ptr %a, i64 %n, i32 %start, i32 %inv) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = icmp slt <vscale x 4 x i32> [[WIDE_LOAD]], splat (i32 3)
 ; NO-VP-NEXT:    [[TMP10]] = or <vscale x 4 x i1> [[VEC_PHI]], [[TMP9]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
@@ -2018,8 +1980,7 @@ define i32 @anyof_fcmp(ptr %a, i64 %n, i32 %start, i32 %inv) {
 ; IF-EVL-NEXT:    [[AVL:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP11]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x float> @llvm.vp.load.nxv4f32.p0(ptr align 4 [[TMP11]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = fcmp fast olt <vscale x 4 x float> [[VP_OP_LOAD]], splat (float 3.000000e+00)
 ; IF-EVL-NEXT:    [[TMP15]] = call <vscale x 4 x i1> @llvm.vp.merge.nxv4i1(<vscale x 4 x i1> [[TMP13]], <vscale x 4 x i1> splat (i1 true), <vscale x 4 x i1> [[VEC_PHI]], i32 [[TMP9]])
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP9]] to i64
@@ -2068,8 +2029,7 @@ define i32 @anyof_fcmp(ptr %a, i64 %n, i32 %start, i32 %inv) {
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x float>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = fcmp fast olt <vscale x 4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00)
 ; NO-VP-NEXT:    [[TMP10]] = or <vscale x 4 x i1> [[VEC_PHI]], [[TMP9]]
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-reverse-load-store.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-reverse-load-store.ll
index 91d94e5..62eef6c 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-reverse-load-store.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-reverse-load-store.ll
@@ -1,12 +1,10 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefix=IF-EVL
 
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefix=NO-VP
 
 define void @reverse_load_store(i64 %startval, ptr noalias %ptr, ptr noalias %ptr2) {
@@ -75,18 +73,61 @@ define void @reverse_load_store(i64 %startval, ptr noalias %ptr, ptr noalias %pt
 ;
 ; NO-VP-LABEL: @reverse_load_store(
 ; NO-VP-NEXT:  entry:
+; NO-VP-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 4
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP1]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; NO-VP:       vector.ph:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[STARTVAL:%.*]], [[N_VEC]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = trunc i64 [[N_VEC]] to i32
 ; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
+; NO-VP:       vector.body:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[FOR_BODY]] ]
+; NO-VP-NEXT:    [[OFFSET_IDX:%.*]] = sub i64 [[STARTVAL]], [[INDEX]]
+; NO-VP-NEXT:    [[TMP8:%.*]] = add i64 [[OFFSET_IDX]], -1
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[PTR:%.*]], i64 [[TMP8]]
+; NO-VP-NEXT:    [[TMP10:%.*]] = mul i64 0, [[TMP5]]
+; NO-VP-NEXT:    [[TMP11:%.*]] = sub i64 [[TMP5]], 1
+; NO-VP-NEXT:    [[TMP12:%.*]] = mul i64 -1, [[TMP11]]
+; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i64 [[TMP10]]
+; NO-VP-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i64 [[TMP12]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP14]], align 4
+; NO-VP-NEXT:    [[REVERSE:%.*]] = call <vscale x 4 x i32> @llvm.vector.reverse.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]])
+; NO-VP-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[PTR2:%.*]], i64 [[TMP8]]
+; NO-VP-NEXT:    [[TMP16:%.*]] = mul i64 0, [[TMP5]]
+; NO-VP-NEXT:    [[TMP17:%.*]] = sub i64 [[TMP5]], 1
+; NO-VP-NEXT:    [[TMP18:%.*]] = mul i64 -1, [[TMP17]]
+; NO-VP-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[TMP15]], i64 [[TMP16]]
+; NO-VP-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[TMP19]], i64 [[TMP18]]
+; NO-VP-NEXT:    [[REVERSE1:%.*]] = call <vscale x 4 x i32> @llvm.vector.reverse.nxv4i32(<vscale x 4 x i32> [[REVERSE]])
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[REVERSE1]], ptr [[TMP20]], align 4
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; NO-VP-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; NO-VP:       middle.block:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label [[LOOPEND:%.*]], label [[SCALAR_PH]]
+; NO-VP:       scalar.ph:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[TMP6]], [[MIDDLE_BLOCK]] ], [ [[STARTVAL]], [[ENTRY:%.*]] ]
+; NO-VP-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i32 [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; NO-VP-NEXT:    br label [[FOR_BODY1:%.*]]
 ; NO-VP:       for.body:
-; NO-VP-NEXT:    [[ADD_PHI:%.*]] = phi i64 [ [[STARTVAL:%.*]], [[ENTRY:%.*]] ], [ [[ADD:%.*]], [[FOR_BODY]] ]
-; NO-VP-NEXT:    [[I:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
+; NO-VP-NEXT:    [[ADD_PHI:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[FOR_BODY1]] ]
+; NO-VP-NEXT:    [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL2]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[FOR_BODY1]] ]
 ; NO-VP-NEXT:    [[ADD]] = add i64 [[ADD_PHI]], -1
-; NO-VP-NEXT:    [[GEPL:%.*]] = getelementptr inbounds i32, ptr [[PTR:%.*]], i64 [[ADD]]
+; NO-VP-NEXT:    [[GEPL:%.*]] = getelementptr inbounds i32, ptr [[PTR]], i64 [[ADD]]
 ; NO-VP-NEXT:    [[TMP:%.*]] = load i32, ptr [[GEPL]], align 4
-; NO-VP-NEXT:    [[GEPS:%.*]] = getelementptr inbounds i32, ptr [[PTR2:%.*]], i64 [[ADD]]
+; NO-VP-NEXT:    [[GEPS:%.*]] = getelementptr inbounds i32, ptr [[PTR2]], i64 [[ADD]]
 ; NO-VP-NEXT:    store i32 [[TMP]], ptr [[GEPS]], align 4
 ; NO-VP-NEXT:    [[INC]] = add i32 [[I]], 1
 ; NO-VP-NEXT:    [[EXITCOND:%.*]] = icmp ne i32 [[INC]], 1024
-; NO-VP-NEXT:    br i1 [[EXITCOND]], label [[FOR_BODY]], label [[LOOPEND:%.*]]
+; NO-VP-NEXT:    br i1 [[EXITCOND]], label [[FOR_BODY1]], label [[LOOPEND]], !llvm.loop [[LOOP3:![0-9]+]]
 ; NO-VP:       loopend:
 ; NO-VP-NEXT:    ret void
 ;
@@ -132,8 +173,7 @@ define void @reverse_load_store_masked(i64 %startval, ptr noalias %ptr, ptr noal
 ; IF-EVL-NEXT:    [[OFFSET_IDX3:%.*]] = trunc i64 [[EVL_BASED_IV]] to i32
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = add i64 [[OFFSET_IDX]], -1
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[PTR:%.*]], i32 [[OFFSET_IDX3]]
-; IF-EVL-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP5]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP5]])
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = icmp slt <vscale x 4 x i32> [[VP_OP_LOAD]], splat (i32 100)
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr i32, ptr [[PTR1:%.*]], i64 [[TMP11]]
 ; IF-EVL-NEXT:    [[TMP26:%.*]] = zext i32 [[TMP5]] to i64
@@ -189,25 +229,74 @@ define void @reverse_load_store_masked(i64 %startval, ptr noalias %ptr, ptr noal
 ;
 ; NO-VP-LABEL: @reverse_load_store_masked(
 ; NO-VP-NEXT:  entry:
+; NO-VP-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 4
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1024, [[TMP1]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[ENTRY:%.*]], label [[VECTOR_PH:%.*]]
+; NO-VP:       vector.ph:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 1024, [[TMP3]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 1024, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4
+; NO-VP-NEXT:    [[TMP6:%.*]] = sub i64 [[STARTVAL1:%.*]], [[N_VEC]]
+; NO-VP-NEXT:    [[TMP7:%.*]] = trunc i64 [[N_VEC]] to i32
 ; NO-VP-NEXT:    br label [[FOR_BODY:%.*]]
+; NO-VP:       vector.body:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[FOR_BODY]] ]
+; NO-VP-NEXT:    [[OFFSET_IDX:%.*]] = sub i64 [[STARTVAL1]], [[INDEX]]
+; NO-VP-NEXT:    [[OFFSET_IDX1:%.*]] = trunc i64 [[INDEX]] to i32
+; NO-VP-NEXT:    [[TMP8:%.*]] = add i64 [[OFFSET_IDX]], -1
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[PTR:%.*]], i32 [[OFFSET_IDX1]]
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP9]], align 4
+; NO-VP-NEXT:    [[TMP10:%.*]] = icmp slt <vscale x 4 x i32> [[WIDE_LOAD]], splat (i32 100)
+; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[PTR1:%.*]], i64 [[TMP8]]
+; NO-VP-NEXT:    [[TMP12:%.*]] = mul i64 0, [[TMP5]]
+; NO-VP-NEXT:    [[TMP13:%.*]] = sub i64 [[TMP5]], 1
+; NO-VP-NEXT:    [[TMP14:%.*]] = mul i64 -1, [[TMP13]]
+; NO-VP-NEXT:    [[TMP15:%.*]] = getelementptr i32, ptr [[TMP11]], i64 [[TMP12]]
+; NO-VP-NEXT:    [[TMP16:%.*]] = getelementptr i32, ptr [[TMP15]], i64 [[TMP14]]
+; NO-VP-NEXT:    [[REVERSE:%.*]] = call <vscale x 4 x i1> @llvm.vector.reverse.nxv4i1(<vscale x 4 x i1> [[TMP10]])
+; NO-VP-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.masked.load.nxv4i32.p0(ptr [[TMP16]], i32 4, <vscale x 4 x i1> [[REVERSE]], <vscale x 4 x i32> poison)
+; NO-VP-NEXT:    [[REVERSE2:%.*]] = call <vscale x 4 x i32> @llvm.vector.reverse.nxv4i32(<vscale x 4 x i32> [[WIDE_MASKED_LOAD]])
+; NO-VP-NEXT:    [[TMP17:%.*]] = getelementptr i32, ptr [[PTR2:%.*]], i64 [[TMP8]]
+; NO-VP-NEXT:    [[TMP18:%.*]] = mul i64 0, [[TMP5]]
+; NO-VP-NEXT:    [[TMP19:%.*]] = sub i64 [[TMP5]], 1
+; NO-VP-NEXT:    [[TMP20:%.*]] = mul i64 -1, [[TMP19]]
+; NO-VP-NEXT:    [[TMP21:%.*]] = getelementptr i32, ptr [[TMP17]], i64 [[TMP18]]
+; NO-VP-NEXT:    [[TMP22:%.*]] = getelementptr i32, ptr [[TMP21]], i64 [[TMP20]]
+; NO-VP-NEXT:    [[REVERSE3:%.*]] = call <vscale x 4 x i1> @llvm.vector.reverse.nxv4i1(<vscale x 4 x i1> [[TMP10]])
+; NO-VP-NEXT:    [[REVERSE4:%.*]] = call <vscale x 4 x i32> @llvm.vector.reverse.nxv4i32(<vscale x 4 x i32> [[REVERSE2]])
+; NO-VP-NEXT:    call void @llvm.masked.store.nxv4i32.p0(<vscale x 4 x i32> [[REVERSE4]], ptr [[TMP22]], i32 4, <vscale x 4 x i1> [[REVERSE3]])
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
+; NO-VP-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP23]], label [[MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; NO-VP:       middle.block:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 1024, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label [[LOOPEND:%.*]], label [[ENTRY]]
+; NO-VP:       scalar.ph:
+; NO-VP-NEXT:    [[STARTVAL:%.*]] = phi i64 [ [[TMP6]], [[MIDDLE_BLOCK]] ], [ [[STARTVAL1]], [[ENTRY1:%.*]] ]
+; NO-VP-NEXT:    [[BC_RESUME_VAL5:%.*]] = phi i32 [ [[TMP7]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1]] ]
+; NO-VP-NEXT:    br label [[FOR_BODY1:%.*]]
 ; NO-VP:       for.body:
-; NO-VP-NEXT:    [[ADD_PHI:%.*]] = phi i64 [ [[STARTVAL:%.*]], [[ENTRY:%.*]] ], [ [[ADD:%.*]], [[FOR_INC:%.*]] ]
-; NO-VP-NEXT:    [[I:%.*]] = phi i32 [ 0, [[ENTRY]] ], [ [[INC:%.*]], [[FOR_INC]] ]
+; NO-VP-NEXT:    [[ADD_PHI:%.*]] = phi i64 [ [[STARTVAL]], [[ENTRY]] ], [ [[ADD:%.*]], [[FOR_INC:%.*]] ]
+; NO-VP-NEXT:    [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL5]], [[ENTRY]] ], [ [[INC:%.*]], [[FOR_INC]] ]
 ; NO-VP-NEXT:    [[ADD]] = add i64 [[ADD_PHI]], -1
-; NO-VP-NEXT:    [[GEPL:%.*]] = getelementptr inbounds i32, ptr [[PTR:%.*]], i32 [[I]]
+; NO-VP-NEXT:    [[GEPL:%.*]] = getelementptr inbounds i32, ptr [[PTR]], i32 [[I]]
 ; NO-VP-NEXT:    [[TMP:%.*]] = load i32, ptr [[GEPL]], align 4
 ; NO-VP-NEXT:    [[CMP1:%.*]] = icmp slt i32 [[TMP]], 100
 ; NO-VP-NEXT:    br i1 [[CMP1]], label [[IF_THEN:%.*]], label [[FOR_INC]]
 ; NO-VP:       if.then:
-; NO-VP-NEXT:    [[GEPL1:%.*]] = getelementptr inbounds i32, ptr [[PTR1:%.*]], i64 [[ADD]]
+; NO-VP-NEXT:    [[GEPL1:%.*]] = getelementptr inbounds i32, ptr [[PTR1]], i64 [[ADD]]
 ; NO-VP-NEXT:    [[V:%.*]] = load i32, ptr [[GEPL1]], align 4
-; NO-VP-NEXT:    [[GEPS:%.*]] = getelementptr inbounds i32, ptr [[PTR2:%.*]], i64 [[ADD]]
+; NO-VP-NEXT:    [[GEPS:%.*]] = getelementptr inbounds i32, ptr [[PTR2]], i64 [[ADD]]
 ; NO-VP-NEXT:    store i32 [[V]], ptr [[GEPS]], align 4
 ; NO-VP-NEXT:    br label [[FOR_INC]]
 ; NO-VP:       for.inc:
 ; NO-VP-NEXT:    [[INC]] = add i32 [[I]], 1
 ; NO-VP-NEXT:    [[EXITCOND:%.*]] = icmp ne i32 [[INC]], 1024
-; NO-VP-NEXT:    br i1 [[EXITCOND]], label [[FOR_BODY]], label [[LOOPEND:%.*]]
+; NO-VP-NEXT:    br i1 [[EXITCOND]], label [[FOR_BODY1]], label [[LOOPEND]], !llvm.loop [[LOOP5:![0-9]+]]
 ; NO-VP:       loopend:
 ; NO-VP-NEXT:    ret void
 ;
@@ -319,20 +408,49 @@ define void @multiple_reverse_vector_pointer(ptr noalias %a, ptr noalias %b, ptr
 ;
 ; NO-VP-LABEL: @multiple_reverse_vector_pointer(
 ; NO-VP-NEXT:  entry:
+; NO-VP-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; NO-VP:       vector.ph:
 ; NO-VP-NEXT:    br label [[LOOP:%.*]]
+; NO-VP:       vector.body:
+; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[LOOP]] ]
+; NO-VP-NEXT:    [[OFFSET_IDX:%.*]] = sub i64 1024, [[INDEX]]
+; NO-VP-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[OFFSET_IDX]]
+; NO-VP-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
+; NO-VP-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 -15
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; NO-VP-NEXT:    [[REVERSE:%.*]] = shufflevector <16 x i8> [[WIDE_LOAD]], <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; NO-VP-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[B:%.*]], <16 x i8> [[REVERSE]]
+; NO-VP-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <16 x i8> @llvm.masked.gather.v16i8.v16p0(<16 x ptr> [[TMP3]], i32 1, <16 x i1> splat (i1 true), <16 x i8> poison)
+; NO-VP-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[C:%.*]], i64 [[OFFSET_IDX]]
+; NO-VP-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP4]], i32 0
+; NO-VP-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[TMP5]], i32 -15
+; NO-VP-NEXT:    [[REVERSE1:%.*]] = shufflevector <16 x i8> [[WIDE_MASKED_GATHER]], <16 x i8> poison, <16 x i32> <i32 15, i32 14, i32 13, i32 12, i32 11, i32 10, i32 9, i32 8, i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
+; NO-VP-NEXT:    store <16 x i8> [[REVERSE1]], ptr [[TMP6]], align 1
+; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[D:%.*]], i64 [[OFFSET_IDX]]
+; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[TMP7]], i32 0
+; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[TMP8]], i32 -15
+; NO-VP-NEXT:    store <16 x i8> [[REVERSE1]], ptr [[TMP9]], align 1
+; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
+; NO-VP-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
+; NO-VP-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP]], !llvm.loop [[LOOP6:![0-9]+]]
+; NO-VP:       middle.block:
+; NO-VP-NEXT:    br label [[SCALAR_PH]]
+; NO-VP:       scalar.ph:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[MIDDLE_BLOCK]] ], [ 1024, [[ENTRY:%.*]] ]
+; NO-VP-NEXT:    br label [[LOOP1:%.*]]
 ; NO-VP:       loop:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 1024, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
-; NO-VP-NEXT:    [[GEP_A:%.*]] = getelementptr i8, ptr [[A:%.*]], i64 [[IV]]
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP1]] ]
+; NO-VP-NEXT:    [[GEP_A:%.*]] = getelementptr i8, ptr [[A]], i64 [[IV]]
 ; NO-VP-NEXT:    [[X:%.*]] = load i8, ptr [[GEP_A]], align 1
-; NO-VP-NEXT:    [[GEP_B:%.*]] = getelementptr i8, ptr [[B:%.*]], i8 [[X]]
+; NO-VP-NEXT:    [[GEP_B:%.*]] = getelementptr i8, ptr [[B]], i8 [[X]]
 ; NO-VP-NEXT:    [[Y:%.*]] = load i8, ptr [[GEP_B]], align 1
-; NO-VP-NEXT:    [[GEP_C:%.*]] = getelementptr i8, ptr [[C:%.*]], i64 [[IV]]
+; NO-VP-NEXT:    [[GEP_C:%.*]] = getelementptr i8, ptr [[C]], i64 [[IV]]
 ; NO-VP-NEXT:    store i8 [[Y]], ptr [[GEP_C]], align 1
-; NO-VP-NEXT:    [[GEP_D:%.*]] = getelementptr i8, ptr [[D:%.*]], i64 [[IV]]
+; NO-VP-NEXT:    [[GEP_D:%.*]] = getelementptr i8, ptr [[D]], i64 [[IV]]
 ; NO-VP-NEXT:    store i8 [[Y]], ptr [[GEP_D]], align 1
 ; NO-VP-NEXT:    [[IV_NEXT]] = add i64 [[IV]], -1
 ; NO-VP-NEXT:    [[CMP_NOT:%.*]] = icmp eq i64 [[IV]], 0
-; NO-VP-NEXT:    br i1 [[CMP_NOT]], label [[EXIT:%.*]], label [[LOOP]]
+; NO-VP-NEXT:    br i1 [[CMP_NOT]], label [[EXIT:%.*]], label [[LOOP1]], !llvm.loop [[LOOP7:![0-9]+]]
 ; NO-VP:       exit:
 ; NO-VP-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-safe-dep-distance.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-safe-dep-distance.ll
index 336c242..a891eef 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-safe-dep-distance.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-safe-dep-distance.ll
@@ -1,12 +1,10 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefix=IF-EVL
 
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck %s --check-prefix=NO-VP
 
 ; Dependence distance between read and write is greater than the trip
@@ -32,12 +30,10 @@ define void @test(ptr %p) {
 ; IF-EVL-NEXT:    [[TMP5:%.*]] = sub i64 200, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP6:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP5]], i32 2, i1 true)
 ; IF-EVL-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP9:%.*]] = getelementptr i64, ptr [[TMP8]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP9]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP6]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 8 [[TMP8]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP6]])
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = add i64 [[EVL_BASED_IV]], 200
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr i64, ptr [[P]], i64 [[TMP10]]
-; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr i64, ptr [[TMP11]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VP_OP_LOAD]], ptr align 8 [[TMP12]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP6]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VP_OP_LOAD]], ptr align 8 [[TMP11]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP6]])
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP6]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP13]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP4]]
@@ -63,17 +59,44 @@ define void @test(ptr %p) {
 ;
 ; NO-VP-LABEL: @test(
 ; NO-VP-NEXT:  entry:
+; NO-VP-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 2
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 200, [[TMP1]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
+; NO-VP:       vector.ph:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 2
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 200, [[TMP3]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 200, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2
 ; NO-VP-NEXT:    br label [[LOOP:%.*]]
-; NO-VP:       loop:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
+; NO-VP:       vector.body:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
 ; NO-VP-NEXT:    [[A1:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[IV]]
-; NO-VP-NEXT:    [[V:%.*]] = load i64, ptr [[A1]], align 8
-; NO-VP-NEXT:    [[OFFSET:%.*]] = add i64 [[IV]], 200
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[A1]], align 8
+; NO-VP-NEXT:    [[TMP7:%.*]] = add i64 [[IV]], 200
+; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[P]], i64 [[TMP7]]
+; NO-VP-NEXT:    store <vscale x 2 x i64> [[WIDE_LOAD]], ptr [[TMP8]], align 8
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw i64 [[IV]], [[TMP5]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP]], !llvm.loop [[LOOP0:![0-9]+]]
+; NO-VP:       middle.block:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 200, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; NO-VP:       scalar.ph:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
+; NO-VP-NEXT:    br label [[LOOP1:%.*]]
+; NO-VP:       loop:
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[LOOP1]] ]
+; NO-VP-NEXT:    [[A3:%.*]] = getelementptr i64, ptr [[P]], i64 [[IV1]]
+; NO-VP-NEXT:    [[V:%.*]] = load i64, ptr [[A3]], align 8
+; NO-VP-NEXT:    [[OFFSET:%.*]] = add i64 [[IV1]], 200
 ; NO-VP-NEXT:    [[A2:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET]]
 ; NO-VP-NEXT:    store i64 [[V]], ptr [[A2]], align 8
-; NO-VP-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
-; NO-VP-NEXT:    [[CMP:%.*]] = icmp ne i64 [[IV]], 199
-; NO-VP-NEXT:    br i1 [[CMP]], label [[LOOP]], label [[EXIT:%.*]]
+; NO-VP-NEXT:    [[IV_NEXT1]] = add i64 [[IV1]], 1
+; NO-VP-NEXT:    [[CMP:%.*]] = icmp ne i64 [[IV1]], 199
+; NO-VP-NEXT:    br i1 [[CMP]], label [[LOOP1]], label [[EXIT]], !llvm.loop [[LOOP3:![0-9]+]]
 ; NO-VP:       exit:
 ; NO-VP-NEXT:    ret void
 ;
@@ -106,19 +129,17 @@ define void @test_may_clobber1(ptr %p) {
 ; IF-EVL:       vector.body:
 ; IF-EVL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; IF-EVL-NEXT:    [[TMP1:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[INDEX]]
-; IF-EVL-NEXT:    [[TMP2:%.*]] = getelementptr i64, ptr [[TMP1]], i32 0
-; IF-EVL-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 32
+; IF-EVL-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 32
 ; IF-EVL-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 100
 ; IF-EVL-NEXT:    [[TMP4:%.*]] = getelementptr i64, ptr [[P]], i64 [[TMP3]]
-; IF-EVL-NEXT:    [[TMP5:%.*]] = getelementptr i64, ptr [[TMP4]], i32 0
-; IF-EVL-NEXT:    store <4 x i64> [[WIDE_LOAD]], ptr [[TMP5]], align 32
+; IF-EVL-NEXT:    store <4 x i64> [[WIDE_LOAD]], ptr [[TMP4]], align 32
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; IF-EVL-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 200
 ; IF-EVL-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; IF-EVL:       middle.block:
-; IF-EVL-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; IF-EVL-NEXT:    br label [[EXIT:%.*]]
 ; IF-EVL:       scalar.ph:
-; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 200, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; IF-EVL-NEXT:    br label [[LOOP:%.*]]
 ; IF-EVL:       loop:
 ; IF-EVL-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -141,19 +162,17 @@ define void @test_may_clobber1(ptr %p) {
 ; NO-VP:       vector.body:
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP1:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP2:%.*]] = getelementptr i64, ptr [[TMP1]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 32
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 32
 ; NO-VP-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 100
 ; NO-VP-NEXT:    [[TMP4:%.*]] = getelementptr i64, ptr [[P]], i64 [[TMP3]]
-; NO-VP-NEXT:    [[TMP5:%.*]] = getelementptr i64, ptr [[TMP4]], i32 0
-; NO-VP-NEXT:    store <4 x i64> [[WIDE_LOAD]], ptr [[TMP5]], align 32
+; NO-VP-NEXT:    store <4 x i64> [[WIDE_LOAD]], ptr [[TMP4]], align 32
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; NO-VP-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 200
-; NO-VP-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; NO-VP-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; NO-VP:       middle.block:
-; NO-VP-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; NO-VP-NEXT:    br label [[EXIT:%.*]]
 ; NO-VP:       scalar.ph:
-; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 200, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; NO-VP-NEXT:    br label [[LOOP:%.*]]
 ; NO-VP:       loop:
 ; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -164,7 +183,7 @@ define void @test_may_clobber1(ptr %p) {
 ; NO-VP-NEXT:    store i64 [[V]], ptr [[A2]], align 32
 ; NO-VP-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; NO-VP-NEXT:    [[CMP:%.*]] = icmp ne i64 [[IV]], 199
-; NO-VP-NEXT:    br i1 [[CMP]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP3:![0-9]+]]
+; NO-VP-NEXT:    br i1 [[CMP]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP5:![0-9]+]]
 ; NO-VP:       exit:
 ; NO-VP-NEXT:    ret void
 ;
@@ -246,19 +265,17 @@ define void @test_may_clobber3(ptr %p) {
 ; IF-EVL:       vector.body:
 ; IF-EVL-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; IF-EVL-NEXT:    [[TMP1:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[INDEX]]
-; IF-EVL-NEXT:    [[TMP2:%.*]] = getelementptr i64, ptr [[TMP1]], i32 0
-; IF-EVL-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 32
+; IF-EVL-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 32
 ; IF-EVL-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 10
 ; IF-EVL-NEXT:    [[TMP4:%.*]] = getelementptr i64, ptr [[P]], i64 [[TMP3]]
-; IF-EVL-NEXT:    [[TMP5:%.*]] = getelementptr i64, ptr [[TMP4]], i32 0
-; IF-EVL-NEXT:    store <2 x i64> [[WIDE_LOAD]], ptr [[TMP5]], align 32
+; IF-EVL-NEXT:    store <2 x i64> [[WIDE_LOAD]], ptr [[TMP4]], align 32
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; IF-EVL-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 200
 ; IF-EVL-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
 ; IF-EVL:       middle.block:
-; IF-EVL-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; IF-EVL-NEXT:    br label [[EXIT:%.*]]
 ; IF-EVL:       scalar.ph:
-; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 200, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; IF-EVL-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; IF-EVL-NEXT:    br label [[LOOP:%.*]]
 ; IF-EVL:       loop:
 ; IF-EVL-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -281,19 +298,17 @@ define void @test_may_clobber3(ptr %p) {
 ; NO-VP:       vector.body:
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP1:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP2:%.*]] = getelementptr i64, ptr [[TMP1]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 32
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 32
 ; NO-VP-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 10
 ; NO-VP-NEXT:    [[TMP4:%.*]] = getelementptr i64, ptr [[P]], i64 [[TMP3]]
-; NO-VP-NEXT:    [[TMP5:%.*]] = getelementptr i64, ptr [[TMP4]], i32 0
-; NO-VP-NEXT:    store <2 x i64> [[WIDE_LOAD]], ptr [[TMP5]], align 32
+; NO-VP-NEXT:    store <2 x i64> [[WIDE_LOAD]], ptr [[TMP4]], align 32
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; NO-VP-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 200
-; NO-VP-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; NO-VP-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; NO-VP:       middle.block:
-; NO-VP-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; NO-VP-NEXT:    br label [[EXIT:%.*]]
 ; NO-VP:       scalar.ph:
-; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 200, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; NO-VP-NEXT:    br label [[LOOP:%.*]]
 ; NO-VP:       loop:
 ; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -304,7 +319,7 @@ define void @test_may_clobber3(ptr %p) {
 ; NO-VP-NEXT:    store i64 [[V]], ptr [[A2]], align 32
 ; NO-VP-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; NO-VP-NEXT:    [[CMP:%.*]] = icmp ne i64 [[IV]], 199
-; NO-VP-NEXT:    br i1 [[CMP]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP5:![0-9]+]]
+; NO-VP-NEXT:    br i1 [[CMP]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP7:![0-9]+]]
 ; NO-VP:       exit:
 ; NO-VP-NEXT:    ret void
 ;
@@ -347,12 +362,10 @@ define void @trivial_due_max_vscale(ptr %p) {
 ; IF-EVL-NEXT:    [[TMP5:%.*]] = sub i64 200, [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP6:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP5]], i32 2, i1 true)
 ; IF-EVL-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP9:%.*]] = getelementptr i64, ptr [[TMP8]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 32 [[TMP9]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP6]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 2 x i64> @llvm.vp.load.nxv2i64.p0(ptr align 32 [[TMP8]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP6]])
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = add i64 [[EVL_BASED_IV]], 8192
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = getelementptr i64, ptr [[P]], i64 [[TMP10]]
-; IF-EVL-NEXT:    [[TMP12:%.*]] = getelementptr i64, ptr [[TMP11]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VP_OP_LOAD]], ptr align 32 [[TMP12]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP6]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[VP_OP_LOAD]], ptr align 32 [[TMP11]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP6]])
 ; IF-EVL-NEXT:    [[TMP13:%.*]] = zext i32 [[TMP6]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP13]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP4]]
@@ -378,17 +391,44 @@ define void @trivial_due_max_vscale(ptr %p) {
 ;
 ; NO-VP-LABEL: @trivial_due_max_vscale(
 ; NO-VP-NEXT:  entry:
+; NO-VP-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 2
+; NO-VP-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 200, [[TMP1]]
+; NO-VP-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
+; NO-VP:       vector.ph:
+; NO-VP-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 2
+; NO-VP-NEXT:    [[N_MOD_VF:%.*]] = urem i64 200, [[TMP3]]
+; NO-VP-NEXT:    [[N_VEC:%.*]] = sub i64 200, [[N_MOD_VF]]
+; NO-VP-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; NO-VP-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2
 ; NO-VP-NEXT:    br label [[LOOP:%.*]]
-; NO-VP:       loop:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
+; NO-VP:       vector.body:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
 ; NO-VP-NEXT:    [[A1:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[IV]]
-; NO-VP-NEXT:    [[V:%.*]] = load i64, ptr [[A1]], align 32
-; NO-VP-NEXT:    [[OFFSET:%.*]] = add i64 [[IV]], 8192
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 2 x i64>, ptr [[A1]], align 32
+; NO-VP-NEXT:    [[TMP7:%.*]] = add i64 [[IV]], 8192
+; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[P]], i64 [[TMP7]]
+; NO-VP-NEXT:    store <vscale x 2 x i64> [[WIDE_LOAD]], ptr [[TMP8]], align 32
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw i64 [[IV]], [[TMP5]]
+; NO-VP-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[IV_NEXT]], [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP]], !llvm.loop [[LOOP8:![0-9]+]]
+; NO-VP:       middle.block:
+; NO-VP-NEXT:    [[CMP_N:%.*]] = icmp eq i64 200, [[N_VEC]]
+; NO-VP-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
+; NO-VP:       scalar.ph:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
+; NO-VP-NEXT:    br label [[LOOP1:%.*]]
+; NO-VP:       loop:
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[LOOP1]] ]
+; NO-VP-NEXT:    [[A3:%.*]] = getelementptr i64, ptr [[P]], i64 [[IV1]]
+; NO-VP-NEXT:    [[V:%.*]] = load i64, ptr [[A3]], align 32
+; NO-VP-NEXT:    [[OFFSET:%.*]] = add i64 [[IV1]], 8192
 ; NO-VP-NEXT:    [[A2:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET]]
 ; NO-VP-NEXT:    store i64 [[V]], ptr [[A2]], align 32
-; NO-VP-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
-; NO-VP-NEXT:    [[CMP:%.*]] = icmp ne i64 [[IV]], 199
-; NO-VP-NEXT:    br i1 [[CMP]], label [[LOOP]], label [[EXIT:%.*]]
+; NO-VP-NEXT:    [[IV_NEXT1]] = add i64 [[IV1]], 1
+; NO-VP-NEXT:    [[CMP:%.*]] = icmp ne i64 [[IV1]], 199
+; NO-VP-NEXT:    br i1 [[CMP]], label [[LOOP1]], label [[EXIT]], !llvm.loop [[LOOP9:![0-9]+]]
 ; NO-VP:       exit:
 ; NO-VP-NEXT:    ret void
 ;
@@ -431,12 +471,10 @@ define void @no_high_lmul_or_interleave(ptr %p) {
 ; IF-EVL-NEXT:    [[SAFE_AVL:%.*]] = select i1 [[TMP9]], i64 [[AVL]], i64 1024
 ; IF-EVL-NEXT:    [[TMP10:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[SAFE_AVL]], i32 1, i1 true)
 ; IF-EVL-NEXT:    [[TMP2:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP3:%.*]] = getelementptr i64, ptr [[TMP2]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 1 x i64> @llvm.vp.load.nxv1i64.p0(ptr align 32 [[TMP3]], <vscale x 1 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 1 x i64> @llvm.vp.load.nxv1i64.p0(ptr align 32 [[TMP2]], <vscale x 1 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP4:%.*]] = add i64 [[EVL_BASED_IV]], 1024
 ; IF-EVL-NEXT:    [[TMP5:%.*]] = getelementptr i64, ptr [[P]], i64 [[TMP4]]
-; IF-EVL-NEXT:    [[TMP6:%.*]] = getelementptr i64, ptr [[TMP5]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv1i64.p0(<vscale x 1 x i64> [[VP_OP_LOAD]], ptr align 32 [[TMP6]], <vscale x 1 x i1> splat (i1 true), i32 [[TMP10]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv1i64.p0(<vscale x 1 x i64> [[VP_OP_LOAD]], ptr align 32 [[TMP5]], <vscale x 1 x i1> splat (i1 true), i32 [[TMP10]])
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP10]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP11]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
@@ -462,17 +500,34 @@ define void @no_high_lmul_or_interleave(ptr %p) {
 ;
 ; NO-VP-LABEL: @no_high_lmul_or_interleave(
 ; NO-VP-NEXT:  entry:
+; NO-VP-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[ENTRY:%.*]]
+; NO-VP:       vector.ph:
 ; NO-VP-NEXT:    br label [[LOOP:%.*]]
-; NO-VP:       loop:
-; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
+; NO-VP:       vector.body:
+; NO-VP-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
 ; NO-VP-NEXT:    [[A1:%.*]] = getelementptr i64, ptr [[P:%.*]], i64 [[IV]]
-; NO-VP-NEXT:    [[V:%.*]] = load i64, ptr [[A1]], align 32
-; NO-VP-NEXT:    [[OFFSET:%.*]] = add i64 [[IV]], 1024
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[A1]], align 32
+; NO-VP-NEXT:    [[TMP1:%.*]] = add i64 [[IV]], 1024
+; NO-VP-NEXT:    [[TMP2:%.*]] = getelementptr i64, ptr [[P]], i64 [[TMP1]]
+; NO-VP-NEXT:    store <4 x i64> [[WIDE_LOAD]], ptr [[TMP2]], align 32
+; NO-VP-NEXT:    [[IV_NEXT]] = add nuw i64 [[IV]], 4
+; NO-VP-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[IV_NEXT]], 3000
+; NO-VP-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP]], !llvm.loop [[LOOP10:![0-9]+]]
+; NO-VP:       middle.block:
+; NO-VP-NEXT:    br label [[SCALAR_PH]]
+; NO-VP:       scalar.ph:
+; NO-VP-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY1:%.*]] ]
+; NO-VP-NEXT:    br label [[LOOP1:%.*]]
+; NO-VP:       loop:
+; NO-VP-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], [[LOOP1]] ]
+; NO-VP-NEXT:    [[A3:%.*]] = getelementptr i64, ptr [[P]], i64 [[IV1]]
+; NO-VP-NEXT:    [[V:%.*]] = load i64, ptr [[A3]], align 32
+; NO-VP-NEXT:    [[OFFSET:%.*]] = add i64 [[IV1]], 1024
 ; NO-VP-NEXT:    [[A2:%.*]] = getelementptr i64, ptr [[P]], i64 [[OFFSET]]
 ; NO-VP-NEXT:    store i64 [[V]], ptr [[A2]], align 32
-; NO-VP-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
-; NO-VP-NEXT:    [[CMP:%.*]] = icmp ne i64 [[IV]], 3001
-; NO-VP-NEXT:    br i1 [[CMP]], label [[LOOP]], label [[EXIT:%.*]]
+; NO-VP-NEXT:    [[IV_NEXT1]] = add i64 [[IV1]], 1
+; NO-VP-NEXT:    [[CMP:%.*]] = icmp ne i64 [[IV1]], 3001
+; NO-VP-NEXT:    br i1 [[CMP]], label [[LOOP1]], label [[EXIT:%.*]], !llvm.loop [[LOOP11:![0-9]+]]
 ; NO-VP:       exit:
 ; NO-VP-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-uniform-store.ll b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-uniform-store.ll
index 984b64c..8ab3817 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-force-tail-with-evl-uniform-store.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/tail-folding-uniform-store.ll
@@ -1,5 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 5
-; RUN: opt < %s --prefer-predicate-over-epilogue=predicate-dont-vectorize --passes=loop-vectorize -mcpu=sifive-p470 -mattr=+v,+f -force-tail-folding-style=data-with-evl -S | FileCheck %s
+; RUN: opt < %s --prefer-predicate-over-epilogue=predicate-else-scalar-epilogue --passes=loop-vectorize -mcpu=sifive-p470 -mattr=+v,+f -S | FileCheck %s
 ; Generated from issue #109468.
 ; In this test case, the vector store with tail mask will transfer to the vp intrinsic with EVL.
 
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/truncate-to-minimal-bitwidth-cost.ll b/llvm/test/Transforms/LoopVectorize/RISCV/truncate-to-minimal-bitwidth-cost.ll
index b56e712..79ec73a 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/truncate-to-minimal-bitwidth-cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/truncate-to-minimal-bitwidth-cost.ll
@@ -26,13 +26,11 @@ define void @test_pr98413_zext_removed(ptr %src, ptr noalias %dst, i64 %x) {
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP7:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[SRC]], i64 [[TMP7]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i16, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i16>, ptr [[TMP9]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i16>, ptr [[TMP8]], align 8
 ; CHECK-NEXT:    [[TMP10:%.*]] = trunc <vscale x 8 x i16> [[WIDE_LOAD]] to <vscale x 8 x i8>
 ; CHECK-NEXT:    [[TMP11:%.*]] = and <vscale x 8 x i8> [[TMP6]], [[TMP10]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[TMP7]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[TMP12]], i32 0
-; CHECK-NEXT:    store <vscale x 8 x i8> [[TMP11]], ptr [[TMP13]], align 1
+; CHECK-NEXT:    store <vscale x 8 x i8> [[TMP11]], ptr [[TMP12]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP7]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -99,13 +97,11 @@ define void @test_pr98413_sext_removed(ptr %src, ptr noalias %dst, i64 %x) {
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP7:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[SRC]], i64 [[TMP7]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i16, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i16>, ptr [[TMP9]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i16>, ptr [[TMP8]], align 8
 ; CHECK-NEXT:    [[TMP10:%.*]] = trunc <vscale x 8 x i16> [[WIDE_LOAD]] to <vscale x 8 x i8>
 ; CHECK-NEXT:    [[TMP11:%.*]] = and <vscale x 8 x i8> [[TMP6]], [[TMP10]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[TMP7]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[TMP12]], i32 0
-; CHECK-NEXT:    store <vscale x 8 x i8> [[TMP11]], ptr [[TMP13]], align 1
+; CHECK-NEXT:    store <vscale x 8 x i8> [[TMP11]], ptr [[TMP12]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP7]], [[TMP5]]
 ; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/truncate-to-minimal-bitwidth-evl-crash.ll b/llvm/test/Transforms/LoopVectorize/RISCV/truncate-to-minimal-bitwidth-evl-crash.ll
index 72afff2..27d7bd0 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/truncate-to-minimal-bitwidth-evl-crash.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/truncate-to-minimal-bitwidth-evl-crash.ll
@@ -1,5 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
-; RUN: opt -passes=loop-vectorize -force-tail-folding-style=data-with-evl -prefer-predicate-over-epilogue=predicate-dont-vectorize -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s
+; RUN: opt -passes=loop-vectorize -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s
 
 ; Make sure we don't crash when transforming a VPWidenCastRecipe created without
 ; an underlying value to an EVL recipe. This occurs in this test via
@@ -26,8 +26,7 @@ define void @truncate_to_minimal_bitwidths_widen_cast_recipe(ptr %src) {
 ; CHECK-NEXT:    [[AVL:%.*]] = sub i64 9, [[EVL_BASED_IV]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 8, i1 true)
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[SRC]], i64 [[EVL_BASED_IV]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 8 x i8> @llvm.vp.load.nxv8i8.p0(ptr align 1 [[TMP6]], <vscale x 8 x i1> splat (i1 true), i32 [[TMP7]])
+; CHECK-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 8 x i8> @llvm.vp.load.nxv8i8.p0(ptr align 1 [[TMP5]], <vscale x 8 x i1> splat (i1 true), i32 [[TMP7]])
 ; CHECK-NEXT:    [[TMP8:%.*]] = zext <vscale x 8 x i8> [[VP_OP_LOAD]] to <vscale x 8 x i16>
 ; CHECK-NEXT:    [[TMP12:%.*]] = mul <vscale x 8 x i16> zeroinitializer, [[TMP8]]
 ; CHECK-NEXT:    [[TMP13:%.*]] = lshr <vscale x 8 x i16> [[TMP12]], splat (i16 1)
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/type-info-cache-evl-crash.ll b/llvm/test/Transforms/LoopVectorize/RISCV/type-info-cache-evl-crash.ll
index 1e1ed49..27abddf 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/type-info-cache-evl-crash.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/type-info-cache-evl-crash.ll
@@ -1,5 +1,5 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
-; RUN: opt -passes=loop-vectorize -force-tail-folding-style=data-with-evl -prefer-predicate-over-epilogue=predicate-dont-vectorize -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s
+; RUN: opt -passes=loop-vectorize -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue -mtriple=riscv64 -mattr=+v -S %s | FileCheck %s
 
 ; This test tries to recreate the conditions for a crash that occurred when the
 ; VPTypeAnalysis cache wasn't cleared after a recipe was erased and clobbered
@@ -37,8 +37,7 @@ define void @type_info_cache_clobber(ptr %dstv, ptr %src, i64 %wide.trip.count)
 ; CHECK-NEXT:    [[AVL:%.*]] = sub i64 [[TMP0]], [[EVL_BASED_IV]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 8, i1 true)
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr i8, ptr [[SRC]], i64 [[EVL_BASED_IV]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[TMP13]], i32 0
-; CHECK-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 8 x i8> @llvm.vp.load.nxv8i8.p0(ptr align 1 [[TMP14]], <vscale x 8 x i1> splat (i1 true), i32 [[TMP11]]), !alias.scope [[META0:![0-9]+]]
+; CHECK-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 8 x i8> @llvm.vp.load.nxv8i8.p0(ptr align 1 [[TMP13]], <vscale x 8 x i1> splat (i1 true), i32 [[TMP11]]), !alias.scope [[META0:![0-9]+]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = zext <vscale x 8 x i8> [[VP_OP_LOAD]] to <vscale x 8 x i32>
 ; CHECK-NEXT:    [[VP_OP:%.*]] = mul <vscale x 8 x i32> [[TMP15]], zeroinitializer
 ; CHECK-NEXT:    [[TMP23:%.*]] = ashr <vscale x 8 x i32> [[TMP15]], zeroinitializer
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/uniform-load-store.ll b/llvm/test/Transforms/LoopVectorize/RISCV/uniform-load-store.ll
index c6955f1..d3c2ccf 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/uniform-load-store.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/uniform-load-store.ll
@@ -1,7 +1,7 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 5
 ; RUN: opt < %s -passes=loop-vectorize -scalable-vectorization=on -riscv-v-vector-bits-min=0 -mtriple riscv64-linux-gnu -mattr=+v,+f -S 2>%t | FileCheck %s -check-prefix=SCALABLE
 ; RUN: opt < %s -passes=loop-vectorize -scalable-vectorization=off -mtriple riscv64-linux-gnu -mattr=+v,+f -S 2>%t | FileCheck %s -check-prefix=FIXEDLEN
-; RUN: opt < %s -passes=loop-vectorize -scalable-vectorization=on -riscv-v-vector-bits-min=0 -prefer-predicate-over-epilogue=predicate-dont-vectorize -mtriple riscv64-linux-gnu -mattr=+v,+f -S 2>%t | FileCheck %s -check-prefix=TF-SCALABLE
+; RUN: opt < %s -passes=loop-vectorize -scalable-vectorization=on -riscv-v-vector-bits-min=0 -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue -mtriple riscv64-linux-gnu -mattr=+v,+f -S 2>%t | FileCheck %s -check-prefix=TF-SCALABLE
 
 target datalayout = "e-m:e-p:64:64-i64:64-i128:128-n64-S128"
 target triple = "riscv64"
@@ -28,8 +28,7 @@ define void @uniform_load(ptr noalias nocapture %a, ptr noalias nocapture %b, i6
 ; SCALABLE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP7]], i64 0
 ; SCALABLE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
 ; SCALABLE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; SCALABLE-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP8]], i32 0
-; SCALABLE-NEXT:    store <vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP9]], align 8
+; SCALABLE-NEXT:    store <vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP8]], align 8
 ; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; SCALABLE-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; SCALABLE-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -62,15 +61,14 @@ define void @uniform_load(ptr noalias nocapture %a, ptr noalias nocapture %b, i6
 ; FIXEDLEN-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[TMP1]], i64 0
 ; FIXEDLEN-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
 ; FIXEDLEN-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; FIXEDLEN-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
 ; FIXEDLEN-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 4
-; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP3]], align 8
+; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP2]], align 8
 ; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP4]], align 8
 ; FIXEDLEN-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FIXEDLEN-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXEDLEN-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; FIXEDLEN:       [[MIDDLE_BLOCK]]:
-; FIXEDLEN-NEXT:    br i1 false, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; FIXEDLEN-NEXT:    br label %[[SCALAR_PH]]
 ; FIXEDLEN:       [[SCALAR_PH]]:
 ; FIXEDLEN-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; FIXEDLEN-NEXT:    br label %[[FOR_BODY:.*]]
@@ -81,7 +79,7 @@ define void @uniform_load(ptr noalias nocapture %a, ptr noalias nocapture %b, i6
 ; FIXEDLEN-NEXT:    store i64 [[V]], ptr [[ARRAYIDX]], align 8
 ; FIXEDLEN-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; FIXEDLEN-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025
-; FIXEDLEN-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
+; FIXEDLEN-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; FIXEDLEN:       [[FOR_END]]:
 ; FIXEDLEN-NEXT:    ret void
 ;
@@ -108,10 +106,9 @@ define void @uniform_load(ptr noalias nocapture %a, ptr noalias nocapture %b, i6
 ; TF-SCALABLE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP5]], i64 0
 ; TF-SCALABLE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
 ; TF-SCALABLE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; TF-SCALABLE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP7]], i32 0
-; TF-SCALABLE-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr align 8 [[TMP8]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP6]])
-; TF-SCALABLE-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP6]] to i64
-; TF-SCALABLE-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP10]], [[INDEX]]
+; TF-SCALABLE-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr align 8 [[TMP7]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP6]])
+; TF-SCALABLE-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP6]] to i64
+; TF-SCALABLE-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP8]], [[INDEX]]
 ; TF-SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX1]], [[TMP4]]
 ; TF-SCALABLE-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; TF-SCALABLE-NEXT:    br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -169,8 +166,7 @@ define i64 @uniform_load_outside_use(ptr noalias nocapture %a, ptr noalias nocap
 ; SCALABLE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP6]], i64 0
 ; SCALABLE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
 ; SCALABLE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; SCALABLE-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP8]], i32 0
-; SCALABLE-NEXT:    store <vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP9]], align 8
+; SCALABLE-NEXT:    store <vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP8]], align 8
 ; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; SCALABLE-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; SCALABLE-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -204,15 +200,14 @@ define i64 @uniform_load_outside_use(ptr noalias nocapture %a, ptr noalias nocap
 ; FIXEDLEN-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[TMP1]], i64 0
 ; FIXEDLEN-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
 ; FIXEDLEN-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; FIXEDLEN-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
 ; FIXEDLEN-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 4
-; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP3]], align 8
+; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP2]], align 8
 ; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP4]], align 8
 ; FIXEDLEN-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FIXEDLEN-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXEDLEN-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; FIXEDLEN:       [[MIDDLE_BLOCK]]:
-; FIXEDLEN-NEXT:    br i1 false, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; FIXEDLEN-NEXT:    br label %[[SCALAR_PH]]
 ; FIXEDLEN:       [[SCALAR_PH]]:
 ; FIXEDLEN-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; FIXEDLEN-NEXT:    br label %[[FOR_BODY:.*]]
@@ -223,25 +218,52 @@ define i64 @uniform_load_outside_use(ptr noalias nocapture %a, ptr noalias nocap
 ; FIXEDLEN-NEXT:    store i64 [[V]], ptr [[ARRAYIDX]], align 8
 ; FIXEDLEN-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; FIXEDLEN-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025
-; FIXEDLEN-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; FIXEDLEN-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; FIXEDLEN:       [[FOR_END]]:
-; FIXEDLEN-NEXT:    [[V_LCSSA:%.*]] = phi i64 [ [[V]], %[[FOR_BODY]] ], [ [[TMP1]], %[[MIDDLE_BLOCK]] ]
+; FIXEDLEN-NEXT:    [[V_LCSSA:%.*]] = phi i64 [ [[V]], %[[FOR_BODY]] ]
 ; FIXEDLEN-NEXT:    ret i64 [[V_LCSSA]]
 ;
 ; TF-SCALABLE-LABEL: define i64 @uniform_load_outside_use(
 ; TF-SCALABLE-SAME: ptr noalias captures(none) [[A:%.*]], ptr noalias captures(none) [[B:%.*]], i64 [[N:%.*]]) #[[ATTR0]] {
 ; TF-SCALABLE-NEXT:  [[ENTRY:.*]]:
-; TF-SCALABLE-NEXT:    br label %[[FOR_BODY:.*]]
-; TF-SCALABLE:       [[FOR_BODY]]:
-; TF-SCALABLE-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[FOR_BODY]] ]
+; TF-SCALABLE-NEXT:    [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
+; TF-SCALABLE-NEXT:    [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 2
+; TF-SCALABLE-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 1025, [[TMP1]]
+; TF-SCALABLE-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; TF-SCALABLE:       [[VECTOR_PH]]:
+; TF-SCALABLE-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
+; TF-SCALABLE-NEXT:    [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 2
+; TF-SCALABLE-NEXT:    [[N_MOD_VF:%.*]] = urem i64 1025, [[TMP3]]
+; TF-SCALABLE-NEXT:    [[N_VEC:%.*]] = sub i64 1025, [[N_MOD_VF]]
+; TF-SCALABLE-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
+; TF-SCALABLE-NEXT:    [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2
+; TF-SCALABLE-NEXT:    br label %[[VECTOR_BODY:.*]]
+; TF-SCALABLE:       [[VECTOR_BODY]]:
+; TF-SCALABLE-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; TF-SCALABLE-NEXT:    [[V:%.*]] = load i64, ptr [[B]], align 8
+; TF-SCALABLE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[V]], i64 0
+; TF-SCALABLE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
 ; TF-SCALABLE-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
-; TF-SCALABLE-NEXT:    store i64 [[V]], ptr [[ARRAYIDX]], align 8
-; TF-SCALABLE-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
+; TF-SCALABLE-NEXT:    store <vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[ARRAYIDX]], align 8
+; TF-SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], [[TMP5]]
+; TF-SCALABLE-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; TF-SCALABLE-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; TF-SCALABLE:       [[MIDDLE_BLOCK]]:
+; TF-SCALABLE-NEXT:    [[CMP_N:%.*]] = icmp eq i64 1025, [[N_VEC]]
+; TF-SCALABLE-NEXT:    br i1 [[CMP_N]], label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; TF-SCALABLE:       [[SCALAR_PH]]:
+; TF-SCALABLE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; TF-SCALABLE-NEXT:    br label %[[FOR_BODY:.*]]
+; TF-SCALABLE:       [[FOR_BODY]]:
+; TF-SCALABLE-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[FOR_BODY]] ]
+; TF-SCALABLE-NEXT:    [[V1:%.*]] = load i64, ptr [[B]], align 8
+; TF-SCALABLE-NEXT:    [[ARRAYIDX1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV1]]
+; TF-SCALABLE-NEXT:    store i64 [[V1]], ptr [[ARRAYIDX1]], align 8
+; TF-SCALABLE-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV1]], 1
 ; TF-SCALABLE-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025
-; TF-SCALABLE-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END:.*]], label %[[FOR_BODY]]
+; TF-SCALABLE-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; TF-SCALABLE:       [[FOR_END]]:
-; TF-SCALABLE-NEXT:    [[V_LCSSA:%.*]] = phi i64 [ [[V]], %[[FOR_BODY]] ]
+; TF-SCALABLE-NEXT:    [[V_LCSSA:%.*]] = phi i64 [ [[V1]], %[[FOR_BODY]] ], [ [[V]], %[[MIDDLE_BLOCK]] ]
 ; TF-SCALABLE-NEXT:    ret i64 [[V_LCSSA]]
 ;
 entry:
@@ -292,8 +314,7 @@ define void @conditional_uniform_load(ptr noalias nocapture %a, ptr noalias noca
 ; SCALABLE-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 4 x i64> @llvm.masked.gather.nxv4i64.nxv4p0(<vscale x 4 x ptr> [[BROADCAST_SPLAT]], i32 8, <vscale x 4 x i1> [[TMP10]], <vscale x 4 x i64> poison)
 ; SCALABLE-NEXT:    [[PREDPHI:%.*]] = select <vscale x 4 x i1> [[TMP10]], <vscale x 4 x i64> [[WIDE_MASKED_GATHER]], <vscale x 4 x i64> zeroinitializer
 ; SCALABLE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; SCALABLE-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i64, ptr [[TMP12]], i32 0
-; SCALABLE-NEXT:    store <vscale x 4 x i64> [[PREDPHI]], ptr [[TMP13]], align 8
+; SCALABLE-NEXT:    store <vscale x 4 x i64> [[PREDPHI]], ptr [[TMP12]], align 8
 ; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; SCALABLE-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 4 x i64> [[VEC_IND]], [[DOTSPLAT]]
 ; SCALABLE-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -340,16 +361,15 @@ define void @conditional_uniform_load(ptr noalias nocapture %a, ptr noalias noca
 ; FIXEDLEN-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP1]], <4 x i64> [[WIDE_MASKED_GATHER]], <4 x i64> zeroinitializer
 ; FIXEDLEN-NEXT:    [[PREDPHI2:%.*]] = select <4 x i1> [[TMP2]], <4 x i64> [[WIDE_MASKED_GATHER1]], <4 x i64> zeroinitializer
 ; FIXEDLEN-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; FIXEDLEN-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP3]], i32 0
 ; FIXEDLEN-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP3]], i32 4
-; FIXEDLEN-NEXT:    store <4 x i64> [[PREDPHI]], ptr [[TMP4]], align 8
+; FIXEDLEN-NEXT:    store <4 x i64> [[PREDPHI]], ptr [[TMP3]], align 8
 ; FIXEDLEN-NEXT:    store <4 x i64> [[PREDPHI2]], ptr [[TMP5]], align 8
 ; FIXEDLEN-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FIXEDLEN-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[STEP_ADD]], splat (i64 4)
 ; FIXEDLEN-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXEDLEN-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; FIXEDLEN:       [[MIDDLE_BLOCK]]:
-; FIXEDLEN-NEXT:    br i1 false, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; FIXEDLEN-NEXT:    br label %[[SCALAR_PH]]
 ; FIXEDLEN:       [[SCALAR_PH]]:
 ; FIXEDLEN-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; FIXEDLEN-NEXT:    br label %[[FOR_BODY:.*]]
@@ -366,7 +386,7 @@ define void @conditional_uniform_load(ptr noalias nocapture %a, ptr noalias noca
 ; FIXEDLEN-NEXT:    store i64 [[PHI]], ptr [[ARRAYIDX]], align 8
 ; FIXEDLEN-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; FIXEDLEN-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025
-; FIXEDLEN-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
+; FIXEDLEN-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
 ; FIXEDLEN:       [[FOR_END]]:
 ; FIXEDLEN-NEXT:    ret void
 ;
@@ -405,14 +425,13 @@ define void @conditional_uniform_load(ptr noalias nocapture %a, ptr noalias noca
 ; TF-SCALABLE-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <vscale x 4 x i64> @llvm.vp.gather.nxv4i64.nxv4p0(<vscale x 4 x ptr> align 8 [[BROADCAST_SPLAT]], <vscale x 4 x i1> [[TMP10]], i32 [[TMP7]])
 ; TF-SCALABLE-NEXT:    [[PREDPHI:%.*]] = select <vscale x 4 x i1> [[TMP9]], <vscale x 4 x i64> [[WIDE_MASKED_GATHER]], <vscale x 4 x i64> zeroinitializer
 ; TF-SCALABLE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; TF-SCALABLE-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i64, ptr [[TMP12]], i32 0
-; TF-SCALABLE-NEXT:    call void @llvm.vp.store.nxv4i64.p0(<vscale x 4 x i64> [[PREDPHI]], ptr align 8 [[TMP13]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP7]])
+; TF-SCALABLE-NEXT:    call void @llvm.vp.store.nxv4i64.p0(<vscale x 4 x i64> [[PREDPHI]], ptr align 8 [[TMP12]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP7]])
 ; TF-SCALABLE-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP7]] to i64
 ; TF-SCALABLE-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP15]], [[INDEX]]
 ; TF-SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX1]], [[TMP4]]
 ; TF-SCALABLE-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 4 x i64> [[VEC_IND]], [[DOTSPLAT]]
 ; TF-SCALABLE-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; TF-SCALABLE-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; TF-SCALABLE-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
 ; TF-SCALABLE:       [[MIDDLE_BLOCK]]:
 ; TF-SCALABLE-NEXT:    br label %[[FOR_END:.*]]
 ; TF-SCALABLE:       [[SCALAR_PH]]:
@@ -431,7 +450,7 @@ define void @conditional_uniform_load(ptr noalias nocapture %a, ptr noalias noca
 ; TF-SCALABLE-NEXT:    store i64 [[PHI]], ptr [[ARRAYIDX]], align 8
 ; TF-SCALABLE-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; TF-SCALABLE-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025
-; TF-SCALABLE-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; TF-SCALABLE-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; TF-SCALABLE:       [[FOR_END]]:
 ; TF-SCALABLE-NEXT:    ret void
 ;
@@ -480,8 +499,7 @@ define void @uniform_load_unaligned(ptr noalias nocapture %a, ptr noalias nocapt
 ; SCALABLE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP6]], i64 0
 ; SCALABLE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
 ; SCALABLE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; SCALABLE-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP8]], i32 0
-; SCALABLE-NEXT:    store <vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP9]], align 8
+; SCALABLE-NEXT:    store <vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP8]], align 8
 ; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; SCALABLE-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; SCALABLE-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
@@ -514,15 +532,14 @@ define void @uniform_load_unaligned(ptr noalias nocapture %a, ptr noalias nocapt
 ; FIXEDLEN-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[TMP1]], i64 0
 ; FIXEDLEN-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
 ; FIXEDLEN-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; FIXEDLEN-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
 ; FIXEDLEN-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 4
-; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP3]], align 8
+; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP2]], align 8
 ; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP4]], align 8
 ; FIXEDLEN-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FIXEDLEN-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXEDLEN-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; FIXEDLEN:       [[MIDDLE_BLOCK]]:
-; FIXEDLEN-NEXT:    br i1 false, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; FIXEDLEN-NEXT:    br label %[[SCALAR_PH]]
 ; FIXEDLEN:       [[SCALAR_PH]]:
 ; FIXEDLEN-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; FIXEDLEN-NEXT:    br label %[[FOR_BODY:.*]]
@@ -533,7 +550,7 @@ define void @uniform_load_unaligned(ptr noalias nocapture %a, ptr noalias nocapt
 ; FIXEDLEN-NEXT:    store i64 [[V]], ptr [[ARRAYIDX]], align 8
 ; FIXEDLEN-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; FIXEDLEN-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025
-; FIXEDLEN-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; FIXEDLEN-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
 ; FIXEDLEN:       [[FOR_END]]:
 ; FIXEDLEN-NEXT:    ret void
 ;
@@ -560,13 +577,12 @@ define void @uniform_load_unaligned(ptr noalias nocapture %a, ptr noalias nocapt
 ; TF-SCALABLE-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 2 x i64> poison, i64 [[TMP5]], i64 0
 ; TF-SCALABLE-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
 ; TF-SCALABLE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; TF-SCALABLE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP7]], i32 0
-; TF-SCALABLE-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr align 8 [[TMP8]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP6]])
-; TF-SCALABLE-NEXT:    [[TMP10:%.*]] = zext i32 [[TMP6]] to i64
-; TF-SCALABLE-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP10]], [[INDEX]]
+; TF-SCALABLE-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr align 8 [[TMP7]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP6]])
+; TF-SCALABLE-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP6]] to i64
+; TF-SCALABLE-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP8]], [[INDEX]]
 ; TF-SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX1]], [[TMP4]]
 ; TF-SCALABLE-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; TF-SCALABLE-NEXT:    br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
+; TF-SCALABLE-NEXT:    br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
 ; TF-SCALABLE:       [[MIDDLE_BLOCK]]:
 ; TF-SCALABLE-NEXT:    br label %[[FOR_END:.*]]
 ; TF-SCALABLE:       [[SCALAR_PH]]:
@@ -579,7 +595,7 @@ define void @uniform_load_unaligned(ptr noalias nocapture %a, ptr noalias nocapt
 ; TF-SCALABLE-NEXT:    store i64 [[V]], ptr [[ARRAYIDX]], align 8
 ; TF-SCALABLE-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; TF-SCALABLE-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025
-; TF-SCALABLE-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; TF-SCALABLE-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; TF-SCALABLE:       [[FOR_END]]:
 ; TF-SCALABLE-NEXT:    ret void
 ;
@@ -621,8 +637,7 @@ define void @uniform_store(ptr noalias nocapture %a, ptr noalias nocapture %b, i
 ; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; SCALABLE-NEXT:    store i64 [[V]], ptr [[B]], align 8
 ; SCALABLE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; SCALABLE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP7]], i32 0
-; SCALABLE-NEXT:    store <vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP8]], align 8
+; SCALABLE-NEXT:    store <vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP7]], align 8
 ; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; SCALABLE-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; SCALABLE-NEXT:    br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
@@ -655,15 +670,14 @@ define void @uniform_store(ptr noalias nocapture %a, ptr noalias nocapture %b, i
 ; FIXEDLEN-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; FIXEDLEN-NEXT:    store i64 [[V]], ptr [[B]], align 8
 ; FIXEDLEN-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; FIXEDLEN-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
 ; FIXEDLEN-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 4
-; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP2]], align 8
+; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP1]], align 8
 ; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP3]], align 8
 ; FIXEDLEN-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FIXEDLEN-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXEDLEN-NEXT:    br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; FIXEDLEN:       [[MIDDLE_BLOCK]]:
-; FIXEDLEN-NEXT:    br i1 false, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; FIXEDLEN-NEXT:    br label %[[SCALAR_PH]]
 ; FIXEDLEN:       [[SCALAR_PH]]:
 ; FIXEDLEN-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; FIXEDLEN-NEXT:    br label %[[FOR_BODY:.*]]
@@ -674,7 +688,7 @@ define void @uniform_store(ptr noalias nocapture %a, ptr noalias nocapture %b, i
 ; FIXEDLEN-NEXT:    store i64 [[V]], ptr [[ARRAYIDX]], align 8
 ; FIXEDLEN-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; FIXEDLEN-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025
-; FIXEDLEN-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
+; FIXEDLEN-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ; FIXEDLEN:       [[FOR_END]]:
 ; FIXEDLEN-NEXT:    ret void
 ;
@@ -701,13 +715,12 @@ define void @uniform_store(ptr noalias nocapture %a, ptr noalias nocapture %b, i
 ; TF-SCALABLE-NEXT:    [[TMP5:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
 ; TF-SCALABLE-NEXT:    store i64 [[V]], ptr [[B]], align 8
 ; TF-SCALABLE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; TF-SCALABLE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i32 0
-; TF-SCALABLE-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr align 8 [[TMP7]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
-; TF-SCALABLE-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP5]] to i64
-; TF-SCALABLE-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP9]], [[INDEX]]
+; TF-SCALABLE-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr align 8 [[TMP6]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
+; TF-SCALABLE-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP5]] to i64
+; TF-SCALABLE-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP7]], [[INDEX]]
 ; TF-SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX1]], [[TMP4]]
 ; TF-SCALABLE-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; TF-SCALABLE-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; TF-SCALABLE-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ; TF-SCALABLE:       [[MIDDLE_BLOCK]]:
 ; TF-SCALABLE-NEXT:    br label %[[FOR_END:.*]]
 ; TF-SCALABLE:       [[SCALAR_PH]]:
@@ -720,7 +733,7 @@ define void @uniform_store(ptr noalias nocapture %a, ptr noalias nocapture %b, i
 ; TF-SCALABLE-NEXT:    store i64 [[V]], ptr [[ARRAYIDX]], align 8
 ; TF-SCALABLE-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; TF-SCALABLE-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025
-; TF-SCALABLE-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; TF-SCALABLE-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; TF-SCALABLE:       [[FOR_END]]:
 ; TF-SCALABLE-NEXT:    ret void
 ;
@@ -774,8 +787,7 @@ define void @uniform_store_of_loop_varying(ptr noalias nocapture %a, ptr noalias
 ; SCALABLE-NEXT:    [[TMP15:%.*]] = extractelement <vscale x 2 x i64> [[TMP9]], i32 [[TMP14]]
 ; SCALABLE-NEXT:    store i64 [[TMP15]], ptr [[B]], align 8
 ; SCALABLE-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP10]]
-; SCALABLE-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i64, ptr [[TMP16]], i32 0
-; SCALABLE-NEXT:    store <vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP17]], align 8
+; SCALABLE-NEXT:    store <vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP16]], align 8
 ; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; SCALABLE-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; SCALABLE-NEXT:    br i1 [[TMP18]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
@@ -812,15 +824,14 @@ define void @uniform_store_of_loop_varying(ptr noalias nocapture %a, ptr noalias
 ; FIXEDLEN-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 7
 ; FIXEDLEN-NEXT:    store i64 [[TMP4]], ptr [[B]], align 8
 ; FIXEDLEN-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; FIXEDLEN-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 0
 ; FIXEDLEN-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 4
-; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP6]], align 8
+; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP5]], align 8
 ; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP7]], align 8
 ; FIXEDLEN-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FIXEDLEN-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXEDLEN-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; FIXEDLEN:       [[MIDDLE_BLOCK]]:
-; FIXEDLEN-NEXT:    br i1 false, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; FIXEDLEN-NEXT:    br label %[[SCALAR_PH]]
 ; FIXEDLEN:       [[SCALAR_PH]]:
 ; FIXEDLEN-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; FIXEDLEN-NEXT:    br label %[[FOR_BODY:.*]]
@@ -831,7 +842,7 @@ define void @uniform_store_of_loop_varying(ptr noalias nocapture %a, ptr noalias
 ; FIXEDLEN-NEXT:    store i64 [[V]], ptr [[ARRAYIDX]], align 8
 ; FIXEDLEN-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; FIXEDLEN-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025
-; FIXEDLEN-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
+; FIXEDLEN-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
 ; FIXEDLEN:       [[FOR_END]]:
 ; FIXEDLEN-NEXT:    ret void
 ;
@@ -868,14 +879,13 @@ define void @uniform_store_of_loop_varying(ptr noalias nocapture %a, ptr noalias
 ; TF-SCALABLE-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 2 x i64> [[BROADCAST_SPLATINSERT1]], <vscale x 2 x i64> poison, <vscale x 2 x i32> zeroinitializer
 ; TF-SCALABLE-NEXT:    call void @llvm.vp.scatter.nxv2i64.nxv2p0(<vscale x 2 x i64> [[VEC_IND]], <vscale x 2 x ptr> align 8 [[BROADCAST_SPLAT]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP9]])
 ; TF-SCALABLE-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; TF-SCALABLE-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i32 0
-; TF-SCALABLE-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[BROADCAST_SPLAT3]], ptr align 8 [[TMP11]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP9]])
-; TF-SCALABLE-NEXT:    [[TMP14:%.*]] = zext i32 [[TMP9]] to i64
-; TF-SCALABLE-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP14]], [[INDEX]]
+; TF-SCALABLE-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[BROADCAST_SPLAT3]], ptr align 8 [[TMP10]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP9]])
+; TF-SCALABLE-NEXT:    [[TMP11:%.*]] = zext i32 [[TMP9]] to i64
+; TF-SCALABLE-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP11]], [[INDEX]]
 ; TF-SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX1]], [[TMP4]]
 ; TF-SCALABLE-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT2]]
 ; TF-SCALABLE-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; TF-SCALABLE-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
+; TF-SCALABLE-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
 ; TF-SCALABLE:       [[MIDDLE_BLOCK]]:
 ; TF-SCALABLE-NEXT:    br label %[[FOR_END:.*]]
 ; TF-SCALABLE:       [[SCALAR_PH]]:
@@ -888,7 +898,7 @@ define void @uniform_store_of_loop_varying(ptr noalias nocapture %a, ptr noalias
 ; TF-SCALABLE-NEXT:    store i64 [[V]], ptr [[ARRAYIDX]], align 8
 ; TF-SCALABLE-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; TF-SCALABLE-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025
-; TF-SCALABLE-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
+; TF-SCALABLE-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; TF-SCALABLE:       [[FOR_END]]:
 ; TF-SCALABLE-NEXT:    ret void
 ;
@@ -940,8 +950,7 @@ define void @conditional_uniform_store(ptr noalias nocapture %a, ptr noalias noc
 ; SCALABLE-NEXT:    [[TMP10:%.*]] = icmp ugt <vscale x 2 x i64> [[VEC_IND]], splat (i64 10)
 ; SCALABLE-NEXT:    call void @llvm.masked.scatter.nxv2i64.nxv2p0(<vscale x 2 x i64> [[BROADCAST_SPLAT1]], <vscale x 2 x ptr> [[BROADCAST_SPLAT2]], i32 8, <vscale x 2 x i1> [[TMP10]])
 ; SCALABLE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; SCALABLE-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i64, ptr [[TMP12]], i32 0
-; SCALABLE-NEXT:    store <vscale x 2 x i64> [[BROADCAST_SPLAT1]], ptr [[TMP13]], align 8
+; SCALABLE-NEXT:    store <vscale x 2 x i64> [[BROADCAST_SPLAT1]], ptr [[TMP12]], align 8
 ; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; SCALABLE-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]]
 ; SCALABLE-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -987,16 +996,15 @@ define void @conditional_uniform_store(ptr noalias nocapture %a, ptr noalias noc
 ; FIXEDLEN-NEXT:    call void @llvm.masked.scatter.v4i64.v4p0(<4 x i64> [[BROADCAST_SPLAT]], <4 x ptr> [[BROADCAST_SPLAT2]], i32 8, <4 x i1> [[TMP1]])
 ; FIXEDLEN-NEXT:    call void @llvm.masked.scatter.v4i64.v4p0(<4 x i64> [[BROADCAST_SPLAT]], <4 x ptr> [[BROADCAST_SPLAT2]], i32 8, <4 x i1> [[TMP2]])
 ; FIXEDLEN-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; FIXEDLEN-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP3]], i32 0
 ; FIXEDLEN-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP3]], i32 4
-; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP4]], align 8
+; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP3]], align 8
 ; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP5]], align 8
 ; FIXEDLEN-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FIXEDLEN-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[STEP_ADD]], splat (i64 4)
 ; FIXEDLEN-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXEDLEN-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; FIXEDLEN:       [[MIDDLE_BLOCK]]:
-; FIXEDLEN-NEXT:    br i1 false, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; FIXEDLEN-NEXT:    br label %[[SCALAR_PH]]
 ; FIXEDLEN:       [[SCALAR_PH]]:
 ; FIXEDLEN-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; FIXEDLEN-NEXT:    br label %[[FOR_BODY:.*]]
@@ -1012,7 +1020,7 @@ define void @conditional_uniform_store(ptr noalias nocapture %a, ptr noalias noc
 ; FIXEDLEN-NEXT:    store i64 [[V]], ptr [[ARRAYIDX]], align 8
 ; FIXEDLEN-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; FIXEDLEN-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025
-; FIXEDLEN-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
+; FIXEDLEN-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; FIXEDLEN:       [[FOR_END]]:
 ; FIXEDLEN-NEXT:    ret void
 ;
@@ -1050,14 +1058,13 @@ define void @conditional_uniform_store(ptr noalias nocapture %a, ptr noalias noc
 ; TF-SCALABLE-NEXT:    [[TMP10:%.*]] = icmp ugt <vscale x 2 x i64> [[VEC_IND]], splat (i64 10)
 ; TF-SCALABLE-NEXT:    call void @llvm.vp.scatter.nxv2i64.nxv2p0(<vscale x 2 x i64> [[BROADCAST_SPLAT1]], <vscale x 2 x ptr> align 8 [[BROADCAST_SPLAT2]], <vscale x 2 x i1> [[TMP10]], i32 [[TMP9]])
 ; TF-SCALABLE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; TF-SCALABLE-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i64, ptr [[TMP12]], i32 0
-; TF-SCALABLE-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[BROADCAST_SPLAT1]], ptr align 8 [[TMP13]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP9]])
-; TF-SCALABLE-NEXT:    [[TMP15:%.*]] = zext i32 [[TMP9]] to i64
-; TF-SCALABLE-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP15]], [[INDEX]]
+; TF-SCALABLE-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[BROADCAST_SPLAT1]], ptr align 8 [[TMP12]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP9]])
+; TF-SCALABLE-NEXT:    [[TMP14:%.*]] = zext i32 [[TMP9]] to i64
+; TF-SCALABLE-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP14]], [[INDEX]]
 ; TF-SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX1]], [[TMP4]]
 ; TF-SCALABLE-NEXT:    [[VEC_IND_NEXT]] = add <vscale x 2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]]
-; TF-SCALABLE-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; TF-SCALABLE-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
+; TF-SCALABLE-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; TF-SCALABLE-NEXT:    br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; TF-SCALABLE:       [[MIDDLE_BLOCK]]:
 ; TF-SCALABLE-NEXT:    br label %[[FOR_END:.*]]
 ; TF-SCALABLE:       [[SCALAR_PH]]:
@@ -1075,7 +1082,7 @@ define void @conditional_uniform_store(ptr noalias nocapture %a, ptr noalias noc
 ; TF-SCALABLE-NEXT:    store i64 [[V]], ptr [[ARRAYIDX]], align 8
 ; TF-SCALABLE-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; TF-SCALABLE-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025
-; TF-SCALABLE-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
+; TF-SCALABLE-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; TF-SCALABLE:       [[FOR_END]]:
 ; TF-SCALABLE-NEXT:    ret void
 ;
@@ -1123,8 +1130,7 @@ define void @uniform_store_unaligned(ptr noalias nocapture %a, ptr noalias nocap
 ; SCALABLE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; SCALABLE-NEXT:    store i64 [[V]], ptr [[B]], align 1
 ; SCALABLE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; SCALABLE-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP7]], i32 0
-; SCALABLE-NEXT:    store <vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP8]], align 8
+; SCALABLE-NEXT:    store <vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP7]], align 8
 ; SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; SCALABLE-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; SCALABLE-NEXT:    br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
@@ -1157,15 +1163,14 @@ define void @uniform_store_unaligned(ptr noalias nocapture %a, ptr noalias nocap
 ; FIXEDLEN-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; FIXEDLEN-NEXT:    store i64 [[V]], ptr [[B]], align 1
 ; FIXEDLEN-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; FIXEDLEN-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
 ; FIXEDLEN-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 4
-; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP2]], align 8
+; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP1]], align 8
 ; FIXEDLEN-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP3]], align 8
 ; FIXEDLEN-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FIXEDLEN-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; FIXEDLEN-NEXT:    br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; FIXEDLEN:       [[MIDDLE_BLOCK]]:
-; FIXEDLEN-NEXT:    br i1 false, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; FIXEDLEN-NEXT:    br label %[[SCALAR_PH]]
 ; FIXEDLEN:       [[SCALAR_PH]]:
 ; FIXEDLEN-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; FIXEDLEN-NEXT:    br label %[[FOR_BODY:.*]]
@@ -1176,7 +1181,7 @@ define void @uniform_store_unaligned(ptr noalias nocapture %a, ptr noalias nocap
 ; FIXEDLEN-NEXT:    store i64 [[V]], ptr [[ARRAYIDX]], align 8
 ; FIXEDLEN-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; FIXEDLEN-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025
-; FIXEDLEN-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
+; FIXEDLEN-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
 ; FIXEDLEN:       [[FOR_END]]:
 ; FIXEDLEN-NEXT:    ret void
 ;
@@ -1203,13 +1208,12 @@ define void @uniform_store_unaligned(ptr noalias nocapture %a, ptr noalias nocap
 ; TF-SCALABLE-NEXT:    [[TMP5:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[AVL]], i32 2, i1 true)
 ; TF-SCALABLE-NEXT:    store i64 [[V]], ptr [[B]], align 1
 ; TF-SCALABLE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; TF-SCALABLE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i32 0
-; TF-SCALABLE-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr align 8 [[TMP7]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
-; TF-SCALABLE-NEXT:    [[TMP9:%.*]] = zext i32 [[TMP5]] to i64
-; TF-SCALABLE-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP9]], [[INDEX]]
+; TF-SCALABLE-NEXT:    call void @llvm.vp.store.nxv2i64.p0(<vscale x 2 x i64> [[BROADCAST_SPLAT]], ptr align 8 [[TMP6]], <vscale x 2 x i1> splat (i1 true), i32 [[TMP5]])
+; TF-SCALABLE-NEXT:    [[TMP7:%.*]] = zext i32 [[TMP5]] to i64
+; TF-SCALABLE-NEXT:    [[INDEX_EVL_NEXT]] = add nuw i64 [[TMP7]], [[INDEX]]
 ; TF-SCALABLE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX1]], [[TMP4]]
 ; TF-SCALABLE-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; TF-SCALABLE-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
+; TF-SCALABLE-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
 ; TF-SCALABLE:       [[MIDDLE_BLOCK]]:
 ; TF-SCALABLE-NEXT:    br label %[[FOR_END:.*]]
 ; TF-SCALABLE:       [[SCALAR_PH]]:
@@ -1222,7 +1226,7 @@ define void @uniform_store_unaligned(ptr noalias nocapture %a, ptr noalias nocap
 ; TF-SCALABLE-NEXT:    store i64 [[V]], ptr [[ARRAYIDX]], align 8
 ; TF-SCALABLE-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; TF-SCALABLE-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1025
-; TF-SCALABLE-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; TF-SCALABLE-NEXT:    br i1 [[EXITCOND_NOT]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; TF-SCALABLE:       [[FOR_END]]:
 ; TF-SCALABLE-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-vp-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-vp-intrinsics.ll
index 7b91f10..7537a43 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-vp-intrinsics.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vectorize-vp-intrinsics.ll
@@ -1,12 +1,10 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck --check-prefix=IF-EVL %s
 
 ; RUN: opt -passes=loop-vectorize \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -S < %s | FileCheck --check-prefix=NO-VP %s
 
 define void @foo(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %N) {
@@ -29,15 +27,12 @@ define void @foo(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %N) {
 ; IF-EVL-NEXT:    [[TMP11:%.*]] = sub i64 [[N]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[TMP12:%.*]] = call i32 @llvm.experimental.get.vector.length.i64(i64 [[TMP11]], i32 4, i1 true)
 ; IF-EVL-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP15]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP14]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; IF-EVL-NEXT:    [[VP_OP_LOAD1:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP17]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    [[VP_OP_LOAD1:%.*]] = call <vscale x 4 x i32> @llvm.vp.load.nxv4i32.p0(ptr align 4 [[TMP16]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-NEXT:    [[VP_OP:%.*]] = add nsw <vscale x 4 x i32> [[VP_OP_LOAD1]], [[VP_OP_LOAD]]
 ; IF-EVL-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[EVL_BASED_IV]]
-; IF-EVL-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[TMP18]], i32 0
-; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[VP_OP]], ptr align 4 [[TMP19]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
+; IF-EVL-NEXT:    call void @llvm.vp.store.nxv4i32.p0(<vscale x 4 x i32> [[VP_OP]], ptr align 4 [[TMP18]], <vscale x 4 x i1> splat (i1 true), i32 [[TMP12]])
 ; IF-EVL-NEXT:    [[TMP20:%.*]] = zext i32 [[TMP12]] to i64
 ; IF-EVL-NEXT:    [[INDEX_EVL_NEXT]] = add i64 [[TMP20]], [[EVL_BASED_IV]]
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], [[TMP10]]
@@ -80,15 +75,12 @@ define void @foo(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %N) {
 ; NO-VP:       vector.body:
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
 ; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i32>, ptr [[TMP10]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i32>, ptr [[TMP9]], align 4
 ; NO-VP-NEXT:    [[TMP11:%.*]] = add nsw <vscale x 4 x i32> [[WIDE_LOAD1]], [[WIDE_LOAD]]
 ; NO-VP-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP11]], ptr [[TMP13]], align 4
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP11]], ptr [[TMP12]], align 4
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
 ; NO-VP-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; NO-VP-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vf-will-not-generate-any-vector-insts.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vf-will-not-generate-any-vector-insts.ll
index e7fdfbc..bda9839 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vf-will-not-generate-any-vector-insts.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vf-will-not-generate-any-vector-insts.ll
@@ -31,9 +31,9 @@ define void @vf_will_not_generate_any_vector_insts(ptr %src, ptr %dst) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP1]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[TMP2:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[TMP3:%.*]], %[[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-call-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-call-intrinsics.ll
index 9652351..7f1066c 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-call-intrinsics.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-call-intrinsics.ll
@@ -1,8 +1,7 @@
 ; REQUIRES: asserts
 
 ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefix=IF-EVL %s
 
 define void @vp_smax(ptr %a, ptr %b, ptr %c, i64 %N) {
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
index 3ec48ef..c1b656a 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-cast-intrinsics.ll
@@ -1,7 +1,6 @@
 ; REQUIRES: asserts
 ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefix=IF-EVL %s
 
 define void @vp_sext(ptr %a, ptr %b, i64 %N) {
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-fixed-order-recurrence.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-fixed-order-recurrence.ll
index 7f29213..9900602 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-fixed-order-recurrence.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-fixed-order-recurrence.ll
@@ -1,8 +1,7 @@
 ; REQUIRES: asserts
 
 ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefix=IF-EVL %s
 
 define void @first_order_recurrence(ptr noalias %A, ptr noalias %B, i64 %TC) {
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-reduction.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-reduction.ll
index baf546b..1c9554d 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics-reduction.ll
@@ -1,25 +1,21 @@
 ; REQUIRES: asserts
 
 ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefixes=IF-EVL-OUTLOOP,IF-EVL %s
 
 ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \
 ; RUN: -prefer-inloop-reductions \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefixes=IF-EVL-INLOOP,IF-EVL %s
 
 ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefixes=NO-VP-OUTLOOP %s
 
 ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \
 ; RUN: -prefer-inloop-reductions \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefixes=NO-VP-INLOOP %s
 
 
diff --git a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics.ll b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics.ll
index 97a6130..42a846a 100644
--- a/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics.ll
+++ b/llvm/test/Transforms/LoopVectorize/RISCV/vplan-vp-intrinsics.ll
@@ -1,13 +1,11 @@
 ; REQUIRES: asserts
 
 ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \
-; RUN: -force-tail-folding-style=data-with-evl \
-; RUN: -prefer-predicate-over-epilogue=predicate-dont-vectorize \
+; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefixes=IF-EVL,CHECK %s
 
 ; RUN: opt -passes=loop-vectorize -debug-only=loop-vectorize \
-; RUN: -force-tail-folding-style=none \
-; RUN: -prefer-predicate-over-epilogue=predicate-else-scalar-epilogue \
+; RUN: -prefer-predicate-over-epilogue=scalar-epilogue \
 ; RUN: -mtriple=riscv64 -mattr=+v -riscv-v-vector-bits-max=128 -disable-output < %s 2>&1 | FileCheck --check-prefixes=NO-VP,CHECK %s
 
 define void @foo(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %N) {
diff --git a/llvm/test/Transforms/LoopVectorize/SystemZ/addressing.ll b/llvm/test/Transforms/LoopVectorize/SystemZ/addressing.ll
index c5e4ecc..722786f 100644
--- a/llvm/test/Transforms/LoopVectorize/SystemZ/addressing.ll
+++ b/llvm/test/Transforms/LoopVectorize/SystemZ/addressing.ll
@@ -25,7 +25,7 @@ define i32 @foo(ptr nocapture %A) {
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10000
 ; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -76,7 +76,7 @@ define i32 @foo1(ptr nocapture noalias %A, ptr nocapture %PtrPtr) {
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10000
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
diff --git a/llvm/test/Transforms/LoopVectorize/SystemZ/scalar-steps-with-users-demanding-all-lanes-and-first-lane-only.ll b/llvm/test/Transforms/LoopVectorize/SystemZ/scalar-steps-with-users-demanding-all-lanes-and-first-lane-only.ll
index 61bcbaa..a91bc65 100644
--- a/llvm/test/Transforms/LoopVectorize/SystemZ/scalar-steps-with-users-demanding-all-lanes-and-first-lane-only.ll
+++ b/llvm/test/Transforms/LoopVectorize/SystemZ/scalar-steps-with-users-demanding-all-lanes-and-first-lane-only.ll
@@ -34,8 +34,7 @@ define void @test_scalar_iv_steps_used_by_replicate_and_first_lane_only_vpinst(p
 ; CHECK-NEXT:    [[TMP19:%.*]] = insertelement <4 x i8> [[TMP14]], i8 [[TMP11]], i32 3
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq <4 x i8> [[TMP19]], zeroinitializer
 ; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr [8 x i32], ptr @src, i64 0, i64 4
-; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr i32, ptr [[TMP22]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP23]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP22]], align 4
 ; CHECK-NEXT:    [[TMP24:%.*]] = extractelement <4 x i1> [[TMP20]], i32 0
 ; CHECK-NEXT:    br i1 [[TMP24]], label %[[PRED_STORE_IF:.*]], label %[[PRED_STORE_CONTINUE:.*]]
 ; CHECK:       [[PRED_STORE_IF]]:
@@ -66,9 +65,9 @@ define void @test_scalar_iv_steps_used_by_replicate_and_first_lane_only_vpinst(p
 ; CHECK:       [[PRED_STORE_CONTINUE6]]:
 ; CHECK-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/constant-fold.ll b/llvm/test/Transforms/LoopVectorize/X86/constant-fold.ll
index 86b28c3..c61b1b9 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/constant-fold.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/constant-fold.ll
@@ -19,13 +19,12 @@ define void @f1() {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[TMP0:%.*]] = sext i16 0 to i64
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr [2 x ptr], ptr @b, i16 0, i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr ptr, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <2 x ptr> <ptr @a, ptr @a>, ptr [[TMP2]], align 8
+; CHECK-NEXT:    store <2 x ptr> <ptr @a, ptr @a>, ptr [[TMP1]], align 8
 ; CHECK-NEXT:    br label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[BB3:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[BB3:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 2, [[MIDDLE_BLOCK]] ], [ 0, [[BB1:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, [[BB1:%.*]] ]
 ; CHECK-NEXT:    br label [[BB2:%.*]]
 ; CHECK:       bb2:
 ; CHECK-NEXT:    [[C_1_0:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[_TMP9:%.*]], [[BB2]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/conversion-cost.ll b/llvm/test/Transforms/LoopVectorize/X86/conversion-cost.ll
index 0cbf6eb..9506ad3 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/conversion-cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/conversion-cost.ll
@@ -26,8 +26,7 @@ define i32 @conversion_cost1(i32 %n, ptr nocapture %A, ptr nocapture %B) nounwin
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <32 x i8> [ <i8 3, i8 4, i8 5, i8 6, i8 7, i8 8, i8 9, i8 10, i8 11, i8 12, i8 13, i8 14, i8 15, i8 16, i8 17, i8 18, i8 19, i8 20, i8 21, i8 22, i8 23, i8 24, i8 25, i8 26, i8 27, i8 28, i8 29, i8 30, i8 31, i8 32, i8 33, i8 34>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <32 x i8> [[VEC_IND]], ptr [[TMP6]], align 1
+; CHECK-NEXT:    store <32 x i8> [[VEC_IND]], ptr [[TMP5]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <32 x i8> [[VEC_IND]], splat (i8 32)
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -56,8 +55,7 @@ define i32 @conversion_cost1(i32 %n, ptr nocapture %A, ptr nocapture %B) nounwin
 ; CHECK-NEXT:    [[VEC_IND8:%.*]] = phi <4 x i8> [ [[INDUCTION]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT9:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX10:%.*]] = add i64 3, [[INDEX7]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[OFFSET_IDX10]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[VEC_IND8]], ptr [[TMP11]], align 1
+; CHECK-NEXT:    store <4 x i8> [[VEC_IND8]], ptr [[TMP10]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT11]] = add nuw i64 [[INDEX7]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT9]] = add <4 x i8> [[VEC_IND8]], splat (i8 4)
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT11]], [[N_VEC3]]
@@ -129,11 +127,10 @@ define i32 @conversion_cost2(i32 %n, ptr nocapture %A, ptr nocapture %B) nounwin
 ; CHECK-NEXT:    [[TMP19:%.*]] = sitofp <2 x i64> [[TMP10]] to <2 x float>
 ; CHECK-NEXT:    [[TMP20:%.*]] = sitofp <2 x i64> [[TMP11]] to <2 x float>
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[TMP13]], i32 0
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[TMP13]], i32 2
 ; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[TMP13]], i32 4
 ; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[TMP13]], i32 6
-; CHECK-NEXT:    store <2 x float> [[TMP12]], ptr [[TMP14]], align 4
+; CHECK-NEXT:    store <2 x float> [[TMP12]], ptr [[TMP13]], align 4
 ; CHECK-NEXT:    store <2 x float> [[TMP18]], ptr [[TMP15]], align 4
 ; CHECK-NEXT:    store <2 x float> [[TMP19]], ptr [[TMP16]], align 4
 ; CHECK-NEXT:    store <2 x float> [[TMP20]], ptr [[TMP17]], align 4
diff --git a/llvm/test/Transforms/LoopVectorize/X86/cost-constant-known-via-scev.ll b/llvm/test/Transforms/LoopVectorize/X86/cost-constant-known-via-scev.ll
index 0c5db43..04e0daf 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/cost-constant-known-via-scev.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/cost-constant-known-via-scev.ll
@@ -25,7 +25,7 @@ define i64 @test_foldable_live_in_via_scev() {
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[BIN_RDX:%.*]] = mul <2 x i64> [[TMP1]], [[TMP0]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vector.reduce.mul.v2i64(<2 x i64> [[BIN_RDX]])
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 97, %[[MIDDLE_BLOCK]] ], [ 1, %[[ENTRY]] ]
 ; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP3]], %[[MIDDLE_BLOCK]] ], [ 1, %[[ENTRY]] ]
@@ -36,9 +36,9 @@ define i64 @test_foldable_live_in_via_scev() {
 ; CHECK-NEXT:    [[MUL]] = mul nsw i64 [[RED]], [[ADD]]
 ; CHECK-NEXT:    [[IV_NEXT]] = add nsw i32 [[IV]], 1
 ; CHECK-NEXT:    [[EC:%.*]] = icmp eq i32 [[IV_NEXT]], 100
-; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       [[EXIT]]:
-; CHECK-NEXT:    [[RET:%.*]] = phi i64 [ [[MUL]], %[[LOOP]] ], [ [[TMP3]], %[[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[RET:%.*]] = phi i64 [ [[MUL]], %[[LOOP]] ]
 ; CHECK-NEXT:    ret i64 [[RET]]
 ;
 entry:
@@ -95,10 +95,10 @@ define i64 @second_lshr_operand_zero_via_scev() {
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[BIN_RDX:%.*]] = or <2 x i64> [[TMP11]], [[TMP10]]
 ; CHECK-NEXT:    [[TMP13:%.*]] = call i64 @llvm.vector.reduce.or.v2i64(<2 x i64> [[BIN_RDX]])
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP13]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOPS:.*]]
 ; CHECK:       [[LOOPS]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOPS]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/cost-model.ll b/llvm/test/Transforms/LoopVectorize/X86/cost-model.ll
index 147e949..85b475c 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/cost-model.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/cost-model.ll
@@ -982,7 +982,7 @@ define void @reduction_store(ptr noalias %src, ptr %dst, i1 %x) #2 {
 ; CHECK-NEXT:    [[BIN_RDX:%.*]] = and <4 x i32> [[TMP12]], [[TMP11]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = call i32 @llvm.vector.reduce.and.v4i32(<4 x i32> [[BIN_RDX]])
 ; CHECK-NEXT:    store i32 [[TMP10]], ptr [[DST:%.*]], align 4
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP10]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 24, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
@@ -1001,7 +1001,7 @@ define void @reduction_store(ptr noalias %src, ptr %dst, i1 %x) #2 {
 ; CHECK-NEXT:    store i32 [[RED_NEXT]], ptr [[DST]], align 4
 ; CHECK-NEXT:    [[IV_NEXT]] = add i32 [[IV]], 1
 ; CHECK-NEXT:    [[EC:%.*]] = icmp eq i32 [[IV]], 29
-; CHECK-NEXT:    br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP23:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP23:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
@@ -1045,8 +1045,8 @@ define i64 @live_in_known_1_via_scev() {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ <i64 3, i64 1, i64 1, i64 1>, [[VECTOR_PH]] ], [ [[VEC_PHI]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 8
-; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; CHECK-NEXT:    [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 8
+; CHECK-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP3:%.*]] = call i64 @llvm.vector.reduce.mul.v4i64(<4 x i64> [[VEC_PHI]])
 ; CHECK-NEXT:    br label [[EXIT:%.*]]
@@ -1110,10 +1110,10 @@ define i64 @cost_loop_invariant_recipes(i1 %x, i64 %y) {
 ; CHECK-NEXT:    br i1 true, label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP4:%.*]] = call i64 @llvm.vector.reduce.mul.v2i64(<2 x i64> [[TMP3]])
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 2, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP4]], [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ 1, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT_I_I_I:%.*]], [[LOOP]] ]
@@ -1169,10 +1169,10 @@ define i32 @narrowed_reduction(ptr %a, i1 %cmp) #0 {
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP20:%.*]] = call i1 @llvm.vector.reduce.or.v16i1(<16 x i1> [[TMP5]])
 ; CHECK-NEXT:    [[TMP21:%.*]] = zext i1 [[TMP20]] to i32
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[VEC_EPILOG_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 17, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP21]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP1:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INC:%.*]], [[LOOP1]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/divs-with-tail-folding.ll b/llvm/test/Transforms/LoopVectorize/X86/divs-with-tail-folding.ll
index fd53a4c..d49aca9 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/divs-with-tail-folding.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/divs-with-tail-folding.ll
@@ -37,8 +37,7 @@ define void @sdiv_feeding_gep(ptr %dst, i32 %x, i64 %M, i64 %conv6, i64 %N) {
 ; CHECK-NEXT:    [[TMP24:%.*]] = add i32 [[TMP23]], [[TMP22]]
 ; CHECK-NEXT:    [[TMP25:%.*]] = sext i32 [[TMP24]] to i64
 ; CHECK-NEXT:    [[TMP26:%.*]] = getelementptr double, ptr [[DST]], i64 [[TMP25]]
-; CHECK-NEXT:    [[TMP27:%.*]] = getelementptr double, ptr [[TMP26]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> zeroinitializer, ptr [[TMP27]], i32 8, <4 x i1> [[TMP6]])
+; CHECK-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> zeroinitializer, ptr [[TMP26]], i32 8, <4 x i1> [[TMP6]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP28:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP28]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -151,8 +150,7 @@ define void @sdiv_feeding_gep_predicated(ptr %dst, i32 %x, i64 %M, i64 %conv6, i
 ; CHECK-NEXT:    [[TMP26:%.*]] = add i32 [[TMP25]], [[TMP24]]
 ; CHECK-NEXT:    [[TMP27:%.*]] = sext i32 [[TMP26]] to i64
 ; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr double, ptr [[DST]], i64 [[TMP27]]
-; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr double, ptr [[TMP28]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> zeroinitializer, ptr [[TMP29]], i32 8, <4 x i1> [[TMP8]])
+; CHECK-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> zeroinitializer, ptr [[TMP28]], i32 8, <4 x i1> [[TMP8]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add i64 [[TMP5]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/drop-poison-generating-flags.ll b/llvm/test/Transforms/LoopVectorize/X86/drop-poison-generating-flags.ll
index 49d33d3..84e36cb 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/drop-poison-generating-flags.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/drop-poison-generating-flags.ll
@@ -33,12 +33,10 @@ define void @drop_scalar_nuw_nsw(ptr noalias nocapture readonly %input, ptr %out
 ; CHECK-NEXT:    [[TMP1:%.*]] = xor <4 x i1> [[TMP0]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP2:%.*]] = sub i64 [[INDEX]], 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr float, ptr [[INPUT]], i64 [[TMP2]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr float, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP4]], i32 4, <4 x i1> [[TMP1]], <4 x float> poison), !invariant.load [[META0:![0-9]+]]
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP3]], i32 4, <4 x i1> [[TMP1]], <4 x float> poison), !invariant.load [[META0:![0-9]+]]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP0]], <4 x float> zeroinitializer, <4 x float> [[WIDE_MASKED_LOAD]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[OUTPUT]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP1:![0-9]+]]
@@ -85,12 +83,10 @@ define void @drop_scalar_gep_nusw(ptr noalias nocapture readonly %input, ptr %ou
 ; CHECK-NEXT:    [[TMP1:%.*]] = xor <4 x i1> [[TMP0]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP2:%.*]] = sub i64 [[INDEX]], 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr float, ptr [[INPUT]], i64 [[TMP2]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr float, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP4]], i32 4, <4 x i1> [[TMP1]], <4 x float> poison), !invariant.load [[META0]]
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP3]], i32 4, <4 x i1> [[TMP1]], <4 x float> poison), !invariant.load [[META0]]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP0]], <4 x float> zeroinitializer, <4 x float> [[WIDE_MASKED_LOAD]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr nusw float, ptr [[OUTPUT]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr nusw float, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
@@ -137,12 +133,10 @@ define void @drop_scalar_gep_nuw(ptr noalias nocapture readonly %input, ptr %out
 ; CHECK-NEXT:    [[TMP1:%.*]] = xor <4 x i1> [[TMP0]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP2:%.*]] = sub i64 [[INDEX]], 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr float, ptr [[INPUT]], i64 [[TMP2]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr float, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP4]], i32 4, <4 x i1> [[TMP1]], <4 x float> poison), !invariant.load [[META0]]
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP3]], i32 4, <4 x i1> [[TMP1]], <4 x float> poison), !invariant.load [[META0]]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP0]], <4 x float> zeroinitializer, <4 x float> [[WIDE_MASKED_LOAD]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr nuw float, ptr [[OUTPUT]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr nuw float, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
@@ -190,12 +184,10 @@ define void @drop_nonpred_scalar_nuw_nsw(ptr noalias nocapture readonly %input,
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr float, ptr [[INPUT]], i64 [[TMP0]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq <4 x i64> [[VEC_IND]], zeroinitializer
 ; CHECK-NEXT:    [[TMP3:%.*]] = xor <4 x i1> [[TMP2]], splat (i1 true)
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP4]], i32 4, <4 x i1> [[TMP3]], <4 x float> poison), !invariant.load [[META0]]
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP1]], i32 4, <4 x i1> [[TMP3]], <4 x float> poison), !invariant.load [[META0]]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP2]], <4 x float> zeroinitializer, <4 x float> [[WIDE_MASKED_LOAD]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[OUTPUT]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
@@ -246,8 +238,7 @@ define void @preserve_vector_nuw_nsw(ptr noalias nocapture readonly %input, ptr
 ; CHECK-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <4 x float> @llvm.masked.gather.v4f32.v4p0(<4 x ptr> [[TMP4]], i32 4, <4 x i1> [[TMP1]], <4 x float> poison), !invariant.load [[META0]]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP0]], <4 x float> zeroinitializer, <4 x float> [[WIDE_MASKED_GATHER]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[OUTPUT]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
@@ -295,16 +286,13 @@ define void @drop_vector_nuw_nsw(ptr noalias nocapture readonly %input, ptr %out
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds ptr, ptr [[PTRS]], i64 [[INDEX]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = sub <4 x i64> [[VEC_IND]], splat (i64 1)
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr float, ptr [[INPUT]], <4 x i64> [[TMP2]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds ptr, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <4 x ptr> [[TMP3]], ptr [[TMP4]], align 8
+; CHECK-NEXT:    store <4 x ptr> [[TMP3]], ptr [[TMP1]], align 8
 ; CHECK-NEXT:    [[TMP5:%.*]] = xor <4 x i1> [[TMP0]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <4 x ptr> [[TMP3]], i32 0
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr float, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP7]], i32 4, <4 x i1> [[TMP5]], <4 x float> poison), !invariant.load [[META0]]
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP6]], i32 4, <4 x i1> [[TMP5]], <4 x float> poison), !invariant.load [[META0]]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP0]], <4 x float> zeroinitializer, <4 x float> [[WIDE_MASKED_LOAD]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[OUTPUT]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP9]], align 4
+; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
@@ -369,15 +357,12 @@ define void @drop_nonvector_nuw_nsw_avx1(ptr noalias nocapture readonly %input,
 ; CHECK-NEXT:    [[TMP15:%.*]] = insertelement <4 x ptr> [[TMP14]], ptr [[TMP11]], i32 1
 ; CHECK-NEXT:    [[TMP16:%.*]] = insertelement <4 x ptr> [[TMP15]], ptr [[TMP12]], i32 2
 ; CHECK-NEXT:    [[TMP17:%.*]] = insertelement <4 x ptr> [[TMP16]], ptr [[TMP13]], i32 3
-; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds ptr, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <4 x ptr> [[TMP17]], ptr [[TMP18]], align 8
+; CHECK-NEXT:    store <4 x ptr> [[TMP17]], ptr [[TMP5]], align 8
 ; CHECK-NEXT:    [[TMP19:%.*]] = xor <4 x i1> [[TMP4]], splat (i1 true)
-; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr float, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP20]], i32 4, <4 x i1> [[TMP19]], <4 x float> poison), !invariant.load [[META0]]
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP10]], i32 4, <4 x i1> [[TMP19]], <4 x float> poison), !invariant.load [[META0]]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP4]], <4 x float> zeroinitializer, <4 x float> [[WIDE_MASKED_LOAD]]
 ; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, ptr [[OUTPUT]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[TMP21]], i32 0
-; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP22]], align 4
+; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP21]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
@@ -427,8 +412,7 @@ define void @preserve_nuw_nsw_no_addr(ptr %output) local_unnamed_addr #0 {
 ; CHECK-NEXT:    [[TMP1:%.*]] = sub nuw nsw <4 x i64> [[VEC_IND]], splat (i64 1)
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP0]], <4 x i64> zeroinitializer, <4 x i64> [[TMP1]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[OUTPUT]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    store <4 x i64> [[PREDPHI]], ptr [[TMP3]], align 4
+; CHECK-NEXT:    store <4 x i64> [[PREDPHI]], ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
@@ -476,12 +460,10 @@ define void @drop_scalar_exact(ptr noalias nocapture readonly %input, ptr %outpu
 ; CHECK-NEXT:    [[TMP4:%.*]] = xor <4 x i1> [[TMP3]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP5:%.*]] = sdiv i64 [[INDEX]], 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr float, ptr [[INPUT]], i64 [[TMP5]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr float, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP7]], i32 4, <4 x i1> [[TMP4]], <4 x float> poison), !invariant.load [[META0]]
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x float> @llvm.masked.load.v4f32.p0(ptr [[TMP6]], i32 4, <4 x i1> [[TMP4]], <4 x float> poison), !invariant.load [[META0]]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP3]], <4 x float> zeroinitializer, <4 x float> [[WIDE_MASKED_LOAD]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[OUTPUT]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP9]], align 4
+; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
@@ -530,8 +512,7 @@ define void @drop_zext_nneg(ptr noalias %p, ptr noalias %p1) #0 {
 ; CHECK-NEXT:    [[TMP1:%.*]] = zext <4 x i32> [[VEC_IND]] to <4 x i64>
 ; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x i64> [[TMP1]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr double, ptr [[P]], i64 [[TMP2]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr double, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP4]], i32 8, <4 x i1> [[TMP0]], <4 x double> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP3]], i32 8, <4 x i1> [[TMP0]], <4 x double> poison)
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP0]], <4 x double> [[WIDE_MASKED_LOAD]], <4 x double> zeroinitializer
 ; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <4 x double> [[PREDPHI]], i32 3
 ; CHECK-NEXT:    store double [[TMP5]], ptr [[P1]], align 8
@@ -589,8 +570,7 @@ define void @preserve_vector_exact_no_addr(ptr noalias nocapture readonly %input
 ; CHECK-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <4 x float> @llvm.masked.gather.v4f32.v4p0(<4 x ptr> [[TMP6]], i32 4, <4 x i1> [[TMP4]], <4 x float> poison), !invariant.load [[META0]]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP3]], <4 x float> zeroinitializer, <4 x float> [[WIDE_MASKED_GATHER]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[OUTPUT]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP8]], align 4
+; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
@@ -641,8 +621,7 @@ define void @preserve_exact_no_addr(ptr %output) local_unnamed_addr #0 {
 ; CHECK-NEXT:    [[TMP1:%.*]] = sdiv exact <4 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP0]], <4 x i64> zeroinitializer, <4 x i64> [[TMP1]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[OUTPUT]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    store <4 x i64> [[PREDPHI]], ptr [[TMP3]], align 4
+; CHECK-NEXT:    store <4 x i64> [[PREDPHI]], ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP25:![0-9]+]]
@@ -789,8 +768,7 @@ define void @pr70590_recipe_without_underlying_instr(i64 %n, ptr noalias %dst) {
 ; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = phi <4 x i8> [ [[TMP22]], %[[PRED_LOAD_CONTINUE4]] ], [ [[TMP28]], %[[PRED_LOAD_IF5]] ]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP0]], <4 x i8> zeroinitializer, <4 x i8> [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr i8, ptr [[TMP15]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[PREDPHI]], ptr [[TMP16]], align 4
+; CHECK-NEXT:    store <4 x i8> [[PREDPHI]], ptr [[TMP15]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP27:![0-9]+]]
@@ -889,8 +867,7 @@ define void @recipe_without_underlying_instr_lanes_used(i64 %n, ptr noalias %dst
 ; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <4 x i64> [[PREDPHI7]], i32 3
 ; CHECK-NEXT:    store i64 [[TMP12]], ptr [[AUX]], align 8
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i8, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[PREDPHI]], ptr [[TMP11]], align 4
+; CHECK-NEXT:    store <4 x i8> [[PREDPHI]], ptr [[TMP10]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP29:![0-9]+]]
@@ -940,16 +917,13 @@ define void @Bgep_inbounds_unconditionally_due_to_store(ptr noalias %B, ptr read
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP0]], align 4
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD]], splat (i32 20)
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr float, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr float, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = fadd <4 x float> [[WIDE_LOAD1]], splat (float 2.000000e+00)
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP2]], <4 x float> splat (float 3.300000e+01), <4 x float> [[TMP5]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <4 x float> [[PREDPHI]], ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10000
 ; CHECK-NEXT:    br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP31:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/epilog-vectorization-inductions.ll b/llvm/test/Transforms/LoopVectorize/X86/epilog-vectorization-inductions.ll
index 523f284..7c74981 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/epilog-vectorization-inductions.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/epilog-vectorization-inductions.ll
@@ -44,8 +44,7 @@ define void @test_pr59459(i64 %iv.start, ptr %arr) {
 ; CHECK-NEXT:    [[TMP15:%.*]] = trunc <16 x i32> [[TMP14]] to <16 x i16>
 ; CHECK-NEXT:    [[TMP16:%.*]] = zext i32 [[TMP12]] to i64
 ; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i16, ptr [[ARR:%.*]], i64 [[TMP16]]
-; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr i16, ptr [[TMP17]], i32 0
-; CHECK-NEXT:    store <16 x i16> [[TMP15]], ptr [[TMP18]], align 2
+; CHECK-NEXT:    store <16 x i16> [[TMP15]], ptr [[TMP17]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <16 x i32> [[VEC_IND]], splat (i32 16)
 ; CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -80,8 +79,7 @@ define void @test_pr59459(i64 %iv.start, ptr %arr) {
 ; CHECK-NEXT:    [[TMP26:%.*]] = trunc <4 x i32> [[TMP25]] to <4 x i16>
 ; CHECK-NEXT:    [[TMP27:%.*]] = zext i32 [[TMP23]] to i64
 ; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr i16, ptr [[ARR]], i64 [[TMP27]]
-; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr i16, ptr [[TMP28]], i32 0
-; CHECK-NEXT:    store <4 x i16> [[TMP26]], ptr [[TMP29]], align 2
+; CHECK-NEXT:    store <4 x i16> [[TMP26]], ptr [[TMP28]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT15]] = add nuw i64 [[INDEX8]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT13]] = add <4 x i32> [[VEC_IND12]], splat (i32 4)
 ; CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT15]], [[N_VEC4]]
@@ -166,11 +164,10 @@ define void @test_induction_step_needs_expansion(ptr noalias %j, ptr %k, i64 %l,
 ; CHECK-NEXT:    [[TMP6:%.*]] = sub <16 x i16> [[STEP_ADD_2]], [[BROADCAST_SPLAT3]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = sub <16 x i16> [[STEP_ADD_3]], [[BROADCAST_SPLAT3]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[K:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i16, ptr [[TMP8]], i32 0
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i16, ptr [[TMP8]], i32 16
 ; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i16, ptr [[TMP8]], i32 32
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i16, ptr [[TMP8]], i32 48
-; CHECK-NEXT:    store <16 x i16> [[TMP4]], ptr [[TMP9]], align 2
+; CHECK-NEXT:    store <16 x i16> [[TMP4]], ptr [[TMP8]], align 2
 ; CHECK-NEXT:    store <16 x i16> [[TMP5]], ptr [[TMP10]], align 2
 ; CHECK-NEXT:    store <16 x i16> [[TMP6]], ptr [[TMP21]], align 2
 ; CHECK-NEXT:    store <16 x i16> [[TMP7]], ptr [[TMP12]], align 2
@@ -211,8 +208,7 @@ define void @test_induction_step_needs_expansion(ptr noalias %j, ptr %k, i64 %l,
 ; CHECK-NEXT:    [[VEC_IND20:%.*]] = phi <8 x i16> [ [[INDUCTION17]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT21:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP17:%.*]] = sub <8 x i16> [[VEC_IND20]], [[DOTSPLAT14]]
 ; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i16, ptr [[K]], i64 [[INDEX12]]
-; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i16, ptr [[TMP18]], i32 0
-; CHECK-NEXT:    store <8 x i16> [[TMP17]], ptr [[TMP19]], align 2
+; CHECK-NEXT:    store <8 x i16> [[TMP17]], ptr [[TMP18]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT24]] = add nuw i64 [[INDEX12]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT21]] = add <8 x i16> [[VEC_IND20]], [[BROADCAST_SPLAT23]]
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT24]], [[N_VEC5]]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/fixed-order-recurrence.ll b/llvm/test/Transforms/LoopVectorize/X86/fixed-order-recurrence.ll
index 86fa5d6..fe2ad66 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/fixed-order-recurrence.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/fixed-order-recurrence.ll
@@ -26,18 +26,16 @@ define void @firstorderrec(ptr nocapture noundef readonly %x, ptr noalias nocapt
 ; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <16 x i8> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD1:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[X]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 16
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD1]] = load <16 x i8>, ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <16 x i8> [[VECTOR_RECUR]], <16 x i8> [[WIDE_LOAD]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
 ; CHECK-NEXT:    [[TMP8:%.*]] = shufflevector <16 x i8> [[WIDE_LOAD]], <16 x i8> [[WIDE_LOAD1]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
 ; CHECK-NEXT:    [[TMP9:%.*]] = add <16 x i8> [[WIDE_LOAD]], [[TMP7]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = add <16 x i8> [[WIDE_LOAD1]], [[TMP8]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[Y:%.*]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
 ; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 16
-; CHECK-NEXT:    store <16 x i8> [[TMP9]], ptr [[TMP13]], align 1
+; CHECK-NEXT:    store <16 x i8> [[TMP9]], ptr [[TMP11]], align 1
 ; CHECK-NEXT:    store <16 x i8> [[TMP10]], ptr [[TMP14]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
 ; CHECK-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -121,9 +119,8 @@ define void @thirdorderrec(ptr nocapture noundef readonly %x, ptr noalias nocapt
 ; CHECK-NEXT:    [[VECTOR_RECUR4:%.*]] = phi <16 x i8> [ [[VECTOR_RECUR_INIT3]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[X]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 16
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[WIDE_LOAD5]] = load <16 x i8>, ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <16 x i8> [[VECTOR_RECUR]], <16 x i8> [[WIDE_LOAD]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
 ; CHECK-NEXT:    [[TMP8]] = shufflevector <16 x i8> [[WIDE_LOAD]], <16 x i8> [[WIDE_LOAD5]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
@@ -138,9 +135,8 @@ define void @thirdorderrec(ptr nocapture noundef readonly %x, ptr noalias nocapt
 ; CHECK-NEXT:    [[TMP17:%.*]] = add <16 x i8> [[TMP15]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP18:%.*]] = add <16 x i8> [[TMP16]], [[WIDE_LOAD5]]
 ; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[Y:%.*]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i8, ptr [[TMP19]], i32 0
 ; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i8, ptr [[TMP19]], i32 16
-; CHECK-NEXT:    store <16 x i8> [[TMP17]], ptr [[TMP21]], align 1
+; CHECK-NEXT:    store <16 x i8> [[TMP17]], ptr [[TMP19]], align 1
 ; CHECK-NEXT:    store <16 x i8> [[TMP18]], ptr [[TMP22]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
 ; CHECK-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -313,7 +309,7 @@ define void @for_iv_trunc_optimized(ptr %dst) {
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[STEP_ADD]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT3:%.*]] = extractelement <4 x i32> [[TMP3]], i32 3
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 337, [[MIDDLE_BLOCK]] ], [ 1, [[BB:%.*]] ]
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 1, [[BB]] ]
@@ -328,7 +324,7 @@ define void @for_iv_trunc_optimized(ptr %dst) {
 ; CHECK-NEXT:    store i32 [[FOR_2]], ptr [[DST]], align 4
 ; CHECK-NEXT:    [[ICMP:%.*]] = icmp ult i64 [[IV]], 337
 ; CHECK-NEXT:    [[TRUNC]] = trunc i64 [[IV]] to i32
-; CHECK-NEXT:    br i1 [[ICMP]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP9:![0-9]+]]
+; CHECK-NEXT:    br i1 [[ICMP]], label [[LOOP]], label [[EXIT:%.*]], !llvm.loop [[LOOP9:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/X86/fminimumnum.ll b/llvm/test/Transforms/LoopVectorize/X86/fminimumnum.ll
index 1b359f7..21fc8e4 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/fminimumnum.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/fminimumnum.ll
@@ -22,29 +22,26 @@ define void @fmin32(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw float, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw float, ptr [[TMP2]], i32 4
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw float, ptr [[TMP5]], i32 0
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw float, ptr [[TMP5]], i32 4
-; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP6]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x float>, ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = call <4 x float> @llvm.minimumnum.v4f32(<4 x float> [[WIDE_LOAD]], <4 x float> [[WIDE_LOAD6]])
 ; CHECK-NEXT:    [[TMP9:%.*]] = call <4 x float> @llvm.minimumnum.v4f32(<4 x float> [[WIDE_LOAD5]], <4 x float> [[WIDE_LOAD7]])
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw float, ptr [[TMP10]], i32 0
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw float, ptr [[TMP10]], i32 4
-; CHECK-NEXT:    store <4 x float> [[TMP8]], ptr [[TMP11]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP8]], ptr [[TMP10]], align 4
 ; CHECK-NEXT:    store <4 x float> [[TMP9]], ptr [[TMP12]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; CHECK-NEXT:    br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -103,29 +100,26 @@ define void @fmax32(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw float, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw float, ptr [[TMP2]], i32 4
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw float, ptr [[TMP5]], i32 0
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw float, ptr [[TMP5]], i32 4
-; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP6]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x float>, ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = call <4 x float> @llvm.maximumnum.v4f32(<4 x float> [[WIDE_LOAD]], <4 x float> [[WIDE_LOAD6]])
 ; CHECK-NEXT:    [[TMP9:%.*]] = call <4 x float> @llvm.maximumnum.v4f32(<4 x float> [[WIDE_LOAD5]], <4 x float> [[WIDE_LOAD7]])
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [4096 x float], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw float, ptr [[TMP10]], i32 0
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw float, ptr [[TMP10]], i32 4
-; CHECK-NEXT:    store <4 x float> [[TMP8]], ptr [[TMP11]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP8]], ptr [[TMP10]], align 4
 ; CHECK-NEXT:    store <4 x float> [[TMP9]], ptr [[TMP12]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; CHECK-NEXT:    br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -184,29 +178,26 @@ define void @fmin64(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw double, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw double, ptr [[TMP2]], i32 2
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP3]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP2]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <2 x double>, ptr [[TMP4]], align 8
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw double, ptr [[TMP5]], i32 0
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw double, ptr [[TMP5]], i32 2
-; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <2 x double>, ptr [[TMP6]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <2 x double>, ptr [[TMP5]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <2 x double>, ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[TMP8:%.*]] = call <2 x double> @llvm.minimumnum.v2f64(<2 x double> [[WIDE_LOAD]], <2 x double> [[WIDE_LOAD6]])
 ; CHECK-NEXT:    [[TMP9:%.*]] = call <2 x double> @llvm.minimumnum.v2f64(<2 x double> [[WIDE_LOAD5]], <2 x double> [[WIDE_LOAD7]])
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw double, ptr [[TMP10]], i32 0
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw double, ptr [[TMP10]], i32 2
-; CHECK-NEXT:    store <2 x double> [[TMP8]], ptr [[TMP11]], align 8
+; CHECK-NEXT:    store <2 x double> [[TMP8]], ptr [[TMP10]], align 8
 ; CHECK-NEXT:    store <2 x double> [[TMP9]], ptr [[TMP12]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; CHECK-NEXT:    br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -265,29 +256,26 @@ define void @fmax64(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw double, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw double, ptr [[TMP2]], i32 2
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP3]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP2]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <2 x double>, ptr [[TMP4]], align 8
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw double, ptr [[TMP5]], i32 0
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw double, ptr [[TMP5]], i32 2
-; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <2 x double>, ptr [[TMP6]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <2 x double>, ptr [[TMP5]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <2 x double>, ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[TMP8:%.*]] = call <2 x double> @llvm.maximumnum.v2f64(<2 x double> [[WIDE_LOAD]], <2 x double> [[WIDE_LOAD6]])
 ; CHECK-NEXT:    [[TMP9:%.*]] = call <2 x double> @llvm.maximumnum.v2f64(<2 x double> [[WIDE_LOAD5]], <2 x double> [[WIDE_LOAD7]])
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds nuw [4096 x double], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds nuw double, ptr [[TMP10]], i32 0
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds nuw double, ptr [[TMP10]], i32 2
-; CHECK-NEXT:    store <2 x double> [[TMP8]], ptr [[TMP11]], align 8
+; CHECK-NEXT:    store <2 x double> [[TMP8]], ptr [[TMP10]], align 8
 ; CHECK-NEXT:    store <2 x double> [[TMP9]], ptr [[TMP12]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; CHECK-NEXT:    br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -346,22 +334,19 @@ define void @fmin16(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw half, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x half>, ptr [[TMP3]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x half>, ptr [[TMP2]], align 2
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw half, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <8 x half>, ptr [[TMP5]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <8 x half>, ptr [[TMP4]], align 2
 ; CHECK-NEXT:    [[TMP6:%.*]] = call <8 x half> @llvm.minimumnum.v8f16(<8 x half> [[WIDE_LOAD]], <8 x half> [[WIDE_LOAD5]])
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw half, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <8 x half> [[TMP6]], ptr [[TMP8]], align 2
+; CHECK-NEXT:    store <8 x half> [[TMP6]], ptr [[TMP7]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; CHECK-NEXT:    br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
@@ -420,22 +405,19 @@ define void @fmax16(ptr noundef readonly captures(none) %input1, ptr noundef rea
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[INPUT1]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw half, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x half>, ptr [[TMP3]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x half>, ptr [[TMP2]], align 2
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[INPUT2]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds nuw half, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <8 x half>, ptr [[TMP5]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <8 x half>, ptr [[TMP4]], align 2
 ; CHECK-NEXT:    [[TMP6:%.*]] = call <8 x half> @llvm.maximumnum.v8f16(<8 x half> [[WIDE_LOAD]], <8 x half> [[WIDE_LOAD5]])
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw [4096 x half], ptr [[OUTPUT]], i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw half, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <8 x half> [[TMP6]], ptr [[TMP8]], align 2
+; CHECK-NEXT:    store <8 x half> [[TMP6]], ptr [[TMP7]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; CHECK-NEXT:    br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[FOR_BODY]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/gather_scatter.ll b/llvm/test/Transforms/LoopVectorize/X86/gather_scatter.ll
index c4fc609..07ff8e2 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/gather_scatter.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/gather_scatter.ll
@@ -28,19 +28,16 @@ define void @foo1(ptr noalias %in, ptr noalias %out, ptr noalias %trigger, ptr n
 ; AVX512:       vector.body:
 ; AVX512-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX512-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TRIGGER:%.*]], i64 [[INDEX1]]
-; AVX512-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; AVX512-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i32>, ptr [[TMP2]], align 4
+; AVX512-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i32>, ptr [[TMP1]], align 4
 ; AVX512-NEXT:    [[TMP3:%.*]] = icmp sgt <16 x i32> [[WIDE_LOAD]], zeroinitializer
 ; AVX512-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[INDEX:%.*]], i64 [[INDEX1]]
-; AVX512-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[TMP4]], i32 0
-; AVX512-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p0(ptr [[TMP5]], i32 4, <16 x i1> [[TMP3]], <16 x i32> poison)
+; AVX512-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p0(ptr [[TMP4]], i32 4, <16 x i1> [[TMP3]], <16 x i32> poison)
 ; AVX512-NEXT:    [[TMP6:%.*]] = sext <16 x i32> [[WIDE_MASKED_LOAD]] to <16 x i64>
 ; AVX512-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[IN:%.*]], <16 x i64> [[TMP6]]
 ; AVX512-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <16 x float> @llvm.masked.gather.v16f32.v16p0(<16 x ptr> [[TMP7]], i32 4, <16 x i1> [[TMP3]], <16 x float> poison)
 ; AVX512-NEXT:    [[TMP8:%.*]] = fadd <16 x float> [[WIDE_MASKED_GATHER]], splat (float 5.000000e-01)
 ; AVX512-NEXT:    [[TMP9:%.*]] = getelementptr float, ptr [[OUT:%.*]], i64 [[INDEX1]]
-; AVX512-NEXT:    [[TMP10:%.*]] = getelementptr float, ptr [[TMP9]], i32 0
-; AVX512-NEXT:    call void @llvm.masked.store.v16f32.p0(<16 x float> [[TMP8]], ptr [[TMP10]], i32 4, <16 x i1> [[TMP3]])
+; AVX512-NEXT:    call void @llvm.masked.store.v16f32.p0(<16 x float> [[TMP8]], ptr [[TMP9]], i32 4, <16 x i1> [[TMP3]])
 ; AVX512-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX1]], 16
 ; AVX512-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; AVX512-NEXT:    br i1 [[TMP11]], label [[FOR_END:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -53,19 +50,16 @@ define void @foo1(ptr noalias %in, ptr noalias %out, ptr noalias %trigger, ptr n
 ; FVW2:       vector.body:
 ; FVW2-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; FVW2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TRIGGER:%.*]], i64 [[INDEX1]]
-; FVW2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; FVW2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; FVW2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; FVW2-NEXT:    [[TMP3:%.*]] = icmp sgt <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; FVW2-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[INDEX:%.*]], i64 [[INDEX1]]
-; FVW2-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[TMP4]], i32 0
-; FVW2-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <2 x i32> @llvm.masked.load.v2i32.p0(ptr [[TMP5]], i32 4, <2 x i1> [[TMP3]], <2 x i32> poison)
+; FVW2-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <2 x i32> @llvm.masked.load.v2i32.p0(ptr [[TMP4]], i32 4, <2 x i1> [[TMP3]], <2 x i32> poison)
 ; FVW2-NEXT:    [[TMP6:%.*]] = sext <2 x i32> [[WIDE_MASKED_LOAD]] to <2 x i64>
 ; FVW2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[IN:%.*]], <2 x i64> [[TMP6]]
 ; FVW2-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <2 x float> @llvm.masked.gather.v2f32.v2p0(<2 x ptr> [[TMP7]], i32 4, <2 x i1> [[TMP3]], <2 x float> poison)
 ; FVW2-NEXT:    [[TMP8:%.*]] = fadd <2 x float> [[WIDE_MASKED_GATHER]], splat (float 5.000000e-01)
 ; FVW2-NEXT:    [[TMP9:%.*]] = getelementptr float, ptr [[OUT:%.*]], i64 [[INDEX1]]
-; FVW2-NEXT:    [[TMP10:%.*]] = getelementptr float, ptr [[TMP9]], i32 0
-; FVW2-NEXT:    call void @llvm.masked.store.v2f32.p0(<2 x float> [[TMP8]], ptr [[TMP10]], i32 4, <2 x i1> [[TMP3]])
+; FVW2-NEXT:    call void @llvm.masked.store.v2f32.p0(<2 x float> [[TMP8]], ptr [[TMP9]], i32 4, <2 x i1> [[TMP3]])
 ; FVW2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX1]], 2
 ; FVW2-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; FVW2-NEXT:    br i1 [[TMP11]], label [[FOR_END:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -647,11 +641,9 @@ define void @test_gather_not_profitable_pr48429(i32 %d, ptr readonly noalias %pt
 ; AVX512-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 4
 ; AVX512-NEXT:    [[TMP16:%.*]] = getelementptr i8, ptr [[PTR]], i64 [[OFFSET_IDX]]
 ; AVX512-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[IDXPROM]]
-; AVX512-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i32 0
-; AVX512-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x float>, ptr [[TMP18]], align 4, !alias.scope [[META8:![0-9]+]]
+; AVX512-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x float>, ptr [[TMP17]], align 4, !alias.scope [[META8:![0-9]+]]
 ; AVX512-NEXT:    call void @llvm.masked.scatter.v16f32.v16p0(<16 x float> [[WIDE_LOAD]], <16 x ptr> [[TMP14]], i32 4, <16 x i1> splat (i1 true)), !alias.scope [[META11:![0-9]+]], !noalias [[META13:![0-9]+]]
-; AVX512-NEXT:    [[TMP19:%.*]] = getelementptr float, ptr [[TMP16]], i32 0
-; AVX512-NEXT:    [[WIDE_LOAD8:%.*]] = load <16 x float>, ptr [[TMP19]], align 4, !alias.scope [[META15:![0-9]+]]
+; AVX512-NEXT:    [[WIDE_LOAD8:%.*]] = load <16 x float>, ptr [[TMP16]], align 4, !alias.scope [[META15:![0-9]+]]
 ; AVX512-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, <16 x ptr> [[TMP14]], i64 1
 ; AVX512-NEXT:    call void @llvm.masked.scatter.v16f32.v16p0(<16 x float> [[WIDE_LOAD8]], <16 x ptr> [[TMP20]], i32 4, <16 x i1> splat (i1 true)), !alias.scope [[META11]], !noalias [[META13]]
 ; AVX512-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
@@ -686,17 +678,15 @@ define void @test_gather_not_profitable_pr48429(i32 %d, ptr readonly noalias %pt
 ; AVX512-NEXT:    [[OFFSET_IDX21:%.*]] = mul i64 [[INDEX18]], 4
 ; AVX512-NEXT:    [[TMP28:%.*]] = getelementptr i8, ptr [[PTR]], i64 [[OFFSET_IDX21]]
 ; AVX512-NEXT:    [[TMP29:%.*]] = getelementptr inbounds float, ptr [[TMP28]], i64 [[IDXPROM]]
-; AVX512-NEXT:    [[TMP30:%.*]] = getelementptr inbounds float, ptr [[TMP29]], i32 0
-; AVX512-NEXT:    [[WIDE_LOAD17:%.*]] = load <8 x float>, ptr [[TMP30]], align 4, !alias.scope [[META17:![0-9]+]]
-; AVX512-NEXT:    call void @llvm.masked.scatter.v8f32.v8p0(<8 x float> [[WIDE_LOAD17]], <8 x ptr> [[TMP26]], i32 4, <8 x i1> splat (i1 true)), !alias.scope [[META20:![0-9]+]], !noalias [[META22:![0-9]+]]
-; AVX512-NEXT:    [[TMP31:%.*]] = getelementptr float, ptr [[TMP28]], i32 0
-; AVX512-NEXT:    [[WIDE_LOAD18:%.*]] = load <8 x float>, ptr [[TMP31]], align 4, !alias.scope [[META24:![0-9]+]]
+; AVX512-NEXT:    [[WIDE_LOAD16:%.*]] = load <8 x float>, ptr [[TMP29]], align 4, !alias.scope [[META8]]
+; AVX512-NEXT:    call void @llvm.masked.scatter.v8f32.v8p0(<8 x float> [[WIDE_LOAD16]], <8 x ptr> [[TMP26]], i32 4, <8 x i1> splat (i1 true)), !alias.scope [[META11]], !noalias [[META13]]
+; AVX512-NEXT:    [[WIDE_LOAD17:%.*]] = load <8 x float>, ptr [[TMP28]], align 4, !alias.scope [[META15]]
 ; AVX512-NEXT:    [[TMP32:%.*]] = getelementptr inbounds float, <8 x ptr> [[TMP26]], i64 1
-; AVX512-NEXT:    call void @llvm.masked.scatter.v8f32.v8p0(<8 x float> [[WIDE_LOAD18]], <8 x ptr> [[TMP32]], i32 4, <8 x i1> splat (i1 true)), !alias.scope [[META20]], !noalias [[META22]]
+; AVX512-NEXT:    call void @llvm.masked.scatter.v8f32.v8p0(<8 x float> [[WIDE_LOAD17]], <8 x ptr> [[TMP32]], i32 4, <8 x i1> splat (i1 true)), !alias.scope [[META11]], !noalias [[META13]]
 ; AVX512-NEXT:    [[INDEX_NEXT24]] = add nuw i64 [[INDEX18]], 8
 ; AVX512-NEXT:    [[PTR_IND20]] = getelementptr i8, ptr [[POINTER_PHI19]], i64 512
 ; AVX512-NEXT:    [[TMP33:%.*]] = icmp eq i64 [[INDEX_NEXT24]], [[N_VEC10]]
-; AVX512-NEXT:    br i1 [[TMP33]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP25:![0-9]+]]
+; AVX512-NEXT:    br i1 [[TMP33]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
 ; AVX512:       vec.epilog.middle.block:
 ; AVX512-NEXT:    [[CMP_N17:%.*]] = icmp eq i64 [[TMP3]], [[N_VEC10]]
 ; AVX512-NEXT:    br i1 [[CMP_N17]], label [[FOR_END]], label [[VEC_EPILOG_SCALAR_PH]]
@@ -716,7 +706,7 @@ define void @test_gather_not_profitable_pr48429(i32 %d, ptr readonly noalias %pt
 ; AVX512-NEXT:    [[INCDEC_PTR]] = getelementptr inbounds float, ptr [[PTR_ADDR_012]], i64 1
 ; AVX512-NEXT:    [[ADD_PTR6]] = getelementptr inbounds float, ptr [[DEST_ADDR_011]], i64 16
 ; AVX512-NEXT:    [[CMP_NOT:%.*]] = icmp eq ptr [[INCDEC_PTR]], [[ADD_PTR]]
-; AVX512-NEXT:    br i1 [[CMP_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
+; AVX512-NEXT:    br i1 [[CMP_NOT]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; AVX512:       for.end:
 ; AVX512-NEXT:    ret void
 ;
@@ -775,14 +765,12 @@ define void @test_gather_not_profitable_pr48429(i32 %d, ptr readonly noalias %pt
 ; FVW2-NEXT:    [[TMP19:%.*]] = getelementptr i8, ptr [[DEST]], i64 [[TMP17]]
 ; FVW2-NEXT:    [[TMP20:%.*]] = getelementptr i8, ptr [[DEST]], i64 [[TMP18]]
 ; FVW2-NEXT:    [[TMP21:%.*]] = getelementptr inbounds float, ptr [[TMP16]], i64 [[IDXPROM]]
-; FVW2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds float, ptr [[TMP21]], i32 0
-; FVW2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP22]], align 4, !alias.scope [[META8:![0-9]+]]
+; FVW2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP21]], align 4, !alias.scope [[META8:![0-9]+]]
 ; FVW2-NEXT:    [[TMP23:%.*]] = extractelement <2 x float> [[WIDE_LOAD]], i32 0
 ; FVW2-NEXT:    store float [[TMP23]], ptr [[TMP19]], align 4, !alias.scope [[META11:![0-9]+]], !noalias [[META13:![0-9]+]]
 ; FVW2-NEXT:    [[TMP24:%.*]] = extractelement <2 x float> [[WIDE_LOAD]], i32 1
 ; FVW2-NEXT:    store float [[TMP24]], ptr [[TMP20]], align 4, !alias.scope [[META11]], !noalias [[META13]]
-; FVW2-NEXT:    [[TMP25:%.*]] = getelementptr float, ptr [[TMP16]], i32 0
-; FVW2-NEXT:    [[WIDE_LOAD10:%.*]] = load <2 x float>, ptr [[TMP25]], align 4, !alias.scope [[META15:![0-9]+]]
+; FVW2-NEXT:    [[WIDE_LOAD10:%.*]] = load <2 x float>, ptr [[TMP16]], align 4, !alias.scope [[META15:![0-9]+]]
 ; FVW2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds float, ptr [[TMP19]], i64 1
 ; FVW2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds float, ptr [[TMP20]], i64 1
 ; FVW2-NEXT:    [[TMP28:%.*]] = extractelement <2 x float> [[WIDE_LOAD10]], i32 0
diff --git a/llvm/test/Transforms/LoopVectorize/X86/gep-use-outside-loop.ll b/llvm/test/Transforms/LoopVectorize/X86/gep-use-outside-loop.ll
index 929a2e5..6938ffb 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/gep-use-outside-loop.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/gep-use-outside-loop.ll
@@ -12,17 +12,15 @@ define void @gep_use_in_dead_block(ptr noalias %dst, ptr %src) {
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i16, ptr [[SRC]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i16, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP5]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP4]], align 2
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp ne <4 x i16> [[WIDE_LOAD]], splat (i16 10)
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i16, ptr [[DST]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i16, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v4i16.p0(<4 x i16> zeroinitializer, ptr [[TMP12]], i32 2, <4 x i1> [[TMP7]])
+; CHECK-NEXT:    call void @llvm.masked.store.v4i16.p0(<4 x i16> zeroinitializer, ptr [[TMP8]], i32 2, <4 x i1> [[TMP7]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 4
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], 96
 ; CHECK-NEXT:    br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 96, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
@@ -42,7 +40,7 @@ define void @gep_use_in_dead_block(ptr noalias %dst, ptr %src) {
 ; CHECK:       [[LOOP_LATCH]]:
 ; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; CHECK-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], 99
-; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    ret void
 ;
@@ -86,19 +84,16 @@ define void @gep_use_outside_loop(ptr noalias %dst, ptr %src) {
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i16, ptr [[DST]], <4 x i64> [[VEC_IND]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i16, ptr [[SRC]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i16, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP3]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP2]], align 2
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp ne <4 x i16> [[WIDE_LOAD]], splat (i16 10)
 ; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <4 x ptr> [[TMP1]], i32 0
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i16, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v4i16.p0(<4 x i16> zeroinitializer, ptr [[TMP7]], i32 2, <4 x i1> [[TMP5]])
+; CHECK-NEXT:    call void @llvm.masked.store.v4i16.p0(<4 x i16> zeroinitializer, ptr [[TMP6]], i32 2, <4 x i1> [[TMP5]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 96
 ; CHECK-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <4 x ptr> [[TMP1]], i32 3
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 96, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
@@ -115,9 +110,9 @@ define void @gep_use_outside_loop(ptr noalias %dst, ptr %src) {
 ; CHECK:       [[LOOP_LATCH]]:
 ; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; CHECK-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV_NEXT]], 99
-; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       [[EXIT]]:
-; CHECK-NEXT:    [[GEP_DST_LCSSA:%.*]] = phi ptr [ [[GEP_DST]], %[[LOOP_LATCH]] ], [ [[TMP9]], %[[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[GEP_DST_LCSSA:%.*]] = phi ptr [ [[GEP_DST]], %[[LOOP_LATCH]] ]
 ; CHECK-NEXT:    store i16 0, ptr [[GEP_DST_LCSSA]], align 2
 ; CHECK-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/X86/illegal-parallel-loop-uniform-write.ll b/llvm/test/Transforms/LoopVectorize/X86/illegal-parallel-loop-uniform-write.ll
index 4ed5489..9168ebf 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/illegal-parallel-loop-uniform-write.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/illegal-parallel-loop-uniform-write.ll
@@ -74,8 +74,7 @@ define void @foo(ptr nocapture %a, ptr nocapture %b, i32 %k, i32 %m) #0 {
 ; CHECK-NEXT:    [[TMP13:%.*]] = add i32 [[ADD_US]], [[TMP11]]
 ; CHECK-NEXT:    [[TMP14:%.*]] = sext i32 [[TMP13]] to i64
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP14]]
-; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[TMP15]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP16]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP15]], align 4
 ; CHECK-NEXT:    [[TMP17:%.*]] = add nsw <4 x i32> [[WIDE_LOAD]], splat (i32 1)
 ; CHECK-NEXT:    [[TMP18:%.*]] = extractelement <4 x i32> [[TMP17]], i32 3
 ; CHECK-NEXT:    store i32 [[TMP18]], ptr [[ARRAYIDX7_US]], align 4, !llvm.mem.parallel_loop_access [[META0]]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/imprecise-through-phis.ll b/llvm/test/Transforms/LoopVectorize/X86/imprecise-through-phis.ll
index a62e8ff..7fe4c14 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/imprecise-through-phis.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/imprecise-through-phis.ll
@@ -75,9 +75,8 @@ define double @sumIfVector(ptr nocapture readonly %arr) {
 ; SSE-NEXT:    [[VEC_PHI:%.*]] = phi <2 x double> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PREDPHI:%.*]], [[VECTOR_BODY]] ]
 ; SSE-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x double> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PREDPHI3:%.*]], [[VECTOR_BODY]] ]
 ; SSE-NEXT:    [[TMP2:%.*]] = getelementptr double, ptr [[ARR:%.*]], i32 [[INDEX]]
-; SSE-NEXT:    [[TMP4:%.*]] = getelementptr double, ptr [[TMP2]], i32 0
 ; SSE-NEXT:    [[TMP5:%.*]] = getelementptr double, ptr [[TMP2]], i32 2
-; SSE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP4]], align 8
+; SSE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP2]], align 8
 ; SSE-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x double>, ptr [[TMP5]], align 8
 ; SSE-NEXT:    [[TMP6:%.*]] = fcmp fast une <2 x double> [[WIDE_LOAD]], splat (double 4.200000e+01)
 ; SSE-NEXT:    [[TMP7:%.*]] = fcmp fast une <2 x double> [[WIDE_LOAD2]], splat (double 4.200000e+01)
@@ -91,10 +90,10 @@ define double @sumIfVector(ptr nocapture readonly %arr) {
 ; SSE:       middle.block:
 ; SSE-NEXT:    [[BIN_RDX:%.*]] = fadd fast <2 x double> [[PREDPHI3]], [[PREDPHI]]
 ; SSE-NEXT:    [[TMP11:%.*]] = call fast double @llvm.vector.reduce.fadd.v2f64(double 0.000000e+00, <2 x double> [[BIN_RDX]])
-; SSE-NEXT:    br i1 true, label [[DONE:%.*]], label [[SCALAR_PH]]
+; SSE-NEXT:    br label [[DONE:%.*]]
 ; SSE:       scalar.ph:
-; SSE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 32, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; SSE-NEXT:    [[BC_MERGE_RDX:%.*]] = phi double [ [[TMP11]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
+; SSE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ]
+; SSE-NEXT:    [[BC_MERGE_RDX:%.*]] = phi double [ 0.000000e+00, [[ENTRY]] ]
 ; SSE-NEXT:    br label [[LOOP:%.*]]
 ; SSE:       loop:
 ; SSE-NEXT:    [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[NEXT_ITER:%.*]] ]
@@ -129,11 +128,10 @@ define double @sumIfVector(ptr nocapture readonly %arr) {
 ; AVX-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x double> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PREDPHI8:%.*]], [[VECTOR_BODY]] ]
 ; AVX-NEXT:    [[VEC_PHI3:%.*]] = phi <4 x double> [ zeroinitializer, [[VECTOR_PH]] ], [ [[PREDPHI9:%.*]], [[VECTOR_BODY]] ]
 ; AVX-NEXT:    [[TMP4:%.*]] = getelementptr double, ptr [[ARR:%.*]], i32 [[INDEX]]
-; AVX-NEXT:    [[TMP8:%.*]] = getelementptr double, ptr [[TMP4]], i32 0
 ; AVX-NEXT:    [[TMP9:%.*]] = getelementptr double, ptr [[TMP4]], i32 4
 ; AVX-NEXT:    [[TMP10:%.*]] = getelementptr double, ptr [[TMP4]], i32 8
 ; AVX-NEXT:    [[TMP11:%.*]] = getelementptr double, ptr [[TMP4]], i32 12
-; AVX-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP8]], align 8
+; AVX-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP4]], align 8
 ; AVX-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x double>, ptr [[TMP9]], align 8
 ; AVX-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x double>, ptr [[TMP10]], align 8
 ; AVX-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x double>, ptr [[TMP11]], align 8
@@ -157,10 +155,10 @@ define double @sumIfVector(ptr nocapture readonly %arr) {
 ; AVX-NEXT:    [[BIN_RDX10:%.*]] = fadd fast <4 x double> [[PREDPHI8]], [[BIN_RDX]]
 ; AVX-NEXT:    [[BIN_RDX11:%.*]] = fadd fast <4 x double> [[PREDPHI9]], [[BIN_RDX10]]
 ; AVX-NEXT:    [[TMP21:%.*]] = call fast double @llvm.vector.reduce.fadd.v4f64(double 0.000000e+00, <4 x double> [[BIN_RDX11]])
-; AVX-NEXT:    br i1 true, label [[DONE:%.*]], label [[SCALAR_PH]]
+; AVX-NEXT:    br label [[DONE:%.*]]
 ; AVX:       scalar.ph:
-; AVX-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 32, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; AVX-NEXT:    [[BC_MERGE_RDX:%.*]] = phi double [ [[TMP21]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
+; AVX-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ]
+; AVX-NEXT:    [[BC_MERGE_RDX:%.*]] = phi double [ 0.000000e+00, [[ENTRY]] ]
 ; AVX-NEXT:    br label [[LOOP:%.*]]
 ; AVX:       loop:
 ; AVX-NEXT:    [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[NEXT_ITER:%.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/induction-costs.ll b/llvm/test/Transforms/LoopVectorize/X86/induction-costs.ll
index d445e0d..fcd94f4 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/induction-costs.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/induction-costs.ll
@@ -126,14 +126,12 @@ define void @multiple_truncated_ivs_with_wide_uses(i1 %c, ptr %A, ptr %B) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = select i1 [[C]], <4 x i16> [[VEC_IND]], <4 x i16> splat (i16 10)
 ; CHECK-NEXT:    [[TMP2:%.*]] = select i1 [[C]], <4 x i16> [[STEP_ADD]], <4 x i16> splat (i16 10)
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i16, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i16, ptr [[TMP4]], i32 0
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i16, ptr [[TMP4]], i32 4
-; CHECK-NEXT:    store <4 x i16> [[TMP1]], ptr [[TMP6]], align 2, !alias.scope [[META5:![0-9]+]], !noalias [[META8:![0-9]+]]
+; CHECK-NEXT:    store <4 x i16> [[TMP1]], ptr [[TMP4]], align 2, !alias.scope [[META5:![0-9]+]], !noalias [[META8:![0-9]+]]
 ; CHECK-NEXT:    store <4 x i16> [[TMP2]], ptr [[TMP7]], align 2, !alias.scope [[META5]], !noalias [[META8]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[TMP8]], i32 0
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[TMP8]], i32 4
-; CHECK-NEXT:    store <4 x i32> [[VEC_IND3]], ptr [[TMP10]], align 4, !alias.scope [[META8]]
+; CHECK-NEXT:    store <4 x i32> [[VEC_IND3]], ptr [[TMP8]], align 4, !alias.scope [[META8]]
 ; CHECK-NEXT:    store <4 x i32> [[STEP_ADD4]], ptr [[TMP11]], align 4, !alias.scope [[META8]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[STEP_ADD]], splat (i16 4)
@@ -141,7 +139,7 @@ define void @multiple_truncated_ivs_with_wide_uses(i1 %c, ptr %A, ptr %B) {
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 64, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
@@ -156,7 +154,7 @@ define void @multiple_truncated_ivs_with_wide_uses(i1 %c, ptr %A, ptr %B) {
 ; CHECK-NEXT:    store i32 [[IV_32]], ptr [[GEP_B]], align 4
 ; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; CHECK-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV]], 64
-; CHECK-NEXT:    br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP11:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP11:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
@@ -195,16 +193,15 @@ define void @truncated_ivs_with_wide_and_scalar_uses(i1 %c, ptr %dst) {
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i16, ptr [[DST]], i32 [[TMP0]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = select i1 [[C]], <8 x i16> [[VEC_IND]], <8 x i16> splat (i16 10)
 ; CHECK-NEXT:    [[TMP6:%.*]] = select i1 [[C]], <8 x i16> [[STEP_ADD]], <8 x i16> splat (i16 10)
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i16, ptr [[TMP3]], i32 0
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i16, ptr [[TMP3]], i32 8
-; CHECK-NEXT:    store <8 x i16> [[TMP5]], ptr [[TMP7]], align 2
+; CHECK-NEXT:    store <8 x i16> [[TMP5]], ptr [[TMP3]], align 2
 ; CHECK-NEXT:    store <8 x i16> [[TMP6]], ptr [[TMP8]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <8 x i16> [[STEP_ADD]], splat (i16 8)
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 64, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
@@ -217,7 +214,7 @@ define void @truncated_ivs_with_wide_and_scalar_uses(i1 %c, ptr %dst) {
 ; CHECK-NEXT:    store i16 [[SEL]], ptr [[GEP]], align 2
 ; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; CHECK-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV]], 64
-; CHECK-NEXT:    br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP13:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP13:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
@@ -293,10 +290,9 @@ define void @multiple_pointer_ivs_with_scalar_uses_only(ptr %A, ptr %B) #0 {
 ; CHECK-NEXT:    [[NEXT_GEP20:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP14]]
 ; CHECK-NEXT:    [[NEXT_GEP21:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP15]]
 ; CHECK-NEXT:    [[NEXT_GEP23:%.*]] = getelementptr i8, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr i8, ptr [[NEXT_GEP23]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP18]], align 1, !alias.scope [[META14:![0-9]+]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[NEXT_GEP23]], align 1, !alias.scope [[META14:![0-9]+]]
 ; CHECK-NEXT:    [[TMP19:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i32>
-; CHECK-NEXT:    [[WIDE_LOAD24:%.*]] = load <16 x i8>, ptr [[TMP18]], align 1, !alias.scope [[META14]]
+; CHECK-NEXT:    [[WIDE_LOAD24:%.*]] = load <16 x i8>, ptr [[NEXT_GEP23]], align 1, !alias.scope [[META14]]
 ; CHECK-NEXT:    [[TMP21:%.*]] = zext <16 x i8> [[WIDE_LOAD24]] to <16 x i32>
 ; CHECK-NEXT:    [[TMP22]] = add <16 x i32> [[TMP19]], [[TMP21]]
 ; CHECK-NEXT:    [[TMP23:%.*]] = shufflevector <16 x i32> [[VECTOR_RECUR]], <16 x i32> [[TMP22]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
@@ -339,7 +335,7 @@ define void @multiple_pointer_ivs_with_scalar_uses_only(ptr %A, ptr %B) #0 {
 ; CHECK-NEXT:    br i1 [[TMP42]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <16 x i32> [[TMP22]], i32 15
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ -12, [[MIDDLE_BLOCK]] ], [ 100, [[ENTRY:%.*]] ], [ 100, [[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 2048, [[ENTRY]] ], [ 2048, [[VECTOR_MEMCHECK]] ]
@@ -366,7 +362,7 @@ define void @multiple_pointer_ivs_with_scalar_uses_only(ptr %A, ptr %B) #0 {
 ; CHECK-NEXT:    [[DEC]] = add i32 [[IV_1]], 1
 ; CHECK-NEXT:    [[OUTPTR_0]] = getelementptr i8, ptr [[PTR_IV_1]], i64 2
 ; CHECK-NEXT:    [[CMP30_NOT:%.*]] = icmp eq i32 [[DEC]], 0
-; CHECK-NEXT:    br i1 [[CMP30_NOT]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP20:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP30_NOT]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP20:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
@@ -416,12 +412,11 @@ define i16 @iv_and_step_trunc() {
 ; CHECK-NEXT:    [[VEC_IND_NEXT2]] = add <2 x i16> [[VEC_IND1]], splat (i16 2)
 ; CHECK-NEXT:    br i1 true, label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i16> [[TMP2]], i32 1
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <2 x i16> [[TMP2]], i32 0
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 2, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -485,11 +480,10 @@ define i32 @test_scalar_predicated_cost(i64 %x, i64 %y, ptr %A) #0 {
 ; CHECK-NEXT:    [[TMP21:%.*]] = trunc <8 x i64> [[TMP13]] to <8 x i32>
 ; CHECK-NEXT:    [[TMP22:%.*]] = trunc <8 x i64> [[TMP14]] to <8 x i32>
 ; CHECK-NEXT:    [[TMP23:%.*]] = trunc <8 x i64> [[TMP15]] to <8 x i32>
-; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr i32, ptr [[TMP16]], i32 0
 ; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr i32, ptr [[TMP16]], i32 8
 ; CHECK-NEXT:    [[TMP26:%.*]] = getelementptr i32, ptr [[TMP16]], i32 16
 ; CHECK-NEXT:    [[TMP27:%.*]] = getelementptr i32, ptr [[TMP16]], i32 24
-; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP20]], ptr [[TMP24]], i32 4, <8 x i1> [[TMP8]])
+; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP20]], ptr [[TMP16]], i32 4, <8 x i1> [[TMP8]])
 ; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP21]], ptr [[TMP25]], i32 4, <8 x i1> [[TMP9]])
 ; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP22]], ptr [[TMP26]], i32 4, <8 x i1> [[TMP10]])
 ; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP23]], ptr [[TMP27]], i32 4, <8 x i1> [[TMP11]])
@@ -518,8 +512,7 @@ define i32 @test_scalar_predicated_cost(i64 %x, i64 %y, ptr %A) #0 {
 ; CHECK-NEXT:    [[TMP34:%.*]] = or <4 x i64> [[BROADCAST_SPLAT10]], [[VEC_IND5]]
 ; CHECK-NEXT:    [[TMP35:%.*]] = getelementptr i32, ptr [[A]], i64 [[INDEX4]]
 ; CHECK-NEXT:    [[TMP36:%.*]] = trunc <4 x i64> [[TMP34]] to <4 x i32>
-; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr i32, ptr [[TMP35]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[TMP36]], ptr [[TMP29]], i32 4, <4 x i1> [[TMP33]])
+; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[TMP36]], ptr [[TMP35]], i32 4, <4 x i1> [[TMP33]])
 ; CHECK-NEXT:    [[INDEX_NEXT11]] = add nuw i64 [[INDEX4]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT6]] = add <4 x i64> [[VEC_IND5]], splat (i64 4)
 ; CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT11]], 100
@@ -715,14 +708,13 @@ define void @wombat(i32 %arg, ptr %dst) #1 {
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 4, [[INDEX]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[DST]], i64 [[OFFSET_IDX]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = and <8 x i32> [[VEC_IND]], splat (i32 12)
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i32, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <8 x i32> [[TMP5]], ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <8 x i32> [[TMP5]], ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <8 x i32> [[VEC_IND]], [[DOTSPLAT4]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 56
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 60, [[MIDDLE_BLOCK]] ], [ 4, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL5:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[MUL]], [[ENTRY]] ]
@@ -737,7 +729,7 @@ define void @wombat(i32 %arg, ptr %dst) #1 {
 ; CHECK-NEXT:    [[ADD]] = add i64 [[PHI]], 1
 ; CHECK-NEXT:    [[ICMP:%.*]] = icmp ugt i64 [[PHI]], 65
 ; CHECK-NEXT:    [[TRUNC]] = trunc i64 [[MUL3]] to i32
-; CHECK-NEXT:    br i1 [[ICMP]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP29:![0-9]+]]
+; CHECK-NEXT:    br i1 [[ICMP]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP29:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
@@ -788,14 +780,13 @@ define void @wombat2(i32 %arg, ptr %dst) #1 {
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 4, [[INDEX]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[DST]], i64 [[OFFSET_IDX]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = and <8 x i32> [[VEC_IND]], splat (i32 12)
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i32, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <8 x i32> [[TMP5]], ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <8 x i32> [[TMP5]], ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <8 x i32> [[VEC_IND]], [[DOTSPLAT4]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 56
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 60, [[MIDDLE_BLOCK]] ], [ 4, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL5:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[MUL]], [[ENTRY]] ]
@@ -811,7 +802,7 @@ define void @wombat2(i32 %arg, ptr %dst) #1 {
 ; CHECK-NEXT:    [[ICMP:%.*]] = icmp ugt i64 [[PHI]], 65
 ; CHECK-NEXT:    [[TRUNC_0:%.*]] = trunc i64 [[MUL3]] to i60
 ; CHECK-NEXT:    [[TRUNC_1]] = trunc i60 [[TRUNC_0]] to i32
-; CHECK-NEXT:    br i1 [[ICMP]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP31:![0-9]+]]
+; CHECK-NEXT:    br i1 [[ICMP]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP31:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
@@ -864,14 +855,13 @@ define void @with_dead_use(i32 %arg, ptr %dst) #1 {
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 4, [[INDEX]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[DST]], i64 [[OFFSET_IDX]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = and <8 x i32> [[VEC_IND]], splat (i32 12)
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i32, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <8 x i32> [[TMP5]], ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <8 x i32> [[TMP5]], ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <8 x i32> [[VEC_IND]], [[DOTSPLAT4]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 56
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 60, [[MIDDLE_BLOCK]] ], [ 4, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL5:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[MUL]], [[ENTRY]] ]
@@ -887,7 +877,7 @@ define void @with_dead_use(i32 %arg, ptr %dst) #1 {
 ; CHECK-NEXT:    [[ICMP:%.*]] = icmp ugt i64 [[PHI]], 65
 ; CHECK-NEXT:    [[TRUNC]] = trunc i64 [[MUL3]] to i32
 ; CHECK-NEXT:    [[DEAD_AND:%.*]] = and i32 [[TRUNC]], 123
-; CHECK-NEXT:    br i1 [[ICMP]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP33:![0-9]+]]
+; CHECK-NEXT:    br i1 [[ICMP]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP33:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/X86/induction-step.ll b/llvm/test/Transforms/LoopVectorize/X86/induction-step.ll
index 196c755..fea027d 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/induction-step.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/induction-step.ll
@@ -29,9 +29,8 @@ define i16 @wide_add_induction_step_live_in(ptr %dst, i64 %N, i16 %off) {
 ; CHECK-NEXT:    [[TMP4:%.*]] = add <4 x i16> [[VEC_IND]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = add <4 x i16> [[STEP_ADD]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[DST:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[TMP5]], i32 0
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[TMP5]], i32 4
-; CHECK-NEXT:    store <4 x i16> [[TMP4]], ptr [[TMP6]], align 2
+; CHECK-NEXT:    store <4 x i16> [[TMP4]], ptr [[TMP5]], align 2
 ; CHECK-NEXT:    store <4 x i16> [[TMP9]], ptr [[TMP8]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[STEP_ADD]], [[TMP1]]
@@ -104,9 +103,8 @@ define i16 @wide_sub_induction_step_live_in(ptr %dst, i64 %N, i16 %off) {
 ; CHECK-NEXT:    [[TMP5:%.*]] = sub <4 x i16> [[VEC_IND]], [[BROADCAST_SPLAT2]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = sub <4 x i16> [[STEP_ADD]], [[BROADCAST_SPLAT2]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[DST:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i16, ptr [[TMP6]], i32 0
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i16, ptr [[TMP6]], i32 4
-; CHECK-NEXT:    store <4 x i16> [[TMP5]], ptr [[TMP7]], align 2
+; CHECK-NEXT:    store <4 x i16> [[TMP5]], ptr [[TMP6]], align 2
 ; CHECK-NEXT:    store <4 x i16> [[TMP10]], ptr [[TMP9]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[STEP_ADD]], [[TMP2]]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/interleave-cost.ll b/llvm/test/Transforms/LoopVectorize/X86/interleave-cost.ll
index 7f2544d..6d562be 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/interleave-cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/interleave-cost.ll
@@ -37,9 +37,9 @@ define void @test_free_instructions_feeding_geps_for_interleave_groups(ptr noali
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -523,9 +523,9 @@ define void @interleave_store_double_i64(ptr %dst) {
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 2, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -644,9 +644,9 @@ define void @interleave_store_i64_double_2(ptr %dst) {
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 2, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/interleave-ptradd-with-replicated-operand.ll b/llvm/test/Transforms/LoopVectorize/X86/interleave-ptradd-with-replicated-operand.ll
index 95258e6..7d018ea 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/interleave-ptradd-with-replicated-operand.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/interleave-ptradd-with-replicated-operand.ll
@@ -52,7 +52,6 @@ define ptr @test_interleave_ptradd_with_replicated_op(ptr %m) #0 {
 ; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr i8, ptr [[NEXT_GEP13]], i64 4
 ; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr i8, ptr [[NEXT_GEP14]], i64 4
 ; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr i8, ptr [[NEXT_GEP15]], i64 4
-; CHECK-NEXT:    [[TMP31:%.*]] = getelementptr i8, ptr [[NEXT_GEP16]], i64 4
 ; CHECK-NEXT:    [[TMP32:%.*]] = getelementptr i8, ptr [[TMP27]], i32 -4
 ; CHECK-NEXT:    [[TMP33:%.*]] = getelementptr i8, ptr [[TMP28]], i32 -4
 ; CHECK-NEXT:    [[TMP34:%.*]] = getelementptr i8, ptr [[TMP29]], i32 -4
@@ -109,7 +108,7 @@ define ptr @test_interleave_ptradd_with_replicated_op(ptr %m) #0 {
 ; CHECK-NEXT:    [[TMP56:%.*]] = icmp eq i64 [[INDEX_NEXT]], 96
 ; CHECK-NEXT:    br i1 [[TMP56]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi ptr [ [[IND_END]], %[[MIDDLE_BLOCK]] ], [ [[M]], %[[ENTRY]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL26:%.*]] = phi i32 [ 97, %[[MIDDLE_BLOCK]] ], [ 1, %[[ENTRY]] ]
@@ -126,9 +125,9 @@ define ptr @test_interleave_ptradd_with_replicated_op(ptr %m) #0 {
 ; CHECK-NEXT:    store i32 [[ADD]], ptr [[PTR_IV]], align 4
 ; CHECK-NEXT:    [[IV_NEXT]] = add i32 [[IV]], 1
 ; CHECK-NEXT:    [[TOBOOL_NOT:%.*]] = icmp eq i32 [[IV]], 100
-; CHECK-NEXT:    br i1 [[TOBOOL_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[TOBOOL_NOT]], label %[[EXIT:.*]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       [[EXIT]]:
-; CHECK-NEXT:    [[P_4_LCSSA:%.*]] = phi ptr [ [[P_4]], %[[LOOP]] ], [ [[TMP31]], %[[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[P_4_LCSSA:%.*]] = phi ptr [ [[P_4]], %[[LOOP]] ]
 ; CHECK-NEXT:    ret ptr [[P_4_LCSSA]]
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/X86/interleaving.ll b/llvm/test/Transforms/LoopVectorize/X86/interleaving.ll
index 3d5e9a6..e3e3944 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/interleaving.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/interleaving.ll
@@ -34,7 +34,7 @@ define void @foo(ptr noalias nocapture %a, ptr noalias nocapture readonly %b) {
 ; SSE-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; SSE-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; SSE:       middle.block:
-; SSE-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; SSE-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; SSE:       scalar.ph:
 ; SSE-NEXT:    br label [[FOR_BODY:%.*]]
 ; SSE:       for.cond.cleanup:
@@ -88,7 +88,7 @@ define void @foo(ptr noalias nocapture %a, ptr noalias nocapture readonly %b) {
 ; AVX1-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; AVX1-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; AVX1:       middle.block:
-; AVX1-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; AVX1-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; AVX1:       scalar.ph:
 ; AVX1-NEXT:    br label [[FOR_BODY:%.*]]
 ; AVX1:       for.cond.cleanup:
@@ -142,7 +142,7 @@ define void @foo(ptr noalias nocapture %a, ptr noalias nocapture readonly %b) {
 ; AVX2-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; AVX2-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; AVX2:       middle.block:
-; AVX2-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; AVX2-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; AVX2:       scalar.ph:
 ; AVX2-NEXT:    br label [[FOR_BODY:%.*]]
 ; AVX2:       for.cond.cleanup:
diff --git a/llvm/test/Transforms/LoopVectorize/X86/iv-live-outs.ll b/llvm/test/Transforms/LoopVectorize/X86/iv-live-outs.ll
index 4ddee9b..78f96ca 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/iv-live-outs.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/iv-live-outs.ll
@@ -18,11 +18,10 @@ define i64 @test_pr98660(ptr %dst, i64 %N) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = add i64 [[INDEX]], 1
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr i32, ptr [[DST]], i64 [[TMP5]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr i32, ptr [[TMP9]], i32 0
 ; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr i32, ptr [[TMP9]], i32 8
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr i32, ptr [[TMP9]], i32 16
 ; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr i32, ptr [[TMP9]], i32 24
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP13]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP9]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i32>, ptr [[TMP14]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i32>, ptr [[TMP15]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i32>, ptr [[TMP16]], align 4
@@ -30,7 +29,7 @@ define i64 @test_pr98660(ptr %dst, i64 %N) {
 ; CHECK-NEXT:    [[TMP18:%.*]] = icmp eq <8 x i32> [[WIDE_LOAD1]], zeroinitializer
 ; CHECK-NEXT:    [[TMP19:%.*]] = icmp eq <8 x i32> [[WIDE_LOAD2]], zeroinitializer
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq <8 x i32> [[WIDE_LOAD3]], zeroinitializer
-; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> zeroinitializer, ptr [[TMP13]], i32 4, <8 x i1> [[TMP17]])
+; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> zeroinitializer, ptr [[TMP9]], i32 4, <8 x i1> [[TMP17]])
 ; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> zeroinitializer, ptr [[TMP14]], i32 4, <8 x i1> [[TMP18]])
 ; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> zeroinitializer, ptr [[TMP15]], i32 4, <8 x i1> [[TMP19]])
 ; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> zeroinitializer, ptr [[TMP16]], i32 4, <8 x i1> [[TMP20]])
diff --git a/llvm/test/Transforms/LoopVectorize/X86/limit-vf-by-tripcount.ll b/llvm/test/Transforms/LoopVectorize/X86/limit-vf-by-tripcount.ll
index 2871d4d..1a385b6 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/limit-vf-by-tripcount.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/limit-vf-by-tripcount.ll
@@ -14,16 +14,14 @@ define void @test_tc_17_no_epilogue_vectorization(ptr noalias %src, ptr noalias
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[SRC:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 64
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 64
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[DST:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    store <16 x i8> [[WIDE_LOAD]], ptr [[TMP4]], align 64
+; CHECK-NEXT:    store <16 x i8> [[WIDE_LOAD]], ptr [[TMP3]], align 64
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 16
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 16, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
@@ -35,7 +33,7 @@ define void @test_tc_17_no_epilogue_vectorization(ptr noalias %src, ptr noalias
 ; CHECK-NEXT:    store i8 [[VAL]], ptr [[STADDR]], align 64
 ; CHECK-NEXT:    [[I_NEXT]] = add i64 [[I]], 1
 ; CHECK-NEXT:    [[IS_NEXT:%.*]] = icmp ult i64 [[I_NEXT]], 17
-; CHECK-NEXT:    br i1 [[IS_NEXT]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[IS_NEXT]], label [[LOOP]], label [[EXIT:%.*]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
@@ -67,11 +65,9 @@ define void @test_tc_18(ptr noalias %src, ptr noalias %dst) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[SRC:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 64
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 64
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[DST:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    store <16 x i8> [[WIDE_LOAD]], ptr [[TMP4]], align 64
+; CHECK-NEXT:    store <16 x i8> [[WIDE_LOAD]], ptr [[TMP3]], align 64
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 16
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -85,11 +81,9 @@ define void @test_tc_18(ptr noalias %src, ptr noalias %dst) {
 ; CHECK:       vec.epilog.vector.body:
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i8>, ptr [[TMP8]], align 64
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i8>, ptr [[TMP7]], align 64
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[TMP9]], i32 0
-; CHECK-NEXT:    store <2 x i8> [[WIDE_LOAD2]], ptr [[TMP10]], align 64
+; CHECK-NEXT:    store <2 x i8> [[WIDE_LOAD2]], ptr [[TMP9]], align 64
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 2
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT3]], 18
 ; CHECK-NEXT:    br i1 [[TMP11]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
@@ -138,11 +132,9 @@ define void @test_tc_19(ptr noalias %src, ptr noalias %dst) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[SRC:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 64
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 64
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[DST:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    store <16 x i8> [[WIDE_LOAD]], ptr [[TMP4]], align 64
+; CHECK-NEXT:    store <16 x i8> [[WIDE_LOAD]], ptr [[TMP3]], align 64
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 16
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
@@ -156,11 +148,9 @@ define void @test_tc_19(ptr noalias %src, ptr noalias %dst) {
 ; CHECK:       vec.epilog.vector.body:
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i8>, ptr [[TMP8]], align 64
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i8>, ptr [[TMP7]], align 64
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[TMP9]], i32 0
-; CHECK-NEXT:    store <2 x i8> [[WIDE_LOAD2]], ptr [[TMP10]], align 64
+; CHECK-NEXT:    store <2 x i8> [[WIDE_LOAD2]], ptr [[TMP9]], align 64
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 2
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT3]], 18
 ; CHECK-NEXT:    br i1 [[TMP11]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
@@ -209,20 +199,18 @@ define void @test_tc_20(ptr noalias %src, ptr noalias %dst) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[SRC:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 8
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 12
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 64
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 64
 ; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i8>, ptr [[TMP3]], align 64
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP4]], align 64
 ; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP5]], align 64
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[DST:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[TMP6]], i32 0
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[TMP6]], i32 4
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[TMP6]], i32 8
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[TMP6]], i32 12
-; CHECK-NEXT:    store <4 x i8> [[WIDE_LOAD]], ptr [[TMP7]], align 64
+; CHECK-NEXT:    store <4 x i8> [[WIDE_LOAD]], ptr [[TMP6]], align 64
 ; CHECK-NEXT:    store <4 x i8> [[WIDE_LOAD1]], ptr [[TMP8]], align 64
 ; CHECK-NEXT:    store <4 x i8> [[WIDE_LOAD2]], ptr [[TMP9]], align 64
 ; CHECK-NEXT:    store <4 x i8> [[WIDE_LOAD3]], ptr [[TMP10]], align 64
@@ -239,11 +227,9 @@ define void @test_tc_20(ptr noalias %src, ptr noalias %dst) {
 ; CHECK:       vec.epilog.vector.body:
 ; CHECK-NEXT:    [[INDEX4:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT6:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[INDEX4]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i8>, ptr [[TMP14]], align 64
+; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i8>, ptr [[TMP13]], align 64
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[INDEX4]]
-; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[TMP15]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[WIDE_LOAD5]], ptr [[TMP16]], align 64
+; CHECK-NEXT:    store <4 x i8> [[WIDE_LOAD5]], ptr [[TMP15]], align 64
 ; CHECK-NEXT:    [[INDEX_NEXT6]] = add nuw i64 [[INDEX4]], 4
 ; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT6]], 20
 ; CHECK-NEXT:    br i1 [[TMP17]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
@@ -293,8 +279,7 @@ define void @limit_main_loop_vf_to_avoid_dead_main_vector_loop(ptr noalias %src,
 ; CHECK-NEXT:    [[WIDE_VEC:%.*]] = load <24 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[STRIDED_VEC:%.*]] = shufflevector <24 x i8> [[WIDE_VEC]], <24 x i8> poison, <8 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21>
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[DST:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    store <8 x i8> [[STRIDED_VEC]], ptr [[TMP4]], align 1
+; CHECK-NEXT:    store <8 x i8> [[STRIDED_VEC]], ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 24
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/load-deref-pred.ll b/llvm/test/Transforms/LoopVectorize/X86/load-deref-pred.ll
index f4cd48d..f615e23 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/load-deref-pred.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/load-deref-pred.ll
@@ -39,11 +39,10 @@ define i32 @test_explicit_pred(i64 %len) {
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp slt <4 x i64> [[STEP_ADD1]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp slt <4 x i64> [[STEP_ADD2]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[ALLOCA]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr [[TMP8]], i32 0
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr i32, ptr [[TMP8]], i32 4
 ; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr i32, ptr [[TMP8]], i32 8
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr i32, ptr [[TMP8]], i32 12
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP12]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i32>, ptr [[TMP13]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x i32>, ptr [[TMP14]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x i32>, ptr [[TMP15]], align 4
@@ -64,10 +63,10 @@ define i32 @test_explicit_pred(i64 %len) {
 ; CHECK-NEXT:    [[BIN_RDX13:%.*]] = add <4 x i32> [[TMP18]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX14:%.*]] = add <4 x i32> [[TMP19]], [[BIN_RDX13]]
 ; CHECK-NEXT:    [[TMP21:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX14]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP21]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
@@ -192,11 +191,10 @@ define i32 @test_explicit_pred_generic(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[TMP62:%.*]] = insertelement <4 x i1> [[TMP61]], i1 [[TMP58]], i32 2
 ; CHECK-NEXT:    [[TMP63:%.*]] = insertelement <4 x i1> [[TMP62]], i1 [[TMP59]], i32 3
 ; CHECK-NEXT:    [[TMP64:%.*]] = getelementptr i32, ptr [[ALLOCA]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP68:%.*]] = getelementptr i32, ptr [[TMP64]], i32 0
 ; CHECK-NEXT:    [[TMP69:%.*]] = getelementptr i32, ptr [[TMP64]], i32 4
 ; CHECK-NEXT:    [[TMP70:%.*]] = getelementptr i32, ptr [[TMP64]], i32 8
 ; CHECK-NEXT:    [[TMP71:%.*]] = getelementptr i32, ptr [[TMP64]], i32 12
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP68]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP64]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i32>, ptr [[TMP69]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i32>, ptr [[TMP70]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i32>, ptr [[TMP71]], align 4
@@ -216,10 +214,10 @@ define i32 @test_explicit_pred_generic(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[BIN_RDX10:%.*]] = add <4 x i32> [[TMP74]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX11:%.*]] = add <4 x i32> [[TMP75]], [[BIN_RDX10]]
 ; CHECK-NEXT:    [[TMP77:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX11]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP77]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
@@ -396,10 +394,10 @@ define i32 @test_invariant_address(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[BIN_RDX7:%.*]] = add <4 x i32> [[TMP98]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX8:%.*]] = add <4 x i32> [[TMP99]], [[BIN_RDX7]]
 ; CHECK-NEXT:    [[TMP101:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX8]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP101]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
@@ -667,10 +665,10 @@ define i32 @test_step_narrower_than_access(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[BIN_RDX37:%.*]] = add <4 x i32> [[TMP146]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX38:%.*]] = add <4 x i32> [[TMP147]], [[BIN_RDX37]]
 ; CHECK-NEXT:    [[TMP149:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX38]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP149]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
@@ -802,11 +800,10 @@ define i32 @test_max_trip_count(i64 %len, ptr %test_base, i64 %n) {
 ; CHECK-NEXT:    [[TMP63:%.*]] = insertelement <4 x i1> [[TMP62]], i1 [[TMP59]], i32 2
 ; CHECK-NEXT:    [[TMP64:%.*]] = insertelement <4 x i1> [[TMP63]], i1 [[TMP60]], i32 3
 ; CHECK-NEXT:    [[TMP65:%.*]] = getelementptr i32, ptr [[ALLOCA]], i64 [[TMP1]]
-; CHECK-NEXT:    [[TMP69:%.*]] = getelementptr i32, ptr [[TMP65]], i32 0
 ; CHECK-NEXT:    [[TMP70:%.*]] = getelementptr i32, ptr [[TMP65]], i32 4
 ; CHECK-NEXT:    [[TMP71:%.*]] = getelementptr i32, ptr [[TMP65]], i32 8
 ; CHECK-NEXT:    [[TMP72:%.*]] = getelementptr i32, ptr [[TMP65]], i32 12
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP69]], i32 4, <4 x i1> [[TMP40]], <4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP65]], i32 4, <4 x i1> [[TMP40]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP70]], i32 4, <4 x i1> [[TMP48]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD5:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP71]], i32 4, <4 x i1> [[TMP56]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD6:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP72]], i32 4, <4 x i1> [[TMP64]], <4 x i32> poison)
@@ -962,11 +959,10 @@ define i32 @test_non_zero_start(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[TMP62:%.*]] = insertelement <4 x i1> [[TMP61]], i1 [[TMP58]], i32 2
 ; CHECK-NEXT:    [[TMP63:%.*]] = insertelement <4 x i1> [[TMP62]], i1 [[TMP59]], i32 3
 ; CHECK-NEXT:    [[TMP64:%.*]] = getelementptr i32, ptr [[ALLOCA]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP68:%.*]] = getelementptr i32, ptr [[TMP64]], i32 0
 ; CHECK-NEXT:    [[TMP69:%.*]] = getelementptr i32, ptr [[TMP64]], i32 4
 ; CHECK-NEXT:    [[TMP70:%.*]] = getelementptr i32, ptr [[TMP64]], i32 8
 ; CHECK-NEXT:    [[TMP71:%.*]] = getelementptr i32, ptr [[TMP64]], i32 12
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP68]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP64]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i32>, ptr [[TMP69]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i32>, ptr [[TMP70]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i32>, ptr [[TMP71]], align 4
@@ -986,10 +982,10 @@ define i32 @test_non_zero_start(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[BIN_RDX10:%.*]] = add <4 x i32> [[TMP74]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX11:%.*]] = add <4 x i32> [[TMP75]], [[BIN_RDX10]]
 ; CHECK-NEXT:    [[TMP77:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX11]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 1024, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP77]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
@@ -1230,10 +1226,10 @@ define i32 @test_non_unit_stride(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[BIN_RDX7:%.*]] = add <4 x i32> [[TMP114]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX8:%.*]] = add <4 x i32> [[TMP115]], [[BIN_RDX7]]
 ; CHECK-NEXT:    [[TMP117:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX8]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP117]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
@@ -1359,11 +1355,10 @@ define i32 @neg_off_by_many(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[TMP62:%.*]] = insertelement <4 x i1> [[TMP61]], i1 [[TMP58]], i32 2
 ; CHECK-NEXT:    [[TMP63:%.*]] = insertelement <4 x i1> [[TMP62]], i1 [[TMP59]], i32 3
 ; CHECK-NEXT:    [[TMP64:%.*]] = getelementptr i32, ptr [[ALLOCA]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP68:%.*]] = getelementptr i32, ptr [[TMP64]], i32 0
 ; CHECK-NEXT:    [[TMP69:%.*]] = getelementptr i32, ptr [[TMP64]], i32 4
 ; CHECK-NEXT:    [[TMP70:%.*]] = getelementptr i32, ptr [[TMP64]], i32 8
 ; CHECK-NEXT:    [[TMP71:%.*]] = getelementptr i32, ptr [[TMP64]], i32 12
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP68]], i32 4, <4 x i1> [[TMP39]], <4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP64]], i32 4, <4 x i1> [[TMP39]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP69]], i32 4, <4 x i1> [[TMP47]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD5:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP70]], i32 4, <4 x i1> [[TMP55]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD6:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP71]], i32 4, <4 x i1> [[TMP63]], <4 x i32> poison)
@@ -1383,10 +1378,10 @@ define i32 @neg_off_by_many(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[BIN_RDX10:%.*]] = add <4 x i32> [[TMP74]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX11:%.*]] = add <4 x i32> [[TMP75]], [[BIN_RDX10]]
 ; CHECK-NEXT:    [[TMP77:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX11]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP77]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
@@ -1512,11 +1507,10 @@ define i32 @neg_off_by_one_iteration(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[TMP62:%.*]] = insertelement <4 x i1> [[TMP61]], i1 [[TMP58]], i32 2
 ; CHECK-NEXT:    [[TMP63:%.*]] = insertelement <4 x i1> [[TMP62]], i1 [[TMP59]], i32 3
 ; CHECK-NEXT:    [[TMP64:%.*]] = getelementptr i32, ptr [[ALLOCA]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP68:%.*]] = getelementptr i32, ptr [[TMP64]], i32 0
 ; CHECK-NEXT:    [[TMP69:%.*]] = getelementptr i32, ptr [[TMP64]], i32 4
 ; CHECK-NEXT:    [[TMP70:%.*]] = getelementptr i32, ptr [[TMP64]], i32 8
 ; CHECK-NEXT:    [[TMP71:%.*]] = getelementptr i32, ptr [[TMP64]], i32 12
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP68]], i32 4, <4 x i1> [[TMP39]], <4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP64]], i32 4, <4 x i1> [[TMP39]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP69]], i32 4, <4 x i1> [[TMP47]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD5:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP70]], i32 4, <4 x i1> [[TMP55]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD6:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP71]], i32 4, <4 x i1> [[TMP63]], <4 x i32> poison)
@@ -1536,10 +1530,10 @@ define i32 @neg_off_by_one_iteration(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[BIN_RDX10:%.*]] = add <4 x i32> [[TMP74]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX11:%.*]] = add <4 x i32> [[TMP75]], [[BIN_RDX10]]
 ; CHECK-NEXT:    [[TMP77:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX11]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP77]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
@@ -1665,11 +1659,10 @@ define i32 @neg_off_by_one_byte(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[TMP62:%.*]] = insertelement <4 x i1> [[TMP61]], i1 [[TMP58]], i32 2
 ; CHECK-NEXT:    [[TMP63:%.*]] = insertelement <4 x i1> [[TMP62]], i1 [[TMP59]], i32 3
 ; CHECK-NEXT:    [[TMP64:%.*]] = getelementptr i32, ptr [[ALLOCA]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP68:%.*]] = getelementptr i32, ptr [[TMP64]], i32 0
 ; CHECK-NEXT:    [[TMP69:%.*]] = getelementptr i32, ptr [[TMP64]], i32 4
 ; CHECK-NEXT:    [[TMP70:%.*]] = getelementptr i32, ptr [[TMP64]], i32 8
 ; CHECK-NEXT:    [[TMP71:%.*]] = getelementptr i32, ptr [[TMP64]], i32 12
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP68]], i32 4, <4 x i1> [[TMP39]], <4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP64]], i32 4, <4 x i1> [[TMP39]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP69]], i32 4, <4 x i1> [[TMP47]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD5:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP70]], i32 4, <4 x i1> [[TMP55]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD6:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP71]], i32 4, <4 x i1> [[TMP63]], <4 x i32> poison)
@@ -1689,10 +1682,10 @@ define i32 @neg_off_by_one_byte(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[BIN_RDX10:%.*]] = add <4 x i32> [[TMP74]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX11:%.*]] = add <4 x i32> [[TMP75]], [[BIN_RDX10]]
 ; CHECK-NEXT:    [[TMP77:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX11]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP77]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
@@ -1827,11 +1820,10 @@ define i32 @test_constant_max(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[TMP63:%.*]] = insertelement <4 x i1> [[TMP62]], i1 [[TMP59]], i32 2
 ; CHECK-NEXT:    [[TMP64:%.*]] = insertelement <4 x i1> [[TMP63]], i1 [[TMP60]], i32 3
 ; CHECK-NEXT:    [[TMP65:%.*]] = getelementptr i32, ptr [[ALLOCA]], i64 [[TMP1]]
-; CHECK-NEXT:    [[TMP69:%.*]] = getelementptr i32, ptr [[TMP65]], i32 0
 ; CHECK-NEXT:    [[TMP70:%.*]] = getelementptr i32, ptr [[TMP65]], i32 4
 ; CHECK-NEXT:    [[TMP71:%.*]] = getelementptr i32, ptr [[TMP65]], i32 8
 ; CHECK-NEXT:    [[TMP72:%.*]] = getelementptr i32, ptr [[TMP65]], i32 12
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP69]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP65]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i32>, ptr [[TMP70]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i32>, ptr [[TMP71]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i32>, ptr [[TMP72]], align 4
@@ -1988,11 +1980,10 @@ define i32 @test_allocsize(i64 %len, ptr %test_base) nofree nosync {
 ; CHECK-NEXT:    [[TMP62:%.*]] = insertelement <4 x i1> [[TMP61]], i1 [[TMP58]], i32 2
 ; CHECK-NEXT:    [[TMP63:%.*]] = insertelement <4 x i1> [[TMP62]], i1 [[TMP59]], i32 3
 ; CHECK-NEXT:    [[TMP64:%.*]] = getelementptr i32, ptr [[ALLOCATION]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP68:%.*]] = getelementptr i32, ptr [[TMP64]], i32 0
 ; CHECK-NEXT:    [[TMP69:%.*]] = getelementptr i32, ptr [[TMP64]], i32 4
 ; CHECK-NEXT:    [[TMP70:%.*]] = getelementptr i32, ptr [[TMP64]], i32 8
 ; CHECK-NEXT:    [[TMP71:%.*]] = getelementptr i32, ptr [[TMP64]], i32 12
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP68]], i32 4, <4 x i1> [[TMP39]], <4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP64]], i32 4, <4 x i1> [[TMP39]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP69]], i32 4, <4 x i1> [[TMP47]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD5:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP70]], i32 4, <4 x i1> [[TMP55]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD6:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP71]], i32 4, <4 x i1> [[TMP63]], <4 x i32> poison)
@@ -2012,10 +2003,10 @@ define i32 @test_allocsize(i64 %len, ptr %test_base) nofree nosync {
 ; CHECK-NEXT:    [[BIN_RDX10:%.*]] = add <4 x i32> [[TMP74]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX11:%.*]] = add <4 x i32> [[TMP75]], [[BIN_RDX10]]
 ; CHECK-NEXT:    [[TMP77:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX11]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP77]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
@@ -2142,11 +2133,10 @@ define i32 @test_allocsize_array(i64 %len, ptr %test_base) nofree nosync {
 ; CHECK-NEXT:    [[TMP62:%.*]] = insertelement <4 x i1> [[TMP61]], i1 [[TMP58]], i32 2
 ; CHECK-NEXT:    [[TMP63:%.*]] = insertelement <4 x i1> [[TMP62]], i1 [[TMP59]], i32 3
 ; CHECK-NEXT:    [[TMP64:%.*]] = getelementptr i32, ptr [[ALLOCATION]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP68:%.*]] = getelementptr i32, ptr [[TMP64]], i32 0
 ; CHECK-NEXT:    [[TMP69:%.*]] = getelementptr i32, ptr [[TMP64]], i32 4
 ; CHECK-NEXT:    [[TMP70:%.*]] = getelementptr i32, ptr [[TMP64]], i32 8
 ; CHECK-NEXT:    [[TMP71:%.*]] = getelementptr i32, ptr [[TMP64]], i32 12
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP68]], i32 4, <4 x i1> [[TMP39]], <4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP64]], i32 4, <4 x i1> [[TMP39]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP69]], i32 4, <4 x i1> [[TMP47]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD5:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP70]], i32 4, <4 x i1> [[TMP55]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD6:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP71]], i32 4, <4 x i1> [[TMP63]], <4 x i32> poison)
@@ -2166,10 +2156,10 @@ define i32 @test_allocsize_array(i64 %len, ptr %test_base) nofree nosync {
 ; CHECK-NEXT:    [[BIN_RDX10:%.*]] = add <4 x i32> [[TMP74]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX11:%.*]] = add <4 x i32> [[TMP75]], [[BIN_RDX10]]
 ; CHECK-NEXT:    [[TMP77:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX11]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP77]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
@@ -2306,11 +2296,10 @@ define i32 @test_allocsize_cond_deref(i1 %allzero, ptr %test_base) {
 ; CHECK-NEXT:    [[TMP62:%.*]] = insertelement <4 x i1> [[TMP61]], i1 [[TMP58]], i32 2
 ; CHECK-NEXT:    [[TMP63:%.*]] = insertelement <4 x i1> [[TMP62]], i1 [[TMP59]], i32 3
 ; CHECK-NEXT:    [[TMP64:%.*]] = getelementptr i32, ptr [[ALLOCATION]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP68:%.*]] = getelementptr i32, ptr [[TMP64]], i32 0
 ; CHECK-NEXT:    [[TMP69:%.*]] = getelementptr i32, ptr [[TMP64]], i32 4
 ; CHECK-NEXT:    [[TMP70:%.*]] = getelementptr i32, ptr [[TMP64]], i32 8
 ; CHECK-NEXT:    [[TMP71:%.*]] = getelementptr i32, ptr [[TMP64]], i32 12
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP68]], i32 4, <4 x i1> [[TMP39]], <4 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP64]], i32 4, <4 x i1> [[TMP39]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP69]], i32 4, <4 x i1> [[TMP47]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD5:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP70]], i32 4, <4 x i1> [[TMP55]], <4 x i32> poison)
 ; CHECK-NEXT:    [[WIDE_MASKED_LOAD6:%.*]] = call <4 x i32> @llvm.masked.load.v4i32.p0(ptr [[TMP71]], i32 4, <4 x i1> [[TMP63]], <4 x i32> poison)
@@ -2330,10 +2319,10 @@ define i32 @test_allocsize_cond_deref(i1 %allzero, ptr %test_base) {
 ; CHECK-NEXT:    [[BIN_RDX10:%.*]] = add <4 x i32> [[TMP74]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX11:%.*]] = add <4 x i32> [[TMP75]], [[BIN_RDX10]]
 ; CHECK-NEXT:    [[TMP77:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX11]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[PREHEADER]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP77]], [[MIDDLE_BLOCK]] ], [ 0, [[PREHEADER]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[PREHEADER]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[PREHEADER]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
@@ -2531,7 +2520,7 @@ define i32 @test_stride_three(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[BIN_RDX7:%.*]] = add <4 x i32> [[TMP114]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX8:%.*]] = add <4 x i32> [[TMP115]], [[BIN_RDX7]]
 ; CHECK-NEXT:    [[TMP117:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX8]])
-; CHECK-NEXT:    br i1 false, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 96, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP117]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
@@ -2551,9 +2540,9 @@ define i32 @test_stride_three(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[VAL_PHI:%.*]] = phi i32 [ 0, [[LOOP]] ], [ [[VAL]], [[PRED]] ]
 ; CHECK-NEXT:    [[ACCUM_NEXT]] = add i32 [[ACCUM]], [[VAL_PHI]]
 ; CHECK-NEXT:    [[EXIT:%.*]] = icmp ugt i64 [[IV]], 100
-; CHECK-NEXT:    br i1 [[EXIT]], label [[LOOP_EXIT]], label [[LOOP]], !llvm.loop [[LOOP31:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXIT]], label [[LOOP_EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP31:![0-9]+]]
 ; CHECK:       loop_exit:
-; CHECK-NEXT:    [[ACCUM_NEXT_LCSSA:%.*]] = phi i32 [ [[ACCUM_NEXT]], [[LATCH]] ], [ [[TMP117]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[ACCUM_NEXT_LCSSA:%.*]] = phi i32 [ [[ACCUM_NEXT]], [[LATCH]] ]
 ; CHECK-NEXT:    ret i32 [[ACCUM_NEXT_LCSSA]]
 ;
 entry:
@@ -2660,7 +2649,7 @@ define i32 @test_non_unit_stride_four(i64 %len, ptr %test_base) {
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[TMP57]], [[TMP56]]
 ; CHECK-NEXT:    [[TMP59:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
-; CHECK-NEXT:    br i1 false, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 96, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP59]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
@@ -2680,9 +2669,9 @@ define i32 @test_non_unit_stride_four(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[VAL_PHI:%.*]] = phi i32 [ 0, [[LOOP]] ], [ [[VAL]], [[PRED]] ]
 ; CHECK-NEXT:    [[ACCUM_NEXT]] = add i32 [[ACCUM]], [[VAL_PHI]]
 ; CHECK-NEXT:    [[EXIT:%.*]] = icmp ugt i64 [[IV]], 100
-; CHECK-NEXT:    br i1 [[EXIT]], label [[LOOP_EXIT]], label [[LOOP]], !llvm.loop [[LOOP33:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXIT]], label [[LOOP_EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP33:![0-9]+]]
 ; CHECK:       loop_exit:
-; CHECK-NEXT:    [[ACCUM_NEXT_LCSSA:%.*]] = phi i32 [ [[ACCUM_NEXT]], [[LATCH]] ], [ [[TMP59]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[ACCUM_NEXT_LCSSA:%.*]] = phi i32 [ [[ACCUM_NEXT]], [[LATCH]] ]
 ; CHECK-NEXT:    ret i32 [[ACCUM_NEXT_LCSSA]]
 ;
 entry:
@@ -2853,7 +2842,7 @@ define i32 @test_non_unit_stride_five(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[BIN_RDX7:%.*]] = add <4 x i32> [[TMP114]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX8:%.*]] = add <4 x i32> [[TMP115]], [[BIN_RDX7]]
 ; CHECK-NEXT:    [[TMP117:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX8]])
-; CHECK-NEXT:    br i1 false, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 80, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP117]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
@@ -2873,9 +2862,9 @@ define i32 @test_non_unit_stride_five(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[VAL_PHI:%.*]] = phi i32 [ 0, [[LOOP]] ], [ [[VAL]], [[PRED]] ]
 ; CHECK-NEXT:    [[ACCUM_NEXT]] = add i32 [[ACCUM]], [[VAL_PHI]]
 ; CHECK-NEXT:    [[EXIT:%.*]] = icmp ugt i64 [[IV]], 100
-; CHECK-NEXT:    br i1 [[EXIT]], label [[LOOP_EXIT]], label [[LOOP]], !llvm.loop [[LOOP35:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXIT]], label [[LOOP_EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP35:![0-9]+]]
 ; CHECK:       loop_exit:
-; CHECK-NEXT:    [[ACCUM_NEXT_LCSSA:%.*]] = phi i32 [ [[ACCUM_NEXT]], [[LATCH]] ], [ [[TMP117]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[ACCUM_NEXT_LCSSA:%.*]] = phi i32 [ [[ACCUM_NEXT]], [[LATCH]] ]
 ; CHECK-NEXT:    ret i32 [[ACCUM_NEXT_LCSSA]]
 ;
 entry:
@@ -3046,7 +3035,7 @@ define i32 @test_non_unit_stride_off_by_four_bytes(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[BIN_RDX7:%.*]] = add <4 x i32> [[TMP114]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX8:%.*]] = add <4 x i32> [[TMP115]], [[BIN_RDX7]]
 ; CHECK-NEXT:    [[TMP117:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX8]])
-; CHECK-NEXT:    br i1 false, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 96, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP117]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
@@ -3066,9 +3055,9 @@ define i32 @test_non_unit_stride_off_by_four_bytes(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[VAL_PHI:%.*]] = phi i32 [ 0, [[LOOP]] ], [ [[VAL]], [[PRED]] ]
 ; CHECK-NEXT:    [[ACCUM_NEXT]] = add i32 [[ACCUM]], [[VAL_PHI]]
 ; CHECK-NEXT:    [[EXIT:%.*]] = icmp ugt i64 [[IV]], 100
-; CHECK-NEXT:    br i1 [[EXIT]], label [[LOOP_EXIT]], label [[LOOP]], !llvm.loop [[LOOP37:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXIT]], label [[LOOP_EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP37:![0-9]+]]
 ; CHECK:       loop_exit:
-; CHECK-NEXT:    [[ACCUM_NEXT_LCSSA:%.*]] = phi i32 [ [[ACCUM_NEXT]], [[LATCH]] ], [ [[TMP117]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[ACCUM_NEXT_LCSSA:%.*]] = phi i32 [ [[ACCUM_NEXT]], [[LATCH]] ]
 ; CHECK-NEXT:    ret i32 [[ACCUM_NEXT_LCSSA]]
 ;
 entry:
@@ -3256,7 +3245,7 @@ define i32 @test_non_unit_stride_with_first_iteration_step_access(i64 %len, ptr
 ; CHECK-NEXT:    [[BIN_RDX7:%.*]] = add <4 x i32> [[TMP130]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX8:%.*]] = add <4 x i32> [[TMP131]], [[BIN_RDX7]]
 ; CHECK-NEXT:    [[TMP133:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX8]])
-; CHECK-NEXT:    br i1 false, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 288, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP133]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
@@ -3276,9 +3265,9 @@ define i32 @test_non_unit_stride_with_first_iteration_step_access(i64 %len, ptr
 ; CHECK-NEXT:    [[VAL_PHI:%.*]] = phi i32 [ 0, [[LOOP]] ], [ [[VAL]], [[PRED]] ]
 ; CHECK-NEXT:    [[ACCUM_NEXT]] = add i32 [[ACCUM]], [[VAL_PHI]]
 ; CHECK-NEXT:    [[EXIT:%.*]] = icmp ugt i64 [[IV]], 300
-; CHECK-NEXT:    br i1 [[EXIT]], label [[LOOP_EXIT]], label [[LOOP]], !llvm.loop [[LOOP39:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXIT]], label [[LOOP_EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP39:![0-9]+]]
 ; CHECK:       loop_exit:
-; CHECK-NEXT:    [[ACCUM_NEXT_LCSSA:%.*]] = phi i32 [ [[ACCUM_NEXT]], [[LATCH]] ], [ [[TMP133]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[ACCUM_NEXT_LCSSA:%.*]] = phi i32 [ [[ACCUM_NEXT]], [[LATCH]] ]
 ; CHECK-NEXT:    ret i32 [[ACCUM_NEXT_LCSSA]]
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/X86/masked-store-cost.ll b/llvm/test/Transforms/LoopVectorize/X86/masked-store-cost.ll
index 612f996..af93985 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/masked-store-cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/masked-store-cost.ll
@@ -35,11 +35,10 @@ define i32 @test_scalar_predicated_cost(i64 %x, i64 %y, ptr %A) #0 {
 ; CHECK-NEXT:    [[TMP21:%.*]] = trunc <8 x i64> [[TMP13]] to <8 x i32>
 ; CHECK-NEXT:    [[TMP22:%.*]] = trunc <8 x i64> [[TMP14]] to <8 x i32>
 ; CHECK-NEXT:    [[TMP23:%.*]] = trunc <8 x i64> [[TMP15]] to <8 x i32>
-; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr i32, ptr [[TMP16]], i32 0
 ; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr i32, ptr [[TMP16]], i32 8
 ; CHECK-NEXT:    [[TMP26:%.*]] = getelementptr i32, ptr [[TMP16]], i32 16
 ; CHECK-NEXT:    [[TMP27:%.*]] = getelementptr i32, ptr [[TMP16]], i32 24
-; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP20]], ptr [[TMP24]], i32 4, <8 x i1> [[TMP8]])
+; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP20]], ptr [[TMP16]], i32 4, <8 x i1> [[TMP8]])
 ; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP21]], ptr [[TMP25]], i32 4, <8 x i1> [[TMP9]])
 ; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP22]], ptr [[TMP26]], i32 4, <8 x i1> [[TMP10]])
 ; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP23]], ptr [[TMP27]], i32 4, <8 x i1> [[TMP11]])
@@ -68,8 +67,7 @@ define i32 @test_scalar_predicated_cost(i64 %x, i64 %y, ptr %A) #0 {
 ; CHECK-NEXT:    [[TMP34:%.*]] = or <4 x i64> [[BROADCAST_SPLAT10]], [[VEC_IND5]]
 ; CHECK-NEXT:    [[TMP35:%.*]] = getelementptr i32, ptr [[A]], i64 [[INDEX4]]
 ; CHECK-NEXT:    [[TMP36:%.*]] = trunc <4 x i64> [[TMP34]] to <4 x i32>
-; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr i32, ptr [[TMP35]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[TMP36]], ptr [[TMP29]], i32 4, <4 x i1> [[TMP33]])
+; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> [[TMP36]], ptr [[TMP35]], i32 4, <4 x i1> [[TMP33]])
 ; CHECK-NEXT:    [[INDEX_NEXT11]] = add nuw i64 [[INDEX4]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT6]] = add <4 x i64> [[VEC_IND5]], splat (i64 4)
 ; CHECK-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT11]], 100
@@ -134,13 +132,12 @@ define void @test_scalar_cost_single_store_loop_invariant_cond(ptr %dst, i1 %c)
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[DST]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i32, ptr [[NEXT_GEP]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> zeroinitializer, ptr [[TMP1]], i32 4, <8 x i1> [[BROADCAST_SPLAT]])
+; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> zeroinitializer, ptr [[NEXT_GEP]], i32 4, <8 x i1> [[BROADCAST_SPLAT]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], 24
 ; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi ptr [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[DST]], [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i64 [ 96, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
@@ -156,7 +153,7 @@ define void @test_scalar_cost_single_store_loop_invariant_cond(ptr %dst, i1 %c)
 ; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 4
 ; CHECK-NEXT:    [[PTR_IV_NEXT]] = getelementptr i8, ptr [[DST]], i64 [[IV_NEXT]]
 ; CHECK-NEXT:    [[CMP_NOT:%.*]] = icmp eq i64 [[IV]], 116
-; CHECK-NEXT:    br i1 [[CMP_NOT]], label [[EXIT]], label [[LOOP_HEADER]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP_NOT]], label [[EXIT:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
@@ -204,9 +201,8 @@ define void @test_scalar_cost_single_store_loop_varying_cond(ptr %dst, ptr noali
 ; CHECK-NEXT:    [[STRIDED_VEC5:%.*]] = shufflevector <16 x i32> [[WIDE_VEC4]], <16 x i32> poison, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq <4 x i32> [[STRIDED_VEC]], splat (i32 123)
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq <4 x i32> [[STRIDED_VEC5]], splat (i32 123)
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[NEXT_GEP]], i32 0
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[NEXT_GEP]], i32 4
-; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> zeroinitializer, ptr [[TMP10]], i32 4, <4 x i1> [[TMP8]])
+; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> zeroinitializer, ptr [[NEXT_GEP]], i32 4, <4 x i1> [[TMP8]])
 ; CHECK-NEXT:    call void @llvm.masked.store.v4i32.p0(<4 x i32> zeroinitializer, ptr [[TMP11]], i32 4, <4 x i1> [[TMP9]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 24
diff --git a/llvm/test/Transforms/LoopVectorize/X86/masked_load_store.ll b/llvm/test/Transforms/LoopVectorize/X86/masked_load_store.ll
index 2b6490a..5065dc8 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/masked_load_store.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/masked_load_store.ll
@@ -36,23 +36,20 @@ define void @foo1(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX1:       vector.body:
 ; AVX1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TRIGGER]], i64 [[INDEX]]
-; AVX1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
-; AVX1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP4]], align 4
+; AVX1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP3]], align 4
 ; AVX1-NEXT:    [[TMP5:%.*]] = icmp slt <8 x i32> [[WIDE_LOAD]], splat (i32 100)
 ; AVX1-NEXT:    [[TMP6:%.*]] = getelementptr i32, ptr [[B]], i64 [[INDEX]]
-; AVX1-NEXT:    [[TMP7:%.*]] = getelementptr i32, ptr [[TMP6]], i32 0
-; AVX1-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP7]], i32 4, <8 x i1> [[TMP5]], <8 x i32> poison)
+; AVX1-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP6]], i32 4, <8 x i1> [[TMP5]], <8 x i32> poison)
 ; AVX1-NEXT:    [[TMP8:%.*]] = add nsw <8 x i32> [[WIDE_MASKED_LOAD]], [[WIDE_LOAD]]
 ; AVX1-NEXT:    [[TMP9:%.*]] = getelementptr i32, ptr [[A]], i64 [[INDEX]]
-; AVX1-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[TMP9]], i32 0
-; AVX1-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP8]], ptr [[TMP10]], i32 4, <8 x i1> [[TMP5]])
+; AVX1-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP8]], ptr [[TMP9]], i32 4, <8 x i1> [[TMP5]])
 ; AVX1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; AVX1-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10000
 ; AVX1-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; AVX1:       middle.block:
-; AVX1-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; AVX1-NEXT:    br label [[FOR_END:%.*]]
 ; AVX1:       scalar.ph:
-; AVX1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 10000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; AVX1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
 ; AVX1-NEXT:    br label [[FOR_BODY:%.*]]
 ; AVX1:       for.body:
 ; AVX1-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_INC:%.*]] ]
@@ -94,11 +91,10 @@ define void @foo1(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX2:       vector.body:
 ; AVX2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH1]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TRIGGER]], i64 [[INDEX]]
-; AVX2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
 ; AVX2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 8
 ; AVX2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 16
 ; AVX2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 24
-; AVX2-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP4]], align 4
+; AVX2-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP3]], align 4
 ; AVX2-NEXT:    [[WIDE_LOAD5:%.*]] = load <8 x i32>, ptr [[TMP5]], align 4
 ; AVX2-NEXT:    [[WIDE_LOAD6:%.*]] = load <8 x i32>, ptr [[TMP6]], align 4
 ; AVX2-NEXT:    [[WIDE_LOAD7:%.*]] = load <8 x i32>, ptr [[TMP7]], align 4
@@ -107,11 +103,10 @@ define void @foo1(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX2-NEXT:    [[TMP10:%.*]] = icmp slt <8 x i32> [[WIDE_LOAD6]], splat (i32 100)
 ; AVX2-NEXT:    [[TMP11:%.*]] = icmp slt <8 x i32> [[WIDE_LOAD7]], splat (i32 100)
 ; AVX2-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr [[B]], i64 [[INDEX]]
-; AVX2-NEXT:    [[TMP13:%.*]] = getelementptr i32, ptr [[TMP12]], i32 0
 ; AVX2-NEXT:    [[TMP14:%.*]] = getelementptr i32, ptr [[TMP12]], i32 8
 ; AVX2-NEXT:    [[TMP15:%.*]] = getelementptr i32, ptr [[TMP12]], i32 16
 ; AVX2-NEXT:    [[TMP16:%.*]] = getelementptr i32, ptr [[TMP12]], i32 24
-; AVX2-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP13]], i32 4, <8 x i1> [[TMP8]], <8 x i32> poison)
+; AVX2-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP12]], i32 4, <8 x i1> [[TMP8]], <8 x i32> poison)
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD8:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP14]], i32 4, <8 x i1> [[TMP9]], <8 x i32> poison)
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP15]], i32 4, <8 x i1> [[TMP10]], <8 x i32> poison)
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP16]], i32 4, <8 x i1> [[TMP11]], <8 x i32> poison)
@@ -120,11 +115,10 @@ define void @foo1(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX2-NEXT:    [[TMP19:%.*]] = add nsw <8 x i32> [[WIDE_MASKED_LOAD9]], [[WIDE_LOAD6]]
 ; AVX2-NEXT:    [[TMP20:%.*]] = add nsw <8 x i32> [[WIDE_MASKED_LOAD10]], [[WIDE_LOAD7]]
 ; AVX2-NEXT:    [[TMP21:%.*]] = getelementptr i32, ptr [[A]], i64 [[INDEX]]
-; AVX2-NEXT:    [[TMP22:%.*]] = getelementptr i32, ptr [[TMP21]], i32 0
 ; AVX2-NEXT:    [[TMP23:%.*]] = getelementptr i32, ptr [[TMP21]], i32 8
 ; AVX2-NEXT:    [[TMP24:%.*]] = getelementptr i32, ptr [[TMP21]], i32 16
 ; AVX2-NEXT:    [[TMP25:%.*]] = getelementptr i32, ptr [[TMP21]], i32 24
-; AVX2-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP17]], ptr [[TMP22]], i32 4, <8 x i1> [[TMP8]])
+; AVX2-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP17]], ptr [[TMP21]], i32 4, <8 x i1> [[TMP8]])
 ; AVX2-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP18]], ptr [[TMP23]], i32 4, <8 x i1> [[TMP9]])
 ; AVX2-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP19]], ptr [[TMP24]], i32 4, <8 x i1> [[TMP10]])
 ; AVX2-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP20]], ptr [[TMP25]], i32 4, <8 x i1> [[TMP11]])
@@ -141,16 +135,13 @@ define void @foo1(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX2:       vec.epilog.vector.body:
 ; AVX2-NEXT:    [[INDEX11:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT14:%.*]], [[FOR_BODY]] ]
 ; AVX2-NEXT:    [[TMP38:%.*]] = getelementptr inbounds i32, ptr [[TRIGGER]], i64 [[INDEX11]]
-; AVX2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i32, ptr [[TMP38]], i32 0
-; AVX2-NEXT:    [[WIDE_LOAD12:%.*]] = load <8 x i32>, ptr [[TMP29]], align 4
+; AVX2-NEXT:    [[WIDE_LOAD12:%.*]] = load <8 x i32>, ptr [[TMP38]], align 4
 ; AVX2-NEXT:    [[TMP30:%.*]] = icmp slt <8 x i32> [[WIDE_LOAD12]], splat (i32 100)
 ; AVX2-NEXT:    [[TMP31:%.*]] = getelementptr i32, ptr [[B]], i64 [[INDEX11]]
-; AVX2-NEXT:    [[TMP32:%.*]] = getelementptr i32, ptr [[TMP31]], i32 0
-; AVX2-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP32]], i32 4, <8 x i1> [[TMP30]], <8 x i32> poison)
+; AVX2-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP31]], i32 4, <8 x i1> [[TMP30]], <8 x i32> poison)
 ; AVX2-NEXT:    [[TMP33:%.*]] = add nsw <8 x i32> [[WIDE_MASKED_LOAD13]], [[WIDE_LOAD12]]
 ; AVX2-NEXT:    [[TMP34:%.*]] = getelementptr i32, ptr [[A]], i64 [[INDEX11]]
-; AVX2-NEXT:    [[TMP35:%.*]] = getelementptr i32, ptr [[TMP34]], i32 0
-; AVX2-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP33]], ptr [[TMP35]], i32 4, <8 x i1> [[TMP30]])
+; AVX2-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP33]], ptr [[TMP34]], i32 4, <8 x i1> [[TMP30]])
 ; AVX2-NEXT:    [[INDEX_NEXT14]] = add nuw i64 [[INDEX11]], 8
 ; AVX2-NEXT:    [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT14]], 10000
 ; AVX2-NEXT:    br i1 [[TMP36]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
@@ -199,11 +190,10 @@ define void @foo1(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX512:       vector.body:
 ; AVX512-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX512-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TRIGGER]], i64 [[INDEX]]
-; AVX512-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
 ; AVX512-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 16
 ; AVX512-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 32
 ; AVX512-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 48
-; AVX512-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i32>, ptr [[TMP4]], align 4
+; AVX512-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i32>, ptr [[TMP3]], align 4
 ; AVX512-NEXT:    [[WIDE_LOAD5:%.*]] = load <16 x i32>, ptr [[TMP5]], align 4
 ; AVX512-NEXT:    [[WIDE_LOAD6:%.*]] = load <16 x i32>, ptr [[TMP6]], align 4
 ; AVX512-NEXT:    [[WIDE_LOAD7:%.*]] = load <16 x i32>, ptr [[TMP7]], align 4
@@ -212,11 +202,10 @@ define void @foo1(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX512-NEXT:    [[TMP10:%.*]] = icmp slt <16 x i32> [[WIDE_LOAD6]], splat (i32 100)
 ; AVX512-NEXT:    [[TMP11:%.*]] = icmp slt <16 x i32> [[WIDE_LOAD7]], splat (i32 100)
 ; AVX512-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr [[B]], i64 [[INDEX]]
-; AVX512-NEXT:    [[TMP13:%.*]] = getelementptr i32, ptr [[TMP12]], i32 0
 ; AVX512-NEXT:    [[TMP14:%.*]] = getelementptr i32, ptr [[TMP12]], i32 16
 ; AVX512-NEXT:    [[TMP15:%.*]] = getelementptr i32, ptr [[TMP12]], i32 32
 ; AVX512-NEXT:    [[TMP16:%.*]] = getelementptr i32, ptr [[TMP12]], i32 48
-; AVX512-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p0(ptr [[TMP13]], i32 4, <16 x i1> [[TMP8]], <16 x i32> poison)
+; AVX512-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p0(ptr [[TMP12]], i32 4, <16 x i1> [[TMP8]], <16 x i32> poison)
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD8:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p0(ptr [[TMP14]], i32 4, <16 x i1> [[TMP9]], <16 x i32> poison)
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p0(ptr [[TMP15]], i32 4, <16 x i1> [[TMP10]], <16 x i32> poison)
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p0(ptr [[TMP16]], i32 4, <16 x i1> [[TMP11]], <16 x i32> poison)
@@ -225,11 +214,10 @@ define void @foo1(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX512-NEXT:    [[TMP19:%.*]] = add nsw <16 x i32> [[WIDE_MASKED_LOAD9]], [[WIDE_LOAD6]]
 ; AVX512-NEXT:    [[TMP20:%.*]] = add nsw <16 x i32> [[WIDE_MASKED_LOAD10]], [[WIDE_LOAD7]]
 ; AVX512-NEXT:    [[TMP21:%.*]] = getelementptr i32, ptr [[A]], i64 [[INDEX]]
-; AVX512-NEXT:    [[TMP22:%.*]] = getelementptr i32, ptr [[TMP21]], i32 0
 ; AVX512-NEXT:    [[TMP23:%.*]] = getelementptr i32, ptr [[TMP21]], i32 16
 ; AVX512-NEXT:    [[TMP24:%.*]] = getelementptr i32, ptr [[TMP21]], i32 32
 ; AVX512-NEXT:    [[TMP25:%.*]] = getelementptr i32, ptr [[TMP21]], i32 48
-; AVX512-NEXT:    call void @llvm.masked.store.v16i32.p0(<16 x i32> [[TMP17]], ptr [[TMP22]], i32 4, <16 x i1> [[TMP8]])
+; AVX512-NEXT:    call void @llvm.masked.store.v16i32.p0(<16 x i32> [[TMP17]], ptr [[TMP21]], i32 4, <16 x i1> [[TMP8]])
 ; AVX512-NEXT:    call void @llvm.masked.store.v16i32.p0(<16 x i32> [[TMP18]], ptr [[TMP23]], i32 4, <16 x i1> [[TMP9]])
 ; AVX512-NEXT:    call void @llvm.masked.store.v16i32.p0(<16 x i32> [[TMP19]], ptr [[TMP24]], i32 4, <16 x i1> [[TMP10]])
 ; AVX512-NEXT:    call void @llvm.masked.store.v16i32.p0(<16 x i32> [[TMP20]], ptr [[TMP25]], i32 4, <16 x i1> [[TMP11]])
@@ -246,16 +234,13 @@ define void @foo1(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX512:       vec.epilog.vector.body:
 ; AVX512-NEXT:    [[INDEX11:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT14:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; AVX512-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i32, ptr [[TRIGGER]], i64 [[INDEX11]]
-; AVX512-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i32, ptr [[TMP28]], i32 0
-; AVX512-NEXT:    [[WIDE_LOAD12:%.*]] = load <16 x i32>, ptr [[TMP29]], align 4
+; AVX512-NEXT:    [[WIDE_LOAD12:%.*]] = load <16 x i32>, ptr [[TMP28]], align 4
 ; AVX512-NEXT:    [[TMP30:%.*]] = icmp slt <16 x i32> [[WIDE_LOAD12]], splat (i32 100)
 ; AVX512-NEXT:    [[TMP31:%.*]] = getelementptr i32, ptr [[B]], i64 [[INDEX11]]
-; AVX512-NEXT:    [[TMP32:%.*]] = getelementptr i32, ptr [[TMP31]], i32 0
-; AVX512-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p0(ptr [[TMP32]], i32 4, <16 x i1> [[TMP30]], <16 x i32> poison)
+; AVX512-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p0(ptr [[TMP31]], i32 4, <16 x i1> [[TMP30]], <16 x i32> poison)
 ; AVX512-NEXT:    [[TMP33:%.*]] = add nsw <16 x i32> [[WIDE_MASKED_LOAD13]], [[WIDE_LOAD12]]
 ; AVX512-NEXT:    [[TMP34:%.*]] = getelementptr i32, ptr [[A]], i64 [[INDEX11]]
-; AVX512-NEXT:    [[TMP35:%.*]] = getelementptr i32, ptr [[TMP34]], i32 0
-; AVX512-NEXT:    call void @llvm.masked.store.v16i32.p0(<16 x i32> [[TMP33]], ptr [[TMP35]], i32 4, <16 x i1> [[TMP30]])
+; AVX512-NEXT:    call void @llvm.masked.store.v16i32.p0(<16 x i32> [[TMP33]], ptr [[TMP34]], i32 4, <16 x i1> [[TMP30]])
 ; AVX512-NEXT:    [[INDEX_NEXT14]] = add nuw i64 [[INDEX11]], 16
 ; AVX512-NEXT:    [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT14]], 10000
 ; AVX512-NEXT:    br i1 [[TMP36]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
@@ -332,23 +317,20 @@ define void @foo1_addrspace1(ptr addrspace(1) nocapture %A, ptr addrspace(1) noc
 ; AVX1:       vector.body:
 ; AVX1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr addrspace(1) [[TRIGGER]], i64 [[INDEX]]
-; AVX1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr addrspace(1) [[TMP3]], i32 0
-; AVX1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr addrspace(1) [[TMP4]], align 4
+; AVX1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr addrspace(1) [[TMP3]], align 4
 ; AVX1-NEXT:    [[TMP5:%.*]] = icmp slt <8 x i32> [[WIDE_LOAD]], splat (i32 100)
 ; AVX1-NEXT:    [[TMP6:%.*]] = getelementptr i32, ptr addrspace(1) [[B]], i64 [[INDEX]]
-; AVX1-NEXT:    [[TMP7:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP6]], i32 0
-; AVX1-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p1(ptr addrspace(1) [[TMP7]], i32 4, <8 x i1> [[TMP5]], <8 x i32> poison)
+; AVX1-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p1(ptr addrspace(1) [[TMP6]], i32 4, <8 x i1> [[TMP5]], <8 x i32> poison)
 ; AVX1-NEXT:    [[TMP8:%.*]] = add nsw <8 x i32> [[WIDE_MASKED_LOAD]], [[WIDE_LOAD]]
 ; AVX1-NEXT:    [[TMP9:%.*]] = getelementptr i32, ptr addrspace(1) [[A]], i64 [[INDEX]]
-; AVX1-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP9]], i32 0
-; AVX1-NEXT:    call void @llvm.masked.store.v8i32.p1(<8 x i32> [[TMP8]], ptr addrspace(1) [[TMP10]], i32 4, <8 x i1> [[TMP5]])
+; AVX1-NEXT:    call void @llvm.masked.store.v8i32.p1(<8 x i32> [[TMP8]], ptr addrspace(1) [[TMP9]], i32 4, <8 x i1> [[TMP5]])
 ; AVX1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; AVX1-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10000
 ; AVX1-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; AVX1:       middle.block:
-; AVX1-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; AVX1-NEXT:    br label [[FOR_END:%.*]]
 ; AVX1:       scalar.ph:
-; AVX1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 10000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; AVX1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
 ; AVX1-NEXT:    br label [[FOR_BODY:%.*]]
 ; AVX1:       for.body:
 ; AVX1-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_INC:%.*]] ]
@@ -390,11 +372,10 @@ define void @foo1_addrspace1(ptr addrspace(1) nocapture %A, ptr addrspace(1) noc
 ; AVX2:       vector.body:
 ; AVX2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH1]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr addrspace(1) [[TRIGGER]], i64 [[INDEX]]
-; AVX2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr addrspace(1) [[TMP3]], i32 0
 ; AVX2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr addrspace(1) [[TMP3]], i32 8
 ; AVX2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr addrspace(1) [[TMP3]], i32 16
 ; AVX2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr addrspace(1) [[TMP3]], i32 24
-; AVX2-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr addrspace(1) [[TMP4]], align 4
+; AVX2-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr addrspace(1) [[TMP3]], align 4
 ; AVX2-NEXT:    [[WIDE_LOAD5:%.*]] = load <8 x i32>, ptr addrspace(1) [[TMP5]], align 4
 ; AVX2-NEXT:    [[WIDE_LOAD6:%.*]] = load <8 x i32>, ptr addrspace(1) [[TMP6]], align 4
 ; AVX2-NEXT:    [[WIDE_LOAD7:%.*]] = load <8 x i32>, ptr addrspace(1) [[TMP7]], align 4
@@ -403,11 +384,10 @@ define void @foo1_addrspace1(ptr addrspace(1) nocapture %A, ptr addrspace(1) noc
 ; AVX2-NEXT:    [[TMP10:%.*]] = icmp slt <8 x i32> [[WIDE_LOAD6]], splat (i32 100)
 ; AVX2-NEXT:    [[TMP11:%.*]] = icmp slt <8 x i32> [[WIDE_LOAD7]], splat (i32 100)
 ; AVX2-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr addrspace(1) [[B]], i64 [[INDEX]]
-; AVX2-NEXT:    [[TMP13:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP12]], i32 0
 ; AVX2-NEXT:    [[TMP14:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP12]], i32 8
 ; AVX2-NEXT:    [[TMP15:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP12]], i32 16
 ; AVX2-NEXT:    [[TMP16:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP12]], i32 24
-; AVX2-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p1(ptr addrspace(1) [[TMP13]], i32 4, <8 x i1> [[TMP8]], <8 x i32> poison)
+; AVX2-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p1(ptr addrspace(1) [[TMP12]], i32 4, <8 x i1> [[TMP8]], <8 x i32> poison)
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD8:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p1(ptr addrspace(1) [[TMP14]], i32 4, <8 x i1> [[TMP9]], <8 x i32> poison)
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p1(ptr addrspace(1) [[TMP15]], i32 4, <8 x i1> [[TMP10]], <8 x i32> poison)
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p1(ptr addrspace(1) [[TMP16]], i32 4, <8 x i1> [[TMP11]], <8 x i32> poison)
@@ -416,11 +396,10 @@ define void @foo1_addrspace1(ptr addrspace(1) nocapture %A, ptr addrspace(1) noc
 ; AVX2-NEXT:    [[TMP19:%.*]] = add nsw <8 x i32> [[WIDE_MASKED_LOAD9]], [[WIDE_LOAD6]]
 ; AVX2-NEXT:    [[TMP20:%.*]] = add nsw <8 x i32> [[WIDE_MASKED_LOAD10]], [[WIDE_LOAD7]]
 ; AVX2-NEXT:    [[TMP21:%.*]] = getelementptr i32, ptr addrspace(1) [[A]], i64 [[INDEX]]
-; AVX2-NEXT:    [[TMP22:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP21]], i32 0
 ; AVX2-NEXT:    [[TMP23:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP21]], i32 8
 ; AVX2-NEXT:    [[TMP24:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP21]], i32 16
 ; AVX2-NEXT:    [[TMP25:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP21]], i32 24
-; AVX2-NEXT:    call void @llvm.masked.store.v8i32.p1(<8 x i32> [[TMP17]], ptr addrspace(1) [[TMP22]], i32 4, <8 x i1> [[TMP8]])
+; AVX2-NEXT:    call void @llvm.masked.store.v8i32.p1(<8 x i32> [[TMP17]], ptr addrspace(1) [[TMP21]], i32 4, <8 x i1> [[TMP8]])
 ; AVX2-NEXT:    call void @llvm.masked.store.v8i32.p1(<8 x i32> [[TMP18]], ptr addrspace(1) [[TMP23]], i32 4, <8 x i1> [[TMP9]])
 ; AVX2-NEXT:    call void @llvm.masked.store.v8i32.p1(<8 x i32> [[TMP19]], ptr addrspace(1) [[TMP24]], i32 4, <8 x i1> [[TMP10]])
 ; AVX2-NEXT:    call void @llvm.masked.store.v8i32.p1(<8 x i32> [[TMP20]], ptr addrspace(1) [[TMP25]], i32 4, <8 x i1> [[TMP11]])
@@ -437,16 +416,13 @@ define void @foo1_addrspace1(ptr addrspace(1) nocapture %A, ptr addrspace(1) noc
 ; AVX2:       vec.epilog.vector.body:
 ; AVX2-NEXT:    [[INDEX11:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT14:%.*]], [[FOR_BODY]] ]
 ; AVX2-NEXT:    [[TMP38:%.*]] = getelementptr inbounds i32, ptr addrspace(1) [[TRIGGER]], i64 [[INDEX11]]
-; AVX2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i32, ptr addrspace(1) [[TMP38]], i32 0
-; AVX2-NEXT:    [[WIDE_LOAD12:%.*]] = load <8 x i32>, ptr addrspace(1) [[TMP29]], align 4
+; AVX2-NEXT:    [[WIDE_LOAD12:%.*]] = load <8 x i32>, ptr addrspace(1) [[TMP38]], align 4
 ; AVX2-NEXT:    [[TMP30:%.*]] = icmp slt <8 x i32> [[WIDE_LOAD12]], splat (i32 100)
 ; AVX2-NEXT:    [[TMP31:%.*]] = getelementptr i32, ptr addrspace(1) [[B]], i64 [[INDEX11]]
-; AVX2-NEXT:    [[TMP32:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP31]], i32 0
-; AVX2-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p1(ptr addrspace(1) [[TMP32]], i32 4, <8 x i1> [[TMP30]], <8 x i32> poison)
+; AVX2-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p1(ptr addrspace(1) [[TMP31]], i32 4, <8 x i1> [[TMP30]], <8 x i32> poison)
 ; AVX2-NEXT:    [[TMP33:%.*]] = add nsw <8 x i32> [[WIDE_MASKED_LOAD13]], [[WIDE_LOAD12]]
 ; AVX2-NEXT:    [[TMP34:%.*]] = getelementptr i32, ptr addrspace(1) [[A]], i64 [[INDEX11]]
-; AVX2-NEXT:    [[TMP35:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP34]], i32 0
-; AVX2-NEXT:    call void @llvm.masked.store.v8i32.p1(<8 x i32> [[TMP33]], ptr addrspace(1) [[TMP35]], i32 4, <8 x i1> [[TMP30]])
+; AVX2-NEXT:    call void @llvm.masked.store.v8i32.p1(<8 x i32> [[TMP33]], ptr addrspace(1) [[TMP34]], i32 4, <8 x i1> [[TMP30]])
 ; AVX2-NEXT:    [[INDEX_NEXT14]] = add nuw i64 [[INDEX11]], 8
 ; AVX2-NEXT:    [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT14]], 10000
 ; AVX2-NEXT:    br i1 [[TMP36]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
@@ -495,11 +471,10 @@ define void @foo1_addrspace1(ptr addrspace(1) nocapture %A, ptr addrspace(1) noc
 ; AVX512:       vector.body:
 ; AVX512-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX512-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr addrspace(1) [[TRIGGER]], i64 [[INDEX]]
-; AVX512-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr addrspace(1) [[TMP3]], i32 0
 ; AVX512-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr addrspace(1) [[TMP3]], i32 16
 ; AVX512-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr addrspace(1) [[TMP3]], i32 32
 ; AVX512-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr addrspace(1) [[TMP3]], i32 48
-; AVX512-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i32>, ptr addrspace(1) [[TMP4]], align 4
+; AVX512-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i32>, ptr addrspace(1) [[TMP3]], align 4
 ; AVX512-NEXT:    [[WIDE_LOAD5:%.*]] = load <16 x i32>, ptr addrspace(1) [[TMP5]], align 4
 ; AVX512-NEXT:    [[WIDE_LOAD6:%.*]] = load <16 x i32>, ptr addrspace(1) [[TMP6]], align 4
 ; AVX512-NEXT:    [[WIDE_LOAD7:%.*]] = load <16 x i32>, ptr addrspace(1) [[TMP7]], align 4
@@ -508,11 +483,10 @@ define void @foo1_addrspace1(ptr addrspace(1) nocapture %A, ptr addrspace(1) noc
 ; AVX512-NEXT:    [[TMP10:%.*]] = icmp slt <16 x i32> [[WIDE_LOAD6]], splat (i32 100)
 ; AVX512-NEXT:    [[TMP11:%.*]] = icmp slt <16 x i32> [[WIDE_LOAD7]], splat (i32 100)
 ; AVX512-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr addrspace(1) [[B]], i64 [[INDEX]]
-; AVX512-NEXT:    [[TMP13:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP12]], i32 0
 ; AVX512-NEXT:    [[TMP14:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP12]], i32 16
 ; AVX512-NEXT:    [[TMP15:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP12]], i32 32
 ; AVX512-NEXT:    [[TMP16:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP12]], i32 48
-; AVX512-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p1(ptr addrspace(1) [[TMP13]], i32 4, <16 x i1> [[TMP8]], <16 x i32> poison)
+; AVX512-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p1(ptr addrspace(1) [[TMP12]], i32 4, <16 x i1> [[TMP8]], <16 x i32> poison)
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD8:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p1(ptr addrspace(1) [[TMP14]], i32 4, <16 x i1> [[TMP9]], <16 x i32> poison)
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p1(ptr addrspace(1) [[TMP15]], i32 4, <16 x i1> [[TMP10]], <16 x i32> poison)
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p1(ptr addrspace(1) [[TMP16]], i32 4, <16 x i1> [[TMP11]], <16 x i32> poison)
@@ -521,11 +495,10 @@ define void @foo1_addrspace1(ptr addrspace(1) nocapture %A, ptr addrspace(1) noc
 ; AVX512-NEXT:    [[TMP19:%.*]] = add nsw <16 x i32> [[WIDE_MASKED_LOAD9]], [[WIDE_LOAD6]]
 ; AVX512-NEXT:    [[TMP20:%.*]] = add nsw <16 x i32> [[WIDE_MASKED_LOAD10]], [[WIDE_LOAD7]]
 ; AVX512-NEXT:    [[TMP21:%.*]] = getelementptr i32, ptr addrspace(1) [[A]], i64 [[INDEX]]
-; AVX512-NEXT:    [[TMP22:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP21]], i32 0
 ; AVX512-NEXT:    [[TMP23:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP21]], i32 16
 ; AVX512-NEXT:    [[TMP24:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP21]], i32 32
 ; AVX512-NEXT:    [[TMP25:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP21]], i32 48
-; AVX512-NEXT:    call void @llvm.masked.store.v16i32.p1(<16 x i32> [[TMP17]], ptr addrspace(1) [[TMP22]], i32 4, <16 x i1> [[TMP8]])
+; AVX512-NEXT:    call void @llvm.masked.store.v16i32.p1(<16 x i32> [[TMP17]], ptr addrspace(1) [[TMP21]], i32 4, <16 x i1> [[TMP8]])
 ; AVX512-NEXT:    call void @llvm.masked.store.v16i32.p1(<16 x i32> [[TMP18]], ptr addrspace(1) [[TMP23]], i32 4, <16 x i1> [[TMP9]])
 ; AVX512-NEXT:    call void @llvm.masked.store.v16i32.p1(<16 x i32> [[TMP19]], ptr addrspace(1) [[TMP24]], i32 4, <16 x i1> [[TMP10]])
 ; AVX512-NEXT:    call void @llvm.masked.store.v16i32.p1(<16 x i32> [[TMP20]], ptr addrspace(1) [[TMP25]], i32 4, <16 x i1> [[TMP11]])
@@ -542,16 +515,13 @@ define void @foo1_addrspace1(ptr addrspace(1) nocapture %A, ptr addrspace(1) noc
 ; AVX512:       vec.epilog.vector.body:
 ; AVX512-NEXT:    [[INDEX11:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT14:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; AVX512-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i32, ptr addrspace(1) [[TRIGGER]], i64 [[INDEX11]]
-; AVX512-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i32, ptr addrspace(1) [[TMP28]], i32 0
-; AVX512-NEXT:    [[WIDE_LOAD12:%.*]] = load <16 x i32>, ptr addrspace(1) [[TMP29]], align 4
+; AVX512-NEXT:    [[WIDE_LOAD12:%.*]] = load <16 x i32>, ptr addrspace(1) [[TMP28]], align 4
 ; AVX512-NEXT:    [[TMP30:%.*]] = icmp slt <16 x i32> [[WIDE_LOAD12]], splat (i32 100)
 ; AVX512-NEXT:    [[TMP31:%.*]] = getelementptr i32, ptr addrspace(1) [[B]], i64 [[INDEX11]]
-; AVX512-NEXT:    [[TMP32:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP31]], i32 0
-; AVX512-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p1(ptr addrspace(1) [[TMP32]], i32 4, <16 x i1> [[TMP30]], <16 x i32> poison)
+; AVX512-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p1(ptr addrspace(1) [[TMP31]], i32 4, <16 x i1> [[TMP30]], <16 x i32> poison)
 ; AVX512-NEXT:    [[TMP33:%.*]] = add nsw <16 x i32> [[WIDE_MASKED_LOAD13]], [[WIDE_LOAD12]]
 ; AVX512-NEXT:    [[TMP34:%.*]] = getelementptr i32, ptr addrspace(1) [[A]], i64 [[INDEX11]]
-; AVX512-NEXT:    [[TMP35:%.*]] = getelementptr i32, ptr addrspace(1) [[TMP34]], i32 0
-; AVX512-NEXT:    call void @llvm.masked.store.v16i32.p1(<16 x i32> [[TMP33]], ptr addrspace(1) [[TMP35]], i32 4, <16 x i1> [[TMP30]])
+; AVX512-NEXT:    call void @llvm.masked.store.v16i32.p1(<16 x i32> [[TMP33]], ptr addrspace(1) [[TMP34]], i32 4, <16 x i1> [[TMP30]])
 ; AVX512-NEXT:    [[INDEX_NEXT14]] = add nuw i64 [[INDEX11]], 16
 ; AVX512-NEXT:    [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT14]], 10000
 ; AVX512-NEXT:    br i1 [[TMP36]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
@@ -637,24 +607,21 @@ define void @foo2(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX1:       vector.body:
 ; AVX1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TRIGGER]], i64 [[INDEX]]
-; AVX1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
-; AVX1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP4]], align 4
+; AVX1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP3]], align 4
 ; AVX1-NEXT:    [[TMP5:%.*]] = icmp slt <8 x i32> [[WIDE_LOAD]], splat (i32 100)
 ; AVX1-NEXT:    [[TMP6:%.*]] = getelementptr float, ptr [[B]], i64 [[INDEX]]
-; AVX1-NEXT:    [[TMP7:%.*]] = getelementptr float, ptr [[TMP6]], i32 0
-; AVX1-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP7]], i32 4, <8 x i1> [[TMP5]], <8 x float> poison)
+; AVX1-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP6]], i32 4, <8 x i1> [[TMP5]], <8 x float> poison)
 ; AVX1-NEXT:    [[TMP8:%.*]] = sitofp <8 x i32> [[WIDE_LOAD]] to <8 x float>
 ; AVX1-NEXT:    [[TMP9:%.*]] = fadd <8 x float> [[WIDE_MASKED_LOAD]], [[TMP8]]
 ; AVX1-NEXT:    [[TMP10:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]]
-; AVX1-NEXT:    [[TMP11:%.*]] = getelementptr float, ptr [[TMP10]], i32 0
-; AVX1-NEXT:    call void @llvm.masked.store.v8f32.p0(<8 x float> [[TMP9]], ptr [[TMP11]], i32 4, <8 x i1> [[TMP5]])
+; AVX1-NEXT:    call void @llvm.masked.store.v8f32.p0(<8 x float> [[TMP9]], ptr [[TMP10]], i32 4, <8 x i1> [[TMP5]])
 ; AVX1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; AVX1-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10000
 ; AVX1-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; AVX1:       middle.block:
-; AVX1-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; AVX1-NEXT:    br label [[FOR_END:%.*]]
 ; AVX1:       scalar.ph:
-; AVX1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 10000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; AVX1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
 ; AVX1-NEXT:    br label [[FOR_BODY:%.*]]
 ; AVX1:       for.body:
 ; AVX1-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_INC:%.*]] ]
@@ -697,11 +664,10 @@ define void @foo2(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX2:       vector.body:
 ; AVX2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH1]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TRIGGER]], i64 [[INDEX]]
-; AVX2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
 ; AVX2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 8
 ; AVX2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 16
 ; AVX2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 24
-; AVX2-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP4]], align 4
+; AVX2-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP3]], align 4
 ; AVX2-NEXT:    [[WIDE_LOAD5:%.*]] = load <8 x i32>, ptr [[TMP5]], align 4
 ; AVX2-NEXT:    [[WIDE_LOAD6:%.*]] = load <8 x i32>, ptr [[TMP6]], align 4
 ; AVX2-NEXT:    [[WIDE_LOAD7:%.*]] = load <8 x i32>, ptr [[TMP7]], align 4
@@ -710,11 +676,10 @@ define void @foo2(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX2-NEXT:    [[TMP10:%.*]] = icmp slt <8 x i32> [[WIDE_LOAD6]], splat (i32 100)
 ; AVX2-NEXT:    [[TMP11:%.*]] = icmp slt <8 x i32> [[WIDE_LOAD7]], splat (i32 100)
 ; AVX2-NEXT:    [[TMP12:%.*]] = getelementptr float, ptr [[B]], i64 [[INDEX]]
-; AVX2-NEXT:    [[TMP13:%.*]] = getelementptr float, ptr [[TMP12]], i32 0
 ; AVX2-NEXT:    [[TMP14:%.*]] = getelementptr float, ptr [[TMP12]], i32 8
 ; AVX2-NEXT:    [[TMP15:%.*]] = getelementptr float, ptr [[TMP12]], i32 16
 ; AVX2-NEXT:    [[TMP16:%.*]] = getelementptr float, ptr [[TMP12]], i32 24
-; AVX2-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP13]], i32 4, <8 x i1> [[TMP8]], <8 x float> poison)
+; AVX2-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP12]], i32 4, <8 x i1> [[TMP8]], <8 x float> poison)
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD8:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP14]], i32 4, <8 x i1> [[TMP9]], <8 x float> poison)
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP15]], i32 4, <8 x i1> [[TMP10]], <8 x float> poison)
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP16]], i32 4, <8 x i1> [[TMP11]], <8 x float> poison)
@@ -727,11 +692,10 @@ define void @foo2(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX2-NEXT:    [[TMP23:%.*]] = fadd <8 x float> [[WIDE_MASKED_LOAD9]], [[TMP19]]
 ; AVX2-NEXT:    [[TMP24:%.*]] = fadd <8 x float> [[WIDE_MASKED_LOAD10]], [[TMP20]]
 ; AVX2-NEXT:    [[TMP25:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]]
-; AVX2-NEXT:    [[TMP26:%.*]] = getelementptr float, ptr [[TMP25]], i32 0
 ; AVX2-NEXT:    [[TMP27:%.*]] = getelementptr float, ptr [[TMP25]], i32 8
 ; AVX2-NEXT:    [[TMP28:%.*]] = getelementptr float, ptr [[TMP25]], i32 16
 ; AVX2-NEXT:    [[TMP29:%.*]] = getelementptr float, ptr [[TMP25]], i32 24
-; AVX2-NEXT:    call void @llvm.masked.store.v8f32.p0(<8 x float> [[TMP21]], ptr [[TMP26]], i32 4, <8 x i1> [[TMP8]])
+; AVX2-NEXT:    call void @llvm.masked.store.v8f32.p0(<8 x float> [[TMP21]], ptr [[TMP25]], i32 4, <8 x i1> [[TMP8]])
 ; AVX2-NEXT:    call void @llvm.masked.store.v8f32.p0(<8 x float> [[TMP22]], ptr [[TMP27]], i32 4, <8 x i1> [[TMP9]])
 ; AVX2-NEXT:    call void @llvm.masked.store.v8f32.p0(<8 x float> [[TMP23]], ptr [[TMP28]], i32 4, <8 x i1> [[TMP10]])
 ; AVX2-NEXT:    call void @llvm.masked.store.v8f32.p0(<8 x float> [[TMP24]], ptr [[TMP29]], i32 4, <8 x i1> [[TMP11]])
@@ -748,17 +712,14 @@ define void @foo2(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX2:       vec.epilog.vector.body:
 ; AVX2-NEXT:    [[INDEX11:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT14:%.*]], [[FOR_BODY]] ]
 ; AVX2-NEXT:    [[TMP43:%.*]] = getelementptr inbounds i32, ptr [[TRIGGER]], i64 [[INDEX11]]
-; AVX2-NEXT:    [[TMP33:%.*]] = getelementptr inbounds i32, ptr [[TMP43]], i32 0
-; AVX2-NEXT:    [[WIDE_LOAD12:%.*]] = load <8 x i32>, ptr [[TMP33]], align 4
+; AVX2-NEXT:    [[WIDE_LOAD12:%.*]] = load <8 x i32>, ptr [[TMP43]], align 4
 ; AVX2-NEXT:    [[TMP34:%.*]] = icmp slt <8 x i32> [[WIDE_LOAD12]], splat (i32 100)
 ; AVX2-NEXT:    [[TMP35:%.*]] = getelementptr float, ptr [[B]], i64 [[INDEX11]]
-; AVX2-NEXT:    [[TMP36:%.*]] = getelementptr float, ptr [[TMP35]], i32 0
-; AVX2-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP36]], i32 4, <8 x i1> [[TMP34]], <8 x float> poison)
+; AVX2-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP35]], i32 4, <8 x i1> [[TMP34]], <8 x float> poison)
 ; AVX2-NEXT:    [[TMP37:%.*]] = sitofp <8 x i32> [[WIDE_LOAD12]] to <8 x float>
 ; AVX2-NEXT:    [[TMP38:%.*]] = fadd <8 x float> [[WIDE_MASKED_LOAD13]], [[TMP37]]
 ; AVX2-NEXT:    [[TMP39:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX11]]
-; AVX2-NEXT:    [[TMP40:%.*]] = getelementptr float, ptr [[TMP39]], i32 0
-; AVX2-NEXT:    call void @llvm.masked.store.v8f32.p0(<8 x float> [[TMP38]], ptr [[TMP40]], i32 4, <8 x i1> [[TMP34]])
+; AVX2-NEXT:    call void @llvm.masked.store.v8f32.p0(<8 x float> [[TMP38]], ptr [[TMP39]], i32 4, <8 x i1> [[TMP34]])
 ; AVX2-NEXT:    [[INDEX_NEXT14]] = add nuw i64 [[INDEX11]], 8
 ; AVX2-NEXT:    [[TMP41:%.*]] = icmp eq i64 [[INDEX_NEXT14]], 10000
 ; AVX2-NEXT:    br i1 [[TMP41]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
@@ -808,11 +769,10 @@ define void @foo2(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX512:       vector.body:
 ; AVX512-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX512-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TRIGGER]], i64 [[INDEX]]
-; AVX512-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
 ; AVX512-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 16
 ; AVX512-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 32
 ; AVX512-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 48
-; AVX512-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i32>, ptr [[TMP4]], align 4
+; AVX512-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i32>, ptr [[TMP3]], align 4
 ; AVX512-NEXT:    [[WIDE_LOAD5:%.*]] = load <16 x i32>, ptr [[TMP5]], align 4
 ; AVX512-NEXT:    [[WIDE_LOAD6:%.*]] = load <16 x i32>, ptr [[TMP6]], align 4
 ; AVX512-NEXT:    [[WIDE_LOAD7:%.*]] = load <16 x i32>, ptr [[TMP7]], align 4
@@ -821,11 +781,10 @@ define void @foo2(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX512-NEXT:    [[TMP10:%.*]] = icmp slt <16 x i32> [[WIDE_LOAD6]], splat (i32 100)
 ; AVX512-NEXT:    [[TMP11:%.*]] = icmp slt <16 x i32> [[WIDE_LOAD7]], splat (i32 100)
 ; AVX512-NEXT:    [[TMP12:%.*]] = getelementptr float, ptr [[B]], i64 [[INDEX]]
-; AVX512-NEXT:    [[TMP13:%.*]] = getelementptr float, ptr [[TMP12]], i32 0
 ; AVX512-NEXT:    [[TMP14:%.*]] = getelementptr float, ptr [[TMP12]], i32 16
 ; AVX512-NEXT:    [[TMP15:%.*]] = getelementptr float, ptr [[TMP12]], i32 32
 ; AVX512-NEXT:    [[TMP16:%.*]] = getelementptr float, ptr [[TMP12]], i32 48
-; AVX512-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <16 x float> @llvm.masked.load.v16f32.p0(ptr [[TMP13]], i32 4, <16 x i1> [[TMP8]], <16 x float> poison)
+; AVX512-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <16 x float> @llvm.masked.load.v16f32.p0(ptr [[TMP12]], i32 4, <16 x i1> [[TMP8]], <16 x float> poison)
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD8:%.*]] = call <16 x float> @llvm.masked.load.v16f32.p0(ptr [[TMP14]], i32 4, <16 x i1> [[TMP9]], <16 x float> poison)
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = call <16 x float> @llvm.masked.load.v16f32.p0(ptr [[TMP15]], i32 4, <16 x i1> [[TMP10]], <16 x float> poison)
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <16 x float> @llvm.masked.load.v16f32.p0(ptr [[TMP16]], i32 4, <16 x i1> [[TMP11]], <16 x float> poison)
@@ -838,11 +797,10 @@ define void @foo2(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX512-NEXT:    [[TMP23:%.*]] = fadd <16 x float> [[WIDE_MASKED_LOAD9]], [[TMP19]]
 ; AVX512-NEXT:    [[TMP24:%.*]] = fadd <16 x float> [[WIDE_MASKED_LOAD10]], [[TMP20]]
 ; AVX512-NEXT:    [[TMP25:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX]]
-; AVX512-NEXT:    [[TMP26:%.*]] = getelementptr float, ptr [[TMP25]], i32 0
 ; AVX512-NEXT:    [[TMP27:%.*]] = getelementptr float, ptr [[TMP25]], i32 16
 ; AVX512-NEXT:    [[TMP28:%.*]] = getelementptr float, ptr [[TMP25]], i32 32
 ; AVX512-NEXT:    [[TMP29:%.*]] = getelementptr float, ptr [[TMP25]], i32 48
-; AVX512-NEXT:    call void @llvm.masked.store.v16f32.p0(<16 x float> [[TMP21]], ptr [[TMP26]], i32 4, <16 x i1> [[TMP8]])
+; AVX512-NEXT:    call void @llvm.masked.store.v16f32.p0(<16 x float> [[TMP21]], ptr [[TMP25]], i32 4, <16 x i1> [[TMP8]])
 ; AVX512-NEXT:    call void @llvm.masked.store.v16f32.p0(<16 x float> [[TMP22]], ptr [[TMP27]], i32 4, <16 x i1> [[TMP9]])
 ; AVX512-NEXT:    call void @llvm.masked.store.v16f32.p0(<16 x float> [[TMP23]], ptr [[TMP28]], i32 4, <16 x i1> [[TMP10]])
 ; AVX512-NEXT:    call void @llvm.masked.store.v16f32.p0(<16 x float> [[TMP24]], ptr [[TMP29]], i32 4, <16 x i1> [[TMP11]])
@@ -859,17 +817,14 @@ define void @foo2(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX512:       vec.epilog.vector.body:
 ; AVX512-NEXT:    [[INDEX11:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT14:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; AVX512-NEXT:    [[TMP32:%.*]] = getelementptr inbounds i32, ptr [[TRIGGER]], i64 [[INDEX11]]
-; AVX512-NEXT:    [[TMP33:%.*]] = getelementptr inbounds i32, ptr [[TMP32]], i32 0
-; AVX512-NEXT:    [[WIDE_LOAD12:%.*]] = load <16 x i32>, ptr [[TMP33]], align 4
+; AVX512-NEXT:    [[WIDE_LOAD12:%.*]] = load <16 x i32>, ptr [[TMP32]], align 4
 ; AVX512-NEXT:    [[TMP34:%.*]] = icmp slt <16 x i32> [[WIDE_LOAD12]], splat (i32 100)
 ; AVX512-NEXT:    [[TMP35:%.*]] = getelementptr float, ptr [[B]], i64 [[INDEX11]]
-; AVX512-NEXT:    [[TMP36:%.*]] = getelementptr float, ptr [[TMP35]], i32 0
-; AVX512-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <16 x float> @llvm.masked.load.v16f32.p0(ptr [[TMP36]], i32 4, <16 x i1> [[TMP34]], <16 x float> poison)
+; AVX512-NEXT:    [[WIDE_MASKED_LOAD13:%.*]] = call <16 x float> @llvm.masked.load.v16f32.p0(ptr [[TMP35]], i32 4, <16 x i1> [[TMP34]], <16 x float> poison)
 ; AVX512-NEXT:    [[TMP37:%.*]] = sitofp <16 x i32> [[WIDE_LOAD12]] to <16 x float>
 ; AVX512-NEXT:    [[TMP38:%.*]] = fadd <16 x float> [[WIDE_MASKED_LOAD13]], [[TMP37]]
 ; AVX512-NEXT:    [[TMP39:%.*]] = getelementptr float, ptr [[A]], i64 [[INDEX11]]
-; AVX512-NEXT:    [[TMP40:%.*]] = getelementptr float, ptr [[TMP39]], i32 0
-; AVX512-NEXT:    call void @llvm.masked.store.v16f32.p0(<16 x float> [[TMP38]], ptr [[TMP40]], i32 4, <16 x i1> [[TMP34]])
+; AVX512-NEXT:    call void @llvm.masked.store.v16f32.p0(<16 x float> [[TMP38]], ptr [[TMP39]], i32 4, <16 x i1> [[TMP34]])
 ; AVX512-NEXT:    [[INDEX_NEXT14]] = add nuw i64 [[INDEX11]], 16
 ; AVX512-NEXT:    [[TMP41:%.*]] = icmp eq i64 [[INDEX_NEXT14]], 10000
 ; AVX512-NEXT:    br i1 [[TMP41]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
@@ -959,11 +914,10 @@ define void @foo3(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX1:       vector.body:
 ; AVX1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TRIGGER]], i64 [[INDEX]]
-; AVX1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
 ; AVX1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 4
 ; AVX1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 8
 ; AVX1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 12
-; AVX1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4, !alias.scope [[META8:![0-9]+]]
+; AVX1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4, !alias.scope [[META8:![0-9]+]]
 ; AVX1-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4, !alias.scope [[META8]]
 ; AVX1-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4, !alias.scope [[META8]]
 ; AVX1-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4, !alias.scope [[META8]]
@@ -972,11 +926,10 @@ define void @foo3(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX1-NEXT:    [[TMP8:%.*]] = icmp slt <4 x i32> [[WIDE_LOAD7]], splat (i32 100)
 ; AVX1-NEXT:    [[TMP9:%.*]] = icmp slt <4 x i32> [[WIDE_LOAD8]], splat (i32 100)
 ; AVX1-NEXT:    [[TMP10:%.*]] = getelementptr double, ptr [[B]], i64 [[INDEX]]
-; AVX1-NEXT:    [[TMP11:%.*]] = getelementptr double, ptr [[TMP10]], i32 0
 ; AVX1-NEXT:    [[TMP12:%.*]] = getelementptr double, ptr [[TMP10]], i32 4
 ; AVX1-NEXT:    [[TMP13:%.*]] = getelementptr double, ptr [[TMP10]], i32 8
 ; AVX1-NEXT:    [[TMP14:%.*]] = getelementptr double, ptr [[TMP10]], i32 12
-; AVX1-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP11]], i32 8, <4 x i1> [[TMP6]], <4 x double> poison), !alias.scope [[META11:![0-9]+]]
+; AVX1-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP10]], i32 8, <4 x i1> [[TMP6]], <4 x double> poison), !alias.scope [[META11:![0-9]+]]
 ; AVX1-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP12]], i32 8, <4 x i1> [[TMP7]], <4 x double> poison), !alias.scope [[META11]]
 ; AVX1-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP13]], i32 8, <4 x i1> [[TMP8]], <4 x double> poison), !alias.scope [[META11]]
 ; AVX1-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP14]], i32 8, <4 x i1> [[TMP9]], <4 x double> poison), !alias.scope [[META11]]
@@ -989,11 +942,10 @@ define void @foo3(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX1-NEXT:    [[TMP21:%.*]] = fadd <4 x double> [[WIDE_MASKED_LOAD10]], [[TMP17]]
 ; AVX1-NEXT:    [[TMP22:%.*]] = fadd <4 x double> [[WIDE_MASKED_LOAD11]], [[TMP18]]
 ; AVX1-NEXT:    [[TMP23:%.*]] = getelementptr double, ptr [[A]], i64 [[INDEX]]
-; AVX1-NEXT:    [[TMP24:%.*]] = getelementptr double, ptr [[TMP23]], i32 0
 ; AVX1-NEXT:    [[TMP25:%.*]] = getelementptr double, ptr [[TMP23]], i32 4
 ; AVX1-NEXT:    [[TMP26:%.*]] = getelementptr double, ptr [[TMP23]], i32 8
 ; AVX1-NEXT:    [[TMP27:%.*]] = getelementptr double, ptr [[TMP23]], i32 12
-; AVX1-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> [[TMP19]], ptr [[TMP24]], i32 8, <4 x i1> [[TMP6]]), !alias.scope [[META13:![0-9]+]], !noalias [[META15:![0-9]+]]
+; AVX1-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> [[TMP19]], ptr [[TMP23]], i32 8, <4 x i1> [[TMP6]]), !alias.scope [[META13:![0-9]+]], !noalias [[META15:![0-9]+]]
 ; AVX1-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> [[TMP20]], ptr [[TMP25]], i32 8, <4 x i1> [[TMP7]]), !alias.scope [[META13]], !noalias [[META15]]
 ; AVX1-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> [[TMP21]], ptr [[TMP26]], i32 8, <4 x i1> [[TMP8]]), !alias.scope [[META13]], !noalias [[META15]]
 ; AVX1-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> [[TMP22]], ptr [[TMP27]], i32 8, <4 x i1> [[TMP9]]), !alias.scope [[META13]], !noalias [[META15]]
@@ -1001,9 +953,9 @@ define void @foo3(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX1-NEXT:    [[TMP28:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10000
 ; AVX1-NEXT:    br i1 [[TMP28]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; AVX1:       middle.block:
-; AVX1-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; AVX1-NEXT:    br label [[FOR_END:%.*]]
 ; AVX1:       scalar.ph:
-; AVX1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 10000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; AVX1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
 ; AVX1-NEXT:    br label [[FOR_BODY:%.*]]
 ; AVX1:       for.body:
 ; AVX1-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_INC:%.*]] ]
@@ -1046,11 +998,10 @@ define void @foo3(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX2:       vector.body:
 ; AVX2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TRIGGER]], i64 [[INDEX]]
-; AVX2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
 ; AVX2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 4
 ; AVX2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 8
 ; AVX2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 12
-; AVX2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4, !alias.scope [[META11:![0-9]+]]
+; AVX2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4, !alias.scope [[META11:![0-9]+]]
 ; AVX2-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4, !alias.scope [[META11]]
 ; AVX2-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4, !alias.scope [[META11]]
 ; AVX2-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4, !alias.scope [[META11]]
@@ -1059,11 +1010,10 @@ define void @foo3(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX2-NEXT:    [[TMP8:%.*]] = icmp slt <4 x i32> [[WIDE_LOAD7]], splat (i32 100)
 ; AVX2-NEXT:    [[TMP9:%.*]] = icmp slt <4 x i32> [[WIDE_LOAD8]], splat (i32 100)
 ; AVX2-NEXT:    [[TMP10:%.*]] = getelementptr double, ptr [[B]], i64 [[INDEX]]
-; AVX2-NEXT:    [[TMP11:%.*]] = getelementptr double, ptr [[TMP10]], i32 0
 ; AVX2-NEXT:    [[TMP12:%.*]] = getelementptr double, ptr [[TMP10]], i32 4
 ; AVX2-NEXT:    [[TMP13:%.*]] = getelementptr double, ptr [[TMP10]], i32 8
 ; AVX2-NEXT:    [[TMP14:%.*]] = getelementptr double, ptr [[TMP10]], i32 12
-; AVX2-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP11]], i32 8, <4 x i1> [[TMP6]], <4 x double> poison), !alias.scope [[META14:![0-9]+]]
+; AVX2-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP10]], i32 8, <4 x i1> [[TMP6]], <4 x double> poison), !alias.scope [[META14:![0-9]+]]
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP12]], i32 8, <4 x i1> [[TMP7]], <4 x double> poison), !alias.scope [[META14]]
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP13]], i32 8, <4 x i1> [[TMP8]], <4 x double> poison), !alias.scope [[META14]]
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = call <4 x double> @llvm.masked.load.v4f64.p0(ptr [[TMP14]], i32 8, <4 x i1> [[TMP9]], <4 x double> poison), !alias.scope [[META14]]
@@ -1076,11 +1026,10 @@ define void @foo3(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX2-NEXT:    [[TMP21:%.*]] = fadd <4 x double> [[WIDE_MASKED_LOAD10]], [[TMP17]]
 ; AVX2-NEXT:    [[TMP22:%.*]] = fadd <4 x double> [[WIDE_MASKED_LOAD11]], [[TMP18]]
 ; AVX2-NEXT:    [[TMP23:%.*]] = getelementptr double, ptr [[A]], i64 [[INDEX]]
-; AVX2-NEXT:    [[TMP24:%.*]] = getelementptr double, ptr [[TMP23]], i32 0
 ; AVX2-NEXT:    [[TMP25:%.*]] = getelementptr double, ptr [[TMP23]], i32 4
 ; AVX2-NEXT:    [[TMP26:%.*]] = getelementptr double, ptr [[TMP23]], i32 8
 ; AVX2-NEXT:    [[TMP27:%.*]] = getelementptr double, ptr [[TMP23]], i32 12
-; AVX2-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> [[TMP19]], ptr [[TMP24]], i32 8, <4 x i1> [[TMP6]]), !alias.scope [[META16:![0-9]+]], !noalias [[META18:![0-9]+]]
+; AVX2-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> [[TMP19]], ptr [[TMP23]], i32 8, <4 x i1> [[TMP6]]), !alias.scope [[META16:![0-9]+]], !noalias [[META18:![0-9]+]]
 ; AVX2-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> [[TMP20]], ptr [[TMP25]], i32 8, <4 x i1> [[TMP7]]), !alias.scope [[META16]], !noalias [[META18]]
 ; AVX2-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> [[TMP21]], ptr [[TMP26]], i32 8, <4 x i1> [[TMP8]]), !alias.scope [[META16]], !noalias [[META18]]
 ; AVX2-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> [[TMP22]], ptr [[TMP27]], i32 8, <4 x i1> [[TMP9]]), !alias.scope [[META16]], !noalias [[META18]]
@@ -1088,9 +1037,9 @@ define void @foo3(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX2-NEXT:    [[TMP28:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10000
 ; AVX2-NEXT:    br i1 [[TMP28]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
 ; AVX2:       middle.block:
-; AVX2-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; AVX2-NEXT:    br label [[FOR_END:%.*]]
 ; AVX2:       scalar.ph:
-; AVX2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 10000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; AVX2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
 ; AVX2-NEXT:    br label [[FOR_BODY:%.*]]
 ; AVX2:       for.body:
 ; AVX2-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_INC:%.*]] ]
@@ -1135,11 +1084,10 @@ define void @foo3(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX512:       vector.body:
 ; AVX512-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH1]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX512-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TRIGGER]], i64 [[INDEX]]
-; AVX512-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
 ; AVX512-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 8
 ; AVX512-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 16
 ; AVX512-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 24
-; AVX512-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP2]], align 4, !alias.scope [[META11:![0-9]+]]
+; AVX512-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP1]], align 4, !alias.scope [[META11:![0-9]+]]
 ; AVX512-NEXT:    [[WIDE_LOAD6:%.*]] = load <8 x i32>, ptr [[TMP3]], align 4, !alias.scope [[META11]]
 ; AVX512-NEXT:    [[WIDE_LOAD7:%.*]] = load <8 x i32>, ptr [[TMP4]], align 4, !alias.scope [[META11]]
 ; AVX512-NEXT:    [[WIDE_LOAD8:%.*]] = load <8 x i32>, ptr [[TMP5]], align 4, !alias.scope [[META11]]
@@ -1148,11 +1096,10 @@ define void @foo3(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX512-NEXT:    [[TMP8:%.*]] = icmp slt <8 x i32> [[WIDE_LOAD7]], splat (i32 100)
 ; AVX512-NEXT:    [[TMP9:%.*]] = icmp slt <8 x i32> [[WIDE_LOAD8]], splat (i32 100)
 ; AVX512-NEXT:    [[TMP10:%.*]] = getelementptr double, ptr [[B]], i64 [[INDEX]]
-; AVX512-NEXT:    [[TMP11:%.*]] = getelementptr double, ptr [[TMP10]], i32 0
 ; AVX512-NEXT:    [[TMP12:%.*]] = getelementptr double, ptr [[TMP10]], i32 8
 ; AVX512-NEXT:    [[TMP13:%.*]] = getelementptr double, ptr [[TMP10]], i32 16
 ; AVX512-NEXT:    [[TMP14:%.*]] = getelementptr double, ptr [[TMP10]], i32 24
-; AVX512-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x double> @llvm.masked.load.v8f64.p0(ptr [[TMP11]], i32 8, <8 x i1> [[TMP6]], <8 x double> poison), !alias.scope [[META14:![0-9]+]]
+; AVX512-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x double> @llvm.masked.load.v8f64.p0(ptr [[TMP10]], i32 8, <8 x i1> [[TMP6]], <8 x double> poison), !alias.scope [[META14:![0-9]+]]
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD9:%.*]] = call <8 x double> @llvm.masked.load.v8f64.p0(ptr [[TMP12]], i32 8, <8 x i1> [[TMP7]], <8 x double> poison), !alias.scope [[META14]]
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD10:%.*]] = call <8 x double> @llvm.masked.load.v8f64.p0(ptr [[TMP13]], i32 8, <8 x i1> [[TMP8]], <8 x double> poison), !alias.scope [[META14]]
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD11:%.*]] = call <8 x double> @llvm.masked.load.v8f64.p0(ptr [[TMP14]], i32 8, <8 x i1> [[TMP9]], <8 x double> poison), !alias.scope [[META14]]
@@ -1165,11 +1112,10 @@ define void @foo3(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX512-NEXT:    [[TMP21:%.*]] = fadd <8 x double> [[WIDE_MASKED_LOAD10]], [[TMP17]]
 ; AVX512-NEXT:    [[TMP22:%.*]] = fadd <8 x double> [[WIDE_MASKED_LOAD11]], [[TMP18]]
 ; AVX512-NEXT:    [[TMP23:%.*]] = getelementptr double, ptr [[A]], i64 [[INDEX]]
-; AVX512-NEXT:    [[TMP24:%.*]] = getelementptr double, ptr [[TMP23]], i32 0
 ; AVX512-NEXT:    [[TMP25:%.*]] = getelementptr double, ptr [[TMP23]], i32 8
 ; AVX512-NEXT:    [[TMP26:%.*]] = getelementptr double, ptr [[TMP23]], i32 16
 ; AVX512-NEXT:    [[TMP27:%.*]] = getelementptr double, ptr [[TMP23]], i32 24
-; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> [[TMP19]], ptr [[TMP24]], i32 8, <8 x i1> [[TMP6]]), !alias.scope [[META16:![0-9]+]], !noalias [[META18:![0-9]+]]
+; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> [[TMP19]], ptr [[TMP23]], i32 8, <8 x i1> [[TMP6]]), !alias.scope [[META16:![0-9]+]], !noalias [[META18:![0-9]+]]
 ; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> [[TMP20]], ptr [[TMP25]], i32 8, <8 x i1> [[TMP7]]), !alias.scope [[META16]], !noalias [[META18]]
 ; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> [[TMP21]], ptr [[TMP26]], i32 8, <8 x i1> [[TMP8]]), !alias.scope [[META16]], !noalias [[META18]]
 ; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> [[TMP22]], ptr [[TMP27]], i32 8, <8 x i1> [[TMP9]]), !alias.scope [[META16]], !noalias [[META18]]
@@ -1186,20 +1132,17 @@ define void @foo3(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX512:       vec.epilog.vector.body:
 ; AVX512-NEXT:    [[INDEX12:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT15:%.*]], [[FOR_BODY]] ]
 ; AVX512-NEXT:    [[TMP41:%.*]] = getelementptr inbounds i32, ptr [[TRIGGER]], i64 [[INDEX12]]
-; AVX512-NEXT:    [[TMP31:%.*]] = getelementptr inbounds i32, ptr [[TMP41]], i32 0
-; AVX512-NEXT:    [[WIDE_LOAD13:%.*]] = load <8 x i32>, ptr [[TMP31]], align 4, !alias.scope [[META20:![0-9]+]]
-; AVX512-NEXT:    [[TMP32:%.*]] = icmp slt <8 x i32> [[WIDE_LOAD13]], splat (i32 100)
-; AVX512-NEXT:    [[TMP33:%.*]] = getelementptr double, ptr [[B]], i64 [[INDEX12]]
-; AVX512-NEXT:    [[TMP34:%.*]] = getelementptr double, ptr [[TMP33]], i32 0
-; AVX512-NEXT:    [[WIDE_MASKED_LOAD14:%.*]] = call <8 x double> @llvm.masked.load.v8f64.p0(ptr [[TMP34]], i32 8, <8 x i1> [[TMP32]], <8 x double> poison), !alias.scope [[META23:![0-9]+]]
-; AVX512-NEXT:    [[TMP35:%.*]] = sitofp <8 x i32> [[WIDE_LOAD13]] to <8 x double>
-; AVX512-NEXT:    [[TMP36:%.*]] = fadd <8 x double> [[WIDE_MASKED_LOAD14]], [[TMP35]]
-; AVX512-NEXT:    [[TMP37:%.*]] = getelementptr double, ptr [[A]], i64 [[INDEX12]]
-; AVX512-NEXT:    [[TMP38:%.*]] = getelementptr double, ptr [[TMP37]], i32 0
-; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> [[TMP36]], ptr [[TMP38]], i32 8, <8 x i1> [[TMP32]]), !alias.scope [[META25:![0-9]+]], !noalias [[META27:![0-9]+]]
+; AVX512-NEXT:    [[WIDE_LOAD13:%.*]] = load <8 x i32>, ptr [[TMP41]], align 4, !alias.scope [[META11]]
+; AVX512-NEXT:    [[TMP31:%.*]] = icmp slt <8 x i32> [[WIDE_LOAD13]], splat (i32 100)
+; AVX512-NEXT:    [[TMP32:%.*]] = getelementptr double, ptr [[B]], i64 [[INDEX12]]
+; AVX512-NEXT:    [[WIDE_MASKED_LOAD14:%.*]] = call <8 x double> @llvm.masked.load.v8f64.p0(ptr [[TMP32]], i32 8, <8 x i1> [[TMP31]], <8 x double> poison), !alias.scope [[META14]]
+; AVX512-NEXT:    [[TMP33:%.*]] = sitofp <8 x i32> [[WIDE_LOAD13]] to <8 x double>
+; AVX512-NEXT:    [[TMP34:%.*]] = fadd <8 x double> [[WIDE_MASKED_LOAD14]], [[TMP33]]
+; AVX512-NEXT:    [[TMP35:%.*]] = getelementptr double, ptr [[A]], i64 [[INDEX12]]
+; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> [[TMP34]], ptr [[TMP35]], i32 8, <8 x i1> [[TMP31]]), !alias.scope [[META16]], !noalias [[META18]]
 ; AVX512-NEXT:    [[INDEX_NEXT15]] = add nuw i64 [[INDEX12]], 8
 ; AVX512-NEXT:    [[TMP39:%.*]] = icmp eq i64 [[INDEX_NEXT15]], 10000
-; AVX512-NEXT:    br i1 [[TMP39]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
+; AVX512-NEXT:    br i1 [[TMP39]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
 ; AVX512:       vec.epilog.middle.block:
 ; AVX512-NEXT:    br i1 true, label [[FOR_END]], label [[SCALAR_PH]]
 ; AVX512:       vec.epilog.scalar.ph:
@@ -1222,7 +1165,7 @@ define void @foo3(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX512:       for.inc:
 ; AVX512-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; AVX512-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 10000
-; AVX512-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP29:![0-9]+]]
+; AVX512-NEXT:    br i1 [[EXITCOND]], label [[FOR_END]], label [[FOR_BODY1]], !llvm.loop [[LOOP21:![0-9]+]]
 ; AVX512:       for.end:
 ; AVX512-NEXT:    ret void
 ;
@@ -1312,21 +1255,21 @@ define void @foo4(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX512-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX512-NEXT:    [[VEC_IND:%.*]] = phi <8 x i64> [ <i64 0, i64 16, i64 32, i64 48, i64 64, i64 80, i64 96, i64 112>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX512-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[TRIGGER]], <8 x i64> [[VEC_IND]]
-; AVX512-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <8 x i32> @llvm.masked.gather.v8i32.v8p0(<8 x ptr> [[TMP0]], i32 4, <8 x i1> splat (i1 true), <8 x i32> poison), !alias.scope [[META30:![0-9]+]]
+; AVX512-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <8 x i32> @llvm.masked.gather.v8i32.v8p0(<8 x ptr> [[TMP0]], i32 4, <8 x i1> splat (i1 true), <8 x i32> poison), !alias.scope [[META22:![0-9]+]]
 ; AVX512-NEXT:    [[TMP1:%.*]] = icmp slt <8 x i32> [[WIDE_MASKED_GATHER]], splat (i32 100)
 ; AVX512-NEXT:    [[TMP2:%.*]] = shl nuw nsw <8 x i64> [[VEC_IND]], splat (i64 1)
 ; AVX512-NEXT:    [[TMP3:%.*]] = getelementptr inbounds double, ptr [[B]], <8 x i64> [[TMP2]]
-; AVX512-NEXT:    [[WIDE_MASKED_GATHER6:%.*]] = call <8 x double> @llvm.masked.gather.v8f64.v8p0(<8 x ptr> [[TMP3]], i32 8, <8 x i1> [[TMP1]], <8 x double> poison), !alias.scope [[META33:![0-9]+]]
+; AVX512-NEXT:    [[WIDE_MASKED_GATHER6:%.*]] = call <8 x double> @llvm.masked.gather.v8f64.v8p0(<8 x ptr> [[TMP3]], i32 8, <8 x i1> [[TMP1]], <8 x double> poison), !alias.scope [[META25:![0-9]+]]
 ; AVX512-NEXT:    [[TMP4:%.*]] = sitofp <8 x i32> [[WIDE_MASKED_GATHER]] to <8 x double>
 ; AVX512-NEXT:    [[TMP5:%.*]] = fadd <8 x double> [[WIDE_MASKED_GATHER6]], [[TMP4]]
 ; AVX512-NEXT:    [[TMP6:%.*]] = getelementptr inbounds double, ptr [[A]], <8 x i64> [[VEC_IND]]
-; AVX512-NEXT:    call void @llvm.masked.scatter.v8f64.v8p0(<8 x double> [[TMP5]], <8 x ptr> [[TMP6]], i32 8, <8 x i1> [[TMP1]]), !alias.scope [[META35:![0-9]+]], !noalias [[META37:![0-9]+]]
+; AVX512-NEXT:    call void @llvm.masked.scatter.v8f64.v8p0(<8 x double> [[TMP5]], <8 x ptr> [[TMP6]], i32 8, <8 x i1> [[TMP1]]), !alias.scope [[META27:![0-9]+]], !noalias [[META29:![0-9]+]]
 ; AVX512-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; AVX512-NEXT:    [[VEC_IND_NEXT]] = add <8 x i64> [[VEC_IND]], splat (i64 128)
 ; AVX512-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 624
-; AVX512-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]]
+; AVX512-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
 ; AVX512:       middle.block:
-; AVX512-NEXT:    br i1 false, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; AVX512-NEXT:    br label [[SCALAR_PH]]
 ; AVX512:       scalar.ph:
 ; AVX512-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 9984, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
 ; AVX512-NEXT:    br label [[FOR_BODY:%.*]]
@@ -1348,7 +1291,7 @@ define void @foo4(ptr nocapture %A, ptr nocapture readonly %B, ptr nocapture rea
 ; AVX512:       for.inc:
 ; AVX512-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 16
 ; AVX512-NEXT:    [[CMP:%.*]] = icmp ult i64 [[INDVARS_IV_NEXT]], 10000
-; AVX512-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_END]], !llvm.loop [[LOOP39:![0-9]+]]
+; AVX512-NEXT:    br i1 [[CMP]], label [[FOR_BODY]], label [[FOR_END:%.*]], !llvm.loop [[LOOP31:![0-9]+]]
 ; AVX512:       for.end:
 ; AVX512-NEXT:    ret void
 ;
@@ -1505,9 +1448,9 @@ define void @foo6(ptr nocapture readonly %in, ptr nocapture %out, i32 %size, ptr
 ; AVX2-NEXT:    [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; AVX2-NEXT:    br i1 [[TMP36]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP29:![0-9]+]]
 ; AVX2:       middle.block:
-; AVX2-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; AVX2-NEXT:    br label [[FOR_END:%.*]]
 ; AVX2:       scalar.ph:
-; AVX2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ -1, [[MIDDLE_BLOCK]] ], [ 4095, [[ENTRY:%.*]] ], [ 4095, [[VECTOR_MEMCHECK]] ]
+; AVX2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4095, [[ENTRY:%.*]] ], [ 4095, [[VECTOR_MEMCHECK]] ]
 ; AVX2-NEXT:    br label [[FOR_BODY:%.*]]
 ; AVX2:       for.body:
 ; AVX2-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_INC:%.*]] ]
@@ -1558,13 +1501,13 @@ define void @foo6(ptr nocapture readonly %in, ptr nocapture %out, i32 %size, ptr
 ; AVX512-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 -7
 ; AVX512-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 -24
 ; AVX512-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 -7
-; AVX512-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP3]], align 4, !alias.scope [[META40:![0-9]+]]
+; AVX512-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP3]], align 4, !alias.scope [[META32:![0-9]+]]
 ; AVX512-NEXT:    [[REVERSE:%.*]] = shufflevector <8 x i32> [[WIDE_LOAD]], <8 x i32> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:    [[WIDE_LOAD6:%.*]] = load <8 x i32>, ptr [[TMP5]], align 4, !alias.scope [[META40]]
+; AVX512-NEXT:    [[WIDE_LOAD6:%.*]] = load <8 x i32>, ptr [[TMP5]], align 4, !alias.scope [[META32]]
 ; AVX512-NEXT:    [[REVERSE7:%.*]] = shufflevector <8 x i32> [[WIDE_LOAD6]], <8 x i32> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:    [[WIDE_LOAD8:%.*]] = load <8 x i32>, ptr [[TMP7]], align 4, !alias.scope [[META40]]
+; AVX512-NEXT:    [[WIDE_LOAD8:%.*]] = load <8 x i32>, ptr [[TMP7]], align 4, !alias.scope [[META32]]
 ; AVX512-NEXT:    [[REVERSE9:%.*]] = shufflevector <8 x i32> [[WIDE_LOAD8]], <8 x i32> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:    [[WIDE_LOAD10:%.*]] = load <8 x i32>, ptr [[TMP9]], align 4, !alias.scope [[META40]]
+; AVX512-NEXT:    [[WIDE_LOAD10:%.*]] = load <8 x i32>, ptr [[TMP9]], align 4, !alias.scope [[META32]]
 ; AVX512-NEXT:    [[REVERSE11:%.*]] = shufflevector <8 x i32> [[WIDE_LOAD10]], <8 x i32> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512-NEXT:    [[TMP10:%.*]] = icmp sgt <8 x i32> [[REVERSE]], zeroinitializer
 ; AVX512-NEXT:    [[TMP11:%.*]] = icmp sgt <8 x i32> [[REVERSE7]], zeroinitializer
@@ -1580,16 +1523,16 @@ define void @foo6(ptr nocapture readonly %in, ptr nocapture %out, i32 %size, ptr
 ; AVX512-NEXT:    [[TMP21:%.*]] = getelementptr double, ptr [[TMP14]], i32 -24
 ; AVX512-NEXT:    [[TMP22:%.*]] = getelementptr double, ptr [[TMP21]], i32 -7
 ; AVX512-NEXT:    [[REVERSE12:%.*]] = shufflevector <8 x i1> [[TMP10]], <8 x i1> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x double> @llvm.masked.load.v8f64.p0(ptr [[TMP16]], i32 8, <8 x i1> [[REVERSE12]], <8 x double> poison), !alias.scope [[META43:![0-9]+]]
+; AVX512-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x double> @llvm.masked.load.v8f64.p0(ptr [[TMP16]], i32 8, <8 x i1> [[REVERSE12]], <8 x double> poison), !alias.scope [[META35:![0-9]+]]
 ; AVX512-NEXT:    [[REVERSE13:%.*]] = shufflevector <8 x double> [[WIDE_MASKED_LOAD]], <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512-NEXT:    [[REVERSE14:%.*]] = shufflevector <8 x i1> [[TMP11]], <8 x i1> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:    [[WIDE_MASKED_LOAD15:%.*]] = call <8 x double> @llvm.masked.load.v8f64.p0(ptr [[TMP18]], i32 8, <8 x i1> [[REVERSE14]], <8 x double> poison), !alias.scope [[META43]]
+; AVX512-NEXT:    [[WIDE_MASKED_LOAD15:%.*]] = call <8 x double> @llvm.masked.load.v8f64.p0(ptr [[TMP18]], i32 8, <8 x i1> [[REVERSE14]], <8 x double> poison), !alias.scope [[META35]]
 ; AVX512-NEXT:    [[REVERSE16:%.*]] = shufflevector <8 x double> [[WIDE_MASKED_LOAD15]], <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512-NEXT:    [[REVERSE17:%.*]] = shufflevector <8 x i1> [[TMP12]], <8 x i1> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:    [[WIDE_MASKED_LOAD18:%.*]] = call <8 x double> @llvm.masked.load.v8f64.p0(ptr [[TMP20]], i32 8, <8 x i1> [[REVERSE17]], <8 x double> poison), !alias.scope [[META43]]
+; AVX512-NEXT:    [[WIDE_MASKED_LOAD18:%.*]] = call <8 x double> @llvm.masked.load.v8f64.p0(ptr [[TMP20]], i32 8, <8 x i1> [[REVERSE17]], <8 x double> poison), !alias.scope [[META35]]
 ; AVX512-NEXT:    [[REVERSE19:%.*]] = shufflevector <8 x double> [[WIDE_MASKED_LOAD18]], <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512-NEXT:    [[REVERSE20:%.*]] = shufflevector <8 x i1> [[TMP13]], <8 x i1> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:    [[WIDE_MASKED_LOAD21:%.*]] = call <8 x double> @llvm.masked.load.v8f64.p0(ptr [[TMP22]], i32 8, <8 x i1> [[REVERSE20]], <8 x double> poison), !alias.scope [[META43]]
+; AVX512-NEXT:    [[WIDE_MASKED_LOAD21:%.*]] = call <8 x double> @llvm.masked.load.v8f64.p0(ptr [[TMP22]], i32 8, <8 x i1> [[REVERSE20]], <8 x double> poison), !alias.scope [[META35]]
 ; AVX512-NEXT:    [[REVERSE22:%.*]] = shufflevector <8 x double> [[WIDE_MASKED_LOAD21]], <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
 ; AVX512-NEXT:    [[TMP23:%.*]] = fadd <8 x double> [[REVERSE13]], splat (double 5.000000e-01)
 ; AVX512-NEXT:    [[TMP24:%.*]] = fadd <8 x double> [[REVERSE16]], splat (double 5.000000e-01)
@@ -1605,20 +1548,20 @@ define void @foo6(ptr nocapture readonly %in, ptr nocapture %out, i32 %size, ptr
 ; AVX512-NEXT:    [[TMP34:%.*]] = getelementptr double, ptr [[TMP27]], i32 -24
 ; AVX512-NEXT:    [[TMP35:%.*]] = getelementptr double, ptr [[TMP34]], i32 -7
 ; AVX512-NEXT:    [[REVERSE24:%.*]] = shufflevector <8 x double> [[TMP23]], <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> [[REVERSE24]], ptr [[TMP29]], i32 8, <8 x i1> [[REVERSE12]]), !alias.scope [[META45:![0-9]+]], !noalias [[META47:![0-9]+]]
+; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> [[REVERSE24]], ptr [[TMP29]], i32 8, <8 x i1> [[REVERSE12]]), !alias.scope [[META37:![0-9]+]], !noalias [[META39:![0-9]+]]
 ; AVX512-NEXT:    [[REVERSE26:%.*]] = shufflevector <8 x double> [[TMP24]], <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> [[REVERSE26]], ptr [[TMP31]], i32 8, <8 x i1> [[REVERSE14]]), !alias.scope [[META45]], !noalias [[META47]]
+; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> [[REVERSE26]], ptr [[TMP31]], i32 8, <8 x i1> [[REVERSE14]]), !alias.scope [[META37]], !noalias [[META39]]
 ; AVX512-NEXT:    [[REVERSE28:%.*]] = shufflevector <8 x double> [[TMP25]], <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> [[REVERSE28]], ptr [[TMP33]], i32 8, <8 x i1> [[REVERSE17]]), !alias.scope [[META45]], !noalias [[META47]]
+; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> [[REVERSE28]], ptr [[TMP33]], i32 8, <8 x i1> [[REVERSE17]]), !alias.scope [[META37]], !noalias [[META39]]
 ; AVX512-NEXT:    [[REVERSE30:%.*]] = shufflevector <8 x double> [[TMP26]], <8 x double> poison, <8 x i32> <i32 7, i32 6, i32 5, i32 4, i32 3, i32 2, i32 1, i32 0>
-; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> [[REVERSE30]], ptr [[TMP35]], i32 8, <8 x i1> [[REVERSE20]]), !alias.scope [[META45]], !noalias [[META47]]
+; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> [[REVERSE30]], ptr [[TMP35]], i32 8, <8 x i1> [[REVERSE20]]), !alias.scope [[META37]], !noalias [[META39]]
 ; AVX512-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
 ; AVX512-NEXT:    [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
-; AVX512-NEXT:    br i1 [[TMP36]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP48:![0-9]+]]
+; AVX512-NEXT:    br i1 [[TMP36]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP40:![0-9]+]]
 ; AVX512:       middle.block:
-; AVX512-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; AVX512-NEXT:    br label [[FOR_END:%.*]]
 ; AVX512:       scalar.ph:
-; AVX512-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ -1, [[MIDDLE_BLOCK]] ], [ 4095, [[ENTRY:%.*]] ], [ 4095, [[VECTOR_MEMCHECK]] ]
+; AVX512-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4095, [[ENTRY:%.*]] ], [ 4095, [[VECTOR_MEMCHECK]] ]
 ; AVX512-NEXT:    br label [[FOR_BODY:%.*]]
 ; AVX512:       for.body:
 ; AVX512-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_INC:%.*]] ]
@@ -1636,7 +1579,7 @@ define void @foo6(ptr nocapture readonly %in, ptr nocapture %out, i32 %size, ptr
 ; AVX512:       for.inc:
 ; AVX512-NEXT:    [[INDVARS_IV_NEXT]] = add nsw i64 [[INDVARS_IV]], -1
 ; AVX512-NEXT:    [[CMP:%.*]] = icmp eq i64 [[INDVARS_IV]], 0
-; AVX512-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP49:![0-9]+]]
+; AVX512-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP41:![0-9]+]]
 ; AVX512:       for.end:
 ; AVX512-NEXT:    ret void
 ;
@@ -1694,11 +1637,10 @@ define void @foo7(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX1:       vector.body:
 ; AVX1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TRIGGER:%.*]], i64 [[INDEX]]
-; AVX1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
 ; AVX1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 4
 ; AVX1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 8
 ; AVX1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 12
-; AVX1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; AVX1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
 ; AVX1-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
 ; AVX1-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
 ; AVX1-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP5]], align 1
@@ -1711,11 +1653,10 @@ define void @foo7(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX1-NEXT:    [[TMP16:%.*]] = icmp ne <4 x i8> [[TMP8]], zeroinitializer
 ; AVX1-NEXT:    [[TMP17:%.*]] = icmp ne <4 x i8> [[TMP9]], zeroinitializer
 ; AVX1-NEXT:    [[TMP13:%.*]] = getelementptr ptr, ptr [[IN:%.*]], i64 [[INDEX]]
-; AVX1-NEXT:    [[TMP19:%.*]] = getelementptr ptr, ptr [[TMP13]], i32 0
 ; AVX1-NEXT:    [[TMP20:%.*]] = getelementptr ptr, ptr [[TMP13]], i32 4
 ; AVX1-NEXT:    [[TMP21:%.*]] = getelementptr ptr, ptr [[TMP13]], i32 8
 ; AVX1-NEXT:    [[TMP22:%.*]] = getelementptr ptr, ptr [[TMP13]], i32 12
-; AVX1-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP19]], i32 8, <4 x i1> [[TMP14]], <4 x ptr> poison)
+; AVX1-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP13]], i32 8, <4 x i1> [[TMP14]], <4 x ptr> poison)
 ; AVX1-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP20]], i32 8, <4 x i1> [[TMP15]], <4 x ptr> poison)
 ; AVX1-NEXT:    [[WIDE_MASKED_LOAD5:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP21]], i32 8, <4 x i1> [[TMP16]], <4 x ptr> poison)
 ; AVX1-NEXT:    [[WIDE_MASKED_LOAD6:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP22]], i32 8, <4 x i1> [[TMP17]], <4 x ptr> poison)
@@ -1728,11 +1669,10 @@ define void @foo7(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX1-NEXT:    [[TMP33:%.*]] = select <4 x i1> [[TMP16]], <4 x i1> [[TMP29]], <4 x i1> zeroinitializer
 ; AVX1-NEXT:    [[TMP34:%.*]] = select <4 x i1> [[TMP17]], <4 x i1> [[TMP30]], <4 x i1> zeroinitializer
 ; AVX1-NEXT:    [[TMP35:%.*]] = getelementptr double, ptr [[OUT:%.*]], i64 [[INDEX]]
-; AVX1-NEXT:    [[TMP36:%.*]] = getelementptr double, ptr [[TMP35]], i32 0
 ; AVX1-NEXT:    [[TMP37:%.*]] = getelementptr double, ptr [[TMP35]], i32 4
 ; AVX1-NEXT:    [[TMP38:%.*]] = getelementptr double, ptr [[TMP35]], i32 8
 ; AVX1-NEXT:    [[TMP39:%.*]] = getelementptr double, ptr [[TMP35]], i32 12
-; AVX1-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP36]], i32 8, <4 x i1> [[TMP31]])
+; AVX1-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP35]], i32 8, <4 x i1> [[TMP31]])
 ; AVX1-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP37]], i32 8, <4 x i1> [[TMP32]])
 ; AVX1-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP38]], i32 8, <4 x i1> [[TMP33]])
 ; AVX1-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP39]], i32 8, <4 x i1> [[TMP34]])
@@ -1754,18 +1694,15 @@ define void @foo7(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX1:       vec.epilog.vector.body:
 ; AVX1-NEXT:    [[INDEX10:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDEX_NEXT13:%.*]], [[FOR_BODY]] ]
 ; AVX1-NEXT:    [[TMP56:%.*]] = getelementptr inbounds i8, ptr [[TRIGGER]], i64 [[INDEX10]]
-; AVX1-NEXT:    [[TMP57:%.*]] = getelementptr inbounds i8, ptr [[TMP56]], i32 0
-; AVX1-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x i8>, ptr [[TMP57]], align 1
+; AVX1-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x i8>, ptr [[TMP56]], align 1
 ; AVX1-NEXT:    [[TMP44:%.*]] = and <4 x i8> [[WIDE_LOAD11]], splat (i8 1)
 ; AVX1-NEXT:    [[TMP46:%.*]] = icmp ne <4 x i8> [[TMP44]], zeroinitializer
 ; AVX1-NEXT:    [[TMP47:%.*]] = getelementptr ptr, ptr [[IN]], i64 [[INDEX10]]
-; AVX1-NEXT:    [[TMP48:%.*]] = getelementptr ptr, ptr [[TMP47]], i32 0
-; AVX1-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP48]], i32 8, <4 x i1> [[TMP46]], <4 x ptr> poison)
+; AVX1-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP47]], i32 8, <4 x i1> [[TMP46]], <4 x ptr> poison)
 ; AVX1-NEXT:    [[TMP50:%.*]] = icmp ne <4 x ptr> [[WIDE_MASKED_LOAD12]], zeroinitializer
 ; AVX1-NEXT:    [[TMP51:%.*]] = select <4 x i1> [[TMP46]], <4 x i1> [[TMP50]], <4 x i1> zeroinitializer
 ; AVX1-NEXT:    [[TMP52:%.*]] = getelementptr double, ptr [[OUT]], i64 [[INDEX10]]
-; AVX1-NEXT:    [[TMP53:%.*]] = getelementptr double, ptr [[TMP52]], i32 0
-; AVX1-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP53]], i32 8, <4 x i1> [[TMP51]])
+; AVX1-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP52]], i32 8, <4 x i1> [[TMP51]])
 ; AVX1-NEXT:    [[INDEX_NEXT13]] = add nuw i64 [[INDEX10]], 4
 ; AVX1-NEXT:    [[TMP54:%.*]] = icmp eq i64 [[INDEX_NEXT13]], [[N_VEC9]]
 ; AVX1-NEXT:    br i1 [[TMP54]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
@@ -1818,11 +1755,10 @@ define void @foo7(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX2:       vector.body:
 ; AVX2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TRIGGER:%.*]], i64 [[INDEX]]
-; AVX2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
 ; AVX2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 4
 ; AVX2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 8
 ; AVX2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 12
-; AVX2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; AVX2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
 ; AVX2-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
 ; AVX2-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
 ; AVX2-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP5]], align 1
@@ -1835,11 +1771,10 @@ define void @foo7(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX2-NEXT:    [[TMP16:%.*]] = icmp ne <4 x i8> [[TMP8]], zeroinitializer
 ; AVX2-NEXT:    [[TMP17:%.*]] = icmp ne <4 x i8> [[TMP9]], zeroinitializer
 ; AVX2-NEXT:    [[TMP18:%.*]] = getelementptr ptr, ptr [[IN:%.*]], i64 [[INDEX]]
-; AVX2-NEXT:    [[TMP19:%.*]] = getelementptr ptr, ptr [[TMP18]], i32 0
 ; AVX2-NEXT:    [[TMP20:%.*]] = getelementptr ptr, ptr [[TMP18]], i32 4
 ; AVX2-NEXT:    [[TMP21:%.*]] = getelementptr ptr, ptr [[TMP18]], i32 8
 ; AVX2-NEXT:    [[TMP22:%.*]] = getelementptr ptr, ptr [[TMP18]], i32 12
-; AVX2-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP19]], i32 8, <4 x i1> [[TMP14]], <4 x ptr> poison)
+; AVX2-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP18]], i32 8, <4 x i1> [[TMP14]], <4 x ptr> poison)
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP20]], i32 8, <4 x i1> [[TMP15]], <4 x ptr> poison)
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD5:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP21]], i32 8, <4 x i1> [[TMP16]], <4 x ptr> poison)
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD6:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP22]], i32 8, <4 x i1> [[TMP17]], <4 x ptr> poison)
@@ -1852,11 +1787,10 @@ define void @foo7(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX2-NEXT:    [[TMP33:%.*]] = select <4 x i1> [[TMP16]], <4 x i1> [[TMP29]], <4 x i1> zeroinitializer
 ; AVX2-NEXT:    [[TMP34:%.*]] = select <4 x i1> [[TMP17]], <4 x i1> [[TMP30]], <4 x i1> zeroinitializer
 ; AVX2-NEXT:    [[TMP35:%.*]] = getelementptr double, ptr [[OUT:%.*]], i64 [[INDEX]]
-; AVX2-NEXT:    [[TMP36:%.*]] = getelementptr double, ptr [[TMP35]], i32 0
 ; AVX2-NEXT:    [[TMP37:%.*]] = getelementptr double, ptr [[TMP35]], i32 4
 ; AVX2-NEXT:    [[TMP38:%.*]] = getelementptr double, ptr [[TMP35]], i32 8
 ; AVX2-NEXT:    [[TMP39:%.*]] = getelementptr double, ptr [[TMP35]], i32 12
-; AVX2-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP36]], i32 8, <4 x i1> [[TMP31]])
+; AVX2-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP35]], i32 8, <4 x i1> [[TMP31]])
 ; AVX2-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP37]], i32 8, <4 x i1> [[TMP32]])
 ; AVX2-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP38]], i32 8, <4 x i1> [[TMP33]])
 ; AVX2-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP39]], i32 8, <4 x i1> [[TMP34]])
@@ -1878,18 +1812,15 @@ define void @foo7(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX2:       vec.epilog.vector.body:
 ; AVX2-NEXT:    [[INDEX10:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDEX_NEXT13:%.*]], [[FOR_BODY]] ]
 ; AVX2-NEXT:    [[TMP56:%.*]] = getelementptr inbounds i8, ptr [[TRIGGER]], i64 [[INDEX10]]
-; AVX2-NEXT:    [[TMP57:%.*]] = getelementptr inbounds i8, ptr [[TMP56]], i32 0
-; AVX2-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x i8>, ptr [[TMP57]], align 1
+; AVX2-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x i8>, ptr [[TMP56]], align 1
 ; AVX2-NEXT:    [[TMP44:%.*]] = and <4 x i8> [[WIDE_LOAD11]], splat (i8 1)
 ; AVX2-NEXT:    [[TMP46:%.*]] = icmp ne <4 x i8> [[TMP44]], zeroinitializer
 ; AVX2-NEXT:    [[TMP47:%.*]] = getelementptr ptr, ptr [[IN]], i64 [[INDEX10]]
-; AVX2-NEXT:    [[TMP48:%.*]] = getelementptr ptr, ptr [[TMP47]], i32 0
-; AVX2-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP48]], i32 8, <4 x i1> [[TMP46]], <4 x ptr> poison)
+; AVX2-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP47]], i32 8, <4 x i1> [[TMP46]], <4 x ptr> poison)
 ; AVX2-NEXT:    [[TMP50:%.*]] = icmp ne <4 x ptr> [[WIDE_MASKED_LOAD12]], zeroinitializer
 ; AVX2-NEXT:    [[TMP51:%.*]] = select <4 x i1> [[TMP46]], <4 x i1> [[TMP50]], <4 x i1> zeroinitializer
 ; AVX2-NEXT:    [[TMP52:%.*]] = getelementptr double, ptr [[OUT]], i64 [[INDEX10]]
-; AVX2-NEXT:    [[TMP53:%.*]] = getelementptr double, ptr [[TMP52]], i32 0
-; AVX2-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP53]], i32 8, <4 x i1> [[TMP51]])
+; AVX2-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP52]], i32 8, <4 x i1> [[TMP51]])
 ; AVX2-NEXT:    [[INDEX_NEXT13]] = add nuw i64 [[INDEX10]], 4
 ; AVX2-NEXT:    [[TMP54:%.*]] = icmp eq i64 [[INDEX_NEXT13]], [[N_VEC9]]
 ; AVX2-NEXT:    br i1 [[TMP54]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
@@ -1942,11 +1873,10 @@ define void @foo7(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX512:       vector.body:
 ; AVX512-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX512-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TRIGGER:%.*]], i64 [[INDEX]]
-; AVX512-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
 ; AVX512-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 8
 ; AVX512-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 16
 ; AVX512-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 24
-; AVX512-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP2]], align 1
+; AVX512-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP1]], align 1
 ; AVX512-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i8>, ptr [[TMP3]], align 1
 ; AVX512-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i8>, ptr [[TMP4]], align 1
 ; AVX512-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i8>, ptr [[TMP5]], align 1
@@ -1959,11 +1889,10 @@ define void @foo7(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX512-NEXT:    [[TMP16:%.*]] = icmp ne <8 x i8> [[TMP8]], zeroinitializer
 ; AVX512-NEXT:    [[TMP17:%.*]] = icmp ne <8 x i8> [[TMP9]], zeroinitializer
 ; AVX512-NEXT:    [[TMP18:%.*]] = getelementptr ptr, ptr [[IN:%.*]], i64 [[INDEX]]
-; AVX512-NEXT:    [[TMP19:%.*]] = getelementptr ptr, ptr [[TMP18]], i32 0
 ; AVX512-NEXT:    [[TMP20:%.*]] = getelementptr ptr, ptr [[TMP18]], i32 8
 ; AVX512-NEXT:    [[TMP21:%.*]] = getelementptr ptr, ptr [[TMP18]], i32 16
 ; AVX512-NEXT:    [[TMP22:%.*]] = getelementptr ptr, ptr [[TMP18]], i32 24
-; AVX512-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x ptr> @llvm.masked.load.v8p0.p0(ptr [[TMP19]], i32 8, <8 x i1> [[TMP14]], <8 x ptr> poison)
+; AVX512-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x ptr> @llvm.masked.load.v8p0.p0(ptr [[TMP18]], i32 8, <8 x i1> [[TMP14]], <8 x ptr> poison)
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <8 x ptr> @llvm.masked.load.v8p0.p0(ptr [[TMP20]], i32 8, <8 x i1> [[TMP15]], <8 x ptr> poison)
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD5:%.*]] = call <8 x ptr> @llvm.masked.load.v8p0.p0(ptr [[TMP21]], i32 8, <8 x i1> [[TMP16]], <8 x ptr> poison)
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD6:%.*]] = call <8 x ptr> @llvm.masked.load.v8p0.p0(ptr [[TMP22]], i32 8, <8 x i1> [[TMP17]], <8 x ptr> poison)
@@ -1976,17 +1905,16 @@ define void @foo7(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX512-NEXT:    [[TMP33:%.*]] = select <8 x i1> [[TMP16]], <8 x i1> [[TMP29]], <8 x i1> zeroinitializer
 ; AVX512-NEXT:    [[TMP34:%.*]] = select <8 x i1> [[TMP17]], <8 x i1> [[TMP30]], <8 x i1> zeroinitializer
 ; AVX512-NEXT:    [[TMP35:%.*]] = getelementptr double, ptr [[OUT:%.*]], i64 [[INDEX]]
-; AVX512-NEXT:    [[TMP36:%.*]] = getelementptr double, ptr [[TMP35]], i32 0
 ; AVX512-NEXT:    [[TMP37:%.*]] = getelementptr double, ptr [[TMP35]], i32 8
 ; AVX512-NEXT:    [[TMP38:%.*]] = getelementptr double, ptr [[TMP35]], i32 16
 ; AVX512-NEXT:    [[TMP39:%.*]] = getelementptr double, ptr [[TMP35]], i32 24
-; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> splat (double 5.000000e-01), ptr [[TMP36]], i32 8, <8 x i1> [[TMP31]])
+; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> splat (double 5.000000e-01), ptr [[TMP35]], i32 8, <8 x i1> [[TMP31]])
 ; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> splat (double 5.000000e-01), ptr [[TMP37]], i32 8, <8 x i1> [[TMP32]])
 ; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> splat (double 5.000000e-01), ptr [[TMP38]], i32 8, <8 x i1> [[TMP33]])
 ; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> splat (double 5.000000e-01), ptr [[TMP39]], i32 8, <8 x i1> [[TMP34]])
 ; AVX512-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
 ; AVX512-NEXT:    [[TMP40:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; AVX512-NEXT:    br i1 [[TMP40]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
+; AVX512-NEXT:    br i1 [[TMP40]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP42:![0-9]+]]
 ; AVX512:       middle.block:
 ; AVX512-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
 ; AVX512-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
@@ -2002,21 +1930,18 @@ define void @foo7(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX512:       vec.epilog.vector.body:
 ; AVX512-NEXT:    [[INDEX10:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDEX_NEXT13:%.*]], [[FOR_BODY]] ]
 ; AVX512-NEXT:    [[TMP56:%.*]] = getelementptr inbounds i8, ptr [[TRIGGER]], i64 [[INDEX10]]
-; AVX512-NEXT:    [[TMP57:%.*]] = getelementptr inbounds i8, ptr [[TMP56]], i32 0
-; AVX512-NEXT:    [[WIDE_LOAD11:%.*]] = load <8 x i8>, ptr [[TMP57]], align 1
+; AVX512-NEXT:    [[WIDE_LOAD11:%.*]] = load <8 x i8>, ptr [[TMP56]], align 1
 ; AVX512-NEXT:    [[TMP44:%.*]] = and <8 x i8> [[WIDE_LOAD11]], splat (i8 1)
 ; AVX512-NEXT:    [[TMP46:%.*]] = icmp ne <8 x i8> [[TMP44]], zeroinitializer
 ; AVX512-NEXT:    [[TMP47:%.*]] = getelementptr ptr, ptr [[IN]], i64 [[INDEX10]]
-; AVX512-NEXT:    [[TMP48:%.*]] = getelementptr ptr, ptr [[TMP47]], i32 0
-; AVX512-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <8 x ptr> @llvm.masked.load.v8p0.p0(ptr [[TMP48]], i32 8, <8 x i1> [[TMP46]], <8 x ptr> poison)
+; AVX512-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <8 x ptr> @llvm.masked.load.v8p0.p0(ptr [[TMP47]], i32 8, <8 x i1> [[TMP46]], <8 x ptr> poison)
 ; AVX512-NEXT:    [[TMP50:%.*]] = icmp ne <8 x ptr> [[WIDE_MASKED_LOAD12]], zeroinitializer
 ; AVX512-NEXT:    [[TMP51:%.*]] = select <8 x i1> [[TMP46]], <8 x i1> [[TMP50]], <8 x i1> zeroinitializer
 ; AVX512-NEXT:    [[TMP52:%.*]] = getelementptr double, ptr [[OUT]], i64 [[INDEX10]]
-; AVX512-NEXT:    [[TMP53:%.*]] = getelementptr double, ptr [[TMP52]], i32 0
-; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> splat (double 5.000000e-01), ptr [[TMP53]], i32 8, <8 x i1> [[TMP51]])
+; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> splat (double 5.000000e-01), ptr [[TMP52]], i32 8, <8 x i1> [[TMP51]])
 ; AVX512-NEXT:    [[INDEX_NEXT13]] = add nuw i64 [[INDEX10]], 8
 ; AVX512-NEXT:    [[TMP54:%.*]] = icmp eq i64 [[INDEX_NEXT13]], [[N_VEC9]]
-; AVX512-NEXT:    br i1 [[TMP54]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP51:![0-9]+]]
+; AVX512-NEXT:    br i1 [[TMP54]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP43:![0-9]+]]
 ; AVX512:       vec.epilog.middle.block:
 ; AVX512-NEXT:    [[CMP_N14:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC9]]
 ; AVX512-NEXT:    br i1 [[CMP_N14]], label [[FOR_END_LOOPEXIT]], label [[VEC_EPILOG_SCALAR_PH]]
@@ -2042,7 +1967,7 @@ define void @foo7(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX512:       for.inc:
 ; AVX512-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; AVX512-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]]
-; AVX512-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY1]], !llvm.loop [[LOOP52:![0-9]+]]
+; AVX512-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY1]], !llvm.loop [[LOOP44:![0-9]+]]
 ; AVX512:       for.end.loopexit:
 ; AVX512-NEXT:    br label [[FOR_END]]
 ; AVX512:       for.end:
@@ -2111,11 +2036,10 @@ define void @foo8(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX1:       vector.body:
 ; AVX1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TRIGGER:%.*]], i64 [[INDEX]]
-; AVX1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
 ; AVX1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 4
 ; AVX1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 8
 ; AVX1-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 12
-; AVX1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; AVX1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
 ; AVX1-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
 ; AVX1-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
 ; AVX1-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP5]], align 1
@@ -2128,11 +2052,10 @@ define void @foo8(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX1-NEXT:    [[TMP16:%.*]] = icmp ne <4 x i8> [[TMP8]], zeroinitializer
 ; AVX1-NEXT:    [[TMP17:%.*]] = icmp ne <4 x i8> [[TMP9]], zeroinitializer
 ; AVX1-NEXT:    [[TMP18:%.*]] = getelementptr ptr, ptr [[IN:%.*]], i64 [[INDEX]]
-; AVX1-NEXT:    [[TMP19:%.*]] = getelementptr ptr, ptr [[TMP18]], i32 0
 ; AVX1-NEXT:    [[TMP20:%.*]] = getelementptr ptr, ptr [[TMP18]], i32 4
 ; AVX1-NEXT:    [[TMP21:%.*]] = getelementptr ptr, ptr [[TMP18]], i32 8
 ; AVX1-NEXT:    [[TMP22:%.*]] = getelementptr ptr, ptr [[TMP18]], i32 12
-; AVX1-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP19]], i32 8, <4 x i1> [[TMP14]], <4 x ptr> poison)
+; AVX1-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP18]], i32 8, <4 x i1> [[TMP14]], <4 x ptr> poison)
 ; AVX1-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP20]], i32 8, <4 x i1> [[TMP15]], <4 x ptr> poison)
 ; AVX1-NEXT:    [[WIDE_MASKED_LOAD5:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP21]], i32 8, <4 x i1> [[TMP16]], <4 x ptr> poison)
 ; AVX1-NEXT:    [[WIDE_MASKED_LOAD6:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP22]], i32 8, <4 x i1> [[TMP17]], <4 x ptr> poison)
@@ -2145,11 +2068,10 @@ define void @foo8(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX1-NEXT:    [[TMP33:%.*]] = select <4 x i1> [[TMP16]], <4 x i1> [[TMP29]], <4 x i1> zeroinitializer
 ; AVX1-NEXT:    [[TMP34:%.*]] = select <4 x i1> [[TMP17]], <4 x i1> [[TMP30]], <4 x i1> zeroinitializer
 ; AVX1-NEXT:    [[TMP35:%.*]] = getelementptr double, ptr [[OUT:%.*]], i64 [[INDEX]]
-; AVX1-NEXT:    [[TMP36:%.*]] = getelementptr double, ptr [[TMP35]], i32 0
 ; AVX1-NEXT:    [[TMP37:%.*]] = getelementptr double, ptr [[TMP35]], i32 4
 ; AVX1-NEXT:    [[TMP38:%.*]] = getelementptr double, ptr [[TMP35]], i32 8
 ; AVX1-NEXT:    [[TMP39:%.*]] = getelementptr double, ptr [[TMP35]], i32 12
-; AVX1-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP36]], i32 8, <4 x i1> [[TMP31]])
+; AVX1-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP35]], i32 8, <4 x i1> [[TMP31]])
 ; AVX1-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP37]], i32 8, <4 x i1> [[TMP32]])
 ; AVX1-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP38]], i32 8, <4 x i1> [[TMP33]])
 ; AVX1-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP39]], i32 8, <4 x i1> [[TMP34]])
@@ -2171,18 +2093,15 @@ define void @foo8(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX1:       vec.epilog.vector.body:
 ; AVX1-NEXT:    [[INDEX10:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDEX_NEXT13:%.*]], [[FOR_BODY]] ]
 ; AVX1-NEXT:    [[TMP56:%.*]] = getelementptr inbounds i8, ptr [[TRIGGER]], i64 [[INDEX10]]
-; AVX1-NEXT:    [[TMP57:%.*]] = getelementptr inbounds i8, ptr [[TMP56]], i32 0
-; AVX1-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x i8>, ptr [[TMP57]], align 1
+; AVX1-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x i8>, ptr [[TMP56]], align 1
 ; AVX1-NEXT:    [[TMP44:%.*]] = and <4 x i8> [[WIDE_LOAD11]], splat (i8 1)
 ; AVX1-NEXT:    [[TMP46:%.*]] = icmp ne <4 x i8> [[TMP44]], zeroinitializer
 ; AVX1-NEXT:    [[TMP45:%.*]] = getelementptr ptr, ptr [[IN]], i64 [[INDEX10]]
-; AVX1-NEXT:    [[TMP47:%.*]] = getelementptr ptr, ptr [[TMP45]], i32 0
-; AVX1-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP47]], i32 8, <4 x i1> [[TMP46]], <4 x ptr> poison)
+; AVX1-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP45]], i32 8, <4 x i1> [[TMP46]], <4 x ptr> poison)
 ; AVX1-NEXT:    [[TMP50:%.*]] = icmp ne <4 x ptr> [[WIDE_MASKED_LOAD12]], zeroinitializer
 ; AVX1-NEXT:    [[TMP51:%.*]] = select <4 x i1> [[TMP46]], <4 x i1> [[TMP50]], <4 x i1> zeroinitializer
 ; AVX1-NEXT:    [[TMP52:%.*]] = getelementptr double, ptr [[OUT]], i64 [[INDEX10]]
-; AVX1-NEXT:    [[TMP53:%.*]] = getelementptr double, ptr [[TMP52]], i32 0
-; AVX1-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP53]], i32 8, <4 x i1> [[TMP51]])
+; AVX1-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP52]], i32 8, <4 x i1> [[TMP51]])
 ; AVX1-NEXT:    [[INDEX_NEXT13]] = add nuw i64 [[INDEX10]], 4
 ; AVX1-NEXT:    [[TMP54:%.*]] = icmp eq i64 [[INDEX_NEXT13]], [[N_VEC9]]
 ; AVX1-NEXT:    br i1 [[TMP54]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
@@ -2235,11 +2154,10 @@ define void @foo8(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX2:       vector.body:
 ; AVX2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TRIGGER:%.*]], i64 [[INDEX]]
-; AVX2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
 ; AVX2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 4
 ; AVX2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 8
 ; AVX2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 12
-; AVX2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; AVX2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
 ; AVX2-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
 ; AVX2-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
 ; AVX2-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP5]], align 1
@@ -2252,11 +2170,10 @@ define void @foo8(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX2-NEXT:    [[TMP16:%.*]] = icmp ne <4 x i8> [[TMP8]], zeroinitializer
 ; AVX2-NEXT:    [[TMP12:%.*]] = icmp ne <4 x i8> [[TMP9]], zeroinitializer
 ; AVX2-NEXT:    [[TMP13:%.*]] = getelementptr ptr, ptr [[IN:%.*]], i64 [[INDEX]]
-; AVX2-NEXT:    [[TMP22:%.*]] = getelementptr ptr, ptr [[TMP13]], i32 0
 ; AVX2-NEXT:    [[TMP19:%.*]] = getelementptr ptr, ptr [[TMP13]], i32 4
 ; AVX2-NEXT:    [[TMP20:%.*]] = getelementptr ptr, ptr [[TMP13]], i32 8
 ; AVX2-NEXT:    [[TMP23:%.*]] = getelementptr ptr, ptr [[TMP13]], i32 12
-; AVX2-NEXT:    [[WIDE_MASKED_LOAD6:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP22]], i32 8, <4 x i1> [[TMP17]], <4 x ptr> poison)
+; AVX2-NEXT:    [[WIDE_MASKED_LOAD6:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP13]], i32 8, <4 x i1> [[TMP17]], <4 x ptr> poison)
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP19]], i32 8, <4 x i1> [[TMP15]], <4 x ptr> poison)
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD5:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP20]], i32 8, <4 x i1> [[TMP16]], <4 x ptr> poison)
 ; AVX2-NEXT:    [[WIDE_MASKED_LOAD7:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP23]], i32 8, <4 x i1> [[TMP12]], <4 x ptr> poison)
@@ -2269,11 +2186,10 @@ define void @foo8(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX2-NEXT:    [[TMP33:%.*]] = select <4 x i1> [[TMP16]], <4 x i1> [[TMP29]], <4 x i1> zeroinitializer
 ; AVX2-NEXT:    [[TMP34:%.*]] = select <4 x i1> [[TMP12]], <4 x i1> [[TMP21]], <4 x i1> zeroinitializer
 ; AVX2-NEXT:    [[TMP35:%.*]] = getelementptr double, ptr [[OUT:%.*]], i64 [[INDEX]]
-; AVX2-NEXT:    [[TMP36:%.*]] = getelementptr double, ptr [[TMP35]], i32 0
 ; AVX2-NEXT:    [[TMP37:%.*]] = getelementptr double, ptr [[TMP35]], i32 4
 ; AVX2-NEXT:    [[TMP38:%.*]] = getelementptr double, ptr [[TMP35]], i32 8
 ; AVX2-NEXT:    [[TMP39:%.*]] = getelementptr double, ptr [[TMP35]], i32 12
-; AVX2-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP36]], i32 8, <4 x i1> [[TMP31]])
+; AVX2-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP35]], i32 8, <4 x i1> [[TMP31]])
 ; AVX2-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP37]], i32 8, <4 x i1> [[TMP32]])
 ; AVX2-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP38]], i32 8, <4 x i1> [[TMP33]])
 ; AVX2-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP39]], i32 8, <4 x i1> [[TMP34]])
@@ -2295,18 +2211,15 @@ define void @foo8(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX2:       vec.epilog.vector.body:
 ; AVX2-NEXT:    [[INDEX10:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDEX_NEXT13:%.*]], [[FOR_BODY]] ]
 ; AVX2-NEXT:    [[TMP56:%.*]] = getelementptr inbounds i8, ptr [[TRIGGER]], i64 [[INDEX10]]
-; AVX2-NEXT:    [[TMP57:%.*]] = getelementptr inbounds i8, ptr [[TMP56]], i32 0
-; AVX2-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x i8>, ptr [[TMP57]], align 1
+; AVX2-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x i8>, ptr [[TMP56]], align 1
 ; AVX2-NEXT:    [[TMP44:%.*]] = and <4 x i8> [[WIDE_LOAD11]], splat (i8 1)
 ; AVX2-NEXT:    [[TMP46:%.*]] = icmp ne <4 x i8> [[TMP44]], zeroinitializer
 ; AVX2-NEXT:    [[TMP47:%.*]] = getelementptr ptr, ptr [[IN]], i64 [[INDEX10]]
-; AVX2-NEXT:    [[TMP48:%.*]] = getelementptr ptr, ptr [[TMP47]], i32 0
-; AVX2-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP48]], i32 8, <4 x i1> [[TMP46]], <4 x ptr> poison)
+; AVX2-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <4 x ptr> @llvm.masked.load.v4p0.p0(ptr [[TMP47]], i32 8, <4 x i1> [[TMP46]], <4 x ptr> poison)
 ; AVX2-NEXT:    [[TMP50:%.*]] = icmp ne <4 x ptr> [[WIDE_MASKED_LOAD12]], zeroinitializer
 ; AVX2-NEXT:    [[TMP51:%.*]] = select <4 x i1> [[TMP46]], <4 x i1> [[TMP50]], <4 x i1> zeroinitializer
 ; AVX2-NEXT:    [[TMP52:%.*]] = getelementptr double, ptr [[OUT]], i64 [[INDEX10]]
-; AVX2-NEXT:    [[TMP53:%.*]] = getelementptr double, ptr [[TMP52]], i32 0
-; AVX2-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP53]], i32 8, <4 x i1> [[TMP51]])
+; AVX2-NEXT:    call void @llvm.masked.store.v4f64.p0(<4 x double> splat (double 5.000000e-01), ptr [[TMP52]], i32 8, <4 x i1> [[TMP51]])
 ; AVX2-NEXT:    [[INDEX_NEXT13]] = add nuw i64 [[INDEX10]], 4
 ; AVX2-NEXT:    [[TMP54:%.*]] = icmp eq i64 [[INDEX_NEXT13]], [[N_VEC9]]
 ; AVX2-NEXT:    br i1 [[TMP54]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP35:![0-9]+]]
@@ -2359,11 +2272,10 @@ define void @foo8(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX512:       vector.body:
 ; AVX512-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; AVX512-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TRIGGER:%.*]], i64 [[INDEX]]
-; AVX512-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
 ; AVX512-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 8
 ; AVX512-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 16
 ; AVX512-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 24
-; AVX512-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP2]], align 1
+; AVX512-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP1]], align 1
 ; AVX512-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i8>, ptr [[TMP3]], align 1
 ; AVX512-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x i8>, ptr [[TMP4]], align 1
 ; AVX512-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x i8>, ptr [[TMP5]], align 1
@@ -2376,11 +2288,10 @@ define void @foo8(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX512-NEXT:    [[TMP16:%.*]] = icmp ne <8 x i8> [[TMP8]], zeroinitializer
 ; AVX512-NEXT:    [[TMP17:%.*]] = icmp ne <8 x i8> [[TMP9]], zeroinitializer
 ; AVX512-NEXT:    [[TMP18:%.*]] = getelementptr ptr, ptr [[IN:%.*]], i64 [[INDEX]]
-; AVX512-NEXT:    [[TMP19:%.*]] = getelementptr ptr, ptr [[TMP18]], i32 0
 ; AVX512-NEXT:    [[TMP20:%.*]] = getelementptr ptr, ptr [[TMP18]], i32 8
 ; AVX512-NEXT:    [[TMP21:%.*]] = getelementptr ptr, ptr [[TMP18]], i32 16
 ; AVX512-NEXT:    [[TMP22:%.*]] = getelementptr ptr, ptr [[TMP18]], i32 24
-; AVX512-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x ptr> @llvm.masked.load.v8p0.p0(ptr [[TMP19]], i32 8, <8 x i1> [[TMP14]], <8 x ptr> poison)
+; AVX512-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x ptr> @llvm.masked.load.v8p0.p0(ptr [[TMP18]], i32 8, <8 x i1> [[TMP14]], <8 x ptr> poison)
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD4:%.*]] = call <8 x ptr> @llvm.masked.load.v8p0.p0(ptr [[TMP20]], i32 8, <8 x i1> [[TMP15]], <8 x ptr> poison)
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD5:%.*]] = call <8 x ptr> @llvm.masked.load.v8p0.p0(ptr [[TMP21]], i32 8, <8 x i1> [[TMP16]], <8 x ptr> poison)
 ; AVX512-NEXT:    [[WIDE_MASKED_LOAD6:%.*]] = call <8 x ptr> @llvm.masked.load.v8p0.p0(ptr [[TMP22]], i32 8, <8 x i1> [[TMP17]], <8 x ptr> poison)
@@ -2393,17 +2304,16 @@ define void @foo8(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX512-NEXT:    [[TMP33:%.*]] = select <8 x i1> [[TMP16]], <8 x i1> [[TMP29]], <8 x i1> zeroinitializer
 ; AVX512-NEXT:    [[TMP34:%.*]] = select <8 x i1> [[TMP17]], <8 x i1> [[TMP30]], <8 x i1> zeroinitializer
 ; AVX512-NEXT:    [[TMP35:%.*]] = getelementptr double, ptr [[OUT:%.*]], i64 [[INDEX]]
-; AVX512-NEXT:    [[TMP36:%.*]] = getelementptr double, ptr [[TMP35]], i32 0
 ; AVX512-NEXT:    [[TMP37:%.*]] = getelementptr double, ptr [[TMP35]], i32 8
 ; AVX512-NEXT:    [[TMP38:%.*]] = getelementptr double, ptr [[TMP35]], i32 16
 ; AVX512-NEXT:    [[TMP39:%.*]] = getelementptr double, ptr [[TMP35]], i32 24
-; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> splat (double 5.000000e-01), ptr [[TMP36]], i32 8, <8 x i1> [[TMP31]])
+; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> splat (double 5.000000e-01), ptr [[TMP35]], i32 8, <8 x i1> [[TMP31]])
 ; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> splat (double 5.000000e-01), ptr [[TMP37]], i32 8, <8 x i1> [[TMP32]])
 ; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> splat (double 5.000000e-01), ptr [[TMP38]], i32 8, <8 x i1> [[TMP33]])
 ; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> splat (double 5.000000e-01), ptr [[TMP39]], i32 8, <8 x i1> [[TMP34]])
 ; AVX512-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
 ; AVX512-NEXT:    [[TMP40:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; AVX512-NEXT:    br i1 [[TMP40]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP53:![0-9]+]]
+; AVX512-NEXT:    br i1 [[TMP40]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP45:![0-9]+]]
 ; AVX512:       middle.block:
 ; AVX512-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
 ; AVX512-NEXT:    br i1 [[CMP_N]], label [[FOR_END_LOOPEXIT:%.*]], label [[VEC_EPILOG_ITER_CHECK:%.*]]
@@ -2419,21 +2329,18 @@ define void @foo8(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX512:       vec.epilog.vector.body:
 ; AVX512-NEXT:    [[INDEX10:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDEX_NEXT13:%.*]], [[FOR_BODY]] ]
 ; AVX512-NEXT:    [[TMP56:%.*]] = getelementptr inbounds i8, ptr [[TRIGGER]], i64 [[INDEX10]]
-; AVX512-NEXT:    [[TMP57:%.*]] = getelementptr inbounds i8, ptr [[TMP56]], i32 0
-; AVX512-NEXT:    [[WIDE_LOAD11:%.*]] = load <8 x i8>, ptr [[TMP57]], align 1
+; AVX512-NEXT:    [[WIDE_LOAD11:%.*]] = load <8 x i8>, ptr [[TMP56]], align 1
 ; AVX512-NEXT:    [[TMP44:%.*]] = and <8 x i8> [[WIDE_LOAD11]], splat (i8 1)
 ; AVX512-NEXT:    [[TMP46:%.*]] = icmp ne <8 x i8> [[TMP44]], zeroinitializer
 ; AVX512-NEXT:    [[TMP47:%.*]] = getelementptr ptr, ptr [[IN]], i64 [[INDEX10]]
-; AVX512-NEXT:    [[TMP48:%.*]] = getelementptr ptr, ptr [[TMP47]], i32 0
-; AVX512-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <8 x ptr> @llvm.masked.load.v8p0.p0(ptr [[TMP48]], i32 8, <8 x i1> [[TMP46]], <8 x ptr> poison)
+; AVX512-NEXT:    [[WIDE_MASKED_LOAD12:%.*]] = call <8 x ptr> @llvm.masked.load.v8p0.p0(ptr [[TMP47]], i32 8, <8 x i1> [[TMP46]], <8 x ptr> poison)
 ; AVX512-NEXT:    [[TMP50:%.*]] = icmp ne <8 x ptr> [[WIDE_MASKED_LOAD12]], zeroinitializer
 ; AVX512-NEXT:    [[TMP51:%.*]] = select <8 x i1> [[TMP46]], <8 x i1> [[TMP50]], <8 x i1> zeroinitializer
 ; AVX512-NEXT:    [[TMP52:%.*]] = getelementptr double, ptr [[OUT]], i64 [[INDEX10]]
-; AVX512-NEXT:    [[TMP53:%.*]] = getelementptr double, ptr [[TMP52]], i32 0
-; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> splat (double 5.000000e-01), ptr [[TMP53]], i32 8, <8 x i1> [[TMP51]])
+; AVX512-NEXT:    call void @llvm.masked.store.v8f64.p0(<8 x double> splat (double 5.000000e-01), ptr [[TMP52]], i32 8, <8 x i1> [[TMP51]])
 ; AVX512-NEXT:    [[INDEX_NEXT13]] = add nuw i64 [[INDEX10]], 8
 ; AVX512-NEXT:    [[TMP54:%.*]] = icmp eq i64 [[INDEX_NEXT13]], [[N_VEC9]]
-; AVX512-NEXT:    br i1 [[TMP54]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP54:![0-9]+]]
+; AVX512-NEXT:    br i1 [[TMP54]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP46:![0-9]+]]
 ; AVX512:       vec.epilog.middle.block:
 ; AVX512-NEXT:    [[CMP_N14:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC9]]
 ; AVX512-NEXT:    br i1 [[CMP_N14]], label [[FOR_END_LOOPEXIT]], label [[VEC_EPILOG_SCALAR_PH]]
@@ -2459,7 +2366,7 @@ define void @foo8(ptr noalias nocapture %out, ptr noalias nocapture readonly %in
 ; AVX512:       for.inc:
 ; AVX512-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; AVX512-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], [[WIDE_TRIP_COUNT]]
-; AVX512-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY1]], !llvm.loop [[LOOP55:![0-9]+]]
+; AVX512-NEXT:    br i1 [[EXITCOND]], label [[FOR_END_LOOPEXIT]], label [[FOR_BODY1]], !llvm.loop [[LOOP47:![0-9]+]]
 ; AVX512:       for.end.loopexit:
 ; AVX512-NEXT:    br label [[FOR_END]]
 ; AVX512:       for.end:
diff --git a/llvm/test/Transforms/LoopVectorize/X86/metadata-enable.ll b/llvm/test/Transforms/LoopVectorize/X86/metadata-enable.ll
index c1d3412..f26064a 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/metadata-enable.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/metadata-enable.ll
@@ -1186,24 +1186,22 @@ define i32 @nopragma(ptr noalias nocapture %a, ptr noalias nocapture readonly %b
 ; O1VEC2:       vector.body:
 ; O1VEC2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; O1VEC2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw i32, ptr [[B:%.*]], i64 [[INDEX]]
-; O1VEC2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i32 0
 ; O1VEC2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i32 4
-; O1VEC2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; O1VEC2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; O1VEC2-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
 ; O1VEC2-NEXT:    [[TMP4:%.*]] = add nsw <4 x i32> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
 ; O1VEC2-NEXT:    [[TMP5:%.*]] = add nsw <4 x i32> [[WIDE_LOAD1]], [[BROADCAST_SPLAT]]
 ; O1VEC2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw i32, ptr [[A:%.*]], i64 [[INDEX]]
-; O1VEC2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP6]], i32 0
 ; O1VEC2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP6]], i32 4
-; O1VEC2-NEXT:    store <4 x i32> [[TMP4]], ptr [[TMP7]], align 4
+; O1VEC2-NEXT:    store <4 x i32> [[TMP4]], ptr [[TMP6]], align 4
 ; O1VEC2-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP8]], align 4
 ; O1VEC2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; O1VEC2-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; O1VEC2-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; O1VEC2:       middle.block:
-; O1VEC2-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; O1VEC2-NEXT:    br label [[FOR_END:%.*]]
 ; O1VEC2:       scalar.ph:
-; O1VEC2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 64, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; O1VEC2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; O1VEC2-NEXT:    br label [[FOR_BODY:%.*]]
 ; O1VEC2:       for.body:
 ; O1VEC2-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
@@ -1229,24 +1227,22 @@ define i32 @nopragma(ptr noalias nocapture %a, ptr noalias nocapture readonly %b
 ; OzVEC2:       vector.body:
 ; OzVEC2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; OzVEC2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw i32, ptr [[B:%.*]], i64 [[INDEX]]
-; OzVEC2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i32 0
 ; OzVEC2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i32 4
-; OzVEC2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; OzVEC2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; OzVEC2-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
 ; OzVEC2-NEXT:    [[TMP4:%.*]] = add nsw <4 x i32> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
 ; OzVEC2-NEXT:    [[TMP5:%.*]] = add nsw <4 x i32> [[WIDE_LOAD1]], [[BROADCAST_SPLAT]]
 ; OzVEC2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds nuw i32, ptr [[A:%.*]], i64 [[INDEX]]
-; OzVEC2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP6]], i32 0
 ; OzVEC2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP6]], i32 4
-; OzVEC2-NEXT:    store <4 x i32> [[TMP4]], ptr [[TMP7]], align 4
+; OzVEC2-NEXT:    store <4 x i32> [[TMP4]], ptr [[TMP6]], align 4
 ; OzVEC2-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP8]], align 4
 ; OzVEC2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; OzVEC2-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; OzVEC2-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; OzVEC2:       middle.block:
-; OzVEC2-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; OzVEC2-NEXT:    br label [[FOR_END:%.*]]
 ; OzVEC2:       scalar.ph:
-; OzVEC2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 64, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; OzVEC2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; OzVEC2-NEXT:    br label [[FOR_BODY:%.*]]
 ; OzVEC2:       for.body:
 ; OzVEC2-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/optsize.ll b/llvm/test/Transforms/LoopVectorize/X86/optsize.ll
index 074aeb8..07e2df3 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/optsize.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/optsize.ll
@@ -22,11 +22,10 @@ define i32 @foo_optsize() #0 {
 ; CHECK-NEXT:    [[VEC_IV:%.*]] = add <64 x i32> [[BROADCAST_SPLAT]], <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp ule <64 x i32> [[VEC_IV]], splat (i32 202)
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [32 x i8], ptr @tab, i32 0, i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <64 x i8> @llvm.masked.load.v64i8.p0(ptr [[TMP3]], i32 1, <64 x i1> [[TMP1]], <64 x i8> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <64 x i8> @llvm.masked.load.v64i8.p0(ptr [[TMP2]], i32 1, <64 x i1> [[TMP1]], <64 x i8> poison)
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq <64 x i8> [[WIDE_MASKED_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP5:%.*]] = select <64 x i1> [[TMP4]], <64 x i8> splat (i8 2), <64 x i8> splat (i8 1)
-; CHECK-NEXT:    call void @llvm.masked.store.v64i8.p0(<64 x i8> [[TMP5]], ptr [[TMP3]], i32 1, <64 x i1> [[TMP1]])
+; CHECK-NEXT:    call void @llvm.masked.store.v64i8.p0(<64 x i8> [[TMP5]], ptr [[TMP2]], i32 1, <64 x i1> [[TMP1]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 64
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -60,11 +59,10 @@ define i32 @foo_optsize() #0 {
 ; AUTOVF-NEXT:    [[VEC_IV:%.*]] = add <32 x i32> [[BROADCAST_SPLAT]], <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
 ; AUTOVF-NEXT:    [[TMP1:%.*]] = icmp ule <32 x i32> [[VEC_IV]], splat (i32 202)
 ; AUTOVF-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [32 x i8], ptr @tab, i32 0, i32 [[INDEX]]
-; AUTOVF-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
-; AUTOVF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <32 x i8> @llvm.masked.load.v32i8.p0(ptr [[TMP3]], i32 1, <32 x i1> [[TMP1]], <32 x i8> poison)
+; AUTOVF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <32 x i8> @llvm.masked.load.v32i8.p0(ptr [[TMP2]], i32 1, <32 x i1> [[TMP1]], <32 x i8> poison)
 ; AUTOVF-NEXT:    [[TMP4:%.*]] = icmp eq <32 x i8> [[WIDE_MASKED_LOAD]], zeroinitializer
 ; AUTOVF-NEXT:    [[TMP5:%.*]] = select <32 x i1> [[TMP4]], <32 x i8> splat (i8 2), <32 x i8> splat (i8 1)
-; AUTOVF-NEXT:    call void @llvm.masked.store.v32i8.p0(<32 x i8> [[TMP5]], ptr [[TMP3]], i32 1, <32 x i1> [[TMP1]])
+; AUTOVF-NEXT:    call void @llvm.masked.store.v32i8.p0(<32 x i8> [[TMP5]], ptr [[TMP2]], i32 1, <32 x i1> [[TMP1]])
 ; AUTOVF-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 32
 ; AUTOVF-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 224
 ; AUTOVF-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -120,11 +118,10 @@ define i32 @foo_minsize() #1 {
 ; CHECK-NEXT:    [[VEC_IV:%.*]] = add <64 x i32> [[BROADCAST_SPLAT]], <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31, i32 32, i32 33, i32 34, i32 35, i32 36, i32 37, i32 38, i32 39, i32 40, i32 41, i32 42, i32 43, i32 44, i32 45, i32 46, i32 47, i32 48, i32 49, i32 50, i32 51, i32 52, i32 53, i32 54, i32 55, i32 56, i32 57, i32 58, i32 59, i32 60, i32 61, i32 62, i32 63>
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp ule <64 x i32> [[VEC_IV]], splat (i32 202)
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [32 x i8], ptr @tab, i32 0, i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <64 x i8> @llvm.masked.load.v64i8.p0(ptr [[TMP3]], i32 1, <64 x i1> [[TMP1]], <64 x i8> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <64 x i8> @llvm.masked.load.v64i8.p0(ptr [[TMP2]], i32 1, <64 x i1> [[TMP1]], <64 x i8> poison)
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq <64 x i8> [[WIDE_MASKED_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP5:%.*]] = select <64 x i1> [[TMP4]], <64 x i8> splat (i8 2), <64 x i8> splat (i8 1)
-; CHECK-NEXT:    call void @llvm.masked.store.v64i8.p0(<64 x i8> [[TMP5]], ptr [[TMP3]], i32 1, <64 x i1> [[TMP1]])
+; CHECK-NEXT:    call void @llvm.masked.store.v64i8.p0(<64 x i8> [[TMP5]], ptr [[TMP2]], i32 1, <64 x i1> [[TMP1]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 64
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -158,11 +155,10 @@ define i32 @foo_minsize() #1 {
 ; AUTOVF-NEXT:    [[VEC_IV:%.*]] = add <32 x i32> [[BROADCAST_SPLAT]], <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>
 ; AUTOVF-NEXT:    [[TMP1:%.*]] = icmp ule <32 x i32> [[VEC_IV]], splat (i32 202)
 ; AUTOVF-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [32 x i8], ptr @tab, i32 0, i32 [[INDEX]]
-; AUTOVF-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
-; AUTOVF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <32 x i8> @llvm.masked.load.v32i8.p0(ptr [[TMP3]], i32 1, <32 x i1> [[TMP1]], <32 x i8> poison)
+; AUTOVF-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <32 x i8> @llvm.masked.load.v32i8.p0(ptr [[TMP2]], i32 1, <32 x i1> [[TMP1]], <32 x i8> poison)
 ; AUTOVF-NEXT:    [[TMP4:%.*]] = icmp eq <32 x i8> [[WIDE_MASKED_LOAD]], zeroinitializer
 ; AUTOVF-NEXT:    [[TMP5:%.*]] = select <32 x i1> [[TMP4]], <32 x i8> splat (i8 2), <32 x i8> splat (i8 1)
-; AUTOVF-NEXT:    call void @llvm.masked.store.v32i8.p0(<32 x i8> [[TMP5]], ptr [[TMP3]], i32 1, <32 x i1> [[TMP1]])
+; AUTOVF-NEXT:    call void @llvm.masked.store.v32i8.p0(<32 x i8> [[TMP5]], ptr [[TMP2]], i32 1, <32 x i1> [[TMP1]])
 ; AUTOVF-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 32
 ; AUTOVF-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 224
 ; AUTOVF-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -222,16 +218,15 @@ define void @scev4stride1(ptr noalias nocapture %a, ptr noalias nocapture readon
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], <64 x i32> [[TMP1]]
 ; CHECK-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <64 x i32> @llvm.masked.gather.v64i32.v64p0(<64 x ptr> [[TMP2]], i32 4, <64 x i1> splat (i1 true), <64 x i32> poison)
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    store <64 x i32> [[WIDE_MASKED_GATHER]], ptr [[TMP4]], align 4
+; CHECK-NEXT:    store <64 x i32> [[WIDE_MASKED_GATHER]], ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 64
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <64 x i32> [[VEC_IND]], splat (i32 64)
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END_LOOPEXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 256, [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[FOR_BODY_PREHEADER:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[I_07:%.*]] = phi i32 [ [[INC:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -260,16 +255,15 @@ define void @scev4stride1(ptr noalias nocapture %a, ptr noalias nocapture readon
 ; AUTOVF-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], <8 x i32> [[TMP1]]
 ; AUTOVF-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <8 x i32> @llvm.masked.gather.v8i32.v8p0(<8 x ptr> [[TMP2]], i32 4, <8 x i1> splat (i1 true), <8 x i32> poison)
 ; AUTOVF-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]
-; AUTOVF-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
-; AUTOVF-NEXT:    store <8 x i32> [[WIDE_MASKED_GATHER]], ptr [[TMP4]], align 4
+; AUTOVF-NEXT:    store <8 x i32> [[WIDE_MASKED_GATHER]], ptr [[TMP3]], align 4
 ; AUTOVF-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
 ; AUTOVF-NEXT:    [[VEC_IND_NEXT]] = add <8 x i32> [[VEC_IND]], splat (i32 8)
 ; AUTOVF-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256
 ; AUTOVF-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; AUTOVF:       middle.block:
-; AUTOVF-NEXT:    br i1 true, label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; AUTOVF-NEXT:    br label [[FOR_END_LOOPEXIT:%.*]]
 ; AUTOVF:       scalar.ph:
-; AUTOVF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 256, [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER:%.*]] ]
+; AUTOVF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[FOR_BODY_PREHEADER:%.*]] ]
 ; AUTOVF-NEXT:    br label [[FOR_BODY:%.*]]
 ; AUTOVF:       for.body:
 ; AUTOVF-NEXT:    [[I_07:%.*]] = phi i32 [ [[INC:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/parallel-loops.ll b/llvm/test/Transforms/LoopVectorize/X86/parallel-loops.ll
index f1cbb4a..01d11cc 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/parallel-loops.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/parallel-loops.ll
@@ -108,7 +108,7 @@ define void @parallel_loop(ptr nocapture %a, ptr nocapture %b) nounwind uwtable
 ; CHECK-NEXT:    [[TMP29:%.*]] = icmp eq i64 [[INDEX_NEXT]], 512
 ; CHECK-NEXT:    br i1 [[TMP29]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP2:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
diff --git a/llvm/test/Transforms/LoopVectorize/X86/pr109581-unused-blend.ll b/llvm/test/Transforms/LoopVectorize/X86/pr109581-unused-blend.ll
index 0c99ff6..0cbfb34 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/pr109581-unused-blend.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/pr109581-unused-blend.ll
@@ -46,36 +46,23 @@ define i32 @unused_blend_after_unrolling(ptr %p, i32 %a, i1 %c.1, i16 %x, i16 %y
 ; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <4 x i1> [[TMP0]], i32 0
 ; CHECK-NEXT:    br i1 [[TMP6]], label %[[PRED_SDIV_IF10:.*]], label %[[PRED_SDIV_CONTINUE11:.*]]
 ; CHECK:       [[PRED_SDIV_IF10]]:
-; CHECK-NEXT:    [[TMP7:%.*]] = sdiv i16 [[X]], [[Y]]
-; CHECK-NEXT:    [[TMP8:%.*]] = insertelement <4 x i16> poison, i16 [[TMP7]], i32 0
 ; CHECK-NEXT:    br label %[[PRED_SDIV_CONTINUE11]]
 ; CHECK:       [[PRED_SDIV_CONTINUE11]]:
-; CHECK-NEXT:    [[TMP9:%.*]] = phi <4 x i16> [ poison, %[[PRED_SDIV_CONTINUE9]] ], [ [[TMP8]], %[[PRED_SDIV_IF10]] ]
 ; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <4 x i1> [[TMP0]], i32 1
 ; CHECK-NEXT:    br i1 [[TMP10]], label %[[PRED_SDIV_IF12:.*]], label %[[PRED_SDIV_CONTINUE13:.*]]
 ; CHECK:       [[PRED_SDIV_IF12]]:
-; CHECK-NEXT:    [[TMP11:%.*]] = sdiv i16 [[X]], [[Y]]
-; CHECK-NEXT:    [[TMP12:%.*]] = insertelement <4 x i16> [[TMP9]], i16 [[TMP11]], i32 1
 ; CHECK-NEXT:    br label %[[PRED_SDIV_CONTINUE13]]
 ; CHECK:       [[PRED_SDIV_CONTINUE13]]:
-; CHECK-NEXT:    [[TMP13:%.*]] = phi <4 x i16> [ [[TMP9]], %[[PRED_SDIV_CONTINUE11]] ], [ [[TMP12]], %[[PRED_SDIV_IF12]] ]
 ; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <4 x i1> [[TMP0]], i32 2
 ; CHECK-NEXT:    br i1 [[TMP14]], label %[[PRED_SDIV_IF14:.*]], label %[[PRED_SDIV_CONTINUE15:.*]]
 ; CHECK:       [[PRED_SDIV_IF14]]:
-; CHECK-NEXT:    [[TMP15:%.*]] = sdiv i16 [[X]], [[Y]]
-; CHECK-NEXT:    [[TMP16:%.*]] = insertelement <4 x i16> [[TMP13]], i16 [[TMP15]], i32 2
 ; CHECK-NEXT:    br label %[[PRED_SDIV_CONTINUE15]]
 ; CHECK:       [[PRED_SDIV_CONTINUE15]]:
-; CHECK-NEXT:    [[TMP17:%.*]] = phi <4 x i16> [ [[TMP13]], %[[PRED_SDIV_CONTINUE13]] ], [ [[TMP16]], %[[PRED_SDIV_IF14]] ]
 ; CHECK-NEXT:    [[TMP18:%.*]] = extractelement <4 x i1> [[TMP0]], i32 3
 ; CHECK-NEXT:    br i1 [[TMP18]], label %[[PRED_SDIV_IF16:.*]], label %[[PRED_SDIV_CONTINUE17]]
 ; CHECK:       [[PRED_SDIV_IF16]]:
-; CHECK-NEXT:    [[TMP19:%.*]] = sdiv i16 [[X]], [[Y]]
-; CHECK-NEXT:    [[TMP20:%.*]] = insertelement <4 x i16> [[TMP17]], i16 [[TMP19]], i32 3
 ; CHECK-NEXT:    br label %[[PRED_SDIV_CONTINUE17]]
 ; CHECK:       [[PRED_SDIV_CONTINUE17]]:
-; CHECK-NEXT:    [[TMP21:%.*]] = phi <4 x i16> [ [[TMP17]], %[[PRED_SDIV_CONTINUE15]] ], [ [[TMP20]], %[[PRED_SDIV_IF16]] ]
-; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[BROADCAST_SPLAT]], <4 x i16> zeroinitializer, <4 x i16> [[TMP21]]
 ; CHECK-NEXT:    [[TMP24]] = or <4 x i1> [[VEC_PHI]], [[TMP22]]
 ; CHECK-NEXT:    [[TMP25]] = or <4 x i1> [[VEC_PHI3]], [[TMP22]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
@@ -86,8 +73,7 @@ define i32 @unused_blend_after_unrolling(ptr %p, i32 %a, i1 %c.1, i16 %x, i16 %y
 ; CHECK-NEXT:    [[TMP27:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[BIN_RDX]])
 ; CHECK-NEXT:    [[TMP28:%.*]] = freeze i1 [[TMP27]]
 ; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP28]], i32 0, i32 0
-; CHECK-NEXT:    [[TMP29:%.*]] = extractelement <4 x i16> [[PREDPHI]], i32 3
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 97, %[[MIDDLE_BLOCK]] ], [ 1, %[[ENTRY]] ]
@@ -104,10 +90,10 @@ define i32 @unused_blend_after_unrolling(ptr %p, i32 %a, i1 %c.1, i16 %x, i16 %y
 ; CHECK-NEXT:    [[SEL]] = select i1 [[C]], i32 [[B]], i32 0
 ; CHECK-NEXT:    [[IV_NEXT]] = add i32 [[IV]], 1
 ; CHECK-NEXT:    [[EC:%.*]] = icmp eq i32 [[IV]], 100
-; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       [[EXIT]]:
-; CHECK-NEXT:    [[BLEND_LCSSA:%.*]] = phi i16 [ [[BLEND]], %[[LOOP_LATCH]] ], [ [[TMP29]], %[[MIDDLE_BLOCK]] ]
-; CHECK-NEXT:    [[SEL_LCSSA:%.*]] = phi i32 [ [[SEL]], %[[LOOP_LATCH]] ], [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[BLEND_LCSSA:%.*]] = phi i16 [ [[BLEND]], %[[LOOP_LATCH]] ]
+; CHECK-NEXT:    [[SEL_LCSSA:%.*]] = phi i32 [ [[SEL]], %[[LOOP_LATCH]] ]
 ; CHECK-NEXT:    store i16 [[BLEND_LCSSA]], ptr [[P]], align 2
 ; CHECK-NEXT:    ret i32 [[SEL_LCSSA]]
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/X86/pr131359-dead-for-splice.ll b/llvm/test/Transforms/LoopVectorize/X86/pr131359-dead-for-splice.ll
index c02ec91..18f585b 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/pr131359-dead-for-splice.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/pr131359-dead-for-splice.ll
@@ -23,7 +23,7 @@ define void @no_use() {
 ; CHECK-NEXT:    br i1 [[TMP0]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[STEP_ADD]], i32 3
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 40, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
@@ -33,7 +33,7 @@ define void @no_use() {
 ; CHECK-NEXT:    [[E_0_I]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC_I:%.*]], %[[LOOP]] ]
 ; CHECK-NEXT:    [[INC_I]] = add i32 [[E_0_I]], 1
 ; CHECK-NEXT:    [[EXITCOND_NOT_I:%.*]] = icmp eq i32 [[E_0_I]], 43
-; CHECK-NEXT:    br i1 [[EXITCOND_NOT_I]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT_I]], label %[[EXIT:.*]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    ret void
 ;
@@ -67,7 +67,7 @@ define void @dead_use() {
 ; CHECK-NEXT:    br i1 [[TMP0]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[STEP_ADD]], i32 3
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 40, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
@@ -78,7 +78,7 @@ define void @dead_use() {
 ; CHECK-NEXT:    [[DEAD:%.*]] = add i32 [[D_0_I]], 1
 ; CHECK-NEXT:    [[INC_I]] = add i32 [[E_0_I]], 1
 ; CHECK-NEXT:    [[EXITCOND_NOT_I:%.*]] = icmp eq i32 [[E_0_I]], 43
-; CHECK-NEXT:    br i1 [[EXITCOND_NOT_I]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT_I]], label %[[EXIT:.*]], label %[[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/X86/pr141968-instsimplifyfolder.ll b/llvm/test/Transforms/LoopVectorize/X86/pr141968-instsimplifyfolder.ll
index 9a70ed4..c1d08e1 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/pr141968-instsimplifyfolder.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/pr141968-instsimplifyfolder.ll
@@ -105,9 +105,9 @@ define i8 @pr141968(i1 %cond, i8 %v) {
 ; CHECK-NEXT:    br i1 [[TMP17]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[TMP18:%.*]] = extractelement <16 x i8> [[PREDPHI]], i32 15
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 0, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i8 [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/pr34438.ll b/llvm/test/Transforms/LoopVectorize/X86/pr34438.ll
index 753ef69..df2e35d 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/pr34438.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/pr34438.ll
@@ -16,18 +16,15 @@ define void @small_tc(ptr noalias nocapture %A, ptr noalias nocapture readonly %
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[TMP0:%.*]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x float>, ptr [[TMP1]], align 4, !llvm.access.group [[ACC_GRP0:![0-9]+]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP2:%.*]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x float>, ptr [[TMP3]], align 4, !llvm.access.group [[ACC_GRP0]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x float>, ptr [[TMP0:%.*]], align 4, !llvm.access.group [[ACC_GRP0:![0-9]+]]
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x float>, ptr [[TMP2:%.*]], align 4, !llvm.access.group [[ACC_GRP0]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = fadd fast <8 x float> [[WIDE_LOAD]], [[WIDE_LOAD1]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    store <8 x float> [[TMP4]], ptr [[TMP5]], align 4, !llvm.access.group [[ACC_GRP0]]
+; CHECK-NEXT:    store <8 x float> [[TMP4]], ptr [[TMP2]], align 4, !llvm.access.group [[ACC_GRP0]]
 ; CHECK-NEXT:    br label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 8, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/pr35432.ll b/llvm/test/Transforms/LoopVectorize/X86/pr35432.ll
index 59317fa..639fb86 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/pr35432.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/pr35432.ll
@@ -71,9 +71,8 @@ define i32 @main(ptr %ptr) {
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i32 [[DOTPROMOTED]], [[INDEX]]
 ; CHECK-NEXT:    [[TMP20:%.*]] = add i32 [[OFFSET_IDX]], 1
 ; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i32, ptr [[PTR:%.*]], i32 [[TMP20]]
-; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, ptr [[TMP22]], i32 0
 ; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i32, ptr [[TMP22]], i32 4
-; CHECK-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP24]], align 4
+; CHECK-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP22]], align 4
 ; CHECK-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP25]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
 ; CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/pr36524.ll b/llvm/test/Transforms/LoopVectorize/X86/pr36524.ll
index 52832ed..aad7869 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/pr36524.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/pr36524.ll
@@ -27,17 +27,16 @@ define void @foo(ptr %ptr, ptr %ptr.2) {
 ; CHECK-NEXT:    [[TMP4:%.*]] = add i32 [[TMP0]], 3
 ; CHECK-NEXT:    store i32 [[TMP4]], ptr [[PTR_2]], align 4, !alias.scope [[META0:![0-9]+]], !noalias [[META3:![0-9]+]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[PTR]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <4 x i64> [[VEC_IND]], ptr [[TMP7]], align 8, !alias.scope [[META3]]
+; CHECK-NEXT:    store <4 x i64> [[VEC_IND]], ptr [[TMP6]], align 8, !alias.scope [[META3]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 80
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 80, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i64 [ 82, [[MIDDLE_BLOCK]] ], [ 2, [[ENTRY]] ], [ 2, [[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i64 [ 2, [[ENTRY]] ], [ 2, [[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[CAN_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[CAN_IV_NEXT:%.*]], [[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/pr47437.ll b/llvm/test/Transforms/LoopVectorize/X86/pr47437.ll
index 71e000a..d2f8f22 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/pr47437.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/pr47437.ll
@@ -37,8 +37,7 @@ define void @test_muladd(ptr noalias nocapture %d1, ptr noalias nocapture readon
 ; SSE2-NEXT:    [[TMP11:%.*]] = mul nsw <4 x i32> [[TMP10]], [[TMP9]]
 ; SSE2-NEXT:    [[TMP12:%.*]] = add nsw <4 x i32> [[TMP11]], [[TMP8]]
 ; SSE2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[D1:%.*]], i64 [[INDEX]]
-; SSE2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 0
-; SSE2-NEXT:    store <4 x i32> [[TMP12]], ptr [[TMP14]], align 4
+; SSE2-NEXT:    store <4 x i32> [[TMP12]], ptr [[TMP13]], align 4
 ; SSE2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; SSE2-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; SSE2-NEXT:    br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -125,9 +124,8 @@ define void @test_muladd(ptr noalias nocapture %d1, ptr noalias nocapture readon
 ; SSE41-NEXT:    [[TMP24:%.*]] = add nsw <4 x i32> [[TMP22]], [[TMP16]]
 ; SSE41-NEXT:    [[TMP25:%.*]] = add nsw <4 x i32> [[TMP23]], [[TMP17]]
 ; SSE41-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i32, ptr [[D1:%.*]], i64 [[INDEX]]
-; SSE41-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i32, ptr [[TMP26]], i32 0
 ; SSE41-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i32, ptr [[TMP26]], i32 4
-; SSE41-NEXT:    store <4 x i32> [[TMP24]], ptr [[TMP28]], align 4
+; SSE41-NEXT:    store <4 x i32> [[TMP24]], ptr [[TMP26]], align 4
 ; SSE41-NEXT:    store <4 x i32> [[TMP25]], ptr [[TMP29]], align 4
 ; SSE41-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; SSE41-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -215,9 +213,8 @@ define void @test_muladd(ptr noalias nocapture %d1, ptr noalias nocapture readon
 ; AVX1-NEXT:    [[TMP19:%.*]] = add nsw <4 x i32> [[TMP46]], [[TMP44]]
 ; AVX1-NEXT:    [[TMP20:%.*]] = add nsw <4 x i32> [[TMP47]], [[TMP45]]
 ; AVX1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, ptr [[D1:%.*]], i64 [[INDEX]]
-; AVX1-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i32, ptr [[TMP21]], i32 0
 ; AVX1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i32, ptr [[TMP21]], i32 4
-; AVX1-NEXT:    store <4 x i32> [[TMP19]], ptr [[TMP25]], align 4
+; AVX1-NEXT:    store <4 x i32> [[TMP19]], ptr [[TMP21]], align 4
 ; AVX1-NEXT:    store <4 x i32> [[TMP20]], ptr [[TMP26]], align 4
 ; AVX1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; AVX1-NEXT:    [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -288,8 +285,7 @@ define void @test_muladd(ptr noalias nocapture %d1, ptr noalias nocapture readon
 ; AVX2-NEXT:    [[TMP11:%.*]] = mul nsw <8 x i32> [[TMP10]], [[TMP9]]
 ; AVX2-NEXT:    [[TMP12:%.*]] = add nsw <8 x i32> [[TMP11]], [[TMP8]]
 ; AVX2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[D1:%.*]], i64 [[INDEX]]
-; AVX2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 0
-; AVX2-NEXT:    store <8 x i32> [[TMP12]], ptr [[TMP14]], align 4
+; AVX2-NEXT:    store <8 x i32> [[TMP12]], ptr [[TMP13]], align 4
 ; AVX2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; AVX2-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; AVX2-NEXT:    br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/pr51366-sunk-instruction-used-outside-of-loop.ll b/llvm/test/Transforms/LoopVectorize/X86/pr51366-sunk-instruction-used-outside-of-loop.ll
index db9be20..c8cf9fb 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/pr51366-sunk-instruction-used-outside-of-loop.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/pr51366-sunk-instruction-used-outside-of-loop.ll
@@ -32,14 +32,13 @@ define ptr @test(ptr noalias %src, ptr noalias %dst) {
 ; CHECK-NEXT:    [[TMP13:%.*]] = phi <2 x i32> [ [[TMP9]], [[PRED_LOAD_CONTINUE]] ], [ [[TMP12]], [[PRED_LOAD_IF1]] ]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP3]], <2 x i32> zeroinitializer, <2 x i32> [[TMP13]]
 ; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP15]], align 4
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP14]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label %vector.body
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label %exit, label %scalar.ph
+; CHECK-NEXT:    br label %exit
 ; CHECK:       exit:
 ; CHECK-NEXT:    [[GEP_LCSSA:%.*]] = phi ptr [ %gep.src, %loop.latch ], [ [[TMP2]], %middle.block ]
 ; CHECK-NEXT:    ret ptr [[GEP_LCSSA]]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/pr56319-vector-exit-cond-optimization-epilogue-vectorization.ll b/llvm/test/Transforms/LoopVectorize/X86/pr56319-vector-exit-cond-optimization-epilogue-vectorization.ll
index c405e82..00fe141 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/pr56319-vector-exit-cond-optimization-epilogue-vectorization.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/pr56319-vector-exit-cond-optimization-epilogue-vectorization.ll
@@ -21,8 +21,7 @@ define void @pr56319(ptr noalias %src, ptr noalias %dst) {
 ; CHECK-NEXT:    [[WIDE_VEC:%.*]] = load <96 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[STRIDED_VEC:%.*]] = shufflevector <96 x i8> [[WIDE_VEC]], <96 x i8> poison, <32 x i32> <i32 0, i32 3, i32 6, i32 9, i32 12, i32 15, i32 18, i32 21, i32 24, i32 27, i32 30, i32 33, i32 36, i32 39, i32 42, i32 45, i32 48, i32 51, i32 54, i32 57, i32 60, i32 63, i32 66, i32 69, i32 72, i32 75, i32 78, i32 81, i32 84, i32 87, i32 90, i32 93>
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[DST:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    store <32 x i8> [[STRIDED_VEC]], ptr [[TMP4]], align 1
+; CHECK-NEXT:    store <32 x i8> [[STRIDED_VEC]], ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 32
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -39,8 +38,7 @@ define void @pr56319(ptr noalias %src, ptr noalias %dst) {
 ; CHECK-NEXT:    [[WIDE_VEC2:%.*]] = load <6 x i8>, ptr [[TMP7]], align 1
 ; CHECK-NEXT:    [[STRIDED_VEC3:%.*]] = shufflevector <6 x i8> [[WIDE_VEC2]], <6 x i8> poison, <2 x i32> <i32 0, i32 3>
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[TMP9]], i32 0
-; CHECK-NEXT:    store <2 x i8> [[STRIDED_VEC3]], ptr [[TMP10]], align 1
+; CHECK-NEXT:    store <2 x i8> [[STRIDED_VEC3]], ptr [[TMP9]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT4]] = add nuw i64 [[INDEX1]], 2
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT4]], 36
 ; CHECK-NEXT:    br i1 [[TMP11]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/predicate-switch.ll b/llvm/test/Transforms/LoopVectorize/X86/predicate-switch.ll
index c2668f0..d695de6 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/predicate-switch.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/predicate-switch.ll
@@ -24,12 +24,11 @@ define void @switch_default_to_latch_common_dest(ptr %start, ptr %end) {
 ; COST-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; COST-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
 ; COST-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[START]], i64 [[OFFSET_IDX]]
-; COST-NEXT:    [[TMP6:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 0
-; COST-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP6]], align 1
+; COST-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[NEXT_GEP]], align 1
 ; COST-NEXT:    [[TMP7:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 -12)
 ; COST-NEXT:    [[TMP8:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 13)
 ; COST-NEXT:    [[TMP10:%.*]] = or <4 x i1> [[TMP7]], [[TMP8]]
-; COST-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP6]], i32 1, <4 x i1> [[TMP10]])
+; COST-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP10]])
 ; COST-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; COST-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; COST-NEXT:    br i1 [[TMP11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -77,9 +76,8 @@ define void @switch_default_to_latch_common_dest(ptr %start, ptr %end) {
 ; FORCED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; FORCED-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
 ; FORCED-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[START]], i64 [[OFFSET_IDX]]
-; FORCED-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 0
 ; FORCED-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 4
-; FORCED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP7]], align 1
+; FORCED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[NEXT_GEP]], align 1
 ; FORCED-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i64>, ptr [[TMP8]], align 1
 ; FORCED-NEXT:    [[TMP9:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 -12)
 ; FORCED-NEXT:    [[TMP10:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD4]], splat (i64 -12)
@@ -87,7 +85,7 @@ define void @switch_default_to_latch_common_dest(ptr %start, ptr %end) {
 ; FORCED-NEXT:    [[TMP12:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD4]], splat (i64 13)
 ; FORCED-NEXT:    [[TMP15:%.*]] = or <4 x i1> [[TMP9]], [[TMP11]]
 ; FORCED-NEXT:    [[TMP16:%.*]] = or <4 x i1> [[TMP10]], [[TMP12]]
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP7]], i32 1, <4 x i1> [[TMP15]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP15]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP8]], i32 1, <4 x i1> [[TMP16]])
 ; FORCED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FORCED-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -161,14 +159,13 @@ define void @switch_default_to_latch_common_dest_using_branches(ptr %start, ptr
 ; COST-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; COST-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
 ; COST-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[START]], i64 [[OFFSET_IDX]]
-; COST-NEXT:    [[TMP6:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 0
-; COST-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP6]], align 1
+; COST-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[NEXT_GEP]], align 1
 ; COST-NEXT:    [[TMP7:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 -12)
 ; COST-NEXT:    [[TMP8:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 13)
 ; COST-NEXT:    [[TMP9:%.*]] = xor <4 x i1> [[TMP7]], splat (i1 true)
 ; COST-NEXT:    [[TMP10:%.*]] = select <4 x i1> [[TMP9]], <4 x i1> [[TMP8]], <4 x i1> zeroinitializer
 ; COST-NEXT:    [[TMP11:%.*]] = or <4 x i1> [[TMP10]], [[TMP7]]
-; COST-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP6]], i32 1, <4 x i1> [[TMP11]])
+; COST-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP11]])
 ; COST-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; COST-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; COST-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -217,9 +214,8 @@ define void @switch_default_to_latch_common_dest_using_branches(ptr %start, ptr
 ; FORCED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; FORCED-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
 ; FORCED-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[START]], i64 [[OFFSET_IDX]]
-; FORCED-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 0
 ; FORCED-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 4
-; FORCED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP7]], align 1
+; FORCED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[NEXT_GEP]], align 1
 ; FORCED-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i64>, ptr [[TMP8]], align 1
 ; FORCED-NEXT:    [[TMP9:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 -12)
 ; FORCED-NEXT:    [[TMP10:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD4]], splat (i64 -12)
@@ -231,7 +227,7 @@ define void @switch_default_to_latch_common_dest_using_branches(ptr %start, ptr
 ; FORCED-NEXT:    [[TMP16:%.*]] = select <4 x i1> [[TMP14]], <4 x i1> [[TMP12]], <4 x i1> zeroinitializer
 ; FORCED-NEXT:    [[TMP17:%.*]] = or <4 x i1> [[TMP15]], [[TMP9]]
 ; FORCED-NEXT:    [[TMP18:%.*]] = or <4 x i1> [[TMP16]], [[TMP10]]
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP7]], i32 1, <4 x i1> [[TMP17]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP17]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP8]], i32 1, <4 x i1> [[TMP18]])
 ; FORCED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FORCED-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -341,9 +337,8 @@ define void @switch_all_dests_distinct(ptr %start, ptr %end) {
 ; FORCED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; FORCED-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
 ; FORCED-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[START]], i64 [[OFFSET_IDX]]
-; FORCED-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 0
 ; FORCED-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 4
-; FORCED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP7]], align 1
+; FORCED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[NEXT_GEP]], align 1
 ; FORCED-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i64>, ptr [[TMP8]], align 1
 ; FORCED-NEXT:    [[TMP9:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 -12)
 ; FORCED-NEXT:    [[TMP10:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD4]], splat (i64 -12)
@@ -357,13 +352,13 @@ define void @switch_all_dests_distinct(ptr %start, ptr %end) {
 ; FORCED-NEXT:    [[TMP18:%.*]] = or <4 x i1> [[TMP16]], [[TMP14]]
 ; FORCED-NEXT:    [[TMP19:%.*]] = xor <4 x i1> [[TMP17]], splat (i1 true)
 ; FORCED-NEXT:    [[TMP20:%.*]] = xor <4 x i1> [[TMP18]], splat (i1 true)
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 1), ptr [[TMP7]], i32 1, <4 x i1> [[TMP13]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 1), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP13]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 1), ptr [[TMP8]], i32 1, <4 x i1> [[TMP14]])
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[TMP7]], i32 1, <4 x i1> [[TMP11]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP11]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[TMP8]], i32 1, <4 x i1> [[TMP12]])
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP7]], i32 1, <4 x i1> [[TMP9]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP9]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP8]], i32 1, <4 x i1> [[TMP10]])
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 2), ptr [[TMP7]], i32 1, <4 x i1> [[TMP19]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 2), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP19]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 2), ptr [[TMP8]], i32 1, <4 x i1> [[TMP20]])
 ; FORCED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FORCED-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -460,8 +455,7 @@ define void @switch_all_dests_distinct_variant_using_branches(ptr %start, ptr %e
 ; COST-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; COST-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
 ; COST-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[START]], i64 [[OFFSET_IDX]]
-; COST-NEXT:    [[TMP6:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 0
-; COST-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP6]], align 1
+; COST-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[NEXT_GEP]], align 1
 ; COST-NEXT:    [[TMP7:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 -12)
 ; COST-NEXT:    [[TMP8:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 13)
 ; COST-NEXT:    [[TMP9:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], zeroinitializer
@@ -469,10 +463,10 @@ define void @switch_all_dests_distinct_variant_using_branches(ptr %start, ptr %e
 ; COST-NEXT:    [[TMP11:%.*]] = xor <4 x i1> [[TMP8]], splat (i1 true)
 ; COST-NEXT:    [[TMP12:%.*]] = select <4 x i1> [[TMP10]], <4 x i1> [[TMP11]], <4 x i1> zeroinitializer
 ; COST-NEXT:    [[TMP13:%.*]] = select <4 x i1> [[TMP12]], <4 x i1> [[TMP9]], <4 x i1> zeroinitializer
-; COST-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 1), ptr [[TMP6]], i32 1, <4 x i1> [[TMP13]])
+; COST-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 1), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP13]])
 ; COST-NEXT:    [[TMP14:%.*]] = select <4 x i1> [[TMP10]], <4 x i1> [[TMP8]], <4 x i1> zeroinitializer
-; COST-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[TMP6]], i32 1, <4 x i1> [[TMP14]])
-; COST-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP6]], i32 1, <4 x i1> [[TMP7]])
+; COST-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP14]])
+; COST-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP7]])
 ; COST-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; COST-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; COST-NEXT:    br i1 [[TMP15]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
@@ -533,9 +527,8 @@ define void @switch_all_dests_distinct_variant_using_branches(ptr %start, ptr %e
 ; FORCED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; FORCED-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
 ; FORCED-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[START]], i64 [[OFFSET_IDX]]
-; FORCED-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 0
 ; FORCED-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 4
-; FORCED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP7]], align 1
+; FORCED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[NEXT_GEP]], align 1
 ; FORCED-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i64>, ptr [[TMP8]], align 1
 ; FORCED-NEXT:    [[TMP9:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 -12)
 ; FORCED-NEXT:    [[TMP10:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD4]], splat (i64 -12)
@@ -551,13 +544,13 @@ define void @switch_all_dests_distinct_variant_using_branches(ptr %start, ptr %e
 ; FORCED-NEXT:    [[TMP20:%.*]] = select <4 x i1> [[TMP16]], <4 x i1> [[TMP18]], <4 x i1> zeroinitializer
 ; FORCED-NEXT:    [[TMP21:%.*]] = select <4 x i1> [[TMP19]], <4 x i1> [[TMP13]], <4 x i1> zeroinitializer
 ; FORCED-NEXT:    [[TMP22:%.*]] = select <4 x i1> [[TMP20]], <4 x i1> [[TMP14]], <4 x i1> zeroinitializer
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 1), ptr [[TMP7]], i32 1, <4 x i1> [[TMP21]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 1), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP21]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 1), ptr [[TMP8]], i32 1, <4 x i1> [[TMP22]])
 ; FORCED-NEXT:    [[TMP23:%.*]] = select <4 x i1> [[TMP15]], <4 x i1> [[TMP11]], <4 x i1> zeroinitializer
 ; FORCED-NEXT:    [[TMP24:%.*]] = select <4 x i1> [[TMP16]], <4 x i1> [[TMP12]], <4 x i1> zeroinitializer
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[TMP7]], i32 1, <4 x i1> [[TMP23]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP23]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[TMP8]], i32 1, <4 x i1> [[TMP24]])
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP7]], i32 1, <4 x i1> [[TMP9]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP9]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP8]], i32 1, <4 x i1> [[TMP10]])
 ; FORCED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FORCED-NEXT:    [[TMP25:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -694,9 +687,8 @@ define void @switch_multiple_common_dests(ptr %start, ptr %end) {
 ; FORCED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; FORCED-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
 ; FORCED-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[START]], i64 [[OFFSET_IDX]]
-; FORCED-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 0
 ; FORCED-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 4
-; FORCED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP7]], align 1
+; FORCED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[NEXT_GEP]], align 1
 ; FORCED-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i64>, ptr [[TMP8]], align 1
 ; FORCED-NEXT:    [[TMP23:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 -12)
 ; FORCED-NEXT:    [[TMP24:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD4]], splat (i64 -12)
@@ -718,11 +710,11 @@ define void @switch_multiple_common_dests(ptr %start, ptr %end) {
 ; FORCED-NEXT:    [[TMP38:%.*]] = or <4 x i1> [[TMP28]], [[TMP36]]
 ; FORCED-NEXT:    [[TMP39:%.*]] = xor <4 x i1> [[TMP37]], splat (i1 true)
 ; FORCED-NEXT:    [[TMP40:%.*]] = xor <4 x i1> [[TMP38]], splat (i1 true)
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[TMP7]], i32 1, <4 x i1> [[TMP35]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP35]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[TMP8]], i32 1, <4 x i1> [[TMP36]])
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP7]], i32 1, <4 x i1> [[TMP27]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP27]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP8]], i32 1, <4 x i1> [[TMP28]])
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 2), ptr [[TMP7]], i32 1, <4 x i1> [[TMP39]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 2), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP39]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 2), ptr [[TMP8]], i32 1, <4 x i1> [[TMP40]])
 ; FORCED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FORCED-NEXT:    [[TMP41:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -844,9 +836,8 @@ define void @switch4_default_common_dest_with_case(ptr %start, ptr %end) {
 ; FORCED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; FORCED-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
 ; FORCED-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[START]], i64 [[OFFSET_IDX]]
-; FORCED-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 0
 ; FORCED-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 4
-; FORCED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP7]], align 1
+; FORCED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[NEXT_GEP]], align 1
 ; FORCED-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i64>, ptr [[TMP8]], align 1
 ; FORCED-NEXT:    [[TMP15:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 -12)
 ; FORCED-NEXT:    [[TMP16:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD4]], splat (i64 -12)
@@ -856,11 +847,11 @@ define void @switch4_default_common_dest_with_case(ptr %start, ptr %end) {
 ; FORCED-NEXT:    [[TMP18:%.*]] = or <4 x i1> [[TMP16]], [[TMP12]]
 ; FORCED-NEXT:    [[TMP20:%.*]] = xor <4 x i1> [[TMP17]], splat (i1 true)
 ; FORCED-NEXT:    [[TMP21:%.*]] = xor <4 x i1> [[TMP18]], splat (i1 true)
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[TMP7]], i32 1, <4 x i1> [[TMP11]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP11]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[TMP8]], i32 1, <4 x i1> [[TMP12]])
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP7]], i32 1, <4 x i1> [[TMP15]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP15]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP8]], i32 1, <4 x i1> [[TMP16]])
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 2), ptr [[TMP7]], i32 1, <4 x i1> [[TMP20]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 2), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP20]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 2), ptr [[TMP8]], i32 1, <4 x i1> [[TMP21]])
 ; FORCED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FORCED-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -952,8 +943,7 @@ define void @switch_under_br_default_common_dest_with_case(ptr %start, ptr %end,
 ; COST-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; COST-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
 ; COST-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[START]], i64 [[OFFSET_IDX]]
-; COST-NEXT:    [[TMP6:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 0
-; COST-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP6]], align 1
+; COST-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[NEXT_GEP]], align 1
 ; COST-NEXT:    [[TMP7:%.*]] = icmp ule <4 x i64> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
 ; COST-NEXT:    [[TMP8:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 -12)
 ; COST-NEXT:    [[TMP9:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 13)
@@ -962,9 +952,9 @@ define void @switch_under_br_default_common_dest_with_case(ptr %start, ptr %end,
 ; COST-NEXT:    [[TMP12:%.*]] = or <4 x i1> [[TMP10]], [[TMP11]]
 ; COST-NEXT:    [[TMP13:%.*]] = xor <4 x i1> [[TMP12]], splat (i1 true)
 ; COST-NEXT:    [[TMP14:%.*]] = select <4 x i1> [[TMP7]], <4 x i1> [[TMP13]], <4 x i1> zeroinitializer
-; COST-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[TMP6]], i32 1, <4 x i1> [[TMP11]])
-; COST-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP6]], i32 1, <4 x i1> [[TMP10]])
-; COST-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 2), ptr [[TMP6]], i32 1, <4 x i1> [[TMP14]])
+; COST-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP11]])
+; COST-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP10]])
+; COST-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 2), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP14]])
 ; COST-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; COST-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; COST-NEXT:    br i1 [[TMP16]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
@@ -1024,9 +1014,8 @@ define void @switch_under_br_default_common_dest_with_case(ptr %start, ptr %end,
 ; FORCED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; FORCED-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
 ; FORCED-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[START]], i64 [[OFFSET_IDX]]
-; FORCED-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 0
 ; FORCED-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 4
-; FORCED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP7]], align 1
+; FORCED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[NEXT_GEP]], align 1
 ; FORCED-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i64>, ptr [[TMP8]], align 1
 ; FORCED-NEXT:    [[TMP9:%.*]] = icmp ule <4 x i64> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
 ; FORCED-NEXT:    [[TMP10:%.*]] = icmp ule <4 x i64> [[WIDE_LOAD4]], [[BROADCAST_SPLAT]]
@@ -1044,11 +1033,11 @@ define void @switch_under_br_default_common_dest_with_case(ptr %start, ptr %end,
 ; FORCED-NEXT:    [[TMP22:%.*]] = xor <4 x i1> [[TMP20]], splat (i1 true)
 ; FORCED-NEXT:    [[TMP23:%.*]] = select <4 x i1> [[TMP9]], <4 x i1> [[TMP21]], <4 x i1> zeroinitializer
 ; FORCED-NEXT:    [[TMP24:%.*]] = select <4 x i1> [[TMP10]], <4 x i1> [[TMP22]], <4 x i1> zeroinitializer
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[TMP7]], i32 1, <4 x i1> [[TMP25]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP25]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[TMP8]], i32 1, <4 x i1> [[TMP26]])
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP7]], i32 1, <4 x i1> [[TMP15]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP15]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP8]], i32 1, <4 x i1> [[TMP16]])
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 2), ptr [[TMP7]], i32 1, <4 x i1> [[TMP23]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 2), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP23]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 2), ptr [[TMP8]], i32 1, <4 x i1> [[TMP24]])
 ; FORCED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FORCED-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -1178,9 +1167,8 @@ define void @br_under_switch_default_common_dest_with_case(ptr %start, ptr %end,
 ; FORCED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; FORCED-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
 ; FORCED-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[START]], i64 [[OFFSET_IDX]]
-; FORCED-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 0
 ; FORCED-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 4
-; FORCED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP7]], align 1
+; FORCED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[NEXT_GEP]], align 1
 ; FORCED-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i64>, ptr [[TMP8]], align 1
 ; FORCED-NEXT:    [[TMP9:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 -12)
 ; FORCED-NEXT:    [[TMP10:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD4]], splat (i64 -12)
@@ -1198,15 +1186,15 @@ define void @br_under_switch_default_common_dest_with_case(ptr %start, ptr %end,
 ; FORCED-NEXT:    [[TMP28:%.*]] = select <4 x i1> [[TMP10]], <4 x i1> [[TMP20]], <4 x i1> zeroinitializer
 ; FORCED-NEXT:    [[TMP29:%.*]] = or <4 x i1> [[TMP27]], [[TMP25]]
 ; FORCED-NEXT:    [[TMP30:%.*]] = or <4 x i1> [[TMP28]], [[TMP26]]
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[TMP7]], i32 1, <4 x i1> [[TMP29]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP29]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> zeroinitializer, ptr [[TMP8]], i32 1, <4 x i1> [[TMP30]])
 ; FORCED-NEXT:    [[TMP32:%.*]] = select <4 x i1> [[TMP9]], <4 x i1> [[TMP17]], <4 x i1> zeroinitializer
 ; FORCED-NEXT:    [[TMP33:%.*]] = select <4 x i1> [[TMP10]], <4 x i1> [[TMP18]], <4 x i1> zeroinitializer
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP7]], i32 1, <4 x i1> [[TMP32]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP32]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP8]], i32 1, <4 x i1> [[TMP33]])
 ; FORCED-NEXT:    [[TMP36:%.*]] = or <4 x i1> [[TMP32]], [[TMP15]]
 ; FORCED-NEXT:    [[TMP37:%.*]] = or <4 x i1> [[TMP33]], [[TMP16]]
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 2), ptr [[TMP7]], i32 1, <4 x i1> [[TMP36]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 2), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP36]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 2), ptr [[TMP8]], i32 1, <4 x i1> [[TMP37]])
 ; FORCED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FORCED-NEXT:    [[TMP31:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -1331,9 +1319,8 @@ define void @large_number_of_cases(ptr %start, ptr %end) {
 ; FORCED-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; FORCED-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
 ; FORCED-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[START]], i64 [[OFFSET_IDX]]
-; FORCED-NEXT:    [[TMP7:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 0
 ; FORCED-NEXT:    [[TMP8:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 4
-; FORCED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP7]], align 1
+; FORCED-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[NEXT_GEP]], align 1
 ; FORCED-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i64>, ptr [[TMP8]], align 1
 ; FORCED-NEXT:    [[TMP9:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 1)
 ; FORCED-NEXT:    [[TMP10:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD4]], splat (i64 1)
@@ -1369,7 +1356,7 @@ define void @large_number_of_cases(ptr %start, ptr %end) {
 ; FORCED-NEXT:    [[TMP40:%.*]] = or <4 x i1> [[TMP38]], [[TMP24]]
 ; FORCED-NEXT:    [[TMP57:%.*]] = or <4 x i1> [[TMP39]], [[TMP25]]
 ; FORCED-NEXT:    [[TMP58:%.*]] = or <4 x i1> [[TMP40]], [[TMP26]]
-; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP7]], i32 1, <4 x i1> [[TMP57]])
+; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[NEXT_GEP]], i32 1, <4 x i1> [[TMP57]])
 ; FORCED-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> splat (i64 42), ptr [[TMP8]], i32 1, <4 x i1> [[TMP58]])
 ; FORCED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; FORCED-NEXT:    [[TMP59:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/reduction-crash.ll b/llvm/test/Transforms/LoopVectorize/X86/reduction-crash.ll
index 945d500..3a2d7e7 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/reduction-crash.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/reduction-crash.ll
@@ -34,9 +34,8 @@ define void @pr15344(ptr noalias %ar, ptr noalias %ar2, i32 %exit.limit, i1 %con
 ; CHECK-NEXT:    [[TMP2]] = fadd fast <2 x double> [[VEC_PHI]], splat (double 1.000000e+00)
 ; CHECK-NEXT:    [[TMP3]] = fadd fast <2 x double> [[VEC_PHI2]], splat (double 1.000000e+00)
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[AR2]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP4]], i32 0
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP4]], i32 2
-; CHECK-NEXT:    store <2 x float> splat (float 2.000000e+00), ptr [[TMP5]], align 4, !alias.scope [[META0:![0-9]+]], !noalias [[META3:![0-9]+]]
+; CHECK-NEXT:    store <2 x float> splat (float 2.000000e+00), ptr [[TMP4]], align 4, !alias.scope [[META0:![0-9]+]], !noalias [[META3:![0-9]+]]
 ; CHECK-NEXT:    store <2 x float> splat (float 2.000000e+00), ptr [[TMP6]], align 4, !alias.scope [[META0]], !noalias [[META3]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/reduction-fastmath.ll b/llvm/test/Transforms/LoopVectorize/X86/reduction-fastmath.ll
index 9388aac..0e83cf3 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/reduction-fastmath.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/reduction-fastmath.ll
@@ -60,9 +60,8 @@ define float @reduction_sum_float_fastmath(i32 %n, ptr %array) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr float, ptr [[ARRAY:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr float, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr float, ptr [[TMP2]], i32 4
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[TMP6]] = fadd fast <4 x float> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP7]] = fadd fast <4 x float> [[VEC_PHI1]], [[WIDE_LOAD2]]
@@ -72,10 +71,10 @@ define float @reduction_sum_float_fastmath(i32 %n, ptr %array) {
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[BIN_RDX:%.*]] = fadd fast <4 x float> [[TMP7]], [[TMP6]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = call fast float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> [[BIN_RDX]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT_LOOPEXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP9]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, [[LOOP_PREHEADER]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IDX:%.*]] = phi i32 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -126,9 +125,8 @@ define float @reduction_sum_float_only_reassoc(i32 %n, ptr %array) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr float, ptr [[ARRAY:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr float, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr float, ptr [[TMP2]], i32 4
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[TMP6]] = fadd reassoc <4 x float> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP7]] = fadd reassoc <4 x float> [[VEC_PHI1]], [[WIDE_LOAD2]]
@@ -138,10 +136,10 @@ define float @reduction_sum_float_only_reassoc(i32 %n, ptr %array) {
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[BIN_RDX:%.*]] = fadd reassoc <4 x float> [[TMP7]], [[TMP6]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = call reassoc float @llvm.vector.reduce.fadd.v4f32(float -0.000000e+00, <4 x float> [[BIN_RDX]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT_LOOPEXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP9]], [[MIDDLE_BLOCK]] ], [ -0.000000e+00, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ -0.000000e+00, [[LOOP_PREHEADER]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IDX:%.*]] = phi i32 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -192,9 +190,8 @@ define float @reduction_sum_float_only_reassoc_and_contract(i32 %n, ptr %array)
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x float> [ splat (float -0.000000e+00), [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr float, ptr [[ARRAY:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr float, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr float, ptr [[TMP2]], i32 4
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[TMP6]] = fadd reassoc contract <4 x float> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP7]] = fadd reassoc contract <4 x float> [[VEC_PHI1]], [[WIDE_LOAD2]]
@@ -204,10 +201,10 @@ define float @reduction_sum_float_only_reassoc_and_contract(i32 %n, ptr %array)
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[BIN_RDX:%.*]] = fadd reassoc contract <4 x float> [[TMP7]], [[TMP6]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = call reassoc contract float @llvm.vector.reduce.fadd.v4f32(float -0.000000e+00, <4 x float> [[BIN_RDX]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT_LOOPEXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP9]], [[MIDDLE_BLOCK]] ], [ -0.000000e+00, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ -0.000000e+00, [[LOOP_PREHEADER]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IDX:%.*]] = phi i32 [ [[IDX_INC:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -265,9 +262,8 @@ define float @PR35538(ptr nocapture readonly %a, i32 %N) #0 {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ splat (float -1.000000e+00), [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x float> [ splat (float -1.000000e+00), [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 4
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = fcmp nnan ninf nsz oge <4 x float> [[WIDE_LOAD]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = fcmp nnan ninf nsz oge <4 x float> [[WIDE_LOAD2]], [[VEC_PHI1]]
@@ -347,9 +343,8 @@ define float @PR35538_more_FMF(ptr nocapture readonly %a, i32 %N) #0 {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ splat (float -1.000000e+00), [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x float> [ splat (float -1.000000e+00), [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 4
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = fcmp nnan ninf oge <4 x float> [[WIDE_LOAD]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = fcmp nnan ninf oge <4 x float> [[WIDE_LOAD2]], [[VEC_PHI1]]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/replicate-recipe-with-only-first-lane-used.ll b/llvm/test/Transforms/LoopVectorize/X86/replicate-recipe-with-only-first-lane-used.ll
index e588335..1c6a225 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/replicate-recipe-with-only-first-lane-used.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/replicate-recipe-with-only-first-lane-used.ll
@@ -20,7 +20,7 @@ define void @replicate_udiv_with_only_first_lane_used(i32 %x, ptr %dst, i64 %d)
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 96
 ; CHECK-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 96, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
@@ -36,7 +36,7 @@ define void @replicate_udiv_with_only_first_lane_used(i32 %x, ptr %dst, i64 %d)
 ; CHECK-NEXT:    store i16 0, ptr [[GEP]], align 2
 ; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp ult i64 [[IV_NEXT]], 101
-; CHECK-NEXT:    br i1 [[CMP]], label %[[LOOP_HEADER]], label %[[EXIT]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP]], label %[[LOOP_HEADER]], label %[[EXIT:.*]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    ret void
 ;
@@ -172,7 +172,7 @@ define void @replicate_udiv_with_only_first_lane_used2(i32 %x, ptr %dst, i64 %d)
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 96
 ; CHECK-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 96, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
@@ -188,7 +188,7 @@ define void @replicate_udiv_with_only_first_lane_used2(i32 %x, ptr %dst, i64 %d)
 ; CHECK-NEXT:    store i16 0, ptr [[GEP]], align 2
 ; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp ult i64 [[IV_NEXT]], 101
-; CHECK-NEXT:    br i1 [[CMP]], label %[[LOOP_HEADER]], label %[[EXIT]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP]], label %[[LOOP_HEADER]], label %[[EXIT:.*]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/X86/replicate-uniform-call.ll b/llvm/test/Transforms/LoopVectorize/X86/replicate-uniform-call.ll
index ed1c67c..0b61f20 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/replicate-uniform-call.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/replicate-uniform-call.ll
@@ -53,9 +53,9 @@ define void @smax_call_uniform(ptr %dst, i64 %x) {
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT1:%.*]], %[[LOOP_LATCH:.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/scev-checks-unprofitable.ll b/llvm/test/Transforms/LoopVectorize/X86/scev-checks-unprofitable.ll
index 1bcaa21..440f6e1 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/scev-checks-unprofitable.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/scev-checks-unprofitable.ll
@@ -32,8 +32,7 @@ define void @value_defined_in_loop1_used_for_trip_counts(i32 %start, i1 %c, ptr
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP0:%.*]] = icmp ule <16 x i64> <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11, i64 12, i64 13, i64 14, i64 15>, [[BROADCAST_SPLAT]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[DST]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v16i8.p0(<16 x i8> zeroinitializer, ptr [[TMP2]], i32 1, <16 x i1> [[TMP0]])
+; CHECK-NEXT:    call void @llvm.masked.store.v16i8.p0(<16 x i8> zeroinitializer, ptr [[DST]], i32 1, <16 x i1> [[TMP0]])
 ; CHECK-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    br label %[[EXIT_1_LOOPEXIT1:.*]]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/small-size.ll b/llvm/test/Transforms/LoopVectorize/X86/small-size.ll
index 5917b30..c67b02b 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/small-size.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/small-size.ll
@@ -39,7 +39,7 @@ define void @example1() optsize {
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[TMP7:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[TMP7:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[TMP6:%.*]]
 ; CHECK:       6:
@@ -418,7 +418,7 @@ define void @example23b(ptr noalias nocapture %src, ptr noalias nocapture %dst)
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[TMP5:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[TMP5:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[TMP4:%.*]]
 ; CHECK:       4:
diff --git a/llvm/test/Transforms/LoopVectorize/X86/strided_load_cost.ll b/llvm/test/Transforms/LoopVectorize/X86/strided_load_cost.ll
index ad8f1fb..33b173d 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/strided_load_cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/strided_load_cost.ll
@@ -58,11 +58,10 @@ define i32 @matrix_row_col(ptr nocapture readonly %data, i32 %i, i32 %j) local_u
 ; CHECK-NEXT:    [[TMP30:%.*]] = add i64 [[INDEX]], 30
 ; CHECK-NEXT:    [[TMP31:%.*]] = add i64 [[INDEX]], 31
 ; CHECK-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [100 x i32], ptr [[DATA:%.*]], i64 [[IDXPROM]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP36:%.*]] = getelementptr inbounds i32, ptr [[TMP32]], i32 0
 ; CHECK-NEXT:    [[TMP37:%.*]] = getelementptr inbounds i32, ptr [[TMP32]], i32 8
 ; CHECK-NEXT:    [[TMP38:%.*]] = getelementptr inbounds i32, ptr [[TMP32]], i32 16
 ; CHECK-NEXT:    [[TMP39:%.*]] = getelementptr inbounds i32, ptr [[TMP32]], i32 24
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP36]], align 4, !tbaa [[TBAA1:![0-9]+]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP32]], align 4, !tbaa [[TBAA1:![0-9]+]]
 ; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <8 x i32>, ptr [[TMP37]], align 4, !tbaa [[TBAA1]]
 ; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <8 x i32>, ptr [[TMP38]], align 4, !tbaa [[TBAA1]]
 ; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <8 x i32>, ptr [[TMP39]], align 4, !tbaa [[TBAA1]]
@@ -198,8 +197,7 @@ define i32 @matrix_row_col(ptr nocapture readonly %data, i32 %i, i32 %j) local_u
 ; CHECK-NEXT:    [[TMP174:%.*]] = add i64 [[INDEX9]], 2
 ; CHECK-NEXT:    [[TMP175:%.*]] = add i64 [[INDEX9]], 3
 ; CHECK-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [100 x i32], ptr [[DATA]], i64 [[IDXPROM]], i64 [[TMP172]]
-; CHECK-NEXT:    [[TMP153:%.*]] = getelementptr inbounds i32, ptr [[TMP152]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x i32>, ptr [[TMP153]], align 4, !tbaa [[TBAA1]]
+; CHECK-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x i32>, ptr [[TMP152]], align 4, !tbaa [[TBAA1]]
 ; CHECK-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [100 x i32], ptr [[DATA]], i64 [[TMP172]], i64 [[IDXPROM5]]
 ; CHECK-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [100 x i32], ptr [[DATA]], i64 [[TMP173]], i64 [[IDXPROM5]]
 ; CHECK-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [100 x i32], ptr [[DATA]], i64 [[TMP174]], i64 [[IDXPROM5]]
@@ -290,11 +288,10 @@ define i32 @matrix_row_col(ptr nocapture readonly %data, i32 %i, i32 %j) local_u
 ; MAX-BW-NEXT:    [[TMP30:%.*]] = add i64 [[INDEX]], 30
 ; MAX-BW-NEXT:    [[TMP31:%.*]] = add i64 [[INDEX]], 31
 ; MAX-BW-NEXT:    [[TMP32:%.*]] = getelementptr inbounds [100 x i32], ptr [[DATA:%.*]], i64 [[IDXPROM]], i64 [[TMP0]]
-; MAX-BW-NEXT:    [[TMP36:%.*]] = getelementptr inbounds i32, ptr [[TMP32]], i32 0
 ; MAX-BW-NEXT:    [[TMP37:%.*]] = getelementptr inbounds i32, ptr [[TMP32]], i32 8
 ; MAX-BW-NEXT:    [[TMP38:%.*]] = getelementptr inbounds i32, ptr [[TMP32]], i32 16
 ; MAX-BW-NEXT:    [[TMP39:%.*]] = getelementptr inbounds i32, ptr [[TMP32]], i32 24
-; MAX-BW-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP36]], align 4, !tbaa [[TBAA1:![0-9]+]]
+; MAX-BW-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP32]], align 4, !tbaa [[TBAA1:![0-9]+]]
 ; MAX-BW-NEXT:    [[WIDE_LOAD4:%.*]] = load <8 x i32>, ptr [[TMP37]], align 4, !tbaa [[TBAA1]]
 ; MAX-BW-NEXT:    [[WIDE_LOAD5:%.*]] = load <8 x i32>, ptr [[TMP38]], align 4, !tbaa [[TBAA1]]
 ; MAX-BW-NEXT:    [[WIDE_LOAD6:%.*]] = load <8 x i32>, ptr [[TMP39]], align 4, !tbaa [[TBAA1]]
@@ -430,8 +427,7 @@ define i32 @matrix_row_col(ptr nocapture readonly %data, i32 %i, i32 %j) local_u
 ; MAX-BW-NEXT:    [[TMP174:%.*]] = add i64 [[INDEX9]], 2
 ; MAX-BW-NEXT:    [[TMP175:%.*]] = add i64 [[INDEX9]], 3
 ; MAX-BW-NEXT:    [[TMP152:%.*]] = getelementptr inbounds [100 x i32], ptr [[DATA]], i64 [[IDXPROM]], i64 [[TMP172]]
-; MAX-BW-NEXT:    [[TMP153:%.*]] = getelementptr inbounds i32, ptr [[TMP152]], i32 0
-; MAX-BW-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x i32>, ptr [[TMP153]], align 4, !tbaa [[TBAA1]]
+; MAX-BW-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x i32>, ptr [[TMP152]], align 4, !tbaa [[TBAA1]]
 ; MAX-BW-NEXT:    [[TMP154:%.*]] = getelementptr inbounds [100 x i32], ptr [[DATA]], i64 [[TMP172]], i64 [[IDXPROM5]]
 ; MAX-BW-NEXT:    [[TMP155:%.*]] = getelementptr inbounds [100 x i32], ptr [[DATA]], i64 [[TMP173]], i64 [[IDXPROM5]]
 ; MAX-BW-NEXT:    [[TMP156:%.*]] = getelementptr inbounds [100 x i32], ptr [[DATA]], i64 [[TMP174]], i64 [[IDXPROM5]]
@@ -558,9 +554,9 @@ define void @test(ptr %A, ptr noalias %B) #0 {
 ; CHECK-NEXT:    [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT]], 512
 ; CHECK-NEXT:    br i1 [[TMP36]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
@@ -678,9 +674,9 @@ define void @test(ptr %A, ptr noalias %B) #0 {
 ; MAX-BW-NEXT:    [[TMP68:%.*]] = icmp eq i64 [[INDEX_NEXT]], 512
 ; MAX-BW-NEXT:    br i1 [[TMP68]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; MAX-BW:       middle.block:
-; MAX-BW-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; MAX-BW-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; MAX-BW:       scalar.ph:
-; MAX-BW-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; MAX-BW-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; MAX-BW-NEXT:    br label [[FOR_BODY:%.*]]
 ; MAX-BW:       for.body:
 ; MAX-BW-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_BODY]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/tail_loop_folding.ll b/llvm/test/Transforms/LoopVectorize/X86/tail_loop_folding.ll
index 5fc9e64..5e35c4a 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/tail_loop_folding.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/tail_loop_folding.ll
@@ -18,15 +18,12 @@ define dso_local void @tail_folding_enabled(ptr noalias nocapture %A, ptr noalia
 ; CHECK-NEXT:    [[VEC_IV:%.*]] = add <8 x i64> [[BROADCAST_SPLAT]], <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp ule <8 x i64> [[VEC_IV]], splat (i64 429)
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP3]], i32 4, <8 x i1> [[TMP1]], <8 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP2]], i32 4, <8 x i1> [[TMP1]], <8 x i32> poison)
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP5]], i32 4, <8 x i1> [[TMP1]], <8 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP4]], i32 4, <8 x i1> [[TMP1]], <8 x i32> poison)
 ; CHECK-NEXT:    [[TMP6:%.*]] = add nsw <8 x i32> [[WIDE_MASKED_LOAD1]], [[WIDE_MASKED_LOAD]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP6]], ptr [[TMP8]], i32 4, <8 x i1> [[TMP1]])
+; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP6]], ptr [[TMP7]], i32 4, <8 x i1> [[TMP1]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 432
 ; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -85,15 +82,12 @@ define dso_local void @tail_folding_disabled(ptr noalias nocapture %A, ptr noali
 ; CHECK-NEXT:    [[VEC_IV:%.*]] = add <8 x i64> [[BROADCAST_SPLAT]], <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp ule <8 x i64> [[VEC_IV]], splat (i64 429)
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP3]], i32 4, <8 x i1> [[TMP1]], <8 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP2]], i32 4, <8 x i1> [[TMP1]], <8 x i32> poison)
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP5]], i32 4, <8 x i1> [[TMP1]], <8 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP4]], i32 4, <8 x i1> [[TMP1]], <8 x i32> poison)
 ; CHECK-NEXT:    [[TMP6:%.*]] = add nsw <8 x i32> [[WIDE_MASKED_LOAD1]], [[WIDE_MASKED_LOAD]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP6]], ptr [[TMP8]], i32 4, <8 x i1> [[TMP1]])
+; CHECK-NEXT:    call void @llvm.masked.store.v8i32.p0(<8 x i32> [[TMP6]], ptr [[TMP7]], i32 4, <8 x i1> [[TMP1]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 432
 ; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -170,11 +164,9 @@ define i32 @reduction_i32(ptr nocapture readonly %A, ptr nocapture readonly %B,
 ; CHECK-NEXT:    [[VEC_IV:%.*]] = add <8 x i64> [[BROADCAST_SPLAT2]], <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp ule <8 x i64> [[VEC_IV]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP6]], i32 4, <8 x i1> [[TMP4]], <8 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP5]], i32 4, <8 x i1> [[TMP4]], <8 x i32> poison)
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD3:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP8]], i32 4, <8 x i1> [[TMP4]], <8 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD3:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP7]], i32 4, <8 x i1> [[TMP4]], <8 x i32> poison)
 ; CHECK-NEXT:    [[TMP9:%.*]] = add nsw <8 x i32> [[WIDE_MASKED_LOAD3]], [[WIDE_MASKED_LOAD]]
 ; CHECK-NEXT:    [[TMP10]] = add <8 x i32> [[TMP9]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = select <8 x i1> [[TMP4]], <8 x i32> [[TMP10]], <8 x i32> [[VEC_PHI]]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/transform-narrow-interleave-to-widen-memory.ll b/llvm/test/Transforms/LoopVectorize/X86/transform-narrow-interleave-to-widen-memory.ll
index cb7f0bf..2a3ce03 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/transform-narrow-interleave-to-widen-memory.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/transform-narrow-interleave-to-widen-memory.ll
@@ -104,8 +104,7 @@ define void @test_2xi64(ptr noalias %data, ptr noalias %factor, i64 noundef %n)
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[FACTOR]], i64 [[IV]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[ARRAYIDX]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[ARRAYIDX]], align 8
 ; CHECK-NEXT:    [[TMP3:%.*]] = shl nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP3]]
 ; CHECK-NEXT:    [[WIDE_VEC:%.*]] = load <8 x i64>, ptr [[TMP4]], align 8
@@ -183,8 +182,7 @@ define void @test_2xi64_interleave_loads_order_flipped(ptr noalias %data, ptr no
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[FACTOR]], i64 [[IV]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[ARRAYIDX]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[ARRAYIDX]], align 8
 ; CHECK-NEXT:    [[TMP3:%.*]] = shl nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP3]]
 ; CHECK-NEXT:    [[WIDE_VEC:%.*]] = load <8 x i64>, ptr [[TMP4]], align 8
@@ -262,8 +260,7 @@ define void @test_2xi64_store_order_flipped_1(ptr noalias %data, ptr noalias %fa
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[FACTOR]], i64 [[IV]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[ARRAYIDX]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[ARRAYIDX]], align 8
 ; CHECK-NEXT:    [[TMP3:%.*]] = shl nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP3]]
 ; CHECK-NEXT:    [[WIDE_VEC:%.*]] = load <8 x i64>, ptr [[TMP4]], align 8
@@ -341,8 +338,7 @@ define void @test_2xi64_store_order_flipped_2(ptr noalias %data, ptr noalias %fa
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[FACTOR]], i64 [[IV]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[ARRAYIDX]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[ARRAYIDX]], align 8
 ; CHECK-NEXT:    [[TMP3:%.*]] = shl nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP3]]
 ; CHECK-NEXT:    [[WIDE_VEC:%.*]] = load <8 x i64>, ptr [[TMP4]], align 8
@@ -422,19 +418,17 @@ define void @test_2xi64_different_loads_feeding_fmul(ptr noalias %data, ptr noal
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[GEP_SRC_0:%.*]] = getelementptr inbounds i64, ptr [[SRC_0]], i64 [[IV]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[GEP_SRC_0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[GEP_SRC_0]], align 8
 ; CHECK-NEXT:    [[TMP5:%.*]] = shl nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP5]]
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[GEP_SRC_0]], align 8
 ; CHECK-NEXT:    [[TMP6:%.*]] = mul <4 x i64> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = or disjoint i64 [[TMP5]], 1
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[DATA]], i64 [[TMP7]]
 ; CHECK-NEXT:    [[WIDE_VEC:%.*]] = load <8 x i64>, ptr [[TMP8]], align 8
 ; CHECK-NEXT:    [[STRIDED_VEC:%.*]] = shufflevector <8 x i64> [[WIDE_VEC]], <8 x i64> poison, <4 x i32> <i32 0, i32 2, i32 4, i32 6>
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[SRC_1]], i64 [[IV]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i64>, ptr [[TMP11]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i64>, ptr [[TMP10]], align 8
 ; CHECK-NEXT:    [[TMP12:%.*]] = mul <4 x i64> [[WIDE_LOAD2]], [[STRIDED_VEC]]
 ; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <4 x i64> [[TMP6]], <4 x i64> [[TMP12]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
 ; CHECK-NEXT:    [[INTERLEAVED_VEC:%.*]] = shufflevector <8 x i64> [[TMP14]], <8 x i64> poison, <8 x i32> <i32 0, i32 4, i32 1, i32 5, i32 2, i32 6, i32 3, i32 7>
@@ -509,8 +503,7 @@ define void @test_3xi64(ptr noalias %data, ptr noalias %factor, i64 noundef %n)
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i64, ptr [[FACTOR]], i64 [[IV]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[ARRAYIDX]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[ARRAYIDX]], align 8
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds { i64, i64, i64 }, ptr [[DATA]], i64 [[IV]], i32 0
 ; CHECK-NEXT:    [[WIDE_VEC:%.*]] = load <12 x i64>, ptr [[TMP3]], align 8
 ; CHECK-NEXT:    [[STRIDED_VEC:%.*]] = shufflevector <12 x i64> [[WIDE_VEC]], <12 x i64> poison, <4 x i32> <i32 0, i32 3, i32 6, i32 9>
diff --git a/llvm/test/Transforms/LoopVectorize/X86/uniform_load.ll b/llvm/test/Transforms/LoopVectorize/X86/uniform_load.ll
index 074ea6b..62d08c8 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/uniform_load.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/uniform_load.ll
@@ -25,11 +25,10 @@ define void @foo(ptr nocapture noalias %A, i64 %N) #0 {
 ; CHECK-NEXT:    [[TMP1:%.*]] = load float, ptr @inc, align 4
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <8 x float> poison, float [[TMP1]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x float> [[BROADCAST_SPLATINSERT]], <8 x float> poison, <8 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[A]], i32 0
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[A]], i32 8
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[A]], i32 16
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A]], i32 24
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x float>, ptr [[TMP3]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x float>, ptr [[A]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <8 x float>, ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <8 x float>, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <8 x float>, ptr [[TMP6]], align 4
@@ -37,11 +36,10 @@ define void @foo(ptr nocapture noalias %A, i64 %N) #0 {
 ; CHECK-NEXT:    [[TMP8:%.*]] = fadd <8 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = fadd <8 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD3]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = fadd <8 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD4]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[A]], i32 0
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A]], i32 8
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[A]], i32 16
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[A]], i32 24
-; CHECK-NEXT:    store <8 x float> [[TMP7]], ptr [[TMP14]], align 4
+; CHECK-NEXT:    store <8 x float> [[TMP7]], ptr [[A]], align 4
 ; CHECK-NEXT:    store <8 x float> [[TMP8]], ptr [[TMP11]], align 4
 ; CHECK-NEXT:    store <8 x float> [[TMP9]], ptr [[TMP12]], align 4
 ; CHECK-NEXT:    store <8 x float> [[TMP10]], ptr [[TMP13]], align 4
diff --git a/llvm/test/Transforms/LoopVectorize/X86/uniform_mem_op.ll b/llvm/test/Transforms/LoopVectorize/X86/uniform_mem_op.ll
index 8c91c31..68b4f20 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/uniform_mem_op.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/uniform_mem_op.ll
@@ -16,23 +16,22 @@ define i32 @uniform_load(ptr align(4) %addr) {
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[ADDR:%.*]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; CHECK-NEXT:    br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
-; CHECK-NEXT:    [[LOAD:%.*]] = load i32, ptr [[ADDR]], align 4
+; CHECK-NEXT:    [[LOAD:%.*]] = load i32, ptr [[ADDR:%.*]], align 4
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV]], 4096
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       loopexit:
-; CHECK-NEXT:    [[LOAD_LCSSA:%.*]] = phi i32 [ [[LOAD]], [[FOR_BODY]] ], [ [[TMP0]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[LOAD_LCSSA:%.*]] = phi i32 [ [[LOAD]], [[FOR_BODY]] ]
 ; CHECK-NEXT:    ret i32 [[LOAD_LCSSA]]
 ;
 entry:
@@ -76,7 +75,7 @@ define i32 @uniform_load2(ptr align(4) %addr) {
 ; CHECK-NEXT:    [[BIN_RDX10:%.*]] = add <4 x i32> [[TMP3]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX11:%.*]] = add <4 x i32> [[TMP4]], [[BIN_RDX10]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX11]])
-; CHECK-NEXT:    br i1 false, label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP6]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
@@ -88,9 +87,9 @@ define i32 @uniform_load2(ptr align(4) %addr) {
 ; CHECK-NEXT:    [[ACCUM_NEXT]] = add i32 [[ACCUM]], [[LOAD]]
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV]], 4096
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       loopexit:
-; CHECK-NEXT:    [[ACCUM_NEXT_LCSSA:%.*]] = phi i32 [ [[ACCUM_NEXT]], [[FOR_BODY]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[ACCUM_NEXT_LCSSA:%.*]] = phi i32 [ [[ACCUM_NEXT]], [[FOR_BODY]] ]
 ; CHECK-NEXT:    ret i32 [[ACCUM_NEXT_LCSSA]]
 ;
 entry:
@@ -114,30 +113,27 @@ define i32 @uniform_address(ptr align(4) %addr, i32 %byte_offset) {
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK-NEXT:    [[TMP0:%.*]] = udiv i32 [[BYTE_OFFSET:%.*]], 4
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[ADDR:%.*]], i32 [[TMP0]]
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP8:%.*]] = load i32, ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
-; CHECK-NEXT:    [[OFFSET:%.*]] = udiv i32 [[BYTE_OFFSET]], 4
-; CHECK-NEXT:    [[GEP:%.*]] = getelementptr i32, ptr [[ADDR]], i32 [[OFFSET]]
+; CHECK-NEXT:    [[OFFSET:%.*]] = udiv i32 [[BYTE_OFFSET:%.*]], 4
+; CHECK-NEXT:    [[GEP:%.*]] = getelementptr i32, ptr [[ADDR:%.*]], i32 [[OFFSET]]
 ; CHECK-NEXT:    [[LOAD:%.*]] = load i32, ptr [[GEP]], align 4
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV]], 4096
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP7:![0-9]+]]
 ; CHECK:       loopexit:
-; CHECK-NEXT:    [[LOAD_LCSSA:%.*]] = phi i32 [ [[LOAD]], [[FOR_BODY]] ], [ [[TMP8]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[LOAD_LCSSA:%.*]] = phi i32 [ [[LOAD]], [[FOR_BODY]] ]
 ; CHECK-NEXT:    ret i32 [[LOAD_LCSSA]]
 ;
 entry:
@@ -171,7 +167,7 @@ define void @uniform_store_uniform_value(ptr align(4) %addr) {
 ; CHECK-NEXT:    [[TMP0:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; CHECK-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
@@ -180,7 +176,7 @@ define void @uniform_store_uniform_value(ptr align(4) %addr) {
 ; CHECK-NEXT:    store i32 0, ptr [[ADDR]], align 4
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV]], 4096
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
 ; CHECK:       loopexit:
 ; CHECK-NEXT:    ret void
 ;
@@ -216,7 +212,7 @@ define void @uniform_store_varying_value(ptr align(4) %addr) {
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
@@ -226,7 +222,7 @@ define void @uniform_store_varying_value(ptr align(4) %addr) {
 ; CHECK-NEXT:    store i32 [[IV_I32]], ptr [[ADDR]], align 4
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV]], 4096
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ; CHECK:       loopexit:
 ; CHECK-NEXT:    ret void
 ;
@@ -297,7 +293,7 @@ define void @uniform_copy(ptr %A, ptr %B) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4096
 ; CHECK-NEXT:    br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
@@ -307,7 +303,7 @@ define void @uniform_copy(ptr %A, ptr %B) {
 ; CHECK-NEXT:    store i32 [[LOAD]], ptr [[B]], align 4
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV]], 4096
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; CHECK:       loopexit:
 ; CHECK-NEXT:    ret void
 ;
@@ -408,10 +404,10 @@ define i32 @test_count_bits(ptr %test_base) {
 ; CHECK-NEXT:    [[BIN_RDX13:%.*]] = add <4 x i32> [[TMP38]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX14:%.*]] = add <4 x i32> [[TMP39]], [[BIN_RDX13]]
 ; CHECK-NEXT:    [[TMP41:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX14]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP41]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -486,7 +482,7 @@ define i32 @uniform_load_global() {
 ; CHECK-NEXT:    [[BIN_RDX10:%.*]] = add <4 x i32> [[TMP3]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX11:%.*]] = add <4 x i32> [[TMP4]], [[BIN_RDX10]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX11]])
-; CHECK-NEXT:    br i1 false, label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP6]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
@@ -498,9 +494,9 @@ define i32 @uniform_load_global() {
 ; CHECK-NEXT:    [[ACCUM_NEXT]] = add i32 [[ACCUM]], [[LOAD]]
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV]], 4096
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
 ; CHECK:       loopexit:
-; CHECK-NEXT:    [[ACCUM_NEXT_LCSSA:%.*]] = phi i32 [ [[ACCUM_NEXT]], [[FOR_BODY]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[ACCUM_NEXT_LCSSA:%.*]] = phi i32 [ [[ACCUM_NEXT]], [[FOR_BODY]] ]
 ; CHECK-NEXT:    ret i32 [[ACCUM_NEXT_LCSSA]]
 ;
 entry:
@@ -547,7 +543,7 @@ define i32 @uniform_load_constexpr() {
 ; CHECK-NEXT:    [[BIN_RDX10:%.*]] = add <4 x i32> [[TMP3]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX11:%.*]] = add <4 x i32> [[TMP4]], [[BIN_RDX10]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX11]])
-; CHECK-NEXT:    br i1 false, label [[LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP6]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
@@ -559,9 +555,9 @@ define i32 @uniform_load_constexpr() {
 ; CHECK-NEXT:    [[ACCUM_NEXT]] = add i32 [[ACCUM]], [[LOAD]]
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV]], 4096
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOPEXIT:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
 ; CHECK:       loopexit:
-; CHECK-NEXT:    [[ACCUM_NEXT_LCSSA:%.*]] = phi i32 [ [[ACCUM_NEXT]], [[FOR_BODY]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[ACCUM_NEXT_LCSSA:%.*]] = phi i32 [ [[ACCUM_NEXT]], [[FOR_BODY]] ]
 ; CHECK-NEXT:    ret i32 [[ACCUM_NEXT_LCSSA]]
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/X86/vect.omp.force.small-tc.ll b/llvm/test/Transforms/LoopVectorize/X86/vect.omp.force.small-tc.ll
index d7a3b32..f7eba42 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/vect.omp.force.small-tc.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/vect.omp.force.small-tc.ll
@@ -29,20 +29,18 @@ define void @vectorized(ptr noalias nocapture %A, ptr noalias nocapture readonly
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 8
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 12
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4, !llvm.access.group [[ACC_GRP0:![0-9]+]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4, !llvm.access.group [[ACC_GRP0:![0-9]+]]
 ; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x float>, ptr [[TMP3]], align 4, !llvm.access.group [[ACC_GRP0]]
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP4]], align 4, !llvm.access.group [[ACC_GRP0]]
 ; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x float>, ptr [[TMP5]], align 4, !llvm.access.group [[ACC_GRP0]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 0
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 4
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 8
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 12
-; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x float>, ptr [[TMP7]], align 4, !llvm.access.group [[ACC_GRP0]]
+; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x float>, ptr [[TMP6]], align 4, !llvm.access.group [[ACC_GRP0]]
 ; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x float>, ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP0]]
 ; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP9]], align 4, !llvm.access.group [[ACC_GRP0]]
 ; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x float>, ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP0]]
@@ -50,7 +48,7 @@ define void @vectorized(ptr noalias nocapture %A, ptr noalias nocapture readonly
 ; CHECK-NEXT:    [[TMP12:%.*]] = fadd fast <4 x float> [[WIDE_LOAD1]], [[WIDE_LOAD5]]
 ; CHECK-NEXT:    [[TMP13:%.*]] = fadd fast <4 x float> [[WIDE_LOAD2]], [[WIDE_LOAD6]]
 ; CHECK-NEXT:    [[TMP14:%.*]] = fadd fast <4 x float> [[WIDE_LOAD3]], [[WIDE_LOAD7]]
-; CHECK-NEXT:    store <4 x float> [[TMP11]], ptr [[TMP7]], align 4, !llvm.access.group [[ACC_GRP0]]
+; CHECK-NEXT:    store <4 x float> [[TMP11]], ptr [[TMP6]], align 4, !llvm.access.group [[ACC_GRP0]]
 ; CHECK-NEXT:    store <4 x float> [[TMP12]], ptr [[TMP8]], align 4, !llvm.access.group [[ACC_GRP0]]
 ; CHECK-NEXT:    store <4 x float> [[TMP13]], ptr [[TMP9]], align 4, !llvm.access.group [[ACC_GRP0]]
 ; CHECK-NEXT:    store <4 x float> [[TMP14]], ptr [[TMP10]], align 4, !llvm.access.group [[ACC_GRP0]]
@@ -67,13 +65,11 @@ define void @vectorized(ptr noalias nocapture %A, ptr noalias nocapture readonly
 ; CHECK:       vec.epilog.vector.body:
 ; CHECK-NEXT:    [[INDEX8:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT11:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX8]]
-; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x float>, ptr [[TMP18]], align 4, !llvm.access.group [[ACC_GRP0]]
+; CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x float>, ptr [[TMP17]], align 4, !llvm.access.group [[ACC_GRP0]]
 ; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX8]]
-; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr inbounds float, ptr [[TMP19]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD10:%.*]] = load <4 x float>, ptr [[TMP20]], align 4, !llvm.access.group [[ACC_GRP0]]
+; CHECK-NEXT:    [[WIDE_LOAD10:%.*]] = load <4 x float>, ptr [[TMP19]], align 4, !llvm.access.group [[ACC_GRP0]]
 ; CHECK-NEXT:    [[TMP21:%.*]] = fadd fast <4 x float> [[WIDE_LOAD9]], [[WIDE_LOAD10]]
-; CHECK-NEXT:    store <4 x float> [[TMP21]], ptr [[TMP20]], align 4, !llvm.access.group [[ACC_GRP0]]
+; CHECK-NEXT:    store <4 x float> [[TMP21]], ptr [[TMP19]], align 4, !llvm.access.group [[ACC_GRP0]]
 ; CHECK-NEXT:    [[INDEX_NEXT11]] = add nuw i64 [[INDEX8]], 4
 ; CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT11]], 20
 ; CHECK-NEXT:    br i1 [[TMP22]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
@@ -136,13 +132,11 @@ define void @vectorized1(ptr noalias nocapture %A, ptr noalias nocapture readonl
 ; CHECK-NEXT:    [[VEC_IV:%.*]] = add <8 x i64> [[BROADCAST_SPLAT]], <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp ule <8 x i64> [[VEC_IV]], splat (i64 19)
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP3]], i32 4, <8 x i1> [[TMP1]], <8 x float> poison), !llvm.access.group [[ACC_GRP7:![0-9]+]]
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP2]], i32 4, <8 x i1> [[TMP1]], <8 x float> poison), !llvm.access.group [[ACC_GRP7:![0-9]+]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP5]], i32 4, <8 x i1> [[TMP1]], <8 x float> poison), !llvm.access.group [[ACC_GRP7]]
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD1:%.*]] = call <8 x float> @llvm.masked.load.v8f32.p0(ptr [[TMP4]], i32 4, <8 x i1> [[TMP1]], <8 x float> poison), !llvm.access.group [[ACC_GRP7]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = fadd fast <8 x float> [[WIDE_MASKED_LOAD]], [[WIDE_MASKED_LOAD1]]
-; CHECK-NEXT:    call void @llvm.masked.store.v8f32.p0(<8 x float> [[TMP6]], ptr [[TMP5]], i32 4, <8 x i1> [[TMP1]]), !llvm.access.group [[ACC_GRP7]]
+; CHECK-NEXT:    call void @llvm.masked.store.v8f32.p0(<8 x float> [[TMP6]], ptr [[TMP4]], i32 4, <8 x i1> [[TMP1]]), !llvm.access.group [[ACC_GRP7]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 24
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
@@ -200,20 +194,18 @@ define void @vectorized2(ptr noalias nocapture %A, ptr noalias nocapture readonl
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[B:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x float>, ptr [[TMP2]], align 4, !llvm.access.group [[ACC_GRP7]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x float>, ptr [[TMP1]], align 4, !llvm.access.group [[ACC_GRP7]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x float>, ptr [[TMP4]], align 4, !llvm.access.group [[ACC_GRP7]]
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x float>, ptr [[TMP3]], align 4, !llvm.access.group [[ACC_GRP7]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = fadd fast <8 x float> [[WIDE_LOAD]], [[WIDE_LOAD1]]
-; CHECK-NEXT:    store <8 x float> [[TMP5]], ptr [[TMP4]], align 4, !llvm.access.group [[ACC_GRP7]]
+; CHECK-NEXT:    store <8 x float> [[TMP5]], ptr [[TMP3]], align 4, !llvm.access.group [[ACC_GRP7]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 16
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 16, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/vectorize-force-tail-with-evl.ll b/llvm/test/Transforms/LoopVectorize/X86/vectorize-force-tail-with-evl.ll
index 27321e7..59f2925 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/vectorize-force-tail-with-evl.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/vectorize-force-tail-with-evl.ll
@@ -28,15 +28,12 @@ define void @foo(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %N) {
 ; IF-EVL-NEXT:    [[VEC_IV:%.*]] = add <16 x i64> [[BROADCAST_SPLAT]], <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7, i64 8, i64 9, i64 10, i64 11, i64 12, i64 13, i64 14, i64 15>
 ; IF-EVL-NEXT:    [[TMP1:%.*]] = icmp ule <16 x i64> [[VEC_IV]], [[BROADCAST_SPLAT2]]
 ; IF-EVL-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
-; IF-EVL-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
-; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p0(ptr [[TMP3]], i32 4, <16 x i1> [[TMP1]], <16 x i32> poison)
+; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p0(ptr [[TMP2]], i32 4, <16 x i1> [[TMP1]], <16 x i32> poison)
 ; IF-EVL-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[INDEX]]
-; IF-EVL-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
-; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD3:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p0(ptr [[TMP5]], i32 4, <16 x i1> [[TMP1]], <16 x i32> poison)
+; IF-EVL-NEXT:    [[WIDE_MASKED_LOAD3:%.*]] = call <16 x i32> @llvm.masked.load.v16i32.p0(ptr [[TMP4]], i32 4, <16 x i1> [[TMP1]], <16 x i32> poison)
 ; IF-EVL-NEXT:    [[TMP6:%.*]] = add nsw <16 x i32> [[WIDE_MASKED_LOAD3]], [[WIDE_MASKED_LOAD]]
 ; IF-EVL-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; IF-EVL-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; IF-EVL-NEXT:    call void @llvm.masked.store.v16i32.p0(<16 x i32> [[TMP6]], ptr [[TMP8]], i32 4, <16 x i1> [[TMP1]])
+; IF-EVL-NEXT:    call void @llvm.masked.store.v16i32.p0(<16 x i32> [[TMP6]], ptr [[TMP7]], i32 4, <16 x i1> [[TMP1]])
 ; IF-EVL-NEXT:    [[INDEX_NEXT]] = add i64 [[INDEX]], 16
 ; IF-EVL-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; IF-EVL-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -74,20 +71,18 @@ define void @foo(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %N) {
 ; NO-VP:       vector.body:
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
 ; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 16
 ; NO-VP-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 32
 ; NO-VP-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 48
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i32>, ptr [[TMP8]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i32>, ptr [[TMP4]], align 4
 ; NO-VP-NEXT:    [[WIDE_LOAD2:%.*]] = load <16 x i32>, ptr [[TMP9]], align 4
 ; NO-VP-NEXT:    [[WIDE_LOAD3:%.*]] = load <16 x i32>, ptr [[TMP10]], align 4
 ; NO-VP-NEXT:    [[WIDE_LOAD4:%.*]] = load <16 x i32>, ptr [[TMP11]], align 4
 ; NO-VP-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
 ; NO-VP-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 16
 ; NO-VP-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 32
 ; NO-VP-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 48
-; NO-VP-NEXT:    [[WIDE_LOAD5:%.*]] = load <16 x i32>, ptr [[TMP16]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD5:%.*]] = load <16 x i32>, ptr [[TMP12]], align 4
 ; NO-VP-NEXT:    [[WIDE_LOAD6:%.*]] = load <16 x i32>, ptr [[TMP17]], align 4
 ; NO-VP-NEXT:    [[WIDE_LOAD7:%.*]] = load <16 x i32>, ptr [[TMP18]], align 4
 ; NO-VP-NEXT:    [[WIDE_LOAD8:%.*]] = load <16 x i32>, ptr [[TMP19]], align 4
@@ -96,11 +91,10 @@ define void @foo(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %N) {
 ; NO-VP-NEXT:    [[TMP22:%.*]] = add nsw <16 x i32> [[WIDE_LOAD7]], [[WIDE_LOAD3]]
 ; NO-VP-NEXT:    [[TMP23:%.*]] = add nsw <16 x i32> [[WIDE_LOAD8]], [[WIDE_LOAD4]]
 ; NO-VP-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i32, ptr [[TMP24]], i32 0
 ; NO-VP-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i32, ptr [[TMP24]], i32 16
 ; NO-VP-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i32, ptr [[TMP24]], i32 32
 ; NO-VP-NEXT:    [[TMP31:%.*]] = getelementptr inbounds i32, ptr [[TMP24]], i32 48
-; NO-VP-NEXT:    store <16 x i32> [[TMP20]], ptr [[TMP28]], align 4
+; NO-VP-NEXT:    store <16 x i32> [[TMP20]], ptr [[TMP24]], align 4
 ; NO-VP-NEXT:    store <16 x i32> [[TMP21]], ptr [[TMP29]], align 4
 ; NO-VP-NEXT:    store <16 x i32> [[TMP22]], ptr [[TMP30]], align 4
 ; NO-VP-NEXT:    store <16 x i32> [[TMP23]], ptr [[TMP31]], align 4
@@ -122,15 +116,12 @@ define void @foo(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %N) {
 ; NO-VP:       vec.epilog.vector.body:
 ; NO-VP-NEXT:    [[INDEX12:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT15:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP34:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX12]]
-; NO-VP-NEXT:    [[TMP35:%.*]] = getelementptr inbounds i32, ptr [[TMP34]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD13:%.*]] = load <8 x i32>, ptr [[TMP35]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD13:%.*]] = load <8 x i32>, ptr [[TMP34]], align 4
 ; NO-VP-NEXT:    [[TMP36:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[INDEX12]]
-; NO-VP-NEXT:    [[TMP37:%.*]] = getelementptr inbounds i32, ptr [[TMP36]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD14:%.*]] = load <8 x i32>, ptr [[TMP37]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD14:%.*]] = load <8 x i32>, ptr [[TMP36]], align 4
 ; NO-VP-NEXT:    [[TMP38:%.*]] = add nsw <8 x i32> [[WIDE_LOAD14]], [[WIDE_LOAD13]]
 ; NO-VP-NEXT:    [[TMP39:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX12]]
-; NO-VP-NEXT:    [[TMP40:%.*]] = getelementptr inbounds i32, ptr [[TMP39]], i32 0
-; NO-VP-NEXT:    store <8 x i32> [[TMP38]], ptr [[TMP40]], align 4
+; NO-VP-NEXT:    store <8 x i32> [[TMP38]], ptr [[TMP39]], align 4
 ; NO-VP-NEXT:    [[INDEX_NEXT15]] = add nuw i64 [[INDEX12]], 8
 ; NO-VP-NEXT:    [[TMP41:%.*]] = icmp eq i64 [[INDEX_NEXT15]], [[N_VEC10]]
 ; NO-VP-NEXT:    br i1 [[TMP41]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/vectorize-interleaved-accesses-gap.ll b/llvm/test/Transforms/LoopVectorize/X86/vectorize-interleaved-accesses-gap.ll
index 7b1c60b..e9d85c2 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/vectorize-interleaved-accesses-gap.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/vectorize-interleaved-accesses-gap.ll
@@ -71,8 +71,8 @@ define void @test_pr59090(ptr %l_out, ptr noalias %b) #0 {
 ; CHECK-NEXT:    [[TMP15:%.*]] = and <48 x i1> [[INTERLEAVED_MASK]], <i1 true, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 false, i1 true, i1 false, i1 true, i1 false, i1 false, i1 false>
 ; CHECK-NEXT:    call void @llvm.masked.store.v48i8.p0(<48 x i8> <i8 0, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 0, i8 poison, i8 poison, i8 poison, i8 0, i8 poison, i8 0, i8 poison, i8 poison, i8 poison>, ptr [[TMP13]], i32 1, <48 x i1> [[TMP15]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
-; CHECK-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10008
-; CHECK-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP1:![0-9]+]]
+; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10008
+; CHECK-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP1:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
diff --git a/llvm/test/Transforms/LoopVectorize/X86/widened-value-used-as-scalar-and-first-lane.ll b/llvm/test/Transforms/LoopVectorize/X86/widened-value-used-as-scalar-and-first-lane.ll
index 5c34971..05d08a4 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/widened-value-used-as-scalar-and-first-lane.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/widened-value-used-as-scalar-and-first-lane.ll
@@ -18,11 +18,10 @@ define void @iv.4_used_as_vector_and_first_lane(ptr %src, ptr noalias %dst) {
 ; CHECK-NEXT:    [[STEP_ADD_2:%.*]] = add <4 x i64> [[STEP_ADD]], splat (i64 4)
 ; CHECK-NEXT:    [[STEP_ADD_3:%.*]] = add <4 x i64> [[STEP_ADD_2]], splat (i64 4)
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[SRC]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 4
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 8
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 12
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP8]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i64>, ptr [[TMP9]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i64>, ptr [[TMP10]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i64>, ptr [[TMP11]], align 8
@@ -37,11 +36,10 @@ define void @iv.4_used_as_vector_and_first_lane(ptr %src, ptr noalias %dst) {
 ; CHECK-NEXT:    [[TMP26:%.*]] = extractelement <4 x i64> [[TMP12]], i32 0
 ; CHECK-NEXT:    [[TMP27:%.*]] = add i64 [[TMP26]], 1
 ; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP27]]
-; CHECK-NEXT:    [[TMP32:%.*]] = getelementptr i64, ptr [[TMP28]], i32 0
 ; CHECK-NEXT:    [[TMP33:%.*]] = getelementptr i64, ptr [[TMP28]], i32 4
 ; CHECK-NEXT:    [[TMP34:%.*]] = getelementptr i64, ptr [[TMP28]], i32 8
 ; CHECK-NEXT:    [[TMP35:%.*]] = getelementptr i64, ptr [[TMP28]], i32 12
-; CHECK-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> [[TMP12]], ptr [[TMP32]], i32 4, <4 x i1> [[TMP16]])
+; CHECK-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> [[TMP12]], ptr [[TMP28]], i32 4, <4 x i1> [[TMP16]])
 ; CHECK-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> [[TMP13]], ptr [[TMP33]], i32 4, <4 x i1> [[TMP17]])
 ; CHECK-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> [[TMP14]], ptr [[TMP34]], i32 4, <4 x i1> [[TMP18]])
 ; CHECK-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> [[TMP15]], ptr [[TMP35]], i32 4, <4 x i1> [[TMP19]])
@@ -50,9 +48,9 @@ define void @iv.4_used_as_vector_and_first_lane(ptr %src, ptr noalias %dst) {
 ; CHECK-NEXT:    [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32
 ; CHECK-NEXT:    br i1 [[TMP36]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 32, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; CHECK:       loop.header:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
@@ -109,11 +107,10 @@ define void @iv.4_used_as_first_lane(ptr %src, ptr noalias %dst) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[SRC]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 4
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 8
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 12
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP8]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP9]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i64>, ptr [[TMP10]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i64>, ptr [[TMP11]], align 8
@@ -124,11 +121,10 @@ define void @iv.4_used_as_first_lane(ptr %src, ptr noalias %dst) {
 ; CHECK-NEXT:    [[TMP19:%.*]] = icmp ule <4 x i64> [[WIDE_LOAD3]], splat (i64 128)
 ; CHECK-NEXT:    [[TMP23:%.*]] = add i64 [[TMP15]], 1
 ; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP23]]
-; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr i64, ptr [[TMP24]], i32 0
 ; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr i64, ptr [[TMP24]], i32 4
 ; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr i64, ptr [[TMP24]], i32 8
 ; CHECK-NEXT:    [[TMP31:%.*]] = getelementptr i64, ptr [[TMP24]], i32 12
-; CHECK-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> [[WIDE_LOAD]], ptr [[TMP28]], i32 4, <4 x i1> [[TMP16]])
+; CHECK-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> [[WIDE_LOAD]], ptr [[TMP24]], i32 4, <4 x i1> [[TMP16]])
 ; CHECK-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> [[WIDE_LOAD1]], ptr [[TMP29]], i32 4, <4 x i1> [[TMP17]])
 ; CHECK-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> [[WIDE_LOAD2]], ptr [[TMP30]], i32 4, <4 x i1> [[TMP18]])
 ; CHECK-NEXT:    call void @llvm.masked.store.v4i64.p0(<4 x i64> [[WIDE_LOAD3]], ptr [[TMP31]], i32 4, <4 x i1> [[TMP19]])
@@ -136,9 +132,9 @@ define void @iv.4_used_as_first_lane(ptr %src, ptr noalias %dst) {
 ; CHECK-NEXT:    [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32
 ; CHECK-NEXT:    br i1 [[TMP32]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 32, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; CHECK:       loop.header:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/X86/x86-predication.ll b/llvm/test/Transforms/LoopVectorize/X86/x86-predication.ll
index 93880f5..b0ae40c 100644
--- a/llvm/test/Transforms/LoopVectorize/X86/x86-predication.ll
+++ b/llvm/test/Transforms/LoopVectorize/X86/x86-predication.ll
@@ -19,11 +19,9 @@ define i32 @predicated_sdiv_masked_load(ptr %a, ptr %b, i32 %x, i1 %c) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[PRED_SDIV_CONTINUE2:%.*]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[ENTRY]] ], [ [[TMP17:%.*]], [[PRED_SDIV_CONTINUE2]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i32, ptr [[B:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <2 x i32> @llvm.masked.load.v2i32.p0(ptr [[TMP4]], i32 4, <2 x i1> [[BROADCAST_SPLAT]], <2 x i32> poison)
+; CHECK-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <2 x i32> @llvm.masked.load.v2i32.p0(ptr [[TMP3]], i32 4, <2 x i1> [[BROADCAST_SPLAT]], <2 x i32> poison)
 ; CHECK-NEXT:    br i1 [[C]], label [[PRED_SDIV_IF:%.*]], label [[PRED_SDIV_CONTINUE:%.*]]
 ; CHECK:       pred.sdiv.if:
 ; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <2 x i32> [[WIDE_MASKED_LOAD]], i32 0
@@ -61,11 +59,9 @@ define i32 @predicated_sdiv_masked_load(ptr %a, ptr %b, i32 %x, i1 %c) {
 ; SINK-GATHER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_SDIV_CONTINUE14:%.*]] ]
 ; SINK-GATHER-NEXT:    [[VEC_PHI:%.*]] = phi <8 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP47:%.*]], [[PRED_SDIV_CONTINUE14]] ]
 ; SINK-GATHER-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; SINK-GATHER-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; SINK-GATHER-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP2]], align 4
+; SINK-GATHER-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[TMP1]], align 4
 ; SINK-GATHER-NEXT:    [[TMP3:%.*]] = getelementptr i32, ptr [[B:%.*]], i64 [[INDEX]]
-; SINK-GATHER-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[TMP3]], i32 0
-; SINK-GATHER-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP4]], i32 4, <8 x i1> [[BROADCAST_SPLAT]], <8 x i32> poison)
+; SINK-GATHER-NEXT:    [[WIDE_MASKED_LOAD:%.*]] = call <8 x i32> @llvm.masked.load.v8i32.p0(ptr [[TMP3]], i32 4, <8 x i1> [[BROADCAST_SPLAT]], <8 x i32> poison)
 ; SINK-GATHER-NEXT:    br i1 [[C]], label [[PRED_SDIV_IF:%.*]], label [[PRED_SDIV_CONTINUE:%.*]]
 ; SINK-GATHER:       pred.sdiv.if:
 ; SINK-GATHER-NEXT:    [[TMP6:%.*]] = extractelement <8 x i32> [[WIDE_MASKED_LOAD]], i32 0
@@ -138,10 +134,10 @@ define i32 @predicated_sdiv_masked_load(ptr %a, ptr %b, i32 %x, i1 %c) {
 ; SINK-GATHER-NEXT:    br i1 [[TMP48]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; SINK-GATHER:       middle.block:
 ; SINK-GATHER-NEXT:    [[TMP49:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[TMP47]])
-; SINK-GATHER-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; SINK-GATHER-NEXT:    br label [[FOR_END:%.*]]
 ; SINK-GATHER:       scalar.ph:
-; SINK-GATHER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 10000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; SINK-GATHER-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP49]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; SINK-GATHER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; SINK-GATHER-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; SINK-GATHER-NEXT:    br label [[FOR_BODY:%.*]]
 ; SINK-GATHER:       for.body:
 ; SINK-GATHER-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_INC:%.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/blend-in-header.ll b/llvm/test/Transforms/LoopVectorize/blend-in-header.ll
index 33edaf2..4f33b5b 100644
--- a/llvm/test/Transforms/LoopVectorize/blend-in-header.ll
+++ b/llvm/test/Transforms/LoopVectorize/blend-in-header.ll
@@ -17,7 +17,7 @@ define i64 @pr88297() {
 ; CHECK-NEXT:    [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
@@ -30,9 +30,9 @@ define i64 @pr88297() {
 ; CHECK-NEXT:    [[R:%.*]] = phi i64 [ 1, [[THEN]] ], [ 0, [[LOOP_HEADER]] ]
 ; CHECK-NEXT:    [[IV_NEXT]] = add i32 [[IV]], 1
 ; CHECK-NEXT:    [[ICMP:%.*]] = icmp sgt i32 [[IV]], 1000
-; CHECK-NEXT:    br i1 [[ICMP]], label [[EXIT]], label [[LOOP_HEADER]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[ICMP]], label [[EXIT:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       exit:
-; CHECK-NEXT:    [[R_LCSSA:%.*]] = phi i64 [ [[R]], [[LOOP_LATCH]] ], [ 1, [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[R_LCSSA:%.*]] = phi i64 [ [[R]], [[LOOP_LATCH]] ]
 ; CHECK-NEXT:    ret i64 [[R_LCSSA]]
 ;
 entry:
@@ -68,7 +68,7 @@ define i64 @pr88297_incoming_ops_reordered() {
 ; CHECK-NEXT:    [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
@@ -81,9 +81,9 @@ define i64 @pr88297_incoming_ops_reordered() {
 ; CHECK-NEXT:    [[R:%.*]] = phi i64 [ 0, [[LOOP_HEADER]] ], [ 1, [[THEN]] ]
 ; CHECK-NEXT:    [[IV_NEXT]] = add i32 [[IV]], 1
 ; CHECK-NEXT:    [[ICMP:%.*]] = icmp sgt i32 [[IV]], 1000
-; CHECK-NEXT:    br i1 [[ICMP]], label [[EXIT]], label [[LOOP_HEADER]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    br i1 [[ICMP]], label [[EXIT:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       exit:
-; CHECK-NEXT:    [[R_LCSSA:%.*]] = phi i64 [ [[R]], [[LOOP_LATCH]] ], [ 1, [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[R_LCSSA:%.*]] = phi i64 [ [[R]], [[LOOP_LATCH]] ]
 ; CHECK-NEXT:    ret i64 [[R_LCSSA]]
 ;
 entry:
@@ -115,7 +115,6 @@ define i64 @invar_cond(i1 %c) {
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[C]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i1> [[BROADCAST_SPLATINSERT]], <4 x i1> poison, <4 x i32> zeroinitializer
-; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[BROADCAST_SPLAT]], <4 x i64> zeroinitializer, <4 x i64> splat (i64 1)
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
@@ -123,8 +122,7 @@ define i64 @invar_cond(i1 %c) {
 ; CHECK-NEXT:    [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[TMP1:%.*]] = extractelement <4 x i64> [[PREDPHI]], i32 3
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
@@ -137,9 +135,9 @@ define i64 @invar_cond(i1 %c) {
 ; CHECK-NEXT:    [[R:%.*]] = phi i64 [ 1, [[THEN]] ], [ 0, [[LOOP_HEADER]] ]
 ; CHECK-NEXT:    [[IV_NEXT]] = add i32 [[IV]], 1
 ; CHECK-NEXT:    [[ICMP:%.*]] = icmp sgt i32 [[IV]], 1000
-; CHECK-NEXT:    br i1 [[ICMP]], label [[EXIT]], label [[LOOP_HEADER]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK-NEXT:    br i1 [[ICMP]], label [[EXIT:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP7:![0-9]+]]
 ; CHECK:       exit:
-; CHECK-NEXT:    [[R_LCSSA:%.*]] = phi i64 [ [[R]], [[LOOP_LATCH]] ], [ [[TMP1]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[R_LCSSA:%.*]] = phi i64 [ [[R]], [[LOOP_LATCH]] ]
 ; CHECK-NEXT:    ret i64 [[R_LCSSA]]
 ;
 entry:
@@ -171,7 +169,6 @@ define i64 @invar_cond_incoming_ops_reordered(i1 %c) {
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i1> poison, i1 [[C]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i1> [[BROADCAST_SPLATINSERT]], <4 x i1> poison, <4 x i32> zeroinitializer
-; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[BROADCAST_SPLAT]], <4 x i64> zeroinitializer, <4 x i64> splat (i64 1)
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
@@ -179,8 +176,7 @@ define i64 @invar_cond_incoming_ops_reordered(i1 %c) {
 ; CHECK-NEXT:    [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x i64> [[PREDPHI]], i32 3
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
@@ -193,9 +189,9 @@ define i64 @invar_cond_incoming_ops_reordered(i1 %c) {
 ; CHECK-NEXT:    [[R:%.*]] = phi i64 [ 0, [[LOOP_HEADER]] ], [ 1, [[THEN]] ]
 ; CHECK-NEXT:    [[IV_NEXT]] = add i32 [[IV]], 1
 ; CHECK-NEXT:    [[ICMP:%.*]] = icmp sgt i32 [[IV]], 1000
-; CHECK-NEXT:    br i1 [[ICMP]], label [[EXIT]], label [[LOOP_HEADER]], !llvm.loop [[LOOP9:![0-9]+]]
+; CHECK-NEXT:    br i1 [[ICMP]], label [[EXIT:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP9:![0-9]+]]
 ; CHECK:       exit:
-; CHECK-NEXT:    [[R_LCSSA:%.*]] = phi i64 [ [[R]], [[LOOP_LATCH]] ], [ [[TMP2]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[R_LCSSA:%.*]] = phi i64 [ [[R]], [[LOOP_LATCH]] ]
 ; CHECK-NEXT:    ret i64 [[R_LCSSA]]
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/bsd_regex.ll b/llvm/test/Transforms/LoopVectorize/bsd_regex.ll
index 8397b1e..c95ec0d 100644
--- a/llvm/test/Transforms/LoopVectorize/bsd_regex.ll
+++ b/llvm/test/Transforms/LoopVectorize/bsd_regex.ll
@@ -37,7 +37,7 @@ define i32 @foo(ptr nocapture %A) {
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10000
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
diff --git a/llvm/test/Transforms/LoopVectorize/check-prof-info.ll b/llvm/test/Transforms/LoopVectorize/check-prof-info.ll
index 4074194..d40fd22 100644
--- a/llvm/test/Transforms/LoopVectorize/check-prof-info.ll
+++ b/llvm/test/Transforms/LoopVectorize/check-prof-info.ll
@@ -19,12 +19,12 @@ define void @_Z3foov() {
 ; CHECK:  vector.body:
 ; CHECK:    br i1 [[TMP6:%.*]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF1:![0-9]+]], !llvm.loop [[LOOP2:![0-9]+]]
 ; CHECK:  middle.block:
-; CHECK:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]], !prof [[PROF5:![0-9]+]]
+; CHECK:    br label [[FOR_COND_CLEANUP:%.*]]
 ; CHECK:  scalar.ph:
 ; CHECK:    br label [[FOR_BODY:%.*]]
 ; CHECK:  for.cond.cleanup:
 ; CHECK:  for.body:
-; CHECK:    br i1 [[EXITCOND:%.*]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !prof [[PROF6:![0-9]+]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK:    br i1 [[EXITCOND:%.*]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !prof [[PROF5:![0-9]+]], !llvm.loop [[LOOP6:![0-9]+]]
 ;
 ; CHECK-MASKED-LABEL: @_Z3foov(
 ; CHECK-MASKED:  entry:
@@ -34,19 +34,22 @@ define void @_Z3foov() {
 ; CHECK-MASKED:  vector.body:
 ; CHECK-MASKED:    br i1 [[TMP18:%.*]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF1:![0-9]+]], !llvm.loop [[LOOP2:![0-9]+]]
 ; CHECK-MASKED:  middle.block:
-; CHECK-MASKED:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]], !prof [[PROF5:![0-9]+]]
+; CHECK-MASKED:    br label [[FOR_COND_CLEANUP:%.*]]
 ; CHECK-MASKED:  scalar.ph:
 ; CHECK-MASKED:    br label [[FOR_BODY:%.*]]
 ; CHECK-MASKED:  for.cond.cleanup:
 ; CHECK-MASKED:  for.body:
-; CHECK-MASKED:    br i1 [[EXITCOND:%.*]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !prof [[PROF6:![0-9]+]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK-MASKED:    br i1 [[EXITCOND:%.*]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !prof [[PROF5:![0-9]+]], !llvm.loop [[LOOP6:![0-9]+]]
 ;
 ; CHECK-SCALABLE-LABEL: @_Z3foov(
 ; CHECK-SCALABLE:  entry:
 ; CHECK-SCALABLE:    br i1 [[MIN_ITERS_CHECK:%.*]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]], !prof [[PROF0:![0-9]+]]
 ; CHECK-SCALABLE:  vector.ph:
+; CHECK-SCALABLE:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[TMP9:%.*]], i64 0
+; CHECK-SCALABLE:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
 ; CHECK-SCALABLE:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-SCALABLE:  vector.body:
+; CHECK-SCALABLE:    [[VEC_IND_NEXT:%.*]] = add <vscale x 4 x i32> [[VEC_IND:%.*]], [[BROADCAST_SPLAT]]
 ; CHECK-SCALABLE:    br i1 [[TMP16:%.*]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF1:![0-9]+]], !llvm.loop [[LOOP2:![0-9]+]]
 ; CHECK-SCALABLE:  middle.block:
 ; CHECK-SCALABLE:    br i1 [[CMP_N:%.*]], label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]], !prof [[PROF5:![0-9]+]]
@@ -85,14 +88,14 @@ define void @_Z3foo2v() {
 ; CHECK:  vector.ph:
 ; CHECK:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:  vector.body:
-; CHECK:    br i1 [[TMP6:%.*]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF1]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK:    br i1 [[TMP6:%.*]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF1]], !llvm.loop [[LOOP7:![0-9]+]]
 ; CHECK:  middle.block:
-; CHECK:    br i1 false, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]], !prof [[PROF5]]
+; CHECK:    br label [[SCALAR_PH]]
 ; CHECK:  scalar.ph:
 ; CHECK:    br label [[FOR_BODY:%.*]]
 ; CHECK:  for.cond.cleanup:
 ; CHECK:  for.body:
-; CHECK:    br i1 [[EXITCOND:%.*]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !prof [[PROF9:![0-9]+]], !llvm.loop [[LOOP10:![0-9]+]]
+; CHECK:    br i1 [[EXITCOND:%.*]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_BODY]], !prof [[PROF8:![0-9]+]], !llvm.loop [[LOOP9:![0-9]+]]
 ;
 ; CHECK-MASKED-LABEL: @_Z3foo2v(
 ; CHECK-MASKED:  entry:
@@ -100,21 +103,24 @@ define void @_Z3foo2v() {
 ; CHECK-MASKED:  vector.ph:
 ; CHECK-MASKED:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-MASKED:  vector.body:
-; CHECK-MASKED:    br i1 [[TMP18:%.*]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF1]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK-MASKED:    br i1 [[TMP18:%.*]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF1]], !llvm.loop [[LOOP7:![0-9]+]]
 ; CHECK-MASKED:  middle.block:
-; CHECK-MASKED:    br i1 false, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]], !prof [[PROF5]]
+; CHECK-MASKED:    br label [[SCALAR_PH]]
 ; CHECK-MASKED:  scalar.ph:
 ; CHECK-MASKED:    br label [[FOR_BODY:%.*]]
 ; CHECK-MASKED:  for.cond.cleanup:
 ; CHECK-MASKED:  for.body:
-; CHECK-MASKED:    br i1 [[EXITCOND:%.*]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !prof [[PROF9:![0-9]+]], !llvm.loop [[LOOP10:![0-9]+]]
+; CHECK-MASKED:    br i1 [[EXITCOND:%.*]], label [[FOR_COND_CLEANUP:%.*]], label [[FOR_BODY]], !prof [[PROF8:![0-9]+]], !llvm.loop [[LOOP9:![0-9]+]]
 ;
 ; CHECK-SCALABLE-LABEL: @_Z3foo2v(
 ; CHECK-SCALABLE:  entry:
 ; CHECK-SCALABLE:    br i1 [[MIN_ITERS_CHECK:%.*]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]], !prof [[PROF0]]
 ; CHECK-SCALABLE:  vector.ph:
+; CHECK-SCALABLE:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[TMP9:%.*]], i64 0
+; CHECK-SCALABLE:    [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
 ; CHECK-SCALABLE:    br label [[VECTOR_BODY:%.*]]
 ; CHECK-SCALABLE:  vector.body:
+; CHECK-SCALABLE:    [[VEC_IND_NEXT:%.*]] = add <vscale x 4 x i32> [[VEC_IND:%.*]], [[BROADCAST_SPLAT]]
 ; CHECK-SCALABLE:    br i1 [[TMP16:%.*]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !prof [[PROF1]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK-SCALABLE:  middle.block:
 ; CHECK-SCALABLE:    br i1 [[CMP_N:%.*]], label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]], !prof [[PROF5]]
diff --git a/llvm/test/Transforms/LoopVectorize/constantfolder-infer-correct-gepty.ll b/llvm/test/Transforms/LoopVectorize/constantfolder-infer-correct-gepty.ll
index e6eaa0e..af528ee 100644
--- a/llvm/test/Transforms/LoopVectorize/constantfolder-infer-correct-gepty.ll
+++ b/llvm/test/Transforms/LoopVectorize/constantfolder-infer-correct-gepty.ll
@@ -18,9 +18,9 @@ define void @test(ptr %data) {
 ; CHECK-NEXT:    store i16 [[TMP4]], ptr [[DATA]], align 2, !tbaa [[TBAA4:![0-9]+]]
 ; CHECK-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[END:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[END:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 8, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/constantfolder.ll b/llvm/test/Transforms/LoopVectorize/constantfolder.ll
index df926fa..cfd36bf 100644
--- a/llvm/test/Transforms/LoopVectorize/constantfolder.ll
+++ b/llvm/test/Transforms/LoopVectorize/constantfolder.ll
@@ -15,9 +15,9 @@ define void @const_fold_ptradd(ptr %dst, i64 %d) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP1]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -70,9 +70,9 @@ define void @const_fold_inbounds_ptradd(ptr %dst, i64 %d) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP1]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -126,9 +126,9 @@ define void @const_fold_select(ptr %dst, i64 %d) {
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -181,9 +181,9 @@ define void @const_fold_add_sub_mul_ashr_lshr(ptr %dst, i64 %d) {
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -244,9 +244,9 @@ define void @const_fold_and_or_xor(ptr %dst, i64 %d) {
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -303,9 +303,9 @@ define void @const_fold_cmp_zext(ptr %dst, i64 %d) {
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -360,9 +360,9 @@ define void @const_fold_trunc(ptr %dst, i64 %d) {
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP2]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/create-induction-resume.ll b/llvm/test/Transforms/LoopVectorize/create-induction-resume.ll
index 71273a3..e9c7f75 100644
--- a/llvm/test/Transforms/LoopVectorize/create-induction-resume.ll
+++ b/llvm/test/Transforms/LoopVectorize/create-induction-resume.ll
@@ -44,8 +44,6 @@ define void @test(i32 %arg, i32 %L1.limit, i32 %L2.switch, i1 %c, ptr %dst) {
 ; CHECK:       L2.Inner.header.preheader:
 ; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK-NEXT:    [[TMP3:%.*]] = mul i32 12, [[INDUCTION_IV_LCSSA1]]
-; CHECK-NEXT:    [[IND_END:%.*]] = add i32 1, [[TMP3]]
 ; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[L1_EXIT_VAL]], i64 0
 ; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <4 x i32> [[DOTSPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    [[DOTSPLATINSERT1:%.*]] = insertelement <4 x i32> poison, i32 [[INDUCTION_IV_LCSSA1]], i64 0
@@ -63,17 +61,16 @@ define void @test(i32 %arg, i32 %L1.limit, i32 %L2.switch, i1 %c, ptr %dst) {
 ; CHECK-NEXT:    [[TMP7:%.*]] = sub <4 x i32> [[VEC_IND]], [[DOTSPLAT]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = sext <4 x i32> [[TMP7]] to <4 x i64>
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[DST:%.*]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[TMP9]], i32 0
-; CHECK-NEXT:    store <4 x i64> [[TMP8]], ptr [[TMP10]], align 8
+; CHECK-NEXT:    store <4 x i64> [[TMP8]], ptr [[TMP9]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 12
 ; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[L2_HEADER_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[L2_HEADER_LOOPEXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 1, [[L2_INNER_HEADER_PREHEADER]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i64 [ 13, [[MIDDLE_BLOCK]] ], [ 1, [[L2_INNER_HEADER_PREHEADER]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1, [[L2_INNER_HEADER_PREHEADER]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i64 [ 1, [[L2_INNER_HEADER_PREHEADER]] ]
 ; CHECK-NEXT:    br label [[L2_INNER_HEADER:%.*]]
 ; CHECK:       L2.Inner.header:
 ; CHECK-NEXT:    [[L2_ACCUM:%.*]] = phi i32 [ [[L2_ACCUM_NEXT:%.*]], [[L2_INNER_HEADER]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/dbg-outer-loop-vect.ll b/llvm/test/Transforms/LoopVectorize/dbg-outer-loop-vect.ll
index ff922d4..baa967c 100644
--- a/llvm/test/Transforms/LoopVectorize/dbg-outer-loop-vect.ll
+++ b/llvm/test/Transforms/LoopVectorize/dbg-outer-loop-vect.ll
@@ -30,8 +30,8 @@ define void @foo(ptr %h) !dbg !4 {
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[FOR_COND_CLEANUP32]], label [[FOR_COND5_PREHEADER1]], !dbg [[DBG28]]
 ; CHECK:       vector.latch:
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4, !dbg [[DBG222]]
-; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 20, !dbg [[DBG222]]
-; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !dbg [[DBG222]], !llvm.loop [[LOOP29:![0-9]+]]
+; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 20, !dbg [[DBG21]]
+; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !dbg [[DBG21]], !llvm.loop [[LOOP29:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]], !dbg [[DBG21]]
 ; CHECK:       scalar.ph:
diff --git a/llvm/test/Transforms/LoopVectorize/dead_instructions.ll b/llvm/test/Transforms/LoopVectorize/dead_instructions.ll
index 653baf83..42d45bd 100644
--- a/llvm/test/Transforms/LoopVectorize/dead_instructions.ll
+++ b/llvm/test/Transforms/LoopVectorize/dead_instructions.ll
@@ -25,9 +25,8 @@ define i64 @dead_instructions_01(ptr %a, i64 %n) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP6:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x i64> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP7:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 2
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP4]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i64>, ptr [[TMP5]], align 8
 ; CHECK-NEXT:    [[TMP6]] = add <2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP7]] = add <2 x i64> [[WIDE_LOAD2]], [[VEC_PHI1]]
@@ -45,16 +44,16 @@ define i64 @dead_instructions_01(ptr %a, i64 %n) {
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[I:%.*]] = phi i64 [ [[I_NEXT:%.*]], %[[FOR_BODY]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
-; CHECK-NEXT:    [[R:%.*]] = phi i64 [ [[TMP5:%.*]], %[[FOR_BODY]] ], [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ]
+; CHECK-NEXT:    [[R:%.*]] = phi i64 [ [[TMP6:%.*]], %[[FOR_BODY]] ], [ [[BC_MERGE_RDX]], %[[SCALAR_PH]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[I]]
-; CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr [[TMP2]], align 8
-; CHECK-NEXT:    [[TMP5]] = add i64 [[TMP4]], [[R]]
+; CHECK-NEXT:    [[TMP5:%.*]] = load i64, ptr [[TMP2]], align 8
+; CHECK-NEXT:    [[TMP6]] = add i64 [[TMP5]], [[R]]
 ; CHECK-NEXT:    [[I_NEXT]] = add nuw nsw i64 [[I]], 1
 ; CHECK-NEXT:    [[COND:%.*]] = icmp slt i64 [[I_NEXT]], [[N]]
 ; CHECK-NEXT:    br i1 [[COND]], label %[[FOR_BODY]], label %[[FOR_END]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       [[FOR_END]]:
-; CHECK-NEXT:    [[TMP6:%.*]] = phi i64 [ [[TMP5]], %[[FOR_BODY]] ], [ [[TMP9]], %[[MIDDLE_BLOCK]] ]
-; CHECK-NEXT:    ret i64 [[TMP6]]
+; CHECK-NEXT:    [[TMP7:%.*]] = phi i64 [ [[TMP6]], %[[FOR_BODY]] ], [ [[TMP9]], %[[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    ret i64 [[TMP7]]
 ;
 entry:
   br label %for.body
@@ -149,23 +148,22 @@ define void @dead_load_and_vector_pointer(ptr %a, ptr %b) {
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 2
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP4]], align 8, !alias.scope [[META6:![0-9]+]], !noalias [[META9:![0-9]+]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 8, !alias.scope [[META6:![0-9]+]], !noalias [[META9:![0-9]+]]
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i32>, ptr [[TMP5]], align 8, !alias.scope [[META6]], !noalias [[META9]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = add <2 x i32> [[WIDE_LOAD]], splat (i32 1)
 ; CHECK-NEXT:    [[TMP7:%.*]] = add <2 x i32> [[WIDE_LOAD2]], splat (i32 1)
-; CHECK-NEXT:    store <2 x i32> [[TMP6]], ptr [[TMP4]], align 4, !alias.scope [[META6]], !noalias [[META9]]
+; CHECK-NEXT:    store <2 x i32> [[TMP6]], ptr [[TMP2]], align 4, !alias.scope [[META6]], !noalias [[META9]]
 ; CHECK-NEXT:    store <2 x i32> [[TMP7]], ptr [[TMP5]], align 4, !alias.scope [[META6]], !noalias [[META9]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i32 [[INDEX_NEXT]], 128
 ; CHECK-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 128, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
-; CHECK:       [[EXIT]]:
+; CHECK:       [[EXIT:.*]]:
 ; CHECK-NEXT:    ret void
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[PRIMARY:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[PRIMARY_ADD:%.*]], %[[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/debugloc-optimize-vfuf-term.ll b/llvm/test/Transforms/LoopVectorize/debugloc-optimize-vfuf-term.ll
index 913a1341..9ade6e9 100644
--- a/llvm/test/Transforms/LoopVectorize/debugloc-optimize-vfuf-term.ll
+++ b/llvm/test/Transforms/LoopVectorize/debugloc-optimize-vfuf-term.ll
@@ -14,17 +14,17 @@ define i32 @foo(ptr %p) {
 ; CHECK-NEXT:    store i8 0, ptr [[P]], align 1, !dbg [[DBG3:![0-9]+]]
 ; CHECK-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]], !dbg [[DBG7:![0-9]+]]
+; CHECK-NEXT:    br label %[[EXIT:.*]], !dbg [[DBG3]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 2, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], !dbg [[DBG8:![0-9]+]]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ], !dbg [[DBG7:![0-9]+]]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
-; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], !dbg [[DBG8]]
-; CHECK-NEXT:    [[CONV:%.*]] = trunc i64 0 to i8, !dbg [[DBG9:![0-9]+]]
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], !dbg [[DBG7]]
+; CHECK-NEXT:    [[CONV:%.*]] = trunc i64 0 to i8, !dbg [[DBG8:![0-9]+]]
 ; CHECK-NEXT:    store i8 [[CONV]], ptr [[P]], align 1, !dbg [[DBG3]]
-; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1, !dbg [[DBG10:![0-9]+]]
-; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV]], 1, !dbg [[DBG11:![0-9]+]]
-; CHECK-NEXT:    br i1 [[EXITCOND]], label %[[EXIT]], label %[[LOOP]], !dbg [[DBG7]], !llvm.loop [[LOOP12:![0-9]+]]
+; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1, !dbg [[DBG9:![0-9]+]]
+; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV]], 1, !dbg [[DBG10:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label %[[EXIT]], label %[[LOOP]], !dbg [[DBG11:![0-9]+]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    ret i32 0
 ;
@@ -60,16 +60,16 @@ exit:                              ; preds = %loop
 !11 = !{}
 ;.
 ; CHECK: [[META0:![0-9]+]] = distinct !DICompileUnit(language: DW_LANG_C_plus_plus_14, file: [[META1:![0-9]+]], producer: "{{.*}}clang version {{.*}}", isOptimized: false, runtimeVersion: 0, emissionKind: NoDebug)
-; CHECK: [[META1]] = !DIFile(filename: "test.cpp", directory: {{.*}})
+; CHECK: [[META1]] = !DIFile(filename: "{{.*}}test.cpp", directory: {{.*}})
 ; CHECK: [[DBG3]] = !DILocation(line: 6, scope: [[META4:![0-9]+]])
 ; CHECK: [[META4]] = distinct !DISubprogram(name: "foo", scope: [[META1]], file: [[META1]], line: 11, type: [[META5:![0-9]+]], spFlags: DISPFlagDefinition, unit: [[META0]], retainedNodes: [[META6:![0-9]+]])
 ; CHECK: [[META5]] = distinct !DISubroutineType(types: [[META6]])
 ; CHECK: [[META6]] = !{}
-; CHECK: [[DBG7]] = !DILocation(line: 9, scope: [[META4]])
-; CHECK: [[DBG8]] = !DILocation(line: 4, scope: [[META4]])
-; CHECK: [[DBG9]] = !DILocation(line: 5, scope: [[META4]])
-; CHECK: [[DBG10]] = !DILocation(line: 7, scope: [[META4]])
-; CHECK: [[DBG11]] = !DILocation(line: 8, scope: [[META4]])
+; CHECK: [[DBG7]] = !DILocation(line: 4, scope: [[META4]])
+; CHECK: [[DBG8]] = !DILocation(line: 5, scope: [[META4]])
+; CHECK: [[DBG9]] = !DILocation(line: 7, scope: [[META4]])
+; CHECK: [[DBG10]] = !DILocation(line: 8, scope: [[META4]])
+; CHECK: [[DBG11]] = !DILocation(line: 9, scope: [[META4]])
 ; CHECK: [[LOOP12]] = distinct !{[[LOOP12]], [[META13:![0-9]+]], [[META14:![0-9]+]]}
 ; CHECK: [[META13]] = !{!"llvm.loop.unroll.runtime.disable"}
 ; CHECK: [[META14]] = !{!"llvm.loop.isvectorized", i32 1}
diff --git a/llvm/test/Transforms/LoopVectorize/debugloc.ll b/llvm/test/Transforms/LoopVectorize/debugloc.ll
index e7224ae..40cd6b6 100644
--- a/llvm/test/Transforms/LoopVectorize/debugloc.ll
+++ b/llvm/test/Transforms/LoopVectorize/debugloc.ll
@@ -14,9 +14,9 @@ target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f3
 ; CHECK:   load <2 x i32>, ptr {{.*}}, !dbg ![[LOC1]]
 ; CHECK:   add <2 x i32> {{.*}}, !dbg ![[LOC1]]
 ; CHECK:   add nuw i64 %index, 2, !dbg ![[LOC1]]
-; CHECK:   icmp eq i64 %index.next, %n.vec, !dbg ![[LOC1]]
+; CHECK:   icmp eq i64 %index.next, %n.vec, !dbg ![[BR_LOC:[0-9]+]]
 ; CHECK: middle.block
-; CHECK:   call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> %{{.*}}), !dbg ![[BR_LOC:[0-9]+]]
+; CHECK:   call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> %{{.*}}), !dbg ![[BR_LOC]]
 ; CHECK: for.body
 ; CHECK: br i1{{.*}}, label %for.body,{{.*}}, !dbg ![[BR_LOC]],
 
@@ -141,12 +141,10 @@ define void @test_misc(ptr nocapture %a, ptr noalias %b, i64 %size) !dbg !35 {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %vector.ph ], [ [[INDEX_NEXT:%.*]], %vector.body ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr %a, i64 [[INDEX]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr %b, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP3]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp uge <2 x i32> [[WIDE_LOAD]], splat (i32 10)
 ; CHECK-NEXT:    [[TMP5:%.*]] = select <2 x i1> [[TMP4]], <2 x i32> [[WIDE_LOAD]], <2 x i32> zeroinitializer, !dbg [[LOC6:![0-9]+]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0, !dbg [[LOC7:![0-9]+]]
-; CHECK-NEXT:    store <2 x i32> [[TMP5]], ptr [[TMP6]], align 4, !dbg [[LOC7]]
+; CHECK-NEXT:    store <2 x i32> [[TMP5]], ptr [[TMP2]], align 4, !dbg [[LOC7:![0-9]+]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]],
 ; CHECK-NEXT:    br i1 [[TMP7]], label %middle.block, label %vector.body
diff --git a/llvm/test/Transforms/LoopVectorize/dereferenceable-info-from-assumption-constant-size.ll b/llvm/test/Transforms/LoopVectorize/dereferenceable-info-from-assumption-constant-size.ll
index 7a54519..373c8e0 100644
--- a/llvm/test/Transforms/LoopVectorize/dereferenceable-info-from-assumption-constant-size.ll
+++ b/llvm/test/Transforms/LoopVectorize/dereferenceable-info-from-assumption-constant-size.ll
@@ -19,8 +19,7 @@ define void @deref_assumption_in_header_constant_trip_count(ptr noalias noundef
 ; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x ptr> [[TMP1]], i32 1
 ; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "align"(ptr [[TMP5]], i64 4), "dereferenceable"(ptr [[TMP5]], i64 4) ]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP7]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP10:%.*]] = xor <2 x i1> [[TMP9]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i1> [[TMP10]], i32 0
@@ -43,16 +42,15 @@ define void @deref_assumption_in_header_constant_trip_count(ptr noalias noundef
 ; CHECK-NEXT:    [[TMP17:%.*]] = phi <2 x i32> [ [[TMP12]], %[[PRED_LOAD_CONTINUE]] ], [ [[TMP16]], %[[PRED_LOAD_IF1]] ]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP9]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP17]]
 ; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP31:%.*]] = getelementptr inbounds i32, ptr [[TMP30]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP31]], align 4
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP30]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP32]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -114,8 +112,7 @@ define void @align_deref_assumption_in_header_constant_trip_count_loop_invariant
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = load i32, ptr [[A]], align 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[A]], align 4
@@ -123,15 +120,14 @@ define void @align_deref_assumption_in_header_constant_trip_count_loop_invariant
 ; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <2 x i32> [[TMP6]], i32 [[TMP5]], i32 1
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP3]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP7]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP9]], align 4
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -195,8 +191,7 @@ define void @deref_assumption_too_small_in_header_constant_trip_count(ptr noalia
 ; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x ptr> [[TMP1]], i32 1
 ; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "align"(ptr [[TMP5]], i64 4), "dereferenceable"(ptr [[TMP5]], i64 2) ]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP7]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP10:%.*]] = xor <2 x i1> [[TMP9]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i1> [[TMP10]], i32 0
@@ -219,16 +214,15 @@ define void @deref_assumption_too_small_in_header_constant_trip_count(ptr noalia
 ; CHECK-NEXT:    [[TMP17:%.*]] = phi <2 x i32> [ [[TMP12]], %[[PRED_LOAD_CONTINUE]] ], [ [[TMP16]], %[[PRED_LOAD_IF1]] ]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP9]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP17]]
 ; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP31:%.*]] = getelementptr inbounds i32, ptr [[TMP30]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP31]], align 4
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP30]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP32]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -295,8 +289,7 @@ define void @deref_assumption_in_header_constant_trip_count_align_1(ptr noalias
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x ptr> [[TMP1]], i32 1
 ; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(ptr [[TMP3]], i64 4) ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP5]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP7:%.*]] = xor <2 x i1> [[TMP6]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i1> [[TMP7]], i32 0
@@ -319,16 +312,15 @@ define void @deref_assumption_in_header_constant_trip_count_align_1(ptr noalias
 ; CHECK-NEXT:    [[TMP17:%.*]] = phi <2 x i32> [ [[TMP12]], %[[PRED_LOAD_CONTINUE]] ], [ [[TMP16]], %[[PRED_LOAD_IF1]] ]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP6]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP17]]
 ; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[TMP18]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP19]], align 4
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP18]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -395,8 +387,7 @@ define void @deref_assumption_in_header_constant_trip_count_align_via_arg_attrib
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x ptr> [[TMP1]], i32 1
 ; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(ptr [[TMP3]], i64 4) ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP5]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP7:%.*]] = xor <2 x i1> [[TMP6]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i1> [[TMP7]], i32 0
@@ -419,16 +410,15 @@ define void @deref_assumption_in_header_constant_trip_count_align_via_arg_attrib
 ; CHECK-NEXT:    [[TMP17:%.*]] = phi <2 x i32> [ [[TMP12]], %[[PRED_LOAD_CONTINUE]] ], [ [[TMP16]], %[[PRED_LOAD_IF1]] ]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP6]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP17]]
 ; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[TMP18]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP19]], align 4
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP18]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -495,8 +485,7 @@ define void @deref_assumption_in_header_constant_trip_count_align_not_known(ptr
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <2 x ptr> [[TMP1]], i32 1
 ; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "dereferenceable"(ptr [[TMP3]], i64 4) ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP5]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP7:%.*]] = xor <2 x i1> [[TMP6]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i1> [[TMP7]], i32 0
@@ -519,16 +508,15 @@ define void @deref_assumption_in_header_constant_trip_count_align_not_known(ptr
 ; CHECK-NEXT:    [[TMP17:%.*]] = phi <2 x i32> [ [[TMP12]], %[[PRED_LOAD_CONTINUE]] ], [ [[TMP16]], %[[PRED_LOAD_IF1]] ]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP6]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP17]]
 ; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[TMP18]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP19]], align 4
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP18]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP20]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -590,8 +578,7 @@ define void @deref_assumption_in_then_constant_trip_count(ptr noalias noundef %a
 ; CHECK-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_LOAD_CONTINUE2:.*]] ]
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[PRED_LOAD_CONTINUE2]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = xor <2 x i1> [[TMP3]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[A]], <2 x i64> [[VEC_IND]]
@@ -615,16 +602,15 @@ define void @deref_assumption_in_then_constant_trip_count(ptr noalias noundef %a
 ; CHECK-NEXT:    [[TMP15:%.*]] = phi <2 x i32> [ [[TMP10]], %[[PRED_LOAD_CONTINUE]] ], [ [[TMP14]], %[[PRED_LOAD_IF1]] ]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP3]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP15]]
 ; CHECK-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i32, ptr [[TMP26]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP27]], align 4
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP26]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP28:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP28]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -687,8 +673,7 @@ define void @deref_assumption_in_latch_constant_trip_count(ptr noalias noundef %
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[PRED_LOAD_CONTINUE2]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i32, ptr [[A]], <2 x i64> [[VEC_IND]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP3]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP5:%.*]] = xor <2 x i1> [[TMP4]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <2 x i1> [[TMP5]], i32 0
@@ -717,16 +702,15 @@ define void @deref_assumption_in_latch_constant_trip_count(ptr noalias noundef %
 ; CHECK-NEXT:    [[TMP19:%.*]] = extractelement <2 x ptr> [[TMP1]], i32 1
 ; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "align"(ptr [[TMP29]], i64 4), "dereferenceable"(ptr [[TMP19]], i64 4) ]
 ; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP31:%.*]] = getelementptr inbounds i32, ptr [[TMP30]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP31]], align 4
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP30]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP32]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -796,8 +780,7 @@ define void @deref_assumption_in_header_variable_trip_count(ptr noalias noundef
 ; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x ptr> [[TMP1]], i32 1
 ; CHECK-NEXT:    call void @llvm.assume(i1 true) [ "align"(ptr [[TMP5]], i64 4), "dereferenceable"(ptr [[TMP5]], i64 4) ]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP7]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP10:%.*]] = xor <2 x i1> [[TMP9]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i1> [[TMP10]], i32 0
@@ -820,8 +803,7 @@ define void @deref_assumption_in_header_variable_trip_count(ptr noalias noundef
 ; CHECK-NEXT:    [[TMP17:%.*]] = phi <2 x i32> [ [[TMP12]], %[[PRED_LOAD_CONTINUE]] ], [ [[TMP16]], %[[PRED_LOAD_IF1]] ]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP9]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP17]]
 ; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP31:%.*]] = getelementptr inbounds i32, ptr [[TMP30]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP31]], align 4
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP30]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -892,23 +874,20 @@ define void @deref_assumption_in_preheader_constant_trip_count_align_1(ptr noali
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i32, ptr [[A]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    [[TMP15:%.*]] = load <2 x i32>, ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[TMP15:%.*]] = load <2 x i32>, ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP3]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP15]]
 ; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP17]], align 4
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP16]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP18]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -969,8 +948,7 @@ define void @deref_assumption_too_small_in_preheader_constant_trip_count_align_1
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_LOAD_CONTINUE2:.*]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = xor <2 x i1> [[TMP3]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x i1> [[TMP4]], i32 0
@@ -995,15 +973,14 @@ define void @deref_assumption_too_small_in_preheader_constant_trip_count_align_1
 ; CHECK-NEXT:    [[TMP15:%.*]] = phi <2 x i32> [ [[TMP9]], %[[PRED_LOAD_CONTINUE]] ], [ [[TMP14]], %[[PRED_LOAD_IF1]] ]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP3]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP15]]
 ; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP17]], align 4
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP16]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP18]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -1064,23 +1041,20 @@ define void @align_and_deref_assumption_in_preheader_constant_trip_count_align_4
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[A]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i32>, ptr [[TMP5]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i32>, ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP3]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[WIDE_LOAD1]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP7]], align 4
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -1142,23 +1116,20 @@ define void @deref_assumption_in_preheader_constant_trip_count_align_4_known_via
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[A]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[TMP15:%.*]] = load <2 x i32>, ptr [[TMP4]], align 4
+; CHECK-NEXT:    [[TMP15:%.*]] = load <2 x i32>, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP3]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP15]]
 ; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP17]], align 4
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP16]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP18]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -1219,8 +1190,7 @@ define void @deref_assumption_in_preheader_constant_trip_count_align_4_not_known
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_LOAD_CONTINUE2:.*]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = xor <2 x i1> [[TMP3]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x i1> [[TMP4]], i32 0
@@ -1245,15 +1215,14 @@ define void @deref_assumption_in_preheader_constant_trip_count_align_4_not_known
 ; CHECK-NEXT:    [[TMP15:%.*]] = phi <2 x i32> [ [[TMP9]], %[[PRED_LOAD_CONTINUE]] ], [ [[TMP14]], %[[PRED_LOAD_IF1]] ]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP3]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP15]]
 ; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP17]], align 4
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP16]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP18]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -1314,8 +1283,7 @@ define void @deref_assumption_too_small_in_preheader_constant_trip_count_align_4
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_LOAD_CONTINUE2:.*]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = xor <2 x i1> [[TMP3]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x i1> [[TMP4]], i32 0
@@ -1340,15 +1308,14 @@ define void @deref_assumption_too_small_in_preheader_constant_trip_count_align_4
 ; CHECK-NEXT:    [[TMP15:%.*]] = phi <2 x i32> [ [[TMP9]], %[[PRED_LOAD_CONTINUE]] ], [ [[TMP14]], %[[PRED_LOAD_IF1]] ]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP3]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP15]]
 ; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP17]], align 4
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP16]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP18]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -1411,8 +1378,7 @@ define void @may_free_align_deref_assumption_in_header_constant_trip_count_loop_
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_LOAD_CONTINUE2:.*]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = xor <2 x i1> [[TMP3]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x i1> [[TMP4]], i32 0
@@ -1433,15 +1399,14 @@ define void @may_free_align_deref_assumption_in_header_constant_trip_count_loop_
 ; CHECK-NEXT:    [[TMP11:%.*]] = phi <2 x i32> [ [[TMP12]], %[[PRED_LOAD_CONTINUE]] ], [ [[TMP16]], %[[PRED_LOAD_IF1]] ]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP3]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP11]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP9]], align 4
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -1504,8 +1469,7 @@ define void @may_free_local_ptr_align_deref_assumption_in_header_constant_trip_c
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_LOAD_CONTINUE2:.*]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = xor <2 x i1> [[TMP3]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x i1> [[TMP4]], i32 0
@@ -1526,15 +1490,14 @@ define void @may_free_local_ptr_align_deref_assumption_in_header_constant_trip_c
 ; CHECK-NEXT:    [[TMP12:%.*]] = phi <2 x i32> [ [[TMP8]], %[[PRED_LOAD_CONTINUE]] ], [ [[TMP11]], %[[PRED_LOAD_IF1]] ]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP3]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP12]]
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP14]], align 4
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP13]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP15]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/dereferenceable-info-from-assumption-variable-size.ll b/llvm/test/Transforms/LoopVectorize/dereferenceable-info-from-assumption-variable-size.ll
index e771c40..c8cf2ad 100644
--- a/llvm/test/Transforms/LoopVectorize/dereferenceable-info-from-assumption-variable-size.ll
+++ b/llvm/test/Transforms/LoopVectorize/dereferenceable-info-from-assumption-variable-size.ll
@@ -19,15 +19,12 @@ define void @deref_assumption_in_preheader_non_constant_trip_count_access_i8(ptr
 ; CHECK-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP0]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[TMP3]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[TMP2]], align 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp sge <2 x i8> [[WIDE_LOAD]], zeroinitializer
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i8>, ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP4]], <2 x i8> [[WIDE_LOAD]], <2 x i8> [[WIDE_LOAD1]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <2 x i8> [[PREDPHI]], ptr [[TMP7]], align 1
+; CHECK-NEXT:    store <2 x i8> [[PREDPHI]], ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -102,15 +99,12 @@ define void @deref_assumption_in_preheader_non_constant_trip_count_access_i32(pt
 ; CHECK-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i32, ptr [[A]], i64 [[TMP0]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP3]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i32>, ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i32>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP4]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[WIDE_LOAD1]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP7]], align 1
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -185,8 +179,7 @@ define void @deref_assumption_in_preheader_too_small_non_constant_trip_count_acc
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_LOAD_CONTINUE2:.*]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP3]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP15:%.*]] = xor <2 x i1> [[TMP4]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x i1> [[TMP15]], i32 0
@@ -211,8 +204,7 @@ define void @deref_assumption_in_preheader_too_small_non_constant_trip_count_acc
 ; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = phi <2 x i32> [ [[TMP9]], %[[PRED_LOAD_CONTINUE]] ], [ [[TMP14]], %[[PRED_LOAD_IF1]] ]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP4]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[WIDE_LOAD1]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP7]], align 1
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
@@ -285,8 +277,7 @@ define void @deref_assumption_in_preheader_too_small2_non_constant_trip_count_ac
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_LOAD_CONTINUE2:.*]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP3]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP15:%.*]] = xor <2 x i1> [[TMP4]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x i1> [[TMP15]], i32 0
@@ -311,8 +302,7 @@ define void @deref_assumption_in_preheader_too_small2_non_constant_trip_count_ac
 ; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = phi <2 x i32> [ [[TMP9]], %[[PRED_LOAD_CONTINUE]] ], [ [[TMP14]], %[[PRED_LOAD_IF1]] ]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP4]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[WIDE_LOAD1]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP7]], align 1
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
@@ -387,15 +377,12 @@ define void @deref_assumption_in_preheader_non_constant_trip_count_access_i32_al
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i32, ptr [[A]], i64 [[INDEX]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i32>, ptr [[TMP4]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i32>, ptr [[TMP0]], align 4
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP3]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[WIDE_LOAD1]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP6]], align 1
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP5]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
@@ -470,8 +457,7 @@ define void @deref_assumption_in_preheader_non_constant_trip_count_access_i32_al
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_LOAD_CONTINUE2:.*]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP0]], align 4
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP3:%.*]] = xor <2 x i1> [[TMP2]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x i1> [[TMP3]], i32 0
@@ -496,8 +482,7 @@ define void @deref_assumption_in_preheader_non_constant_trip_count_access_i32_al
 ; CHECK-NEXT:    [[TMP15:%.*]] = phi <2 x i32> [ [[TMP9]], %[[PRED_LOAD_CONTINUE]] ], [ [[TMP14]], %[[PRED_LOAD_IF1]] ]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP2]], <2 x i32> [[WIDE_LOAD]], <2 x i32> [[TMP15]]
 ; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP17]], align 1
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP16]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP18]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/dont-fold-tail-for-const-TC.ll b/llvm/test/Transforms/LoopVectorize/dont-fold-tail-for-const-TC.ll
index 3c556e3d..4f95bdd 100644
--- a/llvm/test/Transforms/LoopVectorize/dont-fold-tail-for-const-TC.ll
+++ b/llvm/test/Transforms/LoopVectorize/dont-fold-tail-for-const-TC.ll
@@ -15,19 +15,18 @@ define dso_local void @constTC(ptr noalias nocapture %A) optsize {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 2
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 4
-; CHECK-NEXT:    store <2 x i32> splat (i32 13), ptr [[TMP6]], align 1
+; CHECK-NEXT:    store <2 x i32> splat (i32 13), ptr [[TMP3]], align 1
 ; CHECK-NEXT:    store <2 x i32> splat (i32 13), ptr [[TMP7]], align 1
 ; CHECK-NEXT:    store <2 x i32> splat (i32 13), ptr [[TMP8]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 6
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1800
 ; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1800, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[RIV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[RIVPLUS1:%.*]], [[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/dont-fold-tail-for-divisible-TC.ll b/llvm/test/Transforms/LoopVectorize/dont-fold-tail-for-divisible-TC.ll
index aa8299b..1936b40 100644
--- a/llvm/test/Transforms/LoopVectorize/dont-fold-tail-for-divisible-TC.ll
+++ b/llvm/test/Transforms/LoopVectorize/dont-fold-tail-for-divisible-TC.ll
@@ -19,8 +19,7 @@ define dso_local void @alignTC(ptr noalias nocapture %A, i32 %n) optsize {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <4 x i32> splat (i32 13), ptr [[TMP2]], align 1
+; CHECK-NEXT:    store <4 x i32> splat (i32 13), ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -83,8 +82,7 @@ define dso_local void @assumeAlignedTC(ptr noalias nocapture %A, i32 %p, i32 %q)
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <4 x i32> splat (i32 13), ptr [[TMP2]], align 1
+; CHECK-NEXT:    store <4 x i32> splat (i32 13), ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/epilog-iv-select-cmp.ll b/llvm/test/Transforms/LoopVectorize/epilog-iv-select-cmp.ll
index 6bf8883..968e107 100644
--- a/llvm/test/Transforms/LoopVectorize/epilog-iv-select-cmp.ll
+++ b/llvm/test/Transforms/LoopVectorize/epilog-iv-select-cmp.ll
@@ -19,8 +19,7 @@ define i64 @select_icmp_const(ptr %a, i64 %n) {
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ splat (i64 -9223372036854775808), %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 3)
 ; CHECK-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i64> [[VEC_IND]], <4 x i64> [[VEC_PHI]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 4
@@ -55,8 +54,7 @@ define i64 @select_icmp_const(ptr %a, i64 %n) {
 ; CHECK-NEXT:    [[VEC_IND5:%.*]] = phi <4 x i64> [ [[INDUCTION]], %[[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT6:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI7:%.*]] = phi <4 x i64> [ [[DOTSPLAT]], %[[VEC_EPILOG_PH]] ], [ [[TMP11:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP7]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x i64>, ptr [[TMP9]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x i64>, ptr [[TMP8]], align 8
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD8]], splat (i64 3)
 ; CHECK-NEXT:    [[TMP11]] = select <4 x i1> [[TMP10]], <4 x i64> [[VEC_IND5]], <4 x i64> [[VEC_PHI7]]
 ; CHECK-NEXT:    [[INDEX_NEXT9]] = add nuw i64 [[TMP7]], 4
@@ -70,12 +68,12 @@ define i64 @select_icmp_const(ptr %a, i64 %n) {
 ; CHECK-NEXT:    [[CMP_N12:%.*]] = icmp eq i64 [[N]], [[N_VEC3]]
 ; CHECK-NEXT:    br i1 [[CMP_N12]], label %[[EXIT]], label %[[VEC_EPILOG_SCALAR_PH]]
 ; CHECK:       [[VEC_EPILOG_SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL13:%.*]] = phi i64 [ [[N_VEC3]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], %[[VEC_EPILOG_ITER_CHECK]] ], [ 0, %[[ITER_CHECK]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX14:%.*]] = phi i64 [ [[RDX_SELECT11]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[RDX_SELECT]], %[[VEC_EPILOG_ITER_CHECK]] ], [ 3, %[[ITER_CHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL15:%.*]] = phi i64 [ [[N_VEC3]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], %[[VEC_EPILOG_ITER_CHECK]] ], [ 0, %[[ITER_CHECK]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX16:%.*]] = phi i64 [ [[RDX_SELECT11]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[RDX_SELECT]], %[[VEC_EPILOG_ITER_CHECK]] ], [ 3, %[[ITER_CHECK]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
-; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL13]], %[[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
-; CHECK-NEXT:    [[RDX:%.*]] = phi i64 [ [[BC_MERGE_RDX14]], %[[VEC_EPILOG_SCALAR_PH]] ], [ [[SEL:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL15]], %[[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[RDX:%.*]] = phi i64 [ [[BC_MERGE_RDX16]], %[[VEC_EPILOG_SCALAR_PH]] ], [ [[SEL:%.*]], %[[LOOP]] ]
 ; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[IV]]
 ; CHECK-NEXT:    [[L:%.*]] = load i64, ptr [[GEP]], align 8
 ; CHECK-NEXT:    [[C:%.*]] = icmp eq i64 [[L]], 3
@@ -123,8 +121,7 @@ define i64 @select_fcmp_const_fast(ptr %a, i64 %n) {
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ splat (i64 -9223372036854775808), %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = fcmp fast ueq <4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00)
 ; CHECK-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i64> [[VEC_IND]], <4 x i64> [[VEC_PHI]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 4
@@ -159,8 +156,7 @@ define i64 @select_fcmp_const_fast(ptr %a, i64 %n) {
 ; CHECK-NEXT:    [[VEC_IND5:%.*]] = phi <4 x i64> [ [[INDUCTION]], %[[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT6:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI7:%.*]] = phi <4 x i64> [ [[DOTSPLAT]], %[[VEC_EPILOG_PH]] ], [ [[TMP11:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP7]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x float>, ptr [[TMP9]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x float>, ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[TMP10:%.*]] = fcmp fast ueq <4 x float> [[WIDE_LOAD8]], splat (float 3.000000e+00)
 ; CHECK-NEXT:    [[TMP11]] = select <4 x i1> [[TMP10]], <4 x i64> [[VEC_IND5]], <4 x i64> [[VEC_PHI7]]
 ; CHECK-NEXT:    [[INDEX_NEXT9]] = add nuw i64 [[TMP7]], 4
@@ -174,12 +170,12 @@ define i64 @select_fcmp_const_fast(ptr %a, i64 %n) {
 ; CHECK-NEXT:    [[CMP_N12:%.*]] = icmp eq i64 [[N]], [[N_VEC3]]
 ; CHECK-NEXT:    br i1 [[CMP_N12]], label %[[EXIT]], label %[[VEC_EPILOG_SCALAR_PH]]
 ; CHECK:       [[VEC_EPILOG_SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL13:%.*]] = phi i64 [ [[N_VEC3]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], %[[VEC_EPILOG_ITER_CHECK]] ], [ 0, %[[ITER_CHECK]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX14:%.*]] = phi i64 [ [[RDX_SELECT11]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[RDX_SELECT]], %[[VEC_EPILOG_ITER_CHECK]] ], [ 2, %[[ITER_CHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL15:%.*]] = phi i64 [ [[N_VEC3]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[N_VEC]], %[[VEC_EPILOG_ITER_CHECK]] ], [ 0, %[[ITER_CHECK]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX16:%.*]] = phi i64 [ [[RDX_SELECT11]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[RDX_SELECT]], %[[VEC_EPILOG_ITER_CHECK]] ], [ 2, %[[ITER_CHECK]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
-; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL13]], %[[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
-; CHECK-NEXT:    [[RDX:%.*]] = phi i64 [ [[BC_MERGE_RDX14]], %[[VEC_EPILOG_SCALAR_PH]] ], [ [[SEL:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL15]], %[[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[RDX:%.*]] = phi i64 [ [[BC_MERGE_RDX16]], %[[VEC_EPILOG_SCALAR_PH]] ], [ [[SEL:%.*]], %[[LOOP]] ]
 ; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[IV]]
 ; CHECK-NEXT:    [[L:%.*]] = load float, ptr [[GEP]], align 4
 ; CHECK-NEXT:    [[C:%.*]] = fcmp fast ueq float [[L]], 3.000000e+00
@@ -233,8 +229,7 @@ define i8 @select_icmp_var_start(ptr %a, i8 %n, i8 %start) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i8> [ splat (i8 -128), %[[VECTOR_PH]] ], [ [[TMP8:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = trunc i32 [[INDEX]] to i8
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[A]], i8 [[TMP4]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP6]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP5]], align 8
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq <4 x i8> [[WIDE_LOAD]], splat (i8 3)
 ; CHECK-NEXT:    [[TMP8]] = select <4 x i1> [[TMP7]], <4 x i8> [[VEC_IND]], <4 x i8> [[VEC_PHI]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
@@ -273,8 +268,7 @@ define i8 @select_icmp_var_start(ptr %a, i8 %n, i8 %start) {
 ; CHECK-NEXT:    [[VEC_PHI7:%.*]] = phi <4 x i8> [ [[DOTSPLAT]], %[[VEC_EPILOG_PH]] ], [ [[TMP17:%.*]], %[[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX4]] to i8
 ; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[A]], i8 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[TMP14]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD10:%.*]] = load <4 x i8>, ptr [[TMP15]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD10:%.*]] = load <4 x i8>, ptr [[TMP14]], align 8
 ; CHECK-NEXT:    [[TMP16:%.*]] = icmp eq <4 x i8> [[WIDE_LOAD10]], splat (i8 3)
 ; CHECK-NEXT:    [[TMP17]] = select <4 x i1> [[TMP16]], <4 x i8> [[VEC_IND5]], <4 x i8> [[VEC_PHI7]]
 ; CHECK-NEXT:    [[INDEX_NEXT11]] = add nuw i32 [[INDEX4]], 4
@@ -288,12 +282,12 @@ define i8 @select_icmp_var_start(ptr %a, i8 %n, i8 %start) {
 ; CHECK-NEXT:    [[CMP_N14:%.*]] = icmp eq i32 [[TMP2]], [[N_VEC3]]
 ; CHECK-NEXT:    br i1 [[CMP_N14]], label %[[EXIT]], label %[[VEC_EPILOG_SCALAR_PH]]
 ; CHECK:       [[VEC_EPILOG_SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL13:%.*]] = phi i8 [ [[TMP13]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[IND_END]], %[[VEC_EPILOG_ITER_CHECK]] ], [ 0, %[[ITER_CHECK]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX14:%.*]] = phi i8 [ [[RDX_SELECT13]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[RDX_SELECT]], %[[VEC_EPILOG_ITER_CHECK]] ], [ [[START]], %[[ITER_CHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL15:%.*]] = phi i8 [ [[TMP13]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[IND_END]], %[[VEC_EPILOG_ITER_CHECK]] ], [ 0, %[[ITER_CHECK]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX16:%.*]] = phi i8 [ [[RDX_SELECT13]], %[[VEC_EPILOG_MIDDLE_BLOCK]] ], [ [[RDX_SELECT]], %[[VEC_EPILOG_ITER_CHECK]] ], [ [[START]], %[[ITER_CHECK]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
-; CHECK-NEXT:    [[IV:%.*]] = phi i8 [ [[BC_RESUME_VAL13]], %[[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
-; CHECK-NEXT:    [[RDX:%.*]] = phi i8 [ [[BC_MERGE_RDX14]], %[[VEC_EPILOG_SCALAR_PH]] ], [ [[SEL:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[IV:%.*]] = phi i8 [ [[BC_RESUME_VAL15]], %[[VEC_EPILOG_SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[RDX:%.*]] = phi i8 [ [[BC_MERGE_RDX16]], %[[VEC_EPILOG_SCALAR_PH]] ], [ [[SEL:%.*]], %[[LOOP]] ]
 ; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i8, ptr [[A]], i8 [[IV]]
 ; CHECK-NEXT:    [[L:%.*]] = load i8, ptr [[GEP]], align 8
 ; CHECK-NEXT:    [[C:%.*]] = icmp eq i8 [[L]], 3
diff --git a/llvm/test/Transforms/LoopVectorize/epilog-vectorization-any-of-reductions.ll b/llvm/test/Transforms/LoopVectorize/epilog-vectorization-any-of-reductions.ll
index 6e62ff8..c5ecf86b 100644
--- a/llvm/test/Transforms/LoopVectorize/epilog-vectorization-any-of-reductions.ll
+++ b/llvm/test/Transforms/LoopVectorize/epilog-vectorization-any-of-reductions.ll
@@ -21,8 +21,7 @@ define i32 @any_of_reduction_epilog(ptr %src, i64 %N) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq <4 x i8> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI]], [[TMP4]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -51,8 +50,7 @@ define i32 @any_of_reduction_epilog(ptr %src, i64 %N) {
 ; CHECK-NEXT:    [[INDEX5:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT8:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI6:%.*]] = phi <4 x i1> [ [[MINMAX_IDENT_SPLAT]], [[VEC_EPILOG_PH]] ], [ [[TMP14:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[INDEX5]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i8>, ptr [[TMP12]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i8>, ptr [[TMP11]], align 1
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq <4 x i8> [[WIDE_LOAD7]], zeroinitializer
 ; CHECK-NEXT:    [[TMP14]] = or <4 x i1> [[VEC_PHI6]], [[TMP13]]
 ; CHECK-NEXT:    [[INDEX_NEXT8]] = add nuw i64 [[INDEX5]], 4
@@ -118,8 +116,7 @@ define i32 @any_of_reduction_epilog_arg_as_start_value(ptr %src, i64 %N, i32 %st
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq <4 x i8> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI]], [[TMP4]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -148,8 +145,7 @@ define i32 @any_of_reduction_epilog_arg_as_start_value(ptr %src, i64 %N, i32 %st
 ; CHECK-NEXT:    [[INDEX5:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT8:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI6:%.*]] = phi <4 x i1> [ [[MINMAX_IDENT_SPLAT]], [[VEC_EPILOG_PH]] ], [ [[TMP14:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[SRC]], i64 [[INDEX5]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i8>, ptr [[TMP12]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i8>, ptr [[TMP11]], align 1
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq <4 x i8> [[WIDE_LOAD7]], zeroinitializer
 ; CHECK-NEXT:    [[TMP14]] = or <4 x i1> [[VEC_PHI6]], [[TMP13]]
 ; CHECK-NEXT:    [[INDEX_NEXT8]] = add nuw i64 [[INDEX5]], 4
diff --git a/llvm/test/Transforms/LoopVectorize/epilog-vectorization-reductions.ll b/llvm/test/Transforms/LoopVectorize/epilog-vectorization-reductions.ll
index c101d6a..2d0d30d 100644
--- a/llvm/test/Transforms/LoopVectorize/epilog-vectorization-reductions.ll
+++ b/llvm/test/Transforms/LoopVectorize/epilog-vectorization-reductions.ll
@@ -20,8 +20,7 @@ define i64 @int_reduction_add(ptr %a, i64 %N) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ <i64 5, i64 0, i64 0, i64 0>, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3]] = add <4 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -45,8 +44,7 @@ define i64 @int_reduction_add(ptr %a, i64 %N) {
 ; CHECK-NEXT:    [[INDEX4:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT7:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI5:%.*]] = phi <4 x i64> [ [[TMP6]], [[VEC_EPILOG_PH]] ], [ [[TMP10:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX4]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i64>, ptr [[TMP9]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i64>, ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[TMP10]] = add <4 x i64> [[WIDE_LOAD6]], [[VEC_PHI5]]
 ; CHECK-NEXT:    [[INDEX_NEXT7]] = add nuw i64 [[INDEX4]], 4
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT7]], [[N_VEC3]]
@@ -108,8 +106,7 @@ define float @fp_reduction_max(ptr noalias %a, i64 %N) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = fcmp fast ogt <4 x float> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP4]] = select fast <4 x i1> [[TMP3]], <4 x float> [[VEC_PHI]], <4 x float> [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -135,8 +132,7 @@ define float @fp_reduction_max(ptr noalias %a, i64 %N) {
 ; CHECK-NEXT:    [[INDEX4:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT7:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI5:%.*]] = phi <4 x float> [ [[MINMAX_IDENT_SPLAT]], [[VEC_EPILOG_PH]] ], [ [[TMP11:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX4]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP9]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[TMP10:%.*]] = fcmp fast ogt <4 x float> [[VEC_PHI5]], [[WIDE_LOAD6]]
 ; CHECK-NEXT:    [[TMP11]] = select fast <4 x i1> [[TMP10]], <4 x float> [[VEC_PHI5]], <4 x float> [[WIDE_LOAD6]]
 ; CHECK-NEXT:    [[INDEX_NEXT7]] = add nuw i64 [[INDEX4]], 4
@@ -198,8 +194,7 @@ define i16 @reduction_or_trunc(ptr noalias nocapture %ptr) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = and <4 x i32> [[VEC_PHI]], splat (i32 65535)
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[PTR:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP3]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP2]], align 2
 ; CHECK-NEXT:    [[TMP4:%.*]] = zext <4 x i16> [[WIDE_LOAD]] to <4 x i32>
 ; CHECK-NEXT:    [[TMP5:%.*]] = or <4 x i32> [[TMP1]], [[TMP4]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = trunc <4 x i32> [[TMP5]] to <4 x i16>
@@ -223,8 +218,7 @@ define i16 @reduction_or_trunc(ptr noalias nocapture %ptr) {
 ; CHECK-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x i32> [ [[TMP12]], [[VEC_EPILOG_PH]] ], [ [[TMP20:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP14:%.*]] = and <4 x i32> [[VEC_PHI2]], splat (i32 65535)
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i16, ptr [[PTR]], i32 [[INDEX1]]
-; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i16, ptr [[TMP15]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i16>, ptr [[TMP16]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i16>, ptr [[TMP15]], align 2
 ; CHECK-NEXT:    [[TMP17:%.*]] = zext <4 x i16> [[WIDE_LOAD3]] to <4 x i32>
 ; CHECK-NEXT:    [[TMP18:%.*]] = or <4 x i32> [[TMP14]], [[TMP17]]
 ; CHECK-NEXT:    [[TMP19:%.*]] = trunc <4 x i32> [[TMP18]] to <4 x i16>
@@ -296,8 +290,7 @@ define float @multiple_fp_rdx(ptr %A, i64 %N) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ <float 1.500000e+01, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x float> [ <float 1.000000e+01, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00>, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3]] = fadd fast <4 x float> [[VEC_PHI2]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP4]] = fmul fast <4 x float> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -326,8 +319,7 @@ define float @multiple_fp_rdx(ptr %A, i64 %N) {
 ; CHECK-NEXT:    [[VEC_PHI7:%.*]] = phi <4 x float> [ [[TMP8]], [[VEC_EPILOG_PH]] ], [ [[TMP14:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI8:%.*]] = phi <4 x float> [ [[TMP9]], [[VEC_EPILOG_PH]] ], [ [[TMP13:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX6]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x float>, ptr [[TMP12]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x float>, ptr [[TMP11]], align 4
 ; CHECK-NEXT:    [[TMP13]] = fadd fast <4 x float> [[VEC_PHI8]], [[WIDE_LOAD9]]
 ; CHECK-NEXT:    [[TMP14]] = fmul fast <4 x float> [[VEC_PHI7]], [[WIDE_LOAD9]]
 ; CHECK-NEXT:    [[INDEX_NEXT10]] = add nuw i64 [[INDEX6]], 4
@@ -404,8 +396,7 @@ define i32 @reduction_phi_start_val(ptr %A, i64 %N) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ [[TMP0]], [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP4]] = sub <4 x i32> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -429,8 +420,7 @@ define i32 @reduction_phi_start_val(ptr %A, i64 %N) {
 ; CHECK-NEXT:    [[INDEX4:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT7:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI5:%.*]] = phi <4 x i32> [ [[TMP7]], [[VEC_EPILOG_PH]] ], [ [[TMP11:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX4]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i32>, ptr [[TMP10]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i32>, ptr [[TMP9]], align 4
 ; CHECK-NEXT:    [[TMP11]] = sub <4 x i32> [[VEC_PHI5]], [[WIDE_LOAD6]]
 ; CHECK-NEXT:    [[INDEX_NEXT7]] = add nuw i64 [[INDEX4]], 4
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT7]], [[N_VEC3]]
diff --git a/llvm/test/Transforms/LoopVectorize/epilog-vectorization-trunc-induction-steps.ll b/llvm/test/Transforms/LoopVectorize/epilog-vectorization-trunc-induction-steps.ll
index 587dd88..6384343 100644
--- a/llvm/test/Transforms/LoopVectorize/epilog-vectorization-trunc-induction-steps.ll
+++ b/llvm/test/Transforms/LoopVectorize/epilog-vectorization-trunc-induction-steps.ll
@@ -27,10 +27,9 @@ define void @trunc_iv_steps_with_epilogue(ptr %A, i64 %N) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = trunc i64 [[INDEX]] to i32
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[A]], i32 [[TMP5]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP8]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP7]], align 1
 ; CHECK-NEXT:    [[TMP9:%.*]] = add <4 x i8> [[WIDE_LOAD]], splat (i8 2)
-; CHECK-NEXT:    store <4 x i8> [[TMP9]], ptr [[TMP8]], align 1
+; CHECK-NEXT:    store <4 x i8> [[TMP9]], ptr [[TMP7]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -50,10 +49,9 @@ define void @trunc_iv_steps_with_epilogue(ptr %A, i64 %N) {
 ; CHECK-NEXT:    [[INDEX5:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT7:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP11:%.*]] = trunc i64 [[INDEX5]] to i32
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr i8, ptr [[A]], i32 [[TMP11]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr i8, ptr [[TMP13]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i8>, ptr [[TMP14]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i8>, ptr [[TMP13]], align 1
 ; CHECK-NEXT:    [[TMP15:%.*]] = add <4 x i8> [[WIDE_LOAD6]], splat (i8 2)
-; CHECK-NEXT:    store <4 x i8> [[TMP15]], ptr [[TMP14]], align 1
+; CHECK-NEXT:    store <4 x i8> [[TMP15]], ptr [[TMP13]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT7]] = add nuw i64 [[INDEX5]], 4
 ; CHECK-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT7]], [[N_VEC3]]
 ; CHECK-NEXT:    br i1 [[TMP16]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/expand-scev-after-invoke.ll b/llvm/test/Transforms/LoopVectorize/expand-scev-after-invoke.ll
index 02e92cd..8556ceb 100644
--- a/llvm/test/Transforms/LoopVectorize/expand-scev-after-invoke.ll
+++ b/llvm/test/Transforms/LoopVectorize/expand-scev-after-invoke.ll
@@ -28,16 +28,15 @@ define void @test(ptr %dst) personality ptr null {
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ [[INDUCTION]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[STEP_ADD:%.*]] = add <4 x i32> [[VEC_IND]], [[TMP1]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 4
-; CHECK-NEXT:    store <4 x i32> [[VEC_IND]], ptr [[TMP5]], align 8
+; CHECK-NEXT:    store <4 x i32> [[VEC_IND]], ptr [[TMP4]], align 8
 ; CHECK-NEXT:    store <4 x i32> [[STEP_ADD]], ptr [[TMP6]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[STEP_ADD]], [[TMP1]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 160
 ; CHECK-NEXT:    br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 160, %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[TMP0]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ]
@@ -50,7 +49,7 @@ define void @test(ptr %dst) personality ptr null {
 ; CHECK-NEXT:    store i32 [[IV_2]], ptr [[GEP]], align 8
 ; CHECK-NEXT:    [[IV_2_NEXT]] = add i32 [[IV_2]], [[STEP]]
 ; CHECK-NEXT:    [[EC:%.*]] = icmp ult i64 [[IV_1]], 161
-; CHECK-NEXT:    br i1 [[EC]], label %[[LOOP]], label %[[EXIT]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label %[[LOOP]], label %[[EXIT:.*]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    ret void
 ; CHECK:       [[LPAD]]:
diff --git a/llvm/test/Transforms/LoopVectorize/extract-from-end-vector-constant.ll b/llvm/test/Transforms/LoopVectorize/extract-from-end-vector-constant.ll
index f1e3ef0..eb0145e 100644
--- a/llvm/test/Transforms/LoopVectorize/extract-from-end-vector-constant.ll
+++ b/llvm/test/Transforms/LoopVectorize/extract-from-end-vector-constant.ll
@@ -13,13 +13,12 @@ define i64 @exit_value_scalar_live_in(ptr %dst, i64 %in) {
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i16, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i16, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <4 x i16> zeroinitializer, ptr [[TMP2]], align 2
+; CHECK-NEXT:    store <4 x i16> zeroinitializer, ptr [[TMP1]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
@@ -29,9 +28,9 @@ define i64 @exit_value_scalar_live_in(ptr %dst, i64 %in) {
 ; CHECK-NEXT:    store i16 0, ptr [[GEP]], align 2
 ; CHECK-NEXT:    [[IV_NEXT_1]] = add i64 [[IV]], 1
 ; CHECK-NEXT:    [[EC:%.*]] = icmp ult i64 [[IV]], 1000
-; CHECK-NEXT:    br i1 [[EC]], label %[[LOOP]], label %[[EXIT]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label %[[LOOP]], label %[[EXIT:.*]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       [[EXIT]]:
-; CHECK-NEXT:    [[RES:%.*]] = phi i64 [ [[IN]], %[[LOOP]] ], [ [[IN]], %[[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[RES:%.*]] = phi i64 [ [[IN]], %[[LOOP]] ]
 ; CHECK-NEXT:    ret i64 [[RES]]
 ;
 entry:
@@ -60,13 +59,12 @@ define <2 x i64> @exit_value_vector_live_in(ptr %dst) {
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i16, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i16, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <4 x i16> zeroinitializer, ptr [[TMP2]], align 2
+; CHECK-NEXT:    store <4 x i16> zeroinitializer, ptr [[TMP1]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
@@ -76,9 +74,9 @@ define <2 x i64> @exit_value_vector_live_in(ptr %dst) {
 ; CHECK-NEXT:    store i16 0, ptr [[GEP]], align 2
 ; CHECK-NEXT:    [[IV_NEXT_1]] = add i64 [[IV]], 1
 ; CHECK-NEXT:    [[EC:%.*]] = icmp ult i64 [[IV]], 1000
-; CHECK-NEXT:    br i1 [[EC]], label %[[LOOP]], label %[[EXIT]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label %[[LOOP]], label %[[EXIT:.*]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       [[EXIT]]:
-; CHECK-NEXT:    [[RES:%.*]] = phi <2 x i64> [ zeroinitializer, %[[LOOP]] ], [ zeroinitializer, %[[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[RES:%.*]] = phi <2 x i64> [ zeroinitializer, %[[LOOP]] ]
 ; CHECK-NEXT:    ret <2 x i64> [[RES]]
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/first-order-recurrence-chains.ll b/llvm/test/Transforms/LoopVectorize/first-order-recurrence-chains.ll
index 03c334b..d3c8c13 100644
--- a/llvm/test/Transforms/LoopVectorize/first-order-recurrence-chains.ll
+++ b/llvm/test/Transforms/LoopVectorize/first-order-recurrence-chains.ll
@@ -1,37 +1,40 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 5
 ; RUN: opt -passes=loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S %s | FileCheck %s
 
-define i16 @test_chained_first_order_recurrences_1(ptr %ptr) {
+define i16 @test_chained_first_order_recurrences_1(ptr %ptr, i64 %n) {
 ; CHECK-LABEL: define i16 @test_chained_first_order_recurrences_1(
-; CHECK-SAME: ptr [[PTR:%.*]]) {
+; CHECK-SAME: ptr [[PTR:%.*]], i64 [[N:%.*]]) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
-; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 22>, %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VECTOR_RECUR1:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 33>, %[[VECTOR_PH]] ], [ [[TMP2:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i16, ptr [[PTR]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP1]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP0]], align 2
 ; CHECK-NEXT:    [[TMP2]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR1]], <4 x i16> [[TMP2]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP4:%.*]] = add <4 x i16> [[TMP2]], [[TMP3]]
-; CHECK-NEXT:    store <4 x i16> [[TMP4]], ptr [[TMP1]], align 2
+; CHECK-NEXT:    store <4 x i16> [[TMP4]], ptr [[TMP0]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT3:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT2:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 2
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI3:%.*]] = extractelement <4 x i16> [[TMP2]], i32 2
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 22, %[[ENTRY]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT4:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT3]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT4:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT2]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[FOR_1:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], %[[SCALAR_PH]] ], [ [[FOR_1_NEXT:%.*]], %[[LOOP]] ]
@@ -42,7 +45,7 @@ define i16 @test_chained_first_order_recurrences_1(ptr %ptr) {
 ; CHECK-NEXT:    [[FOR_1_NEXT]] = load i16, ptr [[GEP_PTR]], align 2
 ; CHECK-NEXT:    [[ADD:%.*]] = add i16 [[FOR_1]], [[FOR_2]]
 ; CHECK-NEXT:    store i16 [[ADD]], ptr [[GEP_PTR]], align 2
-; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    [[FOR_1_LCSSA:%.*]] = phi i16 [ [[FOR_1]], %[[LOOP]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ]
@@ -62,7 +65,7 @@ loop:
   %for.1.next = load i16, ptr %gep.ptr, align 2
   %add = add i16 %for.1, %for.2
   store i16 %add, ptr %gep.ptr
-  %exitcond.not = icmp eq i64 %iv.next, 1000
+  %exitcond.not = icmp eq i64 %iv.next, %n
   br i1 %exitcond.not, label %exit, label %loop
 
 exit:
@@ -70,37 +73,40 @@ exit:
   ret i16 %res
 }
 
-define i16 @test_chained_first_order_recurrences_2(ptr %ptr) {
+define i16 @test_chained_first_order_recurrences_2(ptr %ptr, i64 %n) {
 ; CHECK-LABEL: define i16 @test_chained_first_order_recurrences_2(
-; CHECK-SAME: ptr [[PTR:%.*]]) {
+; CHECK-SAME: ptr [[PTR:%.*]], i64 [[N:%.*]]) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
-; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 33>, %[[VECTOR_PH]] ], [ [[TMP2:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VECTOR_RECUR1:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 22>, %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i16, ptr [[PTR]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP1]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP0]], align 2
 ; CHECK-NEXT:    [[TMP2]] = shufflevector <4 x i16> [[VECTOR_RECUR1]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[TMP2]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP4:%.*]] = add <4 x i16> [[TMP2]], [[TMP3]]
-; CHECK-NEXT:    store <4 x i16> [[TMP4]], ptr [[TMP1]], align 2
+; CHECK-NEXT:    store <4 x i16> [[TMP4]], ptr [[TMP0]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT3:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT2:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[TMP2]], i32 2
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI3:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 2
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT4:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT3]], %[[MIDDLE_BLOCK]] ], [ 22, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT4:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT2]], %[[MIDDLE_BLOCK]] ], [ 22, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[FOR_2:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], %[[SCALAR_PH]] ], [ [[FOR_1:%.*]], %[[LOOP]] ]
@@ -111,7 +117,7 @@ define i16 @test_chained_first_order_recurrences_2(ptr %ptr) {
 ; CHECK-NEXT:    [[FOR_1_NEXT]] = load i16, ptr [[GEP_PTR]], align 2
 ; CHECK-NEXT:    [[ADD:%.*]] = add i16 [[FOR_1]], [[FOR_2]]
 ; CHECK-NEXT:    store i16 [[ADD]], ptr [[GEP_PTR]], align 2
-; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    [[FOR_2_LCSSA:%.*]] = phi i16 [ [[FOR_2]], %[[LOOP]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ]
@@ -131,7 +137,7 @@ loop:
   %for.1.next = load i16, ptr %gep.ptr, align 2
   %add = add i16 %for.1, %for.2
   store i16 %add, ptr %gep.ptr
-  %exitcond.not = icmp eq i64 %iv.next, 1000
+  %exitcond.not = icmp eq i64 %iv.next, %n
   br i1 %exitcond.not, label %exit, label %loop
 
 exit:
@@ -139,12 +145,15 @@ exit:
   ret i16 %res
 }
 
-define i16 @test_chained_first_order_recurrences_3(ptr %ptr) {
+define i16 @test_chained_first_order_recurrences_3(ptr %ptr, i64 %n) {
 ; CHECK-LABEL: define i16 @test_chained_first_order_recurrences_3(
-; CHECK-SAME: ptr [[PTR:%.*]]) {
+; CHECK-SAME: ptr [[PTR:%.*]], i64 [[N:%.*]]) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
-; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
@@ -152,30 +161,30 @@ define i16 @test_chained_first_order_recurrences_3(ptr %ptr) {
 ; CHECK-NEXT:    [[VECTOR_RECUR1:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 33>, %[[VECTOR_PH]] ], [ [[TMP2:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VECTOR_RECUR2:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 33>, %[[VECTOR_PH]] ], [ [[TMP3:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i16, ptr [[PTR]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP1]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP0]], align 2
 ; CHECK-NEXT:    [[TMP2]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP3]] = shufflevector <4 x i16> [[VECTOR_RECUR1]], <4 x i16> [[TMP2]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR2]], <4 x i16> [[TMP3]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP5:%.*]] = add <4 x i16> [[TMP2]], [[TMP3]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = add <4 x i16> [[TMP5]], [[TMP4]]
-; CHECK-NEXT:    store <4 x i16> [[TMP6]], ptr [[TMP1]], align 2
+; CHECK-NEXT:    store <4 x i16> [[TMP6]], ptr [[TMP0]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT5:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT6:%.*]] = extractelement <4 x i16> [[TMP3]], i32 3
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT3:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT4:%.*]] = extractelement <4 x i16> [[TMP3]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 2
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI5:%.*]] = extractelement <4 x i16> [[TMP2]], i32 2
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI6:%.*]] = extractelement <4 x i16> [[TMP3]], i32 2
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 22, %[[ENTRY]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT7:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT5]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT8:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT6]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT7:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT3]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT8:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT4]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[FOR_1:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], %[[SCALAR_PH]] ], [ [[FOR_1_NEXT:%.*]], %[[LOOP]] ]
@@ -188,7 +197,7 @@ define i16 @test_chained_first_order_recurrences_3(ptr %ptr) {
 ; CHECK-NEXT:    [[ADD_1:%.*]] = add i16 [[FOR_1]], [[FOR_2]]
 ; CHECK-NEXT:    [[ADD_2:%.*]] = add i16 [[ADD_1]], [[FOR_3]]
 ; CHECK-NEXT:    store i16 [[ADD_2]], ptr [[GEP_PTR]], align 2
-; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP7:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    [[FOR_1_LCSSA:%.*]] = phi i16 [ [[FOR_1]], %[[LOOP]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ]
@@ -212,7 +221,7 @@ loop:
   %add.1 = add i16 %for.1, %for.2
   %add.2 = add i16 %add.1, %for.3
   store i16 %add.2, ptr %gep.ptr
-  %exitcond.not = icmp eq i64 %iv.next, 1000
+  %exitcond.not = icmp eq i64 %iv.next, %n
   br i1 %exitcond.not, label %exit, label %loop
 
 exit:
@@ -259,35 +268,38 @@ exit:
   ret void
 }
 
-define void @test_first_order_recurrences_incoming_cycle_preheader(ptr %ptr) {
+define void @test_first_order_recurrences_incoming_cycle_preheader(ptr %ptr, i64 %n) {
 ; CHECK-LABEL: define void @test_first_order_recurrences_incoming_cycle_preheader(
-; CHECK-SAME: ptr [[PTR:%.*]]) {
+; CHECK-SAME: ptr [[PTR:%.*]], i64 [[N:%.*]]) {
 ; CHECK-NEXT:  [[ENTRY:.*:]]
 ; CHECK-NEXT:    br label %[[LOOP_1:.*]]
 ; CHECK:       [[LOOP_1]]:
 ; CHECK-NEXT:    br i1 true, label %[[LOOP_PREHEADER:.*]], label %[[LOOP_1]]
 ; CHECK:       [[LOOP_PREHEADER]]:
-; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 0>, %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i16, ptr [[PTR]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP1]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP0]], align 2
 ; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP3:%.*]] = add <4 x i16> [[TMP2]], splat (i16 10)
-; CHECK-NEXT:    store <4 x i16> [[TMP3]], ptr [[TMP1]], align 2
+; CHECK-NEXT:    store <4 x i16> [[TMP3]], ptr [[TMP0]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_PREHEADER]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[FOR_1:%.*]] = phi i16 [ [[FOR_1_NEXT:%.*]], %[[LOOP]] ], [ [[SCALAR_RECUR_INIT]], %[[SCALAR_PH]] ]
@@ -297,7 +309,7 @@ define void @test_first_order_recurrences_incoming_cycle_preheader(ptr %ptr) {
 ; CHECK-NEXT:    [[FOR_1_NEXT]] = load i16, ptr [[GEP_PTR]], align 2
 ; CHECK-NEXT:    [[ADD:%.*]] = add i16 [[FOR_1]], 10
 ; CHECK-NEXT:    store i16 [[ADD]], ptr [[GEP_PTR]], align 2
-; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP9:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    ret void
@@ -317,19 +329,22 @@ loop:
   %for.1.next = load i16, ptr %gep.ptr, align 2
   %add = add i16 %for.1, 10
   store i16 %add, ptr %gep.ptr
-  %exitcond.not = icmp eq i64 %iv.next, 1000
+  %exitcond.not = icmp eq i64 %iv.next, %n
   br i1 %exitcond.not, label %exit, label %loop
 
 exit:
   ret void
 }
 
-define i16 @test_chained_first_order_recurrences_3_reordered_1(ptr %ptr) {
+define i16 @test_chained_first_order_recurrences_3_reordered_1(ptr %ptr, i64 %n) {
 ; CHECK-LABEL: define i16 @test_chained_first_order_recurrences_3_reordered_1(
-; CHECK-SAME: ptr [[PTR:%.*]]) {
+; CHECK-SAME: ptr [[PTR:%.*]], i64 [[N:%.*]]) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
-; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
@@ -337,30 +352,30 @@ define i16 @test_chained_first_order_recurrences_3_reordered_1(ptr %ptr) {
 ; CHECK-NEXT:    [[VECTOR_RECUR1:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 33>, %[[VECTOR_PH]] ], [ [[TMP2:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VECTOR_RECUR2:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 22>, %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i16, ptr [[PTR]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP1]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP0]], align 2
 ; CHECK-NEXT:    [[TMP2]] = shufflevector <4 x i16> [[VECTOR_RECUR2]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP3]] = shufflevector <4 x i16> [[VECTOR_RECUR1]], <4 x i16> [[TMP2]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[TMP3]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP5:%.*]] = add <4 x i16> [[TMP2]], [[TMP3]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = add <4 x i16> [[TMP5]], [[TMP4]]
-; CHECK-NEXT:    store <4 x i16> [[TMP6]], ptr [[TMP1]], align 2
+; CHECK-NEXT:    store <4 x i16> [[TMP6]], ptr [[TMP0]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[TMP3]], i32 3
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT5:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT6:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT3:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT4:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[TMP3]], i32 2
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI5:%.*]] = extractelement <4 x i16> [[TMP2]], i32 2
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI6:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 2
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT7:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT5]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT8:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT6]], %[[MIDDLE_BLOCK]] ], [ 22, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT7:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT3]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT8:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT4]], %[[MIDDLE_BLOCK]] ], [ 22, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[FOR_3:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], %[[SCALAR_PH]] ], [ [[FOR_2:%.*]], %[[LOOP]] ]
@@ -373,7 +388,7 @@ define i16 @test_chained_first_order_recurrences_3_reordered_1(ptr %ptr) {
 ; CHECK-NEXT:    [[ADD_1:%.*]] = add i16 [[FOR_1]], [[FOR_2]]
 ; CHECK-NEXT:    [[ADD_2:%.*]] = add i16 [[ADD_1]], [[FOR_3]]
 ; CHECK-NEXT:    store i16 [[ADD_2]], ptr [[GEP_PTR]], align 2
-; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP11:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    [[FOR_3_LCSSA:%.*]] = phi i16 [ [[FOR_3]], %[[LOOP]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ]
@@ -397,7 +412,7 @@ loop:
   %add.1 = add i16 %for.1, %for.2
   %add.2 = add i16 %add.1, %for.3
   store i16 %add.2, ptr %gep.ptr
-  %exitcond.not = icmp eq i64 %iv.next, 1000
+  %exitcond.not = icmp eq i64 %iv.next, %n
   br i1 %exitcond.not, label %exit, label %loop
 
 exit:
@@ -406,12 +421,15 @@ exit:
   ret i16 %res.2
 }
 
-define i16 @test_chained_first_order_recurrences_3_reordered_2(ptr %ptr) {
+define i16 @test_chained_first_order_recurrences_3_reordered_2(ptr %ptr, i64 %n) {
 ; CHECK-LABEL: define i16 @test_chained_first_order_recurrences_3_reordered_2(
-; CHECK-SAME: ptr [[PTR:%.*]]) {
+; CHECK-SAME: ptr [[PTR:%.*]], i64 [[N:%.*]]) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
-; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
@@ -419,30 +437,30 @@ define i16 @test_chained_first_order_recurrences_3_reordered_2(ptr %ptr) {
 ; CHECK-NEXT:    [[VECTOR_RECUR1:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 33>, %[[VECTOR_PH]] ], [ [[TMP3:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VECTOR_RECUR2:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 22>, %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i16, ptr [[PTR]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP1]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP0]], align 2
 ; CHECK-NEXT:    [[TMP2]] = shufflevector <4 x i16> [[VECTOR_RECUR2]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP3]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[TMP2]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR1]], <4 x i16> [[TMP3]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP5:%.*]] = add <4 x i16> [[TMP2]], [[TMP3]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = add <4 x i16> [[TMP5]], [[TMP4]]
-; CHECK-NEXT:    store <4 x i16> [[TMP6]], ptr [[TMP1]], align 2
+; CHECK-NEXT:    store <4 x i16> [[TMP6]], ptr [[TMP0]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT5:%.*]] = extractelement <4 x i16> [[TMP3]], i32 3
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT6:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT3:%.*]] = extractelement <4 x i16> [[TMP3]], i32 3
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT4:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[TMP2]], i32 2
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI5:%.*]] = extractelement <4 x i16> [[TMP3]], i32 2
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI6:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 2
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT7:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT5]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT8:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT6]], %[[MIDDLE_BLOCK]] ], [ 22, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT7:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT3]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT8:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT4]], %[[MIDDLE_BLOCK]] ], [ 22, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[FOR_2:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], %[[SCALAR_PH]] ], [ [[FOR_1:%.*]], %[[LOOP]] ]
@@ -455,7 +473,7 @@ define i16 @test_chained_first_order_recurrences_3_reordered_2(ptr %ptr) {
 ; CHECK-NEXT:    [[ADD_1:%.*]] = add i16 [[FOR_1]], [[FOR_2]]
 ; CHECK-NEXT:    [[ADD_2:%.*]] = add i16 [[ADD_1]], [[FOR_3]]
 ; CHECK-NEXT:    store i16 [[ADD_2]], ptr [[GEP_PTR]], align 2
-; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP13:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    [[FOR_2_LCSSA:%.*]] = phi i16 [ [[FOR_2]], %[[LOOP]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ]
@@ -479,7 +497,7 @@ loop:
   %add.1 = add i16 %for.1, %for.2
   %add.2 = add i16 %add.1, %for.3
   store i16 %add.2, ptr %gep.ptr
-  %exitcond.not = icmp eq i64 %iv.next, 1000
+  %exitcond.not = icmp eq i64 %iv.next, %n
   br i1 %exitcond.not, label %exit, label %loop
 
 exit:
@@ -488,12 +506,15 @@ exit:
   ret i16 %res.2
 }
 
-define i16 @test_chained_first_order_recurrences_3_for2_no_other_uses(ptr %ptr) {
+define i16 @test_chained_first_order_recurrences_3_for2_no_other_uses(ptr %ptr, i64 %n) {
 ; CHECK-LABEL: define i16 @test_chained_first_order_recurrences_3_for2_no_other_uses(
-; CHECK-SAME: ptr [[PTR:%.*]]) {
+; CHECK-SAME: ptr [[PTR:%.*]], i64 [[N:%.*]]) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
-; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
@@ -501,30 +522,30 @@ define i16 @test_chained_first_order_recurrences_3_for2_no_other_uses(ptr %ptr)
 ; CHECK-NEXT:    [[VECTOR_RECUR1:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 33>, %[[VECTOR_PH]] ], [ [[TMP2:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VECTOR_RECUR2:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 33>, %[[VECTOR_PH]] ], [ [[TMP3:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i16, ptr [[PTR]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP1]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP0]], align 2
 ; CHECK-NEXT:    [[TMP2]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP3]] = shufflevector <4 x i16> [[VECTOR_RECUR1]], <4 x i16> [[TMP2]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR2]], <4 x i16> [[TMP3]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP5:%.*]] = add <4 x i16> [[TMP2]], splat (i16 10)
 ; CHECK-NEXT:    [[TMP6:%.*]] = add <4 x i16> [[TMP5]], [[TMP4]]
-; CHECK-NEXT:    store <4 x i16> [[TMP6]], ptr [[TMP1]], align 2
+; CHECK-NEXT:    store <4 x i16> [[TMP6]], ptr [[TMP0]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT5:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT6:%.*]] = extractelement <4 x i16> [[TMP3]], i32 3
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT3:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT4:%.*]] = extractelement <4 x i16> [[TMP3]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 2
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI5:%.*]] = extractelement <4 x i16> [[TMP2]], i32 2
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI6:%.*]] = extractelement <4 x i16> [[TMP3]], i32 2
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 22, %[[ENTRY]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT7:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT5]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT8:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT6]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT7:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT3]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT8:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT4]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[FOR_1:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], %[[SCALAR_PH]] ], [ [[FOR_1_NEXT:%.*]], %[[LOOP]] ]
@@ -537,7 +558,7 @@ define i16 @test_chained_first_order_recurrences_3_for2_no_other_uses(ptr %ptr)
 ; CHECK-NEXT:    [[ADD_1:%.*]] = add i16 [[FOR_1]], 10
 ; CHECK-NEXT:    [[ADD_2:%.*]] = add i16 [[ADD_1]], [[FOR_3]]
 ; CHECK-NEXT:    store i16 [[ADD_2]], ptr [[GEP_PTR]], align 2
-; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP15:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    [[FOR_1_LCSSA:%.*]] = phi i16 [ [[FOR_1]], %[[LOOP]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ]
@@ -561,7 +582,7 @@ loop:
   %add.1 = add i16 %for.1, 10
   %add.2 = add i16 %add.1, %for.3
   store i16 %add.2, ptr %gep.ptr
-  %exitcond.not = icmp eq i64 %iv.next, 1000
+  %exitcond.not = icmp eq i64 %iv.next, %n
   br i1 %exitcond.not, label %exit, label %loop
 
 exit:
@@ -570,12 +591,15 @@ exit:
   ret i16 %res.2
 }
 
-define i16 @test_chained_first_order_recurrences_3_for1_for2_no_other_uses(ptr %ptr) {
+define i16 @test_chained_first_order_recurrences_3_for1_for2_no_other_uses(ptr %ptr, i64 %n) {
 ; CHECK-LABEL: define i16 @test_chained_first_order_recurrences_3_for1_for2_no_other_uses(
-; CHECK-SAME: ptr [[PTR:%.*]]) {
+; CHECK-SAME: ptr [[PTR:%.*]], i64 [[N:%.*]]) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
-; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
@@ -583,29 +607,29 @@ define i16 @test_chained_first_order_recurrences_3_for1_for2_no_other_uses(ptr %
 ; CHECK-NEXT:    [[VECTOR_RECUR1:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 33>, %[[VECTOR_PH]] ], [ [[TMP2:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VECTOR_RECUR2:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 33>, %[[VECTOR_PH]] ], [ [[TMP3:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i16, ptr [[PTR]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP1]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP0]], align 2
 ; CHECK-NEXT:    [[TMP2]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP3]] = shufflevector <4 x i16> [[VECTOR_RECUR1]], <4 x i16> [[TMP2]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR2]], <4 x i16> [[TMP3]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP5:%.*]] = add <4 x i16> [[TMP4]], splat (i16 10)
-; CHECK-NEXT:    store <4 x i16> [[TMP5]], ptr [[TMP1]], align 2
+; CHECK-NEXT:    store <4 x i16> [[TMP5]], ptr [[TMP0]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 3
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT5:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT6:%.*]] = extractelement <4 x i16> [[TMP3]], i32 3
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT3:%.*]] = extractelement <4 x i16> [[TMP2]], i32 3
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT4:%.*]] = extractelement <4 x i16> [[TMP3]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 2
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI5:%.*]] = extractelement <4 x i16> [[TMP2]], i32 2
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI6:%.*]] = extractelement <4 x i16> [[TMP3]], i32 2
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 22, %[[ENTRY]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT7:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT5]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT8:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT6]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT7:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT3]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT8:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT4]], %[[MIDDLE_BLOCK]] ], [ 33, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[FOR_1:%.*]] = phi i16 [ [[SCALAR_RECUR_INIT]], %[[SCALAR_PH]] ], [ [[FOR_1_NEXT:%.*]], %[[LOOP]] ]
@@ -617,7 +641,7 @@ define i16 @test_chained_first_order_recurrences_3_for1_for2_no_other_uses(ptr %
 ; CHECK-NEXT:    [[FOR_1_NEXT]] = load i16, ptr [[GEP_PTR]], align 2
 ; CHECK-NEXT:    [[ADD_1:%.*]] = add i16 [[FOR_3]], 10
 ; CHECK-NEXT:    store i16 [[ADD_1]], ptr [[GEP_PTR]], align 2
-; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP17:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    [[FOR_1_LCSSA:%.*]] = phi i16 [ [[FOR_1]], %[[LOOP]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ]
@@ -640,7 +664,7 @@ loop:
   %for.1.next = load i16, ptr %gep.ptr, align 2
   %add.1 = add i16 %for.3, 10
   store i16 %add.1, ptr %gep.ptr
-  %exitcond.not = icmp eq i64 %iv.next, 1000
+  %exitcond.not = icmp eq i64 %iv.next, %n
   br i1 %exitcond.not, label %exit, label %loop
 
 exit:
@@ -649,57 +673,51 @@ exit:
   ret i16 %res.2
 }
 
-define double @test_chained_first_order_recurrence_sink_users_1(ptr %ptr) {
+define double @test_chained_first_order_recurrence_sink_users_1(ptr %ptr, i64 %n) {
 ; CHECK-LABEL: define double @test_chained_first_order_recurrence_sink_users_1(
-; CHECK-SAME: ptr [[PTR:%.*]]) {
+; CHECK-SAME: ptr [[PTR:%.*]], i64 [[N:%.*]]) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
-; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK-NEXT:    [[TMP6:%.*]] = add i64 [[N]], -1
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP6]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP6]], 4
+; CHECK-NEXT:    [[INDEX:%.*]] = sub i64 [[TMP6]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x double> [ <double poison, double poison, double poison, double 1.000000e+01>, %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VECTOR_RECUR1:%.*]] = phi <4 x double> [ <double poison, double poison, double poison, double 2.000000e+01>, %[[VECTOR_PH]] ], [ [[TMP2:%.*]], %[[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
-; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds double, ptr [[PTR]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds double, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x double>, ptr [[TMP1]], align 8
-; CHECK-NEXT:    [[TMP2]] = shufflevector <4 x double> [[VECTOR_RECUR]], <4 x double> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x double> [[VECTOR_RECUR1]], <4 x double> [[TMP2]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; CHECK-NEXT:    [[TMP4:%.*]] = fadd <4 x double> splat (double 1.000000e+01), [[TMP3]]
-; CHECK-NEXT:    [[TMP5:%.*]] = fadd <4 x double> [[TMP4]], [[TMP2]]
-; CHECK-NEXT:    store <4 x double> [[TMP5]], ptr [[TMP1]], align 8
-; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 996
-; CHECK-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x double> [[WIDE_LOAD]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT3:%.*]] = extractelement <4 x double> [[TMP2]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x double> [[WIDE_LOAD]], i32 2
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI3:%.*]] = extractelement <4 x double> [[TMP2]], i32 2
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP6]], [[INDEX]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi double [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 1.000000e+01, %[[ENTRY]] ]
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT4:%.*]] = phi double [ [[VECTOR_RECUR_EXTRACT3]], %[[MIDDLE_BLOCK]] ], [ 2.000000e+01, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 997, %[[MIDDLE_BLOCK]] ], [ 1, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[OFFSET_IDX]], %[[MIDDLE_BLOCK]] ], [ 1, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
-; CHECK-NEXT:    [[FOR_1:%.*]] = phi double [ [[SCALAR_RECUR_INIT]], %[[SCALAR_PH]] ], [ [[FOR_1_NEXT:%.*]], %[[LOOP]] ]
-; CHECK-NEXT:    [[FOR_2:%.*]] = phi double [ [[SCALAR_RECUR_INIT4]], %[[SCALAR_PH]] ], [ [[FOR_1]], %[[LOOP]] ]
+; CHECK-NEXT:    [[FOR_2_LCSSA:%.*]] = phi double [ [[SCALAR_RECUR_INIT]], %[[SCALAR_PH]] ], [ [[FOR_1_NEXT:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[FOR_2:%.*]] = phi double [ [[SCALAR_RECUR_INIT4]], %[[SCALAR_PH]] ], [ [[FOR_2_LCSSA]], %[[LOOP]] ]
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
-; CHECK-NEXT:    [[ADD_1:%.*]] = fadd double 1.000000e+01, [[FOR_2]]
-; CHECK-NEXT:    [[ADD_2:%.*]] = fadd double [[ADD_1]], [[FOR_1]]
+; CHECK-NEXT:    [[FOR_1_LCSSA:%.*]] = fadd double 1.000000e+01, [[FOR_2]]
+; CHECK-NEXT:    [[RES:%.*]] = fadd double [[FOR_1_LCSSA]], [[FOR_2_LCSSA]]
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[GEP_PTR:%.*]] = getelementptr inbounds double, ptr [[PTR]], i64 [[IV]]
 ; CHECK-NEXT:    [[FOR_1_NEXT]] = load double, ptr [[GEP_PTR]], align 8
-; CHECK-NEXT:    store double [[ADD_2]], ptr [[GEP_PTR]], align 8
-; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
+; CHECK-NEXT:    store double [[RES]], ptr [[GEP_PTR]], align 8
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP19:![0-9]+]]
 ; CHECK:       [[EXIT]]:
-; CHECK-NEXT:    [[FOR_1_LCSSA:%.*]] = phi double [ [[FOR_1]], %[[LOOP]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ]
-; CHECK-NEXT:    [[FOR_2_LCSSA:%.*]] = phi double [ [[FOR_2]], %[[LOOP]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI3]], %[[MIDDLE_BLOCK]] ]
-; CHECK-NEXT:    [[RES:%.*]] = fadd double [[FOR_1_LCSSA]], [[FOR_2_LCSSA]]
-; CHECK-NEXT:    ret double [[RES]]
+; CHECK-NEXT:    [[FOR_1_LCSSA1:%.*]] = phi double [ [[FOR_2_LCSSA]], %[[LOOP]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[FOR_2_LCSSA1:%.*]] = phi double [ [[FOR_2]], %[[LOOP]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI3]], %[[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[RES1:%.*]] = fadd double [[FOR_1_LCSSA1]], [[FOR_2_LCSSA1]]
+; CHECK-NEXT:    ret double [[RES1]]
 ;
 entry:
   br label %loop
@@ -714,7 +732,7 @@ loop:
   %gep.ptr = getelementptr inbounds double, ptr %ptr, i64 %iv
   %for.1.next  = load double, ptr %gep.ptr, align 8
   store double %add.2, ptr %gep.ptr
-  %exitcond.not = icmp eq i64 %iv.next, 1000
+  %exitcond.not = icmp eq i64 %iv.next, %n
   br i1 %exitcond.not, label %exit, label %loop
 
 exit:
@@ -722,9 +740,9 @@ exit:
   ret double %res
 }
 
-define void @test_first_order_recurrences_and_reduction(ptr %ptr) {
+define void @test_first_order_recurrences_and_reduction(ptr %ptr, i64 %n) {
 ; CHECK-LABEL: define void @test_first_order_recurrences_and_reduction(
-; CHECK-SAME: ptr [[PTR:%.*]]) {
+; CHECK-SAME: ptr [[PTR:%.*]], i64 [[N:%.*]]) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
@@ -738,7 +756,7 @@ define void @test_first_order_recurrences_and_reduction(ptr %ptr) {
 ; CHECK-NEXT:    [[ADD_1:%.*]] = add i16 [[FOR_1]], 10
 ; CHECK-NEXT:    [[RED_NEXT]] = add i16 [[RED]], [[LV]]
 ; CHECK-NEXT:    store i16 [[ADD_1]], ptr [[GEP_PTR]], align 2
-; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[LOOP]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    ret void
@@ -757,19 +775,22 @@ loop:
   %add.1 = add i16 %for.1, 10
   %red.next = add i16 %red, %lv
   store i16 %add.1, ptr %gep.ptr
-  %exitcond.not = icmp eq i64 %iv.next, 1000
+  %exitcond.not = icmp eq i64 %iv.next, %n
   br i1 %exitcond.not, label %exit, label %loop
 
 exit:
   ret void
 }
 
-define i64 @test_first_order_recurrences_and_induction(ptr %ptr) {
+define i64 @test_first_order_recurrences_and_induction(ptr %ptr, i64 %n) {
 ; CHECK-LABEL: define i64 @test_first_order_recurrences_and_induction(
-; CHECK-SAME: ptr [[PTR:%.*]]) {
+; CHECK-SAME: ptr [[PTR:%.*]], i64 [[N:%.*]]) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
-; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
@@ -778,19 +799,19 @@ define i64 @test_first_order_recurrences_and_induction(ptr %ptr) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i64> [[VECTOR_RECUR]], <4 x i64> [[VEC_IND]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[PTR]], i64 [[INDEX]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = add <4 x i64> [[TMP1]], splat (i64 10)
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    store <4 x i64> [[TMP4]], ptr [[TMP3]], align 4
+; CHECK-NEXT:    store <4 x i64> [[TMP4]], ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
-; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i64> [[VEC_IND]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i64> [[VEC_IND]], i32 2
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 22, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[FOR_1:%.*]] = phi i64 [ [[SCALAR_RECUR_INIT]], %[[SCALAR_PH]] ], [ [[IV:%.*]], %[[LOOP]] ]
@@ -799,7 +820,7 @@ define i64 @test_first_order_recurrences_and_induction(ptr %ptr) {
 ; CHECK-NEXT:    [[GEP_PTR:%.*]] = getelementptr inbounds i64, ptr [[PTR]], i64 [[IV]]
 ; CHECK-NEXT:    [[ADD_1:%.*]] = add i64 [[FOR_1]], 10
 ; CHECK-NEXT:    store i64 [[ADD_1]], ptr [[GEP_PTR]], align 4
-; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP21:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    [[FOR_1_LCSSA:%.*]] = phi i64 [ [[FOR_1]], %[[LOOP]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ]
@@ -816,7 +837,7 @@ loop:
   %gep.ptr = getelementptr inbounds i64, ptr %ptr, i64 %iv
   %add.1 = add i64 %for.1, 10
   store i64 %add.1, ptr %gep.ptr
-  %exitcond.not = icmp eq i64 %iv.next, 1000
+  %exitcond.not = icmp eq i64 %iv.next, %n
   br i1 %exitcond.not, label %exit, label %loop
 
 exit:
@@ -825,12 +846,15 @@ exit:
 
 ; Same as @test_first_order_recurrences_and_induction but with order of phis
 ; flipped.
-define i64 @test_first_order_recurrences_and_induction2(ptr %ptr) {
+define i64 @test_first_order_recurrences_and_induction2(ptr %ptr, i64 %n) {
 ; CHECK-LABEL: define i64 @test_first_order_recurrences_and_induction2(
-; CHECK-SAME: ptr [[PTR:%.*]]) {
+; CHECK-SAME: ptr [[PTR:%.*]], i64 [[N:%.*]]) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
-; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
@@ -839,18 +863,18 @@ define i64 @test_first_order_recurrences_and_induction2(ptr %ptr) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i64> [[VECTOR_RECUR]], <4 x i64> [[VEC_IND]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[PTR]], i64 [[INDEX]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = add <4 x i64> [[TMP1]], splat (i64 10)
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    store <4 x i64> [[TMP4]], ptr [[TMP3]], align 4
+; CHECK-NEXT:    store <4 x i64> [[TMP4]], ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
-; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i64> [[VEC_IND]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i64> [[VEC_IND]], i32 2
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 22, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
@@ -860,7 +884,7 @@ define i64 @test_first_order_recurrences_and_induction2(ptr %ptr) {
 ; CHECK-NEXT:    [[GEP_PTR:%.*]] = getelementptr inbounds i64, ptr [[PTR]], i64 [[IV]]
 ; CHECK-NEXT:    [[ADD_1:%.*]] = add i64 [[FOR_1]], 10
 ; CHECK-NEXT:    store i64 [[ADD_1]], ptr [[GEP_PTR]], align 4
-; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP23:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    [[FOR_1_LCSSA:%.*]] = phi i64 [ [[FOR_1]], %[[LOOP]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ]
@@ -876,20 +900,24 @@ loop:
   %gep.ptr = getelementptr inbounds i64, ptr %ptr, i64 %iv
   %add.1 = add i64 %for.1, 10
   store i64 %add.1, ptr %gep.ptr
-  %exitcond.not = icmp eq i64 %iv.next, 1000
+  %exitcond.not = icmp eq i64 %iv.next, %n
   br i1 %exitcond.not, label %exit, label %loop
 
 exit:
   ret i64 %for.1
 }
 
-define ptr @test_first_order_recurrences_and_pointer_induction1(ptr %ptr) {
+define ptr @test_first_order_recurrences_and_pointer_induction1(ptr %ptr, i64 %n) {
 ; CHECK-LABEL: define ptr @test_first_order_recurrences_and_pointer_induction1(
-; CHECK-SAME: ptr [[PTR:%.*]]) {
+; CHECK-SAME: ptr [[PTR:%.*]], i64 [[N:%.*]]) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
-; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
-; CHECK-NEXT:    [[IND_END:%.*]] = getelementptr i8, ptr [[PTR]], i64 4000
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[TMP2:%.*]] = mul i64 [[N_VEC]], 4
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[PTR]], i64 [[TMP2]]
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
@@ -898,21 +926,21 @@ define ptr @test_first_order_recurrences_and_pointer_induction1(ptr %ptr) {
 ; CHECK-NEXT:    [[TMP0]] = getelementptr i8, ptr [[POINTER_PHI]], <4 x i64> <i64 0, i64 4, i64 8, i64 12>
 ; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x ptr> [[VECTOR_RECUR]], <4 x ptr> [[TMP0]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds ptr, ptr [[PTR]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds ptr, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    store <4 x ptr> [[TMP0]], ptr [[TMP4]], align 8
-; CHECK-NEXT:    store <4 x ptr> [[TMP1]], ptr [[TMP4]], align 8
+; CHECK-NEXT:    store <4 x ptr> [[TMP0]], ptr [[TMP3]], align 8
+; CHECK-NEXT:    store <4 x ptr> [[TMP1]], ptr [[TMP3]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i64 16
-; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x ptr> [[TMP0]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x ptr> [[TMP0]], i32 2
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi ptr [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ null, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi ptr [ [[IND_END]], %[[MIDDLE_BLOCK]] ], [ [[PTR]], %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi ptr [ [[TMP6]], %[[MIDDLE_BLOCK]] ], [ [[PTR]], %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -923,7 +951,7 @@ define ptr @test_first_order_recurrences_and_pointer_induction1(ptr %ptr) {
 ; CHECK-NEXT:    store ptr [[PTR_IV]], ptr [[GEP_PTR]], align 8
 ; CHECK-NEXT:    [[PTR_IV_NEXT]] = getelementptr i32, ptr [[PTR_IV]], i64 1
 ; CHECK-NEXT:    store ptr [[FOR_1]], ptr [[GEP_PTR]], align 8
-; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP25:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    [[FOR_1_LCSSA:%.*]] = phi ptr [ [[FOR_1]], %[[LOOP]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ]
@@ -941,7 +969,7 @@ loop:
   store ptr %ptr.iv, ptr %gep.ptr
   %ptr.iv.next = getelementptr i32, ptr %ptr.iv, i64 1
   store ptr %for.1, ptr %gep.ptr
-  %exitcond.not = icmp eq i64 %iv.next, 1000
+  %exitcond.not = icmp eq i64 %iv.next, %n
   br i1 %exitcond.not, label %exit, label %loop
 
 exit:
@@ -950,13 +978,17 @@ exit:
 
 ; same as @test_first_order_recurrences_and_pointer_induction1 but with order
 ; of phis flipped.
-define ptr @test_first_order_recurrences_and_pointer_induction2(ptr %ptr) {
+define ptr @test_first_order_recurrences_and_pointer_induction2(ptr %ptr, i64 %n) {
 ; CHECK-LABEL: define ptr @test_first_order_recurrences_and_pointer_induction2(
-; CHECK-SAME: ptr [[PTR:%.*]]) {
+; CHECK-SAME: ptr [[PTR:%.*]], i64 [[N:%.*]]) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
-; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
-; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[PTR]], i64 4000
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[TMP0:%.*]] = mul i64 [[N_VEC]], 4
+; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[PTR]], i64 [[TMP0]]
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT1:%.*]], %[[VECTOR_BODY]] ]
@@ -965,20 +997,20 @@ define ptr @test_first_order_recurrences_and_pointer_induction2(ptr %ptr) {
 ; CHECK-NEXT:    [[VECTOR_GEP]] = getelementptr i8, ptr [[POINTER_PHI]], <4 x i64> <i64 0, i64 4, i64 8, i64 12>
 ; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x ptr> [[VECTOR_RECUR]], <4 x ptr> [[VECTOR_GEP]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds ptr, ptr [[PTR]], i64 [[INDEX1]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds ptr, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    store <4 x ptr> [[VECTOR_GEP]], ptr [[TMP6]], align 8
-; CHECK-NEXT:    store <4 x ptr> [[TMP1]], ptr [[TMP6]], align 8
+; CHECK-NEXT:    store <4 x ptr> [[VECTOR_GEP]], ptr [[TMP2]], align 8
+; CHECK-NEXT:    store <4 x ptr> [[TMP1]], ptr [[TMP2]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT1]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i64 16
-; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT1]], 1000
-; CHECK-NEXT:    br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT1]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x ptr> [[VECTOR_GEP]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x ptr> [[VECTOR_GEP]], i32 2
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi ptr [ [[TMP0]], %[[MIDDLE_BLOCK]] ], [ [[PTR]], %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi ptr [ [[TMP4]], %[[MIDDLE_BLOCK]] ], [ [[PTR]], %[[ENTRY]] ]
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi ptr [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ null, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
@@ -990,8 +1022,8 @@ define ptr @test_first_order_recurrences_and_pointer_induction2(ptr %ptr) {
 ; CHECK-NEXT:    store ptr [[PTR_IV]], ptr [[TMP3]], align 8
 ; CHECK-NEXT:    [[PTR_IV_NEXT]] = getelementptr i32, ptr [[PTR_IV]], i64 1
 ; CHECK-NEXT:    store ptr [[FOR_1]], ptr [[TMP3]], align 8
-; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; CHECK-NEXT:    br i1 [[TMP5]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP27:![0-9]+]]
+; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP27:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    [[FOR_1_LCSSA:%.*]] = phi ptr [ [[FOR_1]], %[[LOOP]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ]
 ; CHECK-NEXT:    ret ptr [[FOR_1_LCSSA]]
@@ -1008,7 +1040,7 @@ loop:
   store ptr %ptr.iv, ptr %gep.ptr
   %ptr.iv.next = getelementptr i32, ptr %ptr.iv, i64 1
   store ptr %for.1, ptr %gep.ptr
-  %exitcond.not = icmp eq i64 %iv.next, 1000
+  %exitcond.not = icmp eq i64 %iv.next, %n
   br i1 %exitcond.not, label %exit, label %loop
 
 exit:
@@ -1017,12 +1049,15 @@ exit:
 
 ; In this test case, %USE_2_FORS uses 2 different fixed-order recurrences and
 ; it needs to be sunk past the previous value for both recurrences.
-define double @test_resinking_required(ptr %p, ptr noalias %a, ptr noalias %b) {
+define double @test_resinking_required(ptr %p, ptr noalias %a, ptr noalias %b, i64 %n) {
 ; CHECK-LABEL: define double @test_resinking_required(
-; CHECK-SAME: ptr [[P:%.*]], ptr noalias [[A:%.*]], ptr noalias [[B:%.*]]) {
+; CHECK-SAME: ptr [[P:%.*]], ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], i64 [[N:%.*]]) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
-; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[N]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
@@ -1043,17 +1078,18 @@ define double @test_resinking_required(ptr %p, ptr noalias %a, ptr noalias %b) {
 ; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <4 x double> [[TMP8]], i32 3
 ; CHECK-NEXT:    store double [[TMP6]], ptr [[P]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; CHECK-NEXT:    br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
+; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT9:%.*]] = extractelement <4 x double> [[TMP4]], i32 3
+; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x double> [[TMP4]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x double> [[TMP4]], i32 2
-; CHECK-NEXT:    br i1 true, label %[[END:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label %[[END:.*]], label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi double [ [[TMP0]], %[[MIDDLE_BLOCK]] ], [ 0.000000e+00, %[[ENTRY]] ]
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT5:%.*]] = phi double [ [[TMP3]], %[[MIDDLE_BLOCK]] ], [ 0.000000e+00, %[[ENTRY]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT6:%.*]] = phi double [ [[VECTOR_RECUR_EXTRACT9]], %[[MIDDLE_BLOCK]] ], [ 0.000000e+00, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT6:%.*]] = phi double [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 0.000000e+00, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[FOR_1:%.*]] = phi double [ [[L1:%.*]], %[[LOOP]] ], [ [[SCALAR_RECUR_INIT]], %[[SCALAR_PH]] ]
@@ -1067,7 +1103,7 @@ define double @test_resinking_required(ptr %p, ptr noalias %a, ptr noalias %b) {
 ; CHECK-NEXT:    [[L2]] = load double, ptr [[B]], align 8
 ; CHECK-NEXT:    [[ADD:%.*]] = fadd double [[DIV]], [[FOR_3]]
 ; CHECK-NEXT:    store double [[ADD]], ptr [[P]], align 8
-; CHECK-NEXT:    [[COND:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
+; CHECK-NEXT:    [[COND:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; CHECK-NEXT:    br i1 [[COND]], label %[[END]], label %[[LOOP]], !llvm.loop [[LOOP29:![0-9]+]]
 ; CHECK:       [[END]]:
 ; CHECK-NEXT:    [[FOR_1_LCSSA:%.*]] = phi double [ [[FOR_1]], %[[LOOP]] ], [ [[TMP0]], %[[MIDDLE_BLOCK]] ]
@@ -1092,7 +1128,7 @@ Loop:
   %l2 = load double, ptr %b, align 8
   %add = fadd double %div, %for.3
   store double %add, ptr %p, align 8
-  %cond = icmp eq i64 %iv.next, 1000
+  %cond = icmp eq i64 %iv.next, %n
   br i1 %cond, label %End, label %Loop
 
 End:
diff --git a/llvm/test/Transforms/LoopVectorize/first-order-recurrence-complex.ll b/llvm/test/Transforms/LoopVectorize/first-order-recurrence-complex.ll
index ad3c7cb..74df675 100644
--- a/llvm/test/Transforms/LoopVectorize/first-order-recurrence-complex.ll
+++ b/llvm/test/Transforms/LoopVectorize/first-order-recurrence-complex.ll
@@ -25,20 +25,18 @@ define void @can_sink_after_store(i32 %x, ptr %ptr, i64 %tc) local_unnamed_addr
 ; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [257 x i32], ptr @p, i64 0, i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP4:%.*]] = add <4 x i32> [[TMP3]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = add <4 x i32> [[TMP4]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [257 x i32], ptr @q, i64 0, i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP7]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1996
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 3
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ [[DOTPRE]], [[PREHEADER]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1997, [[MIDDLE_BLOCK]] ], [ 1, [[PREHEADER]] ]
@@ -54,7 +52,7 @@ define void @can_sink_after_store(i32 %x, ptr %ptr, i64 %tc) local_unnamed_addr
 ; CHECK-NEXT:    store i32 [[ADD_2]], ptr [[IDX_2]], align 4
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], 2000
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[FOR]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[EXIT:%.*]], label [[FOR]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
@@ -104,20 +102,18 @@ define void @sink_sdiv(i32 %x, ptr %ptr, i64 %tc) local_unnamed_addr #0 {
 ; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [257 x i32], ptr @p, i64 0, i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP4:%.*]] = sdiv <4 x i32> [[TMP3]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = add <4 x i32> [[TMP4]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [257 x i32], ptr @q, i64 0, i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP7]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1996
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 3
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ [[DOTPRE]], [[PREHEADER]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1997, [[MIDDLE_BLOCK]] ], [ 1, [[PREHEADER]] ]
@@ -133,7 +129,7 @@ define void @sink_sdiv(i32 %x, ptr %ptr, i64 %tc) local_unnamed_addr #0 {
 ; CHECK-NEXT:    store i32 [[ADD_2]], ptr [[IDX_2]], align 4
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], 2000
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[FOR]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[EXIT:%.*]], label [[FOR]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
@@ -182,22 +178,20 @@ define void @can_sink_with_additional_user(i32 %x, ptr %ptr, i64 %tc) {
 ; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [257 x i32], ptr @p, i64 0, i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP4:%.*]] = add <4 x i32> [[TMP3]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = add <4 x i32> [[TMP4]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = add <4 x i32> [[TMP4]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = add <4 x i32> [[TMP5]], [[TMP6]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [257 x i32], ptr @q, i64 0, i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP7]], ptr [[TMP9]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP7]], ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1996
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 3
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ [[DOTPRE]], [[PREHEADER]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1997, [[MIDDLE_BLOCK]] ], [ 1, [[PREHEADER]] ]
@@ -215,7 +209,7 @@ define void @can_sink_with_additional_user(i32 %x, ptr %ptr, i64 %tc) {
 ; CHECK-NEXT:    store i32 [[ADD_4]], ptr [[IDX_2]], align 4
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], 2000
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[FOR]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[EXIT:%.*]], label [[FOR]], !llvm.loop [[LOOP7:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
@@ -375,13 +369,12 @@ define void @instruction_with_2_FOR_operands(ptr noalias %A, ptr noalias %B, ptr
 ; CHECK-NEXT:    [[BROADCAST_SPLAT3]] = shufflevector <4 x float> [[BROADCAST_SPLATINSERT2]], <4 x float> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <4 x float> [[VECTOR_RECUR]], <4 x float> [[BROADCAST_SPLAT3]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP6:%.*]] = fmul fast <4 x float> [[TMP5]], [[TMP3]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <4 x float> [[TMP6]], ptr [[TMP7]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP6]], ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[BB74:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi float [ [[TMP4]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[BB:%.*]] ]
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT4:%.*]] = phi float [ [[TMP2]], [[MIDDLE_BLOCK]] ], [ 1.000000e+00, [[BB]] ]
@@ -398,7 +391,7 @@ define void @instruction_with_2_FOR_operands(ptr noalias %A, ptr noalias %B, ptr
 ; CHECK-NEXT:    [[TMP60]] = load float, ptr [[B]], align 4
 ; CHECK-NEXT:    store float [[TMP38]], ptr [[GEP]], align 4
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp slt i64 [[IV]], 1000
-; CHECK-NEXT:    br i1 [[TMP12]], label [[BB13]], label [[BB74]], !llvm.loop [[LOOP9:![0-9]+]]
+; CHECK-NEXT:    br i1 [[TMP12]], label [[BB13]], label [[BB74:%.*]], !llvm.loop [[LOOP9:![0-9]+]]
 ; CHECK:       bb74:
 ; CHECK-NEXT:    ret void
 ;
@@ -444,19 +437,16 @@ define void @instruction_with_2_FOR_operands_and_multiple_other_uses(ptr noalias
 ; CHECK-NEXT:    [[TMP6:%.*]] = fmul fast <4 x float> [[TMP4]], [[TMP2]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = fadd fast <4 x float> [[TMP4]], splat (float 1.000000e+00)
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[DST_1:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <4 x float> [[TMP6]], ptr [[TMP9]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP6]], ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[DST_2:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <4 x float> [[TMP5]], ptr [[TMP11]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP5]], ptr [[TMP10]], align 4
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[DST_3:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP12]], i32 0
-; CHECK-NEXT:    store <4 x float> [[TMP7]], ptr [[TMP13]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP7]], ptr [[TMP12]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi float [ [[TMP3]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[BB:%.*]] ]
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT4:%.*]] = phi float [ [[TMP1]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[BB]] ]
@@ -479,7 +469,7 @@ define void @instruction_with_2_FOR_operands_and_multiple_other_uses(ptr noalias
 ; CHECK-NEXT:    [[GEP_DST_3:%.*]] = getelementptr inbounds float, ptr [[DST_3]], i64 [[IV]]
 ; CHECK-NEXT:    store float [[FOR_1_USE_3]], ptr [[GEP_DST_3]], align 4
 ; CHECK-NEXT:    [[EC:%.*]] = icmp slt i64 [[IV]], 1000
-; CHECK-NEXT:    br i1 [[EC]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP11:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label [[LOOP]], label [[EXIT:%.*]], !llvm.loop [[LOOP11:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
@@ -534,19 +524,16 @@ define void @instruction_with_2_FOR_operands_and_multiple_other_uses_chain(ptr n
 ; CHECK-NEXT:    [[TMP7:%.*]] = fmul fast <4 x float> [[TMP6]], [[TMP2]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = fadd fast <4 x float> [[TMP4]], splat (float 1.000000e+00)
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[DST_1:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[TMP9]], i32 0
-; CHECK-NEXT:    store <4 x float> [[TMP7]], ptr [[TMP10]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP7]], ptr [[TMP9]], align 4
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[DST_2:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    store <4 x float> [[TMP5]], ptr [[TMP12]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP5]], ptr [[TMP11]], align 4
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[DST_3:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds float, ptr [[TMP13]], i32 0
-; CHECK-NEXT:    store <4 x float> [[TMP8]], ptr [[TMP14]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP8]], ptr [[TMP13]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi float [ [[TMP3]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[BB:%.*]] ]
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT4:%.*]] = phi float [ [[TMP1]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[BB]] ]
@@ -570,7 +557,7 @@ define void @instruction_with_2_FOR_operands_and_multiple_other_uses_chain(ptr n
 ; CHECK-NEXT:    [[GEP_DST_3:%.*]] = getelementptr inbounds float, ptr [[DST_3]], i64 [[IV]]
 ; CHECK-NEXT:    store float [[FOR_1_USE_3]], ptr [[GEP_DST_3]], align 4
 ; CHECK-NEXT:    [[EC:%.*]] = icmp slt i64 [[IV]], 1000
-; CHECK-NEXT:    br i1 [[EC]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP13:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label [[LOOP]], label [[EXIT:%.*]], !llvm.loop [[LOOP13:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
@@ -678,10 +665,9 @@ define i16 @multiple_exit(ptr %p, i32 %n) {
 ; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 0>, [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = sext i32 [[INDEX]] to i64
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[P:%.*]], i64 [[TMP4]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP6]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP5]], align 2
 ; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; CHECK-NEXT:    store <4 x i16> [[TMP7]], ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <4 x i16> [[TMP7]], ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
@@ -753,10 +739,9 @@ define i16 @multiple_exit2(ptr %p, i32 %n) {
 ; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ <i16 poison, i16 poison, i16 poison, i16 0>, [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = sext i32 [[INDEX]] to i64
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[P:%.*]], i64 [[TMP4]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP6]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP5]], align 2
 ; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; CHECK-NEXT:    store <4 x i16> [[TMP7]], ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <4 x i16> [[TMP7]], ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
@@ -827,14 +812,13 @@ define void @sink_dominance(ptr %ptr, i32 %N) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i64> [ <i64 poison, i64 poison, i64 poison, i64 0>, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[PTR:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[TMP5]] = zext <4 x i32> [[WIDE_LOAD]] to <4 x i64>
 ; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <4 x i64> [[VECTOR_RECUR]], <4 x i64> [[TMP5]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP7:%.*]] = trunc <4 x i64> [[TMP6]] to <4 x i32>
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp slt <4 x i32> [[TMP7]], splat (i32 213)
 ; CHECK-NEXT:    [[TMP9:%.*]] = select <4 x i1> [[TMP8]], <4 x i32> [[TMP7]], <4 x i32> splat (i32 22)
-; CHECK-NEXT:    store <4 x i32> [[TMP9]], ptr [[TMP4]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP9]], ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
@@ -907,8 +891,7 @@ define void @sink_dominance_2(ptr %ptr, i32 %N) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i64> [ <i64 poison, i64 poison, i64 poison, i64 0>, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[PTR:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[TMP5]] = zext <4 x i32> [[WIDE_LOAD]] to <4 x i64>
 ; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <4 x i64> [[VECTOR_RECUR]], <4 x i64> [[TMP5]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP7:%.*]] = trunc <4 x i64> [[TMP6]] to <4 x i32>
@@ -916,7 +899,7 @@ define void @sink_dominance_2(ptr %ptr, i32 %N) {
 ; CHECK-NEXT:    [[TMP9:%.*]] = mul <4 x i32> [[TMP8]], splat (i32 99)
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp slt <4 x i32> [[TMP7]], splat (i32 213)
 ; CHECK-NEXT:    [[TMP11:%.*]] = select <4 x i1> [[TMP10]], <4 x i32> [[TMP7]], <4 x i32> [[TMP9]]
-; CHECK-NEXT:    store <4 x i32> [[TMP11]], ptr [[TMP4]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP11]], ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
@@ -1031,20 +1014,19 @@ define void @test_for_sink_instruction_after_same_incoming_1(ptr %ptr) {
 ; CHECK-NEXT:    [[VECTOR_RECUR1:%.*]] = phi <4 x double> [ <double poison, double poison, double poison, double 2.000000e+01>, [[VECTOR_PH]] ], [ [[WIDE_LOAD]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds double, ptr [[PTR:%.*]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds double, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x double>, ptr [[TMP2]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x double>, ptr [[TMP1]], align 8
 ; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x double> [[VECTOR_RECUR1]], <4 x double> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP4:%.*]] = fadd <4 x double> splat (double 1.000000e+01), [[TMP3]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <4 x double> [[VECTOR_RECUR]], <4 x double> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP6:%.*]] = fadd <4 x double> [[TMP4]], [[TMP5]]
-; CHECK-NEXT:    store <4 x double> [[TMP6]], ptr [[TMP2]], align 8
+; CHECK-NEXT:    store <4 x double> [[TMP6]], ptr [[TMP1]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 996
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x double> [[WIDE_LOAD]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT2:%.*]] = extractelement <4 x double> [[WIDE_LOAD]], i32 3
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi double [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 1.000000e+01, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT3:%.*]] = phi double [ [[VECTOR_RECUR_EXTRACT2]], [[MIDDLE_BLOCK]] ], [ 2.000000e+01, [[ENTRY]] ]
@@ -1061,7 +1043,7 @@ define void @test_for_sink_instruction_after_same_incoming_1(ptr %ptr) {
 ; CHECK-NEXT:    [[FOR_1_NEXT]] = load double, ptr [[GEP_PTR]], align 8
 ; CHECK-NEXT:    store double [[ADD_2]], ptr [[GEP_PTR]], align 8
 ; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
-; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP23:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP23:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
@@ -1098,20 +1080,19 @@ define void @test_for_sink_instruction_after_same_incoming_2(ptr %ptr) {
 ; CHECK-NEXT:    [[VECTOR_RECUR1:%.*]] = phi <4 x double> [ <double poison, double poison, double poison, double 1.000000e+01>, [[VECTOR_PH]] ], [ [[WIDE_LOAD]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds double, ptr [[PTR:%.*]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds double, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x double>, ptr [[TMP2]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <4 x double>, ptr [[TMP1]], align 8
 ; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x double> [[VECTOR_RECUR1]], <4 x double> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <4 x double> [[VECTOR_RECUR]], <4 x double> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP5:%.*]] = fadd <4 x double> splat (double 1.000000e+01), [[TMP4]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = fadd <4 x double> [[TMP5]], [[TMP3]]
-; CHECK-NEXT:    store <4 x double> [[TMP6]], ptr [[TMP2]], align 8
+; CHECK-NEXT:    store <4 x double> [[TMP6]], ptr [[TMP1]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 996
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x double> [[WIDE_LOAD]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT2:%.*]] = extractelement <4 x double> [[WIDE_LOAD]], i32 3
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi double [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 2.000000e+01, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT3:%.*]] = phi double [ [[VECTOR_RECUR_EXTRACT2]], [[MIDDLE_BLOCK]] ], [ 1.000000e+01, [[ENTRY]] ]
@@ -1128,7 +1109,7 @@ define void @test_for_sink_instruction_after_same_incoming_2(ptr %ptr) {
 ; CHECK-NEXT:    [[FOR_1_NEXT]] = load double, ptr [[GEP_PTR]], align 8
 ; CHECK-NEXT:    store double [[ADD_2]], ptr [[GEP_PTR]], align 8
 ; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 1000
-; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP25:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP25:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/first-order-recurrence-dead-instructions.ll b/llvm/test/Transforms/LoopVectorize/first-order-recurrence-dead-instructions.ll
index 86171e6..71c2da2 100644
--- a/llvm/test/Transforms/LoopVectorize/first-order-recurrence-dead-instructions.ll
+++ b/llvm/test/Transforms/LoopVectorize/first-order-recurrence-dead-instructions.ll
@@ -6,42 +6,18 @@ define i8 @recurrence_phi_with_same_incoming_values_after_simplifications(i8 %fo
 ; CHECK-LABEL: define i8 @recurrence_phi_with_same_incoming_values_after_simplifications(
 ; CHECK-SAME: i8 [[FOR_START:%.*]], ptr [[DST:%.*]]) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
-; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_SCEVCHECK:.*]]
-; CHECK:       [[VECTOR_SCEVCHECK]]:
-; CHECK-NEXT:    br i1 true, label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
-; CHECK:       [[VECTOR_PH]]:
-; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i8> poison, i8 [[FOR_START]], i64 0
-; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i8> [[BROADCAST_SPLATINSERT]], <4 x i8> poison, <4 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <4 x i8> [[BROADCAST_SPLAT]], <4 x i8> [[BROADCAST_SPLAT]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
-; CHECK:       [[VECTOR_BODY]]:
-; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i32 1, [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[DST]], i32 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 4
-; CHECK-NEXT:    store <4 x i8> [[TMP0]], ptr [[TMP2]], align 1
-; CHECK-NEXT:    store <4 x i8> [[TMP0]], ptr [[TMP3]], align 1
-; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
-; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], -8
-; CHECK-NEXT:    br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
-; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
-; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ -7, %[[MIDDLE_BLOCK]] ], [ 1, %[[ENTRY]] ], [ 1, %[[VECTOR_SCEVCHECK]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i8 [ [[FOR_START]], %[[MIDDLE_BLOCK]] ], [ [[FOR_START]], %[[ENTRY]] ], [ [[FOR_START]], %[[VECTOR_SCEVCHECK]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
-; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
-; CHECK-NEXT:    [[FOR:%.*]] = phi i8 [ [[SCALAR_RECUR_INIT]], %[[SCALAR_PH]] ], [ [[FOR_NEXT:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ 1, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[FOR:%.*]] = phi i8 [ [[FOR_START]], %[[ENTRY]] ], [ [[FOR_NEXT:%.*]], %[[LOOP]] ]
 ; CHECK-NEXT:    [[FOR_NEXT]] = and i8 [[FOR_START]], -1
 ; CHECK-NEXT:    [[IV_NEXT]] = add i32 [[IV]], 1
 ; CHECK-NEXT:    [[GEP_DST:%.*]] = getelementptr inbounds i8, ptr [[DST]], i32 [[IV]]
 ; CHECK-NEXT:    store i8 [[FOR]], ptr [[GEP_DST]], align 1
 ; CHECK-NEXT:    [[EC:%.*]] = icmp eq i32 [[IV_NEXT]], 0
-; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP]]
 ; CHECK:       [[EXIT]]:
-; CHECK-NEXT:    [[FOR_NEXT_LCSSA:%.*]] = phi i8 [ [[FOR_NEXT]], %[[LOOP]] ], [ [[FOR_START]], %[[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[FOR_NEXT_LCSSA:%.*]] = phi i8 [ [[FOR_NEXT]], %[[LOOP]] ]
 ; CHECK-NEXT:    ret i8 [[FOR_NEXT_LCSSA]]
 ;
 entry:
@@ -79,21 +55,19 @@ define i32 @sink_after_dead_inst(ptr %A.ptr) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = or <4 x i16> [[TMP0]], [[TMP0]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = zext <4 x i16> [[TMP1]] to <4 x i32>
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i32, ptr [[A_PTR]], i16 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[TMP3]], i32 0
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[TMP3]], i32 4
-; CHECK-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP4]], align 4
+; CHECK-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP3]], align 4
 ; CHECK-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[STEP_ADD]], splat (i16 4)
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 16
-; CHECK-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP2]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP2]], i32 2
-; CHECK-NEXT:    br i1 true, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[FOR_END:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 16, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -108,7 +82,7 @@ define i32 @sink_after_dead_inst(ptr %A.ptr) {
 ; CHECK-NEXT:    [[EXT:%.*]] = zext i1 [[B3]] to i32
 ; CHECK-NEXT:    [[A_GEP:%.*]] = getelementptr i32, ptr [[A_PTR]], i16 [[IV]]
 ; CHECK-NEXT:    store i32 0, ptr [[A_GEP]], align 4
-; CHECK-NEXT:    br i1 [[VEC_DEAD]], label %[[FOR_END]], label %[[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    br i1 [[VEC_DEAD]], label %[[FOR_END]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       [[FOR_END]]:
 ; CHECK-NEXT:    [[FOR_LCSSA:%.*]] = phi i32 [ [[FOR]], %[[LOOP]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ]
 ; CHECK-NEXT:    ret i32 [[FOR_LCSSA]]
@@ -162,18 +136,17 @@ define void @sink_dead_inst(ptr %a) {
 ; CHECK-NEXT:    [[TMP7:%.*]] = sub <4 x i16> [[TMP5]], splat (i16 10)
 ; CHECK-NEXT:    [[TMP8:%.*]] = sub <4 x i16> [[TMP6]], splat (i16 10)
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr i16, ptr [[A]], i16 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i16, ptr [[TMP9]], i32 0
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i16, ptr [[TMP9]], i32 4
-; CHECK-NEXT:    store <4 x i16> [[TMP7]], ptr [[TMP10]], align 2
+; CHECK-NEXT:    store <4 x i16> [[TMP7]], ptr [[TMP9]], align 2
 ; CHECK-NEXT:    store <4 x i16> [[TMP8]], ptr [[TMP11]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[STEP_ADD]], splat (i16 4)
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i32 [[INDEX_NEXT]], 40
-; CHECK-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[TMP4]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT1:%.*]] = extractelement <4 x i32> [[TMP2]], i32 3
-; CHECK-NEXT:    br i1 false, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 13, %[[MIDDLE_BLOCK]] ], [ -27, %[[ENTRY]] ]
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
@@ -190,7 +163,7 @@ define void @sink_dead_inst(ptr %a) {
 ; CHECK-NEXT:    [[REC_1_PREV]] = add i16 [[IV_NEXT]], 5
 ; CHECK-NEXT:    [[GEP:%.*]] = getelementptr i16, ptr [[A]], i16 [[IV]]
 ; CHECK-NEXT:    store i16 [[USE_REC_1]], ptr [[GEP]], align 2
-; CHECK-NEXT:    br i1 [[CMP]], label %[[FOR_END]], label %[[FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP]], label %[[FOR_END:.*]], label %[[FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       [[FOR_END]]:
 ; CHECK-NEXT:    ret void
 ;
@@ -232,10 +205,10 @@ define void @unused_recurrence(ptr %a) {
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[STEP_ADD]], splat (i16 4)
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1024
-; CHECK-NEXT:    br i1 [[TMP2]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK-NEXT:    br i1 [[TMP2]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[TMP1]], i32 3
-; CHECK-NEXT:    br i1 false, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 997, %[[MIDDLE_BLOCK]] ], [ -27, %[[ENTRY]] ]
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
@@ -247,7 +220,7 @@ define void @unused_recurrence(ptr %a) {
 ; CHECK-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
 ; CHECK-NEXT:    [[REC_1_PREV]] = add i16 [[IV_NEXT]], 5
 ; CHECK-NEXT:    [[CMP:%.*]] = icmp eq i16 [[IV]], 1000
-; CHECK-NEXT:    br i1 [[CMP]], label %[[FOR_END]], label %[[FOR_COND]], !llvm.loop [[LOOP9:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP]], label %[[FOR_END:.*]], label %[[FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
 ; CHECK:       [[FOR_END]]:
 ; CHECK-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/first-order-recurrence-interleave-only.ll b/llvm/test/Transforms/LoopVectorize/first-order-recurrence-interleave-only.ll
index 53113b2..d95c487 100644
--- a/llvm/test/Transforms/LoopVectorize/first-order-recurrence-interleave-only.ll
+++ b/llvm/test/Transforms/LoopVectorize/first-order-recurrence-interleave-only.ll
@@ -7,7 +7,6 @@ define float @for_load_interleave_only(ptr %src) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
 ; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
-; CHECK-NEXT:    [[IND_END:%.*]] = getelementptr i8, ptr [[SRC]], i64 16000
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
@@ -16,18 +15,17 @@ define float @for_load_interleave_only(ptr %src) {
 ; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[SRC]], i64 [[OFFSET_IDX]]
 ; CHECK-NEXT:    [[NEXT_GEP2:%.*]] = getelementptr i8, ptr [[SRC]], i64 [[TMP1]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = load float, ptr [[NEXT_GEP]], align 4
-; CHECK-NEXT:    [[TMP3:%.*]] = load float, ptr [[NEXT_GEP2]], align 4
 ; CHECK-NEXT:    store float 0.000000e+00, ptr [[NEXT_GEP]], align 4
 ; CHECK-NEXT:    store float 0.000000e+00, ptr [[NEXT_GEP2]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1001, %[[MIDDLE_BLOCK]] ], [ 1, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi ptr [ [[IND_END]], %[[MIDDLE_BLOCK]] ], [ [[SRC]], %[[ENTRY]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi float [ [[TMP3]], %[[MIDDLE_BLOCK]] ], [ 0.000000e+00, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi ptr [ [[SRC]], %[[ENTRY]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi float [ 0.000000e+00, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/first-order-recurrence-multiply-recurrences.ll b/llvm/test/Transforms/LoopVectorize/first-order-recurrence-multiply-recurrences.ll
index 05eaae5..715ea1c 100644
--- a/llvm/test/Transforms/LoopVectorize/first-order-recurrence-multiply-recurrences.ll
+++ b/llvm/test/Transforms/LoopVectorize/first-order-recurrence-multiply-recurrences.ll
@@ -114,17 +114,16 @@ define void @test_pr54223_sink_after_insertion_order(ptr noalias %a, ptr noalias
 ; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <4 x float> [[VECTOR_RECUR1]], <4 x float> [[BROADCAST_SPLAT3]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP6:%.*]] = fneg <4 x float> [[TMP5]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = call <4 x float> @llvm.fmuladd.v4f32(<4 x float> [[TMP3]], <4 x float> [[TMP6]], <4 x float> zeroinitializer)
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <4 x float> [[TMP7]], ptr [[TMP8]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP7]], ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10000
 ; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 10000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi float [ [[TMP2]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT5:%.*]] = phi float [ [[TMP4]], [[MIDDLE_BLOCK]] ], [ 0.000000e+00, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT5:%.*]] = phi float [ 0.000000e+00, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -235,14 +234,13 @@ define void @test_pr54233_for_depend_on_each_other(ptr noalias %a, ptr noalias %
 ; CHECK-NEXT:    [[TMP7:%.*]] = xor <4 x i32> [[TMP6]], splat (i32 255)
 ; CHECK-NEXT:    [[TMP8:%.*]] = and <4 x i32> [[TMP7]], [[TMP3]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP8]], ptr [[TMP10]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP8]], ptr [[TMP9]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP4]], i32 3
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
@@ -262,7 +260,7 @@ define void @test_pr54233_for_depend_on_each_other(ptr noalias %a, ptr noalias %
 ; CHECK-NEXT:    store i32 [[AND]], ptr [[A_GEP]], align 4
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw i64 [[IV]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV]], 1000
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
@@ -398,13 +396,12 @@ define void @hoist_previous_value_and_operand(ptr %dst, i64 %mask) {
 ; CHECK-NEXT:    [[VECTOR_RECUR1:%.*]] = phi <4 x i32> [ <i32 poison, i32 poison, i32 poison, i32 0>, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[DST:%.*]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
 ; CHECK-NEXT:    [[TMP3:%.*]] = and <4 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP4]] = trunc <4 x i64> [[TMP3]] to <4 x i32>
 ; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[TMP4]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; CHECK-NEXT:    [[TMP6]] = or <4 x i32> [[TMP5]], splat (i32 3)
 ; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR1]], <4 x i32> [[TMP6]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
-; CHECK-NEXT:    store <4 x i32> [[TMP7]], ptr [[TMP2]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP7]], ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 336
@@ -412,7 +409,7 @@ define void @hoist_previous_value_and_operand(ptr %dst, i64 %mask) {
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP4]], i32 3
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT2:%.*]] = extractelement <4 x i32> [[TMP6]], i32 3
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 337, [[MIDDLE_BLOCK]] ], [ 1, [[BB:%.*]] ]
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 1, [[BB]] ]
@@ -429,7 +426,7 @@ define void @hoist_previous_value_and_operand(ptr %dst, i64 %mask) {
 ; CHECK-NEXT:    [[ICMP:%.*]] = icmp ult i64 [[IV]], 337
 ; CHECK-NEXT:    [[A:%.*]] = and i64 [[IV]], [[MASK]]
 ; CHECK-NEXT:    [[TRUNC]] = trunc i64 [[A]] to i32
-; CHECK-NEXT:    br i1 [[ICMP]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK-NEXT:    br i1 [[ICMP]], label [[LOOP]], label [[EXIT:%.*]], !llvm.loop [[LOOP7:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/first-order-recurrence-scalable-vf1.ll b/llvm/test/Transforms/LoopVectorize/first-order-recurrence-scalable-vf1.ll
index f4044a7..2bafa6c 100644
--- a/llvm/test/Transforms/LoopVectorize/first-order-recurrence-scalable-vf1.ll
+++ b/llvm/test/Transforms/LoopVectorize/first-order-recurrence-scalable-vf1.ll
@@ -24,8 +24,7 @@ define i64 @pr97452_scalable_vf1_for_live_out(ptr %src) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 1 x i64> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[SRC]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD]] = load <vscale x 1 x i64>, ptr [[TMP6]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <vscale x 1 x i64>, ptr [[TMP5]], align 8
 ; CHECK-NEXT:    [[TMP7:%.*]] = call <vscale x 1 x i64> @llvm.vector.splice.nxv1i64(<vscale x 1 x i64> [[VECTOR_RECUR]], <vscale x 1 x i64> [[WIDE_LOAD]], i32 -1)
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP2]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -93,12 +92,10 @@ define void @pr97452_scalable_vf1_for_no_live_out(ptr %src, ptr noalias %dst) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 1 x i64> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[SRC]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD]] = load <vscale x 1 x i64>, ptr [[TMP6]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD]] = load <vscale x 1 x i64>, ptr [[TMP5]], align 8
 ; CHECK-NEXT:    [[TMP7:%.*]] = call <vscale x 1 x i64> @llvm.vector.splice.nxv1i64(<vscale x 1 x i64> [[VECTOR_RECUR]], <vscale x 1 x i64> [[WIDE_LOAD]], i32 -1)
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <vscale x 1 x i64> [[TMP7]], ptr [[TMP9]], align 8
+; CHECK-NEXT:    store <vscale x 1 x i64> [[TMP7]], ptr [[TMP8]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP2]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/first-order-recurrence.ll b/llvm/test/Transforms/LoopVectorize/first-order-recurrence.ll
index ff5c29f..3adfcf5 100644
--- a/llvm/test/Transforms/LoopVectorize/first-order-recurrence.ll
+++ b/llvm/test/Transforms/LoopVectorize/first-order-recurrence.ll
@@ -33,18 +33,16 @@ define void @recurrence_1(ptr readonly noalias %a, ptr noalias %b, i32 %n) {
 ; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD1:%.*]], [[VECTOR_BODY]] ]
 ; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add nuw nsw i64 [[INDEX]], 1
 ; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP4]]
-; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 4
-; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4
 ; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD1]] = load <4 x i32>, ptr [[TMP7]], align 4
 ; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = shufflevector <4 x i32> [[WIDE_LOAD]], <4 x i32> [[WIDE_LOAD1]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
 ; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = add <4 x i32> [[WIDE_LOAD]], [[TMP8]]
 ; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = add <4 x i32> [[WIDE_LOAD1]], [[TMP9]]
-; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP10]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP10]], i32 4
-; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP11]], ptr [[TMP13]], align 4
+; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP11]], ptr [[TMP10]], align 4
 ; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP12]], ptr [[TMP14]], align 4
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -147,13 +145,11 @@ define void @recurrence_1(ptr readonly noalias %a, ptr noalias %b, i32 %n) {
 ; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
 ; SINK-AFTER-NEXT:    [[TMP4:%.*]] = add nuw nsw i64 [[INDEX]], 1
 ; SINK-AFTER-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP4]]
-; SINK-AFTER-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i32>, ptr [[TMP6]], align 4
+; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i32>, ptr [[TMP5]], align 4
 ; SINK-AFTER-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; SINK-AFTER-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
 ; SINK-AFTER-NEXT:    [[TMP9:%.*]] = add <4 x i32> [[WIDE_LOAD]], [[TMP7]]
-; SINK-AFTER-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
-; SINK-AFTER-NEXT:    store <4 x i32> [[TMP9]], ptr [[TMP10]], align 4
+; SINK-AFTER-NEXT:    store <4 x i32> [[TMP9]], ptr [[TMP8]], align 4
 ; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; SINK-AFTER-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; SINK-AFTER-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -235,9 +231,8 @@ define i32 @recurrence_2(ptr nocapture readonly %a, i32 %n) {
 ; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ poison, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[VECTOR_BODY]] ]
 ; UNROLL-NO-IC-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i32> [ poison, [[VECTOR_PH]] ], [ [[TMP16:%.*]], [[VECTOR_BODY]] ]
 ; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
-; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 4
-; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
 ; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD2]] = load <4 x i32>, ptr [[TMP4]], align 4
 ; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = shufflevector <4 x i32> [[WIDE_LOAD]], <4 x i32> [[WIDE_LOAD2]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
@@ -375,8 +370,7 @@ define i32 @recurrence_2(ptr nocapture readonly %a, i32 %n) {
 ; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i32> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
 ; SINK-AFTER-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ poison, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; SINK-AFTER-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
-; SINK-AFTER-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
-; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i32>, ptr [[TMP3]], align 4
+; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i32>, ptr [[TMP2]], align 4
 ; SINK-AFTER-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i32> [[VECTOR_RECUR]], <4 x i32> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; SINK-AFTER-NEXT:    [[TMP5:%.*]] = sub nsw <4 x i32> [[WIDE_LOAD]], [[TMP4]]
 ; SINK-AFTER-NEXT:    [[TMP6:%.*]] = icmp sgt <4 x i32> [[TMP5]], zeroinitializer
@@ -491,9 +485,8 @@ define void @recurrence_3(ptr readonly noalias %a, ptr noalias %b, i32 %n, float
 ; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD1:%.*]], [[VECTOR_BODY]] ]
 ; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[OFFSET_IDX]]
-; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[TMP4]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[TMP4]], i32 4
-; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP5]], align 2
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP4]], align 2
 ; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD1]] = load <4 x i16>, ptr [[TMP6]], align 2
 ; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = shufflevector <4 x i16> [[WIDE_LOAD]], <4 x i16> [[WIDE_LOAD1]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
@@ -506,9 +499,8 @@ define void @recurrence_3(ptr readonly noalias %a, ptr noalias %b, i32 %n, float
 ; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = fsub fast <4 x double> [[TMP9]], [[TMP13]]
 ; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = fsub fast <4 x double> [[TMP10]], [[TMP14]]
 ; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[OFFSET_IDX]]
-; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = getelementptr inbounds double, ptr [[TMP17]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = getelementptr inbounds double, ptr [[TMP17]], i32 4
-; UNROLL-NO-IC-NEXT:    store <4 x double> [[TMP15]], ptr [[TMP18]], align 8
+; UNROLL-NO-IC-NEXT:    store <4 x double> [[TMP15]], ptr [[TMP17]], align 8
 ; UNROLL-NO-IC-NEXT:    store <4 x double> [[TMP16]], ptr [[TMP19]], align 8
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -642,16 +634,14 @@ define void @recurrence_3(ptr readonly noalias %a, ptr noalias %b, i32 %n, float
 ; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
 ; SINK-AFTER-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; SINK-AFTER-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[OFFSET_IDX]]
-; SINK-AFTER-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[TMP4]], i32 0
-; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP5]], align 2
+; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP4]], align 2
 ; SINK-AFTER-NEXT:    [[TMP6:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; SINK-AFTER-NEXT:    [[TMP7:%.*]] = sitofp <4 x i16> [[WIDE_LOAD]] to <4 x double>
 ; SINK-AFTER-NEXT:    [[TMP8:%.*]] = sitofp <4 x i16> [[TMP6]] to <4 x double>
 ; SINK-AFTER-NEXT:    [[TMP9:%.*]] = fmul fast <4 x double> [[TMP8]], [[BROADCAST_SPLAT]]
 ; SINK-AFTER-NEXT:    [[TMP10:%.*]] = fsub fast <4 x double> [[TMP7]], [[TMP9]]
 ; SINK-AFTER-NEXT:    [[TMP11:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[OFFSET_IDX]]
-; SINK-AFTER-NEXT:    [[TMP12:%.*]] = getelementptr inbounds double, ptr [[TMP11]], i32 0
-; SINK-AFTER-NEXT:    store <4 x double> [[TMP10]], ptr [[TMP12]], align 8
+; SINK-AFTER-NEXT:    store <4 x double> [[TMP10]], ptr [[TMP11]], align 8
 ; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; SINK-AFTER-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; SINK-AFTER-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
@@ -1203,10 +1193,10 @@ define i64 @constant_folded_previous_value() {
 ; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; UNROLL-NO-IC-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; UNROLL-NO-IC:       middle.block:
-; UNROLL-NO-IC-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_END:%.*]]
 ; UNROLL-NO-IC:       scalar.ph:
-; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ 1, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ 0, [[ENTRY]] ]
 ; UNROLL-NO-IC-NEXT:    br label [[SCALAR_BODY:%.*]]
 ; UNROLL-NO-IC:       scalar.body:
 ; UNROLL-NO-IC-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[SCALAR_BODY]] ]
@@ -1230,10 +1220,10 @@ define i64 @constant_folded_previous_value() {
 ; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; UNROLL-NO-VF-NEXT:    br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; UNROLL-NO-VF:       middle.block:
-; UNROLL-NO-VF-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF-NEXT:    br label [[FOR_END:%.*]]
 ; UNROLL-NO-VF:       scalar.ph:
-; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ 1, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ 0, [[ENTRY]] ]
 ; UNROLL-NO-VF-NEXT:    br label [[SCALAR_BODY:%.*]]
 ; UNROLL-NO-VF:       scalar.body:
 ; UNROLL-NO-VF-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[SCALAR_BODY]] ]
@@ -1257,10 +1247,10 @@ define i64 @constant_folded_previous_value() {
 ; SINK-AFTER-NEXT:    [[TMP0:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; SINK-AFTER-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; SINK-AFTER:       middle.block:
-; SINK-AFTER-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER-NEXT:    br label [[FOR_END:%.*]]
 ; SINK-AFTER:       scalar.ph:
-; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ 1, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ 0, [[ENTRY]] ]
 ; SINK-AFTER-NEXT:    br label [[SCALAR_BODY:%.*]]
 ; SINK-AFTER:       scalar.body:
 ; SINK-AFTER-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[SCALAR_BODY]] ]
@@ -1294,11 +1284,15 @@ for.end:
 ; the first order recurrence phi is used outside the loop, so we require the phi
 ; itself and not its update (addx).
 ; Check the case when unrolled but not vectorized.
-define i32 @extract_second_last_iteration(ptr %cval, i32 %x)  {
+define i32 @extract_second_last_iteration(ptr %cval, i32 %x, i32 %n)  {
 ; UNROLL-NO-IC-LABEL: @extract_second_last_iteration(
 ; UNROLL-NO-IC-NEXT:  entry:
-; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add i32 [[N:%.*]], 1
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP1]], 8
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[TMP1]], 8
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i32 [[TMP1]], [[N_MOD_VF]]
 ; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[X:%.*]], i64 0
 ; UNROLL-NO-IC-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
@@ -1309,14 +1303,15 @@ define i32 @extract_second_last_iteration(ptr %cval, i32 %x)  {
 ; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add <4 x i32> [[STEP_ADD]], [[BROADCAST_SPLAT]]
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
 ; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[STEP_ADD]], splat (i32 4)
-; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], 96
-; UNROLL-NO-IC-NEXT:    br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; UNROLL-NO-IC:       middle.block:
 ; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP0]], i32 3
 ; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP0]], i32 2
-; UNROLL-NO-IC-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP1]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; UNROLL-NO-IC:       scalar.ph:
-; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 96, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
 ; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
 ; UNROLL-NO-IC:       for.body:
@@ -1325,7 +1320,7 @@ define i32 @extract_second_last_iteration(ptr %cval, i32 %x)  {
 ; UNROLL-NO-IC-NEXT:    [[INC]] = add i32 [[INC_PHI]], 1
 ; UNROLL-NO-IC-NEXT:    [[BC:%.*]] = zext i32 [[INC_PHI]] to i64
 ; UNROLL-NO-IC-NEXT:    [[ADDX]] = add i32 [[INC_PHI]], [[X]]
-; UNROLL-NO-IC-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC_PHI]], 95
+; UNROLL-NO-IC-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC_PHI]], [[N]]
 ; UNROLL-NO-IC-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; UNROLL-NO-IC:       for.end:
 ; UNROLL-NO-IC-NEXT:    [[VAL_PHI_LCSSA:%.*]] = phi i32 [ [[VAL_PHI]], [[FOR_BODY]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], [[MIDDLE_BLOCK]] ]
@@ -1333,21 +1328,26 @@ define i32 @extract_second_last_iteration(ptr %cval, i32 %x)  {
 ;
 ; UNROLL-NO-VF-LABEL: @extract_second_last_iteration(
 ; UNROLL-NO-VF-NEXT:  entry:
-; UNROLL-NO-VF-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = add i32 [[N:%.*]], 1
+; UNROLL-NO-VF-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP1]], 2
+; UNROLL-NO-VF-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; UNROLL-NO-VF:       vector.ph:
+; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[TMP1]], 2
+; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i32 [[TMP1]], [[N_MOD_VF]]
 ; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; UNROLL-NO-VF:       vector.body:
 ; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
-; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i32 [[VECTOR_RECUR]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = add i32 [[VECTOR_RECUR]], 1
 ; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = add i32 [[VECTOR_RECUR]], [[X:%.*]]
-; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = add i32 [[TMP0]], [[X]]
+; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = add i32 [[TMP6]], [[X]]
 ; UNROLL-NO-VF-NEXT:    [[TMP3]] = add nuw i32 [[VECTOR_RECUR]], 2
-; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[TMP3]], 96
+; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[TMP3]], [[N_VEC]]
 ; UNROLL-NO-VF-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; UNROLL-NO-VF:       middle.block:
-; UNROLL-NO-VF-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP1]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; UNROLL-NO-VF:       scalar.ph:
-; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 96, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[TMP4]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
 ; UNROLL-NO-VF-NEXT:    br label [[FOR_BODY:%.*]]
 ; UNROLL-NO-VF:       for.body:
@@ -1356,7 +1356,7 @@ define i32 @extract_second_last_iteration(ptr %cval, i32 %x)  {
 ; UNROLL-NO-VF-NEXT:    [[INC]] = add i32 [[INC_PHI]], 1
 ; UNROLL-NO-VF-NEXT:    [[BC:%.*]] = zext i32 [[INC_PHI]] to i64
 ; UNROLL-NO-VF-NEXT:    [[ADDX]] = add i32 [[INC_PHI]], [[X]]
-; UNROLL-NO-VF-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC_PHI]], 95
+; UNROLL-NO-VF-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC_PHI]], [[N]]
 ; UNROLL-NO-VF-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; UNROLL-NO-VF:       for.end:
 ; UNROLL-NO-VF-NEXT:    [[VAL_PHI_LCSSA:%.*]] = phi i32 [ [[VAL_PHI]], [[FOR_BODY]] ], [ [[TMP2]], [[MIDDLE_BLOCK]] ]
@@ -1364,8 +1364,12 @@ define i32 @extract_second_last_iteration(ptr %cval, i32 %x)  {
 ;
 ; SINK-AFTER-LABEL: @extract_second_last_iteration(
 ; SINK-AFTER-NEXT:  entry:
-; SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; SINK-AFTER-NEXT:    [[TMP1:%.*]] = add i32 [[N:%.*]], 1
+; SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP1]], 4
+; SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; SINK-AFTER:       vector.ph:
+; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[TMP1]], 4
+; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i32 [[TMP1]], [[N_MOD_VF]]
 ; SINK-AFTER-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[X:%.*]], i64 0
 ; SINK-AFTER-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
@@ -1375,14 +1379,15 @@ define i32 @extract_second_last_iteration(ptr %cval, i32 %x)  {
 ; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add <4 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
 ; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; SINK-AFTER-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
-; SINK-AFTER-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], 96
-; SINK-AFTER-NEXT:    br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
+; SINK-AFTER-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; SINK-AFTER-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; SINK-AFTER:       middle.block:
 ; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP0]], i32 3
 ; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP0]], i32 2
-; SINK-AFTER-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP1]], [[N_VEC]]
+; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; SINK-AFTER:       scalar.ph:
-; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 96, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
 ; SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
 ; SINK-AFTER:       for.body:
@@ -1391,7 +1396,7 @@ define i32 @extract_second_last_iteration(ptr %cval, i32 %x)  {
 ; SINK-AFTER-NEXT:    [[INC]] = add i32 [[INC_PHI]], 1
 ; SINK-AFTER-NEXT:    [[BC:%.*]] = zext i32 [[INC_PHI]] to i64
 ; SINK-AFTER-NEXT:    [[ADDX]] = add i32 [[INC_PHI]], [[X]]
-; SINK-AFTER-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC_PHI]], 95
+; SINK-AFTER-NEXT:    [[CMP:%.*]] = icmp eq i32 [[INC_PHI]], [[N]]
 ; SINK-AFTER-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; SINK-AFTER:       for.end:
 ; SINK-AFTER-NEXT:    [[VAL_PHI_LCSSA:%.*]] = phi i32 [ [[VAL_PHI]], [[FOR_BODY]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], [[MIDDLE_BLOCK]] ]
@@ -1406,7 +1411,7 @@ for.body:
   %inc = add i32 %inc.phi, 1
   %bc = zext i32 %inc.phi to i64
   %addx = add i32 %inc.phi, %x
-  %cmp = icmp eq i32 %inc.phi, 95
+  %cmp = icmp eq i32 %inc.phi, %n
   br i1 %cmp, label %for.end, label %for.body
 
 for.end:
@@ -1431,13 +1436,21 @@ for.end:
 ; }
 ;
 ;
-define i32 @PR33613(ptr %b, double %j, i32 %d) {
+define i32 @PR33613(ptr %b, double %j, i32 %d, i32 %n) {
 ; UNROLL-NO-IC-LABEL: @PR33613(
 ; UNROLL-NO-IC-NEXT:  entry:
 ; UNROLL-NO-IC-NEXT:    [[IDXPROM:%.*]] = sext i32 [[D:%.*]] to i64
-; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC-NEXT:    [[TMP42:%.*]] = add i32 [[N:%.*]], -1
+; UNROLL-NO-IC-NEXT:    [[TMP45:%.*]] = zext i32 [[TMP42]] to i64
+; UNROLL-NO-IC-NEXT:    [[TMP46:%.*]] = add nuw nsw i64 [[TMP45]], 1
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP46]], 8
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; UNROLL-NO-IC:       vector.ph:
-; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = getelementptr i8, ptr [[B:%.*]], i64 2048000
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP46]], 8
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP46]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[TMP47:%.*]] = mul i64 [[N_VEC]], 200
+; UNROLL-NO-IC-NEXT:    [[IND_END:%.*]] = getelementptr i8, ptr [[B:%.*]], i64 [[TMP47]]
+; UNROLL-NO-IC-NEXT:    [[TMP49:%.*]] = trunc i64 [[N_VEC]] to i32
 ; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x double> poison, double [[J:%.*]], i32 3
 ; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; UNROLL-NO-IC:       vector.body:
@@ -1497,15 +1510,16 @@ define i32 @PR33613(ptr %b, double %j, i32 %d) {
 ; UNROLL-NO-IC-NEXT:    [[TMP40]] = add <4 x i32> [[VEC_PHI]], [[TMP38]]
 ; UNROLL-NO-IC-NEXT:    [[TMP41]] = add <4 x i32> [[VEC_PHI2]], [[TMP39]]
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
-; UNROLL-NO-IC-NEXT:    [[TMP42:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10240
-; UNROLL-NO-IC-NEXT:    br i1 [[TMP42]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; UNROLL-NO-IC-NEXT:    [[TMP48:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP48]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; UNROLL-NO-IC:       middle.block:
 ; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[TMP41]], [[TMP40]]
 ; UNROLL-NO-IC-NEXT:    [[TMP43:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
-; UNROLL-NO-IC-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP46]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
 ; UNROLL-NO-IC:       scalar.ph:
 ; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi ptr [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[B]], [[ENTRY:%.*]] ]
-; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 10240, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[TMP49]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
 ; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP43]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
 ; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi double [ [[TMP27]], [[MIDDLE_BLOCK]] ], [ [[J]], [[ENTRY]] ]
 ; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
@@ -1525,15 +1539,23 @@ define i32 @PR33613(ptr %b, double %j, i32 %d) {
 ; UNROLL-NO-IC-NEXT:    [[A_1]] = add nsw i32 [[A_010]], [[INC]]
 ; UNROLL-NO-IC-NEXT:    [[INC1]] = add nuw nsw i32 [[I_011]], 1
 ; UNROLL-NO-IC-NEXT:    [[ADD_PTR]] = getelementptr inbounds double, ptr [[B_ADDR_012]], i64 25
-; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC1]], 10240
+; UNROLL-NO-IC-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC1]], [[N]]
 ; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
 ;
 ; UNROLL-NO-VF-LABEL: @PR33613(
 ; UNROLL-NO-VF-NEXT:  entry:
 ; UNROLL-NO-VF-NEXT:    [[IDXPROM:%.*]] = sext i32 [[D:%.*]] to i64
-; UNROLL-NO-VF-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i32 [[N:%.*]], -1
+; UNROLL-NO-VF-NEXT:    [[TMP14:%.*]] = zext i32 [[TMP0]] to i64
+; UNROLL-NO-VF-NEXT:    [[TMP16:%.*]] = add nuw nsw i64 [[TMP14]], 1
+; UNROLL-NO-VF-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP16]], 2
+; UNROLL-NO-VF-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; UNROLL-NO-VF:       vector.ph:
-; UNROLL-NO-VF-NEXT:    [[IND_END:%.*]] = getelementptr i8, ptr [[B:%.*]], i64 2048000
+; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP16]], 2
+; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP16]], [[N_MOD_VF]]
+; UNROLL-NO-VF-NEXT:    [[TMP17:%.*]] = mul i64 [[N_VEC]], 200
+; UNROLL-NO-VF-NEXT:    [[IND_END:%.*]] = getelementptr i8, ptr [[B:%.*]], i64 [[TMP17]]
+; UNROLL-NO-VF-NEXT:    [[TMP18:%.*]] = trunc i64 [[N_VEC]] to i32
 ; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; UNROLL-NO-VF:       vector.body:
 ; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
@@ -1557,14 +1579,15 @@ define i32 @PR33613(ptr %b, double %j, i32 %d) {
 ; UNROLL-NO-VF-NEXT:    [[TMP12]] = add i32 [[VEC_PHI]], [[TMP10]]
 ; UNROLL-NO-VF-NEXT:    [[TMP13]] = add i32 [[VEC_PHI2]], [[TMP11]]
 ; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; UNROLL-NO-VF-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10240
-; UNROLL-NO-VF-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; UNROLL-NO-VF-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; UNROLL-NO-VF:       middle.block:
 ; UNROLL-NO-VF-NEXT:    [[BIN_RDX:%.*]] = add i32 [[TMP13]], [[TMP12]]
-; UNROLL-NO-VF-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP16]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
 ; UNROLL-NO-VF:       scalar.ph:
 ; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi ptr [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[B]], [[ENTRY:%.*]] ]
-; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 10240, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[TMP18]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
 ; UNROLL-NO-VF-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[BIN_RDX]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
 ; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi double [ [[TMP5]], [[MIDDLE_BLOCK]] ], [ [[J]], [[ENTRY]] ]
 ; UNROLL-NO-VF-NEXT:    br label [[FOR_BODY:%.*]]
@@ -1584,15 +1607,23 @@ define i32 @PR33613(ptr %b, double %j, i32 %d) {
 ; UNROLL-NO-VF-NEXT:    [[A_1]] = add nsw i32 [[A_010]], [[INC]]
 ; UNROLL-NO-VF-NEXT:    [[INC1]] = add nuw nsw i32 [[I_011]], 1
 ; UNROLL-NO-VF-NEXT:    [[ADD_PTR]] = getelementptr inbounds double, ptr [[B_ADDR_012]], i64 25
-; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC1]], 10240
+; UNROLL-NO-VF-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC1]], [[N]]
 ; UNROLL-NO-VF-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
 ;
 ; SINK-AFTER-LABEL: @PR33613(
 ; SINK-AFTER-NEXT:  entry:
 ; SINK-AFTER-NEXT:    [[IDXPROM:%.*]] = sext i32 [[D:%.*]] to i64
-; SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; SINK-AFTER-NEXT:    [[TMP21:%.*]] = add i32 [[N:%.*]], -1
+; SINK-AFTER-NEXT:    [[TMP24:%.*]] = zext i32 [[TMP21]] to i64
+; SINK-AFTER-NEXT:    [[TMP25:%.*]] = add nuw nsw i64 [[TMP24]], 1
+; SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP25]], 4
+; SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; SINK-AFTER:       vector.ph:
-; SINK-AFTER-NEXT:    [[IND_END:%.*]] = getelementptr i8, ptr [[B:%.*]], i64 2048000
+; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP25]], 4
+; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP25]], [[N_MOD_VF]]
+; SINK-AFTER-NEXT:    [[TMP26:%.*]] = mul i64 [[N_VEC]], 200
+; SINK-AFTER-NEXT:    [[IND_END:%.*]] = getelementptr i8, ptr [[B:%.*]], i64 [[TMP26]]
+; SINK-AFTER-NEXT:    [[TMP28:%.*]] = trunc i64 [[N_VEC]] to i32
 ; SINK-AFTER-NEXT:    [[VECTOR_RECUR_INIT:%.*]] = insertelement <4 x double> poison, double [[J:%.*]], i32 3
 ; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; SINK-AFTER:       vector.body:
@@ -1626,14 +1657,15 @@ define i32 @PR33613(ptr %b, double %j, i32 %d) {
 ; SINK-AFTER-NEXT:    [[TMP19:%.*]] = zext <4 x i1> [[TMP18]] to <4 x i32>
 ; SINK-AFTER-NEXT:    [[TMP20]] = add <4 x i32> [[VEC_PHI]], [[TMP19]]
 ; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; SINK-AFTER-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], 10240
-; SINK-AFTER-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; SINK-AFTER-NEXT:    [[TMP27:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; SINK-AFTER-NEXT:    br i1 [[TMP27]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; SINK-AFTER:       middle.block:
 ; SINK-AFTER-NEXT:    [[TMP22:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP20]])
-; SINK-AFTER-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP25]], [[N_VEC]]
+; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
 ; SINK-AFTER:       scalar.ph:
 ; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi ptr [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[B]], [[ENTRY:%.*]] ]
-; SINK-AFTER-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 10240, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[TMP28]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
 ; SINK-AFTER-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP22]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
 ; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi double [ [[TMP11]], [[MIDDLE_BLOCK]] ], [ [[J]], [[ENTRY]] ]
 ; SINK-AFTER-NEXT:    br label [[FOR_BODY:%.*]]
@@ -1653,7 +1685,7 @@ define i32 @PR33613(ptr %b, double %j, i32 %d) {
 ; SINK-AFTER-NEXT:    [[A_1]] = add nsw i32 [[A_010]], [[INC]]
 ; SINK-AFTER-NEXT:    [[INC1]] = add nuw nsw i32 [[I_011]], 1
 ; SINK-AFTER-NEXT:    [[ADD_PTR]] = getelementptr inbounds double, ptr [[B_ADDR_012]], i64 25
-; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC1]], 10240
+; SINK-AFTER-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[INC1]], [[N]]
 ; SINK-AFTER-NEXT:    br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP]], label [[FOR_BODY]], !llvm.loop [[LOOP17:![0-9]+]]
 ;
 entry:
@@ -1677,7 +1709,7 @@ for.body:
   %a.1 = add nsw i32 %a.010, %inc
   %inc1 = add nuw nsw i32 %i.011, 1
   %add.ptr = getelementptr inbounds double, ptr %b.addr.012, i64 25
-  %exitcond = icmp eq i32 %inc1, 10240
+  %exitcond = icmp eq i32 %inc1, %n
   br i1 %exitcond, label %for.cond.cleanup, label %for.body
 }
 
@@ -1704,9 +1736,8 @@ define void @sink_after(ptr noalias %a, ptr noalias %b, i64 %n) {
 ; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD1:%.*]], [[VECTOR_BODY]] ]
 ; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add nuw nsw i64 [[INDEX]], 1
 ; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[TMP1]]
-; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, ptr [[TMP2]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, ptr [[TMP2]], i32 4
-; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP3]], align 2
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP2]], align 2
 ; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD1]] = load <4 x i16>, ptr [[TMP4]], align 2
 ; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = shufflevector <4 x i16> [[WIDE_LOAD]], <4 x i16> [[WIDE_LOAD1]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
@@ -1717,9 +1748,8 @@ define void @sink_after(ptr noalias %a, ptr noalias %b, i64 %n) {
 ; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = mul nsw <4 x i32> [[TMP9]], [[TMP7]]
 ; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = mul nsw <4 x i32> [[TMP10]], [[TMP8]]
 ; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
-; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 4
-; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP11]], ptr [[TMP14]], align 4
+; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP11]], ptr [[TMP13]], align 4
 ; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP12]], ptr [[TMP15]], align 4
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -1818,15 +1848,13 @@ define void @sink_after(ptr noalias %a, ptr noalias %b, i64 %n) {
 ; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
 ; SINK-AFTER-NEXT:    [[TMP1:%.*]] = add nuw nsw i64 [[INDEX]], 1
 ; SINK-AFTER-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[TMP1]]
-; SINK-AFTER-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, ptr [[TMP2]], i32 0
-; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP3]], align 2
+; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP2]], align 2
 ; SINK-AFTER-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; SINK-AFTER-NEXT:    [[TMP5:%.*]] = sext <4 x i16> [[TMP4]] to <4 x i32>
 ; SINK-AFTER-NEXT:    [[TMP6:%.*]] = sext <4 x i16> [[WIDE_LOAD]] to <4 x i32>
 ; SINK-AFTER-NEXT:    [[TMP7:%.*]] = mul nsw <4 x i32> [[TMP6]], [[TMP5]]
 ; SINK-AFTER-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
-; SINK-AFTER-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
-; SINK-AFTER-NEXT:    store <4 x i32> [[TMP7]], ptr [[TMP9]], align 4
+; SINK-AFTER-NEXT:    store <4 x i32> [[TMP7]], ptr [[TMP8]], align 4
 ; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; SINK-AFTER-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; SINK-AFTER-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
@@ -1923,9 +1951,8 @@ define void @PR34711(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %n) {
 ; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = getelementptr inbounds [2 x i16], ptr [[A]], i64 [[TMP5]], i64 1
 ; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [2 x i16], ptr [[A]], i64 [[TMP6]], i64 1
 ; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = getelementptr inbounds [2 x i16], ptr [[A]], i64 [[TMP7]], i64 1
-; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 4
-; UNROLL-NO-IC-NEXT:    store <4 x i32> splat (i32 7), ptr [[TMP17]], align 4
+; UNROLL-NO-IC-NEXT:    store <4 x i32> splat (i32 7), ptr [[TMP8]], align 4
 ; UNROLL-NO-IC-NEXT:    store <4 x i32> splat (i32 7), ptr [[TMP18]], align 4
 ; UNROLL-NO-IC-NEXT:    [[TMP19:%.*]] = load i16, ptr [[TMP9]], align 2
 ; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = load i16, ptr [[TMP10]], align 2
@@ -1952,9 +1979,8 @@ define void @PR34711(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %n) {
 ; UNROLL-NO-IC-NEXT:    [[TMP41:%.*]] = mul nsw <4 x i32> [[TMP39]], [[TMP37]]
 ; UNROLL-NO-IC-NEXT:    [[TMP42:%.*]] = mul nsw <4 x i32> [[TMP40]], [[TMP38]]
 ; UNROLL-NO-IC-NEXT:    [[TMP43:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[TMP0]]
-; UNROLL-NO-IC-NEXT:    [[TMP44:%.*]] = getelementptr inbounds i32, ptr [[TMP43]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP45:%.*]] = getelementptr inbounds i32, ptr [[TMP43]], i32 4
-; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP41]], ptr [[TMP44]], align 4
+; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP41]], ptr [[TMP43]], align 4
 ; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP42]], ptr [[TMP45]], align 4
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; UNROLL-NO-IC-NEXT:    [[TMP46:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -2065,8 +2091,7 @@ define void @PR34711(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %n) {
 ; SINK-AFTER-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [2 x i16], ptr [[A]], i64 [[TMP1]], i64 1
 ; SINK-AFTER-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [2 x i16], ptr [[A]], i64 [[TMP2]], i64 1
 ; SINK-AFTER-NEXT:    [[TMP8:%.*]] = getelementptr inbounds [2 x i16], ptr [[A]], i64 [[TMP3]], i64 1
-; SINK-AFTER-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
-; SINK-AFTER-NEXT:    store <4 x i32> splat (i32 7), ptr [[TMP9]], align 4
+; SINK-AFTER-NEXT:    store <4 x i32> splat (i32 7), ptr [[TMP4]], align 4
 ; SINK-AFTER-NEXT:    [[TMP10:%.*]] = load i16, ptr [[TMP5]], align 2
 ; SINK-AFTER-NEXT:    [[TMP11:%.*]] = load i16, ptr [[TMP6]], align 2
 ; SINK-AFTER-NEXT:    [[TMP12:%.*]] = load i16, ptr [[TMP7]], align 2
@@ -2080,8 +2105,7 @@ define void @PR34711(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %n) {
 ; SINK-AFTER-NEXT:    [[TMP20:%.*]] = sext <4 x i16> [[TMP17]] to <4 x i32>
 ; SINK-AFTER-NEXT:    [[TMP21:%.*]] = mul nsw <4 x i32> [[TMP20]], [[TMP19]]
 ; SINK-AFTER-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[TMP0]]
-; SINK-AFTER-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, ptr [[TMP22]], i32 0
-; SINK-AFTER-NEXT:    store <4 x i32> [[TMP21]], ptr [[TMP23]], align 4
+; SINK-AFTER-NEXT:    store <4 x i32> [[TMP21]], ptr [[TMP22]], align 4
 ; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; SINK-AFTER-NEXT:    [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; SINK-AFTER-NEXT:    br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
@@ -2158,9 +2182,8 @@ define void @sink_after_with_multiple_users(ptr noalias %a, ptr noalias %b, i64
 ; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD1:%.*]], [[VECTOR_BODY]] ]
 ; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add nuw nsw i64 [[INDEX]], 1
 ; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[TMP1]]
-; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, ptr [[TMP2]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, ptr [[TMP2]], i32 4
-; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP3]], align 2
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP2]], align 2
 ; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD1]] = load <4 x i16>, ptr [[TMP4]], align 2
 ; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = shufflevector <4 x i16> [[WIDE_LOAD]], <4 x i16> [[WIDE_LOAD1]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
@@ -2173,9 +2196,8 @@ define void @sink_after_with_multiple_users(ptr noalias %a, ptr noalias %b, i64
 ; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = mul nsw <4 x i32> [[TMP9]], [[TMP11]]
 ; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = mul nsw <4 x i32> [[TMP10]], [[TMP12]]
 ; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
-; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[TMP15]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP15]], i32 4
-; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP13]], ptr [[TMP16]], align 4
+; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP13]], ptr [[TMP15]], align 4
 ; UNROLL-NO-IC-NEXT:    store <4 x i32> [[TMP14]], ptr [[TMP17]], align 4
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -2278,16 +2300,14 @@ define void @sink_after_with_multiple_users(ptr noalias %a, ptr noalias %b, i64
 ; SINK-AFTER-NEXT:    [[VECTOR_RECUR:%.*]] = phi <4 x i16> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], [[VECTOR_BODY]] ]
 ; SINK-AFTER-NEXT:    [[TMP1:%.*]] = add nuw nsw i64 [[INDEX]], 1
 ; SINK-AFTER-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[TMP1]]
-; SINK-AFTER-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, ptr [[TMP2]], i32 0
-; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP3]], align 2
+; SINK-AFTER-NEXT:    [[WIDE_LOAD]] = load <4 x i16>, ptr [[TMP2]], align 2
 ; SINK-AFTER-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[WIDE_LOAD]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; SINK-AFTER-NEXT:    [[TMP5:%.*]] = sext <4 x i16> [[TMP4]] to <4 x i32>
 ; SINK-AFTER-NEXT:    [[TMP6:%.*]] = add nsw <4 x i32> [[TMP5]], splat (i32 2)
 ; SINK-AFTER-NEXT:    [[TMP7:%.*]] = sext <4 x i16> [[WIDE_LOAD]] to <4 x i32>
 ; SINK-AFTER-NEXT:    [[TMP8:%.*]] = mul nsw <4 x i32> [[TMP6]], [[TMP7]]
 ; SINK-AFTER-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
-; SINK-AFTER-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i32 0
-; SINK-AFTER-NEXT:    store <4 x i32> [[TMP8]], ptr [[TMP10]], align 4
+; SINK-AFTER-NEXT:    store <4 x i32> [[TMP8]], ptr [[TMP9]], align 4
 ; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; SINK-AFTER-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; SINK-AFTER-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
@@ -2465,9 +2485,8 @@ define void @sink_dead_inst(ptr %a) {
 ; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = sub <4 x i16> [[TMP6]], splat (i16 10)
 ; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = sub <4 x i16> [[TMP7]], splat (i16 10)
 ; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = getelementptr i16, ptr [[A:%.*]], i16 [[OFFSET_IDX]]
-; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = getelementptr i16, ptr [[TMP10]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = getelementptr i16, ptr [[TMP10]], i32 4
-; UNROLL-NO-IC-NEXT:    store <4 x i16> [[TMP8]], ptr [[TMP11]], align 2
+; UNROLL-NO-IC-NEXT:    store <4 x i16> [[TMP8]], ptr [[TMP10]], align 2
 ; UNROLL-NO-IC-NEXT:    store <4 x i16> [[TMP9]], ptr [[TMP12]], align 2
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
 ; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[STEP_ADD]], splat (i16 4)
@@ -2476,7 +2495,7 @@ define void @sink_dead_inst(ptr %a) {
 ; UNROLL-NO-IC:       middle.block:
 ; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[TMP5]], i32 3
 ; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT2:%.*]] = extractelement <4 x i32> [[TMP3]], i32 3
-; UNROLL-NO-IC-NEXT:    br i1 false, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC-NEXT:    br label [[SCALAR_PH]]
 ; UNROLL-NO-IC:       scalar.ph:
 ; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 13, [[MIDDLE_BLOCK]] ], [ -27, [[ENTRY:%.*]] ]
 ; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
@@ -2493,7 +2512,7 @@ define void @sink_dead_inst(ptr %a) {
 ; UNROLL-NO-IC-NEXT:    [[REC_1_PREV]] = add i16 [[IV_NEXT]], 5
 ; UNROLL-NO-IC-NEXT:    [[GEP:%.*]] = getelementptr i16, ptr [[A]], i16 [[IV]]
 ; UNROLL-NO-IC-NEXT:    store i16 [[USE_REC_1]], ptr [[GEP]], align 2
-; UNROLL-NO-IC-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP]], label [[FOR_END:%.*]], label [[FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
 ; UNROLL-NO-IC:       for.end:
 ; UNROLL-NO-IC-NEXT:    ret void
 ;
@@ -2523,7 +2542,7 @@ define void @sink_dead_inst(ptr %a) {
 ; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = icmp eq i32 [[INDEX_NEXT]], 42
 ; UNROLL-NO-VF-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
 ; UNROLL-NO-VF:       middle.block:
-; UNROLL-NO-VF-NEXT:    br i1 false, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF-NEXT:    br label [[SCALAR_PH]]
 ; UNROLL-NO-VF:       scalar.ph:
 ; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 15, [[MIDDLE_BLOCK]] ], [ -27, [[ENTRY:%.*]] ]
 ; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[TMP6]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
@@ -2540,7 +2559,7 @@ define void @sink_dead_inst(ptr %a) {
 ; UNROLL-NO-VF-NEXT:    [[REC_1_PREV]] = add i16 [[IV_NEXT]], 5
 ; UNROLL-NO-VF-NEXT:    [[GEP:%.*]] = getelementptr i16, ptr [[A]], i16 [[IV]]
 ; UNROLL-NO-VF-NEXT:    store i16 [[USE_REC_1]], ptr [[GEP]], align 2
-; UNROLL-NO-VF-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP]], label [[FOR_END:%.*]], label [[FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
 ; UNROLL-NO-VF:       for.end:
 ; UNROLL-NO-VF-NEXT:    ret void
 ;
@@ -2561,8 +2580,7 @@ define void @sink_dead_inst(ptr %a) {
 ; SINK-AFTER-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i16> [[VECTOR_RECUR]], <4 x i16> [[TMP3]], <4 x i32> <i32 3, i32 4, i32 5, i32 6>
 ; SINK-AFTER-NEXT:    [[TMP5:%.*]] = sub <4 x i16> [[TMP4]], splat (i16 10)
 ; SINK-AFTER-NEXT:    [[TMP6:%.*]] = getelementptr i16, ptr [[A:%.*]], i16 [[OFFSET_IDX]]
-; SINK-AFTER-NEXT:    [[TMP7:%.*]] = getelementptr i16, ptr [[TMP6]], i32 0
-; SINK-AFTER-NEXT:    store <4 x i16> [[TMP5]], ptr [[TMP7]], align 2
+; SINK-AFTER-NEXT:    store <4 x i16> [[TMP5]], ptr [[TMP6]], align 2
 ; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; SINK-AFTER-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[VEC_IND]], splat (i16 4)
 ; SINK-AFTER-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], 40
@@ -2570,7 +2588,7 @@ define void @sink_dead_inst(ptr %a) {
 ; SINK-AFTER:       middle.block:
 ; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[TMP3]], i32 3
 ; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT2:%.*]] = extractelement <4 x i32> [[TMP2]], i32 3
-; SINK-AFTER-NEXT:    br i1 false, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER-NEXT:    br label [[SCALAR_PH]]
 ; SINK-AFTER:       scalar.ph:
 ; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 13, [[MIDDLE_BLOCK]] ], [ -27, [[ENTRY:%.*]] ]
 ; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
@@ -2587,7 +2605,7 @@ define void @sink_dead_inst(ptr %a) {
 ; SINK-AFTER-NEXT:    [[REC_1_PREV]] = add i16 [[IV_NEXT]], 5
 ; SINK-AFTER-NEXT:    [[GEP:%.*]] = getelementptr i16, ptr [[A]], i16 [[IV]]
 ; SINK-AFTER-NEXT:    store i16 [[USE_REC_1]], ptr [[GEP]], align 2
-; SINK-AFTER-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
+; SINK-AFTER-NEXT:    br i1 [[CMP]], label [[FOR_END:%.*]], label [[FOR_COND]], !llvm.loop [[LOOP25:![0-9]+]]
 ; SINK-AFTER:       for.end:
 ; SINK-AFTER-NEXT:    ret void
 ;
@@ -3335,7 +3353,7 @@ bb:
 
 ; %vec.dead will be marked as dead instruction in the vector loop and no recipe
 ; will be created for it. Make sure a valid sink target is used.
-define i32 @sink_after_dead_inst(ptr %A.ptr) {
+define i32 @sink_after_dead_inst(ptr %A.ptr, i32 %n) {
 ; UNROLL-NO-IC-LABEL: @sink_after_dead_inst(
 ; UNROLL-NO-IC-NEXT:  entry:
 ; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
@@ -3350,21 +3368,19 @@ define i32 @sink_after_dead_inst(ptr %A.ptr) {
 ; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = or <4 x i16> [[TMP1]], [[TMP1]]
 ; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = zext <4 x i16> [[TMP2]] to <4 x i32>
 ; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[A_PTR:%.*]], i16 [[OFFSET_IDX]]
-; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[TMP4]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr i32, ptr [[TMP4]], i32 4
-; UNROLL-NO-IC-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP5]], align 4
+; UNROLL-NO-IC-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP4]], align 4
 ; UNROLL-NO-IC-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP6]], align 4
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
 ; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[STEP_ADD]], splat (i16 4)
 ; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], 16
 ; UNROLL-NO-IC-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP33:![0-9]+]]
 ; UNROLL-NO-IC:       middle.block:
-; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP3]], i32 3
 ; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP3]], i32 2
-; UNROLL-NO-IC-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_END:%.*]]
 ; UNROLL-NO-IC:       scalar.ph:
-; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 16, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; UNROLL-NO-IC-NEXT:    br label [[LOOP:%.*]]
 ; UNROLL-NO-IC:       loop:
 ; UNROLL-NO-IC-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -3392,25 +3408,22 @@ define i32 @sink_after_dead_inst(ptr %A.ptr) {
 ; UNROLL-NO-VF:       vector.body:
 ; UNROLL-NO-VF-NEXT:    [[VECTOR_RECUR:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[VECTOR_BODY]] ]
 ; UNROLL-NO-VF-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[VECTOR_RECUR]] to i16
-; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = add i16 [[OFFSET_IDX]], 1
 ; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = add i16 [[OFFSET_IDX]], 1
-; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = add i16 [[TMP1]], 1
-; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = or i16 [[TMP2]], [[TMP2]]
+; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = add i16 [[OFFSET_IDX]], 1
 ; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = or i16 [[TMP3]], [[TMP3]]
-; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = zext i16 [[TMP4]] to i32
 ; UNROLL-NO-VF-NEXT:    [[TMP10:%.*]] = zext i16 [[TMP5]] to i32
 ; UNROLL-NO-VF-NEXT:    [[TMP8:%.*]] = getelementptr i32, ptr [[A_PTR:%.*]], i16 [[OFFSET_IDX]]
-; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = getelementptr i32, ptr [[A_PTR]], i16 [[TMP1]]
+; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = getelementptr i32, ptr [[A_PTR]], i16 [[TMP2]]
 ; UNROLL-NO-VF-NEXT:    store i32 0, ptr [[TMP8]], align 4
 ; UNROLL-NO-VF-NEXT:    store i32 0, ptr [[TMP9]], align 4
 ; UNROLL-NO-VF-NEXT:    [[TMP7]] = add nuw i32 [[VECTOR_RECUR]], 2
 ; UNROLL-NO-VF-NEXT:    [[TMP11:%.*]] = icmp eq i32 [[TMP7]], 16
 ; UNROLL-NO-VF-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP33:![0-9]+]]
 ; UNROLL-NO-VF:       middle.block:
-; UNROLL-NO-VF-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF-NEXT:    br label [[FOR_END:%.*]]
 ; UNROLL-NO-VF:       scalar.ph:
-; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 16, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[TMP10]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; UNROLL-NO-VF-NEXT:    br label [[LOOP:%.*]]
 ; UNROLL-NO-VF:       loop:
 ; UNROLL-NO-VF-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -3427,7 +3440,7 @@ define i32 @sink_after_dead_inst(ptr %A.ptr) {
 ; UNROLL-NO-VF-NEXT:    store i32 0, ptr [[A_GEP]], align 4
 ; UNROLL-NO-VF-NEXT:    br i1 [[VEC_DEAD]], label [[FOR_END]], label [[LOOP]], !llvm.loop [[LOOP34:![0-9]+]]
 ; UNROLL-NO-VF:       for.end:
-; UNROLL-NO-VF-NEXT:    [[FOR_LCSSA:%.*]] = phi i32 [ [[FOR]], [[LOOP]] ], [ [[TMP6]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-VF-NEXT:    [[FOR_LCSSA:%.*]] = phi i32 [ [[FOR]], [[LOOP]] ], [ [[TMP10]], [[MIDDLE_BLOCK]] ]
 ; UNROLL-NO-VF-NEXT:    ret i32 [[FOR_LCSSA]]
 ;
 ; SINK-AFTER-LABEL: @sink_after_dead_inst(
@@ -3443,19 +3456,17 @@ define i32 @sink_after_dead_inst(ptr %A.ptr) {
 ; SINK-AFTER-NEXT:    [[TMP2:%.*]] = or <4 x i16> [[TMP1]], [[TMP1]]
 ; SINK-AFTER-NEXT:    [[TMP3:%.*]] = zext <4 x i16> [[TMP2]] to <4 x i32>
 ; SINK-AFTER-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[A_PTR:%.*]], i16 [[OFFSET_IDX]]
-; SINK-AFTER-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[TMP4]], i32 0
-; SINK-AFTER-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP5]], align 4
+; SINK-AFTER-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP4]], align 4
 ; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; SINK-AFTER-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[VEC_IND]], splat (i16 4)
 ; SINK-AFTER-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 16
 ; SINK-AFTER-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP33:![0-9]+]]
 ; SINK-AFTER:       middle.block:
-; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP3]], i32 3
 ; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP3]], i32 2
-; SINK-AFTER-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER-NEXT:    br label [[FOR_END:%.*]]
 ; SINK-AFTER:       scalar.ph:
-; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 16, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, [[ENTRY:%.*]] ]
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; SINK-AFTER-NEXT:    br label [[LOOP:%.*]]
 ; SINK-AFTER:       loop:
 ; SINK-AFTER-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -3500,11 +3511,19 @@ for.end:
 
 ; %rec.1 only has %use.rec.1 as use, which can be removed. This enables %rec.1
 ; to be removed also.
-define void @unused_recurrence(ptr %a) {
+define void @unused_recurrence(ptr %a, i16 %n) {
 ; UNROLL-NO-IC-LABEL: @unused_recurrence(
 ; UNROLL-NO-IC-NEXT:  entry:
-; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = add i16 [[N:%.*]], 27
+; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = zext i16 [[TMP4]] to i32
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = add nuw nsw i32 [[TMP5]], 1
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP2]], 8
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[TMP2]], 8
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i32 [[TMP2]], [[N_MOD_VF]]
+; UNROLL-NO-IC-NEXT:    [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i16
+; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = add i16 -27, [[DOTCAST]]
 ; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; UNROLL-NO-IC:       vector.body:
 ; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
@@ -3514,13 +3533,14 @@ define void @unused_recurrence(ptr %a) {
 ; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = add <4 x i16> [[TMP0]], splat (i16 5)
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 8
 ; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[STEP_ADD]], splat (i16 4)
-; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1024
-; UNROLL-NO-IC-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP35:![0-9]+]]
+; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP35:![0-9]+]]
 ; UNROLL-NO-IC:       middle.block:
 ; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[TMP1]], i32 3
-; UNROLL-NO-IC-NEXT:    br i1 false, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP2]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; UNROLL-NO-IC:       scalar.ph:
-; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 997, [[MIDDLE_BLOCK]] ], [ -27, [[ENTRY:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ [[TMP3]], [[MIDDLE_BLOCK]] ], [ -27, [[ENTRY:%.*]] ]
 ; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
 ; UNROLL-NO-IC-NEXT:    br label [[FOR_COND:%.*]]
 ; UNROLL-NO-IC:       for.cond:
@@ -3529,30 +3549,39 @@ define void @unused_recurrence(ptr %a) {
 ; UNROLL-NO-IC-NEXT:    [[USE_REC_1:%.*]] = sub i16 [[REC_1]], 10
 ; UNROLL-NO-IC-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
 ; UNROLL-NO-IC-NEXT:    [[REC_1_PREV]] = add i16 [[IV_NEXT]], 5
-; UNROLL-NO-IC-NEXT:    [[CMP:%.*]] = icmp eq i16 [[IV]], 1000
+; UNROLL-NO-IC-NEXT:    [[CMP:%.*]] = icmp eq i16 [[IV]], [[N]]
 ; UNROLL-NO-IC-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
 ; UNROLL-NO-IC:       for.end:
 ; UNROLL-NO-IC-NEXT:    ret void
 ;
 ; UNROLL-NO-VF-LABEL: @unused_recurrence(
 ; UNROLL-NO-VF-NEXT:  entry:
-; UNROLL-NO-VF-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-VF-NEXT:    [[TMP4:%.*]] = add i16 [[N:%.*]], 27
+; UNROLL-NO-VF-NEXT:    [[TMP5:%.*]] = zext i16 [[TMP4]] to i32
+; UNROLL-NO-VF-NEXT:    [[TMP6:%.*]] = add nuw nsw i32 [[TMP5]], 1
+; UNROLL-NO-VF-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP6]], 2
+; UNROLL-NO-VF-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; UNROLL-NO-VF:       vector.ph:
+; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[TMP6]], 2
+; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i32 [[TMP6]], [[N_MOD_VF]]
+; UNROLL-NO-VF-NEXT:    [[DOTCAST1:%.*]] = trunc i32 [[N_VEC]] to i16
+; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = add i16 -27, [[DOTCAST1]]
 ; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; UNROLL-NO-VF:       vector.body:
 ; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; UNROLL-NO-VF-NEXT:    [[DOTCAST:%.*]] = trunc i32 [[INDEX]] to i16
-; UNROLL-NO-VF-NEXT:    [[OFFSET_IDX:%.*]] = add i16 -27, [[DOTCAST]]
-; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = add i16 [[OFFSET_IDX]], 1
-; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = add i16 [[TMP0]], 1
-; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = add i16 [[TMP1]], 5
+; UNROLL-NO-VF-NEXT:    [[DOTCAST2:%.*]] = trunc i32 [[INDEX]] to i16
+; UNROLL-NO-VF-NEXT:    [[OFFSET_IDX:%.*]] = add i16 -27, [[DOTCAST2]]
+; UNROLL-NO-VF-NEXT:    [[TMP8:%.*]] = add i16 [[OFFSET_IDX]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP9:%.*]] = add i16 [[TMP8]], 1
+; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = add i16 [[TMP9]], 5
 ; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
-; UNROLL-NO-VF-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1028
-; UNROLL-NO-VF-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP35:![0-9]+]]
+; UNROLL-NO-VF-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP35:![0-9]+]]
 ; UNROLL-NO-VF:       middle.block:
-; UNROLL-NO-VF-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP6]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; UNROLL-NO-VF:       scalar.ph:
-; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 1001, [[MIDDLE_BLOCK]] ], [ -27, [[ENTRY:%.*]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ [[TMP3]], [[MIDDLE_BLOCK]] ], [ -27, [[ENTRY:%.*]] ]
 ; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[TMP2]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
 ; UNROLL-NO-VF-NEXT:    br label [[FOR_COND:%.*]]
 ; UNROLL-NO-VF:       for.cond:
@@ -3561,30 +3590,39 @@ define void @unused_recurrence(ptr %a) {
 ; UNROLL-NO-VF-NEXT:    [[USE_REC_1:%.*]] = sub i16 [[REC_1]], 10
 ; UNROLL-NO-VF-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
 ; UNROLL-NO-VF-NEXT:    [[REC_1_PREV]] = add i16 [[IV_NEXT]], 5
-; UNROLL-NO-VF-NEXT:    [[CMP:%.*]] = icmp eq i16 [[IV]], 1000
+; UNROLL-NO-VF-NEXT:    [[CMP:%.*]] = icmp eq i16 [[IV]], [[N]]
 ; UNROLL-NO-VF-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
 ; UNROLL-NO-VF:       for.end:
 ; UNROLL-NO-VF-NEXT:    ret void
 ;
 ; SINK-AFTER-LABEL: @unused_recurrence(
 ; SINK-AFTER-NEXT:  entry:
-; SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; SINK-AFTER-NEXT:    [[TMP4:%.*]] = add i16 [[N:%.*]], 27
+; SINK-AFTER-NEXT:    [[TMP5:%.*]] = zext i16 [[TMP4]] to i32
+; SINK-AFTER-NEXT:    [[TMP2:%.*]] = add nuw nsw i32 [[TMP5]], 1
+; SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i32 [[TMP2]], 4
+; SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; SINK-AFTER:       vector.ph:
+; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[TMP2]], 4
+; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i32 [[TMP2]], [[N_MOD_VF]]
+; SINK-AFTER-NEXT:    [[DOTCAST:%.*]] = trunc i32 [[N_VEC]] to i16
+; SINK-AFTER-NEXT:    [[TMP3:%.*]] = add i16 -27, [[DOTCAST]]
 ; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; SINK-AFTER:       vector.body:
 ; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; SINK-AFTER-NEXT:    [[VEC_IND:%.*]] = phi <4 x i16> [ <i16 -27, i16 -26, i16 -25, i16 -24>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
-; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add <4 x i16> [[VEC_IND]], splat (i16 1)
-; SINK-AFTER-NEXT:    [[TMP1:%.*]] = add <4 x i16> [[TMP0]], splat (i16 5)
+; SINK-AFTER-NEXT:    [[TMP7:%.*]] = add <4 x i16> [[VEC_IND]], splat (i16 1)
+; SINK-AFTER-NEXT:    [[TMP1:%.*]] = add <4 x i16> [[TMP7]], splat (i16 5)
 ; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; SINK-AFTER-NEXT:    [[VEC_IND_NEXT]] = add <4 x i16> [[VEC_IND]], splat (i16 4)
-; SINK-AFTER-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1028
-; SINK-AFTER-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP35:![0-9]+]]
+; SINK-AFTER-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; SINK-AFTER-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP35:![0-9]+]]
 ; SINK-AFTER:       middle.block:
 ; SINK-AFTER-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i16> [[TMP1]], i32 3
-; SINK-AFTER-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP2]], [[N_VEC]]
+; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
 ; SINK-AFTER:       scalar.ph:
-; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 1001, [[MIDDLE_BLOCK]] ], [ -27, [[ENTRY:%.*]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ [[TMP3]], [[MIDDLE_BLOCK]] ], [ -27, [[ENTRY:%.*]] ]
 ; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i16 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
 ; SINK-AFTER-NEXT:    br label [[FOR_COND:%.*]]
 ; SINK-AFTER:       for.cond:
@@ -3593,7 +3631,7 @@ define void @unused_recurrence(ptr %a) {
 ; SINK-AFTER-NEXT:    [[USE_REC_1:%.*]] = sub i16 [[REC_1]], 10
 ; SINK-AFTER-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
 ; SINK-AFTER-NEXT:    [[REC_1_PREV]] = add i16 [[IV_NEXT]], 5
-; SINK-AFTER-NEXT:    [[CMP:%.*]] = icmp eq i16 [[IV]], 1000
+; SINK-AFTER-NEXT:    [[CMP:%.*]] = icmp eq i16 [[IV]], [[N]]
 ; SINK-AFTER-NEXT:    br i1 [[CMP]], label [[FOR_END]], label [[FOR_COND]], !llvm.loop [[LOOP36:![0-9]+]]
 ; SINK-AFTER:       for.end:
 ; SINK-AFTER-NEXT:    ret void
@@ -3607,7 +3645,7 @@ for.cond:
   %use.rec.1 = sub i16 %rec.1, 10
   %iv.next= add i16 %iv, 1
   %rec.1.prev = add i16 %iv.next, 5
-  %cmp = icmp eq i16 %iv, 1000
+  %cmp = icmp eq i16 %iv, %n
   br i1 %cmp, label %for.end, label %for.cond
 
 for.end:
@@ -3615,42 +3653,59 @@ for.end:
 }
 
 ; Test case for https://github.com/llvm/llvm-project/issues/95520.
-define i32 @recurence_uniform_load(ptr %src, ptr noalias %dst) {
+define i32 @recurence_uniform_load(ptr %src, ptr noalias %dst, i64 %n) {
 ; UNROLL-NO-IC-LABEL: @recurence_uniform_load(
 ; UNROLL-NO-IC-NEXT:  entry:
-; UNROLL-NO-IC-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = add i64 [[N:%.*]], 1
+; UNROLL-NO-IC-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 8
+; UNROLL-NO-IC-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; UNROLL-NO-IC:       vector.ph:
+; UNROLL-NO-IC-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP0]], 8
+; UNROLL-NO-IC-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF]]
 ; UNROLL-NO-IC-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; UNROLL-NO-IC:       vector.body:
-; UNROLL-NO-IC-NEXT:    br label [[MIDDLE_BLOCK:%.*]]
+; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = load i32, ptr [[SRC:%.*]], align 4
+; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP37:![0-9]+]]
 ; UNROLL-NO-IC:       middle.block:
-; UNROLL-NO-IC-NEXT:    br label [[EXIT:%.*]]
+; UNROLL-NO-IC-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
+; UNROLL-NO-IC-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
 ; UNROLL-NO-IC:       scalar.ph:
-; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
-; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[TMP1]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
 ; UNROLL-NO-IC-NEXT:    br label [[LOOP:%.*]]
 ; UNROLL-NO-IC:       loop:
 ; UNROLL-NO-IC-NEXT:    [[PHI:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[LOOP]] ]
 ; UNROLL-NO-IC-NEXT:    [[RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[LOAD:%.*]], [[LOOP]] ]
 ; UNROLL-NO-IC-NEXT:    [[ADD]] = add i64 [[PHI]], 1
-; UNROLL-NO-IC-NEXT:    [[LOAD]] = load i32, ptr [[SRC:%.*]], align 4
-; UNROLL-NO-IC-NEXT:    [[ICMP:%.*]] = icmp ult i64 [[PHI]], 1
-; UNROLL-NO-IC-NEXT:    br i1 [[ICMP]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP37:![0-9]+]]
+; UNROLL-NO-IC-NEXT:    [[LOAD]] = load i32, ptr [[SRC]], align 4
+; UNROLL-NO-IC-NEXT:    [[ICMP:%.*]] = icmp ult i64 [[PHI]], [[N]]
+; UNROLL-NO-IC-NEXT:    br i1 [[ICMP]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP38:![0-9]+]]
 ; UNROLL-NO-IC:       exit:
 ; UNROLL-NO-IC-NEXT:    ret i32 0
 ;
 ; UNROLL-NO-VF-LABEL: @recurence_uniform_load(
 ; UNROLL-NO-VF-NEXT:  entry:
-; UNROLL-NO-VF-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; UNROLL-NO-VF-NEXT:    [[TMP1:%.*]] = add i64 [[N:%.*]], 1
+; UNROLL-NO-VF-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP1]], 2
+; UNROLL-NO-VF-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; UNROLL-NO-VF:       vector.ph:
+; UNROLL-NO-VF-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP1]], 2
+; UNROLL-NO-VF-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP1]], [[N_MOD_VF]]
 ; UNROLL-NO-VF-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; UNROLL-NO-VF:       vector.body:
+; UNROLL-NO-VF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; UNROLL-NO-VF-NEXT:    [[TMP0:%.*]] = load i32, ptr [[SRC:%.*]], align 4
-; UNROLL-NO-VF-NEXT:    br label [[MIDDLE_BLOCK:%.*]]
+; UNROLL-NO-VF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; UNROLL-NO-VF-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP37:![0-9]+]]
 ; UNROLL-NO-VF:       middle.block:
-; UNROLL-NO-VF-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP1]], [[N_VEC]]
+; UNROLL-NO-VF-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
 ; UNROLL-NO-VF:       scalar.ph:
-; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 2, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; UNROLL-NO-VF-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[TMP0]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
 ; UNROLL-NO-VF-NEXT:    br label [[LOOP:%.*]]
 ; UNROLL-NO-VF:       loop:
@@ -3658,31 +3713,40 @@ define i32 @recurence_uniform_load(ptr %src, ptr noalias %dst) {
 ; UNROLL-NO-VF-NEXT:    [[RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[LOAD:%.*]], [[LOOP]] ]
 ; UNROLL-NO-VF-NEXT:    [[ADD]] = add i64 [[PHI]], 1
 ; UNROLL-NO-VF-NEXT:    [[LOAD]] = load i32, ptr [[SRC]], align 4
-; UNROLL-NO-VF-NEXT:    [[ICMP:%.*]] = icmp ult i64 [[PHI]], 1
-; UNROLL-NO-VF-NEXT:    br i1 [[ICMP]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP37:![0-9]+]]
+; UNROLL-NO-VF-NEXT:    [[ICMP:%.*]] = icmp ult i64 [[PHI]], [[N]]
+; UNROLL-NO-VF-NEXT:    br i1 [[ICMP]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP38:![0-9]+]]
 ; UNROLL-NO-VF:       exit:
 ; UNROLL-NO-VF-NEXT:    ret i32 0
 ;
 ; SINK-AFTER-LABEL: @recurence_uniform_load(
 ; SINK-AFTER-NEXT:  entry:
-; SINK-AFTER-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; SINK-AFTER-NEXT:    [[TMP0:%.*]] = add i64 [[N:%.*]], 1
+; SINK-AFTER-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP0]], 4
+; SINK-AFTER-NEXT:    br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; SINK-AFTER:       vector.ph:
+; SINK-AFTER-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP0]], 4
+; SINK-AFTER-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP0]], [[N_MOD_VF]]
 ; SINK-AFTER-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; SINK-AFTER:       vector.body:
-; SINK-AFTER-NEXT:    br label [[MIDDLE_BLOCK:%.*]]
+; SINK-AFTER-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; SINK-AFTER-NEXT:    [[TMP1:%.*]] = load i32, ptr [[SRC:%.*]], align 4
+; SINK-AFTER-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; SINK-AFTER-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; SINK-AFTER-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP37:![0-9]+]]
 ; SINK-AFTER:       middle.block:
-; SINK-AFTER-NEXT:    br label [[EXIT:%.*]]
+; SINK-AFTER-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP0]], [[N_VEC]]
+; SINK-AFTER-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
 ; SINK-AFTER:       scalar.ph:
-; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
-; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY]] ]
+; SINK-AFTER-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; SINK-AFTER-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[TMP1]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
 ; SINK-AFTER-NEXT:    br label [[LOOP:%.*]]
 ; SINK-AFTER:       loop:
 ; SINK-AFTER-NEXT:    [[PHI:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[LOOP]] ]
 ; SINK-AFTER-NEXT:    [[RECUR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ], [ [[LOAD:%.*]], [[LOOP]] ]
 ; SINK-AFTER-NEXT:    [[ADD]] = add i64 [[PHI]], 1
-; SINK-AFTER-NEXT:    [[LOAD]] = load i32, ptr [[SRC:%.*]], align 4
-; SINK-AFTER-NEXT:    [[ICMP:%.*]] = icmp ult i64 [[PHI]], 1
-; SINK-AFTER-NEXT:    br i1 [[ICMP]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP37:![0-9]+]]
+; SINK-AFTER-NEXT:    [[LOAD]] = load i32, ptr [[SRC]], align 4
+; SINK-AFTER-NEXT:    [[ICMP:%.*]] = icmp ult i64 [[PHI]], [[N]]
+; SINK-AFTER-NEXT:    br i1 [[ICMP]], label [[LOOP]], label [[EXIT]], !llvm.loop [[LOOP38:![0-9]+]]
 ; SINK-AFTER:       exit:
 ; SINK-AFTER-NEXT:    ret i32 0
 ;
@@ -3694,7 +3758,7 @@ loop:
   %recur = phi i32 [ 0, %entry ], [ %load, %loop ]
   %add = add i64 %phi, 1
   %load = load i32, ptr %src, align 4
-  %icmp = icmp ult i64 %phi, 1
+  %icmp = icmp ult i64 %phi, %n
   br i1 %icmp, label %loop, label %exit
 
 exit:
diff --git a/llvm/test/Transforms/LoopVectorize/float-induction.ll b/llvm/test/Transforms/LoopVectorize/float-induction.ll
index fb6e6be..8a3cad0 100644
--- a/llvm/test/Transforms/LoopVectorize/float-induction.ll
+++ b/llvm/test/Transforms/LoopVectorize/float-induction.ll
@@ -1655,7 +1655,7 @@ define i32 @float_induction_with_dbg_on_fadd(ptr %dst) {
 ; VEC4_INTERL1-NEXT:    [[TMP1:%.*]] = icmp eq i64 [[INDEX_NEXT]], 200
 ; VEC4_INTERL1-NEXT:    br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; VEC4_INTERL1:       middle.block:
-; VEC4_INTERL1-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VEC4_INTERL1-NEXT:    br label [[EXIT:%.*]]
 ; VEC4_INTERL1:       scalar.ph:
 ; VEC4_INTERL1-NEXT:    br label [[LOOP:%.*]]
 ; VEC4_INTERL1:       loop:
@@ -1678,7 +1678,7 @@ define i32 @float_induction_with_dbg_on_fadd(ptr %dst) {
 ; VEC4_INTERL2-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], 200
 ; VEC4_INTERL2-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; VEC4_INTERL2:       middle.block:
-; VEC4_INTERL2-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VEC4_INTERL2-NEXT:    br label [[EXIT:%.*]]
 ; VEC4_INTERL2:       scalar.ph:
 ; VEC4_INTERL2-NEXT:    br label [[LOOP:%.*]]
 ; VEC4_INTERL2:       loop:
@@ -1705,7 +1705,7 @@ define i32 @float_induction_with_dbg_on_fadd(ptr %dst) {
 ; VEC1_INTERL2-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 200
 ; VEC1_INTERL2-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; VEC1_INTERL2:       middle.block:
-; VEC1_INTERL2-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VEC1_INTERL2-NEXT:    br label [[EXIT:%.*]]
 ; VEC1_INTERL2:       scalar.ph:
 ; VEC1_INTERL2-NEXT:    br label [[LOOP:%.*]]
 ; VEC1_INTERL2:       loop:
diff --git a/llvm/test/Transforms/LoopVectorize/float-minmax-instruction-flag.ll b/llvm/test/Transforms/LoopVectorize/float-minmax-instruction-flag.ll
index d01984d..2c02f83 100644
--- a/llvm/test/Transforms/LoopVectorize/float-minmax-instruction-flag.ll
+++ b/llvm/test/Transforms/LoopVectorize/float-minmax-instruction-flag.ll
@@ -58,8 +58,7 @@ define float @minloopattr(ptr nocapture readonly %arg) #0 {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ [[MINMAX_IDENT_SPLAT]], [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr float, ptr [[ARG]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = fcmp olt <4 x float> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x float> [[VEC_PHI]], <4 x float> [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -67,10 +66,10 @@ define float @minloopattr(ptr nocapture readonly %arg) #0 {
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP6:%.*]] = call float @llvm.vector.reduce.fmin.v4f32(<4 x float> [[TMP4]])
-; CHECK-NEXT:    br i1 true, label [[OUT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[OUT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 65537, [[MIDDLE_BLOCK]] ], [ 1, [[TOP:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP6]], [[MIDDLE_BLOCK]] ], [ [[T]], [[TOP]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1, [[TOP:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[T]], [[TOP]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[T1:%.*]] = phi i64 [ [[T7:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/fmax-without-fast-math-flags-interleave.ll b/llvm/test/Transforms/LoopVectorize/fmax-without-fast-math-flags-interleave.ll
index a2eddad..720ea1f 100644
--- a/llvm/test/Transforms/LoopVectorize/fmax-without-fast-math-flags-interleave.ll
+++ b/llvm/test/Transforms/LoopVectorize/fmax-without-fast-math-flags-interleave.ll
@@ -53,9 +53,8 @@ define float @fmaxnum(ptr %src, i64 %n) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ splat (float -1.000000e+07), %[[VECTOR_PH]] ], [ [[TMP7:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x float> [ splat (float -1.000000e+07), %[[VECTOR_PH]] ], [ [[TMP8:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw float, ptr [[GEP_SRC]], i32 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw float, ptr [[GEP_SRC]], i32 4
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[GEP_SRC]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP7]] = call <4 x float> @llvm.maxnum.v4f32(<4 x float> [[VEC_PHI]], <4 x float> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[TMP8]] = call <4 x float> @llvm.maxnum.v4f32(<4 x float> [[VEC_PHI1]], <4 x float> [[WIDE_LOAD2]])
diff --git a/llvm/test/Transforms/LoopVectorize/fmax-without-fast-math-flags.ll b/llvm/test/Transforms/LoopVectorize/fmax-without-fast-math-flags.ll
index 1ca5586..3ef37bc 100644
--- a/llvm/test/Transforms/LoopVectorize/fmax-without-fast-math-flags.ll
+++ b/llvm/test/Transforms/LoopVectorize/fmax-without-fast-math-flags.ll
@@ -202,8 +202,7 @@ define float @fmaxnum_1(ptr %src, i64 %n) {
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ splat (float -1.000000e+07), %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw float, ptr [[GEP_SRC]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[GEP_SRC]], align 4
 ; CHECK-NEXT:    [[TMP4]] = call <4 x float> @llvm.maxnum.v4f32(<4 x float> [[WIDE_LOAD]], <4 x float> [[VEC_PHI]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -270,8 +269,7 @@ define float @fmaxnum_2(ptr %src, i64 %n) {
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ splat (float -1.000000e+07), %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw float, ptr [[GEP_SRC]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[GEP_SRC]], align 4
 ; CHECK-NEXT:    [[TMP4]] = call <4 x float> @llvm.maxnum.v4f32(<4 x float> [[VEC_PHI]], <4 x float> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -340,8 +338,7 @@ define float @fmaxnum_induction_starts_at_10(ptr %src, i64 %n) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ splat (float -1.000000e+07), %[[VECTOR_PH]] ], [ [[TMP3:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[IV:%.*]] = add i64 10, [[INDEX]]
 ; CHECK-NEXT:    [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw float, ptr [[GEP_SRC]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[GEP_SRC]], align 4
 ; CHECK-NEXT:    [[TMP3]] = call <4 x float> @llvm.maxnum.v4f32(<4 x float> [[WIDE_LOAD]], <4 x float> [[VEC_PHI]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -411,8 +408,7 @@ define float @fmaxnum_induction_starts_at_value(ptr %src, i64 %start, i64 %n) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ splat (float -1.000000e+07), %[[VECTOR_PH]] ], [ [[TMP3:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[IV:%.*]] = add i64 [[START]], [[INDEX]]
 ; CHECK-NEXT:    [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw float, ptr [[GEP_SRC]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[GEP_SRC]], align 4
 ; CHECK-NEXT:    [[TMP3]] = call <4 x float> @llvm.maxnum.v4f32(<4 x float> [[WIDE_LOAD]], <4 x float> [[VEC_PHI]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/fmin-without-fast-math-flags.ll b/llvm/test/Transforms/LoopVectorize/fmin-without-fast-math-flags.ll
index 68bc8d0..0f688ab 100644
--- a/llvm/test/Transforms/LoopVectorize/fmin-without-fast-math-flags.ll
+++ b/llvm/test/Transforms/LoopVectorize/fmin-without-fast-math-flags.ll
@@ -202,8 +202,7 @@ define float @fminnum_1(ptr %src, i64 %n) {
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ splat (float -1.000000e+07), %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw float, ptr [[GEP_SRC]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[GEP_SRC]], align 4
 ; CHECK-NEXT:    [[TMP4]] = call <4 x float> @llvm.minnum.v4f32(<4 x float> [[WIDE_LOAD]], <4 x float> [[VEC_PHI]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -270,8 +269,7 @@ define float @fminnum_2(ptr %src, i64 %n) {
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ splat (float -1.000000e+07), %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[GEP_SRC:%.*]] = getelementptr inbounds nuw float, ptr [[SRC]], i64 [[IV]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw float, ptr [[GEP_SRC]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[GEP_SRC]], align 4
 ; CHECK-NEXT:    [[TMP4]] = call <4 x float> @llvm.minnum.v4f32(<4 x float> [[VEC_PHI]], <4 x float> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[IV]], 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/forked-pointers.ll b/llvm/test/Transforms/LoopVectorize/forked-pointers.ll
index 6374fba..677163b 100644
--- a/llvm/test/Transforms/LoopVectorize/forked-pointers.ll
+++ b/llvm/test/Transforms/LoopVectorize/forked-pointers.ll
@@ -71,7 +71,7 @@ define dso_local void @forked_ptrs_different_base_same_offset(ptr nocapture read
 ; CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.cond.cleanup:
diff --git a/llvm/test/Transforms/LoopVectorize/fpsat.ll b/llvm/test/Transforms/LoopVectorize/fpsat.ll
index 84753e4..f20e2c8 100644
--- a/llvm/test/Transforms/LoopVectorize/fpsat.ll
+++ b/llvm/test/Transforms/LoopVectorize/fpsat.ll
@@ -23,12 +23,10 @@ define void @signed(ptr %x, ptr %y, i32 %n) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[X]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = call <4 x i32> @llvm.fptosi.sat.v4i32.v4f32(<4 x float> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[Y]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP4]], ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP4]], ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -98,12 +96,10 @@ define void @unsigned(ptr %x, ptr %y, i32 %n) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[X]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = call <4 x i32> @llvm.fptoui.sat.v4i32.v4f32(<4 x float> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[Y]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP4]], ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP4]], ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/if-pred-non-void.ll b/llvm/test/Transforms/LoopVectorize/if-pred-non-void.ll
index d717a3f..4811a77d 100644
--- a/llvm/test/Transforms/LoopVectorize/if-pred-non-void.ll
+++ b/llvm/test/Transforms/LoopVectorize/if-pred-non-void.ll
@@ -44,14 +44,10 @@ define void @test(ptr nocapture %asd, ptr nocapture %aud,
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[AUD]], i64 [[INDEX]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[ASR]], i64 [[INDEX]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[AUR]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP5]], align 4, !alias.scope [[META5:![0-9]+]], !noalias [[META8:![0-9]+]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD23:%.*]] = load <2 x i32>, ptr [[TMP6]], align 4, !alias.scope [[META12:![0-9]+]], !noalias [[META13:![0-9]+]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD24:%.*]] = load <2 x i32>, ptr [[TMP7]], align 4, !alias.scope [[META14:![0-9]+]], !noalias [[META15:![0-9]+]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD25:%.*]] = load <2 x i32>, ptr [[TMP8]], align 4, !alias.scope [[META15]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4, !alias.scope [[META5:![0-9]+]], !noalias [[META8:![0-9]+]]
+; CHECK-NEXT:    [[WIDE_LOAD23:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4, !alias.scope [[META12:![0-9]+]], !noalias [[META13:![0-9]+]]
+; CHECK-NEXT:    [[WIDE_LOAD24:%.*]] = load <2 x i32>, ptr [[TMP3]], align 4, !alias.scope [[META14:![0-9]+]], !noalias [[META15:![0-9]+]]
+; CHECK-NEXT:    [[WIDE_LOAD25:%.*]] = load <2 x i32>, ptr [[TMP4]], align 4, !alias.scope [[META15]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = add nsw <2 x i32> [[WIDE_LOAD]], splat (i32 23)
 ; CHECK-NEXT:    [[TMP10:%.*]] = add nsw <2 x i32> [[WIDE_LOAD23]], splat (i32 24)
 ; CHECK-NEXT:    [[TMP11:%.*]] = add nsw <2 x i32> [[WIDE_LOAD24]], splat (i32 25)
@@ -111,14 +107,10 @@ define void @test(ptr nocapture %asd, ptr nocapture %aud,
 ; CHECK-NEXT:    [[PREDPHI28:%.*]] = select <2 x i1> [[TMP13]], <2 x i32> [[TMP53]], <2 x i32> [[TMP10]]
 ; CHECK-NEXT:    [[PREDPHI29:%.*]] = select <2 x i1> [[TMP13]], <2 x i32> [[TMP54]], <2 x i32> [[TMP11]]
 ; CHECK-NEXT:    [[PREDPHI30:%.*]] = select <2 x i1> [[TMP13]], <2 x i32> [[TMP55]], <2 x i32> [[TMP12]]
-; CHECK-NEXT:    [[TMP57:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP57]], align 4, !alias.scope [[META5]], !noalias [[META8]]
-; CHECK-NEXT:    [[TMP58:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI28]], ptr [[TMP58]], align 4, !alias.scope [[META12]], !noalias [[META13]]
-; CHECK-NEXT:    [[TMP59:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI29]], ptr [[TMP59]], align 4, !alias.scope [[META14]], !noalias [[META15]]
-; CHECK-NEXT:    [[TMP60:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI30]], ptr [[TMP60]], align 4, !alias.scope [[META15]]
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP1]], align 4, !alias.scope [[META5]], !noalias [[META8]]
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI28]], ptr [[TMP2]], align 4, !alias.scope [[META12]], !noalias [[META13]]
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI29]], ptr [[TMP3]], align 4, !alias.scope [[META14]], !noalias [[META15]]
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI30]], ptr [[TMP4]], align 4, !alias.scope [[META15]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP61:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128
 ; CHECK-NEXT:    br i1 [[TMP61]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
@@ -266,9 +258,9 @@ define void @test(ptr nocapture %asd, ptr nocapture %aud,
 ; UNROLL-NO-VF-NEXT:    [[TMP46:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128
 ; UNROLL-NO-VF-NEXT:    br i1 [[TMP46]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; UNROLL-NO-VF:       middle.block:
-; UNROLL-NO-VF-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; UNROLL-NO-VF:       scalar.ph:
-; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 128, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
 ; UNROLL-NO-VF-NEXT:    br label [[FOR_BODY:%.*]]
 ; UNROLL-NO-VF:       for.cond.cleanup:
 ; UNROLL-NO-VF-NEXT:    ret void
@@ -364,11 +356,9 @@ define void @test_scalar2scalar(ptr nocapture %asd, ptr nocapture %bsd) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[PRED_SDIV_CONTINUE4:%.*]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[ASD]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4, !alias.scope [[META20:![0-9]+]], !noalias [[META23:![0-9]+]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4, !alias.scope [[META20:![0-9]+]], !noalias [[META23:![0-9]+]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[BSD]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i32>, ptr [[TMP4]], align 4, !alias.scope [[META23]]
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i32>, ptr [[TMP3]], align 4, !alias.scope [[META23]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = add nsw <2 x i32> [[WIDE_LOAD]], splat (i32 23)
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp slt <2 x i32> [[WIDE_LOAD]], splat (i32 100)
 ; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <2 x i1> [[TMP6]], i32 0
@@ -396,8 +386,7 @@ define void @test_scalar2scalar(ptr nocapture %asd, ptr nocapture %bsd) {
 ; CHECK:       pred.sdiv.continue4:
 ; CHECK-NEXT:    [[TMP24:%.*]] = phi <2 x i32> [ [[TMP15]], [[PRED_SDIV_CONTINUE]] ], [ [[TMP22]], [[PRED_SDIV_IF3]] ]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP6]], <2 x i32> [[TMP24]], <2 x i32> [[TMP5]]
-; CHECK-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP26]], align 4, !alias.scope [[META20]], !noalias [[META23]]
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP1]], align 4, !alias.scope [[META20]], !noalias [[META23]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP27:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128
 ; CHECK-NEXT:    br i1 [[TMP27]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP25:![0-9]+]]
@@ -472,9 +461,9 @@ define void @test_scalar2scalar(ptr nocapture %asd, ptr nocapture %bsd) {
 ; UNROLL-NO-VF-NEXT:    [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128
 ; UNROLL-NO-VF-NEXT:    br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP25:![0-9]+]]
 ; UNROLL-NO-VF:       middle.block:
-; UNROLL-NO-VF-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; UNROLL-NO-VF:       scalar.ph:
-; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 128, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
 ; UNROLL-NO-VF-NEXT:    br label [[FOR_BODY:%.*]]
 ; UNROLL-NO-VF:       for.cond.cleanup:
 ; UNROLL-NO-VF-NEXT:    ret void
@@ -540,16 +529,14 @@ define void @pr30172(ptr nocapture %asd, ptr nocapture %bsd) !dbg !5 {;
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[PRED_SDIV_CONTINUE4:%.*]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[ASD]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4, !alias.scope [[META29:![0-9]+]], !noalias [[META32:![0-9]+]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4, !alias.scope [[META29:![0-9]+]], !noalias [[META32:![0-9]+]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[BSD]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i32>, ptr [[TMP4]], align 4, !alias.scope [[META32]]
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i32>, ptr [[TMP3]], align 4, !alias.scope [[META32]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = add nsw <2 x i32> [[WIDE_LOAD]], splat (i32 23)
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp slt <2 x i32> [[WIDE_LOAD]], splat (i32 100)
-; CHECK-NEXT:    [[TMP7:%.*]] = xor <2 x i1> [[TMP6]], splat (i1 true), !dbg [[DBG34:![0-9]+]]
+; CHECK-NEXT:    [[TMP4:%.*]] = xor <2 x i1> [[TMP6]], splat (i1 true), !dbg [[DBG34:![0-9]+]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp sge <2 x i32> [[WIDE_LOAD]], splat (i32 200)
-; CHECK-NEXT:    [[TMP9:%.*]] = select <2 x i1> [[TMP7]], <2 x i1> [[TMP8]], <2 x i1> zeroinitializer, !dbg [[DBG35:![0-9]+]]
+; CHECK-NEXT:    [[TMP9:%.*]] = select <2 x i1> [[TMP4]], <2 x i1> [[TMP8]], <2 x i1> zeroinitializer, !dbg [[DBG35:![0-9]+]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = or <2 x i1> [[TMP9]], [[TMP6]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = extractelement <2 x i1> [[TMP10]], i32 0
 ; CHECK-NEXT:    br i1 [[TMP11]], label [[PRED_SDIV_IF:%.*]], label [[PRED_SDIV_CONTINUE:%.*]]
@@ -576,10 +563,9 @@ define void @pr30172(ptr nocapture %asd, ptr nocapture %bsd) !dbg !5 {;
 ; CHECK:       pred.sdiv.continue4:
 ; CHECK-NEXT:    [[TMP28:%.*]] = phi <2 x i32> [ [[TMP19]], [[PRED_SDIV_CONTINUE]] ], [ [[TMP26]], [[PRED_SDIV_IF3]] ]
 ; CHECK-NEXT:    [[TMP27:%.*]] = xor <2 x i1> [[TMP8]], splat (i1 true), !dbg [[DBG35]]
-; CHECK-NEXT:    [[TMP30:%.*]] = select <2 x i1> [[TMP7]], <2 x i1> [[TMP27]], <2 x i1> zeroinitializer, !dbg [[DBG35]]
+; CHECK-NEXT:    [[TMP30:%.*]] = select <2 x i1> [[TMP4]], <2 x i1> [[TMP27]], <2 x i1> zeroinitializer, !dbg [[DBG35]]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP30]], <2 x i32> [[TMP5]], <2 x i32> [[TMP28]]
-; CHECK-NEXT:    [[TMP31:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP31]], align 4, !alias.scope [[META29]], !noalias [[META32]]
+; CHECK-NEXT:    store <2 x i32> [[PREDPHI]], ptr [[TMP1]], align 4, !alias.scope [[META29]], !noalias [[META32]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128
 ; CHECK-NEXT:    br i1 [[TMP32]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]
@@ -668,9 +654,9 @@ define void @pr30172(ptr nocapture %asd, ptr nocapture %bsd) !dbg !5 {;
 ; UNROLL-NO-VF-NEXT:    [[TMP34:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128
 ; UNROLL-NO-VF-NEXT:    br i1 [[TMP34]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]
 ; UNROLL-NO-VF:       middle.block:
-; UNROLL-NO-VF-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-VF-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; UNROLL-NO-VF:       scalar.ph:
-; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 128, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; UNROLL-NO-VF-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
 ; UNROLL-NO-VF-NEXT:    br label [[FOR_BODY:%.*]]
 ; UNROLL-NO-VF:       for.cond.cleanup:
 ; UNROLL-NO-VF-NEXT:    ret void
@@ -746,8 +732,7 @@ define i32 @predicated_udiv_scalarized_operand(ptr %a, i1 %c, i32 %x, i64 %n) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE2:%.*]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[PRED_UDIV_CONTINUE2]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    br i1 [[C]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
 ; CHECK:       pred.udiv.if:
 ; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 0
diff --git a/llvm/test/Transforms/LoopVectorize/if-pred-stores.ll b/llvm/test/Transforms/LoopVectorize/if-pred-stores.ll
index 5c1f628b..b971400 100644
--- a/llvm/test/Transforms/LoopVectorize/if-pred-stores.ll
+++ b/llvm/test/Transforms/LoopVectorize/if-pred-stores.ll
@@ -67,9 +67,9 @@ define i32 @test(ptr nocapture %f) #0 {
 ; UNROLL-NOSIMPLIFY-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128
 ; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; UNROLL-NOSIMPLIFY:       middle.block:
-; UNROLL-NOSIMPLIFY-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NOSIMPLIFY-NEXT:    br label [[FOR_END:%.*]]
 ; UNROLL-NOSIMPLIFY:       scalar.ph:
-; UNROLL-NOSIMPLIFY-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 128, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NOSIMPLIFY-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; UNROLL-NOSIMPLIFY-NEXT:    br label [[FOR_BODY:%.*]]
 ; UNROLL-NOSIMPLIFY:       for.body:
 ; UNROLL-NOSIMPLIFY-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_INC:%.*]] ]
@@ -94,8 +94,7 @@ define i32 @test(ptr nocapture %f) #0 {
 ; VEC:       vector.body:
 ; VEC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE2:%.*]] ]
 ; VEC-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[F:%.*]], i64 [[INDEX]]
-; VEC-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; VEC-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; VEC-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; VEC-NEXT:    [[TMP3:%.*]] = icmp sgt <2 x i32> [[WIDE_LOAD]], splat (i32 100)
 ; VEC-NEXT:    [[TMP4:%.*]] = extractelement <2 x i1> [[TMP3]], i32 0
 ; VEC-NEXT:    br i1 [[TMP4]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
@@ -322,8 +321,7 @@ define void @bug18724(i1 %cond, ptr %ptr, i1 %cond.2, i64 %v.1, i32 %v.2) {
 ; VEC-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ [[TMP5]], [[VECTOR_PH]] ], [ [[PREDPHI:%.*]], [[PRED_STORE_CONTINUE2]] ]
 ; VEC-NEXT:    [[OFFSET_IDX:%.*]] = add i64 [[V_1]], [[INDEX]]
 ; VEC-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [768 x i32], ptr [[PTR:%.*]], i64 0, i64 [[OFFSET_IDX]]
-; VEC-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; VEC-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP8]], align 4
+; VEC-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP7]], align 4
 ; VEC-NEXT:    br i1 [[COND_2]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE2]]
 ; VEC:       pred.store.if:
 ; VEC-NEXT:    [[INDVARS_IV3:%.*]] = add i64 [[OFFSET_IDX]], 0
@@ -452,10 +450,10 @@ define void @minimal_bit_widths(i1 %c) {
 ; UNROLL-NOSIMPLIFY-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; UNROLL-NOSIMPLIFY:       middle.block:
-; UNROLL-NOSIMPLIFY-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NOSIMPLIFY-NEXT:    br label [[FOR_END:%.*]]
 ; UNROLL-NOSIMPLIFY:       scalar.ph:
-; UNROLL-NOSIMPLIFY-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; UNROLL-NOSIMPLIFY-NEXT:    [[BC_RESUME_VAL3:%.*]] = phi i64 [ 0, [[MIDDLE_BLOCK]] ], [ 1000, [[ENTRY]] ]
+; UNROLL-NOSIMPLIFY-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NOSIMPLIFY-NEXT:    [[BC_RESUME_VAL3:%.*]] = phi i64 [ 1000, [[ENTRY]] ]
 ; UNROLL-NOSIMPLIFY-NEXT:    br label [[FOR_BODY:%.*]]
 ; UNROLL-NOSIMPLIFY:       for.body:
 ; UNROLL-NOSIMPLIFY-NEXT:    [[TMP1:%.*]] = phi i64 [ [[TMP9:%.*]], [[FOR_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -482,8 +480,7 @@ define void @minimal_bit_widths(i1 %c) {
 ; VEC:       vector.body:
 ; VEC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE2:%.*]] ]
 ; VEC-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr undef, i64 [[INDEX]]
-; VEC-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
-; VEC-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[TMP2]], align 1
+; VEC-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[TMP1]], align 1
 ; VEC-NEXT:    br i1 [[C:%.*]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE2]]
 ; VEC:       pred.store.if:
 ; VEC-NEXT:    [[TMP8:%.*]] = add i64 [[INDEX]], 0
@@ -581,10 +578,10 @@ define void @minimal_bit_widths_with_aliasing_store(i1 %c, ptr %ptr) {
 ; UNROLL-NOSIMPLIFY-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; UNROLL-NOSIMPLIFY:       middle.block:
-; UNROLL-NOSIMPLIFY-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NOSIMPLIFY-NEXT:    br label [[FOR_END:%.*]]
 ; UNROLL-NOSIMPLIFY:       scalar.ph:
-; UNROLL-NOSIMPLIFY-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; UNROLL-NOSIMPLIFY-NEXT:    [[BC_RESUME_VAL3:%.*]] = phi i64 [ 0, [[MIDDLE_BLOCK]] ], [ 1000, [[ENTRY]] ]
+; UNROLL-NOSIMPLIFY-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NOSIMPLIFY-NEXT:    [[BC_RESUME_VAL3:%.*]] = phi i64 [ 1000, [[ENTRY]] ]
 ; UNROLL-NOSIMPLIFY-NEXT:    br label [[FOR_BODY:%.*]]
 ; UNROLL-NOSIMPLIFY:       for.body:
 ; UNROLL-NOSIMPLIFY-NEXT:    [[TMP1:%.*]] = phi i64 [ [[TMP9:%.*]], [[FOR_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -612,9 +609,8 @@ define void @minimal_bit_widths_with_aliasing_store(i1 %c, ptr %ptr) {
 ; VEC:       vector.body:
 ; VEC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE2:%.*]] ]
 ; VEC-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[PTR:%.*]], i64 [[INDEX]]
-; VEC-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[TMP2]], i32 0
-; VEC-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[TMP3]], align 1
-; VEC-NEXT:    store <2 x i8> zeroinitializer, ptr [[TMP3]], align 1
+; VEC-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[TMP2]], align 1
+; VEC-NEXT:    store <2 x i8> zeroinitializer, ptr [[TMP2]], align 1
 ; VEC-NEXT:    br i1 [[C:%.*]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE2]]
 ; VEC:       pred.store.if:
 ; VEC-NEXT:    [[TMP0:%.*]] = add i64 [[INDEX]], 0
@@ -714,7 +710,7 @@ define void @sdiv_with_uniform_ops(i16 %0, i1 %c, ptr %dst) {
 ; UNROLL-NOSIMPLIFY-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 98
 ; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; UNROLL-NOSIMPLIFY:       middle.block:
-; UNROLL-NOSIMPLIFY-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NOSIMPLIFY-NEXT:    br label [[SCALAR_PH]]
 ; UNROLL-NOSIMPLIFY:       scalar.ph:
 ; UNROLL-NOSIMPLIFY-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 99, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY:%.*]] ]
 ; UNROLL-NOSIMPLIFY-NEXT:    br label [[LOOP_HEADER:%.*]]
@@ -728,7 +724,7 @@ define void @sdiv_with_uniform_ops(i16 %0, i1 %c, ptr %dst) {
 ; UNROLL-NOSIMPLIFY:       loop.latch:
 ; UNROLL-NOSIMPLIFY-NEXT:    [[INC]] = add i16 [[IV]], 1
 ; UNROLL-NOSIMPLIFY-NEXT:    [[EC:%.*]] = icmp eq i16 [[INC]], 100
-; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[EC]], label [[EXIT]], label [[LOOP_HEADER]], !llvm.loop [[LOOP11:![0-9]+]]
+; UNROLL-NOSIMPLIFY-NEXT:    br i1 [[EC]], label [[EXIT:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP11:![0-9]+]]
 ; UNROLL-NOSIMPLIFY:       exit:
 ; UNROLL-NOSIMPLIFY-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/if-reduction.ll b/llvm/test/Transforms/LoopVectorize/if-reduction.ll
index b4aba56..65330aa 100644
--- a/llvm/test/Transforms/LoopVectorize/if-reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/if-reduction.ll
@@ -33,8 +33,7 @@ define float @fcmp_0_fadd_select1(ptr noalias %x, i32 %N) nounwind readonly {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP5:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[X]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = fcmp fast ogt <4 x float> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = fadd fast <4 x float> [[WIDE_LOAD]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP5]] = select <4 x i1> [[TMP3]], <4 x float> [[TMP4]], <4 x float> [[VEC_PHI]]
@@ -122,8 +121,7 @@ define double @fcmp_0_fadd_select2(ptr noalias %x, i32 %N) nounwind readonly {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x double> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP5:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds double, ptr [[X]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds double, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = fcmp fast ogt <4 x double> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = fadd fast <4 x double> [[WIDE_LOAD]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP5]] = select <4 x i1> [[TMP3]], <4 x double> [[TMP4]], <4 x double> [[VEC_PHI]]
@@ -214,8 +212,7 @@ define float @fcmp_val_fadd_select1(ptr noalias %x, float %y, i32 %N) nounwind r
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP5:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[X]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = fcmp fast ogt <4 x float> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = fadd fast <4 x float> [[WIDE_LOAD]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP5]] = select <4 x i1> [[TMP3]], <4 x float> [[TMP4]], <4 x float> [[VEC_PHI]]
@@ -306,8 +303,7 @@ define double @fcmp_val_fadd_select2(ptr noalias %x, double %y, i32 %N) nounwind
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x double> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP5:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds double, ptr [[X]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds double, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = fcmp fast ogt <4 x double> [[WIDE_LOAD]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = fadd fast <4 x double> [[WIDE_LOAD]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP5]] = select <4 x i1> [[TMP3]], <4 x double> [[TMP4]], <4 x double> [[VEC_PHI]]
@@ -397,11 +393,9 @@ define float @fcmp_array_elm_fadd_select1(ptr noalias %x, ptr noalias %y, i32 %N
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP7:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[X]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[Y]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = fcmp fast ogt <4 x float> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = fadd fast <4 x float> [[WIDE_LOAD]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP7]] = select <4 x i1> [[TMP5]], <4 x float> [[TMP6]], <4 x float> [[VEC_PHI]]
@@ -495,11 +489,9 @@ define double @fcmp_array_elm_fadd_select2(ptr noalias %x, ptr noalias %y, i32 %
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x double> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP7:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds double, ptr [[X]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds double, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds double, ptr [[Y]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds double, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x double>, ptr [[TMP4]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x double>, ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = fcmp fast ogt <4 x double> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = fadd fast <4 x double> [[WIDE_LOAD]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[TMP7]] = select <4 x i1> [[TMP5]], <4 x double> [[TMP6]], <4 x double> [[VEC_PHI]]
@@ -591,8 +583,7 @@ define float @fcmp_0_fsub_select1(ptr noalias %x, i32 %N) nounwind readonly {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP5:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[X]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = fcmp fast ogt <4 x float> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = fsub fast <4 x float> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP5]] = select <4 x i1> [[TMP3]], <4 x float> [[TMP4]], <4 x float> [[VEC_PHI]]
@@ -742,8 +733,7 @@ define double @fcmp_0_fsub_select2(ptr noalias %x, i32 %N) nounwind readonly {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x double> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP5:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds double, ptr [[X]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds double, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = fcmp fast ogt <4 x double> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = fsub fast <4 x double> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP5]] = select <4 x i1> [[TMP3]], <4 x double> [[TMP4]], <4 x double> [[VEC_PHI]]
@@ -894,8 +884,7 @@ define float @fcmp_0_fmult_select1(ptr noalias %x, i32 %N) nounwind readonly {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ <float 0.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00>, %[[VECTOR_PH]] ], [ [[TMP5:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[X]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = fcmp fast ogt <4 x float> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = fmul fast <4 x float> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP5]] = select <4 x i1> [[TMP3]], <4 x float> [[TMP4]], <4 x float> [[VEC_PHI]]
@@ -1046,8 +1035,7 @@ define double @fcmp_0_fmult_select2(ptr noalias %x, i32 %N) nounwind readonly {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x double> [ <double 0.000000e+00, double 1.000000e+00, double 1.000000e+00, double 1.000000e+00>, %[[VECTOR_PH]] ], [ [[TMP5:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds double, ptr [[X]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds double, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = fcmp fast ogt <4 x double> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = fmul fast <4 x double> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP5]] = select <4 x i1> [[TMP3]], <4 x double> [[TMP4]], <4 x double> [[VEC_PHI]]
@@ -1203,8 +1191,7 @@ define float @fcmp_multi(ptr nocapture readonly %a, i32 %n) nounwind readonly {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP10:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = fcmp ogt <4 x float> [[WIDE_LOAD]], splat (float 1.000000e+00)
 ; CHECK-NEXT:    [[TMP4:%.*]] = xor <4 x i1> [[TMP3]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP5:%.*]] = fcmp olt <4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00)
@@ -1327,8 +1314,7 @@ define float @fcmp_fadd_fsub(ptr nocapture readonly %a, i32 %n) nounwind readonl
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x float> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[PREDPHI1:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = fcmp ogt <4 x float> [[WIDE_LOAD]], splat (float 1.000000e+00)
 ; CHECK-NEXT:    [[TMP4:%.*]] = xor <4 x i1> [[TMP3]], splat (i1 true)
 ; CHECK-NEXT:    [[TMP8:%.*]] = fcmp uge <4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00)
@@ -1586,8 +1572,7 @@ define i64 @fcmp_0_add_select2(ptr noalias %x, i64 %N) nounwind readonly {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP5:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[X]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = fcmp ogt <4 x float> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = add <4 x i64> [[VEC_PHI]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP5]] = select <4 x i1> [[TMP3]], <4 x i64> [[TMP4]], <4 x i64> [[VEC_PHI]]
@@ -1744,8 +1729,7 @@ define i32 @fcmp_0_mult_select1(ptr noalias %x, i32 %N) nounwind readonly {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 1, i32 1>, %[[VECTOR_PH]] ], [ [[TMP5:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[X]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = fcmp ogt <4 x float> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = mul <4 x i32> [[VEC_PHI]], splat (i32 2)
 ; CHECK-NEXT:    [[TMP5]] = select <4 x i1> [[TMP3]], <4 x i32> [[TMP4]], <4 x i32> [[VEC_PHI]]
diff --git a/llvm/test/Transforms/LoopVectorize/induction-multiple-uses-in-same-instruction.ll b/llvm/test/Transforms/LoopVectorize/induction-multiple-uses-in-same-instruction.ll
index 540170a..9cff1cf 100644
--- a/llvm/test/Transforms/LoopVectorize/induction-multiple-uses-in-same-instruction.ll
+++ b/llvm/test/Transforms/LoopVectorize/induction-multiple-uses-in-same-instruction.ll
@@ -24,9 +24,9 @@ define void @multiple_iv_uses_in_same_instruction(ptr %ptr) {
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/induction-step.ll b/llvm/test/Transforms/LoopVectorize/induction-step.ll
index 036d5f5..59f6e8b04f 100644
--- a/llvm/test/Transforms/LoopVectorize/induction-step.ll
+++ b/llvm/test/Transforms/LoopVectorize/induction-step.ll
@@ -47,8 +47,7 @@ define void @induction_with_global(i32 %init, ptr noalias nocapture %A, i32 %N)
 ; CHECK-NEXT:    [[TMP8:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <8 x i32> [ [[INDUCTION4]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP8]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i32 0
-; CHECK-NEXT:    store <8 x i32> [[VEC_IND]], ptr [[TMP10]], align 4
+; CHECK-NEXT:    store <8 x i32> [[VEC_IND]], ptr [[TMP9]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP8]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <8 x i32> [[VEC_IND]], [[DOTSPLAT6]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -137,8 +136,7 @@ define i32 @induction_with_loop_inv(i32 %init, ptr noalias nocapture %A, i32 %N,
 ; CHECK-NEXT:    [[TMP6:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <8 x i32> [ [[INDUCTION4]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP6]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <8 x i32> [[VEC_IND]], ptr [[TMP8]], align 4
+; CHECK-NEXT:    store <8 x i32> [[VEC_IND]], ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP6]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <8 x i32> [[VEC_IND]], [[DOTSPLAT6]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -231,8 +229,7 @@ define void @non_primary_iv_loop_inv_trunc(ptr %a, i64 %n, i64 %step) {
 ; CHECK-NEXT:    [[TMP6:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_IND10:%.*]] = phi <8 x i32> [ [[INDUCTION7]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT11:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP6]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <8 x i32> [[VEC_IND10]], ptr [[TMP8]], align 4
+; CHECK-NEXT:    store <8 x i32> [[VEC_IND10]], ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP6]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT11]] = add <8 x i32> [[VEC_IND10]], [[DOTSPLAT9]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -287,16 +284,15 @@ define void @iv_no_binary_op_in_descriptor(i1 %c, ptr %dst) {
 ; CHECK-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <8 x i64> [ <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[DST]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <8 x i64> [[VEC_IND]], ptr [[TMP2]], align 8
+; CHECK-NEXT:    store <8 x i64> [[VEC_IND]], ptr [[TMP1]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <8 x i64> [[VEC_IND]], splat (i64 8)
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT_P:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -357,8 +353,7 @@ define void @wide_add_induction_step_live_in(ptr %dst, i64 %N, i16 %off) {
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <8 x i16> [ [[INDUCTION]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = add <8 x i16> [[VEC_IND]], [[DOTSPLAT]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[DST]], i64 [[TMP3]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <8 x i16> [[TMP4]], ptr [[TMP6]], align 2
+; CHECK-NEXT:    store <8 x i16> [[TMP4]], ptr [[TMP5]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP3]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <8 x i16> [[VEC_IND]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -368,11 +363,11 @@ define void @wide_add_induction_step_live_in(ptr %dst, i64 %N, i16 %off) {
 ; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL3:%.*]] = phi i16 [ [[TMP0]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL5:%.*]] = phi i16 [ [[TMP0]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
-; CHECK-NEXT:    [[IV_2:%.*]] = phi i16 [ [[BC_RESUME_VAL3]], %[[SCALAR_PH]] ], [ [[ADD:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[IV_2:%.*]] = phi i16 [ [[BC_RESUME_VAL5]], %[[SCALAR_PH]] ], [ [[ADD:%.*]], %[[LOOP]] ]
 ; CHECK-NEXT:    [[ADD]] = add i16 [[IV_2]], [[O_1]]
 ; CHECK-NEXT:    [[GEP_DST:%.*]] = getelementptr inbounds i16, ptr [[DST]], i64 [[IV]]
 ; CHECK-NEXT:    store i16 [[ADD]], ptr [[GEP_DST]], align 2
@@ -428,8 +423,7 @@ define void @wide_sub_induction_step_live_in(ptr %dst, i64 %N, i16 %off) {
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <8 x i16> [ [[INDUCTION]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = sub <8 x i16> [[VEC_IND]], [[DOTSPLAT]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[DST]], i64 [[TMP4]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i16, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <8 x i16> [[TMP5]], ptr [[TMP7]], align 2
+; CHECK-NEXT:    store <8 x i16> [[TMP5]], ptr [[TMP6]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP4]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <8 x i16> [[VEC_IND]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -439,11 +433,11 @@ define void @wide_sub_induction_step_live_in(ptr %dst, i64 %N, i16 %off) {
 ; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL3:%.*]] = phi i16 [ [[TMP1]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL5:%.*]] = phi i16 [ [[TMP1]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
-; CHECK-NEXT:    [[IV_2:%.*]] = phi i16 [ [[BC_RESUME_VAL3]], %[[SCALAR_PH]] ], [ [[SUB:%.*]], %[[LOOP]] ]
+; CHECK-NEXT:    [[IV_2:%.*]] = phi i16 [ [[BC_RESUME_VAL5]], %[[SCALAR_PH]] ], [ [[SUB:%.*]], %[[LOOP]] ]
 ; CHECK-NEXT:    [[SUB]] = sub i16 [[IV_2]], [[O_1]]
 ; CHECK-NEXT:    [[GEP_DST:%.*]] = getelementptr inbounds i16, ptr [[DST]], i64 [[IV]]
 ; CHECK-NEXT:    store i16 [[SUB]], ptr [[GEP_DST]], align 2
diff --git a/llvm/test/Transforms/LoopVectorize/induction.ll b/llvm/test/Transforms/LoopVectorize/induction.ll
index 5a5b06d..77b91cc 100644
--- a/llvm/test/Transforms/LoopVectorize/induction.ll
+++ b/llvm/test/Transforms/LoopVectorize/induction.ll
@@ -27,8 +27,7 @@ define void @multi_int_induction(ptr %A, i32 %N) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 190, i32 191>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP5]], align 4
+; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -156,9 +155,8 @@ define void @multi_int_induction(ptr %A, i32 %N) {
 ; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 190, i32 191>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 2)
 ; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 2
-; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP5]], align 4
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP4]], align 4
 ; UNROLL-NO-IC-NEXT:    store <2 x i32> [[STEP_ADD]], ptr [[TMP6]], align 4
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], splat (i32 2)
@@ -283,15 +281,13 @@ define void @scalar_use(ptr %a, float %b, i64 %offset, i64 %offset2, i64 %n) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], [[OFFSET]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP6]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP8]], align 4, !alias.scope [[META4:![0-9]+]], !noalias [[META7:![0-9]+]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP7]], align 4, !alias.scope [[META4:![0-9]+]], !noalias [[META7:![0-9]+]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX]], [[OFFSET2]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP9]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <2 x float>, ptr [[TMP11]], align 4, !alias.scope [[META7]]
+; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <2 x float>, ptr [[TMP10]], align 4, !alias.scope [[META7]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD4]]
 ; CHECK-NEXT:    [[TMP13:%.*]] = fadd fast <2 x float> [[WIDE_LOAD]], [[TMP12]]
-; CHECK-NEXT:    store <2 x float> [[TMP13]], ptr [[TMP8]], align 4, !alias.scope [[META4]], !noalias [[META7]]
+; CHECK-NEXT:    store <2 x float> [[TMP13]], ptr [[TMP7]], align 4, !alias.scope [[META4]], !noalias [[META7]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
@@ -473,21 +469,19 @@ define void @scalar_use(ptr %a, float %b, i64 %offset, i64 %offset2, i64 %n) {
 ; UNROLL-NO-IC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; UNROLL-NO-IC-NEXT:    [[TMP6:%.*]] = add i64 [[INDEX]], [[OFFSET]]
 ; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP6]]
-; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP7]], i32 2
-; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP8]], align 4, !alias.scope [[META4:![0-9]+]], !noalias [[META7:![0-9]+]]
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP7]], align 4, !alias.scope [[META4:![0-9]+]], !noalias [[META7:![0-9]+]]
 ; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD4:%.*]] = load <2 x float>, ptr [[TMP9]], align 4, !alias.scope [[META4]], !noalias [[META7]]
 ; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX]], [[OFFSET2]]
 ; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP10]]
-; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP11]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP11]], i32 2
-; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD5:%.*]] = load <2 x float>, ptr [[TMP12]], align 4, !alias.scope [[META7]]
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD5:%.*]] = load <2 x float>, ptr [[TMP11]], align 4, !alias.scope [[META7]]
 ; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD6:%.*]] = load <2 x float>, ptr [[TMP13]], align 4, !alias.scope [[META7]]
 ; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD5]]
 ; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = fmul fast <2 x float> [[BROADCAST_SPLAT]], [[WIDE_LOAD6]]
 ; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = fadd fast <2 x float> [[WIDE_LOAD]], [[TMP14]]
 ; UNROLL-NO-IC-NEXT:    [[TMP17:%.*]] = fadd fast <2 x float> [[WIDE_LOAD4]], [[TMP15]]
-; UNROLL-NO-IC-NEXT:    store <2 x float> [[TMP16]], ptr [[TMP8]], align 4, !alias.scope [[META4]], !noalias [[META7]]
+; UNROLL-NO-IC-NEXT:    store <2 x float> [[TMP16]], ptr [[TMP7]], align 4, !alias.scope [[META4]], !noalias [[META7]]
 ; UNROLL-NO-IC-NEXT:    store <2 x float> [[TMP17]], ptr [[TMP9]], align 4, !alias.scope [[META4]], !noalias [[META7]]
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; UNROLL-NO-IC-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -628,8 +622,7 @@ define i64 @scalarize_induction_variable_01(ptr %a, i64 %n) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 8
 ; CHECK-NEXT:    [[TMP3]] = add <2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -750,9 +743,8 @@ define i64 @scalarize_induction_variable_01(ptr %a, i64 %n) {
 ; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; UNROLL-NO-IC-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[VECTOR_BODY]] ]
 ; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
-; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 2
-; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 8
 ; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i64>, ptr [[TMP3]], align 8
 ; UNROLL-NO-IC-NEXT:    [[TMP4]] = add <2 x i64> [[WIDE_LOAD]], [[VEC_PHI]]
 ; UNROLL-NO-IC-NEXT:    [[TMP5]] = add <2 x i64> [[WIDE_LOAD2]], [[VEC_PHI1]]
@@ -1961,8 +1953,7 @@ define i32 @scalarize_induction_variable_05(ptr %a, i32 %x, i1 %c, i32 %n) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_UDIV_CONTINUE2:%.*]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP16:%.*]], [[PRED_UDIV_CONTINUE2]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    br i1 [[C]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
 ; CHECK:       pred.udiv.if:
 ; CHECK-NEXT:    [[TMP4:%.*]] = add i32 [[INDEX]], 0
@@ -2185,9 +2176,8 @@ define i32 @scalarize_induction_variable_05(ptr %a, i32 %x, i1 %c, i32 %n) {
 ; UNROLL-NO-IC-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP32:%.*]], [[PRED_UDIV_CONTINUE8]] ]
 ; UNROLL-NO-IC-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP33:%.*]], [[PRED_UDIV_CONTINUE8]] ]
 ; UNROLL-NO-IC-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]
-; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 2
-; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
+; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP0]], align 4
 ; UNROLL-NO-IC-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
 ; UNROLL-NO-IC-NEXT:    br i1 [[C]], label [[PRED_UDIV_IF:%.*]], label [[PRED_UDIV_CONTINUE:%.*]]
 ; UNROLL-NO-IC:       pred.udiv.if:
@@ -2755,10 +2745,10 @@ define i32 @i8_loop() nounwind readnone ssp uwtable {
 ; CHECK-NEXT:    br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[TMP0]])
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP2]], [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 0, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 1, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[A_0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[A_0_AND:%.*]], [[LOOP]] ]
@@ -2782,7 +2772,7 @@ define i32 @i8_loop() nounwind readnone ssp uwtable {
 ; IND-NEXT:    [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256
 ; IND-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
 ; IND:       middle.block:
-; IND-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; IND-NEXT:    br label [[EXIT:%.*]]
 ; IND:       scalar.ph:
 ; IND-NEXT:    br label [[LOOP:%.*]]
 ; IND:       loop:
@@ -2801,7 +2791,7 @@ define i32 @i8_loop() nounwind readnone ssp uwtable {
 ; UNROLL-NEXT:    [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256
 ; UNROLL-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
 ; UNROLL:       middle.block:
-; UNROLL-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NEXT:    br label [[EXIT:%.*]]
 ; UNROLL:       scalar.ph:
 ; UNROLL-NEXT:    br label [[LOOP:%.*]]
 ; UNROLL:       loop:
@@ -2826,10 +2816,10 @@ define i32 @i8_loop() nounwind readnone ssp uwtable {
 ; UNROLL-NO-IC:       middle.block:
 ; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = and <2 x i32> [[TMP1]], [[TMP0]]
 ; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[BIN_RDX]])
-; UNROLL-NO-IC-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC-NEXT:    br label [[EXIT:%.*]]
 ; UNROLL-NO-IC:       scalar.ph:
-; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP3]], [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY:%.*]] ]
-; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 0, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 1, [[ENTRY:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 0, [[ENTRY]] ]
 ; UNROLL-NO-IC-NEXT:    br label [[LOOP:%.*]]
 ; UNROLL-NO-IC:       loop:
 ; UNROLL-NO-IC-NEXT:    [[A_0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[A_0_AND:%.*]], [[LOOP]] ]
@@ -2853,7 +2843,7 @@ define i32 @i8_loop() nounwind readnone ssp uwtable {
 ; INTERLEAVE-NEXT:    [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256
 ; INTERLEAVE-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
 ; INTERLEAVE:       middle.block:
-; INTERLEAVE-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE-NEXT:    br label [[EXIT:%.*]]
 ; INTERLEAVE:       scalar.ph:
 ; INTERLEAVE-NEXT:    br label [[LOOP:%.*]]
 ; INTERLEAVE:       loop:
@@ -2892,10 +2882,10 @@ define i32 @i16_loop() nounwind readnone ssp uwtable {
 ; CHECK-NEXT:    br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[TMP0]])
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP2]], [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 1, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[A_0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[A_0_AND:%.*]], [[LOOP]] ]
@@ -2919,7 +2909,7 @@ define i32 @i16_loop() nounwind readnone ssp uwtable {
 ; IND-NEXT:    [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 65536
 ; IND-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
 ; IND:       middle.block:
-; IND-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; IND-NEXT:    br label [[EXIT:%.*]]
 ; IND:       scalar.ph:
 ; IND-NEXT:    br label [[LOOP:%.*]]
 ; IND:       loop:
@@ -2938,7 +2928,7 @@ define i32 @i16_loop() nounwind readnone ssp uwtable {
 ; UNROLL-NEXT:    [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 65536
 ; UNROLL-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
 ; UNROLL:       middle.block:
-; UNROLL-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NEXT:    br label [[EXIT:%.*]]
 ; UNROLL:       scalar.ph:
 ; UNROLL-NEXT:    br label [[LOOP:%.*]]
 ; UNROLL:       loop:
@@ -2963,10 +2953,10 @@ define i32 @i16_loop() nounwind readnone ssp uwtable {
 ; UNROLL-NO-IC:       middle.block:
 ; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = and <2 x i32> [[TMP1]], [[TMP0]]
 ; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.and.v2i32(<2 x i32> [[BIN_RDX]])
-; UNROLL-NO-IC-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC-NEXT:    br label [[EXIT:%.*]]
 ; UNROLL-NO-IC:       scalar.ph:
-; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP3]], [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY:%.*]] ]
-; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 1, [[ENTRY:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, [[ENTRY]] ]
 ; UNROLL-NO-IC-NEXT:    br label [[LOOP:%.*]]
 ; UNROLL-NO-IC:       loop:
 ; UNROLL-NO-IC-NEXT:    [[A_0:%.*]] = phi i32 [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ], [ [[A_0_AND:%.*]], [[LOOP]] ]
@@ -2990,7 +2980,7 @@ define i32 @i16_loop() nounwind readnone ssp uwtable {
 ; INTERLEAVE-NEXT:    [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 65536
 ; INTERLEAVE-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
 ; INTERLEAVE:       middle.block:
-; INTERLEAVE-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE-NEXT:    br label [[EXIT:%.*]]
 ; INTERLEAVE:       scalar.ph:
 ; INTERLEAVE-NEXT:    br label [[LOOP:%.*]]
 ; INTERLEAVE:       loop:
@@ -3384,8 +3374,7 @@ define void @wrappingindvars1(i8 %t, i32 %len, ptr %A) {
 ; CHECK-NEXT:    [[DOTCAST4:%.*]] = trunc i32 [[INDEX]] to i8
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i8 [[T]], [[DOTCAST4]]
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i8 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP14]], align 4
+; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP13]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)
 ; CHECK-NEXT:    [[TMP15:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
@@ -3589,9 +3578,8 @@ define void @wrappingindvars1(i8 %t, i32 %len, ptr %A) {
 ; UNROLL-NO-IC-NEXT:    [[DOTCAST4:%.*]] = trunc i32 [[INDEX]] to i8
 ; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = add i8 [[T]], [[DOTCAST4]]
 ; UNROLL-NO-IC-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i8 [[OFFSET_IDX]]
-; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 2
-; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP14]], align 4
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP13]], align 4
 ; UNROLL-NO-IC-NEXT:    store <2 x i32> [[STEP_ADD]], ptr [[TMP15]], align 4
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], splat (i32 2)
@@ -3762,8 +3750,7 @@ define void @wrappingindvars2(i8 %t, i32 %len, ptr %A) {
 ; CHECK-NEXT:    [[DOTCAST4:%.*]] = trunc i32 [[INDEX]] to i8
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i8 [[T]], [[DOTCAST4]]
 ; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i8 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP15]], align 4
+; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP14]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 8)
 ; CHECK-NEXT:    [[TMP16:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
@@ -3976,9 +3963,8 @@ define void @wrappingindvars2(i8 %t, i32 %len, ptr %A) {
 ; UNROLL-NO-IC-NEXT:    [[DOTCAST4:%.*]] = trunc i32 [[INDEX]] to i8
 ; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = add i8 [[T]], [[DOTCAST4]]
 ; UNROLL-NO-IC-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i8 [[OFFSET_IDX]]
-; UNROLL-NO-IC-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i32 2
-; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP15]], align 4
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP14]], align 4
 ; UNROLL-NO-IC-NEXT:    store <2 x i32> [[STEP_ADD]], ptr [[TMP16]], align 4
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], splat (i32 8)
@@ -4126,8 +4112,7 @@ define void @veciv(ptr nocapture %a, i32 %start, i32 %k) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP2]], align 4
+; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
@@ -4232,9 +4217,8 @@ define void @veciv(ptr nocapture %a, i32 %start, i32 %k) {
 ; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 2)
 ; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]
-; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 2
-; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP2]], align 4
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP1]], align 4
 ; UNROLL-NO-IC-NEXT:    store <2 x i32> [[STEP_ADD]], ptr [[TMP3]], align 4
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], splat (i32 2)
@@ -4329,8 +4313,7 @@ define void @trunciv(ptr nocapture %a, i32 %start, i64 %k) {
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = trunc i64 [[INDEX]] to i32
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[TMP5]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP8]], align 4
+; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -4456,9 +4439,8 @@ define void @trunciv(ptr nocapture %a, i32 %start, i64 %k) {
 ; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 2)
 ; UNROLL-NO-IC-NEXT:    [[TMP5:%.*]] = trunc i64 [[INDEX]] to i32
 ; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[TMP5]]
-; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 2
-; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP8]], align 4
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP7]], align 4
 ; UNROLL-NO-IC-NEXT:    store <2 x i32> [[STEP_ADD]], ptr [[TMP9]], align 4
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], splat (i32 2)
@@ -4561,8 +4543,7 @@ define void @nonprimary(ptr nocapture %a, i32 %start, i32 %i, i32 %k) {
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i32 [[I]], [[INDEX]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP3]], align 4
+; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
@@ -4685,9 +4666,8 @@ define void @nonprimary(ptr nocapture %a, i32 %start, i32 %i, i32 %k) {
 ; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 2)
 ; UNROLL-NO-IC-NEXT:    [[OFFSET_IDX:%.*]] = add i32 [[I]], [[INDEX]]
 ; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[OFFSET_IDX]]
-; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 2
-; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP3]], align 4
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP2]], align 4
 ; UNROLL-NO-IC-NEXT:    store <2 x i32> [[STEP_ADD]], ptr [[TMP4]], align 4
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], splat (i32 2)
@@ -4782,8 +4762,7 @@ define void @non_primary_iv_trunc(ptr %a, i64 %n) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP2]], align 4
+; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 4)
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -4902,9 +4881,8 @@ define void @non_primary_iv_trunc(ptr %a, i64 %n) {
 ; UNROLL-NO-IC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 2>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; UNROLL-NO-IC-NEXT:    [[STEP_ADD:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 4)
 ; UNROLL-NO-IC-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; UNROLL-NO-IC-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 2
-; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP2]], align 4
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP1]], align 4
 ; UNROLL-NO-IC-NEXT:    store <2 x i32> [[STEP_ADD]], ptr [[TMP3]], align 4
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], splat (i32 4)
@@ -5037,10 +5015,10 @@ define i32 @PR32419(i32 %a, i16 %b) {
 ; CHECK-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP46:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP17:%.*]] = call i32 @llvm.vector.reduce.or.v2i32(<2 x i32> [[TMP15]])
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[MIDDLE_BLOCK]] ], [ -20, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP17]], [[MIDDLE_BLOCK]] ], [ [[A]], [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ -20, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[A]], [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_INC:%.*]] ]
@@ -5101,7 +5079,7 @@ define i32 @PR32419(i32 %a, i16 %b) {
 ; IND-NEXT:    br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP46:![0-9]+]]
 ; IND:       middle.block:
 ; IND-NEXT:    [[TMP16:%.*]] = call i32 @llvm.vector.reduce.or.v2i32(<2 x i32> [[TMP14]])
-; IND-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; IND-NEXT:    br label [[FOR_END:%.*]]
 ; IND:       scalar.ph:
 ; IND-NEXT:    br label [[FOR_BODY:%.*]]
 ; IND:       for.body:
@@ -5177,7 +5155,7 @@ define i32 @PR32419(i32 %a, i16 %b) {
 ; UNROLL:       middle.block:
 ; UNROLL-NEXT:    [[BIN_RDX:%.*]] = or <2 x i32> [[TMP27]], [[TMP26]]
 ; UNROLL-NEXT:    [[TMP29:%.*]] = call i32 @llvm.vector.reduce.or.v2i32(<2 x i32> [[BIN_RDX]])
-; UNROLL-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NEXT:    br label [[FOR_END:%.*]]
 ; UNROLL:       scalar.ph:
 ; UNROLL-NEXT:    br label [[FOR_BODY:%.*]]
 ; UNROLL:       for.body:
@@ -5257,10 +5235,10 @@ define i32 @PR32419(i32 %a, i16 %b) {
 ; UNROLL-NO-IC:       middle.block:
 ; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = or <2 x i32> [[TMP29]], [[TMP28]]
 ; UNROLL-NO-IC-NEXT:    [[TMP31:%.*]] = call i32 @llvm.vector.reduce.or.v2i32(<2 x i32> [[BIN_RDX]])
-; UNROLL-NO-IC-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC-NEXT:    br label [[FOR_END:%.*]]
 ; UNROLL-NO-IC:       scalar.ph:
-; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[MIDDLE_BLOCK]] ], [ -20, [[ENTRY:%.*]] ]
-; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP31]], [[MIDDLE_BLOCK]] ], [ [[A]], [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ -20, [[ENTRY:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[A]], [[ENTRY]] ]
 ; UNROLL-NO-IC-NEXT:    br label [[FOR_BODY:%.*]]
 ; UNROLL-NO-IC:       for.body:
 ; UNROLL-NO-IC-NEXT:    [[I:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I_NEXT:%.*]], [[FOR_INC:%.*]] ]
@@ -5381,7 +5359,7 @@ define i32 @PR32419(i32 %a, i16 %b) {
 ; INTERLEAVE:       middle.block:
 ; INTERLEAVE-NEXT:    [[BIN_RDX:%.*]] = or <4 x i32> [[TMP47]], [[TMP46]]
 ; INTERLEAVE-NEXT:    [[TMP49:%.*]] = call i32 @llvm.vector.reduce.or.v4i32(<4 x i32> [[BIN_RDX]])
-; INTERLEAVE-NEXT:    br i1 false, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE-NEXT:    br label [[SCALAR_PH]]
 ; INTERLEAVE:       scalar.ph:
 ; INTERLEAVE-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP49]], [[MIDDLE_BLOCK]] ], [ poison, [[ENTRY:%.*]] ]
 ; INTERLEAVE-NEXT:    br label [[FOR_BODY:%.*]]
@@ -5400,10 +5378,9 @@ define i32 @PR32419(i32 %a, i16 %b) {
 ; INTERLEAVE-NEXT:    [[VAR6]] = or i32 [[VAR0]], [[VAR4]]
 ; INTERLEAVE-NEXT:    [[I_NEXT]] = add nsw i32 [[I]], 1
 ; INTERLEAVE-NEXT:    [[COND:%.*]] = icmp eq i32 [[I_NEXT]], 0
-; INTERLEAVE-NEXT:    br i1 [[COND]], label [[FOR_END]], label [[FOR_BODY]], !llvm.loop [[LOOP47:![0-9]+]]
+; INTERLEAVE-NEXT:    br i1 [[COND]], label [[FOR_END:%.*]], label [[FOR_BODY]], !llvm.loop [[LOOP47:![0-9]+]]
 ; INTERLEAVE:       for.end:
-; INTERLEAVE-NEXT:    [[VAR7:%.*]] = phi i32 [ [[VAR6]], [[FOR_INC]] ], [ poison, [[MIDDLE_BLOCK]] ]
-; INTERLEAVE-NEXT:    ret i32 [[VAR7]]
+; INTERLEAVE-NEXT:    ret i32 [[VAR6]]
 ;
 entry:
   br label %for.body
@@ -5470,7 +5447,7 @@ define i64 @trunc_with_first_order_recurrence() {
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[TMP10]])
 ; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[VEC_IND2]], i32 1
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP12]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 113, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY]] ]
@@ -5478,7 +5455,7 @@ define i64 @trunc_with_first_order_recurrence() {
 ; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 42, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       exit:
-; CHECK-NEXT:    [[DOTLCSSA:%.*]] = phi i64 [ [[C23:%.*]], [[LOOP]] ], [ [[TMP12]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[DOTLCSSA:%.*]] = phi i64 [ [[C23:%.*]], [[LOOP]] ]
 ; CHECK-NEXT:    ret i64 [[DOTLCSSA]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[C5:%.*]] = phi i64 [ [[C23]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
@@ -5500,7 +5477,7 @@ define i64 @trunc_with_first_order_recurrence() {
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    [[C24]] = add nuw nsw i32 [[X]], 1
 ; CHECK-NEXT:    [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 114
-; CHECK-NEXT:    br i1 [[EXITCOND_I]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP49:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND_I]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP49:![0-9]+]]
 ;
 ; IND-LABEL: @trunc_with_first_order_recurrence(
 ; IND-NEXT:  entry:
@@ -5534,16 +5511,15 @@ define i64 @trunc_with_first_order_recurrence() {
 ; IND:       middle.block:
 ; IND-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[TMP10]])
 ; IND-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[VEC_IND2]], i64 1
-; IND-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; IND-NEXT:    br label [[SCALAR_PH]]
 ; IND:       scalar.ph:
 ; IND-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP12]], [[MIDDLE_BLOCK]] ], [ poison, [[ENTRY:%.*]] ]
 ; IND-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ poison, [[ENTRY]] ]
 ; IND-NEXT:    br label [[LOOP:%.*]]
 ; IND:       exit:
-; IND-NEXT:    [[DOTLCSSA:%.*]] = phi i64 [ [[C23:%.*]], [[LOOP]] ], [ poison, [[MIDDLE_BLOCK]] ]
-; IND-NEXT:    ret i64 [[DOTLCSSA]]
+; IND-NEXT:    ret i64 [[DOTLCSSA:%.*]]
 ; IND:       loop:
-; IND-NEXT:    [[C5:%.*]] = phi i64 [ [[C23]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; IND-NEXT:    [[C5:%.*]] = phi i64 [ [[DOTLCSSA]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
 ; IND-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ 113, [[SCALAR_PH]] ]
 ; IND-NEXT:    [[X:%.*]] = phi i32 [ [[C24:%.*]], [[LOOP]] ], [ 113, [[SCALAR_PH]] ]
 ; IND-NEXT:    [[Y:%.*]] = phi i32 [ [[C6:%.*]], [[LOOP]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
@@ -5558,11 +5534,11 @@ define i64 @trunc_with_first_order_recurrence() {
 ; IND-NEXT:    [[C14:%.*]] = shl i32 [[INDVARS_IV_TR]], 1
 ; IND-NEXT:    [[C15:%.*]] = add i32 [[C9]], [[C14]]
 ; IND-NEXT:    [[C16:%.*]] = sext i32 [[C15]] to i64
-; IND-NEXT:    [[C23]] = add i64 [[C13]], [[C16]]
+; IND-NEXT:    [[DOTLCSSA]] = add i64 [[C13]], [[C16]]
 ; IND-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; IND-NEXT:    [[C24]] = add nuw nsw i32 [[X]], 1
 ; IND-NEXT:    [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 114
-; IND-NEXT:    br i1 [[EXITCOND_I]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP49:![0-9]+]]
+; IND-NEXT:    br i1 [[EXITCOND_I]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP49:![0-9]+]]
 ;
 ; UNROLL-LABEL: @trunc_with_first_order_recurrence(
 ; UNROLL-NEXT:  entry:
@@ -5612,16 +5588,15 @@ define i64 @trunc_with_first_order_recurrence() {
 ; UNROLL-NEXT:    [[BIN_RDX:%.*]] = add <2 x i64> [[TMP21]], [[TMP20]]
 ; UNROLL-NEXT:    [[TMP23:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[BIN_RDX]])
 ; UNROLL-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[STEP_ADD7]], i64 1
-; UNROLL-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NEXT:    br label [[SCALAR_PH]]
 ; UNROLL:       scalar.ph:
 ; UNROLL-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP23]], [[MIDDLE_BLOCK]] ], [ poison, [[ENTRY:%.*]] ]
 ; UNROLL-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ poison, [[ENTRY]] ]
 ; UNROLL-NEXT:    br label [[LOOP:%.*]]
 ; UNROLL:       exit:
-; UNROLL-NEXT:    [[DOTLCSSA:%.*]] = phi i64 [ [[C23:%.*]], [[LOOP]] ], [ poison, [[MIDDLE_BLOCK]] ]
-; UNROLL-NEXT:    ret i64 [[DOTLCSSA]]
+; UNROLL-NEXT:    ret i64 [[DOTLCSSA:%.*]]
 ; UNROLL:       loop:
-; UNROLL-NEXT:    [[C5:%.*]] = phi i64 [ [[C23]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; UNROLL-NEXT:    [[C5:%.*]] = phi i64 [ [[DOTLCSSA]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
 ; UNROLL-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ 113, [[SCALAR_PH]] ]
 ; UNROLL-NEXT:    [[X:%.*]] = phi i32 [ [[C24:%.*]], [[LOOP]] ], [ 113, [[SCALAR_PH]] ]
 ; UNROLL-NEXT:    [[Y:%.*]] = phi i32 [ [[C6:%.*]], [[LOOP]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
@@ -5636,11 +5611,11 @@ define i64 @trunc_with_first_order_recurrence() {
 ; UNROLL-NEXT:    [[C14:%.*]] = shl i32 [[INDVARS_IV_TR]], 1
 ; UNROLL-NEXT:    [[C15:%.*]] = add i32 [[C9]], [[C14]]
 ; UNROLL-NEXT:    [[C16:%.*]] = sext i32 [[C15]] to i64
-; UNROLL-NEXT:    [[C23]] = add i64 [[C13]], [[C16]]
+; UNROLL-NEXT:    [[DOTLCSSA]] = add i64 [[C13]], [[C16]]
 ; UNROLL-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; UNROLL-NEXT:    [[C24]] = add nuw nsw i32 [[X]], 1
 ; UNROLL-NEXT:    [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 114
-; UNROLL-NEXT:    br i1 [[EXITCOND_I]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP49:![0-9]+]]
+; UNROLL-NEXT:    br i1 [[EXITCOND_I]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP49:![0-9]+]]
 ;
 ; UNROLL-NO-IC-LABEL: @trunc_with_first_order_recurrence(
 ; UNROLL-NO-IC-NEXT:  entry:
@@ -5690,7 +5665,7 @@ define i64 @trunc_with_first_order_recurrence() {
 ; UNROLL-NO-IC-NEXT:    [[BIN_RDX:%.*]] = add <2 x i64> [[TMP21]], [[TMP20]]
 ; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = call i64 @llvm.vector.reduce.add.v2i64(<2 x i64> [[BIN_RDX]])
 ; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[STEP_ADD7]], i32 1
-; UNROLL-NO-IC-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC-NEXT:    br label [[SCALAR_PH]]
 ; UNROLL-NO-IC:       scalar.ph:
 ; UNROLL-NO-IC-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP23]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 113, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY]] ]
@@ -5698,7 +5673,7 @@ define i64 @trunc_with_first_order_recurrence() {
 ; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 42, [[ENTRY]] ]
 ; UNROLL-NO-IC-NEXT:    br label [[LOOP:%.*]]
 ; UNROLL-NO-IC:       exit:
-; UNROLL-NO-IC-NEXT:    [[DOTLCSSA:%.*]] = phi i64 [ [[C23:%.*]], [[LOOP]] ], [ [[TMP23]], [[MIDDLE_BLOCK]] ]
+; UNROLL-NO-IC-NEXT:    [[DOTLCSSA:%.*]] = phi i64 [ [[C23:%.*]], [[LOOP]] ]
 ; UNROLL-NO-IC-NEXT:    ret i64 [[DOTLCSSA]]
 ; UNROLL-NO-IC:       loop:
 ; UNROLL-NO-IC-NEXT:    [[C5:%.*]] = phi i64 [ [[C23]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
@@ -5720,7 +5695,7 @@ define i64 @trunc_with_first_order_recurrence() {
 ; UNROLL-NO-IC-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; UNROLL-NO-IC-NEXT:    [[C24]] = add nuw nsw i32 [[X]], 1
 ; UNROLL-NO-IC-NEXT:    [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 114
-; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND_I]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP49:![0-9]+]]
+; UNROLL-NO-IC-NEXT:    br i1 [[EXITCOND_I]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP49:![0-9]+]]
 ;
 ; INTERLEAVE-LABEL: @trunc_with_first_order_recurrence(
 ; INTERLEAVE-NEXT:  entry:
@@ -5770,16 +5745,15 @@ define i64 @trunc_with_first_order_recurrence() {
 ; INTERLEAVE-NEXT:    [[BIN_RDX:%.*]] = add <4 x i64> [[TMP21]], [[TMP20]]
 ; INTERLEAVE-NEXT:    [[TMP23:%.*]] = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> [[BIN_RDX]])
 ; INTERLEAVE-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[STEP_ADD7]], i64 3
-; INTERLEAVE-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE-NEXT:    br label [[SCALAR_PH]]
 ; INTERLEAVE:       scalar.ph:
 ; INTERLEAVE-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[TMP23]], [[MIDDLE_BLOCK]] ], [ poison, [[ENTRY:%.*]] ]
 ; INTERLEAVE-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ poison, [[ENTRY]] ]
 ; INTERLEAVE-NEXT:    br label [[LOOP:%.*]]
 ; INTERLEAVE:       exit:
-; INTERLEAVE-NEXT:    [[DOTLCSSA:%.*]] = phi i64 [ [[C23:%.*]], [[LOOP]] ], [ poison, [[MIDDLE_BLOCK]] ]
-; INTERLEAVE-NEXT:    ret i64 [[DOTLCSSA]]
+; INTERLEAVE-NEXT:    ret i64 [[DOTLCSSA:%.*]]
 ; INTERLEAVE:       loop:
-; INTERLEAVE-NEXT:    [[C5:%.*]] = phi i64 [ [[C23]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
+; INTERLEAVE-NEXT:    [[C5:%.*]] = phi i64 [ [[DOTLCSSA]], [[LOOP]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
 ; INTERLEAVE-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], [[LOOP]] ], [ 113, [[SCALAR_PH]] ]
 ; INTERLEAVE-NEXT:    [[X:%.*]] = phi i32 [ [[C24:%.*]], [[LOOP]] ], [ 113, [[SCALAR_PH]] ]
 ; INTERLEAVE-NEXT:    [[Y:%.*]] = phi i32 [ [[C6:%.*]], [[LOOP]] ], [ [[SCALAR_RECUR_INIT]], [[SCALAR_PH]] ]
@@ -5794,11 +5768,11 @@ define i64 @trunc_with_first_order_recurrence() {
 ; INTERLEAVE-NEXT:    [[C14:%.*]] = shl i32 [[INDVARS_IV_TR]], 1
 ; INTERLEAVE-NEXT:    [[C15:%.*]] = add i32 [[C9]], [[C14]]
 ; INTERLEAVE-NEXT:    [[C16:%.*]] = sext i32 [[C15]] to i64
-; INTERLEAVE-NEXT:    [[C23]] = add i64 [[C13]], [[C16]]
+; INTERLEAVE-NEXT:    [[DOTLCSSA]] = add i64 [[C13]], [[C16]]
 ; INTERLEAVE-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; INTERLEAVE-NEXT:    [[C24]] = add nuw nsw i32 [[X]], 1
 ; INTERLEAVE-NEXT:    [[EXITCOND_I:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 114
-; INTERLEAVE-NEXT:    br i1 [[EXITCOND_I]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP49:![0-9]+]]
+; INTERLEAVE-NEXT:    br i1 [[EXITCOND_I]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP49:![0-9]+]]
 ;
 entry:
   br label %loop
@@ -5851,19 +5825,17 @@ define void @pr52460_first_order_recurrence_truncated_iv(ptr noalias %src, ptr %
 ; CHECK-NEXT:    [[TMP4:%.*]] = mul nsw <2 x i32> [[BROADCAST_SPLAT]], [[TMP2]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[DST:%.*]], i32 [[TMP0]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = add <2 x i32> [[VEC_IND]], [[TMP4]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[TMP6]], ptr [[TMP7]], align 4
+; CHECK-NEXT:    store <2 x i32> [[TMP6]], ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[VEC_IND]], i32 1
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -5906,7 +5878,7 @@ define void @pr52460_first_order_recurrence_truncated_iv(ptr noalias %src, ptr %
 ; IND-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; IND-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
 ; IND:       middle.block:
-; IND-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; IND-NEXT:    br label [[EXIT:%.*]]
 ; IND:       scalar.ph:
 ; IND-NEXT:    br label [[LOOP:%.*]]
 ; IND:       loop:
@@ -5944,7 +5916,7 @@ define void @pr52460_first_order_recurrence_truncated_iv(ptr noalias %src, ptr %
 ; UNROLL-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; UNROLL-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
 ; UNROLL:       middle.block:
-; UNROLL-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NEXT:    br label [[EXIT:%.*]]
 ; UNROLL:       scalar.ph:
 ; UNROLL-NEXT:    br label [[LOOP:%.*]]
 ; UNROLL:       loop:
@@ -5973,21 +5945,19 @@ define void @pr52460_first_order_recurrence_truncated_iv(ptr noalias %src, ptr %
 ; UNROLL-NO-IC-NEXT:    [[TMP7:%.*]] = getelementptr i32, ptr [[DST:%.*]], i32 [[TMP0]]
 ; UNROLL-NO-IC-NEXT:    [[TMP8:%.*]] = add <2 x i32> [[VEC_IND]], [[TMP5]]
 ; UNROLL-NO-IC-NEXT:    [[TMP9:%.*]] = add <2 x i32> [[STEP_ADD]], [[TMP6]]
-; UNROLL-NO-IC-NEXT:    [[TMP10:%.*]] = getelementptr i32, ptr [[TMP7]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP11:%.*]] = getelementptr i32, ptr [[TMP7]], i32 2
-; UNROLL-NO-IC-NEXT:    store <2 x i32> [[TMP8]], ptr [[TMP10]], align 4
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[TMP8]], ptr [[TMP7]], align 4
 ; UNROLL-NO-IC-NEXT:    store <2 x i32> [[TMP9]], ptr [[TMP11]], align 4
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], splat (i32 2)
 ; UNROLL-NO-IC-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; UNROLL-NO-IC-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
 ; UNROLL-NO-IC:       middle.block:
-; UNROLL-NO-IC-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <2 x i32> [[STEP_ADD]], i32 1
-; UNROLL-NO-IC-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; UNROLL-NO-IC-NEXT:    br label [[EXIT:%.*]]
 ; UNROLL-NO-IC:       scalar.ph:
-; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
-; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; UNROLL-NO-IC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 0, [[ENTRY]] ]
+; UNROLL-NO-IC-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; UNROLL-NO-IC-NEXT:    br label [[LOOP:%.*]]
 ; UNROLL-NO-IC:       loop:
 ; UNROLL-NO-IC-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -6037,7 +6007,7 @@ define void @pr52460_first_order_recurrence_truncated_iv(ptr noalias %src, ptr %
 ; INTERLEAVE-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP50:![0-9]+]]
 ; INTERLEAVE:       middle.block:
 ; INTERLEAVE-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[STEP_ADD]], i64 3
-; INTERLEAVE-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; INTERLEAVE-NEXT:    br label [[SCALAR_PH]]
 ; INTERLEAVE:       scalar.ph:
 ; INTERLEAVE-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], [[MIDDLE_BLOCK]] ], [ poison, [[ENTRY:%.*]] ]
 ; INTERLEAVE-NEXT:    br label [[LOOP:%.*]]
@@ -6056,7 +6026,7 @@ define void @pr52460_first_order_recurrence_truncated_iv(ptr noalias %src, ptr %
 ; INTERLEAVE-NEXT:    [[ADD:%.*]] = add i32 [[MUL]], [[IV_TRUNC]]
 ; INTERLEAVE-NEXT:    store i32 [[ADD]], ptr [[DST_GEP]], align 4
 ; INTERLEAVE-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[TRUNC_IV_NEXT]], 100
-; INTERLEAVE-NEXT:    br i1 [[EXITCOND]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP51:![0-9]+]]
+; INTERLEAVE-NEXT:    br i1 [[EXITCOND]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP51:![0-9]+]]
 ; INTERLEAVE:       exit:
 ; INTERLEAVE-NEXT:    ret void
 ;
@@ -6133,8 +6103,7 @@ define void @test_optimized_cast_induction_feeding_first_order_recurrence(i64 %n
 ; CHECK-NEXT:    [[VEC_IND]] = phi <2 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP20:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>
 ; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, ptr [[PTR:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i32, ptr [[TMP21]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[TMP20]], ptr [[TMP22]], align 4
+; CHECK-NEXT:    store <2 x i32> [[TMP20]], ptr [[TMP21]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], [[DOTSPLAT3]]
 ; CHECK-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -6357,9 +6326,8 @@ define void @test_optimized_cast_induction_feeding_first_order_recurrence(i64 %n
 ; UNROLL-NO-IC-NEXT:    [[TMP20:%.*]] = shufflevector <2 x i32> [[VECTOR_RECUR]], <2 x i32> [[VEC_IND]], <2 x i32> <i32 1, i32 2>
 ; UNROLL-NO-IC-NEXT:    [[TMP21:%.*]] = shufflevector <2 x i32> [[VEC_IND]], <2 x i32> [[STEP_ADD]], <2 x i32> <i32 1, i32 2>
 ; UNROLL-NO-IC-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i32, ptr [[PTR:%.*]], i64 [[INDEX]]
-; UNROLL-NO-IC-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, ptr [[TMP22]], i32 0
 ; UNROLL-NO-IC-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, ptr [[TMP22]], i32 2
-; UNROLL-NO-IC-NEXT:    store <2 x i32> [[TMP20]], ptr [[TMP23]], align 4
+; UNROLL-NO-IC-NEXT:    store <2 x i32> [[TMP20]], ptr [[TMP22]], align 4
 ; UNROLL-NO-IC-NEXT:    store <2 x i32> [[TMP21]], ptr [[TMP24]], align 4
 ; UNROLL-NO-IC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; UNROLL-NO-IC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[STEP_ADD]], [[TMP17]]
diff --git a/llvm/test/Transforms/LoopVectorize/induction_plus.ll b/llvm/test/Transforms/LoopVectorize/induction_plus.ll
index 83490d3..c87dc7b 100644
--- a/llvm/test/Transforms/LoopVectorize/induction_plus.ll
+++ b/llvm/test/Transforms/LoopVectorize/induction_plus.ll
@@ -10,8 +10,7 @@ define void @array_at_plus_one(i32 %n) {
 ; CHECK: [[VEC_IV_TRUNC:%.+]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, %vector.ph ], [ [[VEC_IV_TRUNC_NEXT:%.+]], %vector.body ]
 ; CHECK: [[T2:%.+]] = add nsw i64 %index, 12
 ; CHECK-NEXT: [[GEP:%.+]] = getelementptr inbounds [1024 x i32], ptr @array, i64 0, i64 [[T2]]
-; CHECK-NEXT: [[GEP0:%.+]] = getelementptr inbounds i32, ptr [[GEP]], i32 0
-; CHECK-NEXT: store <4 x i32> [[VEC_IV_TRUNC]], ptr [[GEP0]]
+; CHECK-NEXT: store <4 x i32> [[VEC_IV_TRUNC]], ptr [[GEP]]
 ; CHECK: [[VEC_IV_TRUNC_NEXT]] = add <4 x i32> [[VEC_IV_TRUNC]], splat (i32 4)
 ; CHECK: ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/instruction-only-used-outside-of-loop.ll b/llvm/test/Transforms/LoopVectorize/instruction-only-used-outside-of-loop.ll
index b9d3356..3330f2b 100644
--- a/llvm/test/Transforms/LoopVectorize/instruction-only-used-outside-of-loop.ll
+++ b/llvm/test/Transforms/LoopVectorize/instruction-only-used-outside-of-loop.ll
@@ -11,17 +11,16 @@ define i32 @one_direct_branch(ptr %src) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = xor <4 x i32> splat (i32 25500), [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <4 x i32> [[TMP3]], i32 3
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
@@ -68,17 +67,16 @@ define i32 @two_direct_branch(ptr %src) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = xor <4 x i32> splat (i32 25500), [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <4 x i32> [[TMP3]], i32 3
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
@@ -135,8 +133,7 @@ define i32 @cond_branch(i32 %a, ptr %src) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = xor <4 x i32> splat (i32 25500), [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp ne <4 x i32> [[VEC_IND]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP4]], <4 x i32> [[TMP3]], <4 x i32> splat (i32 10)
@@ -146,9 +143,9 @@ define i32 @cond_branch(i32 %a, ptr %src) {
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <4 x i32> [[PREDPHI]], i32 3
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
@@ -211,9 +208,9 @@ define i32 @optimizable_trunc_used_outside() {
 ; CHECK-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP1:%.*]] = extractelement <4 x i32> [[VEC_IND]], i32 3
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/interleave-and-scalarize-only.ll b/llvm/test/Transforms/LoopVectorize/interleave-and-scalarize-only.ll
index f038700..64caecc 100644
--- a/llvm/test/Transforms/LoopVectorize/interleave-and-scalarize-only.ll
+++ b/llvm/test/Transforms/LoopVectorize/interleave-and-scalarize-only.ll
@@ -284,7 +284,7 @@ define i16 @reduction_with_casts() {
 ; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[BIN_RDX:%.*]] = add i32 [[TMP3]], [[TMP2]]
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label %scalar.ph
+; CHECK-NEXT:    br label %scalar.ph
 ;
 entry:
   br label %loop
diff --git a/llvm/test/Transforms/LoopVectorize/interleave-with-i65-induction.ll b/llvm/test/Transforms/LoopVectorize/interleave-with-i65-induction.ll
index fb00762..8a48209 100644
--- a/llvm/test/Transforms/LoopVectorize/interleave-with-i65-induction.ll
+++ b/llvm/test/Transforms/LoopVectorize/interleave-with-i65-induction.ll
@@ -32,12 +32,11 @@ define void @i65_induction_with_negative_step(ptr %dst) {
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i65 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i64> [[STEP_ADD]], i32 3
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i65 [ -1000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i65 [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/interleaved-accesses-different-insert-position.ll b/llvm/test/Transforms/LoopVectorize/interleaved-accesses-different-insert-position.ll
index a89945f..651210d 100644
--- a/llvm/test/Transforms/LoopVectorize/interleaved-accesses-different-insert-position.ll
+++ b/llvm/test/Transforms/LoopVectorize/interleaved-accesses-different-insert-position.ll
@@ -21,16 +21,15 @@ define void @gep_for_first_member_does_not_dominate_insert_point(ptr %str, ptr n
 ; CHECK-NEXT:    [[STRIDED_VEC2:%.*]] = shufflevector <8 x i8> [[WIDE_VEC]], <8 x i8> poison, <4 x i32> <i32 1, i32 3, i32 5, i32 7>
 ; CHECK-NEXT:    [[TMP5:%.*]] = add <4 x i8> [[STRIDED_VEC2]], [[STRIDED_VEC]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[TMP5]], ptr [[TMP7]], align 1
+; CHECK-NEXT:    store <4 x i8> [[TMP5]], ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i64 [ 200, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll b/llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll
index 80ccc38..0afd2e1 100644
--- a/llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll
+++ b/llvm/test/Transforms/LoopVectorize/interleaved-accesses.ll
@@ -47,7 +47,7 @@ define void @test_array_load2_store2(i32 %C, i32 %D) {
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 512
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -124,7 +124,7 @@ define void @test_struct_array_load3_store3() {
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -207,7 +207,7 @@ define i32 @test_struct_load4(ptr nocapture readonly %S) {
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP4]])
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -280,7 +280,7 @@ define void @test_struct_store4(ptr noalias nocapture readonly %A, ptr noalias n
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.cond.cleanup:
@@ -366,7 +366,7 @@ define void @test_reversed_load2_store2(ptr noalias nocapture readonly %A, ptr n
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.cond.cleanup:
@@ -616,7 +616,7 @@ define void @load_gap_reverse(ptr noalias nocapture %P1, ptr noalias nocapture %
 ; CHECK-NEXT:    [[TMP29:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP29]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -678,7 +678,7 @@ define void @mixed_load2_store2(ptr noalias nocapture readonly %A, ptr noalias n
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 512
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.cond.cleanup:
@@ -750,7 +750,7 @@ define void @mixed_load3_store3(ptr nocapture %A) {
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.cond.cleanup:
@@ -835,7 +835,7 @@ define void @int_float_struct(ptr nocapture readonly %A) #0 {
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP5:%.*]] = call fast float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> [[TMP3]])
 ; CHECK-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP2]])
-; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.cond.cleanup:
diff --git a/llvm/test/Transforms/LoopVectorize/invalidate-scev-at-scope-after-vectorization.ll b/llvm/test/Transforms/LoopVectorize/invalidate-scev-at-scope-after-vectorization.ll
index 3f56978..752a0a0 100644
--- a/llvm/test/Transforms/LoopVectorize/invalidate-scev-at-scope-after-vectorization.ll
+++ b/llvm/test/Transforms/LoopVectorize/invalidate-scev-at-scope-after-vectorization.ll
@@ -16,18 +16,11 @@ define void @test_invalidate_scevs_at_scope(ptr %p) {
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP0:%.*]] = load i32, ptr [[P]], align 4
-; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TMP0]], i64 0
-; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP1:%.*]] = add <4 x i32> [[BROADCAST_SPLAT]], [[VEC_IND]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
-; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP2]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x i32> [[TMP1]], i32 3
-; CHECK-NEXT:    br i1 false, label %[[EXIT_1:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 100, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_1:.*]]
@@ -37,42 +30,44 @@ define void @test_invalidate_scevs_at_scope(ptr %p) {
 ; CHECK-NEXT:    [[ADD_1:%.*]] = add i32 [[TMP4]], [[IV_1]]
 ; CHECK-NEXT:    [[IV_1_NEXT]] = add i32 [[IV_1]], 1
 ; CHECK-NEXT:    [[C_1:%.*]] = icmp eq i32 [[IV_1]], 100
-; CHECK-NEXT:    br i1 [[C_1]], label %[[EXIT_1]], label %[[LOOP_1]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[C_1]], label %[[EXIT_1:.*]], label %[[LOOP_1]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       [[EXIT_1]]:
-; CHECK-NEXT:    [[ADD_LCSSA:%.*]] = phi i32 [ [[ADD_1]], %[[LOOP_1]] ], [ [[TMP3]], %[[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[DOTLCSSA:%.*]] = phi i32 [ [[TMP4]], %[[LOOP_1]] ]
+; CHECK-NEXT:    [[ADD_LCSSA1:%.*]] = phi i32 [ [[ADD_1]], %[[LOOP_1]] ]
+; CHECK-NEXT:    [[ADD_LCSSA:%.*]] = add i32 [[DOTLCSSA]], 100
 ; CHECK-NEXT:    [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[ADD_LCSSA]], i32 100)
-; CHECK-NEXT:    [[TMP5:%.*]] = sub i32 [[SMAX]], [[ADD_LCSSA]]
+; CHECK-NEXT:    [[TMP3:%.*]] = add i32 [[SMAX]], -100
+; CHECK-NEXT:    [[TMP5:%.*]] = sub i32 [[TMP3]], [[DOTLCSSA]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = zext i32 [[TMP5]] to i64
 ; CHECK-NEXT:    [[TMP7:%.*]] = add nuw nsw i64 [[TMP6]], 1
 ; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP7]], 4
-; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH2:.*]], label %[[VECTOR_PH3:.*]]
-; CHECK:       [[VECTOR_PH3]]:
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH1:.*]], label %[[VECTOR_PH2:.*]]
+; CHECK:       [[VECTOR_PH2]]:
 ; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP7]], 4
 ; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP7]], [[N_MOD_VF]]
-; CHECK-NEXT:    br label %[[VECTOR_BODY4:.*]]
-; CHECK:       [[VECTOR_BODY4]]:
-; CHECK-NEXT:    [[INDEX5:%.*]] = phi i64 [ 0, %[[VECTOR_PH3]] ], [ [[INDEX_NEXT8:%.*]], %[[VECTOR_BODY4]] ]
-; CHECK-NEXT:    [[VEC_IND6:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH3]] ], [ [[VEC_IND_NEXT7:%.*]], %[[VECTOR_BODY4]] ]
+; CHECK-NEXT:    br label %[[VECTOR_BODY3:.*]]
+; CHECK:       [[VECTOR_BODY3]]:
+; CHECK-NEXT:    [[INDEX5:%.*]] = phi i64 [ 0, %[[VECTOR_PH2]] ], [ [[INDEX_NEXT8:%.*]], %[[VECTOR_BODY3]] ]
+; CHECK-NEXT:    [[VEC_IND6:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH2]] ], [ [[VEC_IND_NEXT7:%.*]], %[[VECTOR_BODY3]] ]
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[P]], i64 [[INDEX5]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[TMP9]], i32 0
-; CHECK-NEXT:    store <4 x i64> [[VEC_IND6]], ptr [[TMP10]], align 4
+; CHECK-NEXT:    store <4 x i64> [[VEC_IND6]], ptr [[TMP9]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT8]] = add nuw i64 [[INDEX5]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT7]] = add <4 x i64> [[VEC_IND6]], splat (i64 4)
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT8]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP11]], label %[[MIDDLE_BLOCK1:.*]], label %[[VECTOR_BODY4]], !llvm.loop [[LOOP4:![0-9]+]]
-; CHECK:       [[MIDDLE_BLOCK1]]:
+; CHECK-NEXT:    br i1 [[TMP11]], label %[[MIDDLE_BLOCK6:.*]], label %[[VECTOR_BODY3]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK:       [[MIDDLE_BLOCK6]]:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP7]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT_2:.*]], label %[[SCALAR_PH2]]
-; CHECK:       [[SCALAR_PH2]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL9:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK1]] ], [ 0, %[[EXIT_1]] ]
+; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT_2:.*]], label %[[SCALAR_PH1]]
+; CHECK:       [[SCALAR_PH1]]:
+; CHECK-NEXT:    [[BC_RESUME_VAL7:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK6]] ], [ 0, %[[EXIT_1]] ]
 ; CHECK-NEXT:    br label %[[LOOP_2:.*]]
 ; CHECK:       [[LOOP_2]]:
-; CHECK-NEXT:    [[IV_2:%.*]] = phi i64 [ [[BC_RESUME_VAL9]], %[[SCALAR_PH2]] ], [ [[IV_2_NEXT:%.*]], %[[LOOP_2]] ]
+; CHECK-NEXT:    [[IV_2:%.*]] = phi i64 [ [[BC_RESUME_VAL7]], %[[SCALAR_PH1]] ], [ [[IV_2_NEXT:%.*]], %[[LOOP_2]] ]
 ; CHECK-NEXT:    [[IV_2_TRUNC:%.*]] = trunc i64 [[IV_2]] to i32
 ; CHECK-NEXT:    [[IV_2_NEXT]] = add i64 [[IV_2]], 1
 ; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i64, ptr [[P]], i64 [[IV_2]]
 ; CHECK-NEXT:    store i64 [[IV_2]], ptr [[GEP]], align 4
-; CHECK-NEXT:    [[ADD_2:%.*]] = add i32 [[ADD_LCSSA]], [[IV_2_TRUNC]]
+; CHECK-NEXT:    [[ADD_2:%.*]] = add i32 [[ADD_LCSSA1]], [[IV_2_TRUNC]]
 ; CHECK-NEXT:    [[C_2:%.*]] = icmp slt i32 [[ADD_2]], 100
 ; CHECK-NEXT:    br i1 [[C_2]], label %[[LOOP_2]], label %[[EXIT_2]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       [[EXIT_2]]:
diff --git a/llvm/test/Transforms/LoopVectorize/is_fpclass.ll b/llvm/test/Transforms/LoopVectorize/is_fpclass.ll
index 9f82c93d..6eeeace 100644
--- a/llvm/test/Transforms/LoopVectorize/is_fpclass.ll
+++ b/llvm/test/Transforms/LoopVectorize/is_fpclass.ll
@@ -15,15 +15,14 @@ define void @d() {
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr float, ptr @d, i64 [[INDEX]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = call <2 x i1> @llvm.is.fpclass.v2f32(<2 x float> [[BROADCAST_SPLAT]], i32 0)
 ; CHECK-NEXT:    [[TMP3:%.*]] = select <2 x i1> [[TMP2]], <2 x float> zeroinitializer, <2 x float> splat (float 1.000000e+00)
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <2 x float> [[TMP3]], ptr [[TMP4]], align 4
+; CHECK-NEXT:    store <2 x float> [[TMP3]], ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 128, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[I:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[I7:%.*]], [[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/iv-select-cmp-decreasing.ll b/llvm/test/Transforms/LoopVectorize/iv-select-cmp-decreasing.ll
index eb9c1cd..a0068f0 100644
--- a/llvm/test/Transforms/LoopVectorize/iv-select-cmp-decreasing.ll
+++ b/llvm/test/Transforms/LoopVectorize/iv-select-cmp-decreasing.ll
@@ -30,10 +30,10 @@ define i64 @select_decreasing_induction_icmp_const_start(ptr %a) {
 ; IC1VF4-NEXT:    [[TMP6:%.*]] = call i64 @llvm.vector.reduce.smin.v4i64(<4 x i64> [[TMP4]])
 ; IC1VF4-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i64 [[TMP6]], 9223372036854775807
 ; IC1VF4-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i64 [[TMP6]], i64 331
-; IC1VF4-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IC1VF4-NEXT:    br label %[[EXIT:.*]]
 ; IC1VF4:       [[SCALAR_PH]]:
-; IC1VF4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ -1, %[[MIDDLE_BLOCK]] ], [ 19999, %[[ENTRY]] ]
-; IC1VF4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ 331, %[[ENTRY]] ]
+; IC1VF4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 19999, %[[ENTRY]] ]
+; IC1VF4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ 331, %[[ENTRY]] ]
 ; IC1VF4-NEXT:    br label %[[LOOP:.*]]
 ; IC1VF4:       [[LOOP]]:
 ; IC1VF4-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -102,10 +102,10 @@ define i64 @select_decreasing_induction_icmp_const_start(ptr %a) {
 ; IC4VF4-NEXT:    [[TMP18:%.*]] = call i64 @llvm.vector.reduce.smin.v4i64(<4 x i64> [[RDX_MINMAX11]])
 ; IC4VF4-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i64 [[TMP18]], 9223372036854775807
 ; IC4VF4-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i64 [[TMP18]], i64 331
-; IC4VF4-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IC4VF4-NEXT:    br label %[[EXIT:.*]]
 ; IC4VF4:       [[SCALAR_PH]]:
-; IC4VF4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ -1, %[[MIDDLE_BLOCK]] ], [ 19999, %[[ENTRY]] ]
-; IC4VF4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ 331, %[[ENTRY]] ]
+; IC4VF4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 19999, %[[ENTRY]] ]
+; IC4VF4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ 331, %[[ENTRY]] ]
 ; IC4VF4-NEXT:    br label %[[LOOP:.*]]
 ; IC4VF4:       [[LOOP]]:
 ; IC4VF4-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -162,10 +162,10 @@ define i64 @select_decreasing_induction_icmp_const_start(ptr %a) {
 ; IC4VF1-NEXT:    [[RDX_MINMAX5:%.*]] = call i64 @llvm.smin.i64(i64 [[RDX_MINMAX4]], i64 [[TMP18]])
 ; IC4VF1-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i64 [[RDX_MINMAX5]], 9223372036854775807
 ; IC4VF1-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i64 [[RDX_MINMAX5]], i64 331
-; IC4VF1-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IC4VF1-NEXT:    br label %[[EXIT:.*]]
 ; IC4VF1:       [[SCALAR_PH]]:
-; IC4VF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ -1, %[[MIDDLE_BLOCK]] ], [ 19999, %[[ENTRY]] ]
-; IC4VF1-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ 331, %[[ENTRY]] ]
+; IC4VF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 19999, %[[ENTRY]] ]
+; IC4VF1-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ 331, %[[ENTRY]] ]
 ; IC4VF1-NEXT:    br label %[[LOOP:.*]]
 ; IC4VF1:       [[LOOP]]:
 ; IC4VF1-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -232,10 +232,10 @@ define i16 @select_decreasing_induction_icmp_table_i16(i16 noundef %val) {
 ; IC1VF4-NEXT:    [[TMP7:%.*]] = call i16 @llvm.vector.reduce.smin.v4i16(<4 x i16> [[TMP5]])
 ; IC1VF4-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i16 [[TMP7]], 32767
 ; IC1VF4-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i16 [[TMP7]], i16 0
-; IC1VF4-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IC1VF4-NEXT:    br label %[[EXIT:.*]]
 ; IC1VF4:       [[SCALAR_PH]]:
-; IC1VF4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, %[[MIDDLE_BLOCK]] ], [ 12, %[[ENTRY]] ]
-; IC1VF4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i16 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; IC1VF4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 12, %[[ENTRY]] ]
+; IC1VF4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i16 [ 0, %[[ENTRY]] ]
 ; IC1VF4-NEXT:    br label %[[LOOP:.*]]
 ; IC1VF4:       [[LOOP]]:
 ; IC1VF4-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -532,10 +532,10 @@ define i16 @select_decreasing_induction_icmp_table_i16(i16 noundef %val) {
 ; IC4VF1-NEXT:    [[RDX_MINMAX5:%.*]] = call i16 @llvm.smin.i16(i16 [[RDX_MINMAX4]], i16 [[TMP22]])
 ; IC4VF1-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i16 [[RDX_MINMAX5]], 32767
 ; IC4VF1-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i16 [[RDX_MINMAX5]], i16 0
-; IC4VF1-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IC4VF1-NEXT:    br label %[[EXIT:.*]]
 ; IC4VF1:       [[SCALAR_PH]]:
-; IC4VF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, %[[MIDDLE_BLOCK]] ], [ 12, %[[ENTRY]] ]
-; IC4VF1-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i16 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; IC4VF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 12, %[[ENTRY]] ]
+; IC4VF1-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i16 [ 0, %[[ENTRY]] ]
 ; IC4VF1-NEXT:    br label %[[LOOP:.*]]
 ; IC4VF1:       [[LOOP]]:
 ; IC4VF1-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -603,10 +603,10 @@ define i16 @select_decreasing_induction_icmp_table_half(half noundef %val) {
 ; IC1VF4-NEXT:    [[TMP7:%.*]] = call i16 @llvm.vector.reduce.smin.v4i16(<4 x i16> [[TMP5]])
 ; IC1VF4-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i16 [[TMP7]], 32767
 ; IC1VF4-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i16 [[TMP7]], i16 0
-; IC1VF4-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IC1VF4-NEXT:    br label %[[EXIT:.*]]
 ; IC1VF4:       [[SCALAR_PH]]:
-; IC1VF4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, %[[MIDDLE_BLOCK]] ], [ 12, %[[ENTRY]] ]
-; IC1VF4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i16 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; IC1VF4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 12, %[[ENTRY]] ]
+; IC1VF4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i16 [ 0, %[[ENTRY]] ]
 ; IC1VF4-NEXT:    br label %[[LOOP:.*]]
 ; IC1VF4:       [[LOOP]]:
 ; IC1VF4-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -903,10 +903,10 @@ define i16 @select_decreasing_induction_icmp_table_half(half noundef %val) {
 ; IC4VF1-NEXT:    [[RDX_MINMAX5:%.*]] = call i16 @llvm.smin.i16(i16 [[RDX_MINMAX4]], i16 [[TMP22]])
 ; IC4VF1-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i16 [[RDX_MINMAX5]], 32767
 ; IC4VF1-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i16 [[RDX_MINMAX5]], i16 0
-; IC4VF1-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IC4VF1-NEXT:    br label %[[EXIT:.*]]
 ; IC4VF1:       [[SCALAR_PH]]:
-; IC4VF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, %[[MIDDLE_BLOCK]] ], [ 12, %[[ENTRY]] ]
-; IC4VF1-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i16 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; IC4VF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 12, %[[ENTRY]] ]
+; IC4VF1-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i16 [ 0, %[[ENTRY]] ]
 ; IC4VF1-NEXT:    br label %[[LOOP:.*]]
 ; IC4VF1:       [[LOOP]]:
 ; IC4VF1-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -971,10 +971,10 @@ define i64 @select_decreasing_induction_icmp_iv_unsigned(ptr %a) {
 ; IC1VF4-NEXT:    [[TMP6:%.*]] = call i64 @llvm.vector.reduce.umin.v4i64(<4 x i64> [[TMP4]])
 ; IC1VF4-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i64 [[TMP6]], -1
 ; IC1VF4-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i64 [[TMP6]], i64 331
-; IC1VF4-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IC1VF4-NEXT:    br label %[[EXIT:.*]]
 ; IC1VF4:       [[SCALAR_PH]]:
-; IC1VF4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ -1, %[[MIDDLE_BLOCK]] ], [ 9223372036854775807, %[[ENTRY]] ]
-; IC1VF4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ 331, %[[ENTRY]] ]
+; IC1VF4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 9223372036854775807, %[[ENTRY]] ]
+; IC1VF4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ 331, %[[ENTRY]] ]
 ; IC1VF4-NEXT:    br label %[[LOOP:.*]]
 ; IC1VF4:       [[LOOP]]:
 ; IC1VF4-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -1043,10 +1043,10 @@ define i64 @select_decreasing_induction_icmp_iv_unsigned(ptr %a) {
 ; IC4VF4-NEXT:    [[TMP18:%.*]] = call i64 @llvm.vector.reduce.umin.v4i64(<4 x i64> [[RDX_MINMAX11]])
 ; IC4VF4-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i64 [[TMP18]], -1
 ; IC4VF4-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i64 [[TMP18]], i64 331
-; IC4VF4-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IC4VF4-NEXT:    br label %[[EXIT:.*]]
 ; IC4VF4:       [[SCALAR_PH]]:
-; IC4VF4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ -1, %[[MIDDLE_BLOCK]] ], [ 9223372036854775807, %[[ENTRY]] ]
-; IC4VF4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ 331, %[[ENTRY]] ]
+; IC4VF4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 9223372036854775807, %[[ENTRY]] ]
+; IC4VF4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ 331, %[[ENTRY]] ]
 ; IC4VF4-NEXT:    br label %[[LOOP:.*]]
 ; IC4VF4:       [[LOOP]]:
 ; IC4VF4-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -1103,10 +1103,10 @@ define i64 @select_decreasing_induction_icmp_iv_unsigned(ptr %a) {
 ; IC4VF1-NEXT:    [[RDX_MINMAX5:%.*]] = call i64 @llvm.umin.i64(i64 [[RDX_MINMAX4]], i64 [[TMP18]])
 ; IC4VF1-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i64 [[RDX_MINMAX5]], -1
 ; IC4VF1-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i64 [[RDX_MINMAX5]], i64 331
-; IC4VF1-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IC4VF1-NEXT:    br label %[[EXIT:.*]]
 ; IC4VF1:       [[SCALAR_PH]]:
-; IC4VF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ -1, %[[MIDDLE_BLOCK]] ], [ 9223372036854775807, %[[ENTRY]] ]
-; IC4VF1-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ 331, %[[ENTRY]] ]
+; IC4VF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 9223372036854775807, %[[ENTRY]] ]
+; IC4VF1-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ 331, %[[ENTRY]] ]
 ; IC4VF1-NEXT:    br label %[[LOOP:.*]]
 ; IC4VF1:       [[LOOP]]:
 ; IC4VF1-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/iv-select-cmp-nested-loop.ll b/llvm/test/Transforms/LoopVectorize/iv-select-cmp-nested-loop.ll
index c341605..24c5602 100644
--- a/llvm/test/Transforms/LoopVectorize/iv-select-cmp-nested-loop.ll
+++ b/llvm/test/Transforms/LoopVectorize/iv-select-cmp-nested-loop.ll
@@ -25,8 +25,7 @@ define i64 @select_iv_def_from_outer_loop(ptr %a, i64 %start, i64 %n) {
 ; CHECK-VF4IC1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP5:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP0]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
 ; CHECK-VF4IC1-NEXT:    [[TMP4:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 3)
 ; CHECK-VF4IC1-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI]], [[TMP4]]
 ; CHECK-VF4IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -83,11 +82,10 @@ define i64 @select_iv_def_from_outer_loop(ptr %a, i64 %start, i64 %n) {
 ; CHECK-VF4IC4-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP13:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC4-NEXT:    [[VEC_PHI3:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP14:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP0]], i64 [[TMP1]]
-; CHECK-VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i64>, ptr [[TMP5]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i64>, ptr [[TMP6]], align 8
diff --git a/llvm/test/Transforms/LoopVectorize/iv-select-cmp-no-wrap.ll b/llvm/test/Transforms/LoopVectorize/iv-select-cmp-no-wrap.ll
index eefb327..c958ea7 100644
--- a/llvm/test/Transforms/LoopVectorize/iv-select-cmp-no-wrap.ll
+++ b/llvm/test/Transforms/LoopVectorize/iv-select-cmp-no-wrap.ll
@@ -16,11 +16,9 @@ define i64 @select_icmp_nuw_nsw(ptr %a, ptr %b, i64 %ii, i64 %n) {
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ splat (i64 -9223372036854775808), %[[VECTOR_PH]] ], [ [[TMP6:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP9]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP10]], align 8
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP9]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp sgt <4 x i64> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; CHECK-NEXT:    [[TMP6]] = select <4 x i1> [[TMP5]], <4 x i64> [[VEC_IND]], <4 x i64> [[VEC_PHI]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP9]], 4
@@ -88,11 +86,9 @@ define i64 @select_icmp_nsw(ptr %a, ptr %b, i64 %ii, i64 %n) {
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ splat (i64 -9223372036854775808), %[[VECTOR_PH]] ], [ [[TMP6:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP9]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP10]], align 8
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP9]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp sgt <4 x i64> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; CHECK-NEXT:    [[TMP6]] = select <4 x i1> [[TMP5]], <4 x i64> [[VEC_IND]], <4 x i64> [[VEC_PHI]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP9]], 4
diff --git a/llvm/test/Transforms/LoopVectorize/iv-select-cmp-trunc.ll b/llvm/test/Transforms/LoopVectorize/iv-select-cmp-trunc.ll
index 6a2e3df..1054482 100644
--- a/llvm/test/Transforms/LoopVectorize/iv-select-cmp-trunc.ll
+++ b/llvm/test/Transforms/LoopVectorize/iv-select-cmp-trunc.ll
@@ -26,8 +26,7 @@ define i32 @select_icmp_const_truncated_iv_widened_exit(ptr %a, i32 %n) {
 ; CHECK-VF4IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 -2147483648), %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; CHECK-VF4IC1-NEXT:    [[TMP3:%.*]] = icmp sgt <4 x i64> [[WIDE_LOAD]], splat (i64 3)
 ; CHECK-VF4IC1-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i32> [[VEC_IND]], <4 x i32> [[VEC_PHI]]
 ; CHECK-VF4IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -86,11 +85,10 @@ define i32 @select_icmp_const_truncated_iv_widened_exit(ptr %a, i32 %n) {
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_2:%.*]] = add <4 x i32> [[STEP_ADD]], splat (i32 4)
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_3:%.*]] = add <4 x i32> [[STEP_ADD_2]], splat (i32 4)
 ; CHECK-VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i64>, ptr [[TMP5]], align 8
@@ -251,8 +249,7 @@ define i32 @select_icmp_const_truncated_iv_const_exit(ptr %a) {
 ; CHECK-VF4IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 -2147483648), %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; CHECK-VF4IC1-NEXT:    [[TMP3:%.*]] = icmp sgt <4 x i64> [[WIDE_LOAD]], splat (i64 3)
 ; CHECK-VF4IC1-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i32> [[VEC_IND]], <4 x i32> [[VEC_PHI]]
 ; CHECK-VF4IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -263,10 +260,10 @@ define i32 @select_icmp_const_truncated_iv_const_exit(ptr %a) {
 ; CHECK-VF4IC1-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.smax.v4i32(<4 x i32> [[TMP4]])
 ; CHECK-VF4IC1-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i32 [[TMP6]], -2147483648
 ; CHECK-VF4IC1-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i32 [[TMP6]], i32 331
-; CHECK-VF4IC1-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-VF4IC1-NEXT:    br label %[[EXIT:.*]]
 ; CHECK-VF4IC1:       [[SCALAR_PH]]:
-; CHECK-VF4IC1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 20000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-VF4IC1-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ 331, %[[ENTRY]] ]
+; CHECK-VF4IC1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
+; CHECK-VF4IC1-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 331, %[[ENTRY]] ]
 ; CHECK-VF4IC1-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK-VF4IC1:       [[FOR_BODY]]:
 ; CHECK-VF4IC1-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
@@ -300,11 +297,10 @@ define i32 @select_icmp_const_truncated_iv_const_exit(ptr %a) {
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_2:%.*]] = add <4 x i32> [[STEP_ADD]], splat (i32 4)
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_3:%.*]] = add <4 x i32> [[STEP_ADD_2]], splat (i32 4)
 ; CHECK-VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i64>, ptr [[TMP5]], align 8
@@ -327,10 +323,10 @@ define i32 @select_icmp_const_truncated_iv_const_exit(ptr %a) {
 ; CHECK-VF4IC4-NEXT:    [[TMP15:%.*]] = call i32 @llvm.vector.reduce.smax.v4i32(<4 x i32> [[RDX_MINMAX8]])
 ; CHECK-VF4IC4-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i32 [[TMP15]], -2147483648
 ; CHECK-VF4IC4-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i32 [[TMP15]], i32 331
-; CHECK-VF4IC4-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-VF4IC4-NEXT:    br label %[[EXIT:.*]]
 ; CHECK-VF4IC4:       [[SCALAR_PH]]:
-; CHECK-VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 20000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-VF4IC4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ 331, %[[ENTRY]] ]
+; CHECK-VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
+; CHECK-VF4IC4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 331, %[[ENTRY]] ]
 ; CHECK-VF4IC4-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK-VF4IC4:       [[FOR_BODY]]:
 ; CHECK-VF4IC4-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
@@ -391,10 +387,10 @@ define i32 @select_icmp_const_truncated_iv_const_exit(ptr %a) {
 ; CHECK-VF1IC4-NEXT:    [[RDX_MINMAX5:%.*]] = call i32 @llvm.smax.i32(i32 [[RDX_MINMAX4]], i32 [[TMP24]])
 ; CHECK-VF1IC4-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i32 [[RDX_MINMAX5]], -2147483648
 ; CHECK-VF1IC4-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i32 [[RDX_MINMAX5]], i32 331
-; CHECK-VF1IC4-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-VF1IC4-NEXT:    br label %[[EXIT:.*]]
 ; CHECK-VF1IC4:       [[SCALAR_PH]]:
-; CHECK-VF1IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 20000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-VF1IC4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ 331, %[[ENTRY]] ]
+; CHECK-VF1IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
+; CHECK-VF1IC4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 331, %[[ENTRY]] ]
 ; CHECK-VF1IC4-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK-VF1IC4:       [[FOR_BODY]]:
 ; CHECK-VF1IC4-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
@@ -444,8 +440,7 @@ define i32 @select_fcmp_max_valid_const_ub(ptr %a) {
 ; CHECK-VF4IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ splat (i32 -2147483648), %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-VF4IC1-NEXT:    [[TMP3:%.*]] = fcmp fast olt <4 x float> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-VF4IC1-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i32> [[VEC_IND]], <4 x i32> [[VEC_PHI]]
 ; CHECK-VF4IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -456,10 +451,10 @@ define i32 @select_fcmp_max_valid_const_ub(ptr %a) {
 ; CHECK-VF4IC1-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.smax.v4i32(<4 x i32> [[TMP4]])
 ; CHECK-VF4IC1-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i32 [[TMP6]], -2147483648
 ; CHECK-VF4IC1-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i32 [[TMP6]], i32 -1
-; CHECK-VF4IC1-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-VF4IC1-NEXT:    br label %[[EXIT:.*]]
 ; CHECK-VF4IC1:       [[SCALAR_PH]]:
-; CHECK-VF4IC1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 2147483648, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-VF4IC1-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ -1, %[[ENTRY]] ]
+; CHECK-VF4IC1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
+; CHECK-VF4IC1-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ -1, %[[ENTRY]] ]
 ; CHECK-VF4IC1-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK-VF4IC1:       [[FOR_BODY]]:
 ; CHECK-VF4IC1-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
@@ -493,11 +488,10 @@ define i32 @select_fcmp_max_valid_const_ub(ptr %a) {
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_2:%.*]] = add <4 x i32> [[STEP_ADD]], splat (i32 4)
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_3:%.*]] = add <4 x i32> [[STEP_ADD_2]], splat (i32 4)
 ; CHECK-VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
@@ -520,10 +514,10 @@ define i32 @select_fcmp_max_valid_const_ub(ptr %a) {
 ; CHECK-VF4IC4-NEXT:    [[TMP15:%.*]] = call i32 @llvm.vector.reduce.smax.v4i32(<4 x i32> [[RDX_MINMAX8]])
 ; CHECK-VF4IC4-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i32 [[TMP15]], -2147483648
 ; CHECK-VF4IC4-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i32 [[TMP15]], i32 -1
-; CHECK-VF4IC4-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-VF4IC4-NEXT:    br label %[[EXIT:.*]]
 ; CHECK-VF4IC4:       [[SCALAR_PH]]:
-; CHECK-VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 2147483648, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-VF4IC4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ -1, %[[ENTRY]] ]
+; CHECK-VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
+; CHECK-VF4IC4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ -1, %[[ENTRY]] ]
 ; CHECK-VF4IC4-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK-VF4IC4:       [[FOR_BODY]]:
 ; CHECK-VF4IC4-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
@@ -584,10 +578,10 @@ define i32 @select_fcmp_max_valid_const_ub(ptr %a) {
 ; CHECK-VF1IC4-NEXT:    [[RDX_MINMAX5:%.*]] = call i32 @llvm.smax.i32(i32 [[RDX_MINMAX4]], i32 [[TMP24]])
 ; CHECK-VF1IC4-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i32 [[RDX_MINMAX5]], -2147483648
 ; CHECK-VF1IC4-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i32 [[RDX_MINMAX5]], i32 -1
-; CHECK-VF1IC4-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-VF1IC4-NEXT:    br label %[[EXIT:.*]]
 ; CHECK-VF1IC4:       [[SCALAR_PH]]:
-; CHECK-VF1IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 2147483648, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-VF1IC4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ -1, %[[ENTRY]] ]
+; CHECK-VF1IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
+; CHECK-VF1IC4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ -1, %[[ENTRY]] ]
 ; CHECK-VF1IC4-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK-VF1IC4:       [[FOR_BODY]]:
 ; CHECK-VF1IC4-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
@@ -642,8 +636,7 @@ define i32 @select_icmp_truncated_unsigned_iv_range(ptr %a) {
 ; CHECK-VF4IC1-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 2147483646, i32 2147483647, i32 -2147483648, i32 -2147483647>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[IV:%.*]] = add i64 2147483646, [[INDEX]]
 ; CHECK-VF4IC1-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
-; CHECK-VF4IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[ARRAYIDX]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[ARRAYIDX]], align 4
 ; CHECK-VF4IC1-NEXT:    [[TMP2:%.*]] = icmp sgt <4 x i32> [[WIDE_LOAD]], splat (i32 3)
 ; CHECK-VF4IC1-NEXT:    [[TMP3]] = select <4 x i1> [[TMP2]], <4 x i32> [[VEC_IND]], <4 x i32> [[VEC_PHI]]
 ; CHECK-VF4IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -654,10 +647,10 @@ define i32 @select_icmp_truncated_unsigned_iv_range(ptr %a) {
 ; CHECK-VF4IC1-NEXT:    [[TMP5:%.*]] = call i32 @llvm.vector.reduce.umax.v4i32(<4 x i32> [[TMP3]])
 ; CHECK-VF4IC1-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i32 [[TMP5]], 0
 ; CHECK-VF4IC1-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i32 [[TMP5]], i32 331
-; CHECK-VF4IC1-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-VF4IC1-NEXT:    br label %[[EXIT:.*]]
 ; CHECK-VF4IC1:       [[SCALAR_PH]]:
-; CHECK-VF4IC1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4294967294, %[[MIDDLE_BLOCK]] ], [ 2147483646, %[[ENTRY]] ]
-; CHECK-VF4IC1-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ 331, %[[ENTRY]] ]
+; CHECK-VF4IC1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 2147483646, %[[ENTRY]] ]
+; CHECK-VF4IC1-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 331, %[[ENTRY]] ]
 ; CHECK-VF4IC1-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK-VF4IC1:       [[FOR_BODY]]:
 ; CHECK-VF4IC1-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
@@ -692,11 +685,10 @@ define i32 @select_icmp_truncated_unsigned_iv_range(ptr %a) {
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_3:%.*]] = add <4 x i32> [[STEP_ADD_2]], splat (i32 4)
 ; CHECK-VF4IC4-NEXT:    [[OFFSET_IDX:%.*]] = add i64 2147483646, [[INDEX]]
 ; CHECK-VF4IC4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[OFFSET_IDX]]
-; CHECK-VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP0]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
@@ -719,10 +711,10 @@ define i32 @select_icmp_truncated_unsigned_iv_range(ptr %a) {
 ; CHECK-VF4IC4-NEXT:    [[TMP14:%.*]] = call i32 @llvm.vector.reduce.umax.v4i32(<4 x i32> [[RDX_MINMAX8]])
 ; CHECK-VF4IC4-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i32 [[TMP14]], 0
 ; CHECK-VF4IC4-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i32 [[TMP14]], i32 331
-; CHECK-VF4IC4-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-VF4IC4-NEXT:    br label %[[EXIT:.*]]
 ; CHECK-VF4IC4:       [[SCALAR_PH]]:
-; CHECK-VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4294967294, %[[MIDDLE_BLOCK]] ], [ 2147483646, %[[ENTRY]] ]
-; CHECK-VF4IC4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ 331, %[[ENTRY]] ]
+; CHECK-VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 2147483646, %[[ENTRY]] ]
+; CHECK-VF4IC4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 331, %[[ENTRY]] ]
 ; CHECK-VF4IC4-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK-VF4IC4:       [[FOR_BODY]]:
 ; CHECK-VF4IC4-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
@@ -785,10 +777,10 @@ define i32 @select_icmp_truncated_unsigned_iv_range(ptr %a) {
 ; CHECK-VF1IC4-NEXT:    [[RDX_MINMAX6:%.*]] = call i32 @llvm.umax.i32(i32 [[RDX_MINMAX5]], i32 [[TMP22]])
 ; CHECK-VF1IC4-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i32 [[RDX_MINMAX6]], 0
 ; CHECK-VF1IC4-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i32 [[RDX_MINMAX6]], i32 331
-; CHECK-VF1IC4-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-VF1IC4-NEXT:    br label %[[EXIT:.*]]
 ; CHECK-VF1IC4:       [[SCALAR_PH]]:
-; CHECK-VF1IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4294967294, %[[MIDDLE_BLOCK]] ], [ 2147483646, %[[ENTRY]] ]
-; CHECK-VF1IC4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ], [ 331, %[[ENTRY]] ]
+; CHECK-VF1IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 2147483646, %[[ENTRY]] ]
+; CHECK-VF1IC4-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 331, %[[ENTRY]] ]
 ; CHECK-VF1IC4-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK-VF1IC4:       [[FOR_BODY]]:
 ; CHECK-VF1IC4-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/iv-select-cmp.ll b/llvm/test/Transforms/LoopVectorize/iv-select-cmp.ll
index eab5d5e..fcaff55 100644
--- a/llvm/test/Transforms/LoopVectorize/iv-select-cmp.ll
+++ b/llvm/test/Transforms/LoopVectorize/iv-select-cmp.ll
@@ -18,8 +18,7 @@ define i64 @select_icmp_const_1(ptr %a, i64 %n) {
 ; CHECK-VF4IC1-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ splat (i64 -9223372036854775808), %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; CHECK-VF4IC1-NEXT:    [[TMP3:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 3)
 ; CHECK-VF4IC1-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i64> [[VEC_IND]], <4 x i64> [[VEC_PHI]]
 ; CHECK-VF4IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -70,11 +69,10 @@ define i64 @select_icmp_const_1(ptr %a, i64 %n) {
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_2:%.*]] = add <4 x i64> [[STEP_ADD]], splat (i64 4)
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_3:%.*]] = add <4 x i64> [[STEP_ADD_2]], splat (i64 4)
 ; CHECK-VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i64>, ptr [[TMP5]], align 8
@@ -213,8 +211,7 @@ define i64 @select_icmp_const_2(ptr %a, i64 %n) {
 ; CHECK-VF4IC1-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ splat (i64 -9223372036854775808), %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; CHECK-VF4IC1-NEXT:    [[TMP3:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 3)
 ; CHECK-VF4IC1-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i64> [[VEC_PHI]], <4 x i64> [[VEC_IND]]
 ; CHECK-VF4IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -265,11 +262,10 @@ define i64 @select_icmp_const_2(ptr %a, i64 %n) {
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_2:%.*]] = add <4 x i64> [[STEP_ADD]], splat (i64 4)
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_3:%.*]] = add <4 x i64> [[STEP_ADD_2]], splat (i64 4)
 ; CHECK-VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i64>, ptr [[TMP5]], align 8
@@ -408,8 +404,7 @@ define i64 @select_icmp_const_3_variable_rdx_start(ptr %a, i64 %rdx.start, i64 %
 ; CHECK-VF4IC1-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ splat (i64 -9223372036854775808), %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; CHECK-VF4IC1-NEXT:    [[TMP3:%.*]] = icmp eq <4 x i64> [[WIDE_LOAD]], splat (i64 3)
 ; CHECK-VF4IC1-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i64> [[VEC_IND]], <4 x i64> [[VEC_PHI]]
 ; CHECK-VF4IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -460,11 +455,10 @@ define i64 @select_icmp_const_3_variable_rdx_start(ptr %a, i64 %rdx.start, i64 %
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_2:%.*]] = add <4 x i64> [[STEP_ADD]], splat (i64 4)
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_3:%.*]] = add <4 x i64> [[STEP_ADD_2]], splat (i64 4)
 ; CHECK-VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i64>, ptr [[TMP5]], align 8
@@ -603,8 +597,7 @@ define i64 @select_fcmp_const_fast(ptr %a, i64 %n) {
 ; CHECK-VF4IC1-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ splat (i64 -9223372036854775808), %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-VF4IC1-NEXT:    [[TMP3:%.*]] = fcmp fast ueq <4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00)
 ; CHECK-VF4IC1-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i64> [[VEC_IND]], <4 x i64> [[VEC_PHI]]
 ; CHECK-VF4IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -655,11 +648,10 @@ define i64 @select_fcmp_const_fast(ptr %a, i64 %n) {
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_2:%.*]] = add <4 x i64> [[STEP_ADD]], splat (i64 4)
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_3:%.*]] = add <4 x i64> [[STEP_ADD_2]], splat (i64 4)
 ; CHECK-VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
@@ -798,8 +790,7 @@ define i64 @select_fcmp_const(ptr %a, i64 %n) {
 ; CHECK-VF4IC1-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ splat (i64 -9223372036854775808), %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-VF4IC1-NEXT:    [[TMP3:%.*]] = fcmp ueq <4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00)
 ; CHECK-VF4IC1-NEXT:    [[TMP4]] = select <4 x i1> [[TMP3]], <4 x i64> [[VEC_IND]], <4 x i64> [[VEC_PHI]]
 ; CHECK-VF4IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -850,11 +841,10 @@ define i64 @select_fcmp_const(ptr %a, i64 %n) {
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_2:%.*]] = add <4 x i64> [[STEP_ADD]], splat (i64 4)
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_3:%.*]] = add <4 x i64> [[STEP_ADD_2]], splat (i64 4)
 ; CHECK-VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
@@ -993,11 +983,9 @@ define i64 @select_icmp(ptr %a, ptr %b, i64 %rdx.start, i64 %n) {
 ; CHECK-VF4IC1-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ splat (i64 -9223372036854775808), %[[VECTOR_PH]] ], [ [[TMP6:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; CHECK-VF4IC1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP3]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; CHECK-VF4IC1-NEXT:    [[TMP5:%.*]] = icmp sgt <4 x i64> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; CHECK-VF4IC1-NEXT:    [[TMP6]] = select <4 x i1> [[TMP5]], <4 x i64> [[VEC_IND]], <4 x i64> [[VEC_PHI]]
 ; CHECK-VF4IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -1050,20 +1038,18 @@ define i64 @select_icmp(ptr %a, ptr %b, i64 %rdx.start, i64 %n) {
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_2:%.*]] = add <4 x i64> [[STEP_ADD]], splat (i64 4)
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_3:%.*]] = add <4 x i64> [[STEP_ADD_2]], splat (i64 4)
 ; CHECK-VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i64>, ptr [[TMP5]], align 8
 ; CHECK-VF4IC4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i64>, ptr [[TMP7]], align 8
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i64>, ptr [[TMP6]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x i64>, ptr [[TMP8]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x i64>, ptr [[TMP9]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD10:%.*]] = load <4 x i64>, ptr [[TMP10]], align 8
@@ -1216,11 +1202,9 @@ define i64 @select_fcmp(ptr %a, ptr %b, i64 %rdx.start, i64 %n) {
 ; CHECK-VF4IC1-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 0, i64 1, i64 2, i64 3>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ splat (i64 -9223372036854775808), %[[VECTOR_PH]] ], [ [[TMP6:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-VF4IC1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP3]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
 ; CHECK-VF4IC1-NEXT:    [[TMP5:%.*]] = fcmp ogt <4 x float> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; CHECK-VF4IC1-NEXT:    [[TMP6]] = select <4 x i1> [[TMP5]], <4 x i64> [[VEC_IND]], <4 x i64> [[VEC_PHI]]
 ; CHECK-VF4IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -1273,20 +1257,18 @@ define i64 @select_fcmp(ptr %a, ptr %b, i64 %rdx.start, i64 %n) {
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_2:%.*]] = add <4 x i64> [[STEP_ADD]], splat (i64 4)
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_3:%.*]] = add <4 x i64> [[STEP_ADD_2]], splat (i64 4)
 ; CHECK-VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
 ; CHECK-VF4IC4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x float>, ptr [[TMP7]], align 4
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x float>, ptr [[TMP6]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x float>, ptr [[TMP8]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x float>, ptr [[TMP9]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD10:%.*]] = load <4 x float>, ptr [[TMP10]], align 4
@@ -1440,11 +1422,9 @@ define i64 @select_icmp_min_valid_iv_start(ptr %a, ptr %b, i64 %rdx.start, i64 %
 ; CHECK-VF4IC1-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 -9223372036854775807, i64 -9223372036854775806, i64 -9223372036854775805, i64 -9223372036854775804>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ splat (i64 -9223372036854775808), %[[VECTOR_PH]] ], [ [[TMP6:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; CHECK-VF4IC1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP3]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; CHECK-VF4IC1-NEXT:    [[TMP5:%.*]] = icmp sgt <4 x i64> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-VF4IC1-NEXT:    [[TMP6]] = select <4 x i1> [[TMP5]], <4 x i64> [[VEC_IND]], <4 x i64> [[VEC_PHI]]
 ; CHECK-VF4IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -1501,20 +1481,18 @@ define i64 @select_icmp_min_valid_iv_start(ptr %a, ptr %b, i64 %rdx.start, i64 %
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_2:%.*]] = add <4 x i64> [[STEP_ADD]], splat (i64 4)
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_3:%.*]] = add <4 x i64> [[STEP_ADD_2]], splat (i64 4)
 ; CHECK-VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i64>, ptr [[TMP5]], align 8
 ; CHECK-VF4IC4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x i64>, ptr [[TMP7]], align 8
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x i64>, ptr [[TMP6]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x i64>, ptr [[TMP8]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD10:%.*]] = load <4 x i64>, ptr [[TMP9]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x i64>, ptr [[TMP10]], align 8
@@ -1677,11 +1655,9 @@ define i64 @select_icmp_unsigned_iv_range(ptr %a, ptr %b, i64 %rdx.start) {
 ; CHECK-VF4IC1-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 -9223372036854775808, i64 -9223372036854775807, i64 -9223372036854775806, i64 -9223372036854775805>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i64> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP5:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[TMP0]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP0]], align 8
 ; CHECK-VF4IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
 ; CHECK-VF4IC1-NEXT:    [[TMP4:%.*]] = icmp sgt <4 x i64> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; CHECK-VF4IC1-NEXT:    [[TMP5]] = select <4 x i1> [[TMP4]], <4 x i64> [[VEC_IND]], <4 x i64> [[VEC_PHI]]
 ; CHECK-VF4IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -1692,7 +1668,7 @@ define i64 @select_icmp_unsigned_iv_range(ptr %a, ptr %b, i64 %rdx.start) {
 ; CHECK-VF4IC1-NEXT:    [[TMP7:%.*]] = call i64 @llvm.vector.reduce.umax.v4i64(<4 x i64> [[TMP5]])
 ; CHECK-VF4IC1-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i64 [[TMP7]], 0
 ; CHECK-VF4IC1-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i64 [[TMP7]], i64 [[RDX_START]]
-; CHECK-VF4IC1-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-VF4IC1-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK-VF4IC1:       [[SCALAR_PH]]:
 ; CHECK-VF4IC1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ -4, %[[MIDDLE_BLOCK]] ], [ -9223372036854775808, %[[ENTRY]] ]
 ; CHECK-VF4IC1-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i64 [ 9223372036854775804, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
@@ -1711,9 +1687,9 @@ define i64 @select_icmp_unsigned_iv_range(ptr %a, ptr %b, i64 %rdx.start) {
 ; CHECK-VF4IC1-NEXT:    [[INC]] = add nuw nsw i64 [[IV_I]], 1
 ; CHECK-VF4IC1-NEXT:    [[INC3]] = add nsw i64 [[IV_J]], 1
 ; CHECK-VF4IC1-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INC]], 9223372036854775806
-; CHECK-VF4IC1-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
+; CHECK-VF4IC1-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
 ; CHECK-VF4IC1:       [[EXIT]]:
-; CHECK-VF4IC1-NEXT:    [[COND_LCSSA:%.*]] = phi i64 [ [[COND]], %[[FOR_BODY]] ], [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ]
+; CHECK-VF4IC1-NEXT:    [[COND_LCSSA:%.*]] = phi i64 [ [[COND]], %[[FOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    ret i64 [[COND_LCSSA]]
 ;
 ; CHECK-VF4IC4-LABEL: define i64 @select_icmp_unsigned_iv_range(
@@ -1733,20 +1709,18 @@ define i64 @select_icmp_unsigned_iv_range(ptr %a, ptr %b, i64 %rdx.start) {
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_2:%.*]] = add <4 x i64> [[STEP_ADD]], splat (i64 4)
 ; CHECK-VF4IC4-NEXT:    [[STEP_ADD_3:%.*]] = add <4 x i64> [[STEP_ADD_2]], splat (i64 4)
 ; CHECK-VF4IC4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[TMP0]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP0]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP0]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP0]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP0]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i64>, ptr [[TMP3]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
 ; CHECK-VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i64>, ptr [[TMP6]], align 8
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i64>, ptr [[TMP5]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x i64>, ptr [[TMP7]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x i64>, ptr [[TMP8]], align 8
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD10:%.*]] = load <4 x i64>, ptr [[TMP9]], align 8
@@ -1769,7 +1743,7 @@ define i64 @select_icmp_unsigned_iv_range(ptr %a, ptr %b, i64 %rdx.start) {
 ; CHECK-VF4IC4-NEXT:    [[TMP19:%.*]] = call i64 @llvm.vector.reduce.umax.v4i64(<4 x i64> [[RDX_MINMAX12]])
 ; CHECK-VF4IC4-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i64 [[TMP19]], 0
 ; CHECK-VF4IC4-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i64 [[TMP19]], i64 [[RDX_START]]
-; CHECK-VF4IC4-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-VF4IC4-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK-VF4IC4:       [[SCALAR_PH]]:
 ; CHECK-VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ -16, %[[MIDDLE_BLOCK]] ], [ -9223372036854775808, %[[ENTRY]] ]
 ; CHECK-VF4IC4-NEXT:    [[BC_RESUME_VAL13:%.*]] = phi i64 [ 9223372036854775792, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
@@ -1788,9 +1762,9 @@ define i64 @select_icmp_unsigned_iv_range(ptr %a, ptr %b, i64 %rdx.start) {
 ; CHECK-VF4IC4-NEXT:    [[INC]] = add nuw nsw i64 [[IV_I]], 1
 ; CHECK-VF4IC4-NEXT:    [[INC3]] = add nsw i64 [[IV_J]], 1
 ; CHECK-VF4IC4-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INC]], 9223372036854775806
-; CHECK-VF4IC4-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
+; CHECK-VF4IC4-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
 ; CHECK-VF4IC4:       [[EXIT]]:
-; CHECK-VF4IC4-NEXT:    [[COND_LCSSA:%.*]] = phi i64 [ [[COND]], %[[FOR_BODY]] ], [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ]
+; CHECK-VF4IC4-NEXT:    [[COND_LCSSA:%.*]] = phi i64 [ [[COND]], %[[FOR_BODY]] ]
 ; CHECK-VF4IC4-NEXT:    ret i64 [[COND_LCSSA]]
 ;
 ; CHECK-VF1IC4-LABEL: define i64 @select_icmp_unsigned_iv_range(
@@ -1845,7 +1819,7 @@ define i64 @select_icmp_unsigned_iv_range(ptr %a, ptr %b, i64 %rdx.start) {
 ; CHECK-VF1IC4-NEXT:    [[RDX_MINMAX5:%.*]] = call i64 @llvm.umax.i64(i64 [[RDX_MINMAX4]], i64 [[TMP29]])
 ; CHECK-VF1IC4-NEXT:    [[RDX_SELECT_CMP:%.*]] = icmp ne i64 [[RDX_MINMAX5]], 0
 ; CHECK-VF1IC4-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[RDX_SELECT_CMP]], i64 [[RDX_MINMAX5]], i64 [[RDX_START]]
-; CHECK-VF1IC4-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-VF1IC4-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK-VF1IC4:       [[SCALAR_PH]]:
 ; CHECK-VF1IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ -4, %[[MIDDLE_BLOCK]] ], [ -9223372036854775808, %[[ENTRY]] ]
 ; CHECK-VF1IC4-NEXT:    [[BC_RESUME_VAL6:%.*]] = phi i64 [ 9223372036854775804, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
@@ -1864,9 +1838,9 @@ define i64 @select_icmp_unsigned_iv_range(ptr %a, ptr %b, i64 %rdx.start) {
 ; CHECK-VF1IC4-NEXT:    [[INC]] = add nuw nsw i64 [[IV_I1]], 1
 ; CHECK-VF1IC4-NEXT:    [[INC3]] = add nsw i64 [[IV_J]], 1
 ; CHECK-VF1IC4-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INC]], 9223372036854775806
-; CHECK-VF1IC4-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT]], label %[[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
+; CHECK-VF1IC4-NEXT:    br i1 [[EXITCOND_NOT]], label %[[EXIT:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
 ; CHECK-VF1IC4:       [[EXIT]]:
-; CHECK-VF1IC4-NEXT:    [[COND_LCSSA:%.*]] = phi i64 [ [[COND]], %[[FOR_BODY]] ], [ [[RDX_SELECT]], %[[MIDDLE_BLOCK]] ]
+; CHECK-VF1IC4-NEXT:    [[COND_LCSSA:%.*]] = phi i64 [ [[COND]], %[[FOR_BODY]] ]
 ; CHECK-VF1IC4-NEXT:    ret i64 [[COND_LCSSA]]
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/iv_outside_user.ll b/llvm/test/Transforms/LoopVectorize/iv_outside_user.ll
index 7b77354..97d3385 100644
--- a/llvm/test/Transforms/LoopVectorize/iv_outside_user.ll
+++ b/llvm/test/Transforms/LoopVectorize/iv_outside_user.ll
@@ -101,9 +101,9 @@ define i32 @constpre()  {
 ; CHECK-NEXT:    [[TMP0:%.*]] = icmp eq i32 [[INDEX_NEXT]], 16
 ; CHECK-NEXT:    br i1 [[TMP0]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[FOR_END:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, %[[MIDDLE_BLOCK]] ], [ 32, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 32, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[INC_PHI:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
@@ -133,19 +133,19 @@ define ptr @geppre(ptr %ptr) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
 ; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
-; CHECK-NEXT:    [[IND_END:%.*]] = getelementptr i8, ptr [[PTR]], i64 512
+; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[PTR]], i64 512
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; CHECK-NEXT:    [[TMP0:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32
-; CHECK-NEXT:    br i1 [[TMP0]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32
+; CHECK-NEXT:    br i1 [[TMP1]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    [[IND_ESCAPE:%.*]] = getelementptr i8, ptr [[IND_END]], i64 -16
-; CHECK-NEXT:    br i1 true, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    [[IND_ESCAPE:%.*]] = getelementptr i8, ptr [[TMP0]], i64 -16
+; CHECK-NEXT:    br label %[[FOR_END:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 32, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi ptr [ [[IND_END]], %[[MIDDLE_BLOCK]] ], [ [[PTR]], %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi ptr [ [[PTR]], %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[INC_PHI:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[FOR_BODY]] ]
@@ -407,16 +407,15 @@ define i64 @iv_scalar_steps_and_outside_users(ptr %ptr) {
 ; VEC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VEC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VEC-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[PTR]], i64 [[INDEX]]
-; VEC-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; VEC-NEXT:    store <2 x i64> [[VEC_IND]], ptr [[TMP2]], align 4
+; VEC-NEXT:    store <2 x i64> [[VEC_IND]], ptr [[TMP1]], align 4
 ; VEC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VEC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; VEC-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1002
 ; VEC-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
 ; VEC:       [[MIDDLE_BLOCK]]:
-; VEC-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; VEC-NEXT:    br label %[[EXIT:.*]]
 ; VEC:       [[SCALAR_PH]]:
-; VEC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1002, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; VEC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; VEC-NEXT:    br label %[[LOOP:.*]]
 ; VEC:       [[LOOP]]:
 ; VEC-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -446,9 +445,9 @@ define i64 @iv_scalar_steps_and_outside_users(ptr %ptr) {
 ; INTERLEAVE-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1002
 ; INTERLEAVE-NEXT:    br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
 ; INTERLEAVE:       [[MIDDLE_BLOCK]]:
-; INTERLEAVE-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; INTERLEAVE-NEXT:    br label %[[EXIT:.*]]
 ; INTERLEAVE:       [[SCALAR_PH]]:
-; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1002, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; INTERLEAVE-NEXT:    br label %[[LOOP:.*]]
 ; INTERLEAVE:       [[LOOP]]:
 ; INTERLEAVE-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -490,17 +489,16 @@ define i32 @iv_2_dead_in_loop_only_used_outside(ptr %ptr) {
 ; VEC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VEC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VEC-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[PTR]], i64 [[INDEX]]
-; VEC-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; VEC-NEXT:    store <2 x i64> [[VEC_IND]], ptr [[TMP2]], align 4
+; VEC-NEXT:    store <2 x i64> [[VEC_IND]], ptr [[TMP1]], align 4
 ; VEC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VEC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; VEC-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1002
 ; VEC-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
 ; VEC:       [[MIDDLE_BLOCK]]:
-; VEC-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; VEC-NEXT:    br label %[[EXIT:.*]]
 ; VEC:       [[SCALAR_PH]]:
-; VEC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1002, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; VEC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 2004, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; VEC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
+; VEC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
 ; VEC-NEXT:    br label %[[LOOP:.*]]
 ; VEC:       [[LOOP]]:
 ; VEC-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -532,10 +530,10 @@ define i32 @iv_2_dead_in_loop_only_used_outside(ptr %ptr) {
 ; INTERLEAVE-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1002
 ; INTERLEAVE-NEXT:    br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
 ; INTERLEAVE:       [[MIDDLE_BLOCK]]:
-; INTERLEAVE-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; INTERLEAVE-NEXT:    br label %[[EXIT:.*]]
 ; INTERLEAVE:       [[SCALAR_PH]]:
-; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1002, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; INTERLEAVE-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 2004, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
 ; INTERLEAVE-NEXT:    br label %[[LOOP:.*]]
 ; INTERLEAVE:       [[LOOP]]:
 ; INTERLEAVE-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -762,8 +760,7 @@ define float @fp_postinc_use_fadd(float %init, ptr noalias nocapture %A, i64 %N,
 ; VEC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VEC-NEXT:    [[VEC_IND:%.*]] = phi <2 x float> [ [[INDUCTION]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VEC-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; VEC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i32 0
-; VEC-NEXT:    store <2 x float> [[VEC_IND]], ptr [[TMP6]], align 4
+; VEC-NEXT:    store <2 x float> [[VEC_IND]], ptr [[TMP5]], align 4
 ; VEC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VEC-NEXT:    [[VEC_IND_NEXT]] = fadd fast <2 x float> [[VEC_IND]], [[DOTSPLAT4]]
 ; VEC-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -878,8 +875,7 @@ define float @fp_postinc_use_fadd_ops_swapped(float %init, ptr noalias nocapture
 ; VEC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VEC-NEXT:    [[VEC_IND:%.*]] = phi <2 x float> [ [[INDUCTION]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VEC-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; VEC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i32 0
-; VEC-NEXT:    store <2 x float> [[VEC_IND]], ptr [[TMP6]], align 4
+; VEC-NEXT:    store <2 x float> [[VEC_IND]], ptr [[TMP5]], align 4
 ; VEC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VEC-NEXT:    [[VEC_IND_NEXT]] = fadd fast <2 x float> [[VEC_IND]], [[DOTSPLAT4]]
 ; VEC-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -994,8 +990,7 @@ define float @fp_postinc_use_fsub(float %init, ptr noalias nocapture %A, i64 %N,
 ; VEC-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VEC-NEXT:    [[VEC_IND:%.*]] = phi <2 x float> [ [[INDUCTION]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VEC-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; VEC-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP5]], i32 0
-; VEC-NEXT:    store <2 x float> [[VEC_IND]], ptr [[TMP6]], align 4
+; VEC-NEXT:    store <2 x float> [[VEC_IND]], ptr [[TMP5]], align 4
 ; VEC-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VEC-NEXT:    [[VEC_IND_NEXT]] = fsub fast <2 x float> [[VEC_IND]], [[DOTSPLAT4]]
 ; VEC-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -1099,16 +1094,15 @@ define i32 @test_iv_uniform_with_outside_use_scev_simplification(ptr %dst) {
 ; VEC-NEXT:    [[TMP0:%.*]] = add i32 [[INDEX]], 0
 ; VEC-NEXT:    [[TMP6:%.*]] = add i32 [[INDEX]], 1
 ; VEC-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[DST]], i32 [[TMP0]]
-; VEC-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 0
-; VEC-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP2]], align 2
+; VEC-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP1]], align 2
 ; VEC-NEXT:    [[TMP5:%.*]] = add i32 [[STEP_2]], [[TMP6]]
 ; VEC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; VEC-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 8
 ; VEC-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
 ; VEC:       [[MIDDLE_BLOCK]]:
-; VEC-NEXT:    br i1 true, label %[[E_EXIT:.*]], label %[[SCALAR_PH]]
+; VEC-NEXT:    br label %[[E_EXIT:.*]]
 ; VEC:       [[SCALAR_PH]]:
-; VEC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 8, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; VEC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
 ; VEC-NEXT:    br label %[[LOOP:.*]]
 ; VEC:       [[LOOP]]:
 ; VEC-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -1141,9 +1135,9 @@ define i32 @test_iv_uniform_with_outside_use_scev_simplification(ptr %dst) {
 ; INTERLEAVE-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], 8
 ; INTERLEAVE-NEXT:    br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
 ; INTERLEAVE:       [[MIDDLE_BLOCK]]:
-; INTERLEAVE-NEXT:    br i1 true, label %[[E_EXIT:.*]], label %[[SCALAR_PH]]
+; INTERLEAVE-NEXT:    br label %[[E_EXIT:.*]]
 ; INTERLEAVE:       [[SCALAR_PH]]:
-; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 8, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
 ; INTERLEAVE-NEXT:    br label %[[LOOP:.*]]
 ; INTERLEAVE:       [[LOOP]]:
 ; INTERLEAVE-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -1203,9 +1197,9 @@ define i32 @test_iv_uniform_with_outside_use_scev_simplification_2(ptr %dst) {
 ; VEC-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
 ; VEC:       [[MIDDLE_BLOCK]]:
 ; VEC-NEXT:    [[TMP7:%.*]] = extractelement <2 x i32> [[TMP5]], i32 1
-; VEC-NEXT:    br i1 true, label %[[E_EXIT:.*]], label %[[SCALAR_PH]]
+; VEC-NEXT:    br label %[[E_EXIT:.*]]
 ; VEC:       [[SCALAR_PH]]:
-; VEC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 8, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; VEC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
 ; VEC-NEXT:    br label %[[LOOP:.*]]
 ; VEC:       [[LOOP]]:
 ; VEC-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -1241,9 +1235,9 @@ define i32 @test_iv_uniform_with_outside_use_scev_simplification_2(ptr %dst) {
 ; INTERLEAVE-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 4
 ; INTERLEAVE-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
 ; INTERLEAVE:       [[MIDDLE_BLOCK]]:
-; INTERLEAVE-NEXT:    br i1 true, label %[[E_EXIT:.*]], label %[[SCALAR_PH]]
+; INTERLEAVE-NEXT:    br label %[[E_EXIT:.*]]
 ; INTERLEAVE:       [[SCALAR_PH]]:
-; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 8, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
 ; INTERLEAVE-NEXT:    br label %[[LOOP:.*]]
 ; INTERLEAVE:       [[LOOP]]:
 ; INTERLEAVE-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -1286,20 +1280,14 @@ define i32 @iv_ext_used_outside( ptr %dst) {
 ; VEC-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; VEC:       [[VECTOR_BODY]]:
 ; VEC-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; VEC-NEXT:    [[VEC_IND:%.*]] = phi <2 x i16> [ <i16 0, i16 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VEC-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i16
 ; VEC-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw i32, ptr [[DST]], i16 [[OFFSET_IDX]]
-; VEC-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw i32, ptr [[TMP1]], i32 0
-; VEC-NEXT:    store <2 x i32> zeroinitializer, ptr [[TMP2]], align 4
-; VEC-NEXT:    [[TMP5:%.*]] = add nuw nsw <2 x i16> [[VEC_IND]], splat (i16 1)
-; VEC-NEXT:    [[TMP8:%.*]] = extractelement <2 x i16> [[TMP5]], i32 1
-; VEC-NEXT:    [[TMP7:%.*]] = zext nneg i16 [[TMP8]] to i32
+; VEC-NEXT:    store <2 x i32> zeroinitializer, ptr [[TMP1]], align 4
 ; VEC-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
-; VEC-NEXT:    [[VEC_IND_NEXT]] = add <2 x i16> [[VEC_IND]], splat (i16 2)
 ; VEC-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 128
 ; VEC-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
 ; VEC:       [[MIDDLE_BLOCK]]:
-; VEC-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; VEC-NEXT:    br label %[[SCALAR_PH]]
 ; VEC:       [[SCALAR_PH]]:
 ; VEC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 128, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; VEC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 128, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
@@ -1312,9 +1300,9 @@ define i32 @iv_ext_used_outside( ptr %dst) {
 ; VEC-NEXT:    [[IV_1_NEXT]] = add nuw nsw i16 [[IV_1]], 1
 ; VEC-NEXT:    [[IV_1_EXT]] = zext nneg i16 [[IV_1_NEXT]] to i32
 ; VEC-NEXT:    [[EC:%.*]] = icmp samesign ult i16 [[IV_1]], 128
-; VEC-NEXT:    br i1 [[EC]], label %[[LOOP]], label %[[EXIT]], {{!llvm.loop ![0-9]+}}
+; VEC-NEXT:    br i1 [[EC]], label %[[LOOP]], label %[[EXIT:.*]], {{!llvm.loop ![0-9]+}}
 ; VEC:       [[EXIT]]:
-; VEC-NEXT:    [[IV_1_EXT_LCSSA:%.*]] = phi i32 [ [[IV_1_EXT]], %[[LOOP]] ], [ [[TMP7]], %[[MIDDLE_BLOCK]] ]
+; VEC-NEXT:    [[IV_1_EXT_LCSSA:%.*]] = phi i32 [ [[IV_1_EXT]], %[[LOOP]] ]
 ; VEC-NEXT:    ret i32 [[IV_1_EXT_LCSSA]]
 ;
 ; INTERLEAVE-LABEL: define i32 @iv_ext_used_outside(
@@ -1331,13 +1319,11 @@ define i32 @iv_ext_used_outside( ptr %dst) {
 ; INTERLEAVE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw i32, ptr [[DST]], i16 [[TMP1]]
 ; INTERLEAVE-NEXT:    store i32 0, ptr [[TMP2]], align 4
 ; INTERLEAVE-NEXT:    store i32 0, ptr [[TMP3]], align 4
-; INTERLEAVE-NEXT:    [[TMP4:%.*]] = add nuw nsw i16 [[TMP1]], 1
-; INTERLEAVE-NEXT:    [[TMP5:%.*]] = zext nneg i16 [[TMP4]] to i32
 ; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; INTERLEAVE-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 128
 ; INTERLEAVE-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], {{!llvm.loop ![0-9]+}}
 ; INTERLEAVE:       [[MIDDLE_BLOCK]]:
-; INTERLEAVE-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; INTERLEAVE-NEXT:    br label %[[SCALAR_PH]]
 ; INTERLEAVE:       [[SCALAR_PH]]:
 ; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 128, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; INTERLEAVE-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 128, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
@@ -1350,9 +1336,9 @@ define i32 @iv_ext_used_outside( ptr %dst) {
 ; INTERLEAVE-NEXT:    [[IV_1_NEXT]] = add nuw nsw i16 [[IV_1]], 1
 ; INTERLEAVE-NEXT:    [[IV_1_EXT]] = zext nneg i16 [[IV_1_NEXT]] to i32
 ; INTERLEAVE-NEXT:    [[EC:%.*]] = icmp samesign ult i16 [[IV_1]], 128
-; INTERLEAVE-NEXT:    br i1 [[EC]], label %[[LOOP]], label %[[EXIT]], {{!llvm.loop ![0-9]+}}
+; INTERLEAVE-NEXT:    br i1 [[EC]], label %[[LOOP]], label %[[EXIT:.*]], {{!llvm.loop ![0-9]+}}
 ; INTERLEAVE:       [[EXIT]]:
-; INTERLEAVE-NEXT:    [[IV_1_EXT_LCSSA:%.*]] = phi i32 [ [[IV_1_EXT]], %[[LOOP]] ], [ [[TMP5]], %[[MIDDLE_BLOCK]] ]
+; INTERLEAVE-NEXT:    [[IV_1_EXT_LCSSA:%.*]] = phi i32 [ [[IV_1_EXT]], %[[LOOP]] ]
 ; INTERLEAVE-NEXT:    ret i32 [[IV_1_EXT_LCSSA]]
 ;
 entry:
@@ -1390,10 +1376,10 @@ define i64 @test_iv_increment_incremented(ptr %dst) {
 ; VEC-NEXT:    [[TMP6:%.*]] = add i64 [[TMP5]], 1
 ; VEC-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; VEC:       [[MIDDLE_BLOCK]]:
-; VEC-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; VEC-NEXT:    br label %[[EXIT:.*]]
 ; VEC:       [[SCALAR_PH]]:
-; VEC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1, %[[MIDDLE_BLOCK]] ], [ 3, %[[ENTRY]] ]
-; VEC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i64 [ 0, %[[MIDDLE_BLOCK]] ], [ 2, %[[ENTRY]] ]
+; VEC-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, %[[ENTRY]] ]
+; VEC-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i64 [ 2, %[[ENTRY]] ]
 ; VEC-NEXT:    br label %[[LOOP:.*]]
 ; VEC:       [[LOOP]]:
 ; VEC-NEXT:    [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], %[[LOOP]] ]
@@ -1423,10 +1409,10 @@ define i64 @test_iv_increment_incremented(ptr %dst) {
 ; INTERLEAVE-NEXT:    [[TMP3:%.*]] = add i64 [[TMP2]], 1
 ; INTERLEAVE-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; INTERLEAVE:       [[MIDDLE_BLOCK]]:
-; INTERLEAVE-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; INTERLEAVE-NEXT:    br label %[[EXIT:.*]]
 ; INTERLEAVE:       [[SCALAR_PH]]:
-; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1, %[[MIDDLE_BLOCK]] ], [ 3, %[[ENTRY]] ]
-; INTERLEAVE-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i64 [ 0, %[[MIDDLE_BLOCK]] ], [ 2, %[[ENTRY]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, %[[ENTRY]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i64 [ 2, %[[ENTRY]] ]
 ; INTERLEAVE-NEXT:    br label %[[LOOP:.*]]
 ; INTERLEAVE:       [[LOOP]]:
 ; INTERLEAVE-NEXT:    [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], %[[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/lcssa-crashes.ll b/llvm/test/Transforms/LoopVectorize/lcssa-crashes.ll
index 8d88c62..2084833 100644
--- a/llvm/test/Transforms/LoopVectorize/lcssa-crashes.ll
+++ b/llvm/test/Transforms/LoopVectorize/lcssa-crashes.ll
@@ -128,7 +128,7 @@ define i32 @pr57508(ptr %src) {
 ; CHECK-NEXT:    [[TMP0:%.*]] = icmp eq i64 [[INDEX_NEXT]], 2000
 ; CHECK-NEXT:    br i1 [[TMP0]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 2000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 2000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
@@ -139,7 +139,7 @@ define i32 @pr57508(ptr %src) {
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[LOCAL_NEXT]] = add i32 [[LOCAL]], 1
 ; CHECK-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV]], 2000
-; CHECK-NEXT:    br i1 [[EC]], label [[LOOP_EXIT]], label [[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label [[LOOP_EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       loop.exit:
 ; CHECK-NEXT:    unreachable
 ; CHECK:       bb:
diff --git a/llvm/test/Transforms/LoopVectorize/load-deref-pred-align.ll b/llvm/test/Transforms/LoopVectorize/load-deref-pred-align.ll
index 8a326c9..2c7d1bd 100644
--- a/llvm/test/Transforms/LoopVectorize/load-deref-pred-align.ll
+++ b/llvm/test/Transforms/LoopVectorize/load-deref-pred-align.ll
@@ -19,8 +19,7 @@ define i16 @test_access_size_not_multiple_of_align(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_LOAD_CONTINUE2:%.*]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i16> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[PRED_LOAD_CONTINUE2]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TEST_BASE:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp sge <2 x i8> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x i1> [[TMP3]], i32 0
 ; CHECK-NEXT:    br i1 [[TMP4]], label [[PRED_LOAD_IF:%.*]], label [[PRED_LOAD_CONTINUE:%.*]]
@@ -49,10 +48,10 @@ define i16 @test_access_size_not_multiple_of_align(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP17:%.*]] = call i16 @llvm.vector.reduce.add.v2i16(<2 x i16> [[TMP15]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i16 [ [[TMP17]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i16 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
@@ -116,8 +115,7 @@ define i32 @test_access_size_multiple_of_align_but_offset_by_1(i64 %len, ptr %te
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_LOAD_CONTINUE2:%.*]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP15:%.*]], [[PRED_LOAD_CONTINUE2]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TEST_BASE:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp sge <2 x i8> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <2 x i1> [[TMP3]], i32 0
 ; CHECK-NEXT:    br i1 [[TMP4]], label [[PRED_LOAD_IF:%.*]], label [[PRED_LOAD_CONTINUE:%.*]]
@@ -146,10 +144,10 @@ define i32 @test_access_size_multiple_of_align_but_offset_by_1(i64 %len, ptr %te
 ; CHECK-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP17:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[TMP15]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP17]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
@@ -230,12 +228,10 @@ define i32 @loop_requires_scev_predicate(ptr %dest, i32 %end) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE4:%.*]] ]
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[P1]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP10]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP9]], align 4
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp ne <2 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr [[P2]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr i32, ptr [[TMP12]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <2 x i32>, ptr [[TMP13]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <2 x i32>, ptr [[TMP12]], align 4
 ; CHECK-NEXT:    [[TMP14:%.*]] = extractelement <2 x i1> [[TMP11]], i32 0
 ; CHECK-NEXT:    br i1 [[TMP14]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
 ; CHECK:       pred.store.if:
@@ -378,9 +374,9 @@ define void @test_rev_loops_deref_loads(ptr nocapture noundef writeonly %dest) {
 ; CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ -1, [[MIDDLE_BLOCK]] ], [ 1023, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1023, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_INC:%.*]] ]
@@ -490,9 +486,9 @@ define void @test_rev_loops_non_deref_loads(ptr nocapture noundef writeonly %des
 ; CHECK-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ -1, [[MIDDLE_BLOCK]] ], [ 1023, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1023, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_INC:%.*]] ]
@@ -565,8 +561,7 @@ define i16 @test_strided_access(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i16> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TEST_BASE:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp sge <2 x i8> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = mul <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x i64> [[TMP4]], i32 0
@@ -585,10 +580,10 @@ define i16 @test_strided_access(i64 %len, ptr %test_base) {
 ; CHECK-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP15:%.*]] = call i16 @llvm.vector.reduce.add.v2i16(<2 x i16> [[TMP13]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 4096, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i16 [ [[TMP15]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i16 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
@@ -694,9 +689,9 @@ define void @test_rev_loops_strided_deref_loads(ptr nocapture noundef writeonly
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], 512
 ; CHECK-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ -1, [[MIDDLE_BLOCK]] ], [ 511, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 511, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[FOR_INC:%.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/load-deref-pred-neg-off.ll b/llvm/test/Transforms/LoopVectorize/load-deref-pred-neg-off.ll
index c6103f5..34c04de 100644
--- a/llvm/test/Transforms/LoopVectorize/load-deref-pred-neg-off.ll
+++ b/llvm/test/Transforms/LoopVectorize/load-deref-pred-neg-off.ll
@@ -52,10 +52,10 @@ define i8 @test_negative_off(i16 %len, ptr %test_base) {
 ; CHECK-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP20:%.*]] = call i8 @llvm.vector.reduce.add.v2i8(<2 x i8> [[TMP18]])
-; CHECK-NEXT:    br i1 true, label [[LOOP_EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ -988, [[MIDDLE_BLOCK]] ], [ -1000, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i8 [ [[TMP20]], [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ -1000, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i8 [ 0, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LATCH:%.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/load-deref-pred-poison-ub-ops-feeding-pointer.ll b/llvm/test/Transforms/LoopVectorize/load-deref-pred-poison-ub-ops-feeding-pointer.ll
index 3373c6d..70e730f 100644
--- a/llvm/test/Transforms/LoopVectorize/load-deref-pred-poison-ub-ops-feeding-pointer.ll
+++ b/llvm/test/Transforms/LoopVectorize/load-deref-pred-poison-ub-ops-feeding-pointer.ll
@@ -67,7 +67,7 @@ define void @ptr_depends_on_sdiv(ptr noalias %dst, i16 noundef %off) {
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i16> [[VEC_IND]], splat (i16 2)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br [[EXIT:label %.*]]
 ; CHECK:       [[SCALAR_PH]]:
 ;
 entry:
@@ -143,7 +143,7 @@ define void @ptr_depends_on_possibly_poison_value(ptr noalias %dst, i16 %off) {
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i16> [[VEC_IND]], splat (i16 2)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br [[EXIT:label %.*]]
 ; CHECK:       [[SCALAR_PH]]:
 ;
 entry:
@@ -190,8 +190,7 @@ define void @ptr_doesnt_depend_on_poison_or_ub(ptr noalias %dst, i16 noundef %of
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq <2 x i16> [[VEC_IND]], splat (i16 10)
 ; CHECK-NEXT:    [[TMP3:%.*]] = add i16 [[OFFSET_IDX]], [[TMP1]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i64, ptr @src, i16 [[TMP3]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i64, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP4]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <2 x i1> [[TMP2]], i32 0
 ; CHECK-NEXT:    br i1 [[TMP6]], label %[[PRED_STORE_IF:.*]], label %[[PRED_STORE_CONTINUE:.*]]
 ; CHECK:       [[PRED_STORE_IF]]:
@@ -214,7 +213,7 @@ define void @ptr_doesnt_depend_on_poison_or_ub(ptr noalias %dst, i16 noundef %of
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i16> [[VEC_IND]], splat (i16 2)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br [[EXIT:label %.*]]
 ; CHECK:       [[SCALAR_PH]]:
 ;
 entry:
@@ -290,7 +289,7 @@ define void @ptr_depends_on_possibly_poison_value_from_load(ptr noalias %dst) {
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i16> [[VEC_IND]], splat (i16 2)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br [[EXIT:label %.*]]
 ; CHECK:       [[SCALAR_PH]]:
 ;
 entry:
@@ -340,8 +339,7 @@ define void @ptr_depends_on_noundef_load(ptr noalias %dst) {
 ; CHECK-NEXT:    [[TMP3:%.*]] = add i16 [[TMP2]], [[TMP0]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = add i16 [[OFFSET_IDX]], [[TMP3]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i64, ptr @src, i16 [[TMP4]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i64, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP6]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP5]], align 1
 ; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <2 x i1> [[TMP1]], i32 0
 ; CHECK-NEXT:    br i1 [[TMP7]], label %[[PRED_STORE_IF:.*]], label %[[PRED_STORE_CONTINUE:.*]]
 ; CHECK:       [[PRED_STORE_IF]]:
@@ -364,7 +362,7 @@ define void @ptr_depends_on_noundef_load(ptr noalias %dst) {
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i16> [[VEC_IND]], splat (i16 2)
 ; CHECK-NEXT:    br i1 true, label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, [[EXIT:label %.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br [[EXIT:label %.*]]
 ; CHECK:       [[SCALAR_PH]]:
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/load-of-struct-deref-pred.ll b/llvm/test/Transforms/LoopVectorize/load-of-struct-deref-pred.ll
index b418fa7..f99e883c 100644
--- a/llvm/test/Transforms/LoopVectorize/load-of-struct-deref-pred.ll
+++ b/llvm/test/Transforms/LoopVectorize/load-of-struct-deref-pred.ll
@@ -17,24 +17,21 @@ define void @accesses_to_struct_dereferenceable(ptr noalias %dst) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = icmp ult <4 x i32> [[WIDE_LOAD]], zeroinitializer
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr [[STRUCT_FOO:%.*]], ptr @foo, i64 0, i32 1, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr [[STRUCT_FOO]], ptr @foo, i64 0, i32 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4
-; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP2]], <4 x i32> [[WIDE_LOAD2]], <4 x i32> [[WIDE_LOAD1]]
-; CHECK-NEXT:    store <4 x i32> [[PREDPHI]], ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP0]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult <4 x i32> [[WIDE_LOAD]], zeroinitializer
+; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr [[STRUCT_FOO:%.*]], ptr @foo, i64 0, i32 1, i64 [[INDEX]]
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr [[STRUCT_FOO]], ptr @foo, i64 0, i32 0, i64 [[INDEX]]
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
+; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP1]], <4 x i32> [[WIDE_LOAD2]], <4 x i32> [[WIDE_LOAD1]]
+; CHECK-NEXT:    store <4 x i32> [[PREDPHI]], ptr [[TMP0]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32000
-; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32000
+; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 32000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; CHECK:       loop.header:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
@@ -100,61 +97,58 @@ define void @accesses_to_struct_may_not_be_dereferenceable_due_to_loop_bound(ptr
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_LOAD_CONTINUE6:%.*]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = icmp ult <4 x i32> [[WIDE_LOAD]], zeroinitializer
-; CHECK-NEXT:    [[TMP3:%.*]] = xor <4 x i1> [[TMP2]], splat (i1 true)
-; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <4 x i1> [[TMP3]], i32 0
-; CHECK-NEXT:    br i1 [[TMP4]], label [[PRED_LOAD_IF:%.*]], label [[PRED_LOAD_CONTINUE:%.*]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP0]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult <4 x i32> [[WIDE_LOAD]], zeroinitializer
+; CHECK-NEXT:    [[TMP2:%.*]] = xor <4 x i1> [[TMP1]], splat (i1 true)
+; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x i1> [[TMP2]], i32 0
+; CHECK-NEXT:    br i1 [[TMP3]], label [[PRED_LOAD_IF:%.*]], label [[PRED_LOAD_CONTINUE:%.*]]
 ; CHECK:       pred.load.if:
-; CHECK-NEXT:    [[TMP5:%.*]] = add i64 [[INDEX]], 0
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_FOO:%.*]], ptr @foo, i64 0, i32 1, i64 [[TMP5]]
-; CHECK-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP6]], align 4
-; CHECK-NEXT:    [[TMP8:%.*]] = insertelement <4 x i32> poison, i32 [[TMP7]], i32 0
+; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_FOO:%.*]], ptr @foo, i64 0, i32 1, i64 [[TMP4]]
+; CHECK-NEXT:    [[TMP6:%.*]] = load i32, ptr [[TMP5]], align 4
+; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <4 x i32> poison, i32 [[TMP6]], i32 0
 ; CHECK-NEXT:    br label [[PRED_LOAD_CONTINUE]]
 ; CHECK:       pred.load.continue:
-; CHECK-NEXT:    [[TMP9:%.*]] = phi <4 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP8]], [[PRED_LOAD_IF]] ]
-; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <4 x i1> [[TMP3]], i32 1
-; CHECK-NEXT:    br i1 [[TMP10]], label [[PRED_LOAD_IF1:%.*]], label [[PRED_LOAD_CONTINUE2:%.*]]
+; CHECK-NEXT:    [[TMP8:%.*]] = phi <4 x i32> [ poison, [[VECTOR_BODY]] ], [ [[TMP7]], [[PRED_LOAD_IF]] ]
+; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <4 x i1> [[TMP2]], i32 1
+; CHECK-NEXT:    br i1 [[TMP9]], label [[PRED_LOAD_IF1:%.*]], label [[PRED_LOAD_CONTINUE2:%.*]]
 ; CHECK:       pred.load.if1:
-; CHECK-NEXT:    [[TMP11:%.*]] = add i64 [[INDEX]], 1
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_FOO]], ptr @foo, i64 0, i32 1, i64 [[TMP11]]
-; CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[TMP12]], align 4
-; CHECK-NEXT:    [[TMP14:%.*]] = insertelement <4 x i32> [[TMP9]], i32 [[TMP13]], i32 1
+; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX]], 1
+; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_FOO]], ptr @foo, i64 0, i32 1, i64 [[TMP10]]
+; CHECK-NEXT:    [[TMP12:%.*]] = load i32, ptr [[TMP11]], align 4
+; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <4 x i32> [[TMP8]], i32 [[TMP12]], i32 1
 ; CHECK-NEXT:    br label [[PRED_LOAD_CONTINUE2]]
 ; CHECK:       pred.load.continue2:
-; CHECK-NEXT:    [[TMP15:%.*]] = phi <4 x i32> [ [[TMP9]], [[PRED_LOAD_CONTINUE]] ], [ [[TMP14]], [[PRED_LOAD_IF1]] ]
-; CHECK-NEXT:    [[TMP16:%.*]] = extractelement <4 x i1> [[TMP3]], i32 2
-; CHECK-NEXT:    br i1 [[TMP16]], label [[PRED_LOAD_IF3:%.*]], label [[PRED_LOAD_CONTINUE4:%.*]]
+; CHECK-NEXT:    [[TMP14:%.*]] = phi <4 x i32> [ [[TMP8]], [[PRED_LOAD_CONTINUE]] ], [ [[TMP13]], [[PRED_LOAD_IF1]] ]
+; CHECK-NEXT:    [[TMP15:%.*]] = extractelement <4 x i1> [[TMP2]], i32 2
+; CHECK-NEXT:    br i1 [[TMP15]], label [[PRED_LOAD_IF3:%.*]], label [[PRED_LOAD_CONTINUE4:%.*]]
 ; CHECK:       pred.load.if3:
-; CHECK-NEXT:    [[TMP17:%.*]] = add i64 [[INDEX]], 2
-; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [[STRUCT_FOO]], ptr @foo, i64 0, i32 1, i64 [[TMP17]]
-; CHECK-NEXT:    [[TMP19:%.*]] = load i32, ptr [[TMP18]], align 4
-; CHECK-NEXT:    [[TMP20:%.*]] = insertelement <4 x i32> [[TMP15]], i32 [[TMP19]], i32 2
+; CHECK-NEXT:    [[TMP16:%.*]] = add i64 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_FOO]], ptr @foo, i64 0, i32 1, i64 [[TMP16]]
+; CHECK-NEXT:    [[TMP18:%.*]] = load i32, ptr [[TMP17]], align 4
+; CHECK-NEXT:    [[TMP19:%.*]] = insertelement <4 x i32> [[TMP14]], i32 [[TMP18]], i32 2
 ; CHECK-NEXT:    br label [[PRED_LOAD_CONTINUE4]]
 ; CHECK:       pred.load.continue4:
-; CHECK-NEXT:    [[TMP21:%.*]] = phi <4 x i32> [ [[TMP15]], [[PRED_LOAD_CONTINUE2]] ], [ [[TMP20]], [[PRED_LOAD_IF3]] ]
-; CHECK-NEXT:    [[TMP22:%.*]] = extractelement <4 x i1> [[TMP3]], i32 3
-; CHECK-NEXT:    br i1 [[TMP22]], label [[PRED_LOAD_IF5:%.*]], label [[PRED_LOAD_CONTINUE6]]
+; CHECK-NEXT:    [[TMP20:%.*]] = phi <4 x i32> [ [[TMP14]], [[PRED_LOAD_CONTINUE2]] ], [ [[TMP19]], [[PRED_LOAD_IF3]] ]
+; CHECK-NEXT:    [[TMP21:%.*]] = extractelement <4 x i1> [[TMP2]], i32 3
+; CHECK-NEXT:    br i1 [[TMP21]], label [[PRED_LOAD_IF5:%.*]], label [[PRED_LOAD_CONTINUE6]]
 ; CHECK:       pred.load.if5:
-; CHECK-NEXT:    [[TMP23:%.*]] = add i64 [[INDEX]], 3
-; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_FOO]], ptr @foo, i64 0, i32 1, i64 [[TMP23]]
-; CHECK-NEXT:    [[TMP25:%.*]] = load i32, ptr [[TMP24]], align 4
-; CHECK-NEXT:    [[TMP26:%.*]] = insertelement <4 x i32> [[TMP21]], i32 [[TMP25]], i32 3
+; CHECK-NEXT:    [[TMP22:%.*]] = add i64 [[INDEX]], 3
+; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_FOO]], ptr @foo, i64 0, i32 1, i64 [[TMP22]]
+; CHECK-NEXT:    [[TMP24:%.*]] = load i32, ptr [[TMP23]], align 4
+; CHECK-NEXT:    [[TMP25:%.*]] = insertelement <4 x i32> [[TMP20]], i32 [[TMP24]], i32 3
 ; CHECK-NEXT:    br label [[PRED_LOAD_CONTINUE6]]
 ; CHECK:       pred.load.continue6:
-; CHECK-NEXT:    [[TMP27:%.*]] = phi <4 x i32> [ [[TMP21]], [[PRED_LOAD_CONTINUE4]] ], [ [[TMP26]], [[PRED_LOAD_IF5]] ]
-; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr [[STRUCT_FOO]], ptr @foo, i64 0, i32 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr i32, ptr [[TMP28]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i32>, ptr [[TMP29]], align 4
-; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP2]], <4 x i32> [[WIDE_LOAD7]], <4 x i32> [[TMP27]]
-; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[PREDPHI]], ptr [[TMP30]], align 4
+; CHECK-NEXT:    [[TMP26:%.*]] = phi <4 x i32> [ [[TMP20]], [[PRED_LOAD_CONTINUE4]] ], [ [[TMP25]], [[PRED_LOAD_IF5]] ]
+; CHECK-NEXT:    [[TMP27:%.*]] = getelementptr [[STRUCT_FOO]], ptr @foo, i64 0, i32 0, i64 [[INDEX]]
+; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i32>, ptr [[TMP27]], align 4
+; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP1]], <4 x i32> [[WIDE_LOAD7]], <4 x i32> [[TMP26]]
+; CHECK-NEXT:    store <4 x i32> [[PREDPHI]], ptr [[TMP0]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP31:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32000
-; CHECK-NEXT:    br i1 [[TMP31]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK-NEXT:    [[TMP28:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32000
+; CHECK-NEXT:    br i1 [[TMP28]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 32000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
@@ -177,7 +171,7 @@ define void @accesses_to_struct_may_not_be_dereferenceable_due_to_loop_bound(ptr
 ; CHECK-NEXT:    store i32 [[TMP_0]], ptr [[GEP_DST]], align 4
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 32001
-; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[EXIT]], label [[LOOP_HEADER]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[EXIT:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
@@ -222,64 +216,61 @@ define void @accesses_to_struct_may_not_be_dereferenceable_access_size(ptr noali
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_LOAD_CONTINUE6:%.*]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = icmp ult <4 x i32> [[WIDE_LOAD]], zeroinitializer
-; CHECK-NEXT:    [[TMP3:%.*]] = xor <4 x i1> [[TMP2]], splat (i1 true)
-; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <4 x i1> [[TMP3]], i32 0
-; CHECK-NEXT:    br i1 [[TMP4]], label [[PRED_LOAD_IF:%.*]], label [[PRED_LOAD_CONTINUE:%.*]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP0]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult <4 x i32> [[WIDE_LOAD]], zeroinitializer
+; CHECK-NEXT:    [[TMP2:%.*]] = xor <4 x i1> [[TMP1]], splat (i1 true)
+; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x i1> [[TMP2]], i32 0
+; CHECK-NEXT:    br i1 [[TMP3]], label [[PRED_LOAD_IF:%.*]], label [[PRED_LOAD_CONTINUE:%.*]]
 ; CHECK:       pred.load.if:
-; CHECK-NEXT:    [[TMP5:%.*]] = add i64 [[INDEX]], 0
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [[STRUCT_FOO:%.*]], ptr @foo, i64 0, i32 1, i64 [[TMP5]]
-; CHECK-NEXT:    [[TMP7:%.*]] = load i64, ptr [[TMP6]], align 4
-; CHECK-NEXT:    [[TMP8:%.*]] = insertelement <4 x i64> poison, i64 [[TMP7]], i32 0
+; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 0
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_FOO:%.*]], ptr @foo, i64 0, i32 1, i64 [[TMP4]]
+; CHECK-NEXT:    [[TMP6:%.*]] = load i64, ptr [[TMP5]], align 4
+; CHECK-NEXT:    [[TMP7:%.*]] = insertelement <4 x i64> poison, i64 [[TMP6]], i32 0
 ; CHECK-NEXT:    br label [[PRED_LOAD_CONTINUE]]
 ; CHECK:       pred.load.continue:
-; CHECK-NEXT:    [[TMP9:%.*]] = phi <4 x i64> [ poison, [[VECTOR_BODY]] ], [ [[TMP8]], [[PRED_LOAD_IF]] ]
-; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <4 x i1> [[TMP3]], i32 1
-; CHECK-NEXT:    br i1 [[TMP10]], label [[PRED_LOAD_IF1:%.*]], label [[PRED_LOAD_CONTINUE2:%.*]]
+; CHECK-NEXT:    [[TMP8:%.*]] = phi <4 x i64> [ poison, [[VECTOR_BODY]] ], [ [[TMP7]], [[PRED_LOAD_IF]] ]
+; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <4 x i1> [[TMP2]], i32 1
+; CHECK-NEXT:    br i1 [[TMP9]], label [[PRED_LOAD_IF1:%.*]], label [[PRED_LOAD_CONTINUE2:%.*]]
 ; CHECK:       pred.load.if1:
-; CHECK-NEXT:    [[TMP11:%.*]] = add i64 [[INDEX]], 1
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds [[STRUCT_FOO]], ptr @foo, i64 0, i32 1, i64 [[TMP11]]
-; CHECK-NEXT:    [[TMP13:%.*]] = load i64, ptr [[TMP12]], align 4
-; CHECK-NEXT:    [[TMP14:%.*]] = insertelement <4 x i64> [[TMP9]], i64 [[TMP13]], i32 1
+; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX]], 1
+; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_FOO]], ptr @foo, i64 0, i32 1, i64 [[TMP10]]
+; CHECK-NEXT:    [[TMP12:%.*]] = load i64, ptr [[TMP11]], align 4
+; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <4 x i64> [[TMP8]], i64 [[TMP12]], i32 1
 ; CHECK-NEXT:    br label [[PRED_LOAD_CONTINUE2]]
 ; CHECK:       pred.load.continue2:
-; CHECK-NEXT:    [[TMP15:%.*]] = phi <4 x i64> [ [[TMP9]], [[PRED_LOAD_CONTINUE]] ], [ [[TMP14]], [[PRED_LOAD_IF1]] ]
-; CHECK-NEXT:    [[TMP16:%.*]] = extractelement <4 x i1> [[TMP3]], i32 2
-; CHECK-NEXT:    br i1 [[TMP16]], label [[PRED_LOAD_IF3:%.*]], label [[PRED_LOAD_CONTINUE4:%.*]]
+; CHECK-NEXT:    [[TMP14:%.*]] = phi <4 x i64> [ [[TMP8]], [[PRED_LOAD_CONTINUE]] ], [ [[TMP13]], [[PRED_LOAD_IF1]] ]
+; CHECK-NEXT:    [[TMP15:%.*]] = extractelement <4 x i1> [[TMP2]], i32 2
+; CHECK-NEXT:    br i1 [[TMP15]], label [[PRED_LOAD_IF3:%.*]], label [[PRED_LOAD_CONTINUE4:%.*]]
 ; CHECK:       pred.load.if3:
-; CHECK-NEXT:    [[TMP17:%.*]] = add i64 [[INDEX]], 2
-; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [[STRUCT_FOO]], ptr @foo, i64 0, i32 1, i64 [[TMP17]]
-; CHECK-NEXT:    [[TMP19:%.*]] = load i64, ptr [[TMP18]], align 4
-; CHECK-NEXT:    [[TMP20:%.*]] = insertelement <4 x i64> [[TMP15]], i64 [[TMP19]], i32 2
+; CHECK-NEXT:    [[TMP16:%.*]] = add i64 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [[STRUCT_FOO]], ptr @foo, i64 0, i32 1, i64 [[TMP16]]
+; CHECK-NEXT:    [[TMP18:%.*]] = load i64, ptr [[TMP17]], align 4
+; CHECK-NEXT:    [[TMP19:%.*]] = insertelement <4 x i64> [[TMP14]], i64 [[TMP18]], i32 2
 ; CHECK-NEXT:    br label [[PRED_LOAD_CONTINUE4]]
 ; CHECK:       pred.load.continue4:
-; CHECK-NEXT:    [[TMP21:%.*]] = phi <4 x i64> [ [[TMP15]], [[PRED_LOAD_CONTINUE2]] ], [ [[TMP20]], [[PRED_LOAD_IF3]] ]
-; CHECK-NEXT:    [[TMP22:%.*]] = extractelement <4 x i1> [[TMP3]], i32 3
-; CHECK-NEXT:    br i1 [[TMP22]], label [[PRED_LOAD_IF5:%.*]], label [[PRED_LOAD_CONTINUE6]]
+; CHECK-NEXT:    [[TMP20:%.*]] = phi <4 x i64> [ [[TMP14]], [[PRED_LOAD_CONTINUE2]] ], [ [[TMP19]], [[PRED_LOAD_IF3]] ]
+; CHECK-NEXT:    [[TMP21:%.*]] = extractelement <4 x i1> [[TMP2]], i32 3
+; CHECK-NEXT:    br i1 [[TMP21]], label [[PRED_LOAD_IF5:%.*]], label [[PRED_LOAD_CONTINUE6]]
 ; CHECK:       pred.load.if5:
-; CHECK-NEXT:    [[TMP23:%.*]] = add i64 [[INDEX]], 3
-; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds [[STRUCT_FOO]], ptr @foo, i64 0, i32 1, i64 [[TMP23]]
-; CHECK-NEXT:    [[TMP25:%.*]] = load i64, ptr [[TMP24]], align 4
-; CHECK-NEXT:    [[TMP26:%.*]] = insertelement <4 x i64> [[TMP21]], i64 [[TMP25]], i32 3
+; CHECK-NEXT:    [[TMP22:%.*]] = add i64 [[INDEX]], 3
+; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds [[STRUCT_FOO]], ptr @foo, i64 0, i32 1, i64 [[TMP22]]
+; CHECK-NEXT:    [[TMP24:%.*]] = load i64, ptr [[TMP23]], align 4
+; CHECK-NEXT:    [[TMP25:%.*]] = insertelement <4 x i64> [[TMP20]], i64 [[TMP24]], i32 3
 ; CHECK-NEXT:    br label [[PRED_LOAD_CONTINUE6]]
 ; CHECK:       pred.load.continue6:
-; CHECK-NEXT:    [[TMP27:%.*]] = phi <4 x i64> [ [[TMP21]], [[PRED_LOAD_CONTINUE4]] ], [ [[TMP26]], [[PRED_LOAD_IF5]] ]
-; CHECK-NEXT:    [[TMP28:%.*]] = trunc <4 x i64> [[TMP27]] to <4 x i32>
-; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr [[STRUCT_FOO]], ptr @foo, i64 0, i32 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr i32, ptr [[TMP29]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i32>, ptr [[TMP30]], align 4
-; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP2]], <4 x i32> [[WIDE_LOAD7]], <4 x i32> [[TMP28]]
-; CHECK-NEXT:    [[TMP31:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[PREDPHI]], ptr [[TMP31]], align 4
+; CHECK-NEXT:    [[TMP26:%.*]] = phi <4 x i64> [ [[TMP20]], [[PRED_LOAD_CONTINUE4]] ], [ [[TMP25]], [[PRED_LOAD_IF5]] ]
+; CHECK-NEXT:    [[TMP27:%.*]] = trunc <4 x i64> [[TMP26]] to <4 x i32>
+; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr [[STRUCT_FOO]], ptr @foo, i64 0, i32 0, i64 [[INDEX]]
+; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i32>, ptr [[TMP28]], align 4
+; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> [[TMP1]], <4 x i32> [[WIDE_LOAD7]], <4 x i32> [[TMP27]]
+; CHECK-NEXT:    store <4 x i32> [[PREDPHI]], ptr [[TMP0]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32000
-; CHECK-NEXT:    br i1 [[TMP32]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK-NEXT:    [[TMP29:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32000
+; CHECK-NEXT:    br i1 [[TMP29]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 32000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; CHECK:       loop.header:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/loop-form.ll b/llvm/test/Transforms/LoopVectorize/loop-form.ll
index 64b9b47..10b2e70 100644
--- a/llvm/test/Transforms/LoopVectorize/loop-form.ll
+++ b/llvm/test/Transforms/LoopVectorize/loop-form.ll
@@ -19,8 +19,7 @@ define void @bottom_tested(ptr %p, i32 %n) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP2:%.*]] = sext i32 [[INDEX]] to i64
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, ptr [[P:%.*]], i64 [[TMP2]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP4]], align 4
+; CHECK-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -128,8 +127,7 @@ define void @early_exit(ptr %p, i32 %n) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = sext i32 [[INDEX]] to i64
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[P:%.*]], i64 [[TMP4]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -195,8 +193,7 @@ define i32 @early_exit_with_live_out(ptr %ptr) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <2 x i32> splat (i32 10), ptr [[TMP2]], align 4
+; CHECK-NEXT:    store <2 x i32> splat (i32 10), ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 998
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
@@ -330,8 +327,7 @@ define void @multiple_unique_exit(ptr %p, i32 %n) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = sext i32 [[INDEX]] to i64
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[P:%.*]], i64 [[TMP4]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
@@ -410,8 +406,7 @@ define i32 @multiple_unique_exit2(ptr %p, i32 %n) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = sext i32 [[INDEX]] to i64
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[P:%.*]], i64 [[TMP4]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
@@ -492,8 +487,7 @@ define i32 @multiple_unique_exit3(ptr %p, i32 %n) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = sext i32 [[INDEX]] to i64
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[P:%.*]], i64 [[TMP4]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
@@ -575,8 +569,7 @@ define i32 @multiple_exit_blocks(ptr %p, i32 %n) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = sext i32 [[INDEX]] to i64
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[P:%.*]], i64 [[TMP4]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
@@ -662,8 +655,7 @@ define i32 @multiple_exit_blocks2(ptr %p, i32 %n) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = sext i32 [[INDEX]] to i64
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[P:%.*]], i64 [[TMP4]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
@@ -753,8 +745,7 @@ define i32 @multiple_exit_blocks3(ptr %p, i32 %n) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = sext i32 [[INDEX]] to i64
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[P:%.*]], i64 [[TMP4]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
@@ -1077,8 +1068,7 @@ define void @scalar_predication(ptr %addr) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[PRED_STORE_CONTINUE2:%.*]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr float, ptr [[ADDR:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = fcmp une <2 x float> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <2 x i1> [[TMP4]], i32 0
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
@@ -1179,8 +1169,7 @@ define i32 @me_reduction(ptr %addr) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i32, ptr [[ADDR:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3]] = add <2 x i32> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 200
diff --git a/llvm/test/Transforms/LoopVectorize/make-followup-loop-id.ll b/llvm/test/Transforms/LoopVectorize/make-followup-loop-id.ll
index 1d633dd..14a091f 100644
--- a/llvm/test/Transforms/LoopVectorize/make-followup-loop-id.ll
+++ b/llvm/test/Transforms/LoopVectorize/make-followup-loop-id.ll
@@ -66,9 +66,9 @@ define void @f(ptr noundef captures(none) %a, float noundef %x) {
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT_7:%.*]], %[[FOR_BODY]] ]
@@ -112,8 +112,7 @@ define void @f(ptr noundef captures(none) %a, float noundef %x) {
 ; CHECK-NEXT:    [[MUL_7:%.*]] = fmul float [[X]], [[LOAD_7]]
 ; CHECK-NEXT:    store float [[MUL_7]], ptr [[ARRAYIDX_7]], align 4
 ; CHECK-NEXT:    [[IV_NEXT_7]] = add nuw nsw i64 [[IV]], 8
-; CHECK-NEXT:    [[COMP_7:%.*]] = icmp eq i64 [[IV_NEXT_7]], 1024
-; CHECK-NEXT:    br i1 [[COMP_7]], label %[[EXIT_LOOPEXIT:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 true, label %[[EXIT_LOOPEXIT:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       [[EXIT_LOOPEXIT]]:
 ; CHECK-NEXT:    br label %[[EXIT]]
 ; CHECK:       [[EXIT]]:
diff --git a/llvm/test/Transforms/LoopVectorize/metadata.ll b/llvm/test/Transforms/LoopVectorize/metadata.ll
index 287a15e..ce9c624 100644
--- a/llvm/test/Transforms/LoopVectorize/metadata.ll
+++ b/llvm/test/Transforms/LoopVectorize/metadata.ll
@@ -18,14 +18,12 @@ define void @fp_math(ptr nocapture %a, ptr noalias %b, i64 %size) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[INDEX]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds double, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP2]], align 4, !tbaa [[TBAA0:![0-9]+]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP0]], align 4, !tbaa [[TBAA0:![0-9]+]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = fadd <2 x double> [[WIDE_LOAD]], splat (double 9.900000e+01), !fpmath [[META3:![0-9]+]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = fcmp oge <2 x double> [[TMP3]], splat (double 1.000000e+01)
-; CHECK-NEXT:    [[TMP5:%.*]] = select <2 x i1> [[TMP4]], <2 x double> [[WIDE_LOAD]], <2 x double> zeroinitializer, !fpmath [[META3]]
-; CHECK-NEXT:    [[TMP6:%.*]] = fptrunc <2 x double> [[TMP5]] to <2 x float>, !fpmath [[META3]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <2 x float> [[TMP6]], ptr [[TMP7]], align 4, !tbaa [[TBAA0]]
+; CHECK-NEXT:    [[TMP6:%.*]] = select <2 x i1> [[TMP4]], <2 x double> [[WIDE_LOAD]], <2 x double> zeroinitializer, !fpmath [[META3]]
+; CHECK-NEXT:    [[TMP5:%.*]] = fptrunc <2 x double> [[TMP6]] to <2 x float>, !fpmath [[META3]]
+; CHECK-NEXT:    store <2 x float> [[TMP5]], ptr [[TMP1]], align 4, !tbaa [[TBAA0]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -64,22 +62,20 @@ define void @fp_math(ptr nocapture %a, ptr noalias %b, i64 %size) {
 ; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; INTERLEAVE-NEXT:    [[TMP0:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[INDEX]]
 ; INTERLEAVE-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[B]], i64 [[INDEX]]
-; INTERLEAVE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds double, ptr [[TMP0]], i32 0
 ; INTERLEAVE-NEXT:    [[TMP3:%.*]] = getelementptr inbounds double, ptr [[TMP0]], i32 2
-; INTERLEAVE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP2]], align 4, !tbaa [[TBAA0:![0-9]+]]
+; INTERLEAVE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP0]], align 4, !tbaa [[TBAA0:![0-9]+]]
 ; INTERLEAVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x double>, ptr [[TMP3]], align 4, !tbaa [[TBAA0]]
 ; INTERLEAVE-NEXT:    [[TMP4:%.*]] = fadd <2 x double> [[WIDE_LOAD]], splat (double 9.900000e+01), !fpmath [[META3:![0-9]+]]
 ; INTERLEAVE-NEXT:    [[TMP5:%.*]] = fadd <2 x double> [[WIDE_LOAD1]], splat (double 9.900000e+01), !fpmath [[META3]]
 ; INTERLEAVE-NEXT:    [[TMP6:%.*]] = fcmp oge <2 x double> [[TMP4]], splat (double 1.000000e+01)
 ; INTERLEAVE-NEXT:    [[TMP7:%.*]] = fcmp oge <2 x double> [[TMP5]], splat (double 1.000000e+01)
-; INTERLEAVE-NEXT:    [[TMP8:%.*]] = select <2 x i1> [[TMP6]], <2 x double> [[WIDE_LOAD]], <2 x double> zeroinitializer, !fpmath [[META3]]
-; INTERLEAVE-NEXT:    [[TMP9:%.*]] = select <2 x i1> [[TMP7]], <2 x double> [[WIDE_LOAD1]], <2 x double> zeroinitializer, !fpmath [[META3]]
+; INTERLEAVE-NEXT:    [[TMP11:%.*]] = select <2 x i1> [[TMP6]], <2 x double> [[WIDE_LOAD]], <2 x double> zeroinitializer, !fpmath [[META3]]
+; INTERLEAVE-NEXT:    [[TMP8:%.*]] = select <2 x i1> [[TMP7]], <2 x double> [[WIDE_LOAD1]], <2 x double> zeroinitializer, !fpmath [[META3]]
+; INTERLEAVE-NEXT:    [[TMP9:%.*]] = fptrunc <2 x double> [[TMP11]] to <2 x float>, !fpmath [[META3]]
 ; INTERLEAVE-NEXT:    [[TMP10:%.*]] = fptrunc <2 x double> [[TMP8]] to <2 x float>, !fpmath [[META3]]
-; INTERLEAVE-NEXT:    [[TMP11:%.*]] = fptrunc <2 x double> [[TMP9]] to <2 x float>, !fpmath [[META3]]
-; INTERLEAVE-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
 ; INTERLEAVE-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 2
-; INTERLEAVE-NEXT:    store <2 x float> [[TMP10]], ptr [[TMP12]], align 4, !tbaa [[TBAA0]]
-; INTERLEAVE-NEXT:    store <2 x float> [[TMP11]], ptr [[TMP13]], align 4, !tbaa [[TBAA0]]
+; INTERLEAVE-NEXT:    store <2 x float> [[TMP9]], ptr [[TMP1]], align 4, !tbaa [[TBAA0]]
+; INTERLEAVE-NEXT:    store <2 x float> [[TMP10]], ptr [[TMP13]], align 4, !tbaa [[TBAA0]]
 ; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; INTERLEAVE-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; INTERLEAVE-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -137,19 +133,17 @@ define void @widen_call_range(ptr noalias %a, ptr readonly %b) {
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i64, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i64, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 4, !tbaa [[TBAA0]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP0]], align 4, !tbaa [[TBAA0]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = call <2 x i64> @foo_vector_fixed2_nomask(<2 x i64> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    store <2 x i64> [[TMP2]], ptr [[TMP4]], align 4
+; CHECK-NEXT:    store <2 x i64> [[TMP2]], ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -173,24 +167,22 @@ define void @widen_call_range(ptr noalias %a, ptr readonly %b) {
 ; INTERLEAVE:       [[VECTOR_BODY]]:
 ; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; INTERLEAVE-NEXT:    [[TMP0:%.*]] = getelementptr i64, ptr [[B]], i64 [[INDEX]]
-; INTERLEAVE-NEXT:    [[TMP1:%.*]] = getelementptr i64, ptr [[TMP0]], i32 0
 ; INTERLEAVE-NEXT:    [[TMP2:%.*]] = getelementptr i64, ptr [[TMP0]], i32 2
-; INTERLEAVE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 4, !tbaa [[TBAA0]]
+; INTERLEAVE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP0]], align 4, !tbaa [[TBAA0]]
 ; INTERLEAVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i64>, ptr [[TMP2]], align 4, !tbaa [[TBAA0]]
 ; INTERLEAVE-NEXT:    [[TMP3:%.*]] = call <2 x i64> @foo_vector_fixed2_nomask(<2 x i64> [[WIDE_LOAD]])
 ; INTERLEAVE-NEXT:    [[TMP4:%.*]] = call <2 x i64> @foo_vector_fixed2_nomask(<2 x i64> [[WIDE_LOAD1]])
 ; INTERLEAVE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; INTERLEAVE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 0
 ; INTERLEAVE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 2
-; INTERLEAVE-NEXT:    store <2 x i64> [[TMP3]], ptr [[TMP6]], align 4
+; INTERLEAVE-NEXT:    store <2 x i64> [[TMP3]], ptr [[TMP5]], align 4
 ; INTERLEAVE-NEXT:    store <2 x i64> [[TMP4]], ptr [[TMP7]], align 4
 ; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; INTERLEAVE-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; INTERLEAVE-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; INTERLEAVE:       [[MIDDLE_BLOCK]]:
-; INTERLEAVE-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; INTERLEAVE-NEXT:    br label %[[EXIT:.*]]
 ; INTERLEAVE:       [[SCALAR_PH]]:
-; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; INTERLEAVE-NEXT:    br label %[[LOOP:.*]]
 ; INTERLEAVE:       [[LOOP]]:
 ; INTERLEAVE-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -233,19 +225,17 @@ define void @widen_call_fpmath(ptr noalias %a, ptr readonly %b) {
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr double, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr double, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP1]], align 8, !tbaa [[TBAA0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = call <2 x double> @bar_vector_fixed2_nomask(<2 x double> [[WIDE_LOAD]]), !fpmath [[META3]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP0]], align 8, !tbaa [[TBAA0]]
+; CHECK-NEXT:    [[TMP1:%.*]] = call <2 x double> @bar_vector_fixed2_nomask(<2 x double> [[WIDE_LOAD]]), !fpmath [[META3]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds double, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    store <2 x double> [[TMP2]], ptr [[TMP4]], align 8
+; CHECK-NEXT:    store <2 x double> [[TMP1]], ptr [[TMP3]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -269,24 +259,22 @@ define void @widen_call_fpmath(ptr noalias %a, ptr readonly %b) {
 ; INTERLEAVE:       [[VECTOR_BODY]]:
 ; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; INTERLEAVE-NEXT:    [[TMP0:%.*]] = getelementptr double, ptr [[B]], i64 [[INDEX]]
-; INTERLEAVE-NEXT:    [[TMP1:%.*]] = getelementptr double, ptr [[TMP0]], i32 0
 ; INTERLEAVE-NEXT:    [[TMP2:%.*]] = getelementptr double, ptr [[TMP0]], i32 2
-; INTERLEAVE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP1]], align 8, !tbaa [[TBAA0]]
+; INTERLEAVE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP0]], align 8, !tbaa [[TBAA0]]
 ; INTERLEAVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x double>, ptr [[TMP2]], align 8, !tbaa [[TBAA0]]
 ; INTERLEAVE-NEXT:    [[TMP3:%.*]] = call <2 x double> @bar_vector_fixed2_nomask(<2 x double> [[WIDE_LOAD]]), !fpmath [[META3]]
 ; INTERLEAVE-NEXT:    [[TMP4:%.*]] = call <2 x double> @bar_vector_fixed2_nomask(<2 x double> [[WIDE_LOAD1]]), !fpmath [[META3]]
 ; INTERLEAVE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[INDEX]]
-; INTERLEAVE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds double, ptr [[TMP5]], i32 0
 ; INTERLEAVE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds double, ptr [[TMP5]], i32 2
-; INTERLEAVE-NEXT:    store <2 x double> [[TMP3]], ptr [[TMP6]], align 8
+; INTERLEAVE-NEXT:    store <2 x double> [[TMP3]], ptr [[TMP5]], align 8
 ; INTERLEAVE-NEXT:    store <2 x double> [[TMP4]], ptr [[TMP7]], align 8
 ; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; INTERLEAVE-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; INTERLEAVE-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
 ; INTERLEAVE:       [[MIDDLE_BLOCK]]:
-; INTERLEAVE-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; INTERLEAVE-NEXT:    br label %[[EXIT:.*]]
 ; INTERLEAVE:       [[SCALAR_PH]]:
-; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; INTERLEAVE-NEXT:    br label %[[LOOP:.*]]
 ; INTERLEAVE:       [[LOOP]]:
 ; INTERLEAVE-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -329,19 +317,17 @@ define void @widen_intrinsic(ptr noalias %a, ptr readonly %b) {
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i64, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i64, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP0]], align 4
 ; CHECK-NEXT:    [[TMP2:%.*]] = call <2 x i64> @llvm.abs.v2i64(<2 x i64> [[WIDE_LOAD]], i1 true)
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    store <2 x i64> [[TMP2]], ptr [[TMP4]], align 4
+; CHECK-NEXT:    store <2 x i64> [[TMP2]], ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -365,24 +351,22 @@ define void @widen_intrinsic(ptr noalias %a, ptr readonly %b) {
 ; INTERLEAVE:       [[VECTOR_BODY]]:
 ; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; INTERLEAVE-NEXT:    [[TMP0:%.*]] = getelementptr i64, ptr [[B]], i64 [[INDEX]]
-; INTERLEAVE-NEXT:    [[TMP1:%.*]] = getelementptr i64, ptr [[TMP0]], i32 0
 ; INTERLEAVE-NEXT:    [[TMP2:%.*]] = getelementptr i64, ptr [[TMP0]], i32 2
-; INTERLEAVE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 4
+; INTERLEAVE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP0]], align 4
 ; INTERLEAVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i64>, ptr [[TMP2]], align 4
 ; INTERLEAVE-NEXT:    [[TMP3:%.*]] = call <2 x i64> @llvm.abs.v2i64(<2 x i64> [[WIDE_LOAD]], i1 true)
 ; INTERLEAVE-NEXT:    [[TMP4:%.*]] = call <2 x i64> @llvm.abs.v2i64(<2 x i64> [[WIDE_LOAD1]], i1 true)
 ; INTERLEAVE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[INDEX]]
-; INTERLEAVE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 0
 ; INTERLEAVE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 2
-; INTERLEAVE-NEXT:    store <2 x i64> [[TMP3]], ptr [[TMP6]], align 4
+; INTERLEAVE-NEXT:    store <2 x i64> [[TMP3]], ptr [[TMP5]], align 4
 ; INTERLEAVE-NEXT:    store <2 x i64> [[TMP4]], ptr [[TMP7]], align 4
 ; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; INTERLEAVE-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; INTERLEAVE-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP13:![0-9]+]]
 ; INTERLEAVE:       [[MIDDLE_BLOCK]]:
-; INTERLEAVE-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; INTERLEAVE-NEXT:    br label %[[EXIT:.*]]
 ; INTERLEAVE:       [[SCALAR_PH]]:
-; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; INTERLEAVE-NEXT:    br label %[[LOOP:.*]]
 ; INTERLEAVE:       [[LOOP]]:
 ; INTERLEAVE-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -425,19 +409,17 @@ define void @widen_intrinsic_fpmath(ptr noalias %a, ptr readonly %b) {
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr double, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr double, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP1]], align 8, !tbaa [[TBAA0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = call <2 x double> @llvm.sin.v2f64(<2 x double> [[WIDE_LOAD]]), !fpmath [[META3]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP0]], align 8, !tbaa [[TBAA0]]
+; CHECK-NEXT:    [[TMP1:%.*]] = call <2 x double> @llvm.sin.v2f64(<2 x double> [[WIDE_LOAD]]), !fpmath [[META3]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds double, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    store <2 x double> [[TMP2]], ptr [[TMP4]], align 8
+; CHECK-NEXT:    store <2 x double> [[TMP1]], ptr [[TMP3]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -461,24 +443,22 @@ define void @widen_intrinsic_fpmath(ptr noalias %a, ptr readonly %b) {
 ; INTERLEAVE:       [[VECTOR_BODY]]:
 ; INTERLEAVE-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; INTERLEAVE-NEXT:    [[TMP0:%.*]] = getelementptr double, ptr [[B]], i64 [[INDEX]]
-; INTERLEAVE-NEXT:    [[TMP1:%.*]] = getelementptr double, ptr [[TMP0]], i32 0
 ; INTERLEAVE-NEXT:    [[TMP2:%.*]] = getelementptr double, ptr [[TMP0]], i32 2
-; INTERLEAVE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP1]], align 8, !tbaa [[TBAA0]]
+; INTERLEAVE-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP0]], align 8, !tbaa [[TBAA0]]
 ; INTERLEAVE-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x double>, ptr [[TMP2]], align 8, !tbaa [[TBAA0]]
 ; INTERLEAVE-NEXT:    [[TMP3:%.*]] = call <2 x double> @llvm.sin.v2f64(<2 x double> [[WIDE_LOAD]]), !fpmath [[META3]]
 ; INTERLEAVE-NEXT:    [[TMP4:%.*]] = call <2 x double> @llvm.sin.v2f64(<2 x double> [[WIDE_LOAD1]]), !fpmath [[META3]]
 ; INTERLEAVE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[INDEX]]
-; INTERLEAVE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds double, ptr [[TMP5]], i32 0
 ; INTERLEAVE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds double, ptr [[TMP5]], i32 2
-; INTERLEAVE-NEXT:    store <2 x double> [[TMP3]], ptr [[TMP6]], align 8
+; INTERLEAVE-NEXT:    store <2 x double> [[TMP3]], ptr [[TMP5]], align 8
 ; INTERLEAVE-NEXT:    store <2 x double> [[TMP4]], ptr [[TMP7]], align 8
 ; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; INTERLEAVE-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; INTERLEAVE-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; INTERLEAVE:       [[MIDDLE_BLOCK]]:
-; INTERLEAVE-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; INTERLEAVE-NEXT:    br label %[[EXIT:.*]]
 ; INTERLEAVE:       [[SCALAR_PH]]:
-; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; INTERLEAVE-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; INTERLEAVE-NEXT:    br label %[[LOOP:.*]]
 ; INTERLEAVE:       [[LOOP]]:
 ; INTERLEAVE-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -525,12 +505,10 @@ define void @unknown_metadata(ptr nocapture %a, ptr noalias %b, i64 %size) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP3:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT2:%.*]], %[[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[INDEX]], !custom_md [[META2:![0-9]+]]
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds ptr, ptr [[A]], i64 [[INDEX]], !custom_md [[META2:![0-9]+]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], <2 x i64> [[VEC_IND]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = extractelement <2 x ptr> [[TMP2]], i32 0
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[TMP3]], ptr [[TMP4]], align 4
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds ptr, ptr [[TMP0]], i32 0
+; CHECK-NEXT:    store <2 x i32> [[TMP3]], ptr [[TMP1]], align 4
 ; CHECK-NEXT:    store <2 x ptr> [[TMP2]], ptr [[TMP6]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
@@ -575,13 +553,11 @@ define void @unknown_metadata(ptr nocapture %a, ptr noalias %b, i64 %size) {
 ; INTERLEAVE-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B]], <2 x i64> [[VEC_IND]]
 ; INTERLEAVE-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[B]], <2 x i64> [[STEP_ADD]]
 ; INTERLEAVE-NEXT:    [[TMP3:%.*]] = extractelement <2 x ptr> [[TMP1]], i32 0
-; INTERLEAVE-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
 ; INTERLEAVE-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 2
-; INTERLEAVE-NEXT:    store <2 x i32> [[VEC_IND1]], ptr [[TMP4]], align 4
+; INTERLEAVE-NEXT:    store <2 x i32> [[VEC_IND1]], ptr [[TMP3]], align 4
 ; INTERLEAVE-NEXT:    store <2 x i32> [[STEP_ADD3]], ptr [[TMP5]], align 4
-; INTERLEAVE-NEXT:    [[TMP6:%.*]] = getelementptr inbounds ptr, ptr [[TMP0]], i32 0
 ; INTERLEAVE-NEXT:    [[TMP7:%.*]] = getelementptr inbounds ptr, ptr [[TMP0]], i32 2
-; INTERLEAVE-NEXT:    store <2 x ptr> [[TMP1]], ptr [[TMP6]], align 8
+; INTERLEAVE-NEXT:    store <2 x ptr> [[TMP1]], ptr [[TMP0]], align 8
 ; INTERLEAVE-NEXT:    store <2 x ptr> [[TMP2]], ptr [[TMP7]], align 8
 ; INTERLEAVE-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; INTERLEAVE-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[STEP_ADD]], splat (i64 2)
diff --git a/llvm/test/Transforms/LoopVectorize/min-trip-count-known-via-scev.ll b/llvm/test/Transforms/LoopVectorize/min-trip-count-known-via-scev.ll
index 6ced1f1..445ef03 100644
--- a/llvm/test/Transforms/LoopVectorize/min-trip-count-known-via-scev.ll
+++ b/llvm/test/Transforms/LoopVectorize/min-trip-count-known-via-scev.ll
@@ -20,8 +20,7 @@ define i32 @loop_with_at_least_2_iterations_via_guards_order_1(ptr %dst, i32 %n)
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <2 x i32> splat (i32 1), ptr [[TMP2]], align 4
+; CHECK-NEXT:    store <2 x i32> splat (i32 1), ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -88,8 +87,7 @@ define i32 @loop_with_at_least_2_iterations_via_guards_order_2(ptr %dst, i32 %n)
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <2 x i32> splat (i32 1), ptr [[TMP2]], align 4
+; CHECK-NEXT:    store <2 x i32> splat (i32 1), ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -162,8 +160,7 @@ define void @loop_never_executes_precondition_order_1(i64 %start, ptr %dst) {
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ [[INDUCTION]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 [[START]], [[INDEX]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[DST]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    store <2 x i64> [[VEC_IND]], ptr [[TMP3]], align 4
+; CHECK-NEXT:    store <2 x i64> [[VEC_IND]], ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -236,8 +233,7 @@ define void @loop_never_executes_precondition_order_1_predicates_flipped(i64 %st
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ [[INDUCTION]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 [[START]], [[INDEX]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[DST]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    store <2 x i64> [[VEC_IND]], ptr [[TMP3]], align 4
+; CHECK-NEXT:    store <2 x i64> [[VEC_IND]], ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -307,8 +303,7 @@ define void @loop_never_executes_precondition_order_2_predicates_flipped(i64 %st
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ [[INDUCTION]], %[[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 [[START]], [[INDEX]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[DST]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    store <2 x i64> [[VEC_IND]], ptr [[TMP3]], align 4
+; CHECK-NEXT:    store <2 x i64> [[VEC_IND]], ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/minimumnum-maximumnum-reductions.ll b/llvm/test/Transforms/LoopVectorize/minimumnum-maximumnum-reductions.ll
index 6dde2b9..e26fef4 100644
--- a/llvm/test/Transforms/LoopVectorize/minimumnum-maximumnum-reductions.ll
+++ b/llvm/test/Transforms/LoopVectorize/minimumnum-maximumnum-reductions.ll
@@ -14,9 +14,8 @@ define float @maximumnum_intrinsic(ptr readonly %x) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x float> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP3:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x float> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[X]], i32 [[IV]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[GEP]], i32 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[GEP]], i32 2
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[GEP]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP3]] = call <2 x float> @llvm.maximumnum.v2f32(<2 x float> [[VEC_PHI]], <2 x float> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[TMP4]] = call <2 x float> @llvm.maximumnum.v2f32(<2 x float> [[VEC_PHI1]], <2 x float> [[WIDE_LOAD2]])
@@ -26,10 +25,10 @@ define float @maximumnum_intrinsic(ptr readonly %x) {
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[RDX_MINMAX:%.*]] = call <2 x float> @llvm.maximumnum.v2f32(<2 x float> [[TMP3]], <2 x float> [[TMP4]])
 ; CHECK-NEXT:    [[TMP6:%.*]] = call float @llvm.vector.reduce.fmax.v2f32(<2 x float> [[RDX_MINMAX]])
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP6]], %[[MIDDLE_BLOCK]] ], [ 0.000000e+00, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV1:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[LOOP]] ]
@@ -73,9 +72,8 @@ define float @maximumnum_intrinsic_fast(ptr readonly %x) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x float> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP3:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x float> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[X]], i32 [[IV]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[GEP]], i32 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[GEP]], i32 2
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[GEP]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP3]] = call fast <2 x float> @llvm.maximumnum.v2f32(<2 x float> [[VEC_PHI]], <2 x float> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[TMP4]] = call fast <2 x float> @llvm.maximumnum.v2f32(<2 x float> [[VEC_PHI1]], <2 x float> [[WIDE_LOAD2]])
@@ -85,10 +83,10 @@ define float @maximumnum_intrinsic_fast(ptr readonly %x) {
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[RDX_MINMAX:%.*]] = call fast <2 x float> @llvm.maximumnum.v2f32(<2 x float> [[TMP3]], <2 x float> [[TMP4]])
 ; CHECK-NEXT:    [[TMP6:%.*]] = call fast float @llvm.vector.reduce.fmax.v2f32(<2 x float> [[RDX_MINMAX]])
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP6]], %[[MIDDLE_BLOCK]] ], [ 0.000000e+00, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV1:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[LOOP]] ]
@@ -132,9 +130,8 @@ define float @minimumnum_intrinsic(ptr readonly %x) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x float> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP3:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x float> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[X]], i32 [[IV]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[GEP]], i32 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[GEP]], i32 2
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[GEP]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP3]] = call <2 x float> @llvm.minimumnum.v2f32(<2 x float> [[VEC_PHI]], <2 x float> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[TMP4]] = call <2 x float> @llvm.minimumnum.v2f32(<2 x float> [[VEC_PHI1]], <2 x float> [[WIDE_LOAD2]])
@@ -144,10 +141,10 @@ define float @minimumnum_intrinsic(ptr readonly %x) {
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[RDX_MINMAX:%.*]] = call <2 x float> @llvm.minimumnum.v2f32(<2 x float> [[TMP3]], <2 x float> [[TMP4]])
 ; CHECK-NEXT:    [[TMP6:%.*]] = call float @llvm.vector.reduce.fmin.v2f32(<2 x float> [[RDX_MINMAX]])
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP6]], %[[MIDDLE_BLOCK]] ], [ 0.000000e+00, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV1:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[LOOP]] ]
@@ -191,9 +188,8 @@ define float @minimumnum_intrinsic_fast(ptr readonly %x) {
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x float> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP3:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI1:%.*]] = phi <2 x float> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds float, ptr [[X]], i32 [[IV]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[GEP]], i32 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[GEP]], i32 2
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[GEP]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x float>, ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[TMP3]] = call fast <2 x float> @llvm.minimumnum.v2f32(<2 x float> [[VEC_PHI]], <2 x float> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[TMP4]] = call fast <2 x float> @llvm.minimumnum.v2f32(<2 x float> [[VEC_PHI1]], <2 x float> [[WIDE_LOAD2]])
@@ -203,10 +199,10 @@ define float @minimumnum_intrinsic_fast(ptr readonly %x) {
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[RDX_MINMAX:%.*]] = call fast <2 x float> @llvm.minimumnum.v2f32(<2 x float> [[TMP3]], <2 x float> [[TMP4]])
 ; CHECK-NEXT:    [[TMP6:%.*]] = call fast float @llvm.vector.reduce.fmin.v2f32(<2 x float> [[RDX_MINMAX]])
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ [[TMP6]], %[[MIDDLE_BLOCK]] ], [ 0.000000e+00, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi float [ 0.000000e+00, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV1:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INC:%.*]], %[[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/multiple-address-spaces.ll b/llvm/test/Transforms/LoopVectorize/multiple-address-spaces.ll
index ca0edb3..cc246d55 100644
--- a/llvm/test/Transforms/LoopVectorize/multiple-address-spaces.ll
+++ b/llvm/test/Transforms/LoopVectorize/multiple-address-spaces.ll
@@ -33,7 +33,7 @@ define i32 @main() #0 {
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 40000
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
diff --git a/llvm/test/Transforms/LoopVectorize/multiple-strides-vectorization.ll b/llvm/test/Transforms/LoopVectorize/multiple-strides-vectorization.ll
index 3af6ecd..d21621e 100644
--- a/llvm/test/Transforms/LoopVectorize/multiple-strides-vectorization.ll
+++ b/llvm/test/Transforms/LoopVectorize/multiple-strides-vectorization.ll
@@ -65,17 +65,15 @@ define void @Test(ptr nocapture %obj, i64 %z) #0 {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[OBJ]], i64 0, i32 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP12]], align 4, !alias.scope [[META0:![0-9]+]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP11]], align 4, !alias.scope [[META0:![0-9]+]]
 ; CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[TMP9]], align 4, !alias.scope [[META3:![0-9]+]]
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TMP13]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP14:%.*]] = add nsw <4 x i32> [[BROADCAST_SPLAT]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[OBJ]], i64 0, i32 2, i64 [[I]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[TMP15]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x i32>, ptr [[TMP16]], align 4, !alias.scope [[META5:![0-9]+]], !noalias [[META7:![0-9]+]]
+; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x i32>, ptr [[TMP15]], align 4, !alias.scope [[META5:![0-9]+]], !noalias [[META7:![0-9]+]]
 ; CHECK-NEXT:    [[TMP17:%.*]] = add nsw <4 x i32> [[TMP14]], [[WIDE_LOAD8]]
-; CHECK-NEXT:    store <4 x i32> [[TMP17]], ptr [[TMP16]], align 4, !alias.scope [[META5]], !noalias [[META7]]
+; CHECK-NEXT:    store <4 x i32> [[TMP17]], ptr [[TMP15]], align 4, !alias.scope [[META5]], !noalias [[META7]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
@@ -131,17 +129,15 @@ define void @Test(ptr nocapture %obj, i64 %z) #0 {
 ; CHECK-HOIST:       vector.body:
 ; CHECK-HOIST-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-HOIST-NEXT:    [[TMP5:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[OBJ]], i64 0, i32 0, i64 [[INDEX]]
-; CHECK-HOIST-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-HOIST-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4, !alias.scope [[META0:![0-9]+]]
+; CHECK-HOIST-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4, !alias.scope [[META0:![0-9]+]]
 ; CHECK-HOIST-NEXT:    [[TMP7:%.*]] = load i32, ptr [[TMP3]], align 4, !alias.scope [[META3:![0-9]+]]
 ; CHECK-HOIST-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TMP7]], i64 0
 ; CHECK-HOIST-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-HOIST-NEXT:    [[TMP8:%.*]] = add nsw <4 x i32> [[BROADCAST_SPLAT]], [[WIDE_LOAD]]
 ; CHECK-HOIST-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [[STRUCT_S]], ptr [[OBJ]], i64 0, i32 2, i64 [[I]], i64 [[INDEX]]
-; CHECK-HOIST-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i32 0
-; CHECK-HOIST-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i32>, ptr [[TMP10]], align 4, !alias.scope [[META5:![0-9]+]], !noalias [[META7:![0-9]+]]
+; CHECK-HOIST-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i32>, ptr [[TMP9]], align 4, !alias.scope [[META5:![0-9]+]], !noalias [[META7:![0-9]+]]
 ; CHECK-HOIST-NEXT:    [[TMP11:%.*]] = add nsw <4 x i32> [[TMP8]], [[WIDE_LOAD5]]
-; CHECK-HOIST-NEXT:    store <4 x i32> [[TMP11]], ptr [[TMP10]], align 4, !alias.scope [[META5]], !noalias [[META7]]
+; CHECK-HOIST-NEXT:    store <4 x i32> [[TMP11]], ptr [[TMP9]], align 4, !alias.scope [[META5]], !noalias [[META7]]
 ; CHECK-HOIST-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-HOIST-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-HOIST-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/narrow-to-single-scalar.ll b/llvm/test/Transforms/LoopVectorize/narrow-to-single-scalar.ll
index 6a17b5f..11c8991 100644
--- a/llvm/test/Transforms/LoopVectorize/narrow-to-single-scalar.ll
+++ b/llvm/test/Transforms/LoopVectorize/narrow-to-single-scalar.ll
@@ -25,7 +25,7 @@ define void @narrow_select_to_single_scalar(i1 %invar.cond, ptr noalias %A, ptr
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
@@ -39,7 +39,7 @@ define void @narrow_select_to_single_scalar(i1 %invar.cond, ptr noalias %A, ptr
 ; CHECK-NEXT:    store i16 0, ptr [[GEP_C]], align 1
 ; CHECK-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
 ; CHECK-NEXT:    [[EC:%.*]] = icmp ne i16 [[IV]], 1024
-; CHECK-NEXT:    br i1 [[EC]], label %[[LOOP_HEADER]], label %[[EXIT]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label %[[LOOP_HEADER]], label %[[EXIT:.*]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/no-fold-tail-by-masking-iv-external-uses.ll b/llvm/test/Transforms/LoopVectorize/no-fold-tail-by-masking-iv-external-uses.ll
index 7d33f62..404ef09 100644
--- a/llvm/test/Transforms/LoopVectorize/no-fold-tail-by-masking-iv-external-uses.ll
+++ b/llvm/test/Transforms/LoopVectorize/no-fold-tail-by-masking-iv-external-uses.ll
@@ -37,8 +37,7 @@ define i32 @test(ptr %arr, i64 %n) {
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP17:%.*]] = add nsw i64 [[OFFSET_IDX]], -1
 ; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[ARR]], i64 [[TMP17]]
-; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[TMP18]], i32 0
-; CHECK-NEXT:    store <4 x i32> splat (i32 65), ptr [[TMP19]], align 4
+; CHECK-NEXT:    store <4 x i32> splat (i32 65), ptr [[TMP18]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/no_outside_user.ll b/llvm/test/Transforms/LoopVectorize/no_outside_user.ll
index ba85bb4..02f3242 100644
--- a/llvm/test/Transforms/LoopVectorize/no_outside_user.ll
+++ b/llvm/test/Transforms/LoopVectorize/no_outside_user.ll
@@ -28,9 +28,9 @@ define i32 @test1()  {
 ; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[TMP1]], 2
 ; CHECK-NEXT:    [[N_VEC:%.*]] = sub i32 [[TMP1]], [[N_MOD_VF]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = add i32 [[B_PROMOTED]], [[N_VEC]]
-; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[B_PROMOTED]], i64 0
-; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[INDUCTION:%.*]] = add <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[B_PROMOTED]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    [[INDUCTION:%.*]] = add <2 x i32> [[BROADCAST_SPLAT]], <i32 0, i32 1>
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
@@ -263,8 +263,8 @@ define i32 @test4(i32 %N)  {
 ; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i32 [[TMP1]], 2
 ; CHECK-NEXT:    [[N_VEC:%.*]] = sub i32 [[TMP1]], [[N_MOD_VF]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = add i32 [[B_PROMOTED]], [[N_VEC]]
-; CHECK-NEXT:    [[DOTSPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[B_PROMOTED]], i64 0
-; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[DOTSPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i32> poison, i32 [[B_PROMOTED]], i64 0
+; CHECK-NEXT:    [[DOTSPLAT:%.*]] = shufflevector <2 x i32> [[BROADCAST_SPLATINSERT]], <2 x i32> poison, <2 x i32> zeroinitializer
 ; CHECK-NEXT:    [[INDUCTION:%.*]] = add <2 x i32> [[DOTSPLAT]], <i32 0, i32 1>
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
@@ -669,15 +669,12 @@ define i32 @sum_arrays_outside_use(ptr %B, ptr %A, ptr %C, i32 %N)  {
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i32 [[B_PROMOTED]], [[INDEX]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = sext i32 [[OFFSET_IDX]] to i64
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[TMP6]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP8]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP6]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <2 x i32>, ptr [[TMP10]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <2 x i32>, ptr [[TMP9]], align 4
 ; CHECK-NEXT:    [[TMP11:%.*]] = add nsw <2 x i32> [[WIDE_LOAD]], [[WIDE_LOAD5]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[C]], i64 [[TMP6]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[TMP11]], ptr [[TMP13]], align 4
+; CHECK-NEXT:    store <2 x i32> [[TMP11]], ptr [[TMP12]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP14]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
@@ -742,17 +739,15 @@ define i32 @non_uniform_live_out() {
 ; CHECK-NEXT:    [[TMP0:%.*]] = add <2 x i32> [[VEC_IND]], splat (i32 7)
 ; CHECK-NEXT:    [[TMP1:%.*]] = extractelement <2 x i32> [[TMP0]], i32 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds [32 x i8], ptr @tab, i32 0, i32 [[TMP1]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[TMP3]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[TMP2]], align 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = add <2 x i8> [[WIDE_LOAD]], splat (i8 1)
-; CHECK-NEXT:    store <2 x i8> [[TMP4]], ptr [[TMP3]], align 1
+; CHECK-NEXT:    store <2 x i8> [[TMP4]], ptr [[TMP2]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], 20000
 ; CHECK-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <2 x i32> [[TMP0]], i32 1
-; CHECK-NEXT:    br i1 false, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 20000, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
@@ -765,9 +760,9 @@ define i32 @non_uniform_live_out() {
 ; CHECK-NEXT:    store i8 [[BUMP]], ptr [[ARRAYIDX]], align 1
 ; CHECK-NEXT:    [[INC]] = add nsw i32 [[I_08]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[I_08]], 20000
-; CHECK-NEXT:    br i1 [[EXITCOND]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label %[[FOR_END:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; CHECK:       [[FOR_END]]:
-; CHECK-NEXT:    [[LCSSA:%.*]] = phi i32 [ [[I_09]], %[[FOR_BODY]] ], [ [[TMP6]], %[[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[LCSSA:%.*]] = phi i32 [ [[I_09]], %[[FOR_BODY]] ]
 ; CHECK-NEXT:    [[ARRAYIDX_OUT:%.*]] = getelementptr inbounds [32 x i8], ptr @tab, i32 0, i32 [[LCSSA]]
 ; CHECK-NEXT:    store i8 42, ptr [[ARRAYIDX_OUT]], align 1
 ; CHECK-NEXT:    ret i32 0
diff --git a/llvm/test/Transforms/LoopVectorize/opaque-ptr.ll b/llvm/test/Transforms/LoopVectorize/opaque-ptr.ll
index 6f3736c..496285a 100644
--- a/llvm/test/Transforms/LoopVectorize/opaque-ptr.ll
+++ b/llvm/test/Transforms/LoopVectorize/opaque-ptr.ll
@@ -49,12 +49,10 @@ define void @test_ptr_iv_no_inbounds(ptr %p1.start, ptr %p2.start, ptr %p1.end)
 ; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[P1_START]], i64 [[OFFSET_IDX]]
 ; CHECK-NEXT:    [[OFFSET_IDX10:%.*]] = mul i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[NEXT_GEP11:%.*]] = getelementptr i8, ptr [[P2_START]], i64 [[OFFSET_IDX10]]
-; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr float, ptr [[NEXT_GEP]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP17]], align 4, !alias.scope [[META0:![0-9]+]], !noalias [[META3:![0-9]+]]
-; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr float, ptr [[NEXT_GEP11]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD12:%.*]] = load <2 x float>, ptr [[TMP18]], align 4, !alias.scope [[META3]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[NEXT_GEP]], align 4, !alias.scope [[META0:![0-9]+]], !noalias [[META3:![0-9]+]]
+; CHECK-NEXT:    [[WIDE_LOAD12:%.*]] = load <2 x float>, ptr [[NEXT_GEP11]], align 4, !alias.scope [[META3]]
 ; CHECK-NEXT:    [[TMP19:%.*]] = fadd <2 x float> [[WIDE_LOAD]], [[WIDE_LOAD12]]
-; CHECK-NEXT:    store <2 x float> [[TMP19]], ptr [[TMP17]], align 4, !alias.scope [[META0]], !noalias [[META3]]
+; CHECK-NEXT:    store <2 x float> [[TMP19]], ptr [[NEXT_GEP]], align 4, !alias.scope [[META0]], !noalias [[META3]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
@@ -137,12 +135,10 @@ define void @test_ptr_iv_with_inbounds(ptr %p1.start, ptr %p2.start, ptr %p1.end
 ; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[P1_START]], i64 [[OFFSET_IDX]]
 ; CHECK-NEXT:    [[OFFSET_IDX8:%.*]] = mul i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[NEXT_GEP9:%.*]] = getelementptr i8, ptr [[P2_START]], i64 [[OFFSET_IDX8]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr float, ptr [[NEXT_GEP]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP13]], align 4, !alias.scope [[META9:![0-9]+]], !noalias [[META12:![0-9]+]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr float, ptr [[NEXT_GEP9]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD10:%.*]] = load <2 x float>, ptr [[TMP14]], align 4, !alias.scope [[META12]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[NEXT_GEP]], align 4, !alias.scope [[META9:![0-9]+]], !noalias [[META12:![0-9]+]]
+; CHECK-NEXT:    [[WIDE_LOAD10:%.*]] = load <2 x float>, ptr [[NEXT_GEP9]], align 4, !alias.scope [[META12]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = fadd <2 x float> [[WIDE_LOAD]], [[WIDE_LOAD10]]
-; CHECK-NEXT:    store <2 x float> [[TMP15]], ptr [[TMP13]], align 4, !alias.scope [[META9]], !noalias [[META12]]
+; CHECK-NEXT:    store <2 x float> [[TMP15]], ptr [[NEXT_GEP]], align 4, !alias.scope [[META9]], !noalias [[META12]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
@@ -208,8 +204,7 @@ define void @store_pointer_induction(ptr %start, ptr %end) {
 ; CHECK-NEXT:    [[POINTER_PHI:%.*]] = phi ptr [ [[START]], [[VECTOR_PH]] ], [ [[PTR_IND:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <2 x i64> <i64 0, i64 8>
 ; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <2 x ptr> [[TMP5]], i32 0
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr ptr, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <2 x ptr> [[TMP5]], ptr [[TMP7]], align 4
+; CHECK-NEXT:    store <2 x ptr> [[TMP5]], ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i64 16
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/optimal-epilog-vectorization-liveout.ll b/llvm/test/Transforms/LoopVectorize/optimal-epilog-vectorization-liveout.ll
index 6695450..ee74f22 100644
--- a/llvm/test/Transforms/LoopVectorize/optimal-epilog-vectorization-liveout.ll
+++ b/llvm/test/Transforms/LoopVectorize/optimal-epilog-vectorization-liveout.ll
@@ -29,11 +29,9 @@ define signext i32 @f1(ptr noalias %A, ptr noalias %B, i32 signext %n) {
 ; VF-TWO-CHECK:       vector.body:
 ; VF-TWO-CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; VF-TWO-CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; VF-TWO-CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; VF-TWO-CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
-; VF-TWO-CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
-; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i32>, ptr [[TMP4]], align 4
+; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <2 x i32>, ptr [[TMP3]], align 4
 ; VF-TWO-CHECK-NEXT:    [[TMP5:%.*]] = add nsw <2 x i32> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; VF-TWO-CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VF-TWO-CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -54,11 +52,9 @@ define signext i32 @f1(ptr noalias %A, ptr noalias %B, i32 signext %n) {
 ; VF-TWO-CHECK:       vec.epilog.vector.body:
 ; VF-TWO-CHECK-NEXT:    [[INDEX6:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT9:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; VF-TWO-CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX6]]
-; VF-TWO-CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i32 0
-; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <2 x i32>, ptr [[TMP10]], align 4
+; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <2 x i32>, ptr [[TMP9]], align 4
 ; VF-TWO-CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX6]]
-; VF-TWO-CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <2 x i32>, ptr [[TMP12]], align 4
+; VF-TWO-CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <2 x i32>, ptr [[TMP11]], align 4
 ; VF-TWO-CHECK-NEXT:    [[TMP13:%.*]] = add nsw <2 x i32> [[WIDE_LOAD7]], [[WIDE_LOAD8]]
 ; VF-TWO-CHECK-NEXT:    [[INDEX_NEXT9]] = add nuw i64 [[INDEX6]], 2
 ; VF-TWO-CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT9]], [[N_VEC4]]
diff --git a/llvm/test/Transforms/LoopVectorize/optimal-epilog-vectorization.ll b/llvm/test/Transforms/LoopVectorize/optimal-epilog-vectorization.ll
index 1480bc9..a79a8dd 100644
--- a/llvm/test/Transforms/LoopVectorize/optimal-epilog-vectorization.ll
+++ b/llvm/test/Transforms/LoopVectorize/optimal-epilog-vectorization.ll
@@ -31,15 +31,12 @@ define dso_local void @f1(ptr noalias %aa, ptr noalias %bb, ptr noalias %cc, i32
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[BB:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[CC:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = fadd fast <4 x float> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[AA:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <4 x float> [[TMP5]], ptr [[TMP7]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP5]], ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -58,15 +55,12 @@ define dso_local void @f1(ptr noalias %aa, ptr noalias %bb, ptr noalias %cc, i32
 ; CHECK:       vec.epilog.vector.body:
 ; CHECK-NEXT:    [[INDEX6:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT9:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[BB]], i64 [[INDEX6]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds float, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x float>, ptr [[TMP11]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x float>, ptr [[TMP10]], align 4
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds float, ptr [[CC]], i64 [[INDEX6]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds float, ptr [[TMP12]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x float>, ptr [[TMP13]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x float>, ptr [[TMP12]], align 4
 ; CHECK-NEXT:    [[TMP14:%.*]] = fadd fast <4 x float> [[WIDE_LOAD7]], [[WIDE_LOAD8]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds float, ptr [[AA]], i64 [[INDEX6]]
-; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds float, ptr [[TMP15]], i32 0
-; CHECK-NEXT:    store <4 x float> [[TMP14]], ptr [[TMP16]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP14]], ptr [[TMP15]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT9]] = add nuw i64 [[INDEX6]], 4
 ; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT9]], [[N_VEC4]]
 ; CHECK-NEXT:    br i1 [[TMP17]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
@@ -187,8 +181,7 @@ define dso_local signext i32 @f2(ptr noalias %A, ptr noalias %B, i32 signext %n)
 ; CHECK-NEXT:    [[REVERSE:%.*]] = shufflevector <4 x float> [[WIDE_LOAD]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; CHECK-NEXT:    [[TMP16:%.*]] = fadd fast <4 x float> [[REVERSE]], splat (float 1.000000e+00)
 ; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds float, ptr [[TMP17]], i32 0
-; CHECK-NEXT:    store <4 x float> [[TMP16]], ptr [[TMP18]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP16]], ptr [[TMP17]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
@@ -219,8 +212,7 @@ define dso_local signext i32 @f2(ptr noalias %A, ptr noalias %B, i32 signext %n)
 ; CHECK-NEXT:    [[REVERSE10:%.*]] = shufflevector <4 x float> [[WIDE_LOAD9]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; CHECK-NEXT:    [[TMP28:%.*]] = fadd fast <4 x float> [[REVERSE10]], splat (float 1.000000e+00)
 ; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX7]]
-; CHECK-NEXT:    [[TMP30:%.*]] = getelementptr inbounds float, ptr [[TMP29]], i32 0
-; CHECK-NEXT:    store <4 x float> [[TMP28]], ptr [[TMP30]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP28]], ptr [[TMP29]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT11]] = add nuw i64 [[INDEX7]], 4
 ; CHECK-NEXT:    [[TMP31:%.*]] = icmp eq i64 [[INDEX_NEXT11]], [[N_VEC3]]
 ; CHECK-NEXT:    br i1 [[TMP31]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
@@ -326,8 +318,7 @@ define void @f3(ptr noalias %A, i64 %n) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <4 x i8> splat (i8 1), ptr [[TMP2]], align 1
+; CHECK-NEXT:    store <4 x i8> splat (i8 1), ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
@@ -346,8 +337,7 @@ define void @f3(ptr noalias %A, i64 %n) {
 ; CHECK:       vec.epilog.vector.body:
 ; CHECK-NEXT:    [[INDEX5:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT6:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX5]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <4 x i8> splat (i8 1), ptr [[TMP6]], align 1
+; CHECK-NEXT:    store <4 x i8> splat (i8 1), ptr [[TMP5]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT6]] = add nuw i64 [[INDEX5]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT6]], [[N_VEC3]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP9:![0-9]+]]
@@ -383,8 +373,7 @@ define void @f3(ptr noalias %A, i64 %n) {
 ; CHECK-PROFITABLE-BY-DEFAULT:       vector.body:
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store <4 x i8> splat (i8 1), ptr [[TMP2]], align 1
+; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store <4 x i8> splat (i8 1), ptr [[TMP1]], align 1
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -403,8 +392,7 @@ define void @f3(ptr noalias %A, i64 %n) {
 ; CHECK-PROFITABLE-BY-DEFAULT:       vec.epilog.vector.body:
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX5:%.*]] = phi i64 [ [[VEC_EPILOG_RESUME_VAL]], [[VEC_EPILOG_PH]] ], [ [[INDEX_NEXT6:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX5]]
-; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 0
-; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store <2 x i8> splat (i8 1), ptr [[TMP6]], align 1
+; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store <2 x i8> splat (i8 1), ptr [[TMP5]], align 1
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX_NEXT6]] = add nuw i64 [[INDEX5]], 2
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT6]], [[N_VEC3]]
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    br i1 [[TMP7]], label [[VEC_EPILOG_MIDDLE_BLOCK:%.*]], label [[VEC_EPILOG_VECTOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
@@ -477,8 +465,7 @@ define void @induction_resume_value_requires_non_trivial_scev_expansion(ptr %dst
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i8> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[DST:%.*]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[VEC_IND]], ptr [[TMP6]], align 1
+; CHECK-NEXT:    store <4 x i8> [[VEC_IND]], ptr [[TMP5]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i8> [[VEC_IND]], [[DOTSPLAT2]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 84
@@ -507,8 +494,7 @@ define void @induction_resume_value_requires_non_trivial_scev_expansion(ptr %dst
 ; CHECK-NEXT:    [[VEC_IND15:%.*]] = phi <4 x i8> [ [[INDUCTION12]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT16:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX17:%.*]] = add i64 1, [[INDEX7]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[OFFSET_IDX17]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[VEC_IND15]], ptr [[TMP12]], align 1
+; CHECK-NEXT:    store <4 x i8> [[VEC_IND15]], ptr [[TMP11]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT18]] = add nuw i64 [[INDEX7]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT16]] = add <4 x i8> [[VEC_IND15]], [[DOTSPLAT14]]
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT18]], 84
@@ -565,8 +551,7 @@ define void @induction_resume_value_requires_non_trivial_scev_expansion(ptr %dst
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[VEC_IND:%.*]] = phi <4 x i8> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[DST:%.*]], i64 [[OFFSET_IDX]]
-; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 0
-; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store <4 x i8> [[VEC_IND]], ptr [[TMP6]], align 1
+; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store <4 x i8> [[VEC_IND]], ptr [[TMP5]], align 1
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[VEC_IND_NEXT]] = add <4 x i8> [[VEC_IND]], [[DOTSPLAT2]]
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 84
@@ -595,8 +580,7 @@ define void @induction_resume_value_requires_non_trivial_scev_expansion(ptr %dst
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[VEC_IND15:%.*]] = phi <2 x i8> [ [[INDUCTION12]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT16:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[OFFSET_IDX17:%.*]] = add i64 1, [[INDEX7]]
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[OFFSET_IDX17]]
-; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
-; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store <2 x i8> [[VEC_IND15]], ptr [[TMP12]], align 1
+; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store <2 x i8> [[VEC_IND15]], ptr [[TMP11]], align 1
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX_NEXT18]] = add nuw i64 [[INDEX7]], 2
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[VEC_IND_NEXT16]] = add <2 x i8> [[VEC_IND15]], [[DOTSPLAT14]]
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT18]], 84
@@ -667,8 +651,7 @@ define void @f4(ptr noalias %A, i32 signext %n) {
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = trunc <4 x i32> [[VEC_IND]] to <4 x i8>
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[TMP1]], ptr [[TMP3]], align 1
+; CHECK-NEXT:    store <4 x i8> [[TMP1]], ptr [[TMP2]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -694,8 +677,7 @@ define void @f4(ptr noalias %A, i32 signext %n) {
 ; CHECK-NEXT:    [[VEC_IND7:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT8:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = trunc <4 x i32> [[VEC_IND7]] to <4 x i8>
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX6]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[TMP7]], ptr [[TMP9]], align 1
+; CHECK-NEXT:    store <4 x i8> [[TMP7]], ptr [[TMP8]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT9]] = add nuw i64 [[INDEX6]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT8]] = add <4 x i32> [[VEC_IND7]], splat (i32 4)
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT9]], [[N_VEC3]]
@@ -735,8 +717,7 @@ define void @f4(ptr noalias %A, i32 signext %n) {
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ <i32 0, i32 1, i32 2, i32 3>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP1:%.*]] = trunc <4 x i32> [[VEC_IND]] to <4 x i8>
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[INDEX]]
-; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
-; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store <4 x i8> [[TMP1]], ptr [[TMP3]], align 1
+; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store <4 x i8> [[TMP1]], ptr [[TMP2]], align 1
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -762,8 +743,7 @@ define void @f4(ptr noalias %A, i32 signext %n) {
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[VEC_IND7:%.*]] = phi <2 x i32> [ [[INDUCTION]], [[VEC_EPILOG_PH]] ], [ [[VEC_IND_NEXT8:%.*]], [[VEC_EPILOG_VECTOR_BODY]] ]
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP7:%.*]] = trunc <2 x i32> [[VEC_IND7]] to <2 x i8>
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX6]]
-; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i32 0
-; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store <2 x i8> [[TMP7]], ptr [[TMP9]], align 1
+; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    store <2 x i8> [[TMP7]], ptr [[TMP8]], align 1
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[INDEX_NEXT9]] = add nuw i64 [[INDEX6]], 2
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[VEC_IND_NEXT8]] = add <2 x i32> [[VEC_IND7]], splat (i32 2)
 ; CHECK-PROFITABLE-BY-DEFAULT-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT9]], [[N_VEC3]]
diff --git a/llvm/test/Transforms/LoopVectorize/optsize.ll b/llvm/test/Transforms/LoopVectorize/optsize.ll
index 38e436e..f0d026b 100644
--- a/llvm/test/Transforms/LoopVectorize/optsize.ll
+++ b/llvm/test/Transforms/LoopVectorize/optsize.ll
@@ -200,16 +200,15 @@ define i32 @foo_pgso() !prof !14 {
 ; NPGSO:       [[VECTOR_BODY]]:
 ; NPGSO-NEXT:    [[TMP0:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; NPGSO-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [32 x i8], ptr @tab, i32 0, i32 [[TMP0]]
-; NPGSO-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; NPGSO-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; NPGSO-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
 ; NPGSO-NEXT:    [[TMP3:%.*]] = icmp eq <4 x i8> [[WIDE_LOAD]], zeroinitializer
 ; NPGSO-NEXT:    [[TMP4:%.*]] = select <4 x i1> [[TMP3]], <4 x i8> splat (i8 2), <4 x i8> splat (i8 1)
-; NPGSO-NEXT:    store <4 x i8> [[TMP4]], ptr [[TMP2]], align 1
+; NPGSO-NEXT:    store <4 x i8> [[TMP4]], ptr [[TMP1]], align 1
 ; NPGSO-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[TMP0]], 4
 ; NPGSO-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], 200
 ; NPGSO-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; NPGSO:       [[MIDDLE_BLOCK]]:
-; NPGSO-NEXT:    br i1 false, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; NPGSO-NEXT:    br label %[[SCALAR_PH]]
 ; NPGSO:       [[SCALAR_PH]]:
 ; NPGSO-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 200, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; NPGSO-NEXT:    br label %[[FOR_BODY:.*]]
@@ -222,7 +221,7 @@ define i32 @foo_pgso() !prof !14 {
 ; NPGSO-NEXT:    store i8 [[DOT]], ptr [[ARRAYIDX]], align 1
 ; NPGSO-NEXT:    [[INC]] = add nsw i32 [[I_08]], 1
 ; NPGSO-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[I_08]], 202
-; NPGSO-NEXT:    br i1 [[EXITCOND]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; NPGSO-NEXT:    br i1 [[EXITCOND]], label %[[FOR_END:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; NPGSO:       [[FOR_END]]:
 ; NPGSO-NEXT:    ret i32 0
 ;
@@ -274,9 +273,9 @@ define void @pr43371() optsize {
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 756
 ; CHECK-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP28:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[FOR_COND_CLEANUP28:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 756, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY29:.*]]
 ; CHECK:       [[FOR_COND_CLEANUP28]]:
 ; CHECK-NEXT:    unreachable
@@ -312,9 +311,9 @@ define void @pr43371() optsize {
 ; PGSO-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 756
 ; PGSO-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP15:![0-9]+]]
 ; PGSO:       [[MIDDLE_BLOCK]]:
-; PGSO-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP28:.*]], label %[[SCALAR_PH]]
+; PGSO-NEXT:    br label %[[FOR_COND_CLEANUP28:.*]]
 ; PGSO:       [[SCALAR_PH]]:
-; PGSO-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 756, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; PGSO-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, %[[ENTRY]] ]
 ; PGSO-NEXT:    br label %[[FOR_BODY29:.*]]
 ; PGSO:       [[FOR_COND_CLEANUP28]]:
 ; PGSO-NEXT:    unreachable
@@ -350,9 +349,9 @@ define void @pr43371() optsize {
 ; NPGSO-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 756
 ; NPGSO-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
 ; NPGSO:       [[MIDDLE_BLOCK]]:
-; NPGSO-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP28:.*]], label %[[SCALAR_PH]]
+; NPGSO-NEXT:    br label %[[FOR_COND_CLEANUP28:.*]]
 ; NPGSO:       [[SCALAR_PH]]:
-; NPGSO-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 756, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; NPGSO-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, %[[ENTRY]] ]
 ; NPGSO-NEXT:    br label %[[FOR_BODY29:.*]]
 ; NPGSO:       [[FOR_COND_CLEANUP28]]:
 ; NPGSO-NEXT:    unreachable
@@ -411,9 +410,9 @@ define void @pr43371_pgso() !prof !14 {
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 756
 ; CHECK-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP28:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[FOR_COND_CLEANUP28:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 756, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY29:.*]]
 ; CHECK:       [[FOR_COND_CLEANUP28]]:
 ; CHECK-NEXT:    unreachable
@@ -449,9 +448,9 @@ define void @pr43371_pgso() !prof !14 {
 ; PGSO-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 756
 ; PGSO-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP19:![0-9]+]]
 ; PGSO:       [[MIDDLE_BLOCK]]:
-; PGSO-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP28:.*]], label %[[SCALAR_PH]]
+; PGSO-NEXT:    br label %[[FOR_COND_CLEANUP28:.*]]
 ; PGSO:       [[SCALAR_PH]]:
-; PGSO-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 756, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; PGSO-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, %[[ENTRY]] ]
 ; PGSO-NEXT:    br label %[[FOR_BODY29:.*]]
 ; PGSO:       [[FOR_COND_CLEANUP28]]:
 ; PGSO-NEXT:    unreachable
@@ -479,15 +478,14 @@ define void @pr43371_pgso() !prof !14 {
 ; NPGSO-NEXT:    [[TMP1:%.*]] = add i16 undef, [[OFFSET_IDX]]
 ; NPGSO-NEXT:    [[TMP2:%.*]] = zext i16 [[TMP1]] to i32
 ; NPGSO-NEXT:    [[TMP3:%.*]] = getelementptr [2592 x i16], ptr @cm_array, i32 0, i32 [[TMP2]]
-; NPGSO-NEXT:    [[TMP4:%.*]] = getelementptr i16, ptr [[TMP3]], i32 0
-; NPGSO-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP4]], align 1
+; NPGSO-NEXT:    store <2 x i16> zeroinitializer, ptr [[TMP3]], align 1
 ; NPGSO-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; NPGSO-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], 756
 ; NPGSO-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
 ; NPGSO:       [[MIDDLE_BLOCK]]:
-; NPGSO-NEXT:    br i1 true, label %[[FOR_COND_CLEANUP28:.*]], label %[[SCALAR_PH]]
+; NPGSO-NEXT:    br label %[[FOR_COND_CLEANUP28:.*]]
 ; NPGSO:       [[SCALAR_PH]]:
-; NPGSO-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 756, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
+; NPGSO-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
 ; NPGSO-NEXT:    br label %[[FOR_BODY29:.*]]
 ; NPGSO:       [[FOR_COND_CLEANUP28]]:
 ; NPGSO-NEXT:    unreachable
@@ -628,8 +626,6 @@ define i32 @pr45526_pgso() !prof !14 {
 ; NPGSO-NEXT:    br i1 [[TMP1]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
 ; NPGSO:       [[MIDDLE_BLOCK]]:
 ; NPGSO-NEXT:    [[VECTOR_RECUR_EXTRACT:%.*]] = extractelement <4 x i32> [[TMP0]], i32 3
-; NPGSO-NEXT:    [[VECTOR_RECUR_EXTRACT_FOR_PHI:%.*]] = extractelement <4 x i32> [[TMP0]], i32 2
-; NPGSO-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
 ; NPGSO:       [[SCALAR_PH]]:
 ; NPGSO-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 508, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; NPGSO-NEXT:    [[SCALAR_RECUR_INIT:%.*]] = phi i32 [ [[VECTOR_RECUR_EXTRACT]], %[[MIDDLE_BLOCK]] ], [ 5, %[[ENTRY]] ]
@@ -639,9 +635,9 @@ define i32 @pr45526_pgso() !prof !14 {
 ; NPGSO-NEXT:    [[FOR:%.*]] = phi i32 [ [[SCALAR_RECUR_INIT]], %[[SCALAR_PH]] ], [ [[PIVPLUS1]], %[[LOOP]] ]
 ; NPGSO-NEXT:    [[PIVPLUS1]] = add nuw nsw i32 [[PIV]], 1
 ; NPGSO-NEXT:    [[COND:%.*]] = icmp ult i32 [[PIV]], 510
-; NPGSO-NEXT:    br i1 [[COND]], label %[[LOOP]], label %[[EXIT]], !llvm.loop [[LOOP24:![0-9]+]]
+; NPGSO-NEXT:    br i1 [[COND]], label %[[LOOP]], label %[[EXIT:.*]], !llvm.loop [[LOOP24:![0-9]+]]
 ; NPGSO:       [[EXIT]]:
-; NPGSO-NEXT:    [[FOR_LCSSA:%.*]] = phi i32 [ [[FOR]], %[[LOOP]] ], [ [[VECTOR_RECUR_EXTRACT_FOR_PHI]], %[[MIDDLE_BLOCK]] ]
+; NPGSO-NEXT:    [[FOR_LCSSA:%.*]] = phi i32 [ [[FOR]], %[[LOOP]] ]
 ; NPGSO-NEXT:    ret i32 [[FOR_LCSSA]]
 ;
 entry:
@@ -842,13 +838,12 @@ define void @stride1_pgso(ptr noalias %B, i32 %BStride) !prof !14 {
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP0:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[B]], i32 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <2 x i16> splat (i16 42), ptr [[TMP2]], align 4
+; CHECK-NEXT:    store <2 x i16> splat (i16 42), ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[TMP0]], 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 false, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
@@ -859,7 +854,7 @@ define void @stride1_pgso(ptr noalias %B, i32 %BStride) !prof !14 {
 ; CHECK-NEXT:    store i16 42, ptr [[GEPOFB]], align 4
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i32 [[IV]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[IV_NEXT]], 1025
-; CHECK-NEXT:    br i1 [[EXITCOND]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label %[[FOR_END:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
 ; CHECK:       [[FOR_END]]:
 ; CHECK-NEXT:    ret void
 ;
@@ -875,13 +870,12 @@ define void @stride1_pgso(ptr noalias %B, i32 %BStride) !prof !14 {
 ; PGSO:       [[VECTOR_BODY]]:
 ; PGSO-NEXT:    [[TMP0:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; PGSO-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[B]], i32 [[TMP0]]
-; PGSO-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 0
-; PGSO-NEXT:    store <2 x i16> splat (i16 42), ptr [[TMP2]], align 4
+; PGSO-NEXT:    store <2 x i16> splat (i16 42), ptr [[TMP1]], align 4
 ; PGSO-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[TMP0]], 2
 ; PGSO-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1024
 ; PGSO-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
 ; PGSO:       [[MIDDLE_BLOCK]]:
-; PGSO-NEXT:    br i1 false, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; PGSO-NEXT:    br label %[[SCALAR_PH]]
 ; PGSO:       [[SCALAR_PH]]:
 ; PGSO-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
 ; PGSO-NEXT:    br label %[[FOR_BODY:.*]]
@@ -892,7 +886,7 @@ define void @stride1_pgso(ptr noalias %B, i32 %BStride) !prof !14 {
 ; PGSO-NEXT:    store i16 42, ptr [[GEPOFB]], align 4
 ; PGSO-NEXT:    [[IV_NEXT]] = add nuw nsw i32 [[IV]], 1
 ; PGSO-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[IV_NEXT]], 1025
-; PGSO-NEXT:    br i1 [[EXITCOND]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; PGSO-NEXT:    br i1 [[EXITCOND]], label %[[FOR_END:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
 ; PGSO:       [[FOR_END]]:
 ; PGSO-NEXT:    ret void
 ;
@@ -908,13 +902,12 @@ define void @stride1_pgso(ptr noalias %B, i32 %BStride) !prof !14 {
 ; NPGSO:       [[VECTOR_BODY]]:
 ; NPGSO-NEXT:    [[TMP0:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; NPGSO-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[B]], i32 [[TMP0]]
-; NPGSO-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 0
-; NPGSO-NEXT:    store <2 x i16> splat (i16 42), ptr [[TMP2]], align 4
+; NPGSO-NEXT:    store <2 x i16> splat (i16 42), ptr [[TMP1]], align 4
 ; NPGSO-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[TMP0]], 2
 ; NPGSO-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1024
 ; NPGSO-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP27:![0-9]+]]
 ; NPGSO:       [[MIDDLE_BLOCK]]:
-; NPGSO-NEXT:    br i1 false, label %[[FOR_END:.*]], label %[[SCALAR_PH]]
+; NPGSO-NEXT:    br label %[[SCALAR_PH]]
 ; NPGSO:       [[SCALAR_PH]]:
 ; NPGSO-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
 ; NPGSO-NEXT:    br label %[[FOR_BODY:.*]]
@@ -925,7 +918,7 @@ define void @stride1_pgso(ptr noalias %B, i32 %BStride) !prof !14 {
 ; NPGSO-NEXT:    store i16 42, ptr [[GEPOFB]], align 4
 ; NPGSO-NEXT:    [[IV_NEXT]] = add nuw nsw i32 [[IV]], 1
 ; NPGSO-NEXT:    [[EXITCOND:%.*]] = icmp eq i32 [[IV_NEXT]], 1025
-; NPGSO-NEXT:    br i1 [[EXITCOND]], label %[[FOR_END]], label %[[FOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
+; NPGSO-NEXT:    br i1 [[EXITCOND]], label %[[FOR_END:.*]], label %[[FOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
 ; NPGSO:       [[FOR_END]]:
 ; NPGSO-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/phi-cost.ll b/llvm/test/Transforms/LoopVectorize/phi-cost.ll
index aee80c9..667c6a5 100644
--- a/llvm/test/Transforms/LoopVectorize/phi-cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/phi-cost.ll
@@ -189,7 +189,7 @@ define i32 @red_phi_0(i32 %start, ptr %src) {
 ; CHECK-NEXT:    br i1 [[TMP1]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[TMP0]])
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
diff --git a/llvm/test/Transforms/LoopVectorize/pointer-induction-index-width-smaller-than-iv-width.ll b/llvm/test/Transforms/LoopVectorize/pointer-induction-index-width-smaller-than-iv-width.ll
index 4c04d96d..1bc98f9 100644
--- a/llvm/test/Transforms/LoopVectorize/pointer-induction-index-width-smaller-than-iv-width.ll
+++ b/llvm/test/Transforms/LoopVectorize/pointer-induction-index-width-smaller-than-iv-width.ll
@@ -22,14 +22,12 @@ define void @wide_ptr_induction_index_width_smaller_than_iv_width(ptr noalias %s
 ; CHECK-NEXT:    [[TMP3:%.*]] = add i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 3
 ; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <4 x ptr> [[VECTOR_GEP]], i32 0
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i64, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP6]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP5]], align 1
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[DST_0]], i64 [[TMP1]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[DST_0]], i64 [[TMP2]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[DST_0]], i64 [[TMP3]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[DST_0]], i64 [[TMP4]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <4 x i64> [[WIDE_LOAD]], ptr [[TMP11]], align 8
+; CHECK-NEXT:    store <4 x i64> [[WIDE_LOAD]], ptr [[TMP7]], align 8
 ; CHECK-NEXT:    store ptr [[TMP5]], ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <4 x ptr> [[VECTOR_GEP]], i32 1
 ; CHECK-NEXT:    store ptr [[TMP12]], ptr [[TMP8]], align 8
@@ -42,7 +40,7 @@ define void @wide_ptr_induction_index_width_smaller_than_iv_width(ptr noalias %s
 ; CHECK-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP15]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi ptr [ [[TMP0]], %[[MIDDLE_BLOCK]] ], [ [[SRC]], %[[ENTRY]] ]
@@ -58,7 +56,7 @@ define void @wide_ptr_induction_index_width_smaller_than_iv_width(ptr noalias %s
 ; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; CHECK-NEXT:    [[PTR_IV_NEXT]] = getelementptr i8, ptr [[PTR_IV]], i32 8
 ; CHECK-NEXT:    [[EC:%.*]] = icmp ult i64 [[IV]], 100
-; CHECK-NEXT:    br i1 [[EC]], label %[[LOOP]], label %[[EXIT]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label %[[LOOP]], label %[[EXIT:.*]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/pointer-induction-unroll.ll b/llvm/test/Transforms/LoopVectorize/pointer-induction-unroll.ll
index 97907b2..f36292b 100644
--- a/llvm/test/Transforms/LoopVectorize/pointer-induction-unroll.ll
+++ b/llvm/test/Transforms/LoopVectorize/pointer-induction-unroll.ll
@@ -60,7 +60,7 @@ define void @non_constant_scalar_expansion(i32 %0, ptr %call) {
 ; STRIDED-NEXT:    [[TMP23:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4294967264
 ; STRIDED-NEXT:    br i1 [[TMP23]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; STRIDED:       middle.block:
-; STRIDED-NEXT:    br i1 false, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; STRIDED-NEXT:    br label [[SCALAR_PH]]
 ; STRIDED:       scalar.ph:
 ; STRIDED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ -2, [[MIDDLE_BLOCK]] ], [ 30, [[ENTRY:%.*]] ]
 ; STRIDED-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi ptr [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ null, [[ENTRY]] ]
@@ -73,7 +73,7 @@ define void @non_constant_scalar_expansion(i32 %0, ptr %call) {
 ; STRIDED-NEXT:    store ptr [[P_0]], ptr [[ARRAYIDX]], align 4
 ; STRIDED-NEXT:    [[INC]] = add i32 [[TMP24]], 1
 ; STRIDED-NEXT:    [[TOBOOL_NOT:%.*]] = icmp eq i32 [[TMP24]], 0
-; STRIDED-NEXT:    br i1 [[TOBOOL_NOT]], label [[FOR_END]], label [[FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
+; STRIDED-NEXT:    br i1 [[TOBOOL_NOT]], label [[FOR_END:%.*]], label [[FOR_COND]], !llvm.loop [[LOOP3:![0-9]+]]
 ; STRIDED:       for.end:
 ; STRIDED-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/pointer-induction.ll b/llvm/test/Transforms/LoopVectorize/pointer-induction.ll
index a4f2b07..d1ae9ce 100644
--- a/llvm/test/Transforms/LoopVectorize/pointer-induction.ll
+++ b/llvm/test/Transforms/LoopVectorize/pointer-induction.ll
@@ -146,13 +146,11 @@ define void @pointer_induction_used_as_vector(ptr noalias %start.1, ptr noalias
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[START_1]], i64 [[OFFSET_IDX]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, <4 x ptr> [[VECTOR_GEP]], i64 1
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr ptr, ptr [[NEXT_GEP]], i32 0
-; CHECK-NEXT:    store <4 x ptr> [[TMP2]], ptr [[TMP3]], align 8
+; CHECK-NEXT:    store <4 x ptr> [[TMP2]], ptr [[NEXT_GEP]], align 8
 ; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <4 x ptr> [[VECTOR_GEP]], i32 0
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = add <4 x i8> [[WIDE_LOAD]], splat (i8 1)
-; CHECK-NEXT:    store <4 x i8> [[TMP6]], ptr [[TMP5]], align 1
+; CHECK-NEXT:    store <4 x i8> [[TMP6]], ptr [[TMP4]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i64 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -240,14 +238,13 @@ define void @non_constant_vector_expansion(i32 %0, ptr %call) {
 ; STRIDED-NEXT:    [[VECTOR_GEP:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <4 x i64> [[TMP4]]
 ; STRIDED-NEXT:    [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32
 ; STRIDED-NEXT:    [[TMP6:%.*]] = getelementptr ptr, ptr [[CALL:%.*]], i32 [[OFFSET_IDX]]
-; STRIDED-NEXT:    [[TMP7:%.*]] = getelementptr ptr, ptr [[TMP6]], i32 0
-; STRIDED-NEXT:    store <4 x ptr> [[VECTOR_GEP]], ptr [[TMP7]], align 4
+; STRIDED-NEXT:    store <4 x ptr> [[VECTOR_GEP]], ptr [[TMP6]], align 4
 ; STRIDED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; STRIDED-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i64 [[TMP3]]
 ; STRIDED-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; STRIDED-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; STRIDED:       middle.block:
-; STRIDED-NEXT:    br i1 false, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; STRIDED-NEXT:    br label [[SCALAR_PH]]
 ; STRIDED:       scalar.ph:
 ; STRIDED-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
 ; STRIDED-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi ptr [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ null, [[ENTRY]] ]
@@ -260,7 +257,7 @@ define void @non_constant_vector_expansion(i32 %0, ptr %call) {
 ; STRIDED-NEXT:    store ptr [[P_0]], ptr [[ARRAYIDX]], align 4
 ; STRIDED-NEXT:    [[INC]] = add i32 [[TMP9]], 1
 ; STRIDED-NEXT:    [[TOBOOL_NOT:%.*]] = icmp eq i32 [[TMP9]], 100
-; STRIDED-NEXT:    br i1 [[TOBOOL_NOT]], label [[FOR_END]], label [[FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
+; STRIDED-NEXT:    br i1 [[TOBOOL_NOT]], label [[FOR_END:%.*]], label [[FOR_COND]], !llvm.loop [[LOOP7:![0-9]+]]
 ; STRIDED:       for.end:
 ; STRIDED-NEXT:    ret void
 ;
@@ -311,11 +308,9 @@ define void @outside_lattice(ptr noalias %p, ptr noalias %q, i32 %n) {
 ; DEFAULT-NEXT:    [[VECTOR_GEP:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <4 x i64> <i64 0, i64 4, i64 8, i64 12>
 ; DEFAULT-NEXT:    [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32
 ; DEFAULT-NEXT:    [[TMP5:%.*]] = getelementptr inbounds ptr, ptr [[P]], i32 [[OFFSET_IDX]]
-; DEFAULT-NEXT:    [[TMP6:%.*]] = getelementptr inbounds ptr, ptr [[TMP5]], i32 0
-; DEFAULT-NEXT:    store <4 x ptr> [[VECTOR_GEP]], ptr [[TMP6]], align 8
+; DEFAULT-NEXT:    store <4 x ptr> [[VECTOR_GEP]], ptr [[TMP5]], align 8
 ; DEFAULT-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[Q:%.*]], i32 [[OFFSET_IDX]]
-; DEFAULT-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; DEFAULT-NEXT:    store <4 x i32> [[VEC_IND]], ptr [[TMP8]], align 4
+; DEFAULT-NEXT:    store <4 x i32> [[VEC_IND]], ptr [[TMP7]], align 4
 ; DEFAULT-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; DEFAULT-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
 ; DEFAULT-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i64 16
@@ -367,11 +362,9 @@ define void @outside_lattice(ptr noalias %p, ptr noalias %q, i32 %n) {
 ; STRIDED-NEXT:    [[VECTOR_GEP:%.*]] = getelementptr i8, ptr [[POINTER_PHI]], <4 x i64> <i64 0, i64 4, i64 8, i64 12>
 ; STRIDED-NEXT:    [[OFFSET_IDX:%.*]] = trunc i64 [[INDEX]] to i32
 ; STRIDED-NEXT:    [[TMP5:%.*]] = getelementptr inbounds ptr, ptr [[P]], i32 [[OFFSET_IDX]]
-; STRIDED-NEXT:    [[TMP6:%.*]] = getelementptr inbounds ptr, ptr [[TMP5]], i32 0
-; STRIDED-NEXT:    store <4 x ptr> [[VECTOR_GEP]], ptr [[TMP6]], align 8
+; STRIDED-NEXT:    store <4 x ptr> [[VECTOR_GEP]], ptr [[TMP5]], align 8
 ; STRIDED-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[Q:%.*]], i32 [[OFFSET_IDX]]
-; STRIDED-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; STRIDED-NEXT:    store <4 x i32> [[VEC_IND]], ptr [[TMP8]], align 4
+; STRIDED-NEXT:    store <4 x i32> [[VEC_IND]], ptr [[TMP7]], align 4
 ; STRIDED-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; STRIDED-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
 ; STRIDED-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i64 16
diff --git a/llvm/test/Transforms/LoopVectorize/pr30654-phiscev-sext-trunc.ll b/llvm/test/Transforms/LoopVectorize/pr30654-phiscev-sext-trunc.ll
index b2da8c7..c491477 100644
--- a/llvm/test/Transforms/LoopVectorize/pr30654-phiscev-sext-trunc.ll
+++ b/llvm/test/Transforms/LoopVectorize/pr30654-phiscev-sext-trunc.ll
@@ -81,8 +81,7 @@ define void @doit1(i32 %n, i32 %step) local_unnamed_addr {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr inbounds [250 x i32], ptr @a, i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, ptr [[TMP20]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[VEC_IND]], ptr [[TMP21]], align 4
+; CHECK-NEXT:    store <4 x i32> [[VEC_IND]], ptr [[TMP20]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], [[DOTSPLAT3]]
 ; CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -203,8 +202,7 @@ define void @doit2(i32 %n, i32 %step) local_unnamed_addr  {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds [250 x i32], ptr @a, i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[TMP19]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[VEC_IND]], ptr [[TMP20]], align 4
+; CHECK-NEXT:    store <4 x i32> [[VEC_IND]], ptr [[TMP19]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], [[DOTSPLAT3]]
 ; CHECK-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -397,8 +395,7 @@ define void @doit4(i32 %n, i8 signext %cstep) local_unnamed_addr {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i32> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds [250 x i32], ptr @a, i64 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[TMP17]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[VEC_IND]], ptr [[TMP18]], align 4
+; CHECK-NEXT:    store <4 x i32> [[VEC_IND]], ptr [[TMP17]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], [[DOTSPLAT3]]
 ; CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/pr35773.ll b/llvm/test/Transforms/LoopVectorize/pr35773.ll
index 2f14655..b716515 100644
--- a/llvm/test/Transforms/LoopVectorize/pr35773.ll
+++ b/llvm/test/Transforms/LoopVectorize/pr35773.ll
@@ -14,8 +14,7 @@ define void @doit1(ptr %ptr) {
 ; CHECK-NEXT:    [[I8_IV_NEXT]] = add <4 x i8> [[I8_IV]], [[IV_FROM_TRUNC]]
 
 ; CHECK-NEXT:    [[GEP1:%.+]] = getelementptr inbounds i32, ptr %ptr, i32 [[MAIN_IV]]
-; CHECK-NEXT:    [[GEP2:%.+]] = getelementptr inbounds i32, ptr [[GEP1]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[I32_IV]], ptr [[GEP2]], align 4
+; CHECK-NEXT:    store <4 x i32> [[I32_IV]], ptr [[GEP1]], align 4
 
 ; CHECK-NEXT:    [[MAIN_IV_NEXT]] = add nuw i32 [[MAIN_IV]], 4
 ; CHECK-NEXT:    [[I32_IV_NEXT]] = add <4 x i32> [[I32_IV]], splat (i32 36)
diff --git a/llvm/test/Transforms/LoopVectorize/pr39417-optsize-scevchecks.ll b/llvm/test/Transforms/LoopVectorize/pr39417-optsize-scevchecks.ll
index fde93d7..83646e9 100644
--- a/llvm/test/Transforms/LoopVectorize/pr39417-optsize-scevchecks.ll
+++ b/llvm/test/Transforms/LoopVectorize/pr39417-optsize-scevchecks.ll
@@ -58,14 +58,13 @@ define void @scev4stride1(ptr noalias nocapture %a, ptr noalias nocapture readon
 ; CHECK-NEXT:    [[TMP19:%.*]] = insertelement <4 x i32> [[TMP18]], i32 [[TMP15]], i32 2
 ; CHECK-NEXT:    [[TMP20:%.*]] = insertelement <4 x i32> [[TMP19]], i32 [[TMP16]], i32 3
 ; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i32, ptr [[TMP21]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP20]], ptr [[TMP22]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP20]], ptr [[TMP21]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
 ; CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop !0
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END_LOOPEXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END_LOOPEXIT:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK:       for.body:
 ; CHECK:       for.end.loopexit:
diff --git a/llvm/test/Transforms/LoopVectorize/pr44488-predication.ll b/llvm/test/Transforms/LoopVectorize/pr44488-predication.ll
index e7b87b0..315ea12 100644
--- a/llvm/test/Transforms/LoopVectorize/pr44488-predication.ll
+++ b/llvm/test/Transforms/LoopVectorize/pr44488-predication.ll
@@ -44,9 +44,9 @@ define i16 @test_true_and_false_branch_equal() {
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i32 [[INDEX_NEXT]], 12
 ; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 111, [[MIDDLE_BLOCK]] ], [ 99, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 99, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[I_07:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC7:%.*]], [[FOR_LATCH:%.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/pr45259.ll b/llvm/test/Transforms/LoopVectorize/pr45259.ll
index c7f2d7a..fade726 100644
--- a/llvm/test/Transforms/LoopVectorize/pr45259.ll
+++ b/llvm/test/Transforms/LoopVectorize/pr45259.ll
@@ -48,8 +48,7 @@ define i8 @widget(ptr %arr, i8 %t9) {
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[ARR]], i8 [[TMP12]]
 ; CHECK-NEXT:    [[TMP14:%.*]] = icmp slt <4 x i8> [[TMP11]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = zext <4 x i1> [[TMP14]] to <4 x i8>
-; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[TMP13]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[TMP15]], ptr [[TMP16]], align 1
+; CHECK-NEXT:    store <4 x i8> [[TMP15]], ptr [[TMP13]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i8> [[VEC_IND]], splat (i8 4)
 ; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/pr45679-fold-tail-by-masking.ll b/llvm/test/Transforms/LoopVectorize/pr45679-fold-tail-by-masking.ll
index b6aea9c..c044cc0 100644
--- a/llvm/test/Transforms/LoopVectorize/pr45679-fold-tail-by-masking.ll
+++ b/llvm/test/Transforms/LoopVectorize/pr45679-fold-tail-by-masking.ll
@@ -253,8 +253,8 @@ define void @load_variant(ptr noalias %a, ptr noalias %b) {
 ; CHECK:       pred.store.continue6:
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i8> [[VEC_IND]], splat (i8 4)
-; CHECK-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], 16
-; CHECK-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], 16
+; CHECK-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
@@ -320,8 +320,8 @@ define void @load_variant(ptr noalias %a, ptr noalias %b) {
 ; VF2UF2:       pred.store.continue6:
 ; VF2UF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; VF2UF2-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[STEP_ADD]], splat (i64 2)
-; VF2UF2-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], 16
-; VF2UF2-NEXT:    br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; VF2UF2-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], 16
+; VF2UF2-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; VF2UF2:       middle.block:
 ; VF2UF2-NEXT:    br label [[FOR_END:%.*]]
 ; VF2UF2:       scalar.ph:
diff --git a/llvm/test/Transforms/LoopVectorize/pr47343-expander-lcssa-after-cfg-update.ll b/llvm/test/Transforms/LoopVectorize/pr47343-expander-lcssa-after-cfg-update.ll
index 6aaa443..7e1a449 100644
--- a/llvm/test/Transforms/LoopVectorize/pr47343-expander-lcssa-after-cfg-update.ll
+++ b/llvm/test/Transforms/LoopVectorize/pr47343-expander-lcssa-after-cfg-update.ll
@@ -47,14 +47,14 @@ define void @f() {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    store i32 0, ptr @f.e, align 1, !alias.scope [[META0:![0-9]+]], !noalias [[META3:![0-9]+]]
-; CHECK-NEXT:    store i8 10, ptr [[TMP0]], align 1
+; CHECK-NEXT:    store i8 10, ptr [[TMP0]], align 1, !alias.scope [[META3]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 500
 ; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 500, [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[LOOP_PREHEADER]] ], [ 0, [[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[IV_NEXT:%.*]], [[LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/pr50686.ll b/llvm/test/Transforms/LoopVectorize/pr50686.ll
index fb7947e..14bcfde 100644
--- a/llvm/test/Transforms/LoopVectorize/pr50686.ll
+++ b/llvm/test/Transforms/LoopVectorize/pr50686.ll
@@ -18,26 +18,25 @@ define void @m(ptr nocapture %p, ptr nocapture %p2, i32 %q) {
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[TMP1:%.*]] = load i32, ptr [[P2]], align 4, !alias.scope !0
+; CHECK-NEXT:    [[TMP1:%.*]] = load i32, ptr [[P2]], align 4, !alias.scope [[META0:![0-9]+]]
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i32> poison, i32 [[TMP1]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP2:%.*]] = sub nsw <4 x i32> zeroinitializer, [[BROADCAST_SPLAT]]
-; CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[ARRAYIDX9_1]], align 4, !alias.scope !0
+; CHECK-NEXT:    [[TMP3:%.*]] = load i32, ptr [[ARRAYIDX9_1]], align 4, !alias.scope [[META0]]
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT2:%.*]] = insertelement <4 x i32> poison, i32 [[TMP3]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT3:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT2]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = sub nsw <4 x i32> [[TMP2]], [[BROADCAST_SPLAT3]]
-; CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[ARRAYIDX9_2]], align 4, !alias.scope !0
+; CHECK-NEXT:    [[TMP5:%.*]] = load i32, ptr [[ARRAYIDX9_2]], align 4, !alias.scope [[META0]]
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT4:%.*]] = insertelement <4 x i32> poison, i32 [[TMP5]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT5:%.*]] = shufflevector <4 x i32> [[BROADCAST_SPLATINSERT4]], <4 x i32> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP6:%.*]] = sub nsw <4 x i32> [[TMP4]], [[BROADCAST_SPLAT5]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[P]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP6]], ptr [[TMP8]], align 4, !alias.scope !3, !noalias !0
+; CHECK-NEXT:    store <4 x i32> [[TMP6]], ptr [[TMP7]], align 4, !alias.scope [[META3:![0-9]+]], !noalias [[META0]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 60
-; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 60
+; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[FOR_END17:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 60, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label [[FOR_COND5:%.*]]
@@ -53,7 +52,7 @@ define void @m(ptr nocapture %p, ptr nocapture %p2, i32 %q) {
 ; CHECK-NEXT:    store i32 [[SUB_2]], ptr [[ARRAYIDX14]], align 4
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add nuw nsw i64 [[INDVARS_IV]], 1
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDVARS_IV_NEXT]], 63
-; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_END17]], label [[FOR_COND5]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND]], label [[FOR_END17:%.*]], label [[FOR_COND5]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       for.end17:
 ; CHECK-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/pr55167-fold-tail-live-out.ll b/llvm/test/Transforms/LoopVectorize/pr55167-fold-tail-live-out.ll
index bd96471..a28bdb8 100644
--- a/llvm/test/Transforms/LoopVectorize/pr55167-fold-tail-live-out.ll
+++ b/llvm/test/Transforms/LoopVectorize/pr55167-fold-tail-live-out.ll
@@ -34,10 +34,10 @@ define i32 @test(i32 %a, i1 %c.1, i1 %c.2 ) #0 {
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP10:%.*]] = call i32 @llvm.vector.reduce.add.v2i32(<2 x i32> [[PREDPHI7]])
 ; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <2 x i32> [[PREDPHI5]], i32 1
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 182, [[MIDDLE_BLOCK]] ], [ 6, [[BB:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP10]], [[MIDDLE_BLOCK]] ], [ 35902, [[BB]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 6, [[BB:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 35902, [[BB]] ]
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; CHECK:       loop.header:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/pr58811-scev-expansion.ll b/llvm/test/Transforms/LoopVectorize/pr58811-scev-expansion.ll
index af1c146..98da2b2 100644
--- a/llvm/test/Transforms/LoopVectorize/pr58811-scev-expansion.ll
+++ b/llvm/test/Transforms/LoopVectorize/pr58811-scev-expansion.ll
@@ -31,8 +31,7 @@ define void @test1_pr58811() {
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 196
 ; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[IND_ESCAPE:%.*]] = sub i32 [[IND_END]], [[INDUCTION_IV_LCSSA]]
-; CHECK-NEXT:    br i1 false, label [[LOOP_3_PREHEADER:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 196, [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_2_PREHEADER]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_2_PREHEADER]] ]
@@ -43,9 +42,9 @@ define void @test1_pr58811() {
 ; CHECK-NEXT:    [[IV_4_NEXT]] = sub i32 [[IV_4]], [[IV_1_LCSSA]]
 ; CHECK-NEXT:    [[IV_3_NEXT]] = add i16 [[IV_3]], 1
 ; CHECK-NEXT:    [[CMP88_1:%.*]] = icmp ult i16 [[IV_3]], 198
-; CHECK-NEXT:    br i1 [[CMP88_1]], label [[LOOP_2]], label [[LOOP_3_PREHEADER]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP88_1]], label [[LOOP_2]], label [[LOOP_3_PREHEADER:%.*]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       loop.3.preheader:
-; CHECK-NEXT:    [[IV_4_LCSSA:%.*]] = phi i32 [ [[IV_4]], [[LOOP_2]] ], [ [[IND_ESCAPE]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[IV_4_LCSSA:%.*]] = phi i32 [ [[IV_4]], [[LOOP_2]] ]
 ; CHECK-NEXT:    br label [[LOOP_3:%.*]]
 ; CHECK:       loop.3:
 ; CHECK-NEXT:    [[IV_5:%.*]] = phi i32 [ [[SUB93_2]], [[LOOP_3]] ], [ 0, [[LOOP_3_PREHEADER]] ]
@@ -126,8 +125,7 @@ define void @test2_pr58811() {
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 196
 ; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[IND_ESCAPE:%.*]] = sub i32 [[IND_END]], [[INDUCTION_IV_LCSSA]]
-; CHECK-NEXT:    br i1 false, label [[LOOP_4_PREHEADER:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 196, [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_3_PREHEADER]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_3_PREHEADER]] ]
@@ -138,9 +136,9 @@ define void @test2_pr58811() {
 ; CHECK-NEXT:    [[SUB93_1]] = sub i32 [[IV_5]], [[IV_2_LCSSA]]
 ; CHECK-NEXT:    [[INC_1]] = add i16 [[IV_4]], 1
 ; CHECK-NEXT:    [[CMP88_1:%.*]] = icmp ult i16 [[IV_4]], 198
-; CHECK-NEXT:    br i1 [[CMP88_1]], label [[LOOP_3]], label [[LOOP_4_PREHEADER]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP88_1]], label [[LOOP_3]], label [[LOOP_4_PREHEADER:%.*]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       loop.4.preheader:
-; CHECK-NEXT:    [[IV_5_LCSSA:%.*]] = phi i32 [ [[IV_5]], [[LOOP_3]] ], [ [[IND_ESCAPE]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[IV_5_LCSSA:%.*]] = phi i32 [ [[IV_5]], [[LOOP_3]] ]
 ; CHECK-NEXT:    br label [[LOOP_4]]
 ; CHECK:       loop.4:
 ; CHECK-NEXT:    [[IV_6:%.*]] = phi i32 [ [[SUB93_2]], [[LOOP_4]] ], [ 0, [[LOOP_4_PREHEADER]] ]
@@ -204,8 +202,7 @@ define void @test3_pr58811() {
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], 196
 ; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[IND_ESCAPE:%.*]] = sub i32 [[IND_END]], [[TMP3]]
-; CHECK-NEXT:    br i1 false, label [[LOOP_4_PREHEADER:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 196, [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_3_PREHEADER]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ 0, [[LOOP_3_PREHEADER]] ]
@@ -216,9 +213,9 @@ define void @test3_pr58811() {
 ; CHECK-NEXT:    [[SUB93_1]] = sub i32 [[IV_4]], [[ADD101_LCSSA]]
 ; CHECK-NEXT:    [[INC_1]] = add i16 [[IV_3]], 1
 ; CHECK-NEXT:    [[CMP88_1:%.*]] = icmp ult i16 [[IV_3]], 198
-; CHECK-NEXT:    br i1 [[CMP88_1]], label [[LOOP_3]], label [[LOOP_4_PREHEADER]], !llvm.loop [[LOOP7:![0-9]+]]
+; CHECK-NEXT:    br i1 [[CMP88_1]], label [[LOOP_3]], label [[LOOP_4_PREHEADER:%.*]], !llvm.loop [[LOOP7:![0-9]+]]
 ; CHECK:       loop.4.preheader:
-; CHECK-NEXT:    [[IV_4_LCSSA:%.*]] = phi i32 [ [[IV_4]], [[LOOP_3]] ], [ [[IND_ESCAPE]], [[MIDDLE_BLOCK]] ]
+; CHECK-NEXT:    [[IV_4_LCSSA:%.*]] = phi i32 [ [[IV_4]], [[LOOP_3]] ]
 ; CHECK-NEXT:    br label [[LOOP_4:%.*]]
 ; CHECK:       loop.4:
 ; CHECK-NEXT:    [[IV_5:%.*]] = phi i32 [ [[SUB93_2]], [[LOOP_4]] ], [ 0, [[LOOP_4_PREHEADER]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/pr66616.ll b/llvm/test/Transforms/LoopVectorize/pr66616.ll
index b5f9e99..a39fd47 100644
--- a/llvm/test/Transforms/LoopVectorize/pr66616.ll
+++ b/llvm/test/Transforms/LoopVectorize/pr66616.ll
@@ -21,9 +21,9 @@ define void @pr66616(ptr %ptr) {
 ; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x i32> [[TMP1]], i32 3
-; CHECK-NEXT:    br i1 true, label [[PREHEADER:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[PREHEADER:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 0, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_1:%.*]]
 ; CHECK:       loop.1:
 ; CHECK-NEXT:    [[IV_1:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INC:%.*]], [[LOOP_1]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/predicate-switch.ll b/llvm/test/Transforms/LoopVectorize/predicate-switch.ll
index c4a10e1..724aed8 100644
--- a/llvm/test/Transforms/LoopVectorize/predicate-switch.ll
+++ b/llvm/test/Transforms/LoopVectorize/predicate-switch.ll
@@ -22,8 +22,7 @@ define void @switch4_default_common_dest_with_case(ptr %start, ptr %end) {
 ; IC1-NEXT:    [[TMP2:%.*]] = add i64 [[INDEX]], 1
 ; IC1-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[START]], i64 [[TMP1]]
 ; IC1-NEXT:    [[NEXT_GEP3:%.*]] = getelementptr i8, ptr [[START]], i64 [[TMP2]]
-; IC1-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
-; IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[TMP3]], align 1
+; IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[NEXT_GEP]], align 1
 ; IC1-NEXT:    [[TMP7:%.*]] = icmp eq <2 x i8> [[WIDE_LOAD]], splat (i8 -12)
 ; IC1-NEXT:    [[TMP4:%.*]] = icmp eq <2 x i8> [[WIDE_LOAD]], splat (i8 13)
 ; IC1-NEXT:    [[TMP11:%.*]] = or <2 x i1> [[TMP7]], [[TMP4]]
@@ -120,9 +119,8 @@ define void @switch4_default_common_dest_with_case(ptr %start, ptr %end) {
 ; IC2-NEXT:    [[NEXT_GEP3:%.*]] = getelementptr i8, ptr [[START]], i64 [[TMP2]]
 ; IC2-NEXT:    [[NEXT_GEP4:%.*]] = getelementptr i8, ptr [[START]], i64 [[TMP3]]
 ; IC2-NEXT:    [[NEXT_GEP5:%.*]] = getelementptr i8, ptr [[START]], i64 [[TMP4]]
-; IC2-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
 ; IC2-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 2
-; IC2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[TMP5]], align 1
+; IC2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[NEXT_GEP]], align 1
 ; IC2-NEXT:    [[WIDE_LOAD6:%.*]] = load <2 x i8>, ptr [[TMP6]], align 1
 ; IC2-NEXT:    [[TMP13:%.*]] = icmp eq <2 x i8> [[WIDE_LOAD]], splat (i8 -12)
 ; IC2-NEXT:    [[TMP14:%.*]] = icmp eq <2 x i8> [[WIDE_LOAD6]], splat (i8 -12)
@@ -340,21 +338,21 @@ define void @switch_to_header(ptr %start) {
 ; IC1-NEXT:  [[ENTRY:.*]]:
 ; IC1-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; IC1:       [[LOOP_HEADER]]:
-; IC1-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[IF_THEN:.*]] ]
+; IC1-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[IF_THEN1:.*]] ]
 ; IC1-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; IC1-NEXT:    switch i64 [[IV]], label %[[LOOP_LATCH:.*]] [
-; IC1-NEXT:      i64 120, label %[[IF_THEN]]
+; IC1-NEXT:      i64 120, label %[[IF_THEN1]]
 ; IC1-NEXT:      i64 100, label %[[LOOP_LATCH]]
 ; IC1-NEXT:    ]
-; IC1:       [[IF_THEN]]:
+; IC1:       [[IF_THEN1]]:
 ; IC1-NEXT:    br label %[[LOOP_HEADER]]
-; IC1:       [[IF_THEN1:.*:]]
+; IC1:       [[IF_THEN:.*:]]
 ; IC1-NEXT:    [[GEP:%.*]] = getelementptr inbounds i64, ptr [[START]], i64 poison
 ; IC1-NEXT:    store i64 42, ptr [[GEP]], align 1
 ; IC1-NEXT:    unreachable
 ; IC1:       [[LOOP_LATCH]]:
 ; IC1-NEXT:    [[CMP:%.*]] = icmp eq i64 [[IV_NEXT]], 100
-; IC1-NEXT:    br i1 [[CMP]], label %[[EXIT:.*]], label %[[IF_THEN]]
+; IC1-NEXT:    br i1 [[CMP]], label %[[EXIT:.*]], label %[[IF_THEN1]]
 ; IC1:       [[EXIT]]:
 ; IC1-NEXT:    ret void
 ;
@@ -363,21 +361,21 @@ define void @switch_to_header(ptr %start) {
 ; IC2-NEXT:  [[ENTRY:.*]]:
 ; IC2-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; IC2:       [[LOOP_HEADER]]:
-; IC2-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[IF_THEN:.*]] ]
+; IC2-NEXT:    [[IV:%.*]] = phi i64 [ 0, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[IF_THEN1:.*]] ]
 ; IC2-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; IC2-NEXT:    switch i64 [[IV]], label %[[LOOP_LATCH:.*]] [
-; IC2-NEXT:      i64 120, label %[[IF_THEN]]
+; IC2-NEXT:      i64 120, label %[[IF_THEN1]]
 ; IC2-NEXT:      i64 100, label %[[LOOP_LATCH]]
 ; IC2-NEXT:    ]
-; IC2:       [[IF_THEN]]:
+; IC2:       [[IF_THEN1]]:
 ; IC2-NEXT:    br label %[[LOOP_HEADER]]
-; IC2:       [[IF_THEN1:.*:]]
+; IC2:       [[IF_THEN:.*:]]
 ; IC2-NEXT:    [[GEP:%.*]] = getelementptr inbounds i64, ptr [[START]], i64 poison
 ; IC2-NEXT:    store i64 42, ptr [[GEP]], align 1
 ; IC2-NEXT:    unreachable
 ; IC2:       [[LOOP_LATCH]]:
 ; IC2-NEXT:    [[CMP:%.*]] = icmp eq i64 [[IV_NEXT]], 100
-; IC2-NEXT:    br i1 [[CMP]], label %[[EXIT:.*]], label %[[IF_THEN]]
+; IC2-NEXT:    br i1 [[CMP]], label %[[EXIT:.*]], label %[[IF_THEN1]]
 ; IC2:       [[EXIT]]:
 ; IC2-NEXT:    ret void
 ;
@@ -415,15 +413,14 @@ define void @switch_all_to_default(ptr %start) {
 ; IC1:       [[VECTOR_BODY]]:
 ; IC1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[START]], i64 [[INDEX]]
-; IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; IC1-NEXT:    store <2 x i64> splat (i64 42), ptr [[TMP2]], align 1
+; IC1-NEXT:    store <2 x i64> splat (i64 42), ptr [[TMP1]], align 1
 ; IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; IC1-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; IC1-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; IC1:       [[MIDDLE_BLOCK]]:
-; IC1-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IC1-NEXT:    br label %[[EXIT:.*]]
 ; IC1:       [[SCALAR_PH]]:
-; IC1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; IC1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; IC1-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; IC1:       [[LOOP_HEADER]]:
 ; IC1-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -449,17 +446,16 @@ define void @switch_all_to_default(ptr %start) {
 ; IC2:       [[VECTOR_BODY]]:
 ; IC2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; IC2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[START]], i64 [[INDEX]]
-; IC2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
 ; IC2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 2
-; IC2-NEXT:    store <2 x i64> splat (i64 42), ptr [[TMP4]], align 1
+; IC2-NEXT:    store <2 x i64> splat (i64 42), ptr [[TMP2]], align 1
 ; IC2-NEXT:    store <2 x i64> splat (i64 42), ptr [[TMP5]], align 1
 ; IC2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; IC2-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
 ; IC2-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; IC2:       [[MIDDLE_BLOCK]]:
-; IC2-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; IC2-NEXT:    br label %[[EXIT:.*]]
 ; IC2:       [[SCALAR_PH]]:
-; IC2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; IC2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; IC2-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; IC2:       [[LOOP_HEADER]]:
 ; IC2-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/predicatedinst-loop-invariant.ll b/llvm/test/Transforms/LoopVectorize/predicatedinst-loop-invariant.ll
index 0a97510..cd44c3d 100644
--- a/llvm/test/Transforms/LoopVectorize/predicatedinst-loop-invariant.ll
+++ b/llvm/test/Transforms/LoopVectorize/predicatedinst-loop-invariant.ll
@@ -261,3 +261,55 @@ loop.latch:                                       ; preds = %cond.false, %loop.h
 exit:                                             ; preds = %loop.latch
   ret void
 }
+
+; Test case for https://github.com/llvm/llvm-project/issues/149347.
+; FIXME: Currently mis-compiles.
+define void @test_store_to_invariant_address_needs_mask_due_to_low_trip_count(ptr %dst) {
+; CHECK-LABEL: define void @test_store_to_invariant_address_needs_mask_due_to_low_trip_count(
+; CHECK-SAME: ptr [[DST:%.*]]) {
+; CHECK-NEXT:  [[ENTRY:.*]]:
+; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
+; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
+; CHECK:       [[VECTOR_BODY]]:
+; CHECK-NEXT:    store i32 0, ptr [[DST]], align 4
+; CHECK-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
+; CHECK:       [[MIDDLE_BLOCK]]:
+; CHECK-NEXT:    br label %[[EXIT:.*]]
+; CHECK:       [[SCALAR_PH]]:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
+; CHECK:       [[LOOP_HEADER]]:
+; CHECK-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
+; CHECK-NEXT:    br i1 true, label %[[LOOP_LATCH]], label %[[ELSE:.*]]
+; CHECK:       [[ELSE]]:
+; CHECK-NEXT:    br label %[[LOOP_LATCH]]
+; CHECK:       [[LOOP_LATCH]]:
+; CHECK-NEXT:    [[MERGE:%.*]] = phi i32 [ 1, %[[LOOP_HEADER]] ], [ 0, %[[ELSE]] ]
+; CHECK-NEXT:    store i32 [[MERGE]], ptr [[DST]], align 4
+; CHECK-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
+; CHECK-NEXT:    [[EC:%.*]] = icmp eq i16 [[IV_NEXT]], 3
+; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK:       [[EXIT]]:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br label %loop.header
+
+loop.header:
+  %iv = phi i16 [ 0, %entry ], [ %iv.next, %loop.latch ]
+  br i1 true, label %loop.latch, label %else
+
+else:
+  br label %loop.latch
+
+loop.latch:
+  %merge = phi i32 [ 1, %loop.header ], [ 0, %else ]
+  store i32 %merge, ptr %dst, align 4
+  %iv.next = add i16 %iv, 1
+  %ec = icmp eq i16 %iv.next, 3
+  br i1 %ec, label %exit, label %loop.header
+
+exit:
+  ret void
+}
diff --git a/llvm/test/Transforms/LoopVectorize/preserve-or-disjoint.ll b/llvm/test/Transforms/LoopVectorize/preserve-or-disjoint.ll
index 3323844..b6156a8 100644
--- a/llvm/test/Transforms/LoopVectorize/preserve-or-disjoint.ll
+++ b/llvm/test/Transforms/LoopVectorize/preserve-or-disjoint.ll
@@ -15,12 +15,11 @@ define void @generate_disjoint_flags(i64 %n, ptr noalias %x) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[X]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = or disjoint <4 x i32> [[WIDE_LOAD]], splat (i32 1)
 ; CHECK-NEXT:    [[TMP4:%.*]] = or <4 x i32> [[WIDE_LOAD]], splat (i32 3)
 ; CHECK-NEXT:    [[TMP5:%.*]] = add nuw nsw <4 x i32> [[TMP3]], [[TMP4]]
-; CHECK-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP2]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/reduction-inloop-min-max.ll b/llvm/test/Transforms/LoopVectorize/reduction-inloop-min-max.ll
index 3488f52..c1cc9e5 100644
--- a/llvm/test/Transforms/LoopVectorize/reduction-inloop-min-max.ll
+++ b/llvm/test/Transforms/LoopVectorize/reduction-inloop-min-max.ll
@@ -19,7 +19,7 @@ define i32 @reduction_smin(ptr nocapture %A, ptr nocapture %B) {
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -66,7 +66,7 @@ define i32 @reduction_smin_select_ops_flipped(ptr nocapture %A, ptr nocapture %B
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -113,7 +113,7 @@ define i32 @reduction_smin_intrinsic(ptr nocapture %A, ptr nocapture %B) {
 ; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.smin.v4i32(<4 x i32> [[TMP1]])
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -161,7 +161,7 @@ define i32 @reduction_umax(ptr nocapture %A, ptr nocapture %B) {
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -208,7 +208,7 @@ define i32 @reduction_umax_select_ops_flipped(ptr nocapture %A, ptr nocapture %B
 ; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -255,7 +255,7 @@ define i32 @reduction_umax_intrinsic(ptr nocapture %A, ptr nocapture %B) {
 ; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.umax.v4i32(<4 x i32> [[TMP1]])
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
diff --git a/llvm/test/Transforms/LoopVectorize/reduction-inloop-pred.ll b/llvm/test/Transforms/LoopVectorize/reduction-inloop-pred.ll
index fbe3a7a..795605d 100644
--- a/llvm/test/Transforms/LoopVectorize/reduction-inloop-pred.ll
+++ b/llvm/test/Transforms/LoopVectorize/reduction-inloop-pred.ll
@@ -1370,7 +1370,7 @@ define float @reduction_conditional(ptr %A, ptr %B, ptr %C, float %S) {
 ; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP13:%.*]] = call fast float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> [[PREDPHI3]])
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
diff --git a/llvm/test/Transforms/LoopVectorize/reduction-inloop-uf4.ll b/llvm/test/Transforms/LoopVectorize/reduction-inloop-uf4.ll
index d9819ae..12d83eb 100644
--- a/llvm/test/Transforms/LoopVectorize/reduction-inloop-uf4.ll
+++ b/llvm/test/Transforms/LoopVectorize/reduction-inloop-uf4.ll
@@ -38,7 +38,7 @@ define i32 @reduction_sum_single(ptr noalias nocapture %A) {
 ; CHECK-NEXT:    [[BIN_RDX:%.*]] = add i32 [[TMP7]], [[TMP5]]
 ; CHECK-NEXT:    [[BIN_RDX7:%.*]] = add i32 [[TMP9]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX8:%.*]] = add i32 [[TMP11]], [[BIN_RDX7]]
-; CHECK-NEXT:    br i1 true, label [[DOT_CRIT_EDGE:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[DOT_CRIT_EDGE:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[DOTLR_PH:%.*]]
 ; CHECK:       .lr.ph:
@@ -118,7 +118,7 @@ define i64 @reduction_sum_chain(ptr noalias %p, ptr noalias %q) {
 ; CHECK-NEXT:    [[BIN_RDX:%.*]] = add i64 [[TMP19]], [[TMP17]]
 ; CHECK-NEXT:    [[BIN_RDX11:%.*]] = add i64 [[TMP21]], [[BIN_RDX]]
 ; CHECK-NEXT:    [[BIN_RDX12:%.*]] = add i64 [[TMP23]], [[BIN_RDX11]]
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
diff --git a/llvm/test/Transforms/LoopVectorize/reduction-inloop.ll b/llvm/test/Transforms/LoopVectorize/reduction-inloop.ll
index e762c9f..b302868 100644
--- a/llvm/test/Transforms/LoopVectorize/reduction-inloop.ll
+++ b/llvm/test/Transforms/LoopVectorize/reduction-inloop.ll
@@ -20,7 +20,7 @@ define i32 @reduction_sum_single(ptr noalias nocapture %A) {
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[DOT_CRIT_EDGE:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[DOT_CRIT_EDGE:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[DOTLR_PH:%.*]]
 ; CHECK:       .lr.ph:
@@ -73,7 +73,7 @@ define i32 @reduction_sum(ptr noalias nocapture %A, ptr noalias nocapture %B) {
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[DOT_CRIT_EDGE:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[DOT_CRIT_EDGE:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[DOTLR_PH:%.*]]
 ; CHECK:       .lr.ph:
@@ -124,7 +124,7 @@ define i32 @reduction_sum_const(ptr noalias nocapture %A) {
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[DOT_CRIT_EDGE:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[DOT_CRIT_EDGE:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[DOTLR_PH:%.*]]
 ; CHECK:       .lr.ph:
@@ -178,7 +178,7 @@ define i32 @reduction_prod(ptr noalias nocapture %A, ptr noalias nocapture %B) {
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[DOT_CRIT_EDGE:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[DOT_CRIT_EDGE:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[DOTLR_PH:%.*]]
 ; CHECK:       .lr.ph:
@@ -235,7 +235,7 @@ define i32 @reduction_mix(ptr noalias nocapture %A, ptr noalias nocapture %B) {
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[DOT_CRIT_EDGE:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[DOT_CRIT_EDGE:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[DOTLR_PH:%.*]]
 ; CHECK:       .lr.ph:
@@ -289,7 +289,7 @@ define i32 @reduction_mul(ptr noalias nocapture %A, ptr noalias nocapture %B) {
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[DOT_CRIT_EDGE:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[DOT_CRIT_EDGE:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[DOTLR_PH:%.*]]
 ; CHECK:       .lr.ph:
@@ -340,7 +340,7 @@ define i32 @start_at_non_zero(ptr nocapture %in, ptr nocapture %coeff, ptr nocap
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -392,7 +392,7 @@ define i32 @reduction_and(ptr nocapture %A, ptr nocapture %B) {
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -443,7 +443,7 @@ define i32 @reduction_or(ptr nocapture %A, ptr nocapture %B) {
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -494,7 +494,7 @@ define i32 @reduction_xor(ptr nocapture %A, ptr nocapture %B) {
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -544,7 +544,7 @@ define float @reduction_fadd(ptr nocapture %A, ptr nocapture %B) {
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -596,7 +596,7 @@ define float @reduction_fmul(ptr nocapture %A, ptr nocapture %B) {
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -645,7 +645,7 @@ define i32 @reduction_sub_lhs(ptr noalias nocapture %A) {
 ; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP1]])
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -705,7 +705,7 @@ define float @reduction_conditional(ptr %A, ptr %B, ptr %C, float %S) {
 ; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP13:%.*]] = call fast float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> [[PREDPHI3]])
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -836,7 +836,7 @@ define i32 @reduction_predicated(ptr noalias nocapture %A, ptr noalias nocapture
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[DOT_CRIT_EDGE:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[DOT_CRIT_EDGE:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[DOTLR_PH:%.*]]
 ; CHECK:       .lr.ph:
@@ -887,7 +887,7 @@ define i8 @reduction_add_trunc(ptr noalias nocapture %A) {
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP4:%.*]] = call i8 @llvm.vector.reduce.add.v4i8(<4 x i8> [[TMP2]])
-; CHECK-NEXT:    br i1 true, label [[DOT_CRIT_EDGE:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[DOT_CRIT_EDGE:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[DOTLR_PH:%.*]]
 ; CHECK:       .lr.ph:
@@ -936,7 +936,7 @@ define i8 @reduction_and_trunc(ptr noalias nocapture %A) {
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP4:%.*]] = call i8 @llvm.vector.reduce.and.v4i8(<4 x i8> [[TMP2]])
-; CHECK-NEXT:    br i1 true, label [[DOT_CRIT_EDGE:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[DOT_CRIT_EDGE:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[DOTLR_PH:%.*]]
 ; CHECK:       .lr.ph:
@@ -1445,7 +1445,7 @@ define i32 @predicated_or_dominates_reduction(ptr %b) {
 ; CHECK-NEXT:    [[TMP49:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP49]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP44:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.cond.cleanup:
diff --git a/llvm/test/Transforms/LoopVectorize/reduction-odd-interleave-counts.ll b/llvm/test/Transforms/LoopVectorize/reduction-odd-interleave-counts.ll
index c4f51d9..daf4cba 100644
--- a/llvm/test/Transforms/LoopVectorize/reduction-odd-interleave-counts.ll
+++ b/llvm/test/Transforms/LoopVectorize/reduction-odd-interleave-counts.ll
@@ -8,10 +8,9 @@ define i32 @reduction_sum(i64 %n, ptr noalias nocapture %A) {
 ; UF3-NEXT:   [[SUM1:%.+]] = phi <4 x i32> [ zeroinitializer, %vector.ph ], [ [[SUM1_NEXT:%.+]], %vector.body ]
 ; UF3-NEXT:   [[SUM2:%.+]] = phi <4 x i32> [ zeroinitializer, %vector.ph ], [ [[SUM2_NEXT:%.+]],  %vector.body ]
 ; UF3-NEXT:   [[GEP0:%.+]] = getelementptr inbounds i32, ptr %A, i64 [[IV]]
-; UF3-NEXT:   [[L_GEP0:%.+]] = getelementptr inbounds i32, ptr [[GEP0]], i32 0
 ; UF3-NEXT:   [[L_GEP1:%.+]] = getelementptr inbounds i32, ptr [[GEP0]], i32 4
 ; UF3-NEXT:   [[L_GEP2:%.+]] = getelementptr inbounds i32, ptr [[GEP0]], i32 8
-; UF3-NEXT:   [[L0:%.+]] = load <4 x i32>, ptr [[L_GEP0]], align 4
+; UF3-NEXT:   [[L0:%.+]] = load <4 x i32>, ptr [[GEP0]], align 4
 ; UF3-NEXT:   [[L1:%.+]] = load <4 x i32>, ptr [[L_GEP1]], align 4
 ; UF3-NEXT:   [[L2:%.+]] = load <4 x i32>, ptr [[L_GEP2]], align 4
 ; UF3-NEXT:   [[SUM0_NEXT]] = add <4 x i32> [[SUM0]], [[L0]]
@@ -35,12 +34,11 @@ define i32 @reduction_sum(i64 %n, ptr noalias nocapture %A) {
 ; UF5-NEXT:   [[SUM3:%.+]] = phi <4 x i32> [ zeroinitializer, %vector.ph ], [ [[SUM3_NEXT:%.+]], %vector.body ]
 ; UF5-NEXT:   [[SUM4:%.+]] = phi <4 x i32> [ zeroinitializer, %vector.ph ], [ [[SUM4_NEXT:%.+]], %vector.body ]
 ; UF5-NEXT:   [[GEP0:%.+]] = getelementptr inbounds i32, ptr %A, i64 [[IV]]
-; UF5-NEXT:   [[L_GEP0:%.+]] = getelementptr inbounds i32, ptr [[GEP0]], i32 0
 ; UF5-NEXT:   [[L_GEP1:%.+]] = getelementptr inbounds i32, ptr [[GEP0]], i32 4
 ; UF5-NEXT:   [[L_GEP2:%.+]] = getelementptr inbounds i32, ptr [[GEP0]], i32 8
 ; UF5-NEXT:   [[L_GEP3:%.+]] = getelementptr inbounds i32, ptr [[GEP0]], i32 12
 ; UF5-NEXT:   [[L_GEP4:%.+]] = getelementptr inbounds i32, ptr [[GEP0]], i32 16
-; UF5-NEXT:   [[L0:%.+]] = load <4 x i32>, ptr [[L_GEP0]], align 4
+; UF5-NEXT:   [[L0:%.+]] = load <4 x i32>, ptr [[GEP0]], align 4
 ; UF5-NEXT:   [[L1:%.+]] = load <4 x i32>, ptr [[L_GEP1]], align 4
 ; UF5-NEXT:   [[L2:%.+]] = load <4 x i32>, ptr [[L_GEP2]], align 4
 ; UF5-NEXT:   [[L3:%.+]] = load <4 x i32>, ptr [[L_GEP3]], align 4
diff --git a/llvm/test/Transforms/LoopVectorize/reduction-with-invariant-store.ll b/llvm/test/Transforms/LoopVectorize/reduction-with-invariant-store.ll
index 7579fbc..b8c2405 100644
--- a/llvm/test/Transforms/LoopVectorize/reduction-with-invariant-store.ll
+++ b/llvm/test/Transforms/LoopVectorize/reduction-with-invariant-store.ll
@@ -15,8 +15,7 @@ target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f3
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY:%.*]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[SRC:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP4]] = add <4 x i32> [[VEC_PHI]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
@@ -24,7 +23,7 @@ target datalayout = "e-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f3
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP6:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP4]])
 ; CHECK-NEXT:    store i32 [[TMP6]], ptr [[GEP_DST:%.*]], align 4
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH:%.*]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 define void @reduc_store(ptr %dst, ptr readonly %src) {
 entry:
   %gep.dst = getelementptr inbounds i32, ptr %dst, i64 42
@@ -253,7 +252,7 @@ for.end:
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP36:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[TMP34]])
 ; CHECK-NEXT:    store i32 [[TMP36]], ptr [[GEP_DST:%.*]], align 4
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH:%.*]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 define void @reduc_store_inside_unrolled(ptr %dst, ptr readonly %src) {
 entry:
   %gep.dst = getelementptr inbounds i32, ptr %dst, i64 42
@@ -522,7 +521,7 @@ define void @test_drop_poison_generating_dead_recipe(ptr %dst) {
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i64 @llvm.vector.reduce.add.v4i64(<4 x i64> [[TMP0]])
 ; CHECK-NEXT:    store i64 [[TMP2]], ptr [[DST:%.*]], align 8
-; CHECK-NEXT:    br i1 false, label %exit, label %scalar.ph
+; CHECK-NEXT:    br label %scalar.ph
 ; CHECK:       scalar.ph:
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/reduction.ll b/llvm/test/Transforms/LoopVectorize/reduction.ll
index 757be04..aa1ac25 100644
--- a/llvm/test/Transforms/LoopVectorize/reduction.ll
+++ b/llvm/test/Transforms/LoopVectorize/reduction.ll
@@ -776,7 +776,7 @@ define float @reduction_conditional(ptr %A, ptr %B, ptr %C, float %S) {
 ; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP13:%.*]] = call fast float @llvm.vector.reduce.fadd.v4f32(float 0.000000e+00, <4 x float> [[PREDPHI3]])
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
diff --git a/llvm/test/Transforms/LoopVectorize/remarks-reduction-inloop.ll b/llvm/test/Transforms/LoopVectorize/remarks-reduction-inloop.ll
index 581ccbf..be0e0d1 100644
--- a/llvm/test/Transforms/LoopVectorize/remarks-reduction-inloop.ll
+++ b/llvm/test/Transforms/LoopVectorize/remarks-reduction-inloop.ll
@@ -16,18 +16,17 @@ define i32 @reduction_sum(ptr noalias nocapture %A, ptr noalias nocapture %B) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[TMP4]] = add i32 [[TMP3]], [[VEC_PHI]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 256, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP4]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[BODY:.*]]
 ; CHECK:       [[BODY]]:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[INDVARS_IV_NEXT:%.*]], %[[BODY]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/reuse-lcssa-phi-scev-expansion.ll b/llvm/test/Transforms/LoopVectorize/reuse-lcssa-phi-scev-expansion.ll
index 1fc34db..4612545 100644
--- a/llvm/test/Transforms/LoopVectorize/reuse-lcssa-phi-scev-expansion.ll
+++ b/llvm/test/Transforms/LoopVectorize/reuse-lcssa-phi-scev-expansion.ll
@@ -104,27 +104,23 @@ define void @runtime_checks_ptr_inductions(ptr %dst.1, ptr %dst.2, i1 %c) {
 ; CHECK-LABEL: define void @runtime_checks_ptr_inductions(
 ; CHECK-SAME: ptr [[DST_1:%.*]], ptr [[DST_2:%.*]], i1 [[C:%.*]]) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
-; CHECK-NEXT:    [[DST_11:%.*]] = ptrtoint ptr [[DST_1]] to i64
 ; CHECK-NEXT:    br label %[[LOOP_1:.*]]
 ; CHECK:       [[LOOP_1]]:
-; CHECK-NEXT:    [[INDVAR:%.*]] = phi i64 [ [[INDVAR_NEXT:%.*]], %[[LOOP_1]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    [[PTR_IV_1:%.*]] = phi ptr [ [[DST_1]], %[[ENTRY]] ], [ [[PTR_IV_1_NEXT:%.*]], %[[LOOP_1]] ]
 ; CHECK-NEXT:    [[CALL:%.*]] = call i32 @val()
 ; CHECK-NEXT:    [[SEL_DST:%.*]] = select i1 [[C]], ptr [[DST_1]], ptr [[DST_2]]
 ; CHECK-NEXT:    [[PTR_IV_1_NEXT]] = getelementptr i8, ptr [[PTR_IV_1]], i64 1
 ; CHECK-NEXT:    [[EC_1:%.*]] = icmp eq i32 [[CALL]], 0
-; CHECK-NEXT:    [[INDVAR_NEXT]] = add i64 [[INDVAR]], 1
 ; CHECK-NEXT:    br i1 [[EC_1]], label %[[LOOP_2_HEADER_PREHEADER:.*]], label %[[LOOP_1]]
 ; CHECK:       [[LOOP_2_HEADER_PREHEADER]]:
-; CHECK-NEXT:    [[SEL_DST_LCSSA2:%.*]] = phi ptr [ [[SEL_DST]], %[[LOOP_1]] ]
-; CHECK-NEXT:    [[INDVAR_LCSSA:%.*]] = phi i64 [ [[INDVAR]], %[[LOOP_1]] ]
+; CHECK-NEXT:    [[SEL_DST_LCSSA1:%.*]] = phi ptr [ [[SEL_DST]], %[[LOOP_1]] ]
 ; CHECK-NEXT:    [[PTR_IV_1_LCSSA:%.*]] = phi ptr [ [[PTR_IV_1]], %[[LOOP_1]] ]
 ; CHECK-NEXT:    [[SEL_DST_LCSSA:%.*]] = phi ptr [ [[SEL_DST]], %[[LOOP_1]] ]
-; CHECK-NEXT:    [[SEL_DST_LCSSA23:%.*]] = ptrtoint ptr [[SEL_DST_LCSSA2]] to i64
+; CHECK-NEXT:    [[SEL_DST_LCSSA12:%.*]] = ptrtoint ptr [[SEL_DST_LCSSA1]] to i64
 ; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
 ; CHECK:       [[VECTOR_MEMCHECK]]:
-; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[INDVAR_LCSSA]], [[DST_11]]
-; CHECK-NEXT:    [[TMP1:%.*]] = sub i64 [[TMP0]], [[SEL_DST_LCSSA23]]
+; CHECK-NEXT:    [[TMP0:%.*]] = ptrtoint ptr [[PTR_IV_1_LCSSA]] to i64
+; CHECK-NEXT:    [[TMP1:%.*]] = sub i64 [[TMP0]], [[SEL_DST_LCSSA12]]
 ; CHECK-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP1]], 2
 ; CHECK-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
@@ -135,10 +131,8 @@ define void @runtime_checks_ptr_inductions(ptr %dst.1, ptr %dst.2, i1 %c) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[PTR_IV_1_LCSSA]], i64 [[INDEX]]
 ; CHECK-NEXT:    [[NEXT_GEP4:%.*]] = getelementptr i8, ptr [[SEL_DST_LCSSA]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[NEXT_GEP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[TMP4]], align 1
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
-; CHECK-NEXT:    store <2 x i8> [[WIDE_LOAD]], ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i8>, ptr [[NEXT_GEP4]], align 1
+; CHECK-NEXT:    store <2 x i8> [[WIDE_LOAD]], ptr [[NEXT_GEP]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1022
 ; CHECK-NEXT:    br i1 [[TMP6]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -146,13 +140,13 @@ define void @runtime_checks_ptr_inductions(ptr %dst.1, ptr %dst.2, i1 %c) {
 ; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1023, %[[MIDDLE_BLOCK]] ], [ 1, %[[LOOP_2_HEADER_PREHEADER]] ], [ 1, %[[VECTOR_MEMCHECK]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL5:%.*]] = phi ptr [ [[TMP2]], %[[MIDDLE_BLOCK]] ], [ [[PTR_IV_1_LCSSA]], %[[LOOP_2_HEADER_PREHEADER]] ], [ [[PTR_IV_1_LCSSA]], %[[VECTOR_MEMCHECK]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL6:%.*]] = phi ptr [ [[TMP3]], %[[MIDDLE_BLOCK]] ], [ [[SEL_DST_LCSSA]], %[[LOOP_2_HEADER_PREHEADER]] ], [ [[SEL_DST_LCSSA]], %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL4:%.*]] = phi ptr [ [[TMP2]], %[[MIDDLE_BLOCK]] ], [ [[PTR_IV_1_LCSSA]], %[[LOOP_2_HEADER_PREHEADER]] ], [ [[PTR_IV_1_LCSSA]], %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL5:%.*]] = phi ptr [ [[TMP3]], %[[MIDDLE_BLOCK]] ], [ [[SEL_DST_LCSSA]], %[[LOOP_2_HEADER_PREHEADER]] ], [ [[SEL_DST_LCSSA]], %[[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label %[[LOOP_2_HEADER:.*]]
 ; CHECK:       [[LOOP_2_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[DEC7:%.*]], %[[LOOP_2_LATCH:.*]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
-; CHECK-NEXT:    [[PTR_IV_2:%.*]] = phi ptr [ [[PTR_IV_2_NEXT:%.*]], %[[LOOP_2_LATCH]] ], [ [[BC_RESUME_VAL5]], %[[SCALAR_PH]] ]
-; CHECK-NEXT:    [[PTR_IV_3:%.*]] = phi ptr [ [[PTR_IV_3_NEXT:%.*]], %[[LOOP_2_LATCH]] ], [ [[BC_RESUME_VAL6]], %[[SCALAR_PH]] ]
+; CHECK-NEXT:    [[PTR_IV_2:%.*]] = phi ptr [ [[PTR_IV_2_NEXT:%.*]], %[[LOOP_2_LATCH]] ], [ [[BC_RESUME_VAL4]], %[[SCALAR_PH]] ]
+; CHECK-NEXT:    [[PTR_IV_3:%.*]] = phi ptr [ [[PTR_IV_3_NEXT:%.*]], %[[LOOP_2_LATCH]] ], [ [[BC_RESUME_VAL5]], %[[SCALAR_PH]] ]
 ; CHECK-NEXT:    [[EC_2:%.*]] = icmp eq i32 [[IV]], 1024
 ; CHECK-NEXT:    br i1 [[EC_2]], label %[[EXIT:.*]], label %[[LOOP_2_LATCH]]
 ; CHECK:       [[LOOP_2_LATCH]]:
@@ -213,10 +207,8 @@ define void @expand_diff_scev_unknown(ptr %dst, i1 %invar.c, i32 %step) mustprog
 ; CHECK-NEXT:    br i1 [[INVAR_C]], label %[[LOOP_2_PREHEADER:.*]], label %[[LOOP_1]]
 ; CHECK:       [[LOOP_2_PREHEADER]]:
 ; CHECK-NEXT:    [[INDVAR_LCSSA1:%.*]] = phi i32 [ [[INDVAR]], %[[LOOP_1]] ]
-; CHECK-NEXT:    [[INDVAR_LCSSA:%.*]] = phi i32 [ [[INDVAR]], %[[LOOP_1]] ]
 ; CHECK-NEXT:    [[IV_1_LCSSA:%.*]] = phi i32 [ [[IV_1]], %[[LOOP_1]] ]
-; CHECK-NEXT:    [[TMP0:%.*]] = shl i32 [[STEP]], 1
-; CHECK-NEXT:    [[TMP1:%.*]] = add i32 [[INDVAR_LCSSA]], [[TMP0]]
+; CHECK-NEXT:    [[TMP1:%.*]] = add i32 [[IV_1_LCSSA]], [[STEP]]
 ; CHECK-NEXT:    [[SMAX:%.*]] = call i32 @llvm.smax.i32(i32 [[TMP1]], i32 0)
 ; CHECK-NEXT:    [[TMP2:%.*]] = mul i32 [[STEP]], -2
 ; CHECK-NEXT:    [[TMP3:%.*]] = mul i32 [[INDVAR_LCSSA1]], -1
@@ -242,8 +234,7 @@ define void @expand_diff_scev_unknown(ptr %dst, i1 %invar.c, i32 %step) mustprog
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i32 [[IV_1_LCSSA]], [[INDEX]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[DST]], i32 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    store <2 x i32> zeroinitializer, ptr [[TMP12]], align 4
+; CHECK-NEXT:    store <2 x i32> zeroinitializer, ptr [[TMP11]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP13]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
@@ -289,71 +280,59 @@ define void @expand_diff_neg_ptrtoint_expr(ptr %src, ptr %start) {
 ; CHECK-SAME: ptr [[SRC:%.*]], ptr [[START:%.*]]) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
 ; CHECK-NEXT:    [[SRC2:%.*]] = ptrtoint ptr [[SRC]] to i64
-; CHECK-NEXT:    [[START1:%.*]] = ptrtoint ptr [[START]] to i64
 ; CHECK-NEXT:    br label %[[LOOP_1:.*]]
 ; CHECK:       [[LOOP_1]]:
-; CHECK-NEXT:    [[INDVAR:%.*]] = phi i64 [ [[INDVAR_NEXT:%.*]], %[[LOOP_1]] ], [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 1, %[[ENTRY]] ], [ [[IV_NEXT:%.*]], %[[LOOP_1]] ]
 ; CHECK-NEXT:    [[PTR_IV_1:%.*]] = phi ptr [ [[START]], %[[ENTRY]] ], [ [[PTR_IV_1_NEXT:%.*]], %[[LOOP_1]] ]
 ; CHECK-NEXT:    [[PTR_IV_1_NEXT]] = getelementptr i8, ptr [[PTR_IV_1]], i64 8
 ; CHECK-NEXT:    call void @foo()
 ; CHECK-NEXT:    [[IV_NEXT]] = add i64 [[IV]], 1
 ; CHECK-NEXT:    [[EC_1:%.*]] = icmp eq i64 [[IV_NEXT]], 32
-; CHECK-NEXT:    [[INDVAR_NEXT]] = add i64 [[INDVAR]], 1
 ; CHECK-NEXT:    br i1 [[EC_1]], label %[[LOOP_2_PREHEADER:.*]], label %[[LOOP_1]]
 ; CHECK:       [[LOOP_2_PREHEADER]]:
-; CHECK-NEXT:    [[INDVAR_LCSSA:%.*]] = phi i64 [ [[INDVAR]], %[[LOOP_1]] ]
 ; CHECK-NEXT:    [[PTR_IV_1_NEXT_LCSSA:%.*]] = phi ptr [ [[PTR_IV_1_NEXT]], %[[LOOP_1]] ]
 ; CHECK-NEXT:    br label %[[LOOP_2:.*]]
 ; CHECK:       [[LOOP_2]]:
-; CHECK-NEXT:    [[INDVAR3:%.*]] = phi i64 [ 0, %[[LOOP_2_PREHEADER]] ], [ [[INDVAR_NEXT4:%.*]], %[[LOOP_2]] ]
 ; CHECK-NEXT:    [[IV_1:%.*]] = phi i64 [ [[IV_NEXT_1:%.*]], %[[LOOP_2]] ], [ 1, %[[LOOP_2_PREHEADER]] ]
 ; CHECK-NEXT:    [[PTR_IV_2:%.*]] = phi ptr [ [[PTR_IV_2_NEXT:%.*]], %[[LOOP_2]] ], [ [[PTR_IV_1_NEXT_LCSSA]], %[[LOOP_2_PREHEADER]] ]
 ; CHECK-NEXT:    call void @bar()
 ; CHECK-NEXT:    [[PTR_IV_2_NEXT]] = getelementptr i8, ptr [[PTR_IV_2]], i64 8
 ; CHECK-NEXT:    [[IV_NEXT_1]] = add i64 [[IV_1]], 1
 ; CHECK-NEXT:    [[EC_2:%.*]] = icmp eq i64 [[IV_NEXT_1]], 32
-; CHECK-NEXT:    [[INDVAR_NEXT4]] = add i64 [[INDVAR3]], 1
 ; CHECK-NEXT:    br i1 [[EC_2]], label %[[LOOP_3_PREHEADER:.*]], label %[[LOOP_2]]
 ; CHECK:       [[LOOP_3_PREHEADER]]:
-; CHECK-NEXT:    [[INDVAR3_LCSSA:%.*]] = phi i64 [ [[INDVAR3]], %[[LOOP_2]] ]
-; CHECK-NEXT:    [[PTR_IV_2_NEXT_LCSSA:%.*]] = phi ptr [ [[PTR_IV_2_NEXT]], %[[LOOP_2]] ]
+; CHECK-NEXT:    [[TMP1:%.*]] = phi ptr [ [[PTR_IV_2_NEXT]], %[[LOOP_2]] ]
 ; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
 ; CHECK:       [[VECTOR_MEMCHECK]]:
-; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[START1]], 16
-; CHECK-NEXT:    [[TMP1:%.*]] = sub i64 [[TMP0]], [[SRC2]]
-; CHECK-NEXT:    [[TMP2:%.*]] = shl nuw nsw i64 [[INDVAR_LCSSA]], 3
-; CHECK-NEXT:    [[TMP3:%.*]] = add i64 [[TMP2]], [[TMP1]]
-; CHECK-NEXT:    [[TMP4:%.*]] = shl nuw nsw i64 [[INDVAR3_LCSSA]], 3
-; CHECK-NEXT:    [[TMP5:%.*]] = add i64 [[TMP4]], [[TMP3]]
-; CHECK-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP5]], 16
+; CHECK-NEXT:    [[TMP0:%.*]] = sub i64 0, [[SRC2]]
+; CHECK-NEXT:    [[TMP5:%.*]] = ptrtoint ptr [[TMP1]] to i64
+; CHECK-NEXT:    [[TMP2:%.*]] = add i64 [[TMP5]], [[TMP0]]
+; CHECK-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP2]], 16
 ; CHECK-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[PTR_IV_2_NEXT_LCSSA]], i64 -16
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[TMP1]], i64 -16
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; CHECK-NEXT:    [[OFFSET_IDX5:%.*]] = mul i64 [[INDEX]], 8
-; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[PTR_IV_2_NEXT_LCSSA]], i64 [[OFFSET_IDX5]]
+; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[TMP1]], i64 [[OFFSET_IDX5]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = add i64 [[OFFSET_IDX]], -1
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr double, ptr [[SRC]], i64 [[TMP7]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr i64, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP9]], align 8
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 0
-; CHECK-NEXT:    store <2 x i64> [[WIDE_LOAD]], ptr [[TMP10]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP8]], align 8
+; CHECK-NEXT:    store <2 x i64> [[WIDE_LOAD]], ptr [[NEXT_GEP]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], -2
 ; CHECK-NEXT:    br i1 [[TMP11]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 false, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
 ; CHECK:       [[SCALAR_PH]]:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ -1, %[[MIDDLE_BLOCK]] ], [ 1, %[[LOOP_3_PREHEADER]] ], [ 1, %[[VECTOR_MEMCHECK]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL6:%.*]] = phi ptr [ [[TMP6]], %[[MIDDLE_BLOCK]] ], [ [[PTR_IV_2_NEXT_LCSSA]], %[[LOOP_3_PREHEADER]] ], [ [[PTR_IV_2_NEXT_LCSSA]], %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL3:%.*]] = phi ptr [ [[TMP3]], %[[MIDDLE_BLOCK]] ], [ [[TMP1]], %[[LOOP_3_PREHEADER]] ], [ [[TMP1]], %[[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label %[[LOOP_3:.*]]
 ; CHECK:       [[LOOP_3]]:
 ; CHECK-NEXT:    [[IV_2:%.*]] = phi i64 [ [[IV_NEXT_2:%.*]], %[[LOOP_3]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
-; CHECK-NEXT:    [[PTR_IV_3:%.*]] = phi ptr [ [[PTR_IV_3_NEXT:%.*]], %[[LOOP_3]] ], [ [[BC_RESUME_VAL6]], %[[SCALAR_PH]] ]
+; CHECK-NEXT:    [[PTR_IV_3:%.*]] = phi ptr [ [[PTR_IV_3_NEXT:%.*]], %[[LOOP_3]] ], [ [[BC_RESUME_VAL3]], %[[SCALAR_PH]] ]
 ; CHECK-NEXT:    [[TMP12:%.*]] = add i64 [[IV_2]], -1
 ; CHECK-NEXT:    [[GEP_SRC:%.*]] = getelementptr double, ptr [[SRC]], i64 [[TMP12]]
 ; CHECK-NEXT:    [[L:%.*]] = load i64, ptr [[GEP_SRC]], align 8
@@ -361,7 +340,7 @@ define void @expand_diff_neg_ptrtoint_expr(ptr %src, ptr %start) {
 ; CHECK-NEXT:    store i64 [[L]], ptr [[PTR_IV_3]], align 8
 ; CHECK-NEXT:    [[IV_NEXT_2]] = add i64 [[IV_2]], 1
 ; CHECK-NEXT:    [[EC_3:%.*]] = icmp eq i64 [[IV_NEXT_2]], 0
-; CHECK-NEXT:    br i1 [[EC_3]], label %[[EXIT]], label %[[LOOP_3]], !llvm.loop [[LOOP9:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC_3]], label %[[EXIT:.*]], label %[[LOOP_3]], !llvm.loop [[LOOP9:![0-9]+]]
 ; CHECK:       [[EXIT]]:
 ; CHECK-NEXT:    ret void
 ;
@@ -402,6 +381,93 @@ exit:
   ret void
 }
 
+
+declare i1 @cond()
+
+define void @scev_exp_reuse_const_add(ptr %dst, ptr %src) {
+; CHECK-LABEL: define void @scev_exp_reuse_const_add(
+; CHECK-SAME: ptr [[DST:%.*]], ptr [[SRC:%.*]]) {
+; CHECK-NEXT:  [[ENTRY:.*]]:
+; CHECK-NEXT:    [[SRC2:%.*]] = ptrtoint ptr [[SRC]] to i64
+; CHECK-NEXT:    [[DST1:%.*]] = ptrtoint ptr [[DST]] to i64
+; CHECK-NEXT:    br label %[[LOOP_1:.*]]
+; CHECK:       [[LOOP_1]]:
+; CHECK-NEXT:    [[INDVAR:%.*]] = phi i64 [ [[INDVAR_NEXT:%.*]], %[[LOOP_1]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[PTR_IV_1:%.*]] = phi ptr [ [[DST]], %[[ENTRY]] ], [ [[PTR_IV_1_NEXT:%.*]], %[[LOOP_1]] ]
+; CHECK-NEXT:    [[PTR_IV_1_NEXT]] = getelementptr i8, ptr [[PTR_IV_1]], i64 2
+; CHECK-NEXT:    [[C:%.*]] = call i1 @cond()
+; CHECK-NEXT:    [[INDVAR_NEXT]] = add i64 [[INDVAR]], 1
+; CHECK-NEXT:    br i1 [[C]], label %[[LOOP_2_PH:.*]], label %[[LOOP_1]]
+; CHECK:       [[LOOP_2_PH]]:
+; CHECK-NEXT:    [[INDVAR_LCSSA:%.*]] = phi i64 [ [[INDVAR]], %[[LOOP_1]] ]
+; CHECK-NEXT:    [[PTR_IV_1_NEXT_LCSSA:%.*]] = phi ptr [ [[PTR_IV_1_NEXT]], %[[LOOP_1]] ]
+; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_MEMCHECK:.*]]
+; CHECK:       [[VECTOR_MEMCHECK]]:
+; CHECK-NEXT:    [[TMP0:%.*]] = sub i64 [[DST1]], [[SRC2]]
+; CHECK-NEXT:    [[TMP1:%.*]] = shl i64 [[INDVAR_LCSSA]], 1
+; CHECK-NEXT:    [[TMP2:%.*]] = add i64 [[TMP1]], [[TMP0]]
+; CHECK-NEXT:    [[DIFF_CHECK:%.*]] = icmp ult i64 [[TMP2]], 4
+; CHECK-NEXT:    br i1 [[DIFF_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[PTR_IV_1_NEXT_LCSSA]], i64 80
+; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
+; CHECK:       [[VECTOR_BODY]]:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], 2
+; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[PTR_IV_1_NEXT_LCSSA]], i64 [[OFFSET_IDX]]
+; CHECK-NEXT:    [[TMP4:%.*]] = add i64 [[INDEX]], 1
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i16, ptr [[SRC]], i64 [[TMP4]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i16>, ptr [[TMP5]], align 2
+; CHECK-NEXT:    store <2 x i16> [[WIDE_LOAD]], ptr [[NEXT_GEP]], align 2
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
+; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 40
+; CHECK-NEXT:    br i1 [[TMP8]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; CHECK:       [[MIDDLE_BLOCK]]:
+; CHECK-NEXT:    br label %[[SCALAR_PH]]
+; CHECK:       [[SCALAR_PH]]:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 40, %[[MIDDLE_BLOCK]] ], [ 0, %[[LOOP_2_PH]] ], [ 0, %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL3:%.*]] = phi ptr [ [[TMP3]], %[[MIDDLE_BLOCK]] ], [ [[PTR_IV_1_NEXT_LCSSA]], %[[LOOP_2_PH]] ], [ [[PTR_IV_1_NEXT_LCSSA]], %[[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    br label %[[LOOP_2:.*]]
+; CHECK:       [[LOOP_2]]:
+; CHECK-NEXT:    [[IV_1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_2_NEXT:%.*]], %[[LOOP_2]] ]
+; CHECK-NEXT:    [[PTR_IV_2:%.*]] = phi ptr [ [[BC_RESUME_VAL3]], %[[SCALAR_PH]] ], [ [[PTR_IV_2_NEXT:%.*]], %[[LOOP_2]] ]
+; CHECK-NEXT:    [[IV_2_NEXT]] = add i64 [[IV_1]], 1
+; CHECK-NEXT:    [[GEP_SRC_1:%.*]] = getelementptr i16, ptr [[SRC]], i64 [[IV_2_NEXT]]
+; CHECK-NEXT:    [[L:%.*]] = load i16, ptr [[GEP_SRC_1]], align 2
+; CHECK-NEXT:    [[PTR_IV_2_NEXT]] = getelementptr i8, ptr [[PTR_IV_2]], i64 2
+; CHECK-NEXT:    store i16 [[L]], ptr [[PTR_IV_2]], align 2
+; CHECK-NEXT:    [[EC:%.*]] = icmp eq i64 [[IV_1]], 40
+; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT:.*]], label %[[LOOP_2]], !llvm.loop [[LOOP11:![0-9]+]]
+; CHECK:       [[EXIT]]:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br label %loop.1
+
+loop.1:
+  %ptr.iv.1 = phi ptr [ %dst, %entry ], [ %ptr.iv.1.next, %loop.1 ]
+  %ptr.iv.1.next = getelementptr i8, ptr %ptr.iv.1, i64 2
+  %c = call i1 @cond()
+  br i1 %c, label %loop.2.ph, label %loop.1
+
+loop.2.ph:
+  br label %loop.2
+
+loop.2:
+  %iv.1 = phi i64 [ 0, %loop.2.ph ], [ %iv.2.next, %loop.2 ]
+  %ptr.iv.2 = phi ptr [ %ptr.iv.1.next, %loop.2.ph ], [ %ptr.iv.2.next, %loop.2 ]
+  %iv.2.next = add i64 %iv.1, 1
+  %gep.src.1 = getelementptr i16, ptr %src, i64 %iv.2.next
+  %l = load i16, ptr %gep.src.1, align 2
+  %ptr.iv.2.next = getelementptr i8, ptr %ptr.iv.2, i64 2
+  store i16 %l, ptr %ptr.iv.2, align 2
+  %ec = icmp eq i64 %iv.1, 40
+  br i1 %ec, label %exit, label %loop.2
+
+exit:
+  ret void
+}
+
 ; Function Attrs: nocallback nofree nosync nounwind speculatable willreturn memory(none)
 declare double @llvm.cos.f64(double) #0
 
diff --git a/llvm/test/Transforms/LoopVectorize/reverse_induction.ll b/llvm/test/Transforms/LoopVectorize/reverse_induction.ll
index 8c32ce20..cea16c9 100644
--- a/llvm/test/Transforms/LoopVectorize/reverse_induction.ll
+++ b/llvm/test/Transforms/LoopVectorize/reverse_induction.ll
@@ -12,7 +12,6 @@ define i32 @reverse_induction_i64(i64 %startval, ptr %ptr) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
 ; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
-; CHECK-NEXT:    [[IND_END:%.*]] = sub i64 [[STARTVAL]], 1024
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
@@ -37,11 +36,11 @@ define i32 @reverse_induction_i64(i64 %startval, ptr %ptr) {
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[TMP11]], [[TMP10]]
 ; CHECK-NEXT:    [[TMP13:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
-; CHECK-NEXT:    br i1 true, label %[[LOOPEND:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[LOOPEND:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], %[[MIDDLE_BLOCK]] ], [ [[STARTVAL]], %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL4:%.*]] = phi i32 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP13]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[STARTVAL]], %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL4:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[ADD_I7:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[ADD_I:%.*]], %[[FOR_BODY]] ]
@@ -84,7 +83,6 @@ define i32 @reverse_induction_i128(i128 %startval, ptr %ptr) {
 ; CHECK-NEXT:  [[ENTRY:.*]]:
 ; CHECK-NEXT:    br i1 false, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
-; CHECK-NEXT:    [[IND_END:%.*]] = sub i128 [[STARTVAL]], 1024
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i128 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
@@ -109,11 +107,11 @@ define i32 @reverse_induction_i128(i128 %startval, ptr %ptr) {
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[TMP11]], [[TMP10]]
 ; CHECK-NEXT:    [[TMP13:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
-; CHECK-NEXT:    br i1 true, label %[[LOOPEND:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[LOOPEND:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i128 [ [[IND_END]], %[[MIDDLE_BLOCK]] ], [ [[STARTVAL]], %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL4:%.*]] = phi i32 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP13]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i128 [ [[STARTVAL]], %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL4:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[ADD_I7:%.*]] = phi i128 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[ADD_I:%.*]], %[[FOR_BODY]] ]
@@ -165,7 +163,6 @@ define i32 @reverse_induction_i16(i16 %startval, ptr %ptr) {
 ; CHECK-NEXT:    [[TMP3:%.*]] = or i1 [[TMP2]], [[MUL_OVERFLOW]]
 ; CHECK-NEXT:    br i1 [[TMP3]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
 ; CHECK:       [[VECTOR_PH]]:
-; CHECK-NEXT:    [[IND_END:%.*]] = sub i16 [[STARTVAL]], 1024
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
@@ -191,11 +188,11 @@ define i32 @reverse_induction_i16(i16 %startval, ptr %ptr) {
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[BIN_RDX:%.*]] = add <4 x i32> [[TMP15]], [[TMP14]]
 ; CHECK-NEXT:    [[TMP17:%.*]] = call i32 @llvm.vector.reduce.add.v4i32(<4 x i32> [[BIN_RDX]])
-; CHECK-NEXT:    br i1 true, label %[[LOOPEND:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[LOOPEND:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ [[IND_END]], %[[MIDDLE_BLOCK]] ], [ [[STARTVAL]], %[[ENTRY]] ], [ [[STARTVAL]], %[[VECTOR_SCEVCHECK]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL4:%.*]] = phi i32 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP17]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ [[STARTVAL]], %[[ENTRY]] ], [ [[STARTVAL]], %[[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL4:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ 0, %[[VECTOR_SCEVCHECK]] ]
 ; CHECK-NEXT:    br label %[[FOR_BODY:.*]]
 ; CHECK:       [[FOR_BODY]]:
 ; CHECK-NEXT:    [[ADD_I7:%.*]] = phi i16 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[ADD_I:%.*]], %[[FOR_BODY]] ]
@@ -278,10 +275,10 @@ define void @reverse_forward_induction_i64_i8() {
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[WHILE_END:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[WHILE_END:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ -1, %[[MIDDLE_BLOCK]] ], [ 1023, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i8 [ 0, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1023, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i8 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[WHILE_BODY:.*]]
 ; CHECK:       [[WHILE_BODY]]:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[WHILE_BODY]] ]
@@ -346,10 +343,10 @@ define void @reverse_forward_induction_i64_i8_signed() {
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP12]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[WHILE_END:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[WHILE_END:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ -1, %[[MIDDLE_BLOCK]] ], [ 1023, %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i8 [ -127, %[[MIDDLE_BLOCK]] ], [ -127, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1023, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi i8 [ -127, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[WHILE_BODY:.*]]
 ; CHECK:       [[WHILE_BODY]]:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], %[[WHILE_BODY]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/runtime-check-known-true.ll b/llvm/test/Transforms/LoopVectorize/runtime-check-known-true.ll
new file mode 100644
index 0000000..830bd92
--- /dev/null
+++ b/llvm/test/Transforms/LoopVectorize/runtime-check-known-true.ll
@@ -0,0 +1,92 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --check-globals none --version 5
+; RUN: opt -p loop-vectorize -force-vector-width=4 -S %s | FileCheck %s
+
+define void @test_runtime_check_known_false_after_construction(ptr %start.1, ptr %start.2, ptr %end) {
+; CHECK-LABEL: define void @test_runtime_check_known_false_after_construction(
+; CHECK-SAME: ptr [[START_1:%.*]], ptr [[START_2:%.*]], ptr [[END:%.*]]) {
+; CHECK-NEXT:  [[ENTRY:.*]]:
+; CHECK-NEXT:    [[END1:%.*]] = ptrtoint ptr [[END]] to i64
+; CHECK-NEXT:    [[GEP_START_2:%.*]] = getelementptr i8, ptr [[START_2]], i64 8
+; CHECK-NEXT:    [[START_1_INT:%.*]] = ptrtoint ptr [[START_1]] to i64
+; CHECK-NEXT:    [[START_2_INT:%.*]] = ptrtoint ptr [[GEP_START_2]] to i64
+; CHECK-NEXT:    [[DIFF:%.*]] = sub i64 [[START_1_INT]], [[START_2_INT]]
+; CHECK-NEXT:    [[START_2_DIFF:%.*]] = getelementptr i8, ptr [[START_2]], i64 [[DIFF]]
+; CHECK-NEXT:    [[TMP0:%.*]] = add i64 [[END1]], 8
+; CHECK-NEXT:    [[TMP1:%.*]] = sub i64 [[TMP0]], [[START_1_INT]]
+; CHECK-NEXT:    [[TMP2:%.*]] = mul i64 [[TMP1]], 2305843009213693951
+; CHECK-NEXT:    [[TMP3:%.*]] = lshr i64 [[TMP2]], 3
+; CHECK-NEXT:    [[TMP4:%.*]] = add nuw nsw i64 [[TMP3]], 1
+; CHECK-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[TMP4]], 4
+; CHECK-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_SCEVCHECK:.*]]
+; CHECK:       [[VECTOR_SCEVCHECK]]:
+; CHECK-NEXT:    [[TMP5:%.*]] = trunc i64 [[END1]] to i3
+; CHECK-NEXT:    [[TMP6:%.*]] = trunc i64 [[START_1_INT]] to i3
+; CHECK-NEXT:    [[TMP7:%.*]] = sub i3 [[TMP5]], [[TMP6]]
+; CHECK-NEXT:    [[TMP8:%.*]] = zext i3 [[TMP7]] to i64
+; CHECK-NEXT:    [[IDENT_CHECK:%.*]] = icmp ne i64 [[TMP8]], 0
+; CHECK-NEXT:    br i1 [[IDENT_CHECK]], label %[[SCALAR_PH]], label %[[VECTOR_PH:.*]]
+; CHECK:       [[VECTOR_PH]]:
+; CHECK-NEXT:    [[N_MOD_VF:%.*]] = urem i64 [[TMP4]], 4
+; CHECK-NEXT:    [[N_VEC:%.*]] = sub i64 [[TMP4]], [[N_MOD_VF]]
+; CHECK-NEXT:    [[TMP9:%.*]] = mul i64 [[N_VEC]], -8
+; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr i8, ptr [[START_1]], i64 [[TMP9]]
+; CHECK-NEXT:    [[TMP11:%.*]] = mul i64 [[N_VEC]], -8
+; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i8, ptr [[START_2_DIFF]], i64 [[TMP11]]
+; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
+; CHECK:       [[VECTOR_BODY]]:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], -8
+; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[START_1]], i64 [[OFFSET_IDX]]
+; CHECK-NEXT:    [[OFFSET_IDX2:%.*]] = mul i64 [[INDEX]], -8
+; CHECK-NEXT:    [[NEXT_GEP3:%.*]] = getelementptr i8, ptr [[START_2_DIFF]], i64 [[OFFSET_IDX2]]
+; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr i64, ptr [[NEXT_GEP3]], i32 0
+; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr i64, ptr [[TMP13]], i32 -3
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP14]], align 8
+; CHECK-NEXT:    [[REVERSE:%.*]] = shufflevector <4 x i64> [[WIDE_LOAD]], <4 x i64> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr i64, ptr [[NEXT_GEP]], i32 0
+; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr i64, ptr [[TMP15]], i32 -3
+; CHECK-NEXT:    [[REVERSE4:%.*]] = shufflevector <4 x i64> [[REVERSE]], <4 x i64> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; CHECK-NEXT:    store <4 x i64> [[REVERSE4]], ptr [[TMP16]], align 8
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP17]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK:       [[MIDDLE_BLOCK]]:
+; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP4]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK:       [[SCALAR_PH]]:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi ptr [ [[TMP10]], %[[MIDDLE_BLOCK]] ], [ [[START_1]], %[[ENTRY]] ], [ [[START_1]], %[[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL5:%.*]] = phi ptr [ [[TMP12]], %[[MIDDLE_BLOCK]] ], [ [[START_2_DIFF]], %[[ENTRY]] ], [ [[START_2_DIFF]], %[[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT:    br label %[[LOOP:.*]]
+; CHECK:       [[LOOP]]:
+; CHECK-NEXT:    [[PTR_IV_1:%.*]] = phi ptr [ [[PTR_IV_1_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
+; CHECK-NEXT:    [[PTR_IV_2:%.*]] = phi ptr [ [[PTR_IV_2_NEXT:%.*]], %[[LOOP]] ], [ [[BC_RESUME_VAL5]], %[[SCALAR_PH]] ]
+; CHECK-NEXT:    [[PTR_IV_2_NEXT]] = getelementptr i8, ptr [[PTR_IV_2]], i64 -8
+; CHECK-NEXT:    [[PTR_IV_1_NEXT]] = getelementptr i8, ptr [[PTR_IV_1]], i64 -8
+; CHECK-NEXT:    [[L:%.*]] = load i64, ptr [[PTR_IV_2]], align 8
+; CHECK-NEXT:    store i64 [[L]], ptr [[PTR_IV_1]], align 8
+; CHECK-NEXT:    [[EC:%.*]] = icmp eq ptr [[PTR_IV_2]], [[END]]
+; CHECK-NEXT:    br i1 [[EC]], label %[[EXIT]], label %[[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK:       [[EXIT]]:
+; CHECK-NEXT:    ret void
+;
+entry:
+  %gep.start.2 = getelementptr i8, ptr %start.2, i64 8
+  %start.1.int = ptrtoint ptr %start.1 to i64
+  %start.2.int = ptrtoint ptr %gep.start.2 to i64
+  %diff = sub i64 %start.1.int, %start.2.int
+  %start.2.diff = getelementptr i8, ptr %start.2, i64 %diff
+  br label %loop
+
+loop:
+  %ptr.iv.1 = phi ptr [ %ptr.iv.1.next, %loop ], [ %start.1, %entry ]
+  %ptr.iv.2 = phi ptr [ %ptr.iv.2.next, %loop ], [ %start.2.diff, %entry ]
+  %ptr.iv.2.next = getelementptr i8, ptr %ptr.iv.2, i64 -8
+  %ptr.iv.1.next = getelementptr i8, ptr %ptr.iv.1, i64 -8
+  %l = load i64, ptr %ptr.iv.2, align 8
+  store i64 %l, ptr %ptr.iv.1, align 8
+  %ec = icmp eq ptr %ptr.iv.2, %end
+  br i1 %ec, label %exit, label %loop
+
+exit:
+  ret void
+}
diff --git a/llvm/test/Transforms/LoopVectorize/runtime-check-needed-but-empty.ll b/llvm/test/Transforms/LoopVectorize/runtime-check-needed-but-empty.ll
index db88eaa..44d2925 100644
--- a/llvm/test/Transforms/LoopVectorize/runtime-check-needed-but-empty.ll
+++ b/llvm/test/Transforms/LoopVectorize/runtime-check-needed-but-empty.ll
@@ -17,12 +17,10 @@ define void @test(ptr %A, i32 %x) {
 ; CHECK-NEXT:    [[TMP4:%.*]] = trunc i64 [[TMP3]] to i32
 ; CHECK-NEXT:    [[TMP5:%.*]] = zext i32 [[TMP4]] to i64
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[A:%.*]], i64 [[TMP5]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP7]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = zext i32 [[TMP0]] to i64
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[TMP8]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds float, ptr [[TMP9]], i32 0
-; CHECK-NEXT:    store <4 x float> [[WIDE_LOAD]], ptr [[TMP10]], align 4
+; CHECK-NEXT:    store <4 x float> [[WIDE_LOAD]], ptr [[TMP9]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[OFFSET_IDX]], 4
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], undef
 ; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/runtime-check-small-clamped-bounds.ll b/llvm/test/Transforms/LoopVectorize/runtime-check-small-clamped-bounds.ll
index a47037c4..1035642 100644
--- a/llvm/test/Transforms/LoopVectorize/runtime-check-small-clamped-bounds.ll
+++ b/llvm/test/Transforms/LoopVectorize/runtime-check-small-clamped-bounds.ll
@@ -33,12 +33,10 @@ define void @load_clamped_index(ptr %A, ptr %B, i32 %N) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = urem i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[TMP4]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = add <4 x i32> [[WIDE_LOAD]], splat (i32 10)
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[B]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP7]], ptr [[TMP9]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP7]], ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -105,12 +103,10 @@ define void @store_clamped_index(ptr %A, ptr %B, i32 %N) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = urem i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[B]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = add <4 x i32> [[WIDE_LOAD]], splat (i32 10)
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[TMP4]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP7]], ptr [[TMP9]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP7]], ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -185,12 +181,10 @@ define void @load_clamped_index_offset_1(ptr %A, ptr %B, i32 %N) {
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i32 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP11:%.*]] = urem i32 [[OFFSET_IDX]], 4
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[A]], i32 [[TMP11]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP13]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP12]], align 4
 ; CHECK-NEXT:    [[TMP14:%.*]] = add <4 x i32> [[WIDE_LOAD]], splat (i32 10)
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[B]], i32 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[TMP15]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP14]], ptr [[TMP16]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP14]], ptr [[TMP15]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
@@ -369,10 +363,9 @@ define void @clamped_index_equal_dependence(ptr %A, ptr %B, i32 %N) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP3:%.*]] = urem i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i32 [[TMP3]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = add <4 x i32> [[WIDE_LOAD]], splat (i32 10)
-; CHECK-NEXT:    store <4 x i32> [[TMP6]], ptr [[TMP5]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP6]], ptr [[TMP4]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/runtime-check.ll b/llvm/test/Transforms/LoopVectorize/runtime-check.ll
index 22d9a53..c5838fe 100644
--- a/llvm/test/Transforms/LoopVectorize/runtime-check.ll
+++ b/llvm/test/Transforms/LoopVectorize/runtime-check.ll
@@ -429,7 +429,7 @@ define dso_local void @forced_optsize(ptr noalias nocapture readonly %x_p, ptr n
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128
 ; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP35:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_COND_CLEANUP:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_COND_CLEANUP:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.cond.cleanup:
diff --git a/llvm/test/Transforms/LoopVectorize/runtime-checks-difference-simplifications.ll b/llvm/test/Transforms/LoopVectorize/runtime-checks-difference-simplifications.ll
index 6bf47c5..1d64499 100644
--- a/llvm/test/Transforms/LoopVectorize/runtime-checks-difference-simplifications.ll
+++ b/llvm/test/Transforms/LoopVectorize/runtime-checks-difference-simplifications.ll
@@ -60,52 +60,40 @@ define void @test_large_number_of_group(ptr %dst, i64 %off, i64 %N) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP11:%.*]] = add nsw i64 [[INDEX]], -5
 ; CHECK-NEXT:    [[TMP12:%.*]] = add i64 [[TMP11]], [[OFF]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr i64, ptr [[DST:%.*]], i64 [[TMP12]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr double, ptr [[TMP13]], i32 0
+; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr i64, ptr [[DST:%.*]], i64 [[TMP12]]
 ; CHECK-NEXT:    store <4 x double> zeroinitializer, ptr [[TMP14]], align 8
 ; CHECK-NEXT:    [[TMP15:%.*]] = add i64 [[TMP11]], [[OFF_MUL_2]]
-; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP15]]
-; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr double, ptr [[TMP16]], i32 0
+; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP15]]
 ; CHECK-NEXT:    store <4 x double> zeroinitializer, ptr [[TMP17]], align 8
 ; CHECK-NEXT:    [[TMP18:%.*]] = add i64 [[TMP11]], [[OFF_MUL_3]]
-; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP18]]
-; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr double, ptr [[TMP19]], i32 0
+; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP18]]
 ; CHECK-NEXT:    store <4 x double> zeroinitializer, ptr [[TMP20]], align 8
 ; CHECK-NEXT:    [[TMP21:%.*]] = add i64 [[TMP11]], [[OFF_MUL_4]]
-; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP21]]
-; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr double, ptr [[TMP22]], i32 0
+; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP21]]
 ; CHECK-NEXT:    store <4 x double> zeroinitializer, ptr [[TMP23]], align 8
 ; CHECK-NEXT:    [[TMP24:%.*]] = add i64 [[TMP11]], [[OFF_MUL_5]]
-; CHECK-NEXT:    [[TMP25:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP24]]
-; CHECK-NEXT:    [[TMP26:%.*]] = getelementptr double, ptr [[TMP25]], i32 0
+; CHECK-NEXT:    [[TMP26:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP24]]
 ; CHECK-NEXT:    store <4 x double> zeroinitializer, ptr [[TMP26]], align 8
 ; CHECK-NEXT:    [[TMP27:%.*]] = add i64 [[TMP11]], [[OFF_MUL_6]]
-; CHECK-NEXT:    [[TMP28:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP27]]
-; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr double, ptr [[TMP28]], i32 0
+; CHECK-NEXT:    [[TMP29:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP27]]
 ; CHECK-NEXT:    store <4 x double> zeroinitializer, ptr [[TMP29]], align 8
 ; CHECK-NEXT:    [[TMP30:%.*]] = add i64 [[TMP11]], [[OFF_MUL_7]]
-; CHECK-NEXT:    [[TMP31:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP30]]
-; CHECK-NEXT:    [[TMP32:%.*]] = getelementptr double, ptr [[TMP31]], i32 0
+; CHECK-NEXT:    [[TMP32:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP30]]
 ; CHECK-NEXT:    store <4 x double> zeroinitializer, ptr [[TMP32]], align 8
 ; CHECK-NEXT:    [[TMP33:%.*]] = add i64 [[TMP11]], [[OFF_MUL_8]]
-; CHECK-NEXT:    [[TMP34:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP33]]
-; CHECK-NEXT:    [[TMP35:%.*]] = getelementptr double, ptr [[TMP34]], i32 0
+; CHECK-NEXT:    [[TMP35:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP33]]
 ; CHECK-NEXT:    store <4 x double> zeroinitializer, ptr [[TMP35]], align 8
 ; CHECK-NEXT:    [[TMP36:%.*]] = add i64 [[TMP11]], [[OFF_MUL_9]]
-; CHECK-NEXT:    [[TMP37:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP36]]
-; CHECK-NEXT:    [[TMP38:%.*]] = getelementptr double, ptr [[TMP37]], i32 0
+; CHECK-NEXT:    [[TMP38:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP36]]
 ; CHECK-NEXT:    store <4 x double> zeroinitializer, ptr [[TMP38]], align 8
 ; CHECK-NEXT:    [[TMP39:%.*]] = add i64 [[TMP11]], [[OFF_MUL_10]]
-; CHECK-NEXT:    [[TMP40:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP39]]
-; CHECK-NEXT:    [[TMP41:%.*]] = getelementptr double, ptr [[TMP40]], i32 0
+; CHECK-NEXT:    [[TMP41:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP39]]
 ; CHECK-NEXT:    store <4 x double> zeroinitializer, ptr [[TMP41]], align 8
 ; CHECK-NEXT:    [[TMP42:%.*]] = add i64 [[TMP11]], [[OFF_MUL_11]]
-; CHECK-NEXT:    [[TMP43:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP42]]
-; CHECK-NEXT:    [[TMP44:%.*]] = getelementptr double, ptr [[TMP43]], i32 0
+; CHECK-NEXT:    [[TMP44:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP42]]
 ; CHECK-NEXT:    store <4 x double> zeroinitializer, ptr [[TMP44]], align 8
 ; CHECK-NEXT:    [[TMP45:%.*]] = add i64 [[TMP11]], [[OFF_MUL_12]]
-; CHECK-NEXT:    [[TMP46:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP45]]
-; CHECK-NEXT:    [[TMP47:%.*]] = getelementptr double, ptr [[TMP46]], i32 0
+; CHECK-NEXT:    [[TMP47:%.*]] = getelementptr i64, ptr [[DST]], i64 [[TMP45]]
 ; CHECK-NEXT:    store <4 x double> zeroinitializer, ptr [[TMP47]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP48:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -242,20 +230,17 @@ define void @check_creation_order(ptr %a, ptr %b, i32 %m) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr double, ptr [[INVARIANT_GEP]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr double, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP4]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x double>, ptr [[TMP3]], align 8
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds double, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x double>, ptr [[TMP6]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x double>, ptr [[TMP5]], align 8
 ; CHECK-NEXT:    [[TMP7:%.*]] = fadd <4 x double> [[WIDE_LOAD]], [[WIDE_LOAD4]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds double, ptr [[A]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds double, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <4 x double> [[TMP7]], ptr [[TMP9]], align 8
+; CHECK-NEXT:    store <4 x double> [[TMP7]], ptr [[TMP8]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 31996
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 31996, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
@@ -270,7 +255,7 @@ define void @check_creation_order(ptr %a, ptr %b, i32 %m) {
 ; CHECK-NEXT:    store double [[ADD3]], ptr [[GEP_A]], align 8
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i64 [[IV]], 1
 ; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[IV_NEXT]], 31999
-; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EXITCOND_NOT]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/runtime-checks-hoist.ll b/llvm/test/Transforms/LoopVectorize/runtime-checks-hoist.ll
index cdf04cd..2590ccb 100644
--- a/llvm/test/Transforms/LoopVectorize/runtime-checks-hoist.ll
+++ b/llvm/test/Transforms/LoopVectorize/runtime-checks-hoist.ll
@@ -69,15 +69,13 @@ define void @diff_checks(ptr nocapture noundef writeonly %dst, ptr nocapture nou
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP13:%.*]] = add nuw nsw i64 [[INDEX]], [[TMP10]]
 ; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP13]]
-; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[TMP14]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP15]], align 4, !alias.scope [[META0:![0-9]+]]
-; CHECK-NEXT:    [[TMP16:%.*]] = add nsw i64 [[INDEX]], [[TMP11]]
-; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP16]]
-; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[TMP17]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[WIDE_LOAD]], ptr [[TMP18]], align 4, !alias.scope [[META3:![0-9]+]], !noalias [[META0]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP14]], align 4, !alias.scope [[META0:![0-9]+]]
+; CHECK-NEXT:    [[TMP15:%.*]] = add nsw i64 [[INDEX]], [[TMP11]]
+; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP15]]
+; CHECK-NEXT:    store <4 x i32> [[WIDE_LOAD]], ptr [[TMP16]], align 4, !alias.scope [[META3:![0-9]+]], !noalias [[META0]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[WIDE_N]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[INNER_EXIT]], label [[SCALAR_PH]]
@@ -86,12 +84,12 @@ define void @diff_checks(ptr nocapture noundef writeonly %dst, ptr nocapture nou
 ; CHECK-NEXT:    br label [[INNER_LOOP:%.*]]
 ; CHECK:       inner.loop:
 ; CHECK-NEXT:    [[IV_INNER:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_INNER_NEXT:%.*]], [[INNER_LOOP]] ]
-; CHECK-NEXT:    [[TMP20:%.*]] = add nuw nsw i64 [[IV_INNER]], [[TMP10]]
-; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP20]]
-; CHECK-NEXT:    [[TMP21:%.*]] = load i32, ptr [[ARRAYIDX_US]], align 4
-; CHECK-NEXT:    [[TMP22:%.*]] = add nsw i64 [[IV_INNER]], [[TMP11]]
-; CHECK-NEXT:    [[ARRAYIDX9_US:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP22]]
-; CHECK-NEXT:    store i32 [[TMP21]], ptr [[ARRAYIDX9_US]], align 4
+; CHECK-NEXT:    [[TMP18:%.*]] = add nuw nsw i64 [[IV_INNER]], [[TMP10]]
+; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP18]]
+; CHECK-NEXT:    [[TMP19:%.*]] = load i32, ptr [[ARRAYIDX_US]], align 4
+; CHECK-NEXT:    [[TMP20:%.*]] = add nsw i64 [[IV_INNER]], [[TMP11]]
+; CHECK-NEXT:    [[ARRAYIDX9_US:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP20]]
+; CHECK-NEXT:    store i32 [[TMP19]], ptr [[ARRAYIDX9_US]], align 4
 ; CHECK-NEXT:    [[IV_INNER_NEXT]] = add nuw nsw i64 [[IV_INNER]], 1
 ; CHECK-NEXT:    [[INNER_EXIT_COND:%.*]] = icmp eq i64 [[IV_INNER_NEXT]], [[WIDE_N]]
 ; CHECK-NEXT:    br i1 [[INNER_EXIT_COND]], label [[INNER_EXIT]], label [[INNER_LOOP]], !llvm.loop [[LOOP8:![0-9]+]]
@@ -189,16 +187,14 @@ define void @full_checks(ptr nocapture noundef %dst, ptr nocapture noundef reado
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = add nuw nsw i64 [[INDEX]], [[TMP3]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP4]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4, !alias.scope [[META9:![0-9]+]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP4]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP8]], align 4, !alias.scope [[META12:![0-9]+]], !noalias [[META9]]
-; CHECK-NEXT:    [[TMP9:%.*]] = add nsw <4 x i32> [[WIDE_LOAD2]], [[WIDE_LOAD]]
-; CHECK-NEXT:    store <4 x i32> [[TMP9]], ptr [[TMP8]], align 4, !alias.scope [[META12]], !noalias [[META9]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4, !alias.scope [[META9:![0-9]+]]
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP4]]
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4, !alias.scope [[META12:![0-9]+]], !noalias [[META9]]
+; CHECK-NEXT:    [[TMP7:%.*]] = add nsw <4 x i32> [[WIDE_LOAD2]], [[WIDE_LOAD]]
+; CHECK-NEXT:    store <4 x i32> [[TMP7]], ptr [[TMP6]], align 4, !alias.scope [[META12]], !noalias [[META9]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
+; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[WIDE_N]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[INNER_EXIT]], label [[SCALAR_PH]]
@@ -207,12 +203,12 @@ define void @full_checks(ptr nocapture noundef %dst, ptr nocapture noundef reado
 ; CHECK-NEXT:    br label [[INNER_LOOP:%.*]]
 ; CHECK:       inner.loop:
 ; CHECK-NEXT:    [[IV_INNER:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_INNER_NEXT:%.*]], [[INNER_LOOP]] ]
-; CHECK-NEXT:    [[TMP11:%.*]] = add nuw nsw i64 [[IV_INNER]], [[TMP3]]
-; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP11]]
-; CHECK-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX_US]], align 4
-; CHECK-NEXT:    [[ARRAYIDX8_US:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP11]]
-; CHECK-NEXT:    [[TMP13:%.*]] = load i32, ptr [[ARRAYIDX8_US]], align 4
-; CHECK-NEXT:    [[ADD9_US:%.*]] = add nsw i32 [[TMP13]], [[TMP12]]
+; CHECK-NEXT:    [[TMP9:%.*]] = add nuw nsw i64 [[IV_INNER]], [[TMP3]]
+; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP9]]
+; CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[ARRAYIDX_US]], align 4
+; CHECK-NEXT:    [[ARRAYIDX8_US:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP9]]
+; CHECK-NEXT:    [[TMP11:%.*]] = load i32, ptr [[ARRAYIDX8_US]], align 4
+; CHECK-NEXT:    [[ADD9_US:%.*]] = add nsw i32 [[TMP11]], [[TMP10]]
 ; CHECK-NEXT:    store i32 [[ADD9_US]], ptr [[ARRAYIDX8_US]], align 4
 ; CHECK-NEXT:    [[IV_INNER_NEXT]] = add nuw nsw i64 [[IV_INNER]], 1
 ; CHECK-NEXT:    [[INNER_EXIT_COND:%.*]] = icmp eq i64 [[IV_INNER_NEXT]], [[WIDE_N]]
@@ -319,17 +315,15 @@ define void @full_checks_diff_strides(ptr nocapture noundef %dst, ptr nocapture
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP9:%.*]] = add nuw nsw i64 [[INDEX]], [[TMP7]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP9]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP11]], align 4, !alias.scope [[META16:![0-9]+]]
-; CHECK-NEXT:    [[TMP12:%.*]] = add nuw nsw i64 [[INDEX]], [[TMP8]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP12]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP14]], align 4, !alias.scope [[META19:![0-9]+]], !noalias [[META16]]
-; CHECK-NEXT:    [[TMP15:%.*]] = add nsw <4 x i32> [[WIDE_LOAD2]], [[WIDE_LOAD]]
-; CHECK-NEXT:    store <4 x i32> [[TMP15]], ptr [[TMP14]], align 4, !alias.scope [[META19]], !noalias [[META16]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP10]], align 4, !alias.scope [[META16:![0-9]+]]
+; CHECK-NEXT:    [[TMP11:%.*]] = add nuw nsw i64 [[INDEX]], [[TMP8]]
+; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP11]]
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP12]], align 4, !alias.scope [[META19:![0-9]+]], !noalias [[META16]]
+; CHECK-NEXT:    [[TMP13:%.*]] = add nsw <4 x i32> [[WIDE_LOAD2]], [[WIDE_LOAD]]
+; CHECK-NEXT:    store <4 x i32> [[TMP13]], ptr [[TMP12]], align 4, !alias.scope [[META19]], !noalias [[META16]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
+; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[WIDE_N]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[INNER_EXIT]], label [[SCALAR_PH]]
@@ -338,13 +332,13 @@ define void @full_checks_diff_strides(ptr nocapture noundef %dst, ptr nocapture
 ; CHECK-NEXT:    br label [[INNER_LOOP:%.*]]
 ; CHECK:       inner.loop:
 ; CHECK-NEXT:    [[IV_INNER:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_INNER_NEXT:%.*]], [[INNER_LOOP]] ]
-; CHECK-NEXT:    [[TMP17:%.*]] = add nuw nsw i64 [[IV_INNER]], [[TMP7]]
-; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP17]]
-; CHECK-NEXT:    [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX_US]], align 4
-; CHECK-NEXT:    [[TMP19:%.*]] = add nuw nsw i64 [[IV_INNER]], [[TMP8]]
-; CHECK-NEXT:    [[ARRAYIDX8_US:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP19]]
-; CHECK-NEXT:    [[TMP20:%.*]] = load i32, ptr [[ARRAYIDX8_US]], align 4
-; CHECK-NEXT:    [[ADD9_US:%.*]] = add nsw i32 [[TMP20]], [[TMP18]]
+; CHECK-NEXT:    [[TMP15:%.*]] = add nuw nsw i64 [[IV_INNER]], [[TMP7]]
+; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP15]]
+; CHECK-NEXT:    [[TMP16:%.*]] = load i32, ptr [[ARRAYIDX_US]], align 4
+; CHECK-NEXT:    [[TMP17:%.*]] = add nuw nsw i64 [[IV_INNER]], [[TMP8]]
+; CHECK-NEXT:    [[ARRAYIDX8_US:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP17]]
+; CHECK-NEXT:    [[TMP18:%.*]] = load i32, ptr [[ARRAYIDX8_US]], align 4
+; CHECK-NEXT:    [[ADD9_US:%.*]] = add nsw i32 [[TMP18]], [[TMP16]]
 ; CHECK-NEXT:    store i32 [[ADD9_US]], ptr [[ARRAYIDX8_US]], align 4
 ; CHECK-NEXT:    [[IV_INNER_NEXT]] = add nuw nsw i64 [[IV_INNER]], 1
 ; CHECK-NEXT:    [[INNER_EXIT_COND:%.*]] = icmp eq i64 [[IV_INNER_NEXT]], [[WIDE_N]]
@@ -435,15 +429,13 @@ define void @diff_checks_src_start_invariant(ptr nocapture noundef writeonly %ds
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP7]], align 4
-; CHECK-NEXT:    [[TMP8:%.*]] = add nuw nsw i64 [[INDEX]], [[TMP5]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP8]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[WIDE_LOAD]], ptr [[TMP10]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4
+; CHECK-NEXT:    [[TMP7:%.*]] = add nuw nsw i64 [[INDEX]], [[TMP5]]
+; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP7]]
+; CHECK-NEXT:    store <4 x i32> [[WIDE_LOAD]], ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
+; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP23:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[WIDE_N]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[INNER_LOOP_EXIT]], label [[SCALAR_PH]]
@@ -453,10 +445,10 @@ define void @diff_checks_src_start_invariant(ptr nocapture noundef writeonly %ds
 ; CHECK:       inner.loop:
 ; CHECK-NEXT:    [[IV_INNER:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_INNER_NEXT:%.*]], [[INNER_LOOP]] ]
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[IV_INNER]]
-; CHECK-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX_US]], align 4
-; CHECK-NEXT:    [[TMP13:%.*]] = add nuw nsw i64 [[IV_INNER]], [[TMP5]]
-; CHECK-NEXT:    [[ARRAYIDX6_US:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP13]]
-; CHECK-NEXT:    store i32 [[TMP12]], ptr [[ARRAYIDX6_US]], align 4
+; CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[ARRAYIDX_US]], align 4
+; CHECK-NEXT:    [[TMP11:%.*]] = add nuw nsw i64 [[IV_INNER]], [[TMP5]]
+; CHECK-NEXT:    [[ARRAYIDX6_US:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP11]]
+; CHECK-NEXT:    store i32 [[TMP10]], ptr [[ARRAYIDX6_US]], align 4
 ; CHECK-NEXT:    [[IV_INNER_NEXT]] = add nuw nsw i64 [[IV_INNER]], 1
 ; CHECK-NEXT:    [[INNER_EXIT_COND:%.*]] = icmp eq i64 [[IV_INNER_NEXT]], [[WIDE_N]]
 ; CHECK-NEXT:    br i1 [[INNER_EXIT_COND]], label [[INNER_LOOP_EXIT]], label [[INNER_LOOP]], !llvm.loop [[LOOP24:![0-9]+]]
@@ -546,17 +538,15 @@ define void @full_checks_src_start_invariant(ptr nocapture noundef %dst, ptr noc
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4, !alias.scope [[META25:![0-9]+]]
-; CHECK-NEXT:    [[TMP7:%.*]] = add nuw nsw i64 [[INDEX]], [[TMP4]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP7]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP9]], align 4, !alias.scope [[META28:![0-9]+]], !noalias [[META25]]
-; CHECK-NEXT:    [[TMP10:%.*]] = add nsw <4 x i32> [[WIDE_LOAD2]], [[WIDE_LOAD]]
-; CHECK-NEXT:    store <4 x i32> [[TMP10]], ptr [[TMP9]], align 4, !alias.scope [[META28]], !noalias [[META25]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4, !alias.scope [[META25:![0-9]+]]
+; CHECK-NEXT:    [[TMP6:%.*]] = add nuw nsw i64 [[INDEX]], [[TMP4]]
+; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP6]]
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP7]], align 4, !alias.scope [[META28:![0-9]+]], !noalias [[META25]]
+; CHECK-NEXT:    [[TMP8:%.*]] = add nsw <4 x i32> [[WIDE_LOAD2]], [[WIDE_LOAD]]
+; CHECK-NEXT:    store <4 x i32> [[TMP8]], ptr [[TMP7]], align 4, !alias.scope [[META28]], !noalias [[META25]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
+; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[WIDE_N]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[INNER_LOOP_EXIT]], label [[SCALAR_PH]]
@@ -566,11 +556,11 @@ define void @full_checks_src_start_invariant(ptr nocapture noundef %dst, ptr noc
 ; CHECK:       inner.loop:
 ; CHECK-NEXT:    [[IV_INNER:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_INNER_NEXT:%.*]], [[INNER_LOOP]] ]
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[IV_INNER]]
-; CHECK-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX_US]], align 4
-; CHECK-NEXT:    [[TMP13:%.*]] = add nuw nsw i64 [[IV_INNER]], [[TMP4]]
-; CHECK-NEXT:    [[ARRAYIDX6_US:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP13]]
-; CHECK-NEXT:    [[TMP14:%.*]] = load i32, ptr [[ARRAYIDX6_US]], align 4
-; CHECK-NEXT:    [[ADD7_US:%.*]] = add nsw i32 [[TMP14]], [[TMP12]]
+; CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[ARRAYIDX_US]], align 4
+; CHECK-NEXT:    [[TMP11:%.*]] = add nuw nsw i64 [[IV_INNER]], [[TMP4]]
+; CHECK-NEXT:    [[ARRAYIDX6_US:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP11]]
+; CHECK-NEXT:    [[TMP12:%.*]] = load i32, ptr [[ARRAYIDX6_US]], align 4
+; CHECK-NEXT:    [[ADD7_US:%.*]] = add nsw i32 [[TMP12]], [[TMP10]]
 ; CHECK-NEXT:    store i32 [[ADD7_US]], ptr [[ARRAYIDX6_US]], align 4
 ; CHECK-NEXT:    [[IV_INNER_NEXT]] = add nuw nsw i64 [[IV_INNER]], 1
 ; CHECK-NEXT:    [[INNER_EXIT_COND:%.*]] = icmp eq i64 [[IV_INNER_NEXT]], [[WIDE_N]]
@@ -690,17 +680,15 @@ define void @triple_nested_loop_mixed_access(ptr nocapture noundef %dst, ptr noc
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP19:%.*]] = add nuw nsw i64 [[INDEX]], [[TMP15]]
 ; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP19]]
-; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, ptr [[TMP20]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP21]], align 4, !alias.scope [[META32:![0-9]+]]
-; CHECK-NEXT:    [[TMP22:%.*]] = add nuw nsw i64 [[TMP18]], [[INDEX]]
-; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP22]]
-; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, ptr [[TMP23]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i32>, ptr [[TMP24]], align 4, !alias.scope [[META35:![0-9]+]], !noalias [[META32]]
-; CHECK-NEXT:    [[TMP25:%.*]] = add nsw <4 x i32> [[WIDE_LOAD4]], [[WIDE_LOAD]]
-; CHECK-NEXT:    store <4 x i32> [[TMP25]], ptr [[TMP24]], align 4, !alias.scope [[META35]], !noalias [[META32]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP20]], align 4, !alias.scope [[META32:![0-9]+]]
+; CHECK-NEXT:    [[TMP21:%.*]] = add nuw nsw i64 [[TMP18]], [[INDEX]]
+; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP21]]
+; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i32>, ptr [[TMP22]], align 4, !alias.scope [[META35:![0-9]+]], !noalias [[META32]]
+; CHECK-NEXT:    [[TMP23:%.*]] = add nsw <4 x i32> [[WIDE_LOAD4]], [[WIDE_LOAD]]
+; CHECK-NEXT:    store <4 x i32> [[TMP23]], ptr [[TMP22]], align 4, !alias.scope [[META35]], !noalias [[META32]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP37:![0-9]+]]
+; CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP37:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[INNER_LOOP_END]], label [[SCALAR_PH]]
@@ -709,13 +697,13 @@ define void @triple_nested_loop_mixed_access(ptr nocapture noundef %dst, ptr noc
 ; CHECK-NEXT:    br label [[INNER_LOOP:%.*]]
 ; CHECK:       inner.loop:
 ; CHECK-NEXT:    [[INNER_IV:%.*]] = phi i64 [ [[INNER_IV_NEXT:%.*]], [[INNER_LOOP]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
-; CHECK-NEXT:    [[TMP27:%.*]] = add nuw nsw i64 [[INNER_IV]], [[TMP15]]
-; CHECK-NEXT:    [[ARRAYIDX_US_US_US:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP27]]
-; CHECK-NEXT:    [[TMP28:%.*]] = load i32, ptr [[ARRAYIDX_US_US_US]], align 4
-; CHECK-NEXT:    [[TMP29:%.*]] = add nuw nsw i64 [[TMP18]], [[INNER_IV]]
-; CHECK-NEXT:    [[ARRAYIDX17_US_US_US:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP29]]
-; CHECK-NEXT:    [[TMP30:%.*]] = load i32, ptr [[ARRAYIDX17_US_US_US]], align 4
-; CHECK-NEXT:    [[ADD18_US_US_US:%.*]] = add nsw i32 [[TMP30]], [[TMP28]]
+; CHECK-NEXT:    [[TMP25:%.*]] = add nuw nsw i64 [[INNER_IV]], [[TMP15]]
+; CHECK-NEXT:    [[ARRAYIDX_US_US_US:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP25]]
+; CHECK-NEXT:    [[TMP26:%.*]] = load i32, ptr [[ARRAYIDX_US_US_US]], align 4
+; CHECK-NEXT:    [[TMP27:%.*]] = add nuw nsw i64 [[TMP18]], [[INNER_IV]]
+; CHECK-NEXT:    [[ARRAYIDX17_US_US_US:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP27]]
+; CHECK-NEXT:    [[TMP28:%.*]] = load i32, ptr [[ARRAYIDX17_US_US_US]], align 4
+; CHECK-NEXT:    [[ADD18_US_US_US:%.*]] = add nsw i32 [[TMP28]], [[TMP26]]
 ; CHECK-NEXT:    store i32 [[ADD18_US_US_US]], ptr [[ARRAYIDX17_US_US_US]], align 4
 ; CHECK-NEXT:    [[INNER_IV_NEXT]] = add nuw nsw i64 [[INNER_IV]], 1
 ; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INNER_IV_NEXT]], [[WIDE_TRIP_COUNT]]
@@ -847,17 +835,15 @@ define void @uncomputable_outer_tc(ptr nocapture noundef %dst, ptr nocapture nou
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP12:%.*]] = add nsw i64 [[INDEX]], [[TMP10]]
 ; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP12]]
-; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP13]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP14]], align 4, !alias.scope [[META39:![0-9]+]]
-; CHECK-NEXT:    [[TMP15:%.*]] = add nsw i64 [[INDEX]], [[TMP11]]
-; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP15]]
-; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP16]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i32>, ptr [[TMP17]], align 4, !alias.scope [[META42:![0-9]+]], !noalias [[META39]]
-; CHECK-NEXT:    [[TMP18:%.*]] = add nsw <4 x i32> [[WIDE_LOAD4]], [[WIDE_LOAD]]
-; CHECK-NEXT:    store <4 x i32> [[TMP18]], ptr [[TMP17]], align 4, !alias.scope [[META42]], !noalias [[META39]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP13]], align 4, !alias.scope [[META39:![0-9]+]]
+; CHECK-NEXT:    [[TMP14:%.*]] = add nsw i64 [[INDEX]], [[TMP11]]
+; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP14]]
+; CHECK-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i32>, ptr [[TMP15]], align 4, !alias.scope [[META42:![0-9]+]], !noalias [[META39]]
+; CHECK-NEXT:    [[TMP16:%.*]] = add nsw <4 x i32> [[WIDE_LOAD4]], [[WIDE_LOAD]]
+; CHECK-NEXT:    store <4 x i32> [[TMP16]], ptr [[TMP15]], align 4, !alias.scope [[META42]], !noalias [[META39]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP44:![0-9]+]]
+; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP44:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[INNER_LOOP_EXIT]], label [[SCALAR_PH]]
@@ -866,13 +852,13 @@ define void @uncomputable_outer_tc(ptr nocapture noundef %dst, ptr nocapture nou
 ; CHECK-NEXT:    br label [[INNER_LOOP:%.*]]
 ; CHECK:       inner.loop:
 ; CHECK-NEXT:    [[INNER_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INNER_IV_NEXT:%.*]], [[INNER_LOOP]] ]
-; CHECK-NEXT:    [[TMP20:%.*]] = add nsw i64 [[INNER_IV]], [[TMP10]]
-; CHECK-NEXT:    [[ARRAYIDX5_US:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP20]]
-; CHECK-NEXT:    [[TMP21:%.*]] = load i32, ptr [[ARRAYIDX5_US]], align 4
-; CHECK-NEXT:    [[TMP22:%.*]] = add nsw i64 [[INNER_IV]], [[TMP11]]
-; CHECK-NEXT:    [[ARRAYIDX10_US:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP22]]
-; CHECK-NEXT:    [[TMP23:%.*]] = load i32, ptr [[ARRAYIDX10_US]], align 4
-; CHECK-NEXT:    [[ADD11_US:%.*]] = add nsw i32 [[TMP23]], [[TMP21]]
+; CHECK-NEXT:    [[TMP18:%.*]] = add nsw i64 [[INNER_IV]], [[TMP10]]
+; CHECK-NEXT:    [[ARRAYIDX5_US:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP18]]
+; CHECK-NEXT:    [[TMP19:%.*]] = load i32, ptr [[ARRAYIDX5_US]], align 4
+; CHECK-NEXT:    [[TMP20:%.*]] = add nsw i64 [[INNER_IV]], [[TMP11]]
+; CHECK-NEXT:    [[ARRAYIDX10_US:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP20]]
+; CHECK-NEXT:    [[TMP21:%.*]] = load i32, ptr [[ARRAYIDX10_US]], align 4
+; CHECK-NEXT:    [[ADD11_US:%.*]] = add nsw i32 [[TMP21]], [[TMP19]]
 ; CHECK-NEXT:    store i32 [[ADD11_US]], ptr [[ARRAYIDX10_US]], align 4
 ; CHECK-NEXT:    [[INNER_IV_NEXT]] = add nuw nsw i64 [[INNER_IV]], 1
 ; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INNER_IV_NEXT]], [[WIDE_TRIP_COUNT]]
@@ -880,8 +866,8 @@ define void @uncomputable_outer_tc(ptr nocapture noundef %dst, ptr nocapture nou
 ; CHECK:       inner.loop.exit:
 ; CHECK-NEXT:    [[OUTER_IV_NEXT]] = add i64 [[OUTER_IV]], 1
 ; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i8, ptr [[STR]], i64 [[OUTER_IV_NEXT]]
-; CHECK-NEXT:    [[TMP24:%.*]] = load i8, ptr [[ARRAYIDX_US]], align 1
-; CHECK-NEXT:    [[CMP_NOT_US:%.*]] = icmp eq i8 [[TMP24]], 0
+; CHECK-NEXT:    [[TMP22:%.*]] = load i8, ptr [[ARRAYIDX_US]], align 1
+; CHECK-NEXT:    [[CMP_NOT_US:%.*]] = icmp eq i8 [[TMP22]], 0
 ; CHECK-NEXT:    br i1 [[CMP_NOT_US]], label [[WHILE_END_LOOPEXIT:%.*]], label [[OUTER_LOOP]]
 ; CHECK:       while.end.loopexit:
 ; CHECK-NEXT:    br label [[WHILE_END]]
@@ -1176,17 +1162,15 @@ define void @decreasing_outer_iv(ptr nocapture noundef %dst, ptr nocapture nound
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP19:%.*]] = add nsw i64 [[INDEX]], [[TMP15]]
 ; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP19]]
-; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, ptr [[TMP20]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP21]], align 4, !alias.scope [[META53:![0-9]+]]
-; CHECK-NEXT:    [[TMP22:%.*]] = add nsw i64 [[INDEX]], [[TMP16]]
-; CHECK-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP22]]
-; CHECK-NEXT:    [[TMP24:%.*]] = getelementptr inbounds i32, ptr [[TMP23]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i32>, ptr [[TMP24]], align 4, !alias.scope [[META56:![0-9]+]], !noalias [[META53]]
-; CHECK-NEXT:    [[TMP25:%.*]] = add nsw <4 x i32> [[WIDE_LOAD5]], [[WIDE_LOAD]]
-; CHECK-NEXT:    store <4 x i32> [[TMP25]], ptr [[TMP24]], align 4, !alias.scope [[META56]], !noalias [[META53]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP20]], align 4, !alias.scope [[META53:![0-9]+]]
+; CHECK-NEXT:    [[TMP21:%.*]] = add nsw i64 [[INDEX]], [[TMP16]]
+; CHECK-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP21]]
+; CHECK-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i32>, ptr [[TMP22]], align 4, !alias.scope [[META56:![0-9]+]], !noalias [[META53]]
+; CHECK-NEXT:    [[TMP23:%.*]] = add nsw <4 x i32> [[WIDE_LOAD5]], [[WIDE_LOAD]]
+; CHECK-NEXT:    store <4 x i32> [[TMP23]], ptr [[TMP22]], align 4, !alias.scope [[META56]], !noalias [[META53]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP26]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP58:![0-9]+]]
+; CHECK-NEXT:    [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP24]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP58:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[INNER_LOOP_EXIT]], label [[SCALAR_PH]]
@@ -1195,13 +1179,13 @@ define void @decreasing_outer_iv(ptr nocapture noundef %dst, ptr nocapture nound
 ; CHECK-NEXT:    br label [[INNER_LOOP:%.*]]
 ; CHECK:       inner.loop:
 ; CHECK-NEXT:    [[INNER_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INNER_IV_NEXT:%.*]], [[INNER_LOOP]] ]
-; CHECK-NEXT:    [[TMP27:%.*]] = add nsw i64 [[INNER_IV]], [[TMP15]]
-; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP27]]
-; CHECK-NEXT:    [[TMP28:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
-; CHECK-NEXT:    [[TMP29:%.*]] = add nsw i64 [[INNER_IV]], [[TMP16]]
-; CHECK-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP29]]
-; CHECK-NEXT:    [[TMP30:%.*]] = load i32, ptr [[ARRAYIDX8]], align 4
-; CHECK-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP30]], [[TMP28]]
+; CHECK-NEXT:    [[TMP25:%.*]] = add nsw i64 [[INNER_IV]], [[TMP15]]
+; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP25]]
+; CHECK-NEXT:    [[TMP26:%.*]] = load i32, ptr [[ARRAYIDX]], align 4
+; CHECK-NEXT:    [[TMP27:%.*]] = add nsw i64 [[INNER_IV]], [[TMP16]]
+; CHECK-NEXT:    [[ARRAYIDX8:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP27]]
+; CHECK-NEXT:    [[TMP28:%.*]] = load i32, ptr [[ARRAYIDX8]], align 4
+; CHECK-NEXT:    [[ADD9:%.*]] = add nsw i32 [[TMP28]], [[TMP26]]
 ; CHECK-NEXT:    store i32 [[ADD9]], ptr [[ARRAYIDX8]], align 4
 ; CHECK-NEXT:    [[INNER_IV_NEXT]] = add nuw nsw i64 [[INNER_IV]], 1
 ; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INNER_IV_NEXT]], [[WIDE_TRIP_COUNT]]
@@ -1330,17 +1314,15 @@ define void @unknown_inner_stride(ptr nocapture noundef %dst, ptr nocapture noun
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP14:%.*]] = add nsw i64 [[INDEX]], [[TMP11]]
 ; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP14]]
-; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[TMP15]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP16]], align 4, !alias.scope [[META60:![0-9]+]]
-; CHECK-NEXT:    [[TMP17:%.*]] = add nsw i64 [[INDEX]], [[TMP12]]
-; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP17]]
-; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[TMP18]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i32>, ptr [[TMP19]], align 4, !alias.scope [[META63:![0-9]+]], !noalias [[META60]]
-; CHECK-NEXT:    [[TMP20:%.*]] = add nsw <4 x i32> [[WIDE_LOAD3]], [[WIDE_LOAD]]
-; CHECK-NEXT:    store <4 x i32> [[TMP20]], ptr [[TMP19]], align 4, !alias.scope [[META63]], !noalias [[META60]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP15]], align 4, !alias.scope [[META60:![0-9]+]]
+; CHECK-NEXT:    [[TMP16:%.*]] = add nsw i64 [[INDEX]], [[TMP12]]
+; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP16]]
+; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i32>, ptr [[TMP17]], align 4, !alias.scope [[META63:![0-9]+]], !noalias [[META60]]
+; CHECK-NEXT:    [[TMP18:%.*]] = add nsw <4 x i32> [[WIDE_LOAD3]], [[WIDE_LOAD]]
+; CHECK-NEXT:    store <4 x i32> [[TMP18]], ptr [[TMP17]], align 4, !alias.scope [[META63]], !noalias [[META60]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP65:![0-9]+]]
+; CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP65:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[WIDE_TRIP_COUNT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[INNER_LOOP_EXIT]], label [[SCALAR_PH]]
@@ -1349,15 +1331,15 @@ define void @unknown_inner_stride(ptr nocapture noundef %dst, ptr nocapture noun
 ; CHECK-NEXT:    br label [[INNER_LOOP:%.*]]
 ; CHECK:       inner.loop:
 ; CHECK-NEXT:    [[INNER_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INNER_IV_NEXT:%.*]], [[INNER_LOOP]] ]
-; CHECK-NEXT:    [[TMP22:%.*]] = mul nsw i64 [[INNER_IV]], [[TMP0]]
-; CHECK-NEXT:    [[TMP23:%.*]] = add nsw i64 [[TMP22]], [[TMP11]]
-; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP23]]
-; CHECK-NEXT:    [[TMP24:%.*]] = load i32, ptr [[ARRAYIDX_US]], align 4
-; CHECK-NEXT:    [[TMP25:%.*]] = mul nsw i64 [[INNER_IV]], [[TMP1]]
-; CHECK-NEXT:    [[TMP26:%.*]] = add nsw i64 [[TMP25]], [[TMP12]]
-; CHECK-NEXT:    [[ARRAYIDX11_US:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP26]]
-; CHECK-NEXT:    [[TMP27:%.*]] = load i32, ptr [[ARRAYIDX11_US]], align 4
-; CHECK-NEXT:    [[ADD12_US:%.*]] = add nsw i32 [[TMP27]], [[TMP24]]
+; CHECK-NEXT:    [[TMP20:%.*]] = mul nsw i64 [[INNER_IV]], [[TMP0]]
+; CHECK-NEXT:    [[TMP21:%.*]] = add nsw i64 [[TMP20]], [[TMP11]]
+; CHECK-NEXT:    [[ARRAYIDX_US:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP21]]
+; CHECK-NEXT:    [[TMP22:%.*]] = load i32, ptr [[ARRAYIDX_US]], align 4
+; CHECK-NEXT:    [[TMP23:%.*]] = mul nsw i64 [[INNER_IV]], [[TMP1]]
+; CHECK-NEXT:    [[TMP24:%.*]] = add nsw i64 [[TMP23]], [[TMP12]]
+; CHECK-NEXT:    [[ARRAYIDX11_US:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP24]]
+; CHECK-NEXT:    [[TMP25:%.*]] = load i32, ptr [[ARRAYIDX11_US]], align 4
+; CHECK-NEXT:    [[ADD12_US:%.*]] = add nsw i32 [[TMP25]], [[TMP22]]
 ; CHECK-NEXT:    store i32 [[ADD12_US]], ptr [[ARRAYIDX11_US]], align 4
 ; CHECK-NEXT:    [[INNER_IV_NEXT]] = add nuw nsw i64 [[INNER_IV]], 1
 ; CHECK-NEXT:    [[EXITCOND_NOT:%.*]] = icmp eq i64 [[INNER_IV_NEXT]], [[WIDE_TRIP_COUNT]]
@@ -1446,15 +1428,13 @@ define void @nested_loop_start_of_inner_ptr_addrec_is_same_outer_addrec(ptr noca
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = add nuw nsw i64 [[INDEX]], [[MUL]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[TMP1]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP1]]
-; CHECK-NEXT:    [[TMP5:%.*]] = add nsw <4 x i32> [[WIDE_LOAD]], splat (i32 10)
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <4 x i32> [[TMP5]], ptr [[TMP6]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[TMP1]]
+; CHECK-NEXT:    [[TMP4:%.*]] = add nsw <4 x i32> [[WIDE_LOAD]], splat (i32 10)
+; CHECK-NEXT:    store <4 x i32> [[TMP4]], ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP67:![0-9]+]]
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP67:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[INNER_EXIT]], label [[SCALAR_PH]]
@@ -1538,15 +1518,14 @@ define void @stride_check_known_via_loop_guard(ptr %C, ptr %A, i32 %Acols) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = load double, ptr [[ARRAYIDX_US]], align 8, !alias.scope [[META69:![0-9]+]]
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x double> poison, double [[TMP1]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x double> [[BROADCAST_SPLATINSERT]], <4 x double> poison, <4 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds double, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    store <4 x double> [[BROADCAST_SPLAT]], ptr [[TMP2]], align 8, !alias.scope [[META72:![0-9]+]], !noalias [[META69]]
+; CHECK-NEXT:    store <4 x double> [[BROADCAST_SPLAT]], ptr [[TMP0]], align 8, !alias.scope [[META72:![0-9]+]], !noalias [[META69]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
-; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP74:![0-9]+]]
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP74:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[OUTER_LATCH]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[OUTER_LATCH]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[OUTER_HEADER]] ], [ 0, [[VECTOR_MEMCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[OUTER_HEADER]] ], [ 0, [[VECTOR_MEMCHECK]] ]
 ; CHECK-NEXT:    br label [[INNER:%.*]]
 ; CHECK:       inner:
 ; CHECK-NEXT:    [[INNER_IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INNER_IV_NEXT:%.*]], [[INNER]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/scalable-first-order-recurrence.ll b/llvm/test/Transforms/LoopVectorize/scalable-first-order-recurrence.ll
index 23a3fc1..bfc0a48 100644
--- a/llvm/test/Transforms/LoopVectorize/scalable-first-order-recurrence.ll
+++ b/llvm/test/Transforms/LoopVectorize/scalable-first-order-recurrence.ll
@@ -48,13 +48,11 @@ define i32 @recurrence_1(ptr nocapture readonly %a, ptr nocapture %b, i32 %n) {
 ; CHECK-VF4UF1-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i32> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4UF1-NEXT:    [[TMP17:%.*]] = add nuw nsw i64 [[INDEX]], 1
 ; CHECK-VF4UF1-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP17]]
-; CHECK-VF4UF1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[TMP18]], i32 0
-; CHECK-VF4UF1-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i32>, ptr [[TMP19]], align 4
+; CHECK-VF4UF1-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i32>, ptr [[TMP18]], align 4
 ; CHECK-VF4UF1-NEXT:    [[TMP20:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR]], <vscale x 4 x i32> [[WIDE_LOAD]], i32 -1)
 ; CHECK-VF4UF1-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
 ; CHECK-VF4UF1-NEXT:    [[TMP22:%.*]] = add <vscale x 4 x i32> [[WIDE_LOAD]], [[TMP20]]
-; CHECK-VF4UF1-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i32, ptr [[TMP21]], i32 0
-; CHECK-VF4UF1-NEXT:    store <vscale x 4 x i32> [[TMP22]], ptr [[TMP23]], align 4
+; CHECK-VF4UF1-NEXT:    store <vscale x 4 x i32> [[TMP22]], ptr [[TMP21]], align 4
 ; CHECK-VF4UF1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP13]]
 ; CHECK-VF4UF1-NEXT:    [[TMP24:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-VF4UF1-NEXT:    br i1 [[TMP24]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -115,22 +113,20 @@ define i32 @recurrence_1(ptr nocapture readonly %a, ptr nocapture %b, i32 %n) {
 ; CHECK-VF4UF2-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i32> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD3:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4UF2-NEXT:    [[TMP17:%.*]] = add nuw nsw i64 [[INDEX]], 1
 ; CHECK-VF4UF2-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[TMP17]]
-; CHECK-VF4UF2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[TMP18]], i32 0
 ; CHECK-VF4UF2-NEXT:    [[TMP20:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-VF4UF2-NEXT:    [[TMP21:%.*]] = mul nuw i64 [[TMP20]], 4
 ; CHECK-VF4UF2-NEXT:    [[TMP22:%.*]] = getelementptr inbounds i32, ptr [[TMP18]], i64 [[TMP21]]
-; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP19]], align 4
+; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP18]], align 4
 ; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD3]] = load <vscale x 4 x i32>, ptr [[TMP22]], align 4
 ; CHECK-VF4UF2-NEXT:    [[TMP23:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR]], <vscale x 4 x i32> [[WIDE_LOAD]], i32 -1)
 ; CHECK-VF4UF2-NEXT:    [[TMP24:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[WIDE_LOAD3]], i32 -1)
 ; CHECK-VF4UF2-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
 ; CHECK-VF4UF2-NEXT:    [[TMP26:%.*]] = add <vscale x 4 x i32> [[WIDE_LOAD]], [[TMP23]]
 ; CHECK-VF4UF2-NEXT:    [[TMP27:%.*]] = add <vscale x 4 x i32> [[WIDE_LOAD3]], [[TMP24]]
-; CHECK-VF4UF2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i32, ptr [[TMP25]], i32 0
 ; CHECK-VF4UF2-NEXT:    [[TMP29:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-VF4UF2-NEXT:    [[TMP30:%.*]] = mul nuw i64 [[TMP29]], 4
 ; CHECK-VF4UF2-NEXT:    [[TMP31:%.*]] = getelementptr inbounds i32, ptr [[TMP25]], i64 [[TMP30]]
-; CHECK-VF4UF2-NEXT:    store <vscale x 4 x i32> [[TMP26]], ptr [[TMP28]], align 4
+; CHECK-VF4UF2-NEXT:    store <vscale x 4 x i32> [[TMP26]], ptr [[TMP25]], align 4
 ; CHECK-VF4UF2-NEXT:    store <vscale x 4 x i32> [[TMP27]], ptr [[TMP31]], align 4
 ; CHECK-VF4UF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP13]]
 ; CHECK-VF4UF2-NEXT:    [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -214,8 +210,7 @@ define i32 @recurrence_2(ptr nocapture readonly %a, i32 %n) {
 ; CHECK-VF4UF1-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i32> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4UF1-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ undef, %[[VECTOR_PH]] ], [ [[TMP17:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4UF1-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4UF1-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[TMP10]], i32 0
-; CHECK-VF4UF1-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i32>, ptr [[TMP11]], align 4
+; CHECK-VF4UF1-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i32>, ptr [[TMP10]], align 4
 ; CHECK-VF4UF1-NEXT:    [[TMP12:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR]], <vscale x 4 x i32> [[WIDE_LOAD]], i32 -1)
 ; CHECK-VF4UF1-NEXT:    [[TMP13:%.*]] = sub nsw <vscale x 4 x i32> [[WIDE_LOAD]], [[TMP12]]
 ; CHECK-VF4UF1-NEXT:    [[TMP14:%.*]] = icmp sgt <vscale x 4 x i32> [[TMP13]], zeroinitializer
@@ -277,11 +272,10 @@ define i32 @recurrence_2(ptr nocapture readonly %a, i32 %n) {
 ; CHECK-VF4UF2-NEXT:    [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ undef, %[[VECTOR_PH]] ], [ [[TMP25:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4UF2-NEXT:    [[VEC_PHI1:%.*]] = phi <vscale x 4 x i32> [ undef, %[[VECTOR_PH]] ], [ [[TMP26:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4UF2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4UF2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[TMP10]], i32 0
 ; CHECK-VF4UF2-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-VF4UF2-NEXT:    [[TMP13:%.*]] = mul nuw i64 [[TMP12]], 4
 ; CHECK-VF4UF2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i32, ptr [[TMP10]], i64 [[TMP13]]
-; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP11]], align 4
+; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP10]], align 4
 ; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD2]] = load <vscale x 4 x i32>, ptr [[TMP14]], align 4
 ; CHECK-VF4UF2-NEXT:    [[TMP15:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[VECTOR_RECUR]], <vscale x 4 x i32> [[WIDE_LOAD]], i32 -1)
 ; CHECK-VF4UF2-NEXT:    [[TMP16:%.*]] = call <vscale x 4 x i32> @llvm.vector.splice.nxv4i32(<vscale x 4 x i32> [[WIDE_LOAD]], <vscale x 4 x i32> [[WIDE_LOAD2]], i32 -1)
@@ -409,16 +403,14 @@ define void @recurrence_3(ptr nocapture readonly %a, ptr nocapture %b, i32 %n, f
 ; CHECK-VF4UF1-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i16> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4UF1-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; CHECK-VF4UF1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[OFFSET_IDX]]
-; CHECK-VF4UF1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i16, ptr [[TMP19]], i32 0
-; CHECK-VF4UF1-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i16>, ptr [[TMP20]], align 2, !alias.scope [[META6:![0-9]+]]
+; CHECK-VF4UF1-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i16>, ptr [[TMP19]], align 2, !alias.scope [[META6:![0-9]+]]
 ; CHECK-VF4UF1-NEXT:    [[TMP21:%.*]] = call <vscale x 4 x i16> @llvm.vector.splice.nxv4i16(<vscale x 4 x i16> [[VECTOR_RECUR]], <vscale x 4 x i16> [[WIDE_LOAD]], i32 -1)
 ; CHECK-VF4UF1-NEXT:    [[TMP22:%.*]] = sitofp <vscale x 4 x i16> [[WIDE_LOAD]] to <vscale x 4 x double>
 ; CHECK-VF4UF1-NEXT:    [[TMP23:%.*]] = sitofp <vscale x 4 x i16> [[TMP21]] to <vscale x 4 x double>
 ; CHECK-VF4UF1-NEXT:    [[TMP24:%.*]] = fmul fast <vscale x 4 x double> [[TMP23]], [[BROADCAST_SPLAT]]
 ; CHECK-VF4UF1-NEXT:    [[TMP25:%.*]] = fsub fast <vscale x 4 x double> [[TMP22]], [[TMP24]]
 ; CHECK-VF4UF1-NEXT:    [[TMP26:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[OFFSET_IDX]]
-; CHECK-VF4UF1-NEXT:    [[TMP27:%.*]] = getelementptr inbounds double, ptr [[TMP26]], i32 0
-; CHECK-VF4UF1-NEXT:    store <vscale x 4 x double> [[TMP25]], ptr [[TMP27]], align 8, !alias.scope [[META9:![0-9]+]], !noalias [[META6]]
+; CHECK-VF4UF1-NEXT:    store <vscale x 4 x double> [[TMP25]], ptr [[TMP26]], align 8, !alias.scope [[META9:![0-9]+]], !noalias [[META6]]
 ; CHECK-VF4UF1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP14]]
 ; CHECK-VF4UF1-NEXT:    [[TMP28:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-VF4UF1-NEXT:    br i1 [[TMP28]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP11:![0-9]+]]
@@ -489,11 +481,10 @@ define void @recurrence_3(ptr nocapture readonly %a, ptr nocapture %b, i32 %n, f
 ; CHECK-VF4UF2-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i16> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4UF2-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; CHECK-VF4UF2-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[OFFSET_IDX]]
-; CHECK-VF4UF2-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i16, ptr [[TMP19]], i32 0
 ; CHECK-VF4UF2-NEXT:    [[TMP21:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-VF4UF2-NEXT:    [[TMP22:%.*]] = mul nuw i64 [[TMP21]], 4
 ; CHECK-VF4UF2-NEXT:    [[TMP23:%.*]] = getelementptr inbounds i16, ptr [[TMP19]], i64 [[TMP22]]
-; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP20]], align 2, !alias.scope [[META6:![0-9]+]]
+; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP19]], align 2, !alias.scope [[META6:![0-9]+]]
 ; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD4]] = load <vscale x 4 x i16>, ptr [[TMP23]], align 2, !alias.scope [[META6]]
 ; CHECK-VF4UF2-NEXT:    [[TMP24:%.*]] = call <vscale x 4 x i16> @llvm.vector.splice.nxv4i16(<vscale x 4 x i16> [[VECTOR_RECUR]], <vscale x 4 x i16> [[WIDE_LOAD]], i32 -1)
 ; CHECK-VF4UF2-NEXT:    [[TMP25:%.*]] = call <vscale x 4 x i16> @llvm.vector.splice.nxv4i16(<vscale x 4 x i16> [[WIDE_LOAD]], <vscale x 4 x i16> [[WIDE_LOAD4]], i32 -1)
@@ -506,11 +497,10 @@ define void @recurrence_3(ptr nocapture readonly %a, ptr nocapture %b, i32 %n, f
 ; CHECK-VF4UF2-NEXT:    [[TMP32:%.*]] = fsub fast <vscale x 4 x double> [[TMP26]], [[TMP30]]
 ; CHECK-VF4UF2-NEXT:    [[TMP33:%.*]] = fsub fast <vscale x 4 x double> [[TMP27]], [[TMP31]]
 ; CHECK-VF4UF2-NEXT:    [[TMP34:%.*]] = getelementptr inbounds double, ptr [[B]], i64 [[OFFSET_IDX]]
-; CHECK-VF4UF2-NEXT:    [[TMP35:%.*]] = getelementptr inbounds double, ptr [[TMP34]], i32 0
 ; CHECK-VF4UF2-NEXT:    [[TMP36:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-VF4UF2-NEXT:    [[TMP37:%.*]] = mul nuw i64 [[TMP36]], 4
 ; CHECK-VF4UF2-NEXT:    [[TMP38:%.*]] = getelementptr inbounds double, ptr [[TMP34]], i64 [[TMP37]]
-; CHECK-VF4UF2-NEXT:    store <vscale x 4 x double> [[TMP32]], ptr [[TMP35]], align 8, !alias.scope [[META9:![0-9]+]], !noalias [[META6]]
+; CHECK-VF4UF2-NEXT:    store <vscale x 4 x double> [[TMP32]], ptr [[TMP34]], align 8, !alias.scope [[META9:![0-9]+]], !noalias [[META6]]
 ; CHECK-VF4UF2-NEXT:    store <vscale x 4 x double> [[TMP33]], ptr [[TMP38]], align 8, !alias.scope [[META9]], !noalias [[META6]]
 ; CHECK-VF4UF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP14]]
 ; CHECK-VF4UF2-NEXT:    [[TMP39:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
@@ -800,15 +790,13 @@ define void @sink_after(ptr %a, ptr %b, i64 %n) {
 ; CHECK-VF4UF1-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i16> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4UF1-NEXT:    [[TMP12:%.*]] = add nuw nsw i64 [[INDEX]], 1
 ; CHECK-VF4UF1-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[TMP12]]
-; CHECK-VF4UF1-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i16, ptr [[TMP13]], i32 0
-; CHECK-VF4UF1-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i16>, ptr [[TMP14]], align 2, !alias.scope [[META17:![0-9]+]]
+; CHECK-VF4UF1-NEXT:    [[WIDE_LOAD]] = load <vscale x 4 x i16>, ptr [[TMP13]], align 2, !alias.scope [[META17:![0-9]+]]
 ; CHECK-VF4UF1-NEXT:    [[TMP15:%.*]] = call <vscale x 4 x i16> @llvm.vector.splice.nxv4i16(<vscale x 4 x i16> [[VECTOR_RECUR]], <vscale x 4 x i16> [[WIDE_LOAD]], i32 -1)
 ; CHECK-VF4UF1-NEXT:    [[TMP16:%.*]] = sext <vscale x 4 x i16> [[TMP15]] to <vscale x 4 x i32>
 ; CHECK-VF4UF1-NEXT:    [[TMP17:%.*]] = sext <vscale x 4 x i16> [[WIDE_LOAD]] to <vscale x 4 x i32>
 ; CHECK-VF4UF1-NEXT:    [[TMP18:%.*]] = mul nsw <vscale x 4 x i32> [[TMP17]], [[TMP16]]
 ; CHECK-VF4UF1-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
-; CHECK-VF4UF1-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i32, ptr [[TMP19]], i32 0
-; CHECK-VF4UF1-NEXT:    store <vscale x 4 x i32> [[TMP18]], ptr [[TMP20]], align 4, !alias.scope [[META20:![0-9]+]], !noalias [[META17]]
+; CHECK-VF4UF1-NEXT:    store <vscale x 4 x i32> [[TMP18]], ptr [[TMP19]], align 4, !alias.scope [[META20:![0-9]+]], !noalias [[META17]]
 ; CHECK-VF4UF1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
 ; CHECK-VF4UF1-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-VF4UF1-NEXT:    br i1 [[TMP21]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
@@ -861,11 +849,10 @@ define void @sink_after(ptr %a, ptr %b, i64 %n) {
 ; CHECK-VF4UF2-NEXT:    [[VECTOR_RECUR:%.*]] = phi <vscale x 4 x i16> [ [[VECTOR_RECUR_INIT]], %[[VECTOR_PH]] ], [ [[WIDE_LOAD3:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4UF2-NEXT:    [[TMP12:%.*]] = add nuw nsw i64 [[INDEX]], 1
 ; CHECK-VF4UF2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i16, ptr [[A]], i64 [[TMP12]]
-; CHECK-VF4UF2-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i16, ptr [[TMP13]], i32 0
 ; CHECK-VF4UF2-NEXT:    [[TMP15:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-VF4UF2-NEXT:    [[TMP16:%.*]] = mul nuw i64 [[TMP15]], 4
 ; CHECK-VF4UF2-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i16, ptr [[TMP13]], i64 [[TMP16]]
-; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP14]], align 2, !alias.scope [[META17:![0-9]+]]
+; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i16>, ptr [[TMP13]], align 2, !alias.scope [[META17:![0-9]+]]
 ; CHECK-VF4UF2-NEXT:    [[WIDE_LOAD3]] = load <vscale x 4 x i16>, ptr [[TMP17]], align 2, !alias.scope [[META17]]
 ; CHECK-VF4UF2-NEXT:    [[TMP18:%.*]] = call <vscale x 4 x i16> @llvm.vector.splice.nxv4i16(<vscale x 4 x i16> [[VECTOR_RECUR]], <vscale x 4 x i16> [[WIDE_LOAD]], i32 -1)
 ; CHECK-VF4UF2-NEXT:    [[TMP19:%.*]] = call <vscale x 4 x i16> @llvm.vector.splice.nxv4i16(<vscale x 4 x i16> [[WIDE_LOAD]], <vscale x 4 x i16> [[WIDE_LOAD3]], i32 -1)
@@ -876,11 +863,10 @@ define void @sink_after(ptr %a, ptr %b, i64 %n) {
 ; CHECK-VF4UF2-NEXT:    [[TMP24:%.*]] = mul nsw <vscale x 4 x i32> [[TMP22]], [[TMP20]]
 ; CHECK-VF4UF2-NEXT:    [[TMP25:%.*]] = mul nsw <vscale x 4 x i32> [[TMP23]], [[TMP21]]
 ; CHECK-VF4UF2-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i32, ptr [[B]], i64 [[INDEX]]
-; CHECK-VF4UF2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i32, ptr [[TMP26]], i32 0
 ; CHECK-VF4UF2-NEXT:    [[TMP28:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-VF4UF2-NEXT:    [[TMP29:%.*]] = mul nuw i64 [[TMP28]], 4
 ; CHECK-VF4UF2-NEXT:    [[TMP30:%.*]] = getelementptr inbounds i32, ptr [[TMP26]], i64 [[TMP29]]
-; CHECK-VF4UF2-NEXT:    store <vscale x 4 x i32> [[TMP24]], ptr [[TMP27]], align 4, !alias.scope [[META20:![0-9]+]], !noalias [[META17]]
+; CHECK-VF4UF2-NEXT:    store <vscale x 4 x i32> [[TMP24]], ptr [[TMP26]], align 4, !alias.scope [[META20:![0-9]+]], !noalias [[META17]]
 ; CHECK-VF4UF2-NEXT:    store <vscale x 4 x i32> [[TMP25]], ptr [[TMP30]], align 4, !alias.scope [[META20]], !noalias [[META17]]
 ; CHECK-VF4UF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP8]]
 ; CHECK-VF4UF2-NEXT:    [[TMP31:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
diff --git a/llvm/test/Transforms/LoopVectorize/scalable-iv-outside-user.ll b/llvm/test/Transforms/LoopVectorize/scalable-iv-outside-user.ll
index b0029a4..1ec2993 100644
--- a/llvm/test/Transforms/LoopVectorize/scalable-iv-outside-user.ll
+++ b/llvm/test/Transforms/LoopVectorize/scalable-iv-outside-user.ll
@@ -32,11 +32,10 @@ define i32 @iv_live_out_wide(ptr %dst) {
 ; CHECK-NEXT:    [[BROADCAST_SPLAT2:%.*]] = shufflevector <vscale x 2 x i32> [[BROADCAST_SPLATINSERT1]], <vscale x 2 x i32> poison, <vscale x 2 x i32> zeroinitializer
 ; CHECK-NEXT:    [[STEP_ADD:%.*]] = add <vscale x 2 x i32> [[VEC_IND]], [[BROADCAST_SPLAT2]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i16, ptr [[DST]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i16, ptr [[TMP10]], i32 0
 ; CHECK-NEXT:    [[TMP12:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP13:%.*]] = mul nuw i64 [[TMP12]], 2
 ; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i16, ptr [[TMP10]], i64 [[TMP13]]
-; CHECK-NEXT:    store <vscale x 2 x i16> zeroinitializer, ptr [[TMP11]], align 2
+; CHECK-NEXT:    store <vscale x 2 x i16> zeroinitializer, ptr [[TMP10]], align 2
 ; CHECK-NEXT:    store <vscale x 2 x i16> zeroinitializer, ptr [[TMP14]], align 2
 ; CHECK-NEXT:    [[TMP15:%.*]] = add <vscale x 2 x i32> [[BROADCAST_SPLAT]], [[STEP_ADD]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], [[TMP6]]
diff --git a/llvm/test/Transforms/LoopVectorize/scalable-lifetime.ll b/llvm/test/Transforms/LoopVectorize/scalable-lifetime.ll
index 4a1d7a2..7aac9d1 100644
--- a/llvm/test/Transforms/LoopVectorize/scalable-lifetime.ll
+++ b/llvm/test/Transforms/LoopVectorize/scalable-lifetime.ll
@@ -27,12 +27,11 @@ define void @test(ptr %d) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    call void @llvm.lifetime.end.p0(i64 4096, ptr [[ARR]])
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[D]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <vscale x 2 x i32> splat (i32 100), ptr [[TMP7]], align 8
+; CHECK-NEXT:    store <vscale x 2 x i32> splat (i32 100), ptr [[TMP6]], align 8
 ; CHECK-NEXT:    call void @llvm.lifetime.start.p0(i64 4096, ptr [[ARR]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
-; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 128, [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
@@ -43,7 +42,7 @@ define void @test(ptr %d) {
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
 ; CHECK-NEXT:    call void @llvm.lifetime.end.p0(i64 4096, ptr [[ARR]])
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[D]], i64 [[INDVARS_IV]]
-; CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[ARRAYIDX]], align 8
+; CHECK-NEXT:    [[TMP8:%.*]] = load i32, ptr [[ARRAYIDX]], align 8
 ; CHECK-NEXT:    store i32 100, ptr [[ARRAYIDX]], align 8
 ; CHECK-NEXT:    call void @llvm.lifetime.start.p0(i64 4096, ptr [[ARR]])
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
@@ -98,12 +97,11 @@ define void @testloopvariant(ptr %d) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    call void @llvm.lifetime.end.p0(i64 4096, ptr [[ARR]])
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[D]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <vscale x 2 x i32> splat (i32 100), ptr [[TMP7]], align 8
+; CHECK-NEXT:    store <vscale x 2 x i32> splat (i32 100), ptr [[TMP6]], align 8
 ; CHECK-NEXT:    call void @llvm.lifetime.start.p0(i64 4096, ptr [[ARR]])
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
-; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 128, [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[FOR_END:%.*]], label [[SCALAR_PH]]
@@ -112,10 +110,10 @@ define void @testloopvariant(ptr %d) {
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[INDVARS_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[INDVARS_IV_NEXT:%.*]], [[FOR_BODY]] ]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr [1024 x i32], ptr [[ARR]], i32 0, i64 [[INDVARS_IV]]
+; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr [1024 x i32], ptr [[ARR]], i32 0, i64 [[INDVARS_IV]]
 ; CHECK-NEXT:    call void @llvm.lifetime.end.p0(i64 4096, ptr [[ARR]])
 ; CHECK-NEXT:    [[ARRAYIDX:%.*]] = getelementptr inbounds i32, ptr [[D]], i64 [[INDVARS_IV]]
-; CHECK-NEXT:    [[TMP10:%.*]] = load i32, ptr [[ARRAYIDX]], align 8
+; CHECK-NEXT:    [[TMP9:%.*]] = load i32, ptr [[ARRAYIDX]], align 8
 ; CHECK-NEXT:    store i32 100, ptr [[ARRAYIDX]], align 8
 ; CHECK-NEXT:    call void @llvm.lifetime.start.p0(i64 4096, ptr [[ARR]])
 ; CHECK-NEXT:    [[INDVARS_IV_NEXT]] = add i64 [[INDVARS_IV]], 1
diff --git a/llvm/test/Transforms/LoopVectorize/scalable-reduction-inloop.ll b/llvm/test/Transforms/LoopVectorize/scalable-reduction-inloop.ll
index e901d98..ba337aa 100644
--- a/llvm/test/Transforms/LoopVectorize/scalable-reduction-inloop.ll
+++ b/llvm/test/Transforms/LoopVectorize/scalable-reduction-inloop.ll
@@ -25,11 +25,10 @@ define i8 @reduction_add_trunc(ptr noalias nocapture %A) {
 ; CHECK-NEXT:    [[TMP14:%.*]] = and <vscale x 8 x i32> [[VEC_PHI]], splat (i32 255)
 ; CHECK-NEXT:    [[TMP15:%.*]] = and <vscale x 8 x i32> [[VEC_PHI1]], splat (i32 255)
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i32 0
 ; CHECK-NEXT:    [[TMP10:%.*]] = call i64 @llvm.vscale.i64()
 ; CHECK-NEXT:    [[TMP11:%.*]] = mul nuw i64 [[TMP10]], 8
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP8]], i64 [[TMP11]]
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP9]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <vscale x 8 x i8>, ptr [[TMP12]], align 4
 ; CHECK-NEXT:    [[TMP26:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i32>
 ; CHECK-NEXT:    [[TMP27:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD2]] to <vscale x 8 x i32>
diff --git a/llvm/test/Transforms/LoopVectorize/scalar_after_vectorization.ll b/llvm/test/Transforms/LoopVectorize/scalar_after_vectorization.ll
index 16b0593..7811b17 100644
--- a/llvm/test/Transforms/LoopVectorize/scalar_after_vectorization.ll
+++ b/llvm/test/Transforms/LoopVectorize/scalar_after_vectorization.ll
@@ -24,9 +24,8 @@ target datalayout = "e-m:e-i64:64-i128:128-n32:64-S128"
 ; NO-IC:   %[[T4:.+]] = add nuw nsw i64 [[OFFSET_IDX]], %tmp0
 ; NO-IC:   %[[T6:.+]] = sub nsw i64 %[[T4]], %x
 ; NO-IC:   %[[T8:.+]] = getelementptr inbounds i32, ptr %a, i64 %[[T6]]
-; NO-IC:   %[[T10:.+]] = getelementptr inbounds i32, ptr %[[T8]], i32 0
 ; NO-IC:   %[[T12:.+]] = getelementptr inbounds i32, ptr %[[T8]], i32 4
-; NO-IC:   load <4 x i32>, ptr %[[T10]], align 4
+; NO-IC:   load <4 x i32>, ptr %[[T8]], align 4
 ; NO-IC:   load <4 x i32>, ptr %[[T12]], align 4
 ; NO-IC:   br {{.*}}, label %middle.block, label %vector.body
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/scev-exit-phi-invalidation.ll b/llvm/test/Transforms/LoopVectorize/scev-exit-phi-invalidation.ll
index a808c8b..70772dc 100644
--- a/llvm/test/Transforms/LoopVectorize/scev-exit-phi-invalidation.ll
+++ b/llvm/test/Transforms/LoopVectorize/scev-exit-phi-invalidation.ll
@@ -19,9 +19,9 @@ define void @test_pr63368(i1 %c, ptr %A) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = icmp eq i32 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP1]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT_1:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT_1:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_1_HEADER:%.*]]
 ; CHECK:       loop.1.header:
 ; CHECK-NEXT:    [[IV_1:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_1_NEXT:%.*]], [[LOOP_1_LATCH:%.*]] ]
@@ -57,11 +57,10 @@ define void @test_pr63368(i1 %c, ptr %A) {
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX5]] to i8
 ; CHECK-NEXT:    [[TMP10:%.*]] = add i8 [[OFFSET_IDX]], 1
 ; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr i8, ptr [[A]], i8 [[TMP10]]
-; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i8, ptr [[TMP11]], i32 0
-; CHECK-NEXT:    store <4 x i8> zeroinitializer, ptr [[TMP12]], align 1
+; CHECK-NEXT:    store <4 x i8> zeroinitializer, ptr [[TMP11]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT6]] = add nuw i32 [[INDEX5]], 4
-; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i32 [[INDEX_NEXT6]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK7:%.*]], label [[VECTOR_BODY4]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i32 [[INDEX_NEXT6]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK7:%.*]], label [[VECTOR_BODY4]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block7:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP2]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT_2:%.*]], label [[SCALAR_PH2]]
diff --git a/llvm/test/Transforms/LoopVectorize/scev-predicate-reasoning.ll b/llvm/test/Transforms/LoopVectorize/scev-predicate-reasoning.ll
index a5cf45c..b2acc64 100644
--- a/llvm/test/Transforms/LoopVectorize/scev-predicate-reasoning.ll
+++ b/llvm/test/Transforms/LoopVectorize/scev-predicate-reasoning.ll
@@ -46,9 +46,9 @@ define void @step_direction_unknown(i32 %arg, ptr %dst) {
 ; CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[PHI:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[ADD2:%.*]], [[LOOP]] ]
@@ -104,14 +104,13 @@ define void @integer_induction_wraps_scev_predicate_known(i32 %x, ptr %call, ptr
 ; CHECK-NEXT:    [[DOTCAST:%.*]] = trunc i64 [[INDEX]] to i32
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i32 30, [[DOTCAST]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr ptr, ptr [[CALL]], i32 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr ptr, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <4 x ptr> [[VECTOR_GEP]], ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <4 x ptr> [[VECTOR_GEP]], ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[PTR_IND]] = getelementptr i8, ptr [[POINTER_PHI]], i64 [[TMP3]]
-; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 992
-; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 992
+; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1022, [[MIDDLE_BLOCK]] ], [ 30, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi ptr [ [[TMP2]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
@@ -124,7 +123,7 @@ define void @integer_induction_wraps_scev_predicate_known(i32 %x, ptr %call, ptr
 ; CHECK-NEXT:    store ptr [[P_0]], ptr [[ARRAYIDX]], align 4
 ; CHECK-NEXT:    [[INC]] = add i32 [[IV]], 1
 ; CHECK-NEXT:    [[TOBOOL_NOT:%.*]] = icmp eq i32 [[IV]], 1024
-; CHECK-NEXT:    br i1 [[TOBOOL_NOT]], label [[FOR_END]], label [[FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    br i1 [[TOBOOL_NOT]], label [[FOR_END:%.*]], label [[FOR_COND]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       for.end:
 ; CHECK-NEXT:    ret void
 ;
@@ -190,14 +189,12 @@ define void @implied_wrap_predicate(ptr %A, ptr %B, ptr %C) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr i64, ptr [[A]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr i64, ptr [[TMP17]], i32 0
+; CHECK-NEXT:    store <4 x i64> zeroinitializer, ptr [[TMP17]], align 4
+; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr i64, ptr [[C]], i64 [[OFFSET_IDX]]
 ; CHECK-NEXT:    store <4 x i64> zeroinitializer, ptr [[TMP18]], align 4
-; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr i64, ptr [[C]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr i64, ptr [[TMP19]], i32 0
-; CHECK-NEXT:    store <4 x i64> zeroinitializer, ptr [[TMP20]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[TMP4]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
@@ -272,11 +269,10 @@ define void @no_signed_wrap_iv_via_btc(ptr %dst, i32 %N) mustprogress {
 ; CHECK-NEXT:    [[TMP3:%.*]] = add i32 [[SUB4]], [[INDEX]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = sext i32 [[TMP3]] to i64
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[DST]], i64 [[TMP4]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
-; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
+; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[CMP_N:%.*]] = icmp eq i32 [[TMP2]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[CMP_N]], label [[OUTER_LOOPEXIT:%.*]], label [[SCALAR_PH]]
diff --git a/llvm/test/Transforms/LoopVectorize/select-cmp-multiuse.ll b/llvm/test/Transforms/LoopVectorize/select-cmp-multiuse.ll
index a4b2f0c..64e12cc 100644
--- a/llvm/test/Transforms/LoopVectorize/select-cmp-multiuse.ll
+++ b/llvm/test/Transforms/LoopVectorize/select-cmp-multiuse.ll
@@ -31,8 +31,7 @@ define i32 @multi_user_cmp(ptr readonly %a, i64 noundef %n) {
 ; CHECK-VF4-IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-VF4-IC1-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-VF4-IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4-IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-VF4-IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-VF4-IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-VF4-IC1-NEXT:    [[TMP3:%.*]] = fcmp olt <4 x float> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-VF4-IC1-NEXT:    [[TMP4]] = or <4 x i1> [[VEC_PHI1]], [[TMP3]]
 ; CHECK-VF4-IC1-NEXT:    [[TMP5:%.*]] = xor <4 x i1> [[TMP3]], splat (i1 true)
@@ -89,9 +88,8 @@ define i32 @multi_user_cmp(ptr readonly %a, i64 noundef %n) {
 ; CHECK-VF4-IC2-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-VF4-IC2-NEXT:    [[VEC_PHI3:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-VF4-IC2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4-IC2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
 ; CHECK-VF4-IC2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 4
-; CHECK-VF4-IC2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
+; CHECK-VF4-IC2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
 ; CHECK-VF4-IC2-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
 ; CHECK-VF4-IC2-NEXT:    [[TMP6:%.*]] = fcmp olt <4 x float> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-VF4-IC2-NEXT:    [[TMP7:%.*]] = fcmp olt <4 x float> [[WIDE_LOAD4]], zeroinitializer
@@ -252,8 +250,7 @@ define i32 @multi_user_cmp_int(ptr readonly %a, i64 noundef %n) {
 ; CHECK-VF4-IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-VF4-IC1-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-VF4-IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4-IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-VF4-IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-VF4-IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-VF4-IC1-NEXT:    [[TMP3:%.*]] = icmp slt <4 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-VF4-IC1-NEXT:    [[TMP4]] = or <4 x i1> [[VEC_PHI1]], [[TMP3]]
 ; CHECK-VF4-IC1-NEXT:    [[TMP5:%.*]] = xor <4 x i1> [[TMP3]], splat (i1 true)
@@ -310,9 +307,8 @@ define i32 @multi_user_cmp_int(ptr readonly %a, i64 noundef %n) {
 ; CHECK-VF4-IC2-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-VF4-IC2-NEXT:    [[VEC_PHI3:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-VF4-IC2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4-IC2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 0
 ; CHECK-VF4-IC2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP2]], i32 4
-; CHECK-VF4-IC2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
+; CHECK-VF4-IC2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
 ; CHECK-VF4-IC2-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4
 ; CHECK-VF4-IC2-NEXT:    [[TMP6:%.*]] = icmp slt <4 x i32> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-VF4-IC2-NEXT:    [[TMP7:%.*]] = icmp slt <4 x i32> [[WIDE_LOAD4]], zeroinitializer
@@ -484,8 +480,7 @@ define i32 @multi_user_cmp_branch_use(ptr readonly %a, ptr %b, i64 noundef %n) {
 ; CHECK-VF4-IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP7:%.*]], [[PRED_STORE_CONTINUE8]] ]
 ; CHECK-VF4-IC1-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP5:%.*]], [[PRED_STORE_CONTINUE8]] ]
 ; CHECK-VF4-IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4-IC1-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
-; CHECK-VF4-IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP3]], align 4, !alias.scope [[META6:![0-9]+]]
+; CHECK-VF4-IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4, !alias.scope [[META6:![0-9]+]]
 ; CHECK-VF4-IC1-NEXT:    [[TMP4:%.*]] = fcmp olt <4 x float> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-VF4-IC1-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI2]], [[TMP4]]
 ; CHECK-VF4-IC1-NEXT:    [[TMP6:%.*]] = xor <4 x i1> [[TMP4]], splat (i1 true)
@@ -598,9 +593,8 @@ define i32 @multi_user_cmp_branch_use(ptr readonly %a, ptr %b, i64 noundef %n) {
 ; CHECK-VF4-IC2-NEXT:    [[VEC_PHI3:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[PRED_STORE_CONTINUE19]] ]
 ; CHECK-VF4-IC2-NEXT:    [[VEC_PHI4:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[PRED_STORE_CONTINUE19]] ]
 ; CHECK-VF4-IC2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4-IC2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP3]], i32 0
 ; CHECK-VF4-IC2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds float, ptr [[TMP3]], i32 4
-; CHECK-VF4-IC2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP5]], align 4, !alias.scope [[META6:![0-9]+]]
+; CHECK-VF4-IC2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP3]], align 4, !alias.scope [[META6:![0-9]+]]
 ; CHECK-VF4-IC2-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x float>, ptr [[TMP6]], align 4, !alias.scope [[META6]]
 ; CHECK-VF4-IC2-NEXT:    [[TMP7:%.*]] = fcmp olt <4 x float> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-VF4-IC2-NEXT:    [[TMP8:%.*]] = fcmp olt <4 x float> [[WIDE_LOAD5]], zeroinitializer
@@ -893,8 +887,7 @@ define i32 @multi_user_cmp_branch_use_and_outside_bb_use(ptr readonly %a, i64 no
 ; CHECK-VF4-IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP6:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-VF4-IC1-NEXT:    [[VEC_PHI1:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-VF4-IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4-IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-VF4-IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-VF4-IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-VF4-IC1-NEXT:    [[TMP3:%.*]] = fcmp olt <4 x float> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-VF4-IC1-NEXT:    [[TMP4]] = or <4 x i1> [[VEC_PHI1]], [[TMP3]]
 ; CHECK-VF4-IC1-NEXT:    [[TMP5:%.*]] = xor <4 x i1> [[TMP3]], splat (i1 true)
@@ -954,9 +947,8 @@ define i32 @multi_user_cmp_branch_use_and_outside_bb_use(ptr readonly %a, i64 no
 ; CHECK-VF4-IC2-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-VF4-IC2-NEXT:    [[VEC_PHI3:%.*]] = phi <4 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP9:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-VF4-IC2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[A]], i64 [[INDEX]]
-; CHECK-VF4-IC2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
 ; CHECK-VF4-IC2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 4
-; CHECK-VF4-IC2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
+; CHECK-VF4-IC2-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
 ; CHECK-VF4-IC2-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
 ; CHECK-VF4-IC2-NEXT:    [[TMP6:%.*]] = fcmp olt <4 x float> [[WIDE_LOAD]], zeroinitializer
 ; CHECK-VF4-IC2-NEXT:    [[TMP7:%.*]] = fcmp olt <4 x float> [[WIDE_LOAD4]], zeroinitializer
diff --git a/llvm/test/Transforms/LoopVectorize/select-cmp-predicated.ll b/llvm/test/Transforms/LoopVectorize/select-cmp-predicated.ll
index c17985d..8ab7ea8 100644
--- a/llvm/test/Transforms/LoopVectorize/select-cmp-predicated.ll
+++ b/llvm/test/Transforms/LoopVectorize/select-cmp-predicated.ll
@@ -16,8 +16,7 @@ define i32 @pred_select_const_i32_from_icmp(ptr noalias nocapture readonly %src1
 ; CHECK-VF2IC1-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_LOAD_CONTINUE2:.*]] ]
 ; CHECK-VF2IC1-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[PREDPHI:%.*]], %[[PRED_LOAD_CONTINUE2]] ]
 ; CHECK-VF2IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[SRC1]], i64 [[TMP0]]
-; CHECK-VF2IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-VF2IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
+; CHECK-VF2IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
 ; CHECK-VF2IC1-NEXT:    [[TMP4:%.*]] = icmp sgt <2 x i32> [[WIDE_LOAD]], splat (i32 35)
 ; CHECK-VF2IC1-NEXT:    [[TMP5:%.*]] = extractelement <2 x i1> [[TMP4]], i32 0
 ; CHECK-VF2IC1-NEXT:    br i1 [[TMP5]], label %[[PRED_LOAD_IF:.*]], label %[[PRED_LOAD_CONTINUE:.*]]
diff --git a/llvm/test/Transforms/LoopVectorize/select-cmp.ll b/llvm/test/Transforms/LoopVectorize/select-cmp.ll
index c5d3181..5e48b1f 100644
--- a/llvm/test/Transforms/LoopVectorize/select-cmp.ll
+++ b/llvm/test/Transforms/LoopVectorize/select-cmp.ll
@@ -17,8 +17,7 @@ define i32 @select_const_i32_from_icmp(ptr %v, i64 %n) {
 ; CHECK-VF4IC1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP5:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[V]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-VF4IC1-NEXT:    [[TMP4:%.*]] = icmp ne <4 x i32> [[WIDE_LOAD]], splat (i32 3)
 ; CHECK-VF4IC1-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI]], [[TMP4]]
 ; CHECK-VF4IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -64,11 +63,10 @@ define i32 @select_const_i32_from_icmp(ptr %v, i64 %n) {
 ; CHECK-VF4IC4-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP16:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC4-NEXT:    [[VEC_PHI3:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP17:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[V]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4
@@ -209,8 +207,7 @@ define i32 @select_const_i32_from_icmp2(ptr %v, i64 %n) {
 ; CHECK-VF4IC1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP4:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[V]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-VF4IC1-NEXT:    [[TMP3:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD]], splat (i32 3)
 ; CHECK-VF4IC1-NEXT:    [[TMP4]] = or <4 x i1> [[VEC_PHI]], [[TMP3]]
 ; CHECK-VF4IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -256,11 +253,10 @@ define i32 @select_const_i32_from_icmp2(ptr %v, i64 %n) {
 ; CHECK-VF4IC4-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP12:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC4-NEXT:    [[VEC_PHI3:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP13:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[V]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4
@@ -397,8 +393,7 @@ define i32 @select_i32_from_icmp(ptr %v, i32 %a, i32 %b, i64 %n) {
 ; CHECK-VF4IC1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP5:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[V]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-VF4IC1-NEXT:    [[TMP4:%.*]] = icmp ne <4 x i32> [[WIDE_LOAD]], splat (i32 3)
 ; CHECK-VF4IC1-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI]], [[TMP4]]
 ; CHECK-VF4IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -444,11 +439,10 @@ define i32 @select_i32_from_icmp(ptr %v, i32 %a, i32 %b, i64 %n) {
 ; CHECK-VF4IC4-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP16:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC4-NEXT:    [[VEC_PHI3:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP17:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[V]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i32>, ptr [[TMP3]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i32>, ptr [[TMP4]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4
@@ -589,8 +583,7 @@ define i32 @select_const_i32_from_fcmp_fast(ptr %v, i64 %n) {
 ; CHECK-VF4IC1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP5:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[V]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-VF4IC1-NEXT:    [[TMP4:%.*]] = fcmp fast one <4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00)
 ; CHECK-VF4IC1-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI]], [[TMP4]]
 ; CHECK-VF4IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -636,11 +629,10 @@ define i32 @select_const_i32_from_fcmp_fast(ptr %v, i64 %n) {
 ; CHECK-VF4IC4-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP16:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC4-NEXT:    [[VEC_PHI3:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP17:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[V]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
@@ -781,8 +773,7 @@ define i32 @select_const_i32_from_fcmp(ptr %v, i64 %n) {
 ; CHECK-VF4IC1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[VEC_PHI:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP5:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC1-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[V]], i64 [[INDEX]]
-; CHECK-VF4IC1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-VF4IC1-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-VF4IC1-NEXT:    [[TMP4:%.*]] = fcmp one <4 x float> [[WIDE_LOAD]], splat (float 3.000000e+00)
 ; CHECK-VF4IC1-NEXT:    [[TMP5]] = or <4 x i1> [[VEC_PHI]], [[TMP4]]
 ; CHECK-VF4IC1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
@@ -828,11 +819,10 @@ define i32 @select_const_i32_from_fcmp(ptr %v, i64 %n) {
 ; CHECK-VF4IC4-NEXT:    [[VEC_PHI2:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP16:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC4-NEXT:    [[VEC_PHI3:%.*]] = phi <4 x i1> [ zeroinitializer, %[[VECTOR_PH]] ], [ [[TMP17:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[V]], i64 [[INDEX]]
-; CHECK-VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
 ; CHECK-VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 4
 ; CHECK-VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 8
 ; CHECK-VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 12
-; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
 ; CHECK-VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x float>, ptr [[TMP5]], align 4
diff --git a/llvm/test/Transforms/LoopVectorize/select-neg-cond.ll b/llvm/test/Transforms/LoopVectorize/select-neg-cond.ll
index def239e..285c674 100644
--- a/llvm/test/Transforms/LoopVectorize/select-neg-cond.ll
+++ b/llvm/test/Transforms/LoopVectorize/select-neg-cond.ll
@@ -11,18 +11,17 @@ define void @neg_cond(ptr noalias %p, ptr noalias %q) {
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i32, ptr [[P]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i32, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP0]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp ne <4 x i32> [[WIDE_LOAD]], splat (i32 42)
 ; CHECK-NEXT:    [[TMP4:%.*]] = select <4 x i1> [[TMP3]], <4 x i32> splat (i32 42), <4 x i32> splat (i32 43)
-; CHECK-NEXT:    store <4 x i32> [[TMP4]], ptr [[TMP1]], align 4
+; CHECK-NEXT:    store <4 x i32> [[TMP4]], ptr [[TMP0]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/select-reduction-start-value-may-be-undef-or-poison.ll b/llvm/test/Transforms/LoopVectorize/select-reduction-start-value-may-be-undef-or-poison.ll
index d51f64f..37d75ff 100644
--- a/llvm/test/Transforms/LoopVectorize/select-reduction-start-value-may-be-undef-or-poison.ll
+++ b/llvm/test/Transforms/LoopVectorize/select-reduction-start-value-may-be-undef-or-poison.ll
@@ -11,24 +11,23 @@ define i64 @pr62565_incoming_value_known_undef(i64 %a, ptr %src) {
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP2:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i32 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i32 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = icmp ne <2 x i32> [[WIDE_LOAD]], splat (i32 1)
-; CHECK-NEXT:    [[TMP3]] = or <2 x i1> [[VEC_PHI]], [[TMP2]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP0]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ne <2 x i32> [[WIDE_LOAD]], splat (i32 1)
+; CHECK-NEXT:    [[TMP2]] = or <2 x i1> [[VEC_PHI]], [[TMP1]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
-; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], 32
-; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 32
+; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v2i1(<2 x i1> [[TMP3]])
-; CHECK-NEXT:    [[TMP6:%.*]] = freeze i1 [[TMP5]]
-; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP6]], i64 [[A]], i64 undef
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    [[TMP4:%.*]] = call i1 @llvm.vector.reduce.or.v2i1(<2 x i1> [[TMP2]])
+; CHECK-NEXT:    [[TMP5:%.*]] = freeze i1 [[TMP4]]
+; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP5]], i64 [[A]], i64 undef
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 33, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ], [ undef, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ undef, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[LOOP]] ]
@@ -72,24 +71,23 @@ define i64 @pr62565_incoming_value_known_poison(i64 %a, ptr %src) {
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP2:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i32 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i32 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = icmp ne <2 x i32> [[WIDE_LOAD]], splat (i32 1)
-; CHECK-NEXT:    [[TMP3]] = or <2 x i1> [[VEC_PHI]], [[TMP2]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP0]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ne <2 x i32> [[WIDE_LOAD]], splat (i32 1)
+; CHECK-NEXT:    [[TMP2]] = or <2 x i1> [[VEC_PHI]], [[TMP1]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
-; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], 32
-; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 32
+; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v2i1(<2 x i1> [[TMP3]])
-; CHECK-NEXT:    [[TMP6:%.*]] = freeze i1 [[TMP5]]
-; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP6]], i64 [[A]], i64 poison
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    [[TMP4:%.*]] = call i1 @llvm.vector.reduce.or.v2i1(<2 x i1> [[TMP2]])
+; CHECK-NEXT:    [[TMP5:%.*]] = freeze i1 [[TMP4]]
+; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP5]], i64 [[A]], i64 poison
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 33, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ], [ poison, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ poison, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[LOOP]] ]
@@ -133,24 +131,23 @@ define i64 @pr62565_incoming_value_may_be_poison(i64 %a, ptr %src, i64 %start) {
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
-; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP3:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_PHI:%.*]] = phi <2 x i1> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP2:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i32 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i32 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
-; CHECK-NEXT:    [[TMP2:%.*]] = icmp ne <2 x i32> [[WIDE_LOAD]], splat (i32 1)
-; CHECK-NEXT:    [[TMP3]] = or <2 x i1> [[VEC_PHI]], [[TMP2]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP0]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = icmp ne <2 x i32> [[WIDE_LOAD]], splat (i32 1)
+; CHECK-NEXT:    [[TMP2]] = or <2 x i1> [[VEC_PHI]], [[TMP1]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
-; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], 32
-; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 32
+; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v2i1(<2 x i1> [[TMP3]])
-; CHECK-NEXT:    [[TMP6:%.*]] = freeze i1 [[TMP5]]
-; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP6]], i64 [[A]], i64 [[START]]
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    [[TMP4:%.*]] = call i1 @llvm.vector.reduce.or.v2i1(<2 x i1> [[TMP2]])
+; CHECK-NEXT:    [[TMP5:%.*]] = freeze i1 [[TMP4]]
+; CHECK-NEXT:    [[RDX_SELECT:%.*]] = select i1 [[TMP5]], i64 [[A]], i64 [[START]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 33, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[RDX_SELECT]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i64 [ [[START]], [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[ADD:%.*]], [[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/select-with-fastflags.ll b/llvm/test/Transforms/LoopVectorize/select-with-fastflags.ll
index 56cfc31..e571f69 100644
--- a/llvm/test/Transforms/LoopVectorize/select-with-fastflags.ll
+++ b/llvm/test/Transforms/LoopVectorize/select-with-fastflags.ll
@@ -15,17 +15,14 @@ define void @select_with_fastmath_flags(ptr noalias %a, ptr noalias %b, ptr noal
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds nuw float, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds nuw float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds nuw float, ptr [[C]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds nuw float, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x float>, ptr [[TMP4]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x float>, ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = fcmp fast ogt <4 x float> [[WIDE_LOAD]], [[WIDE_LOAD1]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = fadd fast <4 x float> [[WIDE_LOAD]], splat (float 1.000000e+01)
 ; CHECK-NEXT:    [[TMP7:%.*]] = select fast <4 x i1> [[TMP5]], <4 x float> [[TMP6]], <4 x float> [[WIDE_LOAD1]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds nuw float, ptr [[A]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds nuw float, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <4 x float> [[TMP7]], ptr [[TMP9]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP7]], ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 4
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP10]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/single-early-exit-interleave-hint.ll b/llvm/test/Transforms/LoopVectorize/single-early-exit-interleave-hint.ll
index de8a3c5..c648bed 100644
--- a/llvm/test/Transforms/LoopVectorize/single-early-exit-interleave-hint.ll
+++ b/llvm/test/Transforms/LoopVectorize/single-early-exit-interleave-hint.ll
@@ -1,13 +1,8 @@
 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
-; REQUIRES: asserts
-; RUN: opt -p loop-vectorize -enable-early-exit-vectorization -force-vector-width=4 \
-; RUN:   -debug-only=loop-vectorize -S %s 2>%t | FileCheck --check-prefix=VF4IC4 %s
-; RUN: cat %t | FileCheck --check-prefix=DEBUG %s
+; RUN: opt -p loop-vectorize -enable-early-exit-vectorization -force-vector-width=4 -S %s | FileCheck --check-prefix=VF4IC4 %s
 
 declare void @init_mem(ptr, i64);
 
-; DEBUG: Interleaving not supported for loops with uncountable early exits
-
 define i64 @multi_exiting_to_different_exits_live_in_exit_values() {
 ; VF4IC4-LABEL: define i64 @multi_exiting_to_different_exits_live_in_exit_values() {
 ; VF4IC4-NEXT:  [[ENTRY:.*]]:
@@ -19,22 +14,33 @@ define i64 @multi_exiting_to_different_exits_live_in_exit_values() {
 ; VF4IC4:       [[VECTOR_BODY]]:
 ; VF4IC4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF4IC4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[INDEX]]
-; VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 0
-; VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
+; VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 4
+; VF4IC4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 8
+; VF4IC4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 12
+; VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP0]], align 4
+; VF4IC4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
+; VF4IC4-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP12]], align 4
+; VF4IC4-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i32>, ptr [[TMP13]], align 4
 ; VF4IC4-NEXT:    [[TMP2:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD]], splat (i32 10)
-; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; VF4IC4-NEXT:    [[TMP3:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP2]])
+; VF4IC4-NEXT:    [[TMP6:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD1]], splat (i32 10)
+; VF4IC4-NEXT:    [[TMP7:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD2]], splat (i32 10)
+; VF4IC4-NEXT:    [[TMP8:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD3]], splat (i32 10)
+; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
+; VF4IC4-NEXT:    [[TMP9:%.*]] = or <4 x i1> [[TMP2]], [[TMP6]]
+; VF4IC4-NEXT:    [[TMP10:%.*]] = or <4 x i1> [[TMP9]], [[TMP7]]
+; VF4IC4-NEXT:    [[TMP11:%.*]] = or <4 x i1> [[TMP10]], [[TMP8]]
+; VF4IC4-NEXT:    [[TMP3:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP11]])
 ; VF4IC4-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128
 ; VF4IC4-NEXT:    [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]
 ; VF4IC4-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_SPLIT:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; VF4IC4:       [[MIDDLE_SPLIT]]:
 ; VF4IC4-NEXT:    br i1 [[TMP3]], label %[[VECTOR_EARLY_EXIT:.*]], label %[[MIDDLE_BLOCK:.*]]
 ; VF4IC4:       [[MIDDLE_BLOCK]]:
-; VF4IC4-NEXT:    br i1 true, label %[[E2:.*]], label %[[SCALAR_PH]]
+; VF4IC4-NEXT:    br label %[[E2:.*]]
 ; VF4IC4:       [[VECTOR_EARLY_EXIT]]:
 ; VF4IC4-NEXT:    br label %[[E1:.*]]
 ; VF4IC4:       [[SCALAR_PH]]:
-; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 128, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; VF4IC4-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; VF4IC4:       [[LOOP_HEADER]]:
 ; VF4IC4-NEXT:    [[IV:%.*]] = phi i64 [ [[INC:%.*]], %[[LOOP_LATCH:.*]] ], [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/single-early-exit-interleave.ll b/llvm/test/Transforms/LoopVectorize/single-early-exit-interleave.ll
index 0f99ed5..3f51c72 100644
--- a/llvm/test/Transforms/LoopVectorize/single-early-exit-interleave.ll
+++ b/llvm/test/Transforms/LoopVectorize/single-early-exit-interleave.ll
@@ -14,22 +14,33 @@ define i64 @multi_exiting_to_different_exits_live_in_exit_values() {
 ; VF4IC4:       vector.body:
 ; VF4IC4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; VF4IC4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[SRC]], i64 [[INDEX]]
-; VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 0
-; VF4IC4-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
-; VF4IC4-NEXT:    [[TMP8:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD3]], splat (i32 10)
-; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; VF4IC4-NEXT:    [[TMP3:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP8]])
+; VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 4
+; VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 8
+; VF4IC4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 12
+; VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP0]], align 4
+; VF4IC4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
+; VF4IC4-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP2]], align 4
+; VF4IC4-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i32>, ptr [[TMP12]], align 4
+; VF4IC4-NEXT:    [[TMP8:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD]], splat (i32 10)
+; VF4IC4-NEXT:    [[TMP6:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD1]], splat (i32 10)
+; VF4IC4-NEXT:    [[TMP7:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD2]], splat (i32 10)
+; VF4IC4-NEXT:    [[TMP14:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD3]], splat (i32 10)
+; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
+; VF4IC4-NEXT:    [[TMP9:%.*]] = or <4 x i1> [[TMP8]], [[TMP6]]
+; VF4IC4-NEXT:    [[TMP10:%.*]] = or <4 x i1> [[TMP9]], [[TMP7]]
+; VF4IC4-NEXT:    [[TMP11:%.*]] = or <4 x i1> [[TMP10]], [[TMP14]]
+; VF4IC4-NEXT:    [[TMP3:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP11]])
 ; VF4IC4-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 128
 ; VF4IC4-NEXT:    [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]
 ; VF4IC4-NEXT:    br i1 [[TMP5]], label [[MIDDLE_SPLIT:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; VF4IC4:       middle.split:
 ; VF4IC4-NEXT:    br i1 [[TMP3]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; VF4IC4:       middle.block:
-; VF4IC4-NEXT:    br i1 true, label [[E2:%.*]], label [[SCALAR_PH]]
+; VF4IC4-NEXT:    br label [[E2:%.*]]
 ; VF4IC4:       vector.early.exit:
 ; VF4IC4-NEXT:    br label [[E1:%.*]]
 ; VF4IC4:       scalar.ph:
-; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 128, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; VF4IC4-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; VF4IC4:       loop.header:
 ; VF4IC4-NEXT:    [[IV:%.*]] = phi i64 [ [[INC:%.*]], [[LOOP_LATCH:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -88,28 +99,57 @@ define i64 @same_exit_block_pre_inc_use1() {
 ; VF4IC4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; VF4IC4-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX]]
 ; VF4IC4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 0
-; VF4IC4-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
-; VF4IC4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; VF4IC4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[TMP18]], i32 0
-; VF4IC4-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i8>, ptr [[TMP19]], align 1
+; VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 4
+; VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 8
+; VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 12
+; VF4IC4-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP0]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
+; VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
+; VF4IC4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 4
+; VF4IC4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 8
+; VF4IC4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 12
+; VF4IC4-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i8>, ptr [[TMP17]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i8>, ptr [[TMP18]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x i8>, ptr [[TMP19]], align 1
 ; VF4IC4-NEXT:    [[TMP13:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD3]], [[WIDE_LOAD7]]
-; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; VF4IC4-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP13]])
+; VF4IC4-NEXT:    [[TMP11:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD1]], [[WIDE_LOAD5]]
+; VF4IC4-NEXT:    [[TMP12:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD2]], [[WIDE_LOAD6]]
+; VF4IC4-NEXT:    [[TMP35:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD4]], [[WIDE_LOAD8]]
+; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
+; VF4IC4-NEXT:    [[TMP14:%.*]] = or <4 x i1> [[TMP13]], [[TMP11]]
+; VF4IC4-NEXT:    [[TMP15:%.*]] = or <4 x i1> [[TMP14]], [[TMP12]]
+; VF4IC4-NEXT:    [[TMP16:%.*]] = or <4 x i1> [[TMP15]], [[TMP35]]
+; VF4IC4-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP16]])
 ; VF4IC4-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; VF4IC4-NEXT:    [[TMP7:%.*]] = or i1 [[TMP5]], [[TMP6]]
 ; VF4IC4-NEXT:    br i1 [[TMP7]], label [[MIDDLE_SPLIT:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; VF4IC4:       middle.split:
 ; VF4IC4-NEXT:    br i1 [[TMP5]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; VF4IC4:       middle.block:
-; VF4IC4-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; VF4IC4-NEXT:    br label [[LOOP_END:%.*]]
 ; VF4IC4:       vector.early.exit:
-; VF4IC4-NEXT:    [[TMP8:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP13]], i1 true)
+; VF4IC4-NEXT:    [[TMP20:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP35]], i1 true)
+; VF4IC4-NEXT:    [[TMP21:%.*]] = add i64 12, [[TMP20]]
+; VF4IC4-NEXT:    [[TMP22:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP12]], i1 true)
+; VF4IC4-NEXT:    [[TMP23:%.*]] = add i64 8, [[TMP22]]
+; VF4IC4-NEXT:    [[TMP24:%.*]] = icmp ne i64 [[TMP22]], 4
+; VF4IC4-NEXT:    [[TMP25:%.*]] = select i1 [[TMP24]], i64 [[TMP23]], i64 [[TMP21]]
+; VF4IC4-NEXT:    [[TMP26:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP11]], i1 true)
+; VF4IC4-NEXT:    [[TMP27:%.*]] = add i64 4, [[TMP26]]
+; VF4IC4-NEXT:    [[TMP28:%.*]] = icmp ne i64 [[TMP26]], 4
+; VF4IC4-NEXT:    [[TMP29:%.*]] = select i1 [[TMP28]], i64 [[TMP27]], i64 [[TMP25]]
+; VF4IC4-NEXT:    [[TMP30:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP13]], i1 true)
+; VF4IC4-NEXT:    [[TMP31:%.*]] = add i64 0, [[TMP30]]
+; VF4IC4-NEXT:    [[TMP32:%.*]] = icmp ne i64 [[TMP30]], 4
+; VF4IC4-NEXT:    [[TMP8:%.*]] = select i1 [[TMP32]], i64 [[TMP31]], i64 [[TMP29]]
 ; VF4IC4-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX]], [[TMP8]]
 ; VF4IC4-NEXT:    [[TMP10:%.*]] = add i64 3, [[TMP9]]
 ; VF4IC4-NEXT:    br label [[LOOP_END]]
 ; VF4IC4:       scalar.ph:
-; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; VF4IC4-NEXT:    br label [[LOOP:%.*]]
 ; VF4IC4:       loop:
 ; VF4IC4-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -161,30 +201,53 @@ define ptr @same_exit_block_pre_inc_use1_ivptr() {
 ; VF4IC4-NEXT:    [[PTREND:%.*]] = getelementptr i8, ptr [[P1]], i64 1024
 ; VF4IC4-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; VF4IC4:       vector.ph:
-; VF4IC4-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[P1]], i64 1024
 ; VF4IC4-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; VF4IC4:       vector.body:
 ; VF4IC4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; VF4IC4-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[P1]], i64 [[INDEX]]
-; VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
-; VF4IC4-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
-; VF4IC4-NEXT:    [[TMP17:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD3]], splat (i8 72)
-; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; VF4IC4-NEXT:    [[TMP3:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP17]])
-; VF4IC4-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
-; VF4IC4-NEXT:    [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]
-; VF4IC4-NEXT:    br i1 [[TMP5]], label [[MIDDLE_SPLIT:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; VF4IC4-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 4
+; VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 8
+; VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 12
+; VF4IC4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[NEXT_GEP]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i8>, ptr [[TMP0]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; VF4IC4-NEXT:    [[TMP17:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], splat (i8 72)
+; VF4IC4-NEXT:    [[TMP14:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD1]], splat (i8 72)
+; VF4IC4-NEXT:    [[TMP28:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD2]], splat (i8 72)
+; VF4IC4-NEXT:    [[TMP29:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD3]], splat (i8 72)
+; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
+; VF4IC4-NEXT:    [[TMP13:%.*]] = or <4 x i1> [[TMP17]], [[TMP14]]
+; VF4IC4-NEXT:    [[TMP31:%.*]] = or <4 x i1> [[TMP13]], [[TMP28]]
+; VF4IC4-NEXT:    [[TMP9:%.*]] = or <4 x i1> [[TMP31]], [[TMP29]]
+; VF4IC4-NEXT:    [[TMP10:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP9]])
+; VF4IC4-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
+; VF4IC4-NEXT:    [[TMP12:%.*]] = or i1 [[TMP10]], [[TMP11]]
+; VF4IC4-NEXT:    br i1 [[TMP12]], label [[MIDDLE_SPLIT:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; VF4IC4:       middle.split:
-; VF4IC4-NEXT:    br i1 [[TMP3]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
+; VF4IC4-NEXT:    br i1 [[TMP10]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; VF4IC4:       middle.block:
-; VF4IC4-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; VF4IC4-NEXT:    br label [[LOOP_END:%.*]]
 ; VF4IC4:       vector.early.exit:
-; VF4IC4-NEXT:    [[TMP6:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP17]], i1 true)
+; VF4IC4-NEXT:    [[TMP15:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP29]], i1 true)
+; VF4IC4-NEXT:    [[TMP16:%.*]] = add i64 12, [[TMP15]]
+; VF4IC4-NEXT:    [[TMP30:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP28]], i1 true)
+; VF4IC4-NEXT:    [[TMP18:%.*]] = add i64 8, [[TMP30]]
+; VF4IC4-NEXT:    [[TMP19:%.*]] = icmp ne i64 [[TMP30]], 4
+; VF4IC4-NEXT:    [[TMP20:%.*]] = select i1 [[TMP19]], i64 [[TMP18]], i64 [[TMP16]]
+; VF4IC4-NEXT:    [[TMP21:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP14]], i1 true)
+; VF4IC4-NEXT:    [[TMP22:%.*]] = add i64 4, [[TMP21]]
+; VF4IC4-NEXT:    [[TMP23:%.*]] = icmp ne i64 [[TMP21]], 4
+; VF4IC4-NEXT:    [[TMP24:%.*]] = select i1 [[TMP23]], i64 [[TMP22]], i64 [[TMP20]]
+; VF4IC4-NEXT:    [[TMP25:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP17]], i1 true)
+; VF4IC4-NEXT:    [[TMP26:%.*]] = add i64 0, [[TMP25]]
+; VF4IC4-NEXT:    [[TMP27:%.*]] = icmp ne i64 [[TMP25]], 4
+; VF4IC4-NEXT:    [[TMP6:%.*]] = select i1 [[TMP27]], i64 [[TMP26]], i64 [[TMP24]]
 ; VF4IC4-NEXT:    [[TMP7:%.*]] = add i64 [[INDEX]], [[TMP6]]
 ; VF4IC4-NEXT:    [[TMP8:%.*]] = getelementptr i8, ptr [[P1]], i64 [[TMP7]]
 ; VF4IC4-NEXT:    br label [[LOOP_END]]
 ; VF4IC4:       scalar.ph:
-; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi ptr [ [[TMP0]], [[MIDDLE_BLOCK]] ], [ [[P1]], [[ENTRY:%.*]] ]
+; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi ptr [ [[P1]], [[ENTRY:%.*]] ]
 ; VF4IC4-NEXT:    br label [[LOOP:%.*]]
 ; VF4IC4:       loop:
 ; VF4IC4-NEXT:    [[PTR:%.*]] = phi ptr [ [[PTR_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -235,28 +298,57 @@ define i64 @same_exit_block_post_inc_use() {
 ; VF4IC4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; VF4IC4-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX]]
 ; VF4IC4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 0
-; VF4IC4-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
-; VF4IC4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; VF4IC4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[TMP18]], i32 0
-; VF4IC4-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i8>, ptr [[TMP19]], align 1
+; VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 4
+; VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 8
+; VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 12
+; VF4IC4-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP0]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
+; VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
+; VF4IC4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 4
+; VF4IC4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 8
+; VF4IC4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 12
+; VF4IC4-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i8>, ptr [[TMP17]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i8>, ptr [[TMP18]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x i8>, ptr [[TMP19]], align 1
 ; VF4IC4-NEXT:    [[TMP13:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD3]], [[WIDE_LOAD7]]
-; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; VF4IC4-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP13]])
+; VF4IC4-NEXT:    [[TMP11:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD1]], [[WIDE_LOAD5]]
+; VF4IC4-NEXT:    [[TMP12:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD2]], [[WIDE_LOAD6]]
+; VF4IC4-NEXT:    [[TMP35:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD4]], [[WIDE_LOAD8]]
+; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
+; VF4IC4-NEXT:    [[TMP14:%.*]] = or <4 x i1> [[TMP13]], [[TMP11]]
+; VF4IC4-NEXT:    [[TMP15:%.*]] = or <4 x i1> [[TMP14]], [[TMP12]]
+; VF4IC4-NEXT:    [[TMP16:%.*]] = or <4 x i1> [[TMP15]], [[TMP35]]
+; VF4IC4-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP16]])
 ; VF4IC4-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; VF4IC4-NEXT:    [[TMP7:%.*]] = or i1 [[TMP5]], [[TMP6]]
 ; VF4IC4-NEXT:    br i1 [[TMP7]], label [[MIDDLE_SPLIT:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; VF4IC4:       middle.split:
 ; VF4IC4-NEXT:    br i1 [[TMP5]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; VF4IC4:       middle.block:
-; VF4IC4-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; VF4IC4-NEXT:    br label [[LOOP_END:%.*]]
 ; VF4IC4:       vector.early.exit:
-; VF4IC4-NEXT:    [[TMP8:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP13]], i1 true)
+; VF4IC4-NEXT:    [[TMP20:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP35]], i1 true)
+; VF4IC4-NEXT:    [[TMP21:%.*]] = add i64 12, [[TMP20]]
+; VF4IC4-NEXT:    [[TMP22:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP12]], i1 true)
+; VF4IC4-NEXT:    [[TMP23:%.*]] = add i64 8, [[TMP22]]
+; VF4IC4-NEXT:    [[TMP24:%.*]] = icmp ne i64 [[TMP22]], 4
+; VF4IC4-NEXT:    [[TMP25:%.*]] = select i1 [[TMP24]], i64 [[TMP23]], i64 [[TMP21]]
+; VF4IC4-NEXT:    [[TMP26:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP11]], i1 true)
+; VF4IC4-NEXT:    [[TMP27:%.*]] = add i64 4, [[TMP26]]
+; VF4IC4-NEXT:    [[TMP28:%.*]] = icmp ne i64 [[TMP26]], 4
+; VF4IC4-NEXT:    [[TMP29:%.*]] = select i1 [[TMP28]], i64 [[TMP27]], i64 [[TMP25]]
+; VF4IC4-NEXT:    [[TMP30:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP13]], i1 true)
+; VF4IC4-NEXT:    [[TMP31:%.*]] = add i64 0, [[TMP30]]
+; VF4IC4-NEXT:    [[TMP32:%.*]] = icmp ne i64 [[TMP30]], 4
+; VF4IC4-NEXT:    [[TMP8:%.*]] = select i1 [[TMP32]], i64 [[TMP31]], i64 [[TMP29]]
 ; VF4IC4-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX]], [[TMP8]]
 ; VF4IC4-NEXT:    [[TMP10:%.*]] = add i64 3, [[TMP9]]
 ; VF4IC4-NEXT:    br label [[LOOP_END]]
 ; VF4IC4:       scalar.ph:
-; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; VF4IC4-NEXT:    br label [[LOOP:%.*]]
 ; VF4IC4:       loop:
 ; VF4IC4-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -314,28 +406,57 @@ define i64 @diff_exit_block_pre_inc_use1() {
 ; VF4IC4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; VF4IC4-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX]]
 ; VF4IC4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 0
-; VF4IC4-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
-; VF4IC4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; VF4IC4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[TMP18]], i32 0
-; VF4IC4-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i8>, ptr [[TMP19]], align 1
+; VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 4
+; VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 8
+; VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 12
+; VF4IC4-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP0]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
+; VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
+; VF4IC4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 4
+; VF4IC4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 8
+; VF4IC4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 12
+; VF4IC4-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i8>, ptr [[TMP17]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i8>, ptr [[TMP18]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x i8>, ptr [[TMP19]], align 1
 ; VF4IC4-NEXT:    [[TMP13:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD3]], [[WIDE_LOAD7]]
-; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; VF4IC4-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP13]])
+; VF4IC4-NEXT:    [[TMP11:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD1]], [[WIDE_LOAD5]]
+; VF4IC4-NEXT:    [[TMP12:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD2]], [[WIDE_LOAD6]]
+; VF4IC4-NEXT:    [[TMP35:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD4]], [[WIDE_LOAD8]]
+; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
+; VF4IC4-NEXT:    [[TMP14:%.*]] = or <4 x i1> [[TMP13]], [[TMP11]]
+; VF4IC4-NEXT:    [[TMP15:%.*]] = or <4 x i1> [[TMP14]], [[TMP12]]
+; VF4IC4-NEXT:    [[TMP16:%.*]] = or <4 x i1> [[TMP15]], [[TMP35]]
+; VF4IC4-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP16]])
 ; VF4IC4-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; VF4IC4-NEXT:    [[TMP7:%.*]] = or i1 [[TMP5]], [[TMP6]]
 ; VF4IC4-NEXT:    br i1 [[TMP7]], label [[MIDDLE_SPLIT:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; VF4IC4:       middle.split:
 ; VF4IC4-NEXT:    br i1 [[TMP5]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; VF4IC4:       middle.block:
-; VF4IC4-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; VF4IC4-NEXT:    br label [[LOOP_END:%.*]]
 ; VF4IC4:       vector.early.exit:
-; VF4IC4-NEXT:    [[TMP8:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP13]], i1 true)
+; VF4IC4-NEXT:    [[TMP20:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP35]], i1 true)
+; VF4IC4-NEXT:    [[TMP21:%.*]] = add i64 12, [[TMP20]]
+; VF4IC4-NEXT:    [[TMP22:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP12]], i1 true)
+; VF4IC4-NEXT:    [[TMP23:%.*]] = add i64 8, [[TMP22]]
+; VF4IC4-NEXT:    [[TMP24:%.*]] = icmp ne i64 [[TMP22]], 4
+; VF4IC4-NEXT:    [[TMP25:%.*]] = select i1 [[TMP24]], i64 [[TMP23]], i64 [[TMP21]]
+; VF4IC4-NEXT:    [[TMP26:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP11]], i1 true)
+; VF4IC4-NEXT:    [[TMP27:%.*]] = add i64 4, [[TMP26]]
+; VF4IC4-NEXT:    [[TMP28:%.*]] = icmp ne i64 [[TMP26]], 4
+; VF4IC4-NEXT:    [[TMP29:%.*]] = select i1 [[TMP28]], i64 [[TMP27]], i64 [[TMP25]]
+; VF4IC4-NEXT:    [[TMP30:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP13]], i1 true)
+; VF4IC4-NEXT:    [[TMP31:%.*]] = add i64 0, [[TMP30]]
+; VF4IC4-NEXT:    [[TMP32:%.*]] = icmp ne i64 [[TMP30]], 4
+; VF4IC4-NEXT:    [[TMP8:%.*]] = select i1 [[TMP32]], i64 [[TMP31]], i64 [[TMP29]]
 ; VF4IC4-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX]], [[TMP8]]
 ; VF4IC4-NEXT:    [[TMP10:%.*]] = add i64 3, [[TMP9]]
 ; VF4IC4-NEXT:    br label [[LOOP_EARLY_EXIT:%.*]]
 ; VF4IC4:       scalar.ph:
-; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; VF4IC4-NEXT:    br label [[LOOP:%.*]]
 ; VF4IC4:       loop:
 ; VF4IC4-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -400,28 +521,57 @@ define i64 @diff_exit_block_post_inc_use1() {
 ; VF4IC4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; VF4IC4-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX]]
 ; VF4IC4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 0
-; VF4IC4-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
-; VF4IC4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; VF4IC4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[TMP18]], i32 0
-; VF4IC4-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i8>, ptr [[TMP19]], align 1
+; VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 4
+; VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 8
+; VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 12
+; VF4IC4-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP0]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
+; VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
+; VF4IC4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 4
+; VF4IC4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 8
+; VF4IC4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 12
+; VF4IC4-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i8>, ptr [[TMP17]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i8>, ptr [[TMP18]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x i8>, ptr [[TMP19]], align 1
 ; VF4IC4-NEXT:    [[TMP13:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD3]], [[WIDE_LOAD7]]
-; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; VF4IC4-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP13]])
+; VF4IC4-NEXT:    [[TMP11:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD1]], [[WIDE_LOAD5]]
+; VF4IC4-NEXT:    [[TMP12:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD2]], [[WIDE_LOAD6]]
+; VF4IC4-NEXT:    [[TMP35:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD4]], [[WIDE_LOAD8]]
+; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
+; VF4IC4-NEXT:    [[TMP14:%.*]] = or <4 x i1> [[TMP13]], [[TMP11]]
+; VF4IC4-NEXT:    [[TMP15:%.*]] = or <4 x i1> [[TMP14]], [[TMP12]]
+; VF4IC4-NEXT:    [[TMP16:%.*]] = or <4 x i1> [[TMP15]], [[TMP35]]
+; VF4IC4-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP16]])
 ; VF4IC4-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; VF4IC4-NEXT:    [[TMP7:%.*]] = or i1 [[TMP5]], [[TMP6]]
 ; VF4IC4-NEXT:    br i1 [[TMP7]], label [[MIDDLE_SPLIT:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; VF4IC4:       middle.split:
 ; VF4IC4-NEXT:    br i1 [[TMP5]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; VF4IC4:       middle.block:
-; VF4IC4-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; VF4IC4-NEXT:    br label [[LOOP_END:%.*]]
 ; VF4IC4:       vector.early.exit:
-; VF4IC4-NEXT:    [[TMP8:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP13]], i1 true)
+; VF4IC4-NEXT:    [[TMP20:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP35]], i1 true)
+; VF4IC4-NEXT:    [[TMP21:%.*]] = add i64 12, [[TMP20]]
+; VF4IC4-NEXT:    [[TMP22:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP12]], i1 true)
+; VF4IC4-NEXT:    [[TMP23:%.*]] = add i64 8, [[TMP22]]
+; VF4IC4-NEXT:    [[TMP24:%.*]] = icmp ne i64 [[TMP22]], 4
+; VF4IC4-NEXT:    [[TMP25:%.*]] = select i1 [[TMP24]], i64 [[TMP23]], i64 [[TMP21]]
+; VF4IC4-NEXT:    [[TMP26:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP11]], i1 true)
+; VF4IC4-NEXT:    [[TMP27:%.*]] = add i64 4, [[TMP26]]
+; VF4IC4-NEXT:    [[TMP28:%.*]] = icmp ne i64 [[TMP26]], 4
+; VF4IC4-NEXT:    [[TMP29:%.*]] = select i1 [[TMP28]], i64 [[TMP27]], i64 [[TMP25]]
+; VF4IC4-NEXT:    [[TMP30:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP13]], i1 true)
+; VF4IC4-NEXT:    [[TMP31:%.*]] = add i64 0, [[TMP30]]
+; VF4IC4-NEXT:    [[TMP32:%.*]] = icmp ne i64 [[TMP30]], 4
+; VF4IC4-NEXT:    [[TMP8:%.*]] = select i1 [[TMP32]], i64 [[TMP31]], i64 [[TMP29]]
 ; VF4IC4-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX]], [[TMP8]]
 ; VF4IC4-NEXT:    [[TMP10:%.*]] = add i64 3, [[TMP9]]
 ; VF4IC4-NEXT:    br label [[LOOP_EARLY_EXIT:%.*]]
 ; VF4IC4:       scalar.ph:
-; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; VF4IC4-NEXT:    br label [[LOOP:%.*]]
 ; VF4IC4:       loop:
 ; VF4IC4-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -488,30 +638,73 @@ define i64 @same_exit_block_pre_inc_use1_reverse() {
 ; VF4IC4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
 ; VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 0
 ; VF4IC4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 -3
+; VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 -4
+; VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 -3
+; VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 -8
+; VF4IC4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 -3
+; VF4IC4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 -12
+; VF4IC4-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i8, ptr [[TMP18]], i32 -3
 ; VF4IC4-NEXT:    [[WIDE_LOAD13:%.*]] = load <4 x i8>, ptr [[TMP17]], align 1
 ; VF4IC4-NEXT:    [[REVERSE14:%.*]] = shufflevector <4 x i8> [[WIDE_LOAD13]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; VF4IC4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; VF4IC4-NEXT:    [[REVERSE2:%.*]] = shufflevector <4 x i8> [[WIDE_LOAD1]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; VF4IC4-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP6]], align 1
+; VF4IC4-NEXT:    [[REVERSE4:%.*]] = shufflevector <4 x i8> [[WIDE_LOAD3]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i8>, ptr [[TMP27]], align 1
+; VF4IC4-NEXT:    [[REVERSE6:%.*]] = shufflevector <4 x i8> [[WIDE_LOAD5]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; VF4IC4-NEXT:    [[TMP25:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
 ; VF4IC4-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i8, ptr [[TMP25]], i32 0
 ; VF4IC4-NEXT:    [[TMP46:%.*]] = getelementptr inbounds i8, ptr [[TMP26]], i32 -3
+; VF4IC4-NEXT:    [[TMP41:%.*]] = getelementptr inbounds i8, ptr [[TMP25]], i32 -4
+; VF4IC4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[TMP41]], i32 -3
+; VF4IC4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP25]], i32 -8
+; VF4IC4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[TMP14]], i32 -3
+; VF4IC4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[TMP25]], i32 -12
+; VF4IC4-NEXT:    [[TMP42:%.*]] = getelementptr inbounds i8, ptr [[TMP16]], i32 -3
 ; VF4IC4-NEXT:    [[WIDE_LOAD14:%.*]] = load <4 x i8>, ptr [[TMP46]], align 1
 ; VF4IC4-NEXT:    [[REVERSE15:%.*]] = shufflevector <4 x i8> [[WIDE_LOAD14]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; VF4IC4-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x i8>, ptr [[TMP13]], align 1
+; VF4IC4-NEXT:    [[REVERSE10:%.*]] = shufflevector <4 x i8> [[WIDE_LOAD9]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; VF4IC4-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x i8>, ptr [[TMP15]], align 1
+; VF4IC4-NEXT:    [[REVERSE12:%.*]] = shufflevector <4 x i8> [[WIDE_LOAD11]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; VF4IC4-NEXT:    [[WIDE_LOAD15:%.*]] = load <4 x i8>, ptr [[TMP42]], align 1
+; VF4IC4-NEXT:    [[REVERSE16:%.*]] = shufflevector <4 x i8> [[WIDE_LOAD15]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; VF4IC4-NEXT:    [[TMP21:%.*]] = icmp ne <4 x i8> [[REVERSE14]], [[REVERSE15]]
-; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; VF4IC4-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP21]])
-; VF4IC4-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1020
+; VF4IC4-NEXT:    [[TMP19:%.*]] = icmp ne <4 x i8> [[REVERSE2]], [[REVERSE10]]
+; VF4IC4-NEXT:    [[TMP20:%.*]] = icmp ne <4 x i8> [[REVERSE4]], [[REVERSE12]]
+; VF4IC4-NEXT:    [[TMP43:%.*]] = icmp ne <4 x i8> [[REVERSE6]], [[REVERSE16]]
+; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
+; VF4IC4-NEXT:    [[TMP22:%.*]] = or <4 x i1> [[TMP21]], [[TMP19]]
+; VF4IC4-NEXT:    [[TMP23:%.*]] = or <4 x i1> [[TMP22]], [[TMP20]]
+; VF4IC4-NEXT:    [[TMP24:%.*]] = or <4 x i1> [[TMP23]], [[TMP43]]
+; VF4IC4-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP24]])
+; VF4IC4-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1008
 ; VF4IC4-NEXT:    [[TMP9:%.*]] = or i1 [[TMP7]], [[TMP8]]
 ; VF4IC4-NEXT:    br i1 [[TMP9]], label [[MIDDLE_SPLIT:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; VF4IC4:       middle.split:
 ; VF4IC4-NEXT:    br i1 [[TMP7]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; VF4IC4:       middle.block:
-; VF4IC4-NEXT:    br i1 false, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; VF4IC4-NEXT:    br label [[SCALAR_PH]]
 ; VF4IC4:       vector.early.exit:
-; VF4IC4-NEXT:    [[TMP10:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP21]], i1 true)
+; VF4IC4-NEXT:    [[TMP28:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP43]], i1 true)
+; VF4IC4-NEXT:    [[TMP29:%.*]] = add i64 12, [[TMP28]]
+; VF4IC4-NEXT:    [[TMP30:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP20]], i1 true)
+; VF4IC4-NEXT:    [[TMP31:%.*]] = add i64 8, [[TMP30]]
+; VF4IC4-NEXT:    [[TMP32:%.*]] = icmp ne i64 [[TMP30]], 4
+; VF4IC4-NEXT:    [[TMP33:%.*]] = select i1 [[TMP32]], i64 [[TMP31]], i64 [[TMP29]]
+; VF4IC4-NEXT:    [[TMP34:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP19]], i1 true)
+; VF4IC4-NEXT:    [[TMP35:%.*]] = add i64 4, [[TMP34]]
+; VF4IC4-NEXT:    [[TMP36:%.*]] = icmp ne i64 [[TMP34]], 4
+; VF4IC4-NEXT:    [[TMP37:%.*]] = select i1 [[TMP36]], i64 [[TMP35]], i64 [[TMP33]]
+; VF4IC4-NEXT:    [[TMP38:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP21]], i1 true)
+; VF4IC4-NEXT:    [[TMP39:%.*]] = add i64 0, [[TMP38]]
+; VF4IC4-NEXT:    [[TMP40:%.*]] = icmp ne i64 [[TMP38]], 4
+; VF4IC4-NEXT:    [[TMP10:%.*]] = select i1 [[TMP40]], i64 [[TMP39]], i64 [[TMP37]]
 ; VF4IC4-NEXT:    [[TMP11:%.*]] = add i64 [[INDEX]], [[TMP10]]
 ; VF4IC4-NEXT:    [[TMP12:%.*]] = sub i64 1023, [[TMP11]]
-; VF4IC4-NEXT:    br label [[LOOP_END]]
+; VF4IC4-NEXT:    br label [[LOOP_END:%.*]]
 ; VF4IC4:       scalar.ph:
-; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[MIDDLE_BLOCK]] ], [ 1023, [[ENTRY:%.*]] ]
+; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 15, [[MIDDLE_BLOCK]] ], [ 1023, [[ENTRY:%.*]] ]
 ; VF4IC4-NEXT:    br label [[LOOP:%.*]]
 ; VF4IC4:       loop:
 ; VF4IC4-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -526,7 +719,7 @@ define i64 @same_exit_block_pre_inc_use1_reverse() {
 ; VF4IC4-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], 0
 ; VF4IC4-NEXT:    br i1 [[EXITCOND]], label [[LOOP_END]], label [[LOOP]], !llvm.loop [[LOOP15:![0-9]+]]
 ; VF4IC4:       loop.end:
-; VF4IC4-NEXT:    [[RETVAL:%.*]] = phi i64 [ [[IV]], [[LOOP]] ], [ 1024, [[LOOP_INC]] ], [ 1024, [[MIDDLE_BLOCK]] ], [ [[TMP12]], [[VECTOR_EARLY_EXIT]] ]
+; VF4IC4-NEXT:    [[RETVAL:%.*]] = phi i64 [ [[IV]], [[LOOP]] ], [ 1024, [[LOOP_INC]] ], [ [[TMP12]], [[VECTOR_EARLY_EXIT]] ]
 ; VF4IC4-NEXT:    ret i64 [[RETVAL]]
 ;
 entry:
@@ -569,27 +762,68 @@ define i8 @same_exit_block_use_loaded_value() {
 ; VF4IC4:       vector.body:
 ; VF4IC4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; VF4IC4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[INDEX]]
-; VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 0
-; VF4IC4-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
-; VF4IC4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[INDEX]]
-; VF4IC4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[TMP18]], i32 0
-; VF4IC4-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i8>, ptr [[TMP19]], align 1
-; VF4IC4-NEXT:    [[TMP13:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD3]], [[WIDE_LOAD7]]
-; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; VF4IC4-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP13]])
+; VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 4
+; VF4IC4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 8
+; VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 12
+; VF4IC4-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP0]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD4:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
+; VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[INDEX]]
+; VF4IC4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 4
+; VF4IC4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 8
+; VF4IC4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 12
+; VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD8:%.*]] = load <4 x i8>, ptr [[TMP9]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x i8>, ptr [[TMP10]], align 1
+; VF4IC4-NEXT:    [[WIDE_LOAD7:%.*]] = load <4 x i8>, ptr [[TMP15]], align 1
+; VF4IC4-NEXT:    [[TMP12:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD2]], [[WIDE_LOAD6]]
+; VF4IC4-NEXT:    [[TMP29:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD4]], [[WIDE_LOAD8]]
+; VF4IC4-NEXT:    [[TMP11:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD5]], [[WIDE_LOAD9]]
+; VF4IC4-NEXT:    [[TMP17:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD3]], [[WIDE_LOAD7]]
+; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
+; VF4IC4-NEXT:    [[TMP13:%.*]] = or <4 x i1> [[TMP12]], [[TMP29]]
+; VF4IC4-NEXT:    [[TMP14:%.*]] = or <4 x i1> [[TMP13]], [[TMP11]]
+; VF4IC4-NEXT:    [[TMP16:%.*]] = or <4 x i1> [[TMP14]], [[TMP17]]
+; VF4IC4-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP16]])
 ; VF4IC4-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; VF4IC4-NEXT:    [[TMP7:%.*]] = or i1 [[TMP5]], [[TMP6]]
 ; VF4IC4-NEXT:    br i1 [[TMP7]], label [[MIDDLE_SPLIT:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; VF4IC4:       middle.split:
 ; VF4IC4-NEXT:    br i1 [[TMP5]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; VF4IC4:       middle.block:
-; VF4IC4-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; VF4IC4-NEXT:    br label [[LOOP_END:%.*]]
 ; VF4IC4:       vector.early.exit:
-; VF4IC4-NEXT:    [[FIRST_ACTIVE_LANE1:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP13]], i1 true)
-; VF4IC4-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = extractelement <4 x i8> [[WIDE_LOAD3]], i64 [[FIRST_ACTIVE_LANE1]]
+; VF4IC4-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP17]], i1 true)
+; VF4IC4-NEXT:    [[TMP20:%.*]] = add i64 12, [[FIRST_ACTIVE_LANE]]
+; VF4IC4-NEXT:    [[FIRST_ACTIVE_LANE8:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP11]], i1 true)
+; VF4IC4-NEXT:    [[TMP21:%.*]] = add i64 8, [[FIRST_ACTIVE_LANE8]]
+; VF4IC4-NEXT:    [[TMP22:%.*]] = icmp ne i64 [[FIRST_ACTIVE_LANE8]], 4
+; VF4IC4-NEXT:    [[TMP23:%.*]] = select i1 [[TMP22]], i64 [[TMP21]], i64 [[TMP20]]
+; VF4IC4-NEXT:    [[FIRST_ACTIVE_LANE9:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP29]], i1 true)
+; VF4IC4-NEXT:    [[TMP24:%.*]] = add i64 4, [[FIRST_ACTIVE_LANE9]]
+; VF4IC4-NEXT:    [[TMP25:%.*]] = icmp ne i64 [[FIRST_ACTIVE_LANE9]], 4
+; VF4IC4-NEXT:    [[TMP26:%.*]] = select i1 [[TMP25]], i64 [[TMP24]], i64 [[TMP23]]
+; VF4IC4-NEXT:    [[FIRST_ACTIVE_LANE1:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP12]], i1 true)
+; VF4IC4-NEXT:    [[TMP27:%.*]] = add i64 0, [[FIRST_ACTIVE_LANE1]]
+; VF4IC4-NEXT:    [[TMP28:%.*]] = icmp ne i64 [[FIRST_ACTIVE_LANE1]], 4
+; VF4IC4-NEXT:    [[TMP8:%.*]] = select i1 [[TMP28]], i64 [[TMP27]], i64 [[TMP26]]
+; VF4IC4-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = extractelement <4 x i8> [[WIDE_LOAD2]], i64 [[TMP8]]
+; VF4IC4-NEXT:    [[TMP31:%.*]] = sub i64 [[TMP8]], 4
+; VF4IC4-NEXT:    [[TMP32:%.*]] = extractelement <4 x i8> [[WIDE_LOAD4]], i64 [[TMP31]]
+; VF4IC4-NEXT:    [[TMP33:%.*]] = icmp uge i64 [[TMP8]], 4
+; VF4IC4-NEXT:    [[TMP34:%.*]] = select i1 [[TMP33]], i8 [[TMP32]], i8 [[EARLY_EXIT_VALUE]]
+; VF4IC4-NEXT:    [[TMP35:%.*]] = sub i64 [[TMP8]], 8
+; VF4IC4-NEXT:    [[TMP36:%.*]] = extractelement <4 x i8> [[WIDE_LOAD5]], i64 [[TMP35]]
+; VF4IC4-NEXT:    [[TMP37:%.*]] = icmp uge i64 [[TMP8]], 8
+; VF4IC4-NEXT:    [[TMP38:%.*]] = select i1 [[TMP37]], i8 [[TMP36]], i8 [[TMP34]]
+; VF4IC4-NEXT:    [[TMP39:%.*]] = sub i64 [[TMP8]], 12
+; VF4IC4-NEXT:    [[TMP40:%.*]] = extractelement <4 x i8> [[WIDE_LOAD3]], i64 [[TMP39]]
+; VF4IC4-NEXT:    [[TMP41:%.*]] = icmp uge i64 [[TMP8]], 12
+; VF4IC4-NEXT:    [[TMP42:%.*]] = select i1 [[TMP41]], i8 [[TMP40]], i8 [[TMP38]]
 ; VF4IC4-NEXT:    br label [[LOOP_END]]
 ; VF4IC4:       scalar.ph:
-; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1024, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; VF4IC4-NEXT:    br label [[LOOP:%.*]]
 ; VF4IC4:       loop:
 ; VF4IC4-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -604,7 +838,7 @@ define i8 @same_exit_block_use_loaded_value() {
 ; VF4IC4-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], 1024
 ; VF4IC4-NEXT:    br i1 [[EXITCOND]], label [[LOOP_END]], label [[LOOP]], !llvm.loop [[LOOP17:![0-9]+]]
 ; VF4IC4:       loop.end:
-; VF4IC4-NEXT:    [[RETVAL:%.*]] = phi i8 [ [[LD1]], [[LOOP]] ], [ -1, [[LOOP_INC]] ], [ -1, [[MIDDLE_BLOCK]] ], [ [[EARLY_EXIT_VALUE]], [[VECTOR_EARLY_EXIT]] ]
+; VF4IC4-NEXT:    [[RETVAL:%.*]] = phi i8 [ [[LD1]], [[LOOP]] ], [ -1, [[LOOP_INC]] ], [ -1, [[MIDDLE_BLOCK]] ], [ [[TMP42]], [[VECTOR_EARLY_EXIT]] ]
 ; VF4IC4-NEXT:    ret i8 [[RETVAL]]
 ;
 entry:
@@ -650,29 +884,84 @@ define i8 @same_exit_block_reverse_use_loaded_value() {
 ; VF4IC4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
 ; VF4IC4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 0
 ; VF4IC4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 -3
+; VF4IC4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 -4
+; VF4IC4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 -3
+; VF4IC4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 -8
+; VF4IC4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP5]], i32 -3
+; VF4IC4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 -12
+; VF4IC4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[TMP7]], i32 -3
 ; VF4IC4-NEXT:    [[WIDE_LOAD5:%.*]] = load <4 x i8>, ptr [[TMP8]], align 1
 ; VF4IC4-NEXT:    [[REVERSE6:%.*]] = shufflevector <4 x i8> [[WIDE_LOAD5]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; VF4IC4-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; VF4IC4-NEXT:    [[REVERSE2:%.*]] = shufflevector <4 x i8> [[WIDE_LOAD1]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; VF4IC4-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP6]], align 1
+; VF4IC4-NEXT:    [[REVERSE4:%.*]] = shufflevector <4 x i8> [[WIDE_LOAD3]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; VF4IC4-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i8>, ptr [[TMP11]], align 1
+; VF4IC4-NEXT:    [[REVERSE7:%.*]] = shufflevector <4 x i8> [[WIDE_LOAD6]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; VF4IC4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
 ; VF4IC4-NEXT:    [[TMP38:%.*]] = getelementptr inbounds i8, ptr [[TMP9]], i32 0
 ; VF4IC4-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP38]], i32 -3
+; VF4IC4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[TMP9]], i32 -4
+; VF4IC4-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[TMP12]], i32 -3
+; VF4IC4-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP9]], i32 -8
+; VF4IC4-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[TMP14]], i32 -3
+; VF4IC4-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[TMP9]], i32 -12
+; VF4IC4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[TMP16]], i32 -3
 ; VF4IC4-NEXT:    [[WIDE_LOAD13:%.*]] = load <4 x i8>, ptr [[TMP17]], align 1
 ; VF4IC4-NEXT:    [[REVERSE14:%.*]] = shufflevector <4 x i8> [[WIDE_LOAD13]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; VF4IC4-NEXT:    [[WIDE_LOAD9:%.*]] = load <4 x i8>, ptr [[TMP13]], align 1
+; VF4IC4-NEXT:    [[REVERSE10:%.*]] = shufflevector <4 x i8> [[WIDE_LOAD9]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; VF4IC4-NEXT:    [[WIDE_LOAD11:%.*]] = load <4 x i8>, ptr [[TMP15]], align 1
+; VF4IC4-NEXT:    [[REVERSE12:%.*]] = shufflevector <4 x i8> [[WIDE_LOAD11]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
+; VF4IC4-NEXT:    [[WIDE_LOAD14:%.*]] = load <4 x i8>, ptr [[TMP18]], align 1
+; VF4IC4-NEXT:    [[REVERSE15:%.*]] = shufflevector <4 x i8> [[WIDE_LOAD14]], <4 x i8> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0>
 ; VF4IC4-NEXT:    [[TMP21:%.*]] = icmp ne <4 x i8> [[REVERSE6]], [[REVERSE14]]
-; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; VF4IC4-NEXT:    [[TMP25:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP21]])
-; VF4IC4-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1020
+; VF4IC4-NEXT:    [[TMP19:%.*]] = icmp ne <4 x i8> [[REVERSE2]], [[REVERSE10]]
+; VF4IC4-NEXT:    [[TMP20:%.*]] = icmp ne <4 x i8> [[REVERSE4]], [[REVERSE12]]
+; VF4IC4-NEXT:    [[TMP37:%.*]] = icmp ne <4 x i8> [[REVERSE7]], [[REVERSE15]]
+; VF4IC4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
+; VF4IC4-NEXT:    [[TMP22:%.*]] = or <4 x i1> [[TMP21]], [[TMP19]]
+; VF4IC4-NEXT:    [[TMP23:%.*]] = or <4 x i1> [[TMP22]], [[TMP20]]
+; VF4IC4-NEXT:    [[TMP24:%.*]] = or <4 x i1> [[TMP23]], [[TMP37]]
+; VF4IC4-NEXT:    [[TMP25:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP24]])
+; VF4IC4-NEXT:    [[TMP26:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1008
 ; VF4IC4-NEXT:    [[TMP27:%.*]] = or i1 [[TMP25]], [[TMP26]]
 ; VF4IC4-NEXT:    br i1 [[TMP27]], label [[MIDDLE_SPLIT:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; VF4IC4:       middle.split:
 ; VF4IC4-NEXT:    br i1 [[TMP25]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; VF4IC4:       middle.block:
-; VF4IC4-NEXT:    br i1 false, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; VF4IC4-NEXT:    br label [[SCALAR_PH]]
 ; VF4IC4:       vector.early.exit:
+; VF4IC4-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP37]], i1 true)
+; VF4IC4-NEXT:    [[TMP28:%.*]] = add i64 12, [[FIRST_ACTIVE_LANE]]
+; VF4IC4-NEXT:    [[FIRST_ACTIVE_LANE15:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP20]], i1 true)
+; VF4IC4-NEXT:    [[TMP29:%.*]] = add i64 8, [[FIRST_ACTIVE_LANE15]]
+; VF4IC4-NEXT:    [[TMP30:%.*]] = icmp ne i64 [[FIRST_ACTIVE_LANE15]], 4
+; VF4IC4-NEXT:    [[TMP31:%.*]] = select i1 [[TMP30]], i64 [[TMP29]], i64 [[TMP28]]
+; VF4IC4-NEXT:    [[FIRST_ACTIVE_LANE16:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP19]], i1 true)
+; VF4IC4-NEXT:    [[TMP32:%.*]] = add i64 4, [[FIRST_ACTIVE_LANE16]]
+; VF4IC4-NEXT:    [[TMP33:%.*]] = icmp ne i64 [[FIRST_ACTIVE_LANE16]], 4
+; VF4IC4-NEXT:    [[TMP34:%.*]] = select i1 [[TMP33]], i64 [[TMP32]], i64 [[TMP31]]
 ; VF4IC4-NEXT:    [[FIRST_ACTIVE_LANE1:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP21]], i1 true)
-; VF4IC4-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = extractelement <4 x i8> [[REVERSE6]], i64 [[FIRST_ACTIVE_LANE1]]
-; VF4IC4-NEXT:    br label [[LOOP_END]]
+; VF4IC4-NEXT:    [[TMP35:%.*]] = add i64 0, [[FIRST_ACTIVE_LANE1]]
+; VF4IC4-NEXT:    [[TMP36:%.*]] = icmp ne i64 [[FIRST_ACTIVE_LANE1]], 4
+; VF4IC4-NEXT:    [[TMP10:%.*]] = select i1 [[TMP36]], i64 [[TMP35]], i64 [[TMP34]]
+; VF4IC4-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = extractelement <4 x i8> [[REVERSE6]], i64 [[TMP10]]
+; VF4IC4-NEXT:    [[TMP39:%.*]] = sub i64 [[TMP10]], 4
+; VF4IC4-NEXT:    [[TMP40:%.*]] = extractelement <4 x i8> [[REVERSE2]], i64 [[TMP39]]
+; VF4IC4-NEXT:    [[TMP41:%.*]] = icmp uge i64 [[TMP10]], 4
+; VF4IC4-NEXT:    [[TMP42:%.*]] = select i1 [[TMP41]], i8 [[TMP40]], i8 [[EARLY_EXIT_VALUE]]
+; VF4IC4-NEXT:    [[TMP43:%.*]] = sub i64 [[TMP10]], 8
+; VF4IC4-NEXT:    [[TMP44:%.*]] = extractelement <4 x i8> [[REVERSE4]], i64 [[TMP43]]
+; VF4IC4-NEXT:    [[TMP45:%.*]] = icmp uge i64 [[TMP10]], 8
+; VF4IC4-NEXT:    [[TMP46:%.*]] = select i1 [[TMP45]], i8 [[TMP44]], i8 [[TMP42]]
+; VF4IC4-NEXT:    [[TMP47:%.*]] = sub i64 [[TMP10]], 12
+; VF4IC4-NEXT:    [[TMP48:%.*]] = extractelement <4 x i8> [[REVERSE7]], i64 [[TMP47]]
+; VF4IC4-NEXT:    [[TMP49:%.*]] = icmp uge i64 [[TMP10]], 12
+; VF4IC4-NEXT:    [[TMP50:%.*]] = select i1 [[TMP49]], i8 [[TMP48]], i8 [[TMP46]]
+; VF4IC4-NEXT:    br label [[LOOP_END:%.*]]
 ; VF4IC4:       scalar.ph:
-; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[MIDDLE_BLOCK]] ], [ 1023, [[ENTRY:%.*]] ]
+; VF4IC4-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 15, [[MIDDLE_BLOCK]] ], [ 1023, [[ENTRY:%.*]] ]
 ; VF4IC4-NEXT:    br label [[LOOP:%.*]]
 ; VF4IC4:       loop:
 ; VF4IC4-NEXT:    [[IV:%.*]] = phi i64 [ [[IV_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -687,7 +976,7 @@ define i8 @same_exit_block_reverse_use_loaded_value() {
 ; VF4IC4-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[IV_NEXT]], 0
 ; VF4IC4-NEXT:    br i1 [[EXITCOND]], label [[LOOP_END]], label [[LOOP]], !llvm.loop [[LOOP19:![0-9]+]]
 ; VF4IC4:       loop.end:
-; VF4IC4-NEXT:    [[RETVAL:%.*]] = phi i8 [ [[LD1]], [[LOOP]] ], [ -1, [[LOOP_INC]] ], [ -1, [[MIDDLE_BLOCK]] ], [ [[EARLY_EXIT_VALUE]], [[VECTOR_EARLY_EXIT]] ]
+; VF4IC4-NEXT:    [[RETVAL:%.*]] = phi i8 [ [[LD1]], [[LOOP]] ], [ -1, [[LOOP_INC]] ], [ [[TMP50]], [[VECTOR_EARLY_EXIT]] ]
 ; VF4IC4-NEXT:    ret i8 [[RETVAL]]
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/single-value-blend-phis.ll b/llvm/test/Transforms/LoopVectorize/single-value-blend-phis.ll
index d5a206f..842ff91 100644
--- a/llvm/test/Transforms/LoopVectorize/single-value-blend-phis.ll
+++ b/llvm/test/Transforms/LoopVectorize/single-value-blend-phis.ll
@@ -19,21 +19,19 @@ define void @single_incoming_phi_no_blend_mask(i64 %a, i64 %b) {
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDEX]] to i16
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [32 x i16], ptr @src, i16 0, i16 [[TMP1]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i16>, ptr [[TMP4]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i16>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp sgt <2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP5]], <2 x i16> splat (i16 1), <2 x i16> [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds [32 x i16], ptr @dst, i16 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i16, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <2 x i16> [[PREDPHI]], ptr [[TMP7]], align 2
+; CHECK-NEXT:    store <2 x i16> [[PREDPHI]], ptr [[TMP6]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 32, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; CHECK:       loop.header:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
@@ -103,24 +101,22 @@ define void @single_incoming_phi_with_blend_mask(i64 %a, i64 %b) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = trunc i64 [[INDEX]] to i16
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp ugt <2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr [32 x i16], ptr @src, i16 0, i16 [[TMP1]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i16, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i16>, ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i16>, ptr [[TMP4]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp sle <2 x i64> [[VEC_IND]], [[BROADCAST_SPLAT]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = select <2 x i1> [[TMP3]], <2 x i1> [[TMP6]], <2 x i1> zeroinitializer
 ; CHECK-NEXT:    [[TMP8:%.*]] = xor <2 x i1> [[TMP3]], splat (i1 true)
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP7]], <2 x i16> [[WIDE_LOAD]], <2 x i16> splat (i16 1)
 ; CHECK-NEXT:    [[PREDPHI1:%.*]] = select <2 x i1> [[TMP8]], <2 x i16> zeroinitializer, <2 x i16> [[PREDPHI]]
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds [32 x i16], ptr @dst, i16 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i16, ptr [[TMP9]], i32 0
-; CHECK-NEXT:    store <2 x i16> [[PREDPHI1]], ptr [[TMP10]], align 2
+; CHECK-NEXT:    store <2 x i16> [[PREDPHI1]], ptr [[TMP9]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32
 ; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 32, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; CHECK:       loop.header:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
@@ -200,8 +196,7 @@ define void @multiple_incoming_phi_with_blend_mask(i64 %a, ptr noalias %dst) {
 ; CHECK-NEXT:    [[TMP8:%.*]] = insertelement <2 x i16> poison, i16 [[TMP6]], i32 0
 ; CHECK-NEXT:    [[TMP9:%.*]] = insertelement <2 x i16> [[TMP8]], i16 [[TMP7]], i32 1
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i16, ptr [[DST:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i16, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <2 x i16> [[TMP9]], ptr [[TMP11]], align 2
+; CHECK-NEXT:    store <2 x i16> [[TMP9]], ptr [[TMP10]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[VEC_IND_NEXT2]] = add <2 x i16> [[VEC_IND1]], splat (i16 2)
@@ -209,9 +204,9 @@ define void @multiple_incoming_phi_with_blend_mask(i64 %a, ptr noalias %dst) {
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32
 ; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 32, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; CHECK:       loop.header:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
@@ -301,16 +296,15 @@ define void @single_incoming_needs_predication(i64 %a, i64 %b) {
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select <2 x i1> [[TMP16]], <2 x i16> [[TMP14]], <2 x i16> splat (i16 1)
 ; CHECK-NEXT:    [[PREDPHI3:%.*]] = select <2 x i1> [[TMP17]], <2 x i16> zeroinitializer, <2 x i16> [[PREDPHI]]
 ; CHECK-NEXT:    [[TMP18:%.*]] = getelementptr inbounds [32 x i16], ptr @dst, i16 0, i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i16, ptr [[TMP18]], i32 0
-; CHECK-NEXT:    store <2 x i16> [[PREDPHI3]], ptr [[TMP19]], align 2
+; CHECK-NEXT:    store <2 x i16> [[PREDPHI3]], ptr [[TMP18]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; CHECK-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], 64
 ; CHECK-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 64, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; CHECK:       loop.header:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
@@ -377,16 +371,15 @@ define void @duplicated_incoming_blocks_blend(i32 %x, ptr %ptr) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <2 x i32> [ <i32 0, i32 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i32, ptr [[PTR:%.*]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP2]], align 4
+; CHECK-NEXT:    store <2 x i32> [[VEC_IND]], ptr [[TMP1]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; CHECK:       loop.header:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[ADD_I:%.*]], [[LOOP_LATCH:%.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/single_early_exit.ll b/llvm/test/Transforms/LoopVectorize/single_early_exit.ll
index 2069570..2c0a6f1 100644
--- a/llvm/test/Transforms/LoopVectorize/single_early_exit.ll
+++ b/llvm/test/Transforms/LoopVectorize/single_early_exit.ll
@@ -18,11 +18,9 @@ define i64 @same_exit_block_phi_of_consts() {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]])
@@ -32,11 +30,11 @@ define i64 @same_exit_block_phi_of_consts() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -95,11 +93,9 @@ define i64 @diff_exit_block_phi_of_consts() {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]])
@@ -109,11 +105,11 @@ define i64 @diff_exit_block_phi_of_consts() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    br label [[LOOP_EARLY_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -193,11 +189,9 @@ define i32 @diff_exit_block_needs_scev_check(i32 %end) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[P1]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i32, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP11]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP10]], align 4
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i32, ptr [[P2]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i32, ptr [[TMP12]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP13]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP12]], align 4
 ; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP15:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP14]])
@@ -294,11 +288,11 @@ define i32 @diff_blocks_invariant_early_exit_cond(ptr %s) {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP1]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    br label [[EARLY_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 266, [[MIDDLE_BLOCK]] ], [ -10, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ -10, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[IND:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IND_NEXT:%.*]], [[FOR_INC:%.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/single_early_exit_live_outs.ll b/llvm/test/Transforms/LoopVectorize/single_early_exit_live_outs.ll
index a21666a..940e398 100644
--- a/llvm/test/Transforms/LoopVectorize/single_early_exit_live_outs.ll
+++ b/llvm/test/Transforms/LoopVectorize/single_early_exit_live_outs.ll
@@ -17,11 +17,9 @@ define i64 @same_exit_block_pre_inc_use1() {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]])
@@ -31,14 +29,14 @@ define i64 @same_exit_block_pre_inc_use1() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP6]], i1 true)
 ; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i64 3, [[TMP10]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -97,11 +95,9 @@ define i32 @same_exit_block_pre_inc_use1_iv64_endi32_step2() {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]])
@@ -111,7 +107,7 @@ define i32 @same_exit_block_pre_inc_use1_iv64_endi32_step2() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP6]], i1 true)
 ; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
@@ -120,8 +116,8 @@ define i32 @same_exit_block_pre_inc_use1_iv64_endi32_step2() {
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i32 9, [[TMP11]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL4:%.*]] = phi i32 [ 137, [[MIDDLE_BLOCK]] ], [ 9, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL4:%.*]] = phi i32 [ 9, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -177,14 +173,12 @@ define i32 @same_exit_block_pre_inc_use1_iv128_endi32_step2() {
 ; CHECK-NEXT:    call void @init_mem(ptr [[P1]], i64 1024)
 ; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[P1]], i64 64
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i128 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i128 [[INDEX1]] to i64
 ; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[NEXT_GEP]], align 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], splat (i8 3)
 ; CHECK-NEXT:    [[INDEX_NEXT2]] = add nuw i128 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP4]])
@@ -194,7 +188,7 @@ define i32 @same_exit_block_pre_inc_use1_iv128_endi32_step2() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP4]], i1 true)
 ; CHECK-NEXT:    [[TMP8:%.*]] = zext i64 [[FIRST_ACTIVE_LANE]] to i128
@@ -204,9 +198,9 @@ define i32 @same_exit_block_pre_inc_use1_iv128_endi32_step2() {
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i32 9, [[TMP10]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i128 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL3:%.*]] = phi i32 [ 137, [[MIDDLE_BLOCK]] ], [ 9, [[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL4:%.*]] = phi ptr [ [[TMP0]], [[MIDDLE_BLOCK]] ], [ [[P1]], [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i128 [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL3:%.*]] = phi i32 [ 9, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL4:%.*]] = phi ptr [ [[P1]], [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i128 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -265,11 +259,9 @@ define float @same_exit_block_pre_inc_use1_iv64_endf32() {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]])
@@ -279,7 +271,7 @@ define float @same_exit_block_pre_inc_use1_iv64_endf32() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP6]], i1 true)
 ; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
@@ -288,8 +280,8 @@ define float @same_exit_block_pre_inc_use1_iv64_endf32() {
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = fadd fast float 9.000000e+00, [[TMP11]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL4:%.*]] = phi float [ 7.300000e+01, [[MIDDLE_BLOCK]] ], [ 9.000000e+00, [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL4:%.*]] = phi float [ 9.000000e+00, [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -347,17 +339,14 @@ define ptr @same_exit_block_pre_inc_use1_iv64_endptr() {
 ; CHECK-NEXT:    call void @init_mem(ptr [[P2]], i64 1024)
 ; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[P2]], i64 320
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT7:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP11]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP10]], align 1
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[TMP12]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i8>, ptr [[TMP13]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD6:%.*]] = load <4 x i8>, ptr [[TMP12]], align 1
 ; CHECK-NEXT:    [[TMP15:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD6]]
 ; CHECK-NEXT:    [[INDEX_NEXT7]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP16:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP15]])
@@ -367,7 +356,7 @@ define ptr @same_exit_block_pre_inc_use1_iv64_endptr() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP16]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP15]], i1 true)
 ; CHECK-NEXT:    [[TMP19:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
@@ -375,8 +364,8 @@ define ptr @same_exit_block_pre_inc_use1_iv64_endptr() {
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = getelementptr i8, ptr [[P2]], i64 [[TMP20]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL8:%.*]] = phi ptr [ [[TMP0]], [[MIDDLE_BLOCK]] ], [ [[P2]], [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL8:%.*]] = phi ptr [ [[P2]], [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -433,13 +422,11 @@ define ptr @same_exit_block_pre_inc_use1_ivptr() {
 ; CHECK-NEXT:    [[PTREND:%.*]] = getelementptr i8, ptr [[P1]], i64 1024
 ; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[P1]], i64 1024
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[P1]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP9]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[NEXT_GEP]], align 1
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], splat (i8 72)
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP12:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP11]])
@@ -449,14 +436,14 @@ define ptr @same_exit_block_pre_inc_use1_ivptr() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP12]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP11]], i1 true)
 ; CHECK-NEXT:    [[TMP8:%.*]] = add i64 [[INDEX]], [[FIRST_ACTIVE_LANE]]
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = getelementptr i8, ptr [[P1]], i64 [[TMP8]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi ptr [ [[TMP0]], [[MIDDLE_BLOCK]] ], [ [[P1]], [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi ptr [ [[P1]], [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[PTR:%.*]] = phi ptr [ [[PTR_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -509,11 +496,9 @@ define i64 @same_exit_block_pre_inc1_use_inv_cond(i1 %cond) {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = select i1 [[COND]], <4 x i1> [[TMP5]], <4 x i1> zeroinitializer
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4
@@ -525,14 +510,14 @@ define i64 @same_exit_block_pre_inc1_use_inv_cond(i1 %cond) {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP7]], i1 true)
 ; CHECK-NEXT:    [[TMP11:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i64 3, [[TMP11]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -593,11 +578,9 @@ define i64 @same_exit_block_pre_inc_use1_gep_two_indices() {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds [1024 x i8], ptr [[P1]], i64 0, i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds [1024 x i8], ptr [[P2]], i64 0, i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]])
@@ -607,14 +590,14 @@ define i64 @same_exit_block_pre_inc_use1_gep_two_indices() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP6]], i1 true)
 ; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i64 3, [[TMP10]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -673,11 +656,9 @@ define i64 @same_exit_block_pre_inc_use1_alloca_diff_type() {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]])
@@ -687,14 +668,14 @@ define i64 @same_exit_block_pre_inc_use1_alloca_diff_type() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP6]], i1 true)
 ; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i64 3, [[TMP10]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -753,11 +734,9 @@ define i64 @same_exit_block_pre_inc_use2() {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]])
@@ -767,11 +746,11 @@ define i64 @same_exit_block_pre_inc_use2() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -830,11 +809,9 @@ define i64 @same_exit_block_pre_inc_use3() {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]])
@@ -844,14 +821,14 @@ define i64 @same_exit_block_pre_inc_use3() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP6]], i1 true)
 ; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i64 3, [[TMP10]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -912,8 +889,7 @@ define i64 @same_exit_block_pre_inc_use4() {
 ; CHECK-NEXT:    [[VEC_IND:%.*]] = phi <4 x i64> [ <i64 3, i64 4, i64 5, i64 6>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp uge <4 x i64> [[VEC_IND]], [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[INDEX_NEXT2]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP4]])
@@ -924,14 +900,14 @@ define i64 @same_exit_block_pre_inc_use4() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP4]], i1 true)
 ; CHECK-NEXT:    [[TMP8:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i64 3, [[TMP8]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -986,11 +962,9 @@ define i64 @same_exit_block_post_inc_use() {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP7]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP14:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP13]])
@@ -1000,14 +974,14 @@ define i64 @same_exit_block_post_inc_use() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP14]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP13]], i1 true)
 ; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i64 3, [[TMP10]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -1060,13 +1034,11 @@ define ptr @same_exit_block_post_inc_use1_ivptr() {
 ; CHECK-NEXT:    [[PTREND:%.*]] = getelementptr i8, ptr [[P1]], i64 1024
 ; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
-; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[P1]], i64 1024
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[P1]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[NEXT_GEP]], align 1
 ; CHECK-NEXT:    [[TMP15:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], splat (i8 72)
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP16:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP15]])
@@ -1076,7 +1048,7 @@ define ptr @same_exit_block_post_inc_use1_ivptr() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP16]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP15]], i1 true)
 ; CHECK-NEXT:    [[TMP8:%.*]] = add i64 [[INDEX]], [[FIRST_ACTIVE_LANE]]
@@ -1084,7 +1056,7 @@ define ptr @same_exit_block_post_inc_use1_ivptr() {
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = getelementptr i8, ptr [[P1]], i64 [[TMP9]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi ptr [ [[TMP0]], [[MIDDLE_BLOCK]] ], [ [[P1]], [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi ptr [ [[P1]], [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[PTR:%.*]] = phi ptr [ [[PTR_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -1136,11 +1108,9 @@ define i64 @same_exit_block_post_inc_use2() {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP7]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[TMP17:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP18:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP17]])
@@ -1150,7 +1120,7 @@ define i64 @same_exit_block_post_inc_use2() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP18]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP17]], i1 true)
 ; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
@@ -1158,7 +1128,7 @@ define i64 @same_exit_block_post_inc_use2() {
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i64 3, [[TMP11]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -1217,11 +1187,9 @@ define i64 @diff_exit_block_pre_inc_use1() {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]])
@@ -1231,14 +1199,14 @@ define i64 @diff_exit_block_pre_inc_use1() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP6]], i1 true)
 ; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i64 3, [[TMP10]]
 ; CHECK-NEXT:    br label [[LOOP_EARLY_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -1304,11 +1272,9 @@ define i64 @diff_exit_block_pre_inc_use2() {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]])
@@ -1318,11 +1284,11 @@ define i64 @diff_exit_block_pre_inc_use2() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    br label [[LOOP_EARLY_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -1388,11 +1354,9 @@ define i64 @diff_exit_block_pre_inc_use3() {
 ; CHECK-NEXT:    [[INDEX2:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX2]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD3]]
 ; CHECK-NEXT:    [[INDEX_NEXT4]] = add nuw i64 [[INDEX2]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]])
@@ -1402,14 +1366,14 @@ define i64 @diff_exit_block_pre_inc_use3() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP6]], i1 true)
 ; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX2]], [[FIRST_ACTIVE_LANE]]
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i64 3, [[TMP10]]
 ; CHECK-NEXT:    br label [[LOOP_EARLY_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -1473,11 +1437,9 @@ define i64 @diff_exit_block_post_inc_use1() {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP5]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP7]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP6]], align 1
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP14:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP13]])
@@ -1487,14 +1449,14 @@ define i64 @diff_exit_block_post_inc_use1() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP14]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP13]], i1 true)
 ; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i64 3, [[TMP10]]
 ; CHECK-NEXT:    br label [[LOOP_EARLY_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -1560,11 +1522,9 @@ define i64 @diff_exit_block_post_inc_use2() {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[TMP12]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP13]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP12]], align 1
 ; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[TMP14]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP15]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP14]], align 1
 ; CHECK-NEXT:    [[TMP17:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP18:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP17]])
@@ -1574,7 +1534,7 @@ define i64 @diff_exit_block_post_inc_use2() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP18]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP17]], i1 true)
 ; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
@@ -1582,7 +1542,7 @@ define i64 @diff_exit_block_post_inc_use2() {
 ; CHECK-NEXT:    [[TMP21:%.*]] = add i64 3, [[TMP11]]
 ; CHECK-NEXT:    br label [[LOOP_EARLY_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -1649,11 +1609,9 @@ define i64 @diff_exit_block_post_inc_use3(i64 %start) {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT4:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[TMP14]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP15]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP14]], align 1
 ; CHECK-NEXT:    [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP16]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP17]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD3:%.*]] = load <4 x i8>, ptr [[TMP16]], align 1
 ; CHECK-NEXT:    [[TMP19:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD3]]
 ; CHECK-NEXT:    [[INDEX_NEXT4]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP20:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP19]])
@@ -1664,7 +1622,7 @@ define i64 @diff_exit_block_post_inc_use3(i64 %start) {
 ; CHECK-NEXT:    br i1 [[TMP20]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[IND_ESCAPE:%.*]] = sub i64 [[TMP0]], 1
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP19]], i1 true)
 ; CHECK-NEXT:    [[TMP11:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
@@ -1672,8 +1630,8 @@ define i64 @diff_exit_block_post_inc_use3(i64 %start) {
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i64 [[START]], [[TMP12]]
 ; CHECK-NEXT:    br label [[LOOP_EARLY_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL5:%.*]] = phi i64 [ [[TMP0]], [[MIDDLE_BLOCK]] ], [ [[START]], [[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL5:%.*]] = phi i64 [ [[START]], [[ENTRY]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -1743,8 +1701,7 @@ define i64 @loop_contains_safe_call() {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = call fast <4 x float> @llvm.sqrt.v4f32(<4 x float> [[WIDE_LOAD]])
 ; CHECK-NEXT:    [[TMP5:%.*]] = fcmp fast oge <4 x float> [[TMP3]], splat (float 3.000000e+00)
 ; CHECK-NEXT:    [[INDEX_NEXT2]] = add nuw i64 [[INDEX1]], 4
@@ -1755,14 +1712,14 @@ define i64 @loop_contains_safe_call() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP5]], i1 true)
 ; CHECK-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i64 3, [[TMP9]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -1819,8 +1776,7 @@ define i64 @loop_contains_safe_div() {
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT2:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = udiv <4 x i32> [[WIDE_LOAD]], splat (i32 20000)
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp ne <4 x i32> [[TMP3]], splat (i32 1)
 ; CHECK-NEXT:    [[INDEX_NEXT2]] = add nuw i64 [[INDEX1]], 4
@@ -1831,14 +1787,14 @@ define i64 @loop_contains_safe_div() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP5]], i1 true)
 ; CHECK-NEXT:    [[TMP9:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i64 3, [[TMP9]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -1894,12 +1850,10 @@ define i64 @loop_contains_load_after_early_exit(ptr dereferenceable(1024) align(
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp ne <4 x i32> [[WIDE_LOAD]], splat (i32 1)
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i64, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i64, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i64>, ptr [[TMP5]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i64>, ptr [[TMP4]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]])
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT3]], 64
@@ -1909,14 +1863,14 @@ define i64 @loop_contains_load_after_early_exit(ptr dereferenceable(1024) align(
 ; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <4 x i64> [[WIDE_LOAD2]], i32 3
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP6]], i1 true)
 ; CHECK-NEXT:    [[TMP11:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i64 3, [[TMP11]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
@@ -1991,12 +1945,12 @@ define i64 @same_exit_block_pre_inc_use1_reverse() {
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP9]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP8]], i1 true)
 ; CHECK-NEXT:    [[TMP12:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = sub i64 1023, [[TMP12]]
-; CHECK-NEXT:    br label [[LOOP_END]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[MIDDLE_BLOCK]] ], [ 1023, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP1:%.*]]
@@ -2013,7 +1967,7 @@ define i64 @same_exit_block_pre_inc_use1_reverse() {
 ; CHECK-NEXT:    [[EXITCOND:%.*]] = icmp eq i64 [[INDEX_NEXT]], 0
 ; CHECK-NEXT:    br i1 [[EXITCOND]], label [[LOOP_END]], label [[LOOP1]], !llvm.loop [[LOOP51:![0-9]+]]
 ; CHECK:       loop.end:
-; CHECK-NEXT:    [[RETVAL:%.*]] = phi i64 [ [[INDEX]], [[LOOP1]] ], [ 1024, [[LOOP_INC]] ], [ 1024, [[MIDDLE_BLOCK]] ], [ [[EARLY_EXIT_VALUE]], [[VECTOR_EARLY_EXIT]] ]
+; CHECK-NEXT:    [[RETVAL:%.*]] = phi i64 [ [[INDEX]], [[LOOP1]] ], [ 1024, [[LOOP_INC]] ], [ [[EARLY_EXIT_VALUE]], [[VECTOR_EARLY_EXIT]] ]
 ; CHECK-NEXT:    ret i64 [[RETVAL]]
 ;
 entry:
@@ -2104,11 +2058,9 @@ define i64 @same_exit_block_pre_inc_use1_deref_ptrs(ptr dereferenceable(1024) %p
 ; CHECK-NEXT:    [[INDEX1:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT3:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX1]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[P1]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[P2]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP4]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD2:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp ne <4 x i8> [[WIDE_LOAD]], [[WIDE_LOAD2]]
 ; CHECK-NEXT:    [[INDEX_NEXT3]] = add nuw i64 [[INDEX1]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = call i1 @llvm.vector.reduce.or.v4i1(<4 x i1> [[TMP6]])
@@ -2118,14 +2070,14 @@ define i64 @same_exit_block_pre_inc_use1_deref_ptrs(ptr dereferenceable(1024) %p
 ; CHECK:       middle.split:
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[VECTOR_EARLY_EXIT:%.*]], label [[MIDDLE_BLOCK:%.*]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[LOOP_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[LOOP_END:%.*]]
 ; CHECK:       vector.early.exit:
 ; CHECK-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v4i1(<4 x i1> [[TMP6]], i1 true)
 ; CHECK-NEXT:    [[TMP10:%.*]] = add i64 [[INDEX1]], [[FIRST_ACTIVE_LANE]]
 ; CHECK-NEXT:    [[EARLY_EXIT_VALUE:%.*]] = add i64 3, [[TMP10]]
 ; CHECK-NEXT:    br label [[LOOP_END]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 67, [[MIDDLE_BLOCK]] ], [ 3, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 3, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ [[INDEX_NEXT:%.*]], [[LOOP_INC:%.*]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/single_early_exit_with_outer_loop.ll b/llvm/test/Transforms/LoopVectorize/single_early_exit_with_outer_loop.ll
index 5c1bde0..604e6a8 100644
--- a/llvm/test/Transforms/LoopVectorize/single_early_exit_with_outer_loop.ll
+++ b/llvm/test/Transforms/LoopVectorize/single_early_exit_with_outer_loop.ll
@@ -92,7 +92,7 @@ loop.outer.latch:
 
 define i32 @early_exit_branch_to_outer_header() {
 ; CHECK-LABEL: Loop info for function 'early_exit_branch_to_outer_header':
-; CHECK-NEXT:  Loop at depth 1 containing: %outer.header<header>,%loop.header,%loop.latch<exiting>,%outer.header.loopexit<latch>,%scalar.ph,%vector.ph,%vector.body,%middle.split,%middle.block<exiting>,%vector.early.exit
+; CHECK-NEXT:  Loop at depth 1 containing: %outer.header<header>,%loop.header,%loop.latch<exiting>,%outer.header.loopexit<latch>,%scalar.ph,%vector.ph,%vector.body,%middle.split<exiting>,%vector.early.exit
 ; CHECK-NEXT:    Loop at depth 2 containing: %loop.header<header><exiting>,%loop.latch<latch><exiting>
 ; CHECK-NEXT:    Loop at depth 2 containing: %vector.body<header><latch><exiting>
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/skeleton-lcssa-crash.ll b/llvm/test/Transforms/LoopVectorize/skeleton-lcssa-crash.ll
index 36dbc96..9c14a8c 100644
--- a/llvm/test/Transforms/LoopVectorize/skeleton-lcssa-crash.ll
+++ b/llvm/test/Transforms/LoopVectorize/skeleton-lcssa-crash.ll
@@ -46,8 +46,7 @@ define i16 @test(ptr %arg, i64 %N) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP4:%.*]] = add nuw nsw i64 [[INDEX]], 1
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[L_1]], i64 [[TMP4]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i16>, ptr [[TMP6]], align 2, !alias.scope [[META0:![0-9]+]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i16>, ptr [[TMP5]], align 2, !alias.scope [[META0:![0-9]+]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i16> [[WIDE_LOAD]], i32 1
 ; CHECK-NEXT:    store i16 [[TMP8]], ptr [[L_2]], align 2, !alias.scope [[META3:![0-9]+]], !noalias [[META0]]
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
diff --git a/llvm/test/Transforms/LoopVectorize/strided-accesses-interleave-only.ll b/llvm/test/Transforms/LoopVectorize/strided-accesses-interleave-only.ll
index 1c5a783..971921a 100644
--- a/llvm/test/Transforms/LoopVectorize/strided-accesses-interleave-only.ll
+++ b/llvm/test/Transforms/LoopVectorize/strided-accesses-interleave-only.ll
@@ -21,9 +21,9 @@ define void @test_variable_stride(ptr %dst, i32 %scale) {
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/struct-return-replicate.ll b/llvm/test/Transforms/LoopVectorize/struct-return-replicate.ll
index 717d1f9..1782086 100644
--- a/llvm/test/Transforms/LoopVectorize/struct-return-replicate.ll
+++ b/llvm/test/Transforms/LoopVectorize/struct-return-replicate.ll
@@ -12,8 +12,7 @@ define void @struct_return_1xi64_replicate(ptr noalias %in, ptr noalias writeonl
 ; VF4:       [[VECTOR_BODY]]:
 ; VF4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds float, ptr [[IN]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[TMP0]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
+; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP0]], align 4
 ; VF4-NEXT:    [[TMP2:%.*]] = extractelement <4 x float> [[WIDE_LOAD]], i32 0
 ; VF4-NEXT:    [[TMP3:%.*]] = tail call { i64 } @fn1(float [[TMP2]]) #[[ATTR0:[0-9]+]]
 ; VF4-NEXT:    [[TMP4:%.*]] = extractelement <4 x float> [[WIDE_LOAD]], i32 1
@@ -39,8 +38,7 @@ define void @struct_return_1xi64_replicate(ptr noalias %in, ptr noalias writeonl
 ; VF4-NEXT:    [[TMP24:%.*]] = insertvalue { <4 x i64> } [[TMP20]], <4 x i64> [[TMP23]], 0
 ; VF4-NEXT:    [[TMP25:%.*]] = extractvalue { <4 x i64> } [[TMP24]], 0
 ; VF4-NEXT:    [[TMP26:%.*]] = getelementptr inbounds i64, ptr [[OUT_A]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i64, ptr [[TMP26]], i32 0
-; VF4-NEXT:    store <4 x i64> [[TMP25]], ptr [[TMP27]], align 4
+; VF4-NEXT:    store <4 x i64> [[TMP25]], ptr [[TMP26]], align 4
 ; VF4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; VF4-NEXT:    [[TMP28:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; VF4-NEXT:    br i1 [[TMP28]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -55,9 +53,8 @@ define void @struct_return_1xi64_replicate(ptr noalias %in, ptr noalias writeonl
 ; VF2IC2:       [[VECTOR_BODY]]:
 ; VF2IC2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF2IC2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds float, ptr [[IN]], i64 [[INDEX]]
-; VF2IC2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[TMP0]], i32 0
 ; VF2IC2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP0]], i32 2
-; VF2IC2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP1]], align 4
+; VF2IC2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP0]], align 4
 ; VF2IC2-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x float>, ptr [[TMP2]], align 4
 ; VF2IC2-NEXT:    [[TMP3:%.*]] = extractelement <2 x float> [[WIDE_LOAD]], i32 0
 ; VF2IC2-NEXT:    [[TMP4:%.*]] = tail call { i64 } @fn1(float [[TMP3]]) #[[ATTR0:[0-9]+]]
@@ -84,9 +81,8 @@ define void @struct_return_1xi64_replicate(ptr noalias %in, ptr noalias writeonl
 ; VF2IC2-NEXT:    [[TMP25:%.*]] = extractvalue { <2 x i64> } [[TMP13]], 0
 ; VF2IC2-NEXT:    [[TMP26:%.*]] = extractvalue { <2 x i64> } [[TMP24]], 0
 ; VF2IC2-NEXT:    [[TMP27:%.*]] = getelementptr inbounds i64, ptr [[OUT_A]], i64 [[INDEX]]
-; VF2IC2-NEXT:    [[TMP28:%.*]] = getelementptr inbounds i64, ptr [[TMP27]], i32 0
 ; VF2IC2-NEXT:    [[TMP29:%.*]] = getelementptr inbounds i64, ptr [[TMP27]], i32 2
-; VF2IC2-NEXT:    store <2 x i64> [[TMP25]], ptr [[TMP28]], align 4
+; VF2IC2-NEXT:    store <2 x i64> [[TMP25]], ptr [[TMP27]], align 4
 ; VF2IC2-NEXT:    store <2 x i64> [[TMP26]], ptr [[TMP29]], align 4
 ; VF2IC2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; VF2IC2-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
@@ -122,8 +118,7 @@ define void @struct_return_2xf32_replicate(ptr noalias %in, ptr noalias writeonl
 ; VF4:       [[VECTOR_BODY]]:
 ; VF4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds float, ptr [[IN]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[TMP0]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
+; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP0]], align 4
 ; VF4-NEXT:    [[TMP2:%.*]] = extractelement <4 x float> [[WIDE_LOAD]], i32 0
 ; VF4-NEXT:    [[TMP3:%.*]] = tail call { float, float } @fn2(float [[TMP2]]) #[[ATTR1:[0-9]+]]
 ; VF4-NEXT:    [[TMP4:%.*]] = extractelement <4 x float> [[WIDE_LOAD]], i32 1
@@ -166,11 +161,9 @@ define void @struct_return_2xf32_replicate(ptr noalias %in, ptr noalias writeonl
 ; VF4-NEXT:    [[TMP41:%.*]] = extractvalue { <4 x float>, <4 x float> } [[TMP40]], 0
 ; VF4-NEXT:    [[TMP42:%.*]] = extractvalue { <4 x float>, <4 x float> } [[TMP40]], 1
 ; VF4-NEXT:    [[TMP43:%.*]] = getelementptr inbounds float, ptr [[OUT_A]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP44:%.*]] = getelementptr inbounds float, ptr [[TMP43]], i32 0
-; VF4-NEXT:    store <4 x float> [[TMP41]], ptr [[TMP44]], align 4
+; VF4-NEXT:    store <4 x float> [[TMP41]], ptr [[TMP43]], align 4
 ; VF4-NEXT:    [[TMP45:%.*]] = getelementptr inbounds float, ptr [[OUT_B]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP46:%.*]] = getelementptr inbounds float, ptr [[TMP45]], i32 0
-; VF4-NEXT:    store <4 x float> [[TMP42]], ptr [[TMP46]], align 4
+; VF4-NEXT:    store <4 x float> [[TMP42]], ptr [[TMP45]], align 4
 ; VF4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; VF4-NEXT:    [[TMP47:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; VF4-NEXT:    br i1 [[TMP47]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -185,9 +178,8 @@ define void @struct_return_2xf32_replicate(ptr noalias %in, ptr noalias writeonl
 ; VF2IC2:       [[VECTOR_BODY]]:
 ; VF2IC2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF2IC2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds float, ptr [[IN]], i64 [[INDEX]]
-; VF2IC2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds float, ptr [[TMP0]], i32 0
 ; VF2IC2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[TMP0]], i32 2
-; VF2IC2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP1]], align 4
+; VF2IC2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x float>, ptr [[TMP0]], align 4
 ; VF2IC2-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x float>, ptr [[TMP2]], align 4
 ; VF2IC2-NEXT:    [[TMP3:%.*]] = extractelement <2 x float> [[WIDE_LOAD]], i32 0
 ; VF2IC2-NEXT:    [[TMP4:%.*]] = tail call { float, float } @fn2(float [[TMP3]]) #[[ATTR1:[0-9]+]]
@@ -232,14 +224,12 @@ define void @struct_return_2xf32_replicate(ptr noalias %in, ptr noalias writeonl
 ; VF2IC2-NEXT:    [[TMP43:%.*]] = extractvalue { <2 x float>, <2 x float> } [[TMP21]], 1
 ; VF2IC2-NEXT:    [[TMP44:%.*]] = extractvalue { <2 x float>, <2 x float> } [[TMP40]], 1
 ; VF2IC2-NEXT:    [[TMP45:%.*]] = getelementptr inbounds float, ptr [[OUT_A]], i64 [[INDEX]]
-; VF2IC2-NEXT:    [[TMP46:%.*]] = getelementptr inbounds float, ptr [[TMP45]], i32 0
 ; VF2IC2-NEXT:    [[TMP47:%.*]] = getelementptr inbounds float, ptr [[TMP45]], i32 2
-; VF2IC2-NEXT:    store <2 x float> [[TMP41]], ptr [[TMP46]], align 4
+; VF2IC2-NEXT:    store <2 x float> [[TMP41]], ptr [[TMP45]], align 4
 ; VF2IC2-NEXT:    store <2 x float> [[TMP42]], ptr [[TMP47]], align 4
 ; VF2IC2-NEXT:    [[TMP48:%.*]] = getelementptr inbounds float, ptr [[OUT_B]], i64 [[INDEX]]
-; VF2IC2-NEXT:    [[TMP49:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i32 0
 ; VF2IC2-NEXT:    [[TMP50:%.*]] = getelementptr inbounds float, ptr [[TMP48]], i32 2
-; VF2IC2-NEXT:    store <2 x float> [[TMP43]], ptr [[TMP49]], align 4
+; VF2IC2-NEXT:    store <2 x float> [[TMP43]], ptr [[TMP48]], align 4
 ; VF2IC2-NEXT:    store <2 x float> [[TMP44]], ptr [[TMP50]], align 4
 ; VF2IC2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; VF2IC2-NEXT:    [[TMP51:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
@@ -279,8 +269,7 @@ define void @struct_return_3xi32_replicate(ptr noalias %in, ptr noalias writeonl
 ; VF4:       [[VECTOR_BODY]]:
 ; VF4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF4-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[IN]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 4
+; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP0]], align 4
 ; VF4-NEXT:    [[TMP2:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 0
 ; VF4-NEXT:    [[TMP3:%.*]] = tail call { i32, i32, i32 } @fn3(i32 [[TMP2]]) #[[ATTR2:[0-9]+]]
 ; VF4-NEXT:    [[TMP4:%.*]] = extractelement <4 x i32> [[WIDE_LOAD]], i32 1
@@ -338,16 +327,13 @@ define void @struct_return_3xi32_replicate(ptr noalias %in, ptr noalias writeonl
 ; VF4-NEXT:    [[TMP56:%.*]] = insertvalue { <4 x i32>, <4 x i32>, <4 x i32> } [[TMP52]], <4 x i32> [[TMP55]], 2
 ; VF4-NEXT:    [[TMP57:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32> } [[TMP56]], 0
 ; VF4-NEXT:    [[TMP58:%.*]] = getelementptr inbounds i32, ptr [[DST_A]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP59:%.*]] = getelementptr inbounds i32, ptr [[TMP58]], i32 0
-; VF4-NEXT:    store <4 x i32> [[TMP57]], ptr [[TMP59]], align 4
+; VF4-NEXT:    store <4 x i32> [[TMP57]], ptr [[TMP58]], align 4
 ; VF4-NEXT:    [[TMP60:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32> } [[TMP56]], 1
 ; VF4-NEXT:    [[TMP61:%.*]] = getelementptr inbounds i32, ptr [[DST_B]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP62:%.*]] = getelementptr inbounds i32, ptr [[TMP61]], i32 0
-; VF4-NEXT:    store <4 x i32> [[TMP60]], ptr [[TMP62]], align 4
+; VF4-NEXT:    store <4 x i32> [[TMP60]], ptr [[TMP61]], align 4
 ; VF4-NEXT:    [[TMP63:%.*]] = extractvalue { <4 x i32>, <4 x i32>, <4 x i32> } [[TMP56]], 2
 ; VF4-NEXT:    [[TMP64:%.*]] = getelementptr inbounds i32, ptr [[DST_C]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP65:%.*]] = getelementptr inbounds i32, ptr [[TMP64]], i32 0
-; VF4-NEXT:    store <4 x i32> [[TMP63]], ptr [[TMP65]], align 4
+; VF4-NEXT:    store <4 x i32> [[TMP63]], ptr [[TMP64]], align 4
 ; VF4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; VF4-NEXT:    [[TMP66:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
 ; VF4-NEXT:    br i1 [[TMP66]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
@@ -362,9 +348,8 @@ define void @struct_return_3xi32_replicate(ptr noalias %in, ptr noalias writeonl
 ; VF2IC2:       [[VECTOR_BODY]]:
 ; VF2IC2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF2IC2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i32, ptr [[IN]], i64 [[INDEX]]
-; VF2IC2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 0
 ; VF2IC2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP0]], i32 2
-; VF2IC2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP1]], align 4
+; VF2IC2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i32>, ptr [[TMP0]], align 4
 ; VF2IC2-NEXT:    [[WIDE_LOAD1:%.*]] = load <2 x i32>, ptr [[TMP2]], align 4
 ; VF2IC2-NEXT:    [[TMP3:%.*]] = extractelement <2 x i32> [[WIDE_LOAD]], i32 0
 ; VF2IC2-NEXT:    [[TMP4:%.*]] = tail call { i32, i32, i32 } @fn3(i32 [[TMP3]]) #[[ATTR2:[0-9]+]]
@@ -423,23 +408,20 @@ define void @struct_return_3xi32_replicate(ptr noalias %in, ptr noalias writeonl
 ; VF2IC2-NEXT:    [[TMP57:%.*]] = extractvalue { <2 x i32>, <2 x i32>, <2 x i32> } [[TMP29]], 0
 ; VF2IC2-NEXT:    [[TMP58:%.*]] = extractvalue { <2 x i32>, <2 x i32>, <2 x i32> } [[TMP56]], 0
 ; VF2IC2-NEXT:    [[TMP59:%.*]] = getelementptr inbounds i32, ptr [[DST_A]], i64 [[INDEX]]
-; VF2IC2-NEXT:    [[TMP60:%.*]] = getelementptr inbounds i32, ptr [[TMP59]], i32 0
 ; VF2IC2-NEXT:    [[TMP61:%.*]] = getelementptr inbounds i32, ptr [[TMP59]], i32 2
-; VF2IC2-NEXT:    store <2 x i32> [[TMP57]], ptr [[TMP60]], align 4
+; VF2IC2-NEXT:    store <2 x i32> [[TMP57]], ptr [[TMP59]], align 4
 ; VF2IC2-NEXT:    store <2 x i32> [[TMP58]], ptr [[TMP61]], align 4
 ; VF2IC2-NEXT:    [[TMP62:%.*]] = extractvalue { <2 x i32>, <2 x i32>, <2 x i32> } [[TMP29]], 1
 ; VF2IC2-NEXT:    [[TMP63:%.*]] = extractvalue { <2 x i32>, <2 x i32>, <2 x i32> } [[TMP56]], 1
 ; VF2IC2-NEXT:    [[TMP64:%.*]] = getelementptr inbounds i32, ptr [[DST_B]], i64 [[INDEX]]
-; VF2IC2-NEXT:    [[TMP65:%.*]] = getelementptr inbounds i32, ptr [[TMP64]], i32 0
 ; VF2IC2-NEXT:    [[TMP66:%.*]] = getelementptr inbounds i32, ptr [[TMP64]], i32 2
-; VF2IC2-NEXT:    store <2 x i32> [[TMP62]], ptr [[TMP65]], align 4
+; VF2IC2-NEXT:    store <2 x i32> [[TMP62]], ptr [[TMP64]], align 4
 ; VF2IC2-NEXT:    store <2 x i32> [[TMP63]], ptr [[TMP66]], align 4
 ; VF2IC2-NEXT:    [[TMP67:%.*]] = extractvalue { <2 x i32>, <2 x i32>, <2 x i32> } [[TMP29]], 2
 ; VF2IC2-NEXT:    [[TMP68:%.*]] = extractvalue { <2 x i32>, <2 x i32>, <2 x i32> } [[TMP56]], 2
 ; VF2IC2-NEXT:    [[TMP69:%.*]] = getelementptr inbounds i32, ptr [[DST_C]], i64 [[INDEX]]
-; VF2IC2-NEXT:    [[TMP70:%.*]] = getelementptr inbounds i32, ptr [[TMP69]], i32 0
 ; VF2IC2-NEXT:    [[TMP71:%.*]] = getelementptr inbounds i32, ptr [[TMP69]], i32 2
-; VF2IC2-NEXT:    store <2 x i32> [[TMP67]], ptr [[TMP70]], align 4
+; VF2IC2-NEXT:    store <2 x i32> [[TMP67]], ptr [[TMP69]], align 4
 ; VF2IC2-NEXT:    store <2 x i32> [[TMP68]], ptr [[TMP71]], align 4
 ; VF2IC2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; VF2IC2-NEXT:    [[TMP72:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
diff --git a/llvm/test/Transforms/LoopVectorize/tail-folding-optimize-vector-induction-width.ll b/llvm/test/Transforms/LoopVectorize/tail-folding-optimize-vector-induction-width.ll
index adc4c8e..efc2b8d 100644
--- a/llvm/test/Transforms/LoopVectorize/tail-folding-optimize-vector-induction-width.ll
+++ b/llvm/test/Transforms/LoopVectorize/tail-folding-optimize-vector-induction-width.ll
@@ -30,8 +30,8 @@ define void @canonical_small_tc_i8(ptr nocapture noundef writeonly %p) {
 ; CHECK:       [[PRED_STORE_CONTINUE2]]:
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i8> [[VEC_IND]], splat (i8 2)
-; CHECK-NEXT:    [[TMP25:%.*]] = icmp eq i64 [[INDEX_NEXT]], 16
-; CHECK-NEXT:    br i1 [[TMP25]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 16
+; CHECK-NEXT:    br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    br label %[[END:.*]]
 ; CHECK:       [[SCALAR_PH]]:
@@ -91,8 +91,8 @@ define void @canonical_upper_limit_i8(ptr nocapture noundef writeonly %p) {
 ; CHECK:       [[PRED_STORE_CONTINUE2]]:
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i8> [[VEC_IND]], splat (i8 2)
-; CHECK-NEXT:    [[TMP25:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
-; CHECK-NEXT:    br i1 [[TMP25]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
+; CHECK-NEXT:    br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    br label %[[END:.*]]
 ; CHECK:       [[SCALAR_PH]]:
@@ -213,8 +213,8 @@ define void @canonical_upper_limit_i16(ptr nocapture noundef writeonly %p) {
 ; CHECK:       [[PRED_STORE_CONTINUE2]]:
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i16> [[VEC_IND]], splat (i16 2)
-; CHECK-NEXT:    [[TMP25:%.*]] = icmp eq i64 [[INDEX_NEXT]], 65536
-; CHECK-NEXT:    br i1 [[TMP25]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 65536
+; CHECK-NEXT:    br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    br label %[[END:.*]]
 ; CHECK:       [[SCALAR_PH]]:
@@ -335,8 +335,8 @@ define void @canonical_upper_limit_i32(ptr nocapture noundef writeonly %p) {
 ; CHECK:       [[PRED_STORE_CONTINUE2]]:
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i32> [[VEC_IND]], splat (i32 2)
-; CHECK-NEXT:    [[TMP25:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4294967296
-; CHECK-NEXT:    br i1 [[TMP25]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
+; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 4294967296
+; CHECK-NEXT:    br i1 [[TMP7]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    br label %[[END:.*]]
 ; CHECK:       [[SCALAR_PH]]:
diff --git a/llvm/test/Transforms/LoopVectorize/trip-count-expansion-may-introduce-ub.ll b/llvm/test/Transforms/LoopVectorize/trip-count-expansion-may-introduce-ub.ll
index 514e858..586804b 100644
--- a/llvm/test/Transforms/LoopVectorize/trip-count-expansion-may-introduce-ub.ll
+++ b/llvm/test/Transforms/LoopVectorize/trip-count-expansion-may-introduce-ub.ll
@@ -23,8 +23,7 @@ define i64 @multi_exit_1_exit_count_with_udiv_by_value_in_header(ptr %dst, i64 %
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -88,8 +87,7 @@ define i64 @multi_exit_1_exit_count_with_udiv_by_constant_in_header(ptr %dst, i6
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
@@ -153,8 +151,7 @@ define i64 @multi_exit_2_exit_count_with_udiv_by_value_in_block_executed_uncondi
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD]], splat (i32 10)
 ; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <4 x i1> [[TMP7]], i32 0
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
@@ -265,8 +262,7 @@ define i64 @multi_exit_2_exit_count_with_udiv_by_constant_in_block_executed_unco
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[A]], i64 [[IV]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP6]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq <4 x i32> [[WIDE_LOAD]], splat (i32 10)
 ; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <4 x i1> [[TMP7]], i32 0
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[PRED_STORE_IF:%.*]], label [[PRED_STORE_CONTINUE:%.*]]
@@ -478,8 +474,7 @@ define i64 @multi_exit_4_exit_count_with_udiv_by_value_in_latch(ptr %dst, i64 %N
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
@@ -549,8 +544,7 @@ define i64 @multi_exit_4_exit_count_with_udiv_by_value_in_latch_call_before_loop
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP9]], align 4
+; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
@@ -620,8 +614,7 @@ define i64 @multi_exit_4_exit_count_with_udiv_by_value_in_latch_loop_may_not_exe
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP9]], align 4
+; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
@@ -691,8 +684,7 @@ define i64 @multi_exit_4_exit_count_with_udiv_by_value_in_latch_different_bounds
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP8]], i32 0
-; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP9]], align 4
+; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP8]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
@@ -759,8 +751,7 @@ define i64 @multi_exit_4_exit_count_with_udiv_by_frozen_value_in_latch(ptr %dst,
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP8]], align 4
+; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
@@ -826,8 +817,7 @@ define i64 @multi_exit_4_exit_count_with_udiv_by_constant_in_latch(ptr %dst, i64
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[LOOP_HEADER]] ]
 ; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[GEP]], i32 0
-; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[GEP]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP20:![0-9]+]]
@@ -887,8 +877,7 @@ define void @single_exit_tc_with_udiv(ptr %dst, i64 %N) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP4]], align 4
+; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP3]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
@@ -949,8 +938,7 @@ define i64 @multi_exit_4_exit_count_with_urem_by_value_in_latch(ptr %dst, i64 %N
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP8]], align 4
+; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP7]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
@@ -1015,8 +1003,7 @@ define i64 @multi_exit_4_exit_count_with_urem_by_constant_in_latch(ptr %dst, i64
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[LOOP_HEADER]] ]
 ; CHECK-NEXT:    [[GEP:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[GEP]], i32 0
-; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[GEP]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[LOOP_HEADER]], !llvm.loop [[LOOP26:![0-9]+]]
@@ -1164,8 +1151,7 @@ define i64 @multi_exit_4_exit_count_with_udiv_by_value_in_latch1(ptr %dst, i64 %
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP6]], align 4
+; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP5]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
@@ -1276,8 +1262,7 @@ define i64 @multi_exit_count_with_udiv_by_value_in_latch_different_bounds_diviso
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP7]], align 4
+; CHECK-NEXT:    store <4 x i32> splat (i32 1), ptr [[TMP6]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/trunc-extended-icmps.ll b/llvm/test/Transforms/LoopVectorize/trunc-extended-icmps.ll
index 26d8e37..a687ecc3 100644
--- a/llvm/test/Transforms/LoopVectorize/trunc-extended-icmps.ll
+++ b/llvm/test/Transforms/LoopVectorize/trunc-extended-icmps.ll
@@ -15,13 +15,12 @@ define i32 @test_icmp_constant_op_zext(ptr %dst) {
 ; CHECK-NEXT:    [[DOTCAST:%.*]] = trunc i32 [[INDEX]] to i16
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i16 1, [[DOTCAST]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[DST]], i16 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <4 x i8> splat (i8 109), ptr [[TMP2]], align 1
+; CHECK-NEXT:    store <4 x i8> splat (i8 109), ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 996
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 997, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
@@ -35,7 +34,7 @@ define i32 @test_icmp_constant_op_zext(ptr %dst) {
 ; CHECK-NEXT:    store i8 [[OR_TRUNC]], ptr [[GEP]], align 1
 ; CHECK-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
 ; CHECK-NEXT:    [[EC:%.*]] = icmp eq i16 [[IV_NEXT]], 1000
-; CHECK-NEXT:    br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP3:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret i32 0
 ;
@@ -78,13 +77,12 @@ define i32 @test_icmp_and_op_zext(ptr %dst, i64 %a) {
 ; CHECK-NEXT:    [[DOTCAST:%.*]] = trunc i32 [[INDEX]] to i16
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i16 1, [[DOTCAST]]
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[DST]], i16 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[TMP4]], ptr [[TMP6]], align 1
+; CHECK-NEXT:    store <4 x i8> [[TMP4]], ptr [[TMP5]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], 996
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 997, [[MIDDLE_BLOCK]] ], [ 1, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
@@ -98,7 +96,7 @@ define i32 @test_icmp_and_op_zext(ptr %dst, i64 %a) {
 ; CHECK-NEXT:    store i8 [[OR_TRUNC]], ptr [[GEP]], align 1
 ; CHECK-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
 ; CHECK-NEXT:    [[EC:%.*]] = icmp eq i16 [[IV_NEXT]], 1000
-; CHECK-NEXT:    br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP5:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret i32 0
 ;
@@ -132,25 +130,22 @@ define void @ext_cmp(ptr %src.1, ptr %src.2, ptr noalias %dst) {
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[SRC_1]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP2]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP1]], align 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp sgt <4 x i16> zeroinitializer, [[WIDE_LOAD]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[SRC_2]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i8>, ptr [[TMP5]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD1:%.*]] = load <4 x i8>, ptr [[TMP4]], align 2
 ; CHECK-NEXT:    [[TMP6:%.*]] = zext <4 x i8> [[WIDE_LOAD1]] to <4 x i16>
 ; CHECK-NEXT:    [[TMP7:%.*]] = select <4 x i1> [[TMP3]], <4 x i16> zeroinitializer, <4 x i16> [[TMP6]]
 ; CHECK-NEXT:    [[TMP8:%.*]] = and <4 x i16> [[TMP7]], zeroinitializer
 ; CHECK-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i16, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i16, ptr [[TMP9]], i32 0
-; CHECK-NEXT:    store <4 x i16> [[TMP8]], ptr [[TMP10]], align 2
+; CHECK-NEXT:    store <4 x i16> [[TMP8]], ptr [[TMP9]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/trunc-loads-p16.ll b/llvm/test/Transforms/LoopVectorize/trunc-loads-p16.ll
index 956a134..66dc785 100644
--- a/llvm/test/Transforms/LoopVectorize/trunc-loads-p16.ll
+++ b/llvm/test/Transforms/LoopVectorize/trunc-loads-p16.ll
@@ -17,20 +17,18 @@ define void @pr77468(ptr noalias %src, ptr noalias %dst, i1 %x) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i16
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i32, ptr [[SRC]], i16 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr i32, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP2]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i32>, ptr [[TMP1]], align 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = trunc <4 x i32> [[WIDE_LOAD]] to <4 x i16>
 ; CHECK-NEXT:    [[TMP5:%.*]] = and <4 x i16> [[TMP3]], [[TMP4]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr i16, ptr [[DST]], i16 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr i16, ptr [[TMP6]], i32 0
-; CHECK-NEXT:    store <4 x i16> [[TMP5]], ptr [[TMP7]], align 2
+; CHECK-NEXT:    store <4 x i16> [[TMP5]], ptr [[TMP6]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i32 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/trunc-reductions.ll b/llvm/test/Transforms/LoopVectorize/trunc-reductions.ll
index 3afd855..a592312 100644
--- a/llvm/test/Transforms/LoopVectorize/trunc-reductions.ll
+++ b/llvm/test/Transforms/LoopVectorize/trunc-reductions.ll
@@ -19,7 +19,7 @@ define i8 @reduction_and_trunc(ptr noalias nocapture %ptr) {
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP4:%.*]] = call i8 @llvm.vector.reduce.and.v8i8(<8 x i8> [[TMP2]])
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -66,7 +66,7 @@ define i16 @reduction_or_trunc(ptr noalias nocapture %ptr) {
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP4:%.*]] = call i16 @llvm.vector.reduce.or.v8i16(<8 x i16> [[TMP2]])
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -113,7 +113,7 @@ define i16 @reduction_xor_trunc(ptr noalias nocapture %ptr) {
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP4:%.*]] = call i16 @llvm.vector.reduce.xor.v8i16(<8 x i16> [[TMP2]])
-; CHECK-NEXT:    br i1 true, label [[FOR_END:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[FOR_END:%.*]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
@@ -325,7 +325,7 @@ define i32 @reduction_and_or(i16 %a, i32 %b, ptr %src) {
 ; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
 ; CHECK-NEXT:    [[TMP4:%.*]] = call i32 @llvm.vector.reduce.or.v8i32(<8 x i32> [[TMP2]])
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_MERGE_RDX:%.*]] = phi i32 [ [[TMP4]], [[MIDDLE_BLOCK]] ], [ poison, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
@@ -338,10 +338,9 @@ define i32 @reduction_and_or(i16 %a, i32 %b, ptr %src) {
 ; CHECK-NEXT:    [[OR]] = or i32 [[OR67]], [[L]]
 ; CHECK-NEXT:    [[IV_NEXT]] = add nuw nsw i32 [[IV]], 1
 ; CHECK-NEXT:    [[TOBOOL_NOT:%.*]] = icmp eq i32 [[IV_NEXT]], 999
-; CHECK-NEXT:    br i1 [[TOBOOL_NOT]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP9:![0-9]+]]
+; CHECK-NEXT:    br i1 [[TOBOOL_NOT]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP9:![0-9]+]]
 ; CHECK:       exit:
-; CHECK-NEXT:    [[OR_LCSSA:%.*]] = phi i32 [ [[OR]], [[LOOP]] ], [ poison, [[MIDDLE_BLOCK]] ]
-; CHECK-NEXT:    ret i32 [[OR_LCSSA]]
+; CHECK-NEXT:    ret i32 [[OR]]
 ;
 entry:
   %ext1 = zext i16 %a to i32
diff --git a/llvm/test/Transforms/LoopVectorize/trunc-shifts.ll b/llvm/test/Transforms/LoopVectorize/trunc-shifts.ll
index b94bd90..10e9ae8 100644
--- a/llvm/test/Transforms/LoopVectorize/trunc-shifts.ll
+++ b/llvm/test/Transforms/LoopVectorize/trunc-shifts.ll
@@ -19,15 +19,14 @@ define void @test_pr47927_lshr_const_shift_ops(ptr %dst, i32 %f) {
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i8
 ; CHECK-NEXT:    [[TMP3:%.*]] = zext i8 [[OFFSET_IDX]] to i64
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[TMP3]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[TMP1]], ptr [[TMP5]], align 8
+; CHECK-NEXT:    store <4 x i8> [[TMP1]], ptr [[TMP4]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -78,15 +77,14 @@ define void @test_shl_const_shift_ops(ptr %dst, i32 %f) {
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i8
 ; CHECK-NEXT:    [[TMP3:%.*]] = zext i8 [[OFFSET_IDX]] to i64
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[TMP3]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[TMP1]], ptr [[TMP5]], align 8
+; CHECK-NEXT:    store <4 x i8> [[TMP1]], ptr [[TMP4]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -137,15 +135,14 @@ define void @test_ashr_const_shift_ops(ptr %dst, i32 %f) {
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i8
 ; CHECK-NEXT:    [[TMP3:%.*]] = zext i8 [[OFFSET_IDX]] to i64
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[TMP3]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[TMP1]], ptr [[TMP5]], align 8
+; CHECK-NEXT:    store <4 x i8> [[TMP1]], ptr [[TMP4]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -192,19 +189,18 @@ define void @test_shl_const_shifted_op(ptr %dst, i32 %f) {
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i8
 ; CHECK-NEXT:    [[TMP1:%.*]] = zext i8 [[OFFSET_IDX]] to i64
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[DST]], i64 [[TMP1]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = zext <4 x i8> [[WIDE_LOAD]] to <4 x i32>
 ; CHECK-NEXT:    [[TMP5:%.*]] = shl <4 x i32> splat (i32 19), [[TMP4]]
 ; CHECK-NEXT:    [[TMP6:%.*]] = trunc <4 x i32> [[TMP5]] to <4 x i8>
-; CHECK-NEXT:    store <4 x i8> [[TMP6]], ptr [[TMP3]], align 8
+; CHECK-NEXT:    store <4 x i8> [[TMP6]], ptr [[TMP2]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -256,19 +252,18 @@ define void @test_lshr_by_18(ptr %A) {
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i8
 ; CHECK-NEXT:    [[TMP1:%.*]] = zext i8 [[OFFSET_IDX]] to i64
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[TMP1]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = zext <4 x i8> [[WIDE_LOAD]] to <4 x i32>
 ; CHECK-NEXT:    [[TMP5:%.*]] = lshr <4 x i32> [[TMP4]], splat (i32 18)
 ; CHECK-NEXT:    [[TMP6:%.*]] = trunc <4 x i32> [[TMP5]] to <4 x i8>
-; CHECK-NEXT:    store <4 x i8> [[TMP6]], ptr [[TMP3]], align 8
+; CHECK-NEXT:    store <4 x i8> [[TMP6]], ptr [[TMP2]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -319,19 +314,18 @@ define void @test_lshr_by_4(ptr %A) {
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i8
 ; CHECK-NEXT:    [[TMP1:%.*]] = zext i8 [[OFFSET_IDX]] to i64
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[TMP1]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP3]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
 ; CHECK-NEXT:    [[TMP4:%.*]] = zext <4 x i8> [[WIDE_LOAD]] to <4 x i16>
 ; CHECK-NEXT:    [[TMP5:%.*]] = lshr <4 x i16> [[TMP4]], splat (i16 4)
 ; CHECK-NEXT:    [[TMP6:%.*]] = trunc <4 x i16> [[TMP5]] to <4 x i8>
-; CHECK-NEXT:    store <4 x i8> [[TMP6]], ptr [[TMP3]], align 8
+; CHECK-NEXT:    store <4 x i8> [[TMP6]], ptr [[TMP2]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i32 [[INDEX_NEXT]], 100
 ; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/uitofp-preserve-nneg.ll b/llvm/test/Transforms/LoopVectorize/uitofp-preserve-nneg.ll
index 24d099c..c678175 100644
--- a/llvm/test/Transforms/LoopVectorize/uitofp-preserve-nneg.ll
+++ b/llvm/test/Transforms/LoopVectorize/uitofp-preserve-nneg.ll
@@ -16,16 +16,15 @@ define void @uitofp_preserve_nneg(ptr %result, i32 %size, float %y) {
 ; CHECK-NEXT:    [[TMP3:%.*]] = fmul <4 x float> [[TMP0]], [[BROADCAST_SPLAT3]]
 ; CHECK-NEXT:    [[INDEX:%.*]] = zext nneg i32 [[INDEX1]] to i64
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds float, ptr [[RESULT:%.*]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds float, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    store <4 x float> [[TMP3]], ptr [[TMP7]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP3]], ptr [[TMP2]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX1]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i32 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[FOR_EXIT:%.*]], label [[FOR_BODY_PREHEADER4]]
+; CHECK-NEXT:    br label [[FOR_EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 256, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[FOR_BODY:%.*]]
 ; CHECK:       for.body:
 ; CHECK-NEXT:    [[TMP4:%.*]] = phi i32 [ [[BC_RESUME_VAL]], [[FOR_BODY_PREHEADER4]] ], [ [[INC:%.*]], [[FOR_BODY]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/uniform-blend.ll b/llvm/test/Transforms/LoopVectorize/uniform-blend.ll
index 67d99dc..85cf925 100644
--- a/llvm/test/Transforms/LoopVectorize/uniform-blend.ll
+++ b/llvm/test/Transforms/LoopVectorize/uniform-blend.ll
@@ -15,15 +15,14 @@ define void @blend_uniform_iv_trunc(i1 %c) {
 ; CHECK-NEXT:    [[TMP0:%.*]] = trunc i64 [[INDEX]] to i16
 ; CHECK-NEXT:    [[TMP6:%.*]] = select i1 [[C]], i16 [[TMP0]], i16 poison
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [32 x i16], ptr @dst, i16 0, i16 [[TMP6]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i16, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <4 x i16> zeroinitializer, ptr [[TMP3]], align 2
+; CHECK-NEXT:    store <4 x i16> zeroinitializer, ptr [[TMP7]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32
 ; CHECK-NEXT:    br i1 [[TMP4]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 32, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -75,15 +74,14 @@ define void @blend_uniform_iv(i1 %c) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP6:%.*]] = select i1 [[C]], i64 [[INDEX]], i64 poison
 ; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds [32 x i16], ptr @dst, i16 0, i64 [[TMP6]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[TMP7]], i32 0
-; CHECK-NEXT:    store <4 x i16> zeroinitializer, ptr [[TMP2]], align 2
+; CHECK-NEXT:    store <4 x i16> zeroinitializer, ptr [[TMP7]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32
 ; CHECK-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 32, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -156,9 +154,9 @@ define void @blend_chain_iv(i1 %c) {
 ; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 32
 ; CHECK-NEXT:    br i1 [[TMP9]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 32, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; CHECK:       [[LOOP_HEADER]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction1.ll b/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction1.ll
index 3aad626..82f2fdd4 100644
--- a/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction1.ll
+++ b/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction1.ll
@@ -15,17 +15,15 @@ define void @ld_div1_step1_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = udiv i64 [[INDEX]], 1
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8
-; CHECK-NEXT:    [[TMP3:%.*]] = add nsw <2 x i64> [[WIDE_LOAD]], splat (i64 42)
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <2 x i64> [[TMP3]], ptr [[TMP5]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = add nsw <2 x i64> [[WIDE_LOAD]], splat (i64 42)
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
+; CHECK-NEXT:    store <2 x i64> [[TMP2]], ptr [[TMP3]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -63,13 +61,12 @@ define void @ld_div2_step1_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i64> poison, i64 [[TMP3]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i64> [[BROADCAST_SPLATINSERT]], <2 x i64> poison, <2 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP5]], align 8
+; CHECK-NEXT:    store <2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP4]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -111,14 +108,13 @@ define void @ld_div3_step1_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP8:%.*]] = insertelement <2 x i64> [[TMP7]], i64 [[TMP6]], i32 1
 ; CHECK-NEXT:    [[TMP9:%.*]] = add nsw <2 x i64> [[TMP8]], splat (i64 42)
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <2 x i64> [[TMP9]], ptr [[TMP11]], align 8
+; CHECK-NEXT:    store <2 x i64> [[TMP9]], ptr [[TMP10]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
-; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -173,7 +169,7 @@ define void @ld_div1_step2_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
 ; CHECK-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -208,20 +204,19 @@ define void @ld_div2_step2_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = add i64 [[OFFSET_IDX]], 2
 ; CHECK-NEXT:    [[TMP2:%.*]] = udiv i64 [[TMP0]], 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP2]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP4]], align 8
-; CHECK-NEXT:    [[TMP5:%.*]] = add nsw <2 x i64> [[WIDE_LOAD]], splat (i64 42)
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
-; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i64> [[TMP5]], i32 0
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP3]], align 8
+; CHECK-NEXT:    [[TMP4:%.*]] = add nsw <2 x i64> [[WIDE_LOAD]], splat (i64 42)
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP0]]
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
+; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <2 x i64> [[TMP4]], i32 0
+; CHECK-NEXT:    store i64 [[TMP7]], ptr [[TMP5]], align 8
+; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i64> [[TMP4]], i32 1
 ; CHECK-NEXT:    store i64 [[TMP8]], ptr [[TMP6]], align 8
-; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <2 x i64> [[TMP5]], i32 1
-; CHECK-NEXT:    store i64 [[TMP9]], ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
-; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
+; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -276,7 +271,7 @@ define void @ld_div3_step2_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
 ; CHECK-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -331,7 +326,7 @@ define void @ld_div1_step3_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; CHECK-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -386,7 +381,7 @@ define void @ld_div2_step3_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; CHECK-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -421,20 +416,19 @@ define void @ld_div3_step3_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP1:%.*]] = add i64 [[OFFSET_IDX]], 3
 ; CHECK-NEXT:    [[TMP2:%.*]] = udiv i64 [[TMP0]], 3
 ; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP2]]
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP3]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP4]], align 8
-; CHECK-NEXT:    [[TMP5:%.*]] = add nsw <2 x i64> [[WIDE_LOAD]], splat (i64 42)
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
-; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i64> [[TMP5]], i32 0
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP3]], align 8
+; CHECK-NEXT:    [[TMP4:%.*]] = add nsw <2 x i64> [[WIDE_LOAD]], splat (i64 42)
+; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP0]]
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
+; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <2 x i64> [[TMP4]], i32 0
+; CHECK-NEXT:    store i64 [[TMP7]], ptr [[TMP5]], align 8
+; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i64> [[TMP4]], i32 1
 ; CHECK-NEXT:    store i64 [[TMP8]], ptr [[TMP6]], align 8
-; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <2 x i64> [[TMP5]], i32 1
-; CHECK-NEXT:    store i64 [[TMP9]], ptr [[TMP7]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
-; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; CHECK-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
+; CHECK-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -467,17 +461,15 @@ define void @ld_div1_step1_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
 ; CHECK-NEXT:    [[TMP0:%.*]] = udiv i64 [[OFFSET_IDX]], 1
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8
-; CHECK-NEXT:    [[TMP3:%.*]] = add nsw <2 x i64> [[WIDE_LOAD]], splat (i64 42)
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <2 x i64> [[TMP3]], ptr [[TMP5]], align 8
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 8
+; CHECK-NEXT:    [[TMP2:%.*]] = add nsw <2 x i64> [[WIDE_LOAD]], splat (i64 42)
+; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[OFFSET_IDX]]
+; CHECK-NEXT:    store <2 x i64> [[TMP2]], ptr [[TMP3]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 998
-; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
+; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 998
+; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -520,14 +512,13 @@ define void @ld_div2_step1_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP8:%.*]] = insertelement <2 x i64> [[TMP7]], i64 [[TMP6]], i32 1
 ; CHECK-NEXT:    [[TMP9:%.*]] = add nsw <2 x i64> [[TMP8]], splat (i64 42)
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <2 x i64> [[TMP9]], ptr [[TMP11]], align 8
+; CHECK-NEXT:    store <2 x i64> [[TMP9]], ptr [[TMP10]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
-; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 998
-; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 998
+; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -570,14 +561,13 @@ define void @ld_div3_step1_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP8:%.*]] = insertelement <2 x i64> [[TMP7]], i64 [[TMP6]], i32 1
 ; CHECK-NEXT:    [[TMP9:%.*]] = add nsw <2 x i64> [[TMP8]], splat (i64 42)
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <2 x i64> [[TMP9]], ptr [[TMP11]], align 8
+; CHECK-NEXT:    store <2 x i64> [[TMP9]], ptr [[TMP10]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
-; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 998
-; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 998
+; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -633,7 +623,7 @@ define void @ld_div1_step2_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], 498
 ; CHECK-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -669,20 +659,19 @@ define void @ld_div2_step2_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP2:%.*]] = add i64 [[OFFSET_IDX]], 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = udiv i64 [[TMP1]], 2
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP3]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP5]], align 8
-; CHECK-NEXT:    [[TMP6:%.*]] = add nsw <2 x i64> [[WIDE_LOAD]], splat (i64 42)
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP2]]
-; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <2 x i64> [[TMP6]], i32 0
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP4]], align 8
+; CHECK-NEXT:    [[TMP5:%.*]] = add nsw <2 x i64> [[WIDE_LOAD]], splat (i64 42)
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
+; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP2]]
+; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i64> [[TMP5]], i32 0
+; CHECK-NEXT:    store i64 [[TMP8]], ptr [[TMP6]], align 8
+; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <2 x i64> [[TMP5]], i32 1
 ; CHECK-NEXT:    store i64 [[TMP9]], ptr [[TMP7]], align 8
-; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <2 x i64> [[TMP6]], i32 1
-; CHECK-NEXT:    store i64 [[TMP10]], ptr [[TMP8]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 498
-; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
+; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 498
+; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -738,7 +727,7 @@ define void @ld_div3_step2_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], 498
 ; CHECK-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -794,7 +783,7 @@ define void @ld_div1_step3_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; CHECK-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -850,7 +839,7 @@ define void @ld_div2_step3_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; CHECK-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -886,20 +875,19 @@ define void @ld_div3_step3_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP2:%.*]] = add i64 [[OFFSET_IDX]], 3
 ; CHECK-NEXT:    [[TMP3:%.*]] = udiv i64 [[TMP1]], 3
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP3]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP5]], align 8
-; CHECK-NEXT:    [[TMP6:%.*]] = add nsw <2 x i64> [[WIDE_LOAD]], splat (i64 42)
-; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
-; CHECK-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP2]]
-; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <2 x i64> [[TMP6]], i32 0
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP4]], align 8
+; CHECK-NEXT:    [[TMP5:%.*]] = add nsw <2 x i64> [[WIDE_LOAD]], splat (i64 42)
+; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
+; CHECK-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP2]]
+; CHECK-NEXT:    [[TMP8:%.*]] = extractelement <2 x i64> [[TMP5]], i32 0
+; CHECK-NEXT:    store i64 [[TMP8]], ptr [[TMP6]], align 8
+; CHECK-NEXT:    [[TMP9:%.*]] = extractelement <2 x i64> [[TMP5]], i32 1
 ; CHECK-NEXT:    store i64 [[TMP9]], ptr [[TMP7]], align 8
-; CHECK-NEXT:    [[TMP10:%.*]] = extractelement <2 x i64> [[TMP6]], i32 1
-; CHECK-NEXT:    store i64 [[TMP10]], ptr [[TMP8]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
-; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]
+; CHECK-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
+; CHECK-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -947,7 +935,7 @@ define void @test_step_is_not_invariant(ptr %A) {
 ; CHECK-NEXT:    [[TMP11:%.*]] = icmp eq i32 [[INDEX_NEXT]], 56
 ; CHECK-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP38:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction1_and.ll b/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction1_and.ll
index 1f331a4..af2b238 100644
--- a/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction1_and.ll
+++ b/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction1_and.ll
@@ -15,17 +15,8 @@ define void @ld_and_neg1_step1_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = and i64 [[INDEX]], -1
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP0]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8
-; CHECK-NEXT:    [[TMP3:%.*]] = add nsw <2 x i64> [[WIDE_LOAD]], splat (i64 42)
-; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <2 x i64> [[TMP3]], ptr [[TMP5]], align 8
-; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -63,13 +54,8 @@ define void @ld_and_neg2_step1_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i64> poison, i64 [[TMP3]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i64> [[BROADCAST_SPLATINSERT]], <2 x i64> poison, <2 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP5]], align 8
-; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -111,14 +97,8 @@ define void @ld_and_neg3_step1_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP8:%.*]] = insertelement <2 x i64> [[TMP7]], i64 [[TMP6]], i32 1
 ; CHECK-NEXT:    [[TMP9:%.*]] = add nsw <2 x i64> [[TMP8]], splat (i64 42)
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <2 x i64> [[TMP9]], ptr [[TMP11]], align 8
-; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
-; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -173,7 +153,7 @@ define void @ld_and_neg1_step2_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
 ; CHECK-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -218,7 +198,7 @@ define void @ld_and_neg2_step2_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
 ; CHECK-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -273,7 +253,7 @@ define void @ld_and_neg1_step3_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; CHECK-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -328,7 +308,7 @@ define void @ld_and_neg2_step3_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; CHECK-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -371,14 +351,8 @@ define void @ld_and_neg2_step1_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP8:%.*]] = insertelement <2 x i64> [[TMP7]], i64 [[TMP6]], i32 1
 ; CHECK-NEXT:    [[TMP9:%.*]] = add nsw <2 x i64> [[TMP8]], splat (i64 42)
 ; CHECK-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i32 0
-; CHECK-NEXT:    store <2 x i64> [[TMP9]], ptr [[TMP11]], align 8
-; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
-; CHECK-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 998
-; CHECK-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -434,7 +408,7 @@ define void @ld_and_neg2_step2_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], 498
 ; CHECK-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -490,7 +464,7 @@ define void @ld_and_neg2_step3_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; CHECK-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -546,7 +520,7 @@ define void @ld_and_neg3_step3_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; CHECK-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction1_div_urem.ll b/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction1_div_urem.ll
index ea8831c..61f511c 100644
--- a/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction1_div_urem.ll
+++ b/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction1_div_urem.ll
@@ -51,14 +51,13 @@ define void @ld_div2_urem3_1(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP33:%.*]] = insertelement <8 x i64> [[TMP32]], i64 [[TMP25]], i32 7
 ; CHECK-NEXT:    [[TMP34:%.*]] = add nsw <8 x i64> [[TMP33]], splat (i64 42)
 ; CHECK-NEXT:    [[TMP35:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP36:%.*]] = getelementptr inbounds i64, ptr [[TMP35]], i32 0
-; CHECK-NEXT:    store <8 x i64> [[TMP34]], ptr [[TMP36]], align 8
+; CHECK-NEXT:    store <8 x i64> [[TMP34]], ptr [[TMP35]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <8 x i64> [[VEC_IND]], splat (i64 8)
-; CHECK-NEXT:    [[TMP37:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; CHECK-NEXT:    br i1 [[TMP37]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-NEXT:    [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    br i1 [[TMP36]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -127,14 +126,13 @@ define void @ld_div2_urem3_2(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP34:%.*]] = insertelement <8 x i64> [[TMP33]], i64 [[TMP26]], i32 7
 ; CHECK-NEXT:    [[TMP35:%.*]] = add nsw <8 x i64> [[TMP34]], splat (i64 42)
 ; CHECK-NEXT:    [[TMP36:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP37:%.*]] = getelementptr inbounds i64, ptr [[TMP36]], i32 0
-; CHECK-NEXT:    store <8 x i64> [[TMP35]], ptr [[TMP37]], align 8
+; CHECK-NEXT:    store <8 x i64> [[TMP35]], ptr [[TMP36]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <8 x i64> [[VEC_IND]], splat (i64 8)
-; CHECK-NEXT:    [[TMP38:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; CHECK-NEXT:    br i1 [[TMP38]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK-NEXT:    [[TMP37:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    br i1 [[TMP37]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -201,14 +199,13 @@ define void @ld_div4(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[TMP32:%.*]] = insertelement <8 x i64> [[TMP31]], i64 [[TMP24]], i32 7
 ; CHECK-NEXT:    [[TMP33:%.*]] = add nsw <8 x i64> [[TMP32]], splat (i64 42)
 ; CHECK-NEXT:    [[TMP34:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP35:%.*]] = getelementptr inbounds i64, ptr [[TMP34]], i32 0
-; CHECK-NEXT:    store <8 x i64> [[TMP33]], ptr [[TMP35]], align 8
+; CHECK-NEXT:    store <8 x i64> [[TMP33]], ptr [[TMP34]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <8 x i64> [[VEC_IND]], splat (i64 8)
-; CHECK-NEXT:    [[TMP36:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; CHECK-NEXT:    br i1 [[TMP36]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK-NEXT:    [[TMP35:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    br i1 [[TMP35]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ;
 entry:
@@ -247,13 +244,12 @@ define void @ld_div8_urem3(ptr noalias %A, ptr noalias %B) {
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <8 x i64> poison, i64 [[TMP4]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <8 x i64> [[BROADCAST_SPLATINSERT]], <8 x i64> poison, <8 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 0
-; CHECK-NEXT:    store <8 x i64> [[BROADCAST_SPLAT]], ptr [[TMP6]], align 8
+; CHECK-NEXT:    store <8 x i64> [[BROADCAST_SPLAT]], ptr [[TMP5]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
-; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction1_lshr.ll b/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction1_lshr.ll
index 1f33f7a..e412d13 100644
--- a/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction1_lshr.ll
+++ b/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction1_lshr.ll
@@ -16,17 +16,15 @@ define void @ld_lshr0_step1_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; VF2-NEXT:    [[TMP0:%.*]] = lshr i64 [[INDEX]], 0
 ; VF2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP0]]
-; VF2-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP2]], align 8
-; VF2-NEXT:    [[TMP3:%.*]] = add nsw <2 x i64> [[WIDE_LOAD]], splat (i64 42)
-; VF2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; VF2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
-; VF2-NEXT:    store <2 x i64> [[TMP3]], ptr [[TMP5]], align 8
+; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP1]], align 8
+; VF2-NEXT:    [[TMP2:%.*]] = add nsw <2 x i64> [[WIDE_LOAD]], splat (i64 42)
+; VF2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
+; VF2-NEXT:    store <2 x i64> [[TMP2]], ptr [[TMP3]], align 8
 ; VF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; VF2-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; VF2-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; VF2-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; VF2-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[EXIT:%.*]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_lshr0_step1_start0_ind1
@@ -39,17 +37,15 @@ define void @ld_lshr0_step1_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; VF4-NEXT:    [[TMP0:%.*]] = lshr i64 [[INDEX]], 0
 ; VF4-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP0]]
-; VF4-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP2]], align 8
-; VF4-NEXT:    [[TMP3:%.*]] = add nsw <4 x i64> [[WIDE_LOAD]], splat (i64 42)
-; VF4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
-; VF4-NEXT:    store <4 x i64> [[TMP3]], ptr [[TMP5]], align 8
+; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP1]], align 8
+; VF4-NEXT:    [[TMP2:%.*]] = add nsw <4 x i64> [[WIDE_LOAD]], splat (i64 42)
+; VF4-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
+; VF4-NEXT:    store <4 x i64> [[TMP2]], ptr [[TMP3]], align 8
 ; VF4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; VF4-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; VF4-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; VF4-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; VF4-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[EXIT:%.*]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -87,13 +83,12 @@ define void @ld_lshr1_step1_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i64> poison, i64 [[TMP3]], i64 0
 ; VF2-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i64> [[BROADCAST_SPLATINSERT]], <2 x i64> poison, <2 x i32> zeroinitializer
 ; VF2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; VF2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
-; VF2-NEXT:    store <2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP5]], align 8
+; VF2-NEXT:    store <2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP4]], align 8
 ; VF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; VF2-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; VF2-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; VF2-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; VF2-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[EXIT:%.*]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_lshr1_step1_start0_ind1
@@ -124,14 +119,13 @@ define void @ld_lshr1_step1_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP16:%.*]] = insertelement <4 x i64> [[TMP15]], i64 [[TMP12]], i32 3
 ; VF4-NEXT:    [[TMP17:%.*]] = add nsw <4 x i64> [[TMP16]], splat (i64 42)
 ; VF4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i64, ptr [[TMP18]], i32 0
-; VF4-NEXT:    store <4 x i64> [[TMP17]], ptr [[TMP19]], align 8
+; VF4-NEXT:    store <4 x i64> [[TMP17]], ptr [[TMP18]], align 8
 ; VF4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; VF4-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
-; VF4-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; VF4-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; VF4-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; VF4-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[EXIT:%.*]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -169,13 +163,12 @@ define void @ld_lshr2_step1_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i64> poison, i64 [[TMP3]], i64 0
 ; VF2-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i64> [[BROADCAST_SPLATINSERT]], <2 x i64> poison, <2 x i32> zeroinitializer
 ; VF2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; VF2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
-; VF2-NEXT:    store <2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP5]], align 8
+; VF2-NEXT:    store <2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP4]], align 8
 ; VF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; VF2-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; VF2-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; VF2-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; VF2-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[EXIT:%.*]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_lshr2_step1_start0_ind1
@@ -193,13 +186,12 @@ define void @ld_lshr2_step1_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <4 x i64> poison, i64 [[TMP3]], i64 0
 ; VF4-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x i64> [[BROADCAST_SPLATINSERT]], <4 x i64> poison, <4 x i32> zeroinitializer
 ; VF4-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
-; VF4-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP5]], align 8
+; VF4-NEXT:    store <4 x i64> [[BROADCAST_SPLAT]], ptr [[TMP4]], align 8
 ; VF4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; VF4-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; VF4-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; VF4-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; VF4-NEXT:    br i1 [[TMP5]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[EXIT:%.*]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -254,7 +246,7 @@ define void @ld_lshr0_step2_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
 ; VF2-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[EXIT:%.*]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_lshr0_step2_start0_ind1
@@ -306,7 +298,7 @@ define void @ld_lshr0_step2_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
 ; VF4-NEXT:    br i1 [[TMP30]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[EXIT:%.*]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -341,20 +333,19 @@ define void @ld_lshr1_step2_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP1:%.*]] = add i64 [[OFFSET_IDX]], 2
 ; VF2-NEXT:    [[TMP2:%.*]] = lshr i64 [[TMP0]], 1
 ; VF2-NEXT:    [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP2]]
-; VF2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP3]], i32 0
-; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP4]], align 8
-; VF2-NEXT:    [[TMP5:%.*]] = add nsw <2 x i64> [[WIDE_LOAD]], splat (i64 42)
-; VF2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP0]]
-; VF2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
-; VF2-NEXT:    [[TMP8:%.*]] = extractelement <2 x i64> [[TMP5]], i32 0
+; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP3]], align 8
+; VF2-NEXT:    [[TMP4:%.*]] = add nsw <2 x i64> [[WIDE_LOAD]], splat (i64 42)
+; VF2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP0]]
+; VF2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
+; VF2-NEXT:    [[TMP7:%.*]] = extractelement <2 x i64> [[TMP4]], i32 0
+; VF2-NEXT:    store i64 [[TMP7]], ptr [[TMP5]], align 8
+; VF2-NEXT:    [[TMP8:%.*]] = extractelement <2 x i64> [[TMP4]], i32 1
 ; VF2-NEXT:    store i64 [[TMP8]], ptr [[TMP6]], align 8
-; VF2-NEXT:    [[TMP9:%.*]] = extractelement <2 x i64> [[TMP5]], i32 1
-; VF2-NEXT:    store i64 [[TMP9]], ptr [[TMP7]], align 8
 ; VF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; VF2-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
-; VF2-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; VF2-NEXT:    [[TMP9:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
+; VF2-NEXT:    br i1 [[TMP9]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[EXIT:%.*]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_lshr1_step2_start0_ind1
@@ -372,26 +363,25 @@ define void @ld_lshr1_step2_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP3:%.*]] = add i64 [[OFFSET_IDX]], 6
 ; VF4-NEXT:    [[TMP4:%.*]] = lshr i64 [[TMP0]], 1
 ; VF4-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP4]]
-; VF4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[TMP5]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP6]], align 8
-; VF4-NEXT:    [[TMP7:%.*]] = add nsw <4 x i64> [[WIDE_LOAD]], splat (i64 42)
-; VF4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP0]]
-; VF4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
-; VF4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP2]]
-; VF4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP3]]
-; VF4-NEXT:    [[TMP12:%.*]] = extractelement <4 x i64> [[TMP7]], i32 0
+; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP5]], align 8
+; VF4-NEXT:    [[TMP6:%.*]] = add nsw <4 x i64> [[WIDE_LOAD]], splat (i64 42)
+; VF4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP0]]
+; VF4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
+; VF4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP2]]
+; VF4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP3]]
+; VF4-NEXT:    [[TMP11:%.*]] = extractelement <4 x i64> [[TMP6]], i32 0
+; VF4-NEXT:    store i64 [[TMP11]], ptr [[TMP7]], align 8
+; VF4-NEXT:    [[TMP12:%.*]] = extractelement <4 x i64> [[TMP6]], i32 1
 ; VF4-NEXT:    store i64 [[TMP12]], ptr [[TMP8]], align 8
-; VF4-NEXT:    [[TMP13:%.*]] = extractelement <4 x i64> [[TMP7]], i32 1
+; VF4-NEXT:    [[TMP13:%.*]] = extractelement <4 x i64> [[TMP6]], i32 2
 ; VF4-NEXT:    store i64 [[TMP13]], ptr [[TMP9]], align 8
-; VF4-NEXT:    [[TMP14:%.*]] = extractelement <4 x i64> [[TMP7]], i32 2
+; VF4-NEXT:    [[TMP14:%.*]] = extractelement <4 x i64> [[TMP6]], i32 3
 ; VF4-NEXT:    store i64 [[TMP14]], ptr [[TMP10]], align 8
-; VF4-NEXT:    [[TMP15:%.*]] = extractelement <4 x i64> [[TMP7]], i32 3
-; VF4-NEXT:    store i64 [[TMP15]], ptr [[TMP11]], align 8
 ; VF4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; VF4-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
-; VF4-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
+; VF4-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
+; VF4-NEXT:    br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[EXIT:%.*]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -446,7 +436,7 @@ define void @ld_lshr0_step3_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF2-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[SCALAR_PH]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_lshr0_step3_start0_ind1
@@ -498,7 +488,7 @@ define void @ld_lshr0_step3_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF4-NEXT:    br i1 [[TMP30]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[SCALAR_PH]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -553,7 +543,7 @@ define void @ld_lshr1_step3_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF2-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[SCALAR_PH]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_lshr1_step3_start0_ind1
@@ -605,7 +595,7 @@ define void @ld_lshr1_step3_start0_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP30:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF4-NEXT:    br i1 [[TMP30]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[SCALAR_PH]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -649,14 +639,13 @@ define void @ld_lshr1_step1_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP8:%.*]] = insertelement <2 x i64> [[TMP7]], i64 [[TMP6]], i32 1
 ; VF2-NEXT:    [[TMP9:%.*]] = add nsw <2 x i64> [[TMP8]], splat (i64 42)
 ; VF2-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[OFFSET_IDX]]
-; VF2-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i32 0
-; VF2-NEXT:    store <2 x i64> [[TMP9]], ptr [[TMP11]], align 8
+; VF2-NEXT:    store <2 x i64> [[TMP9]], ptr [[TMP10]], align 8
 ; VF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VF2-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
-; VF2-NEXT:    [[TMP12:%.*]] = icmp eq i64 [[INDEX_NEXT]], 998
-; VF2-NEXT:    br i1 [[TMP12]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; VF2-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 998
+; VF2-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[SCALAR_PH]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_lshr1_step1_start1_ind1
@@ -688,14 +677,13 @@ define void @ld_lshr1_step1_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP16:%.*]] = insertelement <4 x i64> [[TMP15]], i64 [[TMP12]], i32 3
 ; VF4-NEXT:    [[TMP17:%.*]] = add nsw <4 x i64> [[TMP16]], splat (i64 42)
 ; VF4-NEXT:    [[TMP18:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[OFFSET_IDX]]
-; VF4-NEXT:    [[TMP19:%.*]] = getelementptr inbounds i64, ptr [[TMP18]], i32 0
-; VF4-NEXT:    store <4 x i64> [[TMP17]], ptr [[TMP19]], align 8
+; VF4-NEXT:    store <4 x i64> [[TMP17]], ptr [[TMP18]], align 8
 ; VF4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; VF4-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
-; VF4-NEXT:    [[TMP20:%.*]] = icmp eq i64 [[INDEX_NEXT]], 996
-; VF4-NEXT:    br i1 [[TMP20]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
+; VF4-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], 996
+; VF4-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[SCALAR_PH]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -731,20 +719,19 @@ define void @ld_lshr1_step2_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP2:%.*]] = add i64 [[OFFSET_IDX]], 2
 ; VF2-NEXT:    [[TMP3:%.*]] = lshr i64 [[TMP1]], 1
 ; VF2-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP3]]
-; VF2-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP4]], i32 0
-; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP5]], align 8
-; VF2-NEXT:    [[TMP6:%.*]] = add nsw <2 x i64> [[WIDE_LOAD]], splat (i64 42)
-; VF2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
-; VF2-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP2]]
-; VF2-NEXT:    [[TMP9:%.*]] = extractelement <2 x i64> [[TMP6]], i32 0
+; VF2-NEXT:    [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP4]], align 8
+; VF2-NEXT:    [[TMP5:%.*]] = add nsw <2 x i64> [[WIDE_LOAD]], splat (i64 42)
+; VF2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
+; VF2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP2]]
+; VF2-NEXT:    [[TMP8:%.*]] = extractelement <2 x i64> [[TMP5]], i32 0
+; VF2-NEXT:    store i64 [[TMP8]], ptr [[TMP6]], align 8
+; VF2-NEXT:    [[TMP9:%.*]] = extractelement <2 x i64> [[TMP5]], i32 1
 ; VF2-NEXT:    store i64 [[TMP9]], ptr [[TMP7]], align 8
-; VF2-NEXT:    [[TMP10:%.*]] = extractelement <2 x i64> [[TMP6]], i32 1
-; VF2-NEXT:    store i64 [[TMP10]], ptr [[TMP8]], align 8
 ; VF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; VF2-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], 498
-; VF2-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; VF2-NEXT:    [[TMP10:%.*]] = icmp eq i64 [[INDEX_NEXT]], 498
+; VF2-NEXT:    br i1 [[TMP10]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[SCALAR_PH]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_lshr1_step2_start1_ind1
@@ -763,26 +750,25 @@ define void @ld_lshr1_step2_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP4:%.*]] = add i64 [[OFFSET_IDX]], 6
 ; VF4-NEXT:    [[TMP5:%.*]] = lshr i64 [[TMP1]], 1
 ; VF4-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 [[TMP5]]
-; VF4-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i32 0
-; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP7]], align 8
-; VF4-NEXT:    [[TMP8:%.*]] = add nsw <4 x i64> [[WIDE_LOAD]], splat (i64 42)
-; VF4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
-; VF4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP2]]
-; VF4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP3]]
-; VF4-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP4]]
-; VF4-NEXT:    [[TMP13:%.*]] = extractelement <4 x i64> [[TMP8]], i32 0
+; VF4-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i64>, ptr [[TMP6]], align 8
+; VF4-NEXT:    [[TMP7:%.*]] = add nsw <4 x i64> [[WIDE_LOAD]], splat (i64 42)
+; VF4-NEXT:    [[TMP8:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP1]]
+; VF4-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP2]]
+; VF4-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP3]]
+; VF4-NEXT:    [[TMP11:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[TMP4]]
+; VF4-NEXT:    [[TMP12:%.*]] = extractelement <4 x i64> [[TMP7]], i32 0
+; VF4-NEXT:    store i64 [[TMP12]], ptr [[TMP8]], align 8
+; VF4-NEXT:    [[TMP13:%.*]] = extractelement <4 x i64> [[TMP7]], i32 1
 ; VF4-NEXT:    store i64 [[TMP13]], ptr [[TMP9]], align 8
-; VF4-NEXT:    [[TMP14:%.*]] = extractelement <4 x i64> [[TMP8]], i32 1
+; VF4-NEXT:    [[TMP14:%.*]] = extractelement <4 x i64> [[TMP7]], i32 2
 ; VF4-NEXT:    store i64 [[TMP14]], ptr [[TMP10]], align 8
-; VF4-NEXT:    [[TMP15:%.*]] = extractelement <4 x i64> [[TMP8]], i32 2
+; VF4-NEXT:    [[TMP15:%.*]] = extractelement <4 x i64> [[TMP7]], i32 3
 ; VF4-NEXT:    store i64 [[TMP15]], ptr [[TMP11]], align 8
-; VF4-NEXT:    [[TMP16:%.*]] = extractelement <4 x i64> [[TMP8]], i32 3
-; VF4-NEXT:    store i64 [[TMP16]], ptr [[TMP12]], align 8
 ; VF4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; VF4-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], 496
-; VF4-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
+; VF4-NEXT:    [[TMP16:%.*]] = icmp eq i64 [[INDEX_NEXT]], 496
+; VF4-NEXT:    br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[SCALAR_PH]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -838,7 +824,7 @@ define void @ld_lshr1_step3_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF2-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[SCALAR_PH]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_lshr1_step3_start1_ind1
@@ -891,7 +877,7 @@ define void @ld_lshr1_step3_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP31:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF4-NEXT:    br i1 [[TMP31]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[SCALAR_PH]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -947,7 +933,7 @@ define void @ld_lshr2_step3_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP17:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF2-NEXT:    br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[SCALAR_PH]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_lshr2_step3_start1_ind1
@@ -1000,7 +986,7 @@ define void @ld_lshr2_step3_start1_ind1(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP31:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF4-NEXT:    br i1 [[TMP31]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[SCALAR_PH]]
 ; VF4:       scalar.ph:
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction2.ll b/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction2.ll
index ef62557..ef6ce08 100644
--- a/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction2.ll
+++ b/llvm/test/Transforms/LoopVectorize/uniform_across_vf_induction2.ll
@@ -27,15 +27,14 @@ define void @ld_div1_step1_start0_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP10:%.*]] = insertelement <2 x i64> [[TMP9]], i64 [[TMP8]], i32 1
 ; VF2-NEXT:    [[TMP11:%.*]] = add nsw <2 x i64> [[TMP10]], splat (i64 42)
 ; VF2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; VF2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i64, ptr [[TMP12]], i32 0
-; VF2-NEXT:    store <2 x i64> [[TMP11]], ptr [[TMP13]], align 8
+; VF2-NEXT:    store <2 x i64> [[TMP11]], ptr [[TMP12]], align 8
 ; VF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VF2-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; VF2-NEXT:    [[VEC_IND_NEXT2]] = add <2 x i64> [[VEC_IND1]], splat (i64 2)
-; VF2-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; VF2-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; VF2-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; VF2-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[EXIT:%.*]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_div1_step1_start0_ind2
@@ -69,15 +68,14 @@ define void @ld_div1_step1_start0_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP18:%.*]] = insertelement <4 x i64> [[TMP17]], i64 [[TMP14]], i32 3
 ; VF4-NEXT:    [[TMP19:%.*]] = add nsw <4 x i64> [[TMP18]], splat (i64 42)
 ; VF4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i64, ptr [[TMP20]], i32 0
-; VF4-NEXT:    store <4 x i64> [[TMP19]], ptr [[TMP21]], align 8
+; VF4-NEXT:    store <4 x i64> [[TMP19]], ptr [[TMP20]], align 8
 ; VF4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; VF4-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; VF4-NEXT:    [[VEC_IND_NEXT2]] = add <4 x i64> [[VEC_IND1]], splat (i64 4)
-; VF4-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; VF4-NEXT:    br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; VF4-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; VF4-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[EXIT:%.*]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -121,13 +119,12 @@ define void @ld_div2_step1_start0_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[BROADCAST_SPLATINSERT:%.*]] = insertelement <2 x i64> poison, i64 [[TMP5]], i64 0
 ; VF2-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <2 x i64> [[BROADCAST_SPLATINSERT]], <2 x i64> poison, <2 x i32> zeroinitializer
 ; VF2-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; VF2-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i64, ptr [[TMP6]], i32 0
-; VF2-NEXT:    store <2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP7]], align 8
+; VF2-NEXT:    store <2 x i64> [[BROADCAST_SPLAT]], ptr [[TMP6]], align 8
 ; VF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
-; VF2-NEXT:    [[TMP8:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; VF2-NEXT:    br i1 [[TMP8]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; VF2-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; VF2-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[EXIT:%.*]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_div2_step1_start0_ind2
@@ -161,15 +158,14 @@ define void @ld_div2_step1_start0_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP18:%.*]] = insertelement <4 x i64> [[TMP17]], i64 [[TMP14]], i32 3
 ; VF4-NEXT:    [[TMP19:%.*]] = add nsw <4 x i64> [[TMP18]], splat (i64 42)
 ; VF4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i64, ptr [[TMP20]], i32 0
-; VF4-NEXT:    store <4 x i64> [[TMP19]], ptr [[TMP21]], align 8
+; VF4-NEXT:    store <4 x i64> [[TMP19]], ptr [[TMP20]], align 8
 ; VF4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; VF4-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; VF4-NEXT:    [[VEC_IND_NEXT2]] = add <4 x i64> [[VEC_IND1]], splat (i64 4)
-; VF4-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; VF4-NEXT:    br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; VF4-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; VF4-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[EXIT:%.*]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -218,15 +214,14 @@ define void @ld_div3_step1_start0_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP10:%.*]] = insertelement <2 x i64> [[TMP9]], i64 [[TMP8]], i32 1
 ; VF2-NEXT:    [[TMP11:%.*]] = add nsw <2 x i64> [[TMP10]], splat (i64 42)
 ; VF2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; VF2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i64, ptr [[TMP12]], i32 0
-; VF2-NEXT:    store <2 x i64> [[TMP11]], ptr [[TMP13]], align 8
+; VF2-NEXT:    store <2 x i64> [[TMP11]], ptr [[TMP12]], align 8
 ; VF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VF2-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; VF2-NEXT:    [[VEC_IND_NEXT2]] = add <2 x i64> [[VEC_IND1]], splat (i64 2)
-; VF2-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; VF2-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; VF2-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; VF2-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[EXIT:%.*]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_div3_step1_start0_ind2
@@ -260,15 +255,14 @@ define void @ld_div3_step1_start0_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP18:%.*]] = insertelement <4 x i64> [[TMP17]], i64 [[TMP14]], i32 3
 ; VF4-NEXT:    [[TMP19:%.*]] = add nsw <4 x i64> [[TMP18]], splat (i64 42)
 ; VF4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[INDEX]]
-; VF4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i64, ptr [[TMP20]], i32 0
-; VF4-NEXT:    store <4 x i64> [[TMP19]], ptr [[TMP21]], align 8
+; VF4-NEXT:    store <4 x i64> [[TMP19]], ptr [[TMP20]], align 8
 ; VF4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; VF4-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; VF4-NEXT:    [[VEC_IND_NEXT2]] = add <4 x i64> [[VEC_IND1]], splat (i64 4)
-; VF4-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
-; VF4-NEXT:    br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; VF4-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1000
+; VF4-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[EXIT:%.*]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -331,7 +325,7 @@ define void @ld_div1_step2_start0_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
 ; VF2-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[EXIT:%.*]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_div1_step2_start0_ind2
@@ -387,7 +381,7 @@ define void @ld_div1_step2_start0_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
 ; VF4-NEXT:    br i1 [[TMP32]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[EXIT:%.*]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -450,7 +444,7 @@ define void @ld_div2_step2_start0_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
 ; VF2-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[EXIT:%.*]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_div2_step2_start0_ind2
@@ -506,7 +500,7 @@ define void @ld_div2_step2_start0_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
 ; VF4-NEXT:    br i1 [[TMP32]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[EXIT:%.*]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -569,7 +563,7 @@ define void @ld_div3_step2_start0_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
 ; VF2-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[EXIT:%.*]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_div3_step2_start0_ind2
@@ -625,7 +619,7 @@ define void @ld_div3_step2_start0_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], 500
 ; VF4-NEXT:    br i1 [[TMP32]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[EXIT:%.*]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -688,7 +682,7 @@ define void @ld_div1_step3_start0_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF2-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[SCALAR_PH]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_div1_step3_start0_ind2
@@ -744,7 +738,7 @@ define void @ld_div1_step3_start0_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF4-NEXT:    br i1 [[TMP32]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP14:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[SCALAR_PH]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -807,7 +801,7 @@ define void @ld_div2_step3_start0_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF2-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[SCALAR_PH]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_div2_step3_start0_ind2
@@ -863,7 +857,7 @@ define void @ld_div2_step3_start0_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF4-NEXT:    br i1 [[TMP32]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP16:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[SCALAR_PH]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -926,7 +920,7 @@ define void @ld_div3_step3_start0_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF2-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[SCALAR_PH]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_div3_step3_start0_ind2
@@ -982,7 +976,7 @@ define void @ld_div3_step3_start0_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP32:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF4-NEXT:    br i1 [[TMP32]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP18:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[SCALAR_PH]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -1032,15 +1026,14 @@ define void @ld_div1_step1_start1_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP10:%.*]] = insertelement <2 x i64> [[TMP9]], i64 [[TMP8]], i32 1
 ; VF2-NEXT:    [[TMP11:%.*]] = add nsw <2 x i64> [[TMP10]], splat (i64 42)
 ; VF2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[OFFSET_IDX]]
-; VF2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i64, ptr [[TMP12]], i32 0
-; VF2-NEXT:    store <2 x i64> [[TMP11]], ptr [[TMP13]], align 8
+; VF2-NEXT:    store <2 x i64> [[TMP11]], ptr [[TMP12]], align 8
 ; VF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VF2-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; VF2-NEXT:    [[VEC_IND_NEXT2]] = add <2 x i64> [[VEC_IND1]], splat (i64 2)
-; VF2-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], 998
-; VF2-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
+; VF2-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], 998
+; VF2-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[SCALAR_PH]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_div1_step1_start1_ind2
@@ -1075,15 +1068,14 @@ define void @ld_div1_step1_start1_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP18:%.*]] = insertelement <4 x i64> [[TMP17]], i64 [[TMP14]], i32 3
 ; VF4-NEXT:    [[TMP19:%.*]] = add nsw <4 x i64> [[TMP18]], splat (i64 42)
 ; VF4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[OFFSET_IDX]]
-; VF4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i64, ptr [[TMP20]], i32 0
-; VF4-NEXT:    store <4 x i64> [[TMP19]], ptr [[TMP21]], align 8
+; VF4-NEXT:    store <4 x i64> [[TMP19]], ptr [[TMP20]], align 8
 ; VF4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; VF4-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; VF4-NEXT:    [[VEC_IND_NEXT2]] = add <4 x i64> [[VEC_IND1]], splat (i64 4)
-; VF4-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], 996
-; VF4-NEXT:    br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
+; VF4-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], 996
+; VF4-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP20:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[SCALAR_PH]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -1133,15 +1125,14 @@ define void @ld_div2_step1_start1_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP10:%.*]] = insertelement <2 x i64> [[TMP9]], i64 [[TMP8]], i32 1
 ; VF2-NEXT:    [[TMP11:%.*]] = add nsw <2 x i64> [[TMP10]], splat (i64 42)
 ; VF2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[OFFSET_IDX]]
-; VF2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i64, ptr [[TMP12]], i32 0
-; VF2-NEXT:    store <2 x i64> [[TMP11]], ptr [[TMP13]], align 8
+; VF2-NEXT:    store <2 x i64> [[TMP11]], ptr [[TMP12]], align 8
 ; VF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VF2-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; VF2-NEXT:    [[VEC_IND_NEXT2]] = add <2 x i64> [[VEC_IND1]], splat (i64 2)
-; VF2-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], 998
-; VF2-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; VF2-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], 998
+; VF2-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[SCALAR_PH]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_div2_step1_start1_ind2
@@ -1176,15 +1167,14 @@ define void @ld_div2_step1_start1_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP18:%.*]] = insertelement <4 x i64> [[TMP17]], i64 [[TMP14]], i32 3
 ; VF4-NEXT:    [[TMP19:%.*]] = add nsw <4 x i64> [[TMP18]], splat (i64 42)
 ; VF4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[OFFSET_IDX]]
-; VF4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i64, ptr [[TMP20]], i32 0
-; VF4-NEXT:    store <4 x i64> [[TMP19]], ptr [[TMP21]], align 8
+; VF4-NEXT:    store <4 x i64> [[TMP19]], ptr [[TMP20]], align 8
 ; VF4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; VF4-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; VF4-NEXT:    [[VEC_IND_NEXT2]] = add <4 x i64> [[VEC_IND1]], splat (i64 4)
-; VF4-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], 996
-; VF4-NEXT:    br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
+; VF4-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], 996
+; VF4-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP22:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[SCALAR_PH]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -1234,15 +1224,14 @@ define void @ld_div3_step1_start1_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP10:%.*]] = insertelement <2 x i64> [[TMP9]], i64 [[TMP8]], i32 1
 ; VF2-NEXT:    [[TMP11:%.*]] = add nsw <2 x i64> [[TMP10]], splat (i64 42)
 ; VF2-NEXT:    [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[OFFSET_IDX]]
-; VF2-NEXT:    [[TMP13:%.*]] = getelementptr inbounds i64, ptr [[TMP12]], i32 0
-; VF2-NEXT:    store <2 x i64> [[TMP11]], ptr [[TMP13]], align 8
+; VF2-NEXT:    store <2 x i64> [[TMP11]], ptr [[TMP12]], align 8
 ; VF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 2
 ; VF2-NEXT:    [[VEC_IND_NEXT]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
 ; VF2-NEXT:    [[VEC_IND_NEXT2]] = add <2 x i64> [[VEC_IND1]], splat (i64 2)
-; VF2-NEXT:    [[TMP14:%.*]] = icmp eq i64 [[INDEX_NEXT]], 998
-; VF2-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; VF2-NEXT:    [[TMP13:%.*]] = icmp eq i64 [[INDEX_NEXT]], 998
+; VF2-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[SCALAR_PH]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_div3_step1_start1_ind2
@@ -1277,15 +1266,14 @@ define void @ld_div3_step1_start1_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP18:%.*]] = insertelement <4 x i64> [[TMP17]], i64 [[TMP14]], i32 3
 ; VF4-NEXT:    [[TMP19:%.*]] = add nsw <4 x i64> [[TMP18]], splat (i64 42)
 ; VF4-NEXT:    [[TMP20:%.*]] = getelementptr inbounds i64, ptr [[B]], i64 [[OFFSET_IDX]]
-; VF4-NEXT:    [[TMP21:%.*]] = getelementptr inbounds i64, ptr [[TMP20]], i32 0
-; VF4-NEXT:    store <4 x i64> [[TMP19]], ptr [[TMP21]], align 8
+; VF4-NEXT:    store <4 x i64> [[TMP19]], ptr [[TMP20]], align 8
 ; VF4-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; VF4-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
 ; VF4-NEXT:    [[VEC_IND_NEXT2]] = add <4 x i64> [[VEC_IND1]], splat (i64 4)
-; VF4-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], 996
-; VF4-NEXT:    br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
+; VF4-NEXT:    [[TMP21:%.*]] = icmp eq i64 [[INDEX_NEXT]], 996
+; VF4-NEXT:    br i1 [[TMP21]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP24:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[SCALAR_PH]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -1349,7 +1337,7 @@ define void @ld_div1_step2_start1_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], 498
 ; VF2-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[SCALAR_PH]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_div1_step2_start1_ind2
@@ -1406,7 +1394,7 @@ define void @ld_div1_step2_start1_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP33:%.*]] = icmp eq i64 [[INDEX_NEXT]], 496
 ; VF4-NEXT:    br i1 [[TMP33]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP26:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[SCALAR_PH]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -1470,7 +1458,7 @@ define void @ld_div2_step2_start1_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], 498
 ; VF2-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[SCALAR_PH]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_div2_step2_start1_ind2
@@ -1527,7 +1515,7 @@ define void @ld_div2_step2_start1_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP33:%.*]] = icmp eq i64 [[INDEX_NEXT]], 496
 ; VF4-NEXT:    br i1 [[TMP33]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP28:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[SCALAR_PH]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -1591,7 +1579,7 @@ define void @ld_div3_step2_start1_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], 498
 ; VF2-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[SCALAR_PH]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_div3_step2_start1_ind2
@@ -1648,7 +1636,7 @@ define void @ld_div3_step2_start1_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP33:%.*]] = icmp eq i64 [[INDEX_NEXT]], 496
 ; VF4-NEXT:    br i1 [[TMP33]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP30:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[SCALAR_PH]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -1712,7 +1700,7 @@ define void @ld_div1_step3_start1_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF2-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[SCALAR_PH]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_div1_step3_start1_ind2
@@ -1769,7 +1757,7 @@ define void @ld_div1_step3_start1_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP33:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF4-NEXT:    br i1 [[TMP33]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP32:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[SCALAR_PH]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -1833,7 +1821,7 @@ define void @ld_div2_step3_start1_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF2-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[SCALAR_PH]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_div2_step3_start1_ind2
@@ -1890,7 +1878,7 @@ define void @ld_div2_step3_start1_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP33:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF4-NEXT:    br i1 [[TMP33]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP34:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[SCALAR_PH]]
 ; VF4:       scalar.ph:
 ;
 entry:
@@ -1954,7 +1942,7 @@ define void @ld_div3_step3_start1_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF2-NEXT:    [[TMP19:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF2-NEXT:    br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]
 ; VF2:       middle.block:
-; VF2-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF2-NEXT:    br label [[SCALAR_PH]]
 ; VF2:       scalar.ph:
 ;
 ; VF4-LABEL: define void @ld_div3_step3_start1_ind2
@@ -2011,7 +1999,7 @@ define void @ld_div3_step3_start1_ind2(ptr noalias %A, ptr noalias %B) {
 ; VF4-NEXT:    [[TMP33:%.*]] = icmp eq i64 [[INDEX_NEXT]], 332
 ; VF4-NEXT:    br i1 [[TMP33]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP36:![0-9]+]]
 ; VF4:       middle.block:
-; VF4-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; VF4-NEXT:    br label [[SCALAR_PH]]
 ; VF4:       scalar.ph:
 ;
 entry:
diff --git a/llvm/test/Transforms/LoopVectorize/unused-blend-mask-for-first-operand.ll b/llvm/test/Transforms/LoopVectorize/unused-blend-mask-for-first-operand.ll
index 3b442a9..0541c9d 100644
--- a/llvm/test/Transforms/LoopVectorize/unused-blend-mask-for-first-operand.ll
+++ b/llvm/test/Transforms/LoopVectorize/unused-blend-mask-for-first-operand.ll
@@ -18,15 +18,14 @@ define void @test_not_first_lane_only_constant(ptr %A, ptr noalias %B)  {
 ; CHECK-NEXT:    [[TMP13:%.*]] = load i16, ptr [[B]], align 2
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT5:%.*]] = insertelement <4 x i16> poison, i16 [[TMP13]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT6:%.*]] = shufflevector <4 x i16> [[BROADCAST_SPLATINSERT5]], <4 x i16> poison, <4 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <4 x i16> [[BROADCAST_SPLAT6]], ptr [[TMP2]], align 2
+; CHECK-NEXT:    store <4 x i16> [[BROADCAST_SPLAT6]], ptr [[TMP1]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; CHECK:       loop.header:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
@@ -88,22 +87,21 @@ define void @test_not_first_lane_only_wide_compare(ptr %A, ptr noalias %B, i16 %
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i16
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[A]], i16 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP2]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP1]], align 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 0
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp ult i16 [[TMP3]], [[X]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = select i1 [[TMP4]], ptr poison, ptr [[B]]
 ; CHECK-NEXT:    [[TMP13:%.*]] = load i16, ptr [[TMP12]], align 2
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT5:%.*]] = insertelement <4 x i16> poison, i16 [[TMP13]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT6:%.*]] = shufflevector <4 x i16> [[BROADCAST_SPLATINSERT5]], <4 x i16> poison, <4 x i32> zeroinitializer
-; CHECK-NEXT:    store <4 x i16> [[BROADCAST_SPLAT6]], ptr [[TMP2]], align 2
+; CHECK-NEXT:    store <4 x i16> [[BROADCAST_SPLAT6]], ptr [[TMP1]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; CHECK:       loop.header:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
@@ -171,22 +169,21 @@ define void @test_not_first_lane_only_wide_compare_incoming_order_swapped(ptr %A
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i16
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[A]], i16 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP2]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i16>, ptr [[TMP1]], align 2
 ; CHECK-NEXT:    [[TMP3:%.*]] = extractelement <4 x i16> [[WIDE_LOAD]], i32 0
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp ult i16 [[TMP3]], [[X]]
 ; CHECK-NEXT:    [[PREDPHI:%.*]] = select i1 [[TMP4]], ptr poison, ptr [[B]]
 ; CHECK-NEXT:    [[TMP12:%.*]] = load i16, ptr [[PREDPHI]], align 2
 ; CHECK-NEXT:    [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <4 x i16> poison, i16 [[TMP12]], i64 0
 ; CHECK-NEXT:    [[BROADCAST_SPLAT4:%.*]] = shufflevector <4 x i16> [[BROADCAST_SPLATINSERT3]], <4 x i16> poison, <4 x i32> zeroinitializer
-; CHECK-NEXT:    store <4 x i16> [[BROADCAST_SPLAT4]], ptr [[TMP2]], align 2
+; CHECK-NEXT:    store <4 x i16> [[BROADCAST_SPLAT4]], ptr [[TMP1]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
 ; CHECK-NEXT:    br i1 [[TMP13]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; CHECK:       loop.header:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/use-scalar-epilogue-if-tp-fails.ll b/llvm/test/Transforms/LoopVectorize/use-scalar-epilogue-if-tp-fails.ll
index df8123d..3b34b75 100644
--- a/llvm/test/Transforms/LoopVectorize/use-scalar-epilogue-if-tp-fails.ll
+++ b/llvm/test/Transforms/LoopVectorize/use-scalar-epilogue-if-tp-fails.ll
@@ -28,10 +28,8 @@ define void @basic_loop(ptr nocapture readonly %ptr, i32 %size, ptr %pos) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[PTR]], i32 [[INDEX]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[NEXT_GEP]], i32 1
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[WIDE_LOAD]], ptr [[TMP3]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEXT:    store <4 x i8> [[WIDE_LOAD]], ptr [[NEXT_GEP]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -91,10 +89,8 @@ define void @metadata(ptr nocapture readonly %ptr, i32 %size, ptr %pos) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[PTR]], i32 [[INDEX]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[NEXT_GEP]], i32 1
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP2]], align 1
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
-; CHECK-NEXT:    store <4 x i8> [[WIDE_LOAD]], ptr [[TMP3]], align 1
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x i8>, ptr [[TMP1]], align 1
+; CHECK-NEXT:    store <4 x i8> [[WIDE_LOAD]], ptr [[NEXT_GEP]], align 1
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/vector-intrinsic-call-cost.ll b/llvm/test/Transforms/LoopVectorize/vector-intrinsic-call-cost.ll
index 3d3b6c4..f80fb8c 100644
--- a/llvm/test/Transforms/LoopVectorize/vector-intrinsic-call-cost.ll
+++ b/llvm/test/Transforms/LoopVectorize/vector-intrinsic-call-cost.ll
@@ -4,12 +4,10 @@
 ; CHECK-LABEL: vector.body:
 ; CHECK-NEXT:    [[IDX:%.+]] = phi i32 [ 0, %vector.ph ], [ %index.next, %vector.body ]
 ; CHECK-NEXT:    [[GEP:%.+]] = getelementptr inbounds i16, ptr %src, i32 %index
-; CHECK-NEXT:    [[GEP0:%.+]] = getelementptr inbounds i16, ptr [[GEP]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.+]] = load <4 x i16>, ptr [[GEP0]], align 2
+; CHECK-NEXT:    [[WIDE_LOAD:%.+]] = load <4 x i16>, ptr [[GEP]], align 2
 ; CHECK-NEXT:    [[FSHL:%.+]] = call <4 x i16> @llvm.fshl.v4i16(<4 x i16> [[WIDE_LOAD]], <4 x i16> [[WIDE_LOAD]], <4 x i16> splat (i16 15))
 ; CHECK-NEXT:    [[GEP0:%.+]] = getelementptr inbounds i16, ptr %dst, i32 %index
-; CHECK-NEXT:    [[GEP1:%.+]] = getelementptr inbounds i16, ptr [[GEP0]], i32 0
-; CHECK-NEXT:    store <4 x i16> [[FSHL]], ptr [[GEP1]], align 2
+; CHECK-NEXT:    store <4 x i16> [[FSHL]], ptr [[GEP0]], align 2
 ; CHECK-NEXT:    [[IDX_NEXT:%.+]] = add nuw i32 [[IDX]], 4
 ; CHECK-NEXT:    [[EC:%.+]] = icmp eq i32 [[IDX_NEXT]], %n.vec
 ; CHECK-NEXT:    br i1 [[EC]], label %middle.block, label %vector.body
diff --git a/llvm/test/Transforms/LoopVectorize/vector-loop-backedge-elimination-branch-weights.ll b/llvm/test/Transforms/LoopVectorize/vector-loop-backedge-elimination-branch-weights.ll
index d5acf5c..38dbbbb 100644
--- a/llvm/test/Transforms/LoopVectorize/vector-loop-backedge-elimination-branch-weights.ll
+++ b/llvm/test/Transforms/LoopVectorize/vector-loop-backedge-elimination-branch-weights.ll
@@ -18,10 +18,9 @@ define void @test_tc_between_8_and_17(ptr %A, i64 range(i64 8, 17) %N) {
 ; VF8UF1:       [[VECTOR_BODY]]:
 ; VF8UF1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF8UF1-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[INDEX]]
-; VF8UF1-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
-; VF8UF1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP1]], align 1
+; VF8UF1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[NEXT_GEP]], align 1
 ; VF8UF1-NEXT:    [[TMP2:%.*]] = add nsw <8 x i8> [[WIDE_LOAD]], splat (i8 10)
-; VF8UF1-NEXT:    store <8 x i8> [[TMP2]], ptr [[TMP1]], align 1
+; VF8UF1-NEXT:    store <8 x i8> [[TMP2]], ptr [[NEXT_GEP]], align 1
 ; VF8UF1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; VF8UF1-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; VF8UF1-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !prof [[PROF1:![0-9]+]], !llvm.loop [[LOOP2:![0-9]+]]
@@ -56,20 +55,18 @@ define void @test_tc_between_8_and_17(ptr %A, i64 range(i64 8, 17) %N) {
 ; VF8UF2-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A]], i64 [[N_VEC]]
 ; VF8UF2-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; VF8UF2:       [[VECTOR_BODY]]:
-; VF8UF2-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i32 0
 ; VF8UF2-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[A]], i32 8
-; VF8UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP1]], align 1
+; VF8UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[A]], align 1
 ; VF8UF2-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i8>, ptr [[TMP2]], align 1
 ; VF8UF2-NEXT:    [[TMP3:%.*]] = add nsw <8 x i8> [[WIDE_LOAD]], splat (i8 10)
 ; VF8UF2-NEXT:    [[TMP4:%.*]] = add nsw <8 x i8> [[WIDE_LOAD1]], splat (i8 10)
-; VF8UF2-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[A]], i32 0
 ; VF8UF2-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[A]], i32 8
-; VF8UF2-NEXT:    store <8 x i8> [[TMP3]], ptr [[TMP5]], align 1
+; VF8UF2-NEXT:    store <8 x i8> [[TMP3]], ptr [[A]], align 1
 ; VF8UF2-NEXT:    store <8 x i8> [[TMP4]], ptr [[TMP6]], align 1
 ; VF8UF2-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; VF8UF2:       [[MIDDLE_BLOCK]]:
 ; VF8UF2-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
-; VF8UF2-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; VF8UF2-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]], !prof [[PROF1:![0-9]+]]
 ; VF8UF2:       [[SCALAR_PH]]:
 ; VF8UF2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; VF8UF2-NEXT:    [[BC_RESUME_VAL2:%.*]] = phi ptr [ [[TMP0]], %[[MIDDLE_BLOCK]] ], [ [[A]], %[[ENTRY]] ]
@@ -83,7 +80,7 @@ define void @test_tc_between_8_and_17(ptr %A, i64 range(i64 8, 17) %N) {
 ; VF8UF2-NEXT:    store i8 [[ADD]], ptr [[P_SRC]], align 1
 ; VF8UF2-NEXT:    [[IV_NEXT]] = add nsw i64 [[IV]], 1
 ; VF8UF2-NEXT:    [[CMP:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; VF8UF2-NEXT:    br i1 [[CMP]], label %[[EXIT]], label %[[LOOP]], !prof [[PROF1:![0-9]+]], !llvm.loop [[LOOP2:![0-9]+]]
+; VF8UF2-NEXT:    br i1 [[CMP]], label %[[EXIT]], label %[[LOOP]], !prof [[PROF2:![0-9]+]], !llvm.loop [[LOOP3:![0-9]+]]
 ; VF8UF2:       [[EXIT]]:
 ; VF8UF2-NEXT:    ret void
 ;
@@ -98,15 +95,13 @@ define void @test_tc_between_8_and_17(ptr %A, i64 range(i64 8, 17) %N) {
 ; VF16UF1-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[A]], i64 [[N_VEC]]
 ; VF16UF1-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; VF16UF1:       [[VECTOR_BODY]]:
-; VF16UF1-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i32 0
-; VF16UF1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
+; VF16UF1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[A]], align 1
 ; VF16UF1-NEXT:    [[TMP2:%.*]] = add nsw <16 x i8> [[WIDE_LOAD]], splat (i8 10)
-; VF16UF1-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[A]], i32 0
-; VF16UF1-NEXT:    store <16 x i8> [[TMP2]], ptr [[TMP3]], align 1
+; VF16UF1-NEXT:    store <16 x i8> [[TMP2]], ptr [[A]], align 1
 ; VF16UF1-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; VF16UF1:       [[MIDDLE_BLOCK]]:
 ; VF16UF1-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
-; VF16UF1-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; VF16UF1-NEXT:    br i1 [[CMP_N]], label %[[EXIT:.*]], label %[[SCALAR_PH]], !prof [[PROF1:![0-9]+]]
 ; VF16UF1:       [[SCALAR_PH]]:
 ; VF16UF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[N_VEC]], %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
 ; VF16UF1-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi ptr [ [[TMP0]], %[[MIDDLE_BLOCK]] ], [ [[A]], %[[ENTRY]] ]
@@ -120,7 +115,7 @@ define void @test_tc_between_8_and_17(ptr %A, i64 range(i64 8, 17) %N) {
 ; VF16UF1-NEXT:    store i8 [[ADD]], ptr [[P_SRC]], align 1
 ; VF16UF1-NEXT:    [[IV_NEXT]] = add nsw i64 [[IV]], 1
 ; VF16UF1-NEXT:    [[CMP:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
-; VF16UF1-NEXT:    br i1 [[CMP]], label %[[EXIT]], label %[[LOOP]], !prof [[PROF1:![0-9]+]], !llvm.loop [[LOOP2:![0-9]+]]
+; VF16UF1-NEXT:    br i1 [[CMP]], label %[[EXIT]], label %[[LOOP]], !prof [[PROF2:![0-9]+]], !llvm.loop [[LOOP3:![0-9]+]]
 ; VF16UF1:       [[EXIT]]:
 ; VF16UF1-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/LoopVectorize/vector-loop-backedge-elimination-early-exit.ll b/llvm/test/Transforms/LoopVectorize/vector-loop-backedge-elimination-early-exit.ll
index fdd5e0e..1ad75bb 100644
--- a/llvm/test/Transforms/LoopVectorize/vector-loop-backedge-elimination-early-exit.ll
+++ b/llvm/test/Transforms/LoopVectorize/vector-loop-backedge-elimination-early-exit.ll
@@ -17,8 +17,7 @@ define i8 @test_early_exit_max_tc_less_than_16(ptr dereferenceable(16) %A) nosyn
 ; VF8UF1:       [[VECTOR_BODY]]:
 ; VF8UF1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF8UF1-NEXT:    [[P_SRC:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
-; VF8UF1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[P_SRC]], i32 0
-; VF8UF1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP2]], align 1
+; VF8UF1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[P_SRC]], align 1
 ; VF8UF1-NEXT:    [[TMP3:%.*]] = icmp eq <8 x i8> [[WIDE_LOAD]], zeroinitializer
 ; VF8UF1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; VF8UF1-NEXT:    [[TMP4:%.*]] = call i1 @llvm.vector.reduce.or.v8i1(<8 x i1> [[TMP3]])
@@ -28,11 +27,11 @@ define i8 @test_early_exit_max_tc_less_than_16(ptr dereferenceable(16) %A) nosyn
 ; VF8UF1:       [[MIDDLE_SPLIT]]:
 ; VF8UF1-NEXT:    br i1 [[TMP4]], label %[[VECTOR_EARLY_EXIT:.*]], label %[[MIDDLE_BLOCK:.*]]
 ; VF8UF1:       [[MIDDLE_BLOCK]]:
-; VF8UF1-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; VF8UF1-NEXT:    br label %[[EXIT:.*]]
 ; VF8UF1:       [[VECTOR_EARLY_EXIT]]:
 ; VF8UF1-NEXT:    br label %[[EXIT]]
 ; VF8UF1:       [[SCALAR_PH]]:
-; VF8UF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 16, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; VF8UF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; VF8UF1-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; VF8UF1:       [[LOOP_HEADER]]:
 ; VF8UF1-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -55,35 +54,33 @@ define i8 @test_early_exit_max_tc_less_than_16(ptr dereferenceable(16) %A) nosyn
 ; VF8UF2:       [[VECTOR_PH]]:
 ; VF8UF2-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; VF8UF2:       [[VECTOR_BODY]]:
-; VF8UF2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; VF8UF2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
-; VF8UF2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 0
-; VF8UF2-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i8>, ptr [[TMP1]], align 1
-; VF8UF2-NEXT:    [[TMP6:%.*]] = icmp eq <8 x i8> [[WIDE_LOAD1]], zeroinitializer
-; VF8UF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
-; VF8UF2-NEXT:    [[TMP3:%.*]] = call i1 @llvm.vector.reduce.or.v8i1(<8 x i1> [[TMP6]])
-; VF8UF2-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 16
-; VF8UF2-NEXT:    [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]
-; VF8UF2-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_SPLIT:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; VF8UF2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[A]], i32 8
+; VF8UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[A]], align 1
+; VF8UF2-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i8>, ptr [[TMP0]], align 1
+; VF8UF2-NEXT:    [[TMP1:%.*]] = icmp eq <8 x i8> [[WIDE_LOAD]], zeroinitializer
+; VF8UF2-NEXT:    [[TMP2:%.*]] = icmp eq <8 x i8> [[WIDE_LOAD1]], zeroinitializer
+; VF8UF2-NEXT:    [[TMP3:%.*]] = or <8 x i1> [[TMP1]], [[TMP2]]
+; VF8UF2-NEXT:    [[TMP4:%.*]] = call i1 @llvm.vector.reduce.or.v8i1(<8 x i1> [[TMP3]])
+; VF8UF2-NEXT:    br label %[[MIDDLE_SPLIT:.*]]
 ; VF8UF2:       [[MIDDLE_SPLIT]]:
-; VF8UF2-NEXT:    br i1 [[TMP3]], label %[[VECTOR_EARLY_EXIT:.*]], label %[[MIDDLE_BLOCK:.*]]
+; VF8UF2-NEXT:    br i1 [[TMP4]], label %[[VECTOR_EARLY_EXIT:.*]], label %[[MIDDLE_BLOCK:.*]]
 ; VF8UF2:       [[MIDDLE_BLOCK]]:
-; VF8UF2-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; VF8UF2-NEXT:    br label %[[EXIT:.*]]
 ; VF8UF2:       [[VECTOR_EARLY_EXIT]]:
 ; VF8UF2-NEXT:    br label %[[EXIT]]
 ; VF8UF2:       [[SCALAR_PH]]:
-; VF8UF2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 16, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; VF8UF2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; VF8UF2-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; VF8UF2:       [[LOOP_HEADER]]:
-; VF8UF2-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
-; VF8UF2-NEXT:    [[P_SRC1:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[IV1]]
-; VF8UF2-NEXT:    [[L:%.*]] = load i8, ptr [[P_SRC1]], align 1
+; VF8UF2-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
+; VF8UF2-NEXT:    [[P_SRC:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[IV]]
+; VF8UF2-NEXT:    [[L:%.*]] = load i8, ptr [[P_SRC]], align 1
 ; VF8UF2-NEXT:    [[C:%.*]] = icmp eq i8 [[L]], 0
 ; VF8UF2-NEXT:    br i1 [[C]], label %[[EXIT]], label %[[LOOP_LATCH]]
 ; VF8UF2:       [[LOOP_LATCH]]:
-; VF8UF2-NEXT:    [[IV_NEXT]] = add nsw i64 [[IV1]], 1
+; VF8UF2-NEXT:    [[IV_NEXT]] = add nsw i64 [[IV]], 1
 ; VF8UF2-NEXT:    [[CMP:%.*]] = icmp eq i64 [[IV_NEXT]], 16
-; VF8UF2-NEXT:    br i1 [[CMP]], label %[[EXIT]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP3:![0-9]+]]
+; VF8UF2-NEXT:    br i1 [[CMP]], label %[[EXIT]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP0:![0-9]+]]
 ; VF8UF2:       [[EXIT]]:
 ; VF8UF2-NEXT:    [[RES:%.*]] = phi i8 [ 0, %[[LOOP_HEADER]] ], [ 1, %[[LOOP_LATCH]] ], [ 1, %[[MIDDLE_BLOCK]] ], [ 0, %[[VECTOR_EARLY_EXIT]] ]
 ; VF8UF2-NEXT:    ret i8 [[RES]]
@@ -95,19 +92,18 @@ define i8 @test_early_exit_max_tc_less_than_16(ptr dereferenceable(16) %A) nosyn
 ; VF16UF1:       [[VECTOR_PH]]:
 ; VF16UF1-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; VF16UF1:       [[VECTOR_BODY]]:
-; VF16UF1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[A]], i32 0
-; VF16UF1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; VF16UF1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[A]], align 1
 ; VF16UF1-NEXT:    [[TMP3:%.*]] = icmp eq <16 x i8> [[WIDE_LOAD]], zeroinitializer
 ; VF16UF1-NEXT:    [[TMP4:%.*]] = call i1 @llvm.vector.reduce.or.v16i1(<16 x i1> [[TMP3]])
 ; VF16UF1-NEXT:    br label %[[MIDDLE_SPLIT:.*]]
 ; VF16UF1:       [[MIDDLE_SPLIT]]:
 ; VF16UF1-NEXT:    br i1 [[TMP4]], label %[[VECTOR_EARLY_EXIT:.*]], label %[[MIDDLE_BLOCK:.*]]
 ; VF16UF1:       [[MIDDLE_BLOCK]]:
-; VF16UF1-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; VF16UF1-NEXT:    br label %[[EXIT:.*]]
 ; VF16UF1:       [[VECTOR_EARLY_EXIT]]:
 ; VF16UF1-NEXT:    br label %[[EXIT]]
 ; VF16UF1:       [[SCALAR_PH]]:
-; VF16UF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 16, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; VF16UF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; VF16UF1-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; VF16UF1:       [[LOOP_HEADER]]:
 ; VF16UF1-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -153,8 +149,7 @@ define i64 @test_early_exit_max_tc_less_than_16_with_iv_used_outside(ptr derefer
 ; VF8UF1:       [[VECTOR_BODY]]:
 ; VF8UF1-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF8UF1-NEXT:    [[P_SRC:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
-; VF8UF1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[P_SRC]], i32 0
-; VF8UF1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP2]], align 1
+; VF8UF1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[P_SRC]], align 1
 ; VF8UF1-NEXT:    [[TMP3:%.*]] = icmp eq <8 x i8> [[WIDE_LOAD]], zeroinitializer
 ; VF8UF1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
 ; VF8UF1-NEXT:    [[TMP4:%.*]] = call i1 @llvm.vector.reduce.or.v8i1(<8 x i1> [[TMP3]])
@@ -164,13 +159,13 @@ define i64 @test_early_exit_max_tc_less_than_16_with_iv_used_outside(ptr derefer
 ; VF8UF1:       [[MIDDLE_SPLIT]]:
 ; VF8UF1-NEXT:    br i1 [[TMP4]], label %[[VECTOR_EARLY_EXIT:.*]], label %[[MIDDLE_BLOCK:.*]]
 ; VF8UF1:       [[MIDDLE_BLOCK]]:
-; VF8UF1-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; VF8UF1-NEXT:    br label %[[EXIT:.*]]
 ; VF8UF1:       [[VECTOR_EARLY_EXIT]]:
 ; VF8UF1-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v8i1(<8 x i1> [[TMP3]], i1 true)
 ; VF8UF1-NEXT:    [[TMP8:%.*]] = add i64 [[INDEX]], [[FIRST_ACTIVE_LANE]]
 ; VF8UF1-NEXT:    br label %[[EXIT]]
 ; VF8UF1:       [[SCALAR_PH]]:
-; VF8UF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 16, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; VF8UF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; VF8UF1-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; VF8UF1:       [[LOOP_HEADER]]:
 ; VF8UF1-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -193,26 +188,29 @@ define i64 @test_early_exit_max_tc_less_than_16_with_iv_used_outside(ptr derefer
 ; VF8UF2:       [[VECTOR_PH]]:
 ; VF8UF2-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; VF8UF2:       [[VECTOR_BODY]]:
-; VF8UF2-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
-; VF8UF2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[A]], i64 [[INDEX]]
-; VF8UF2-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[TMP0]], i32 0
-; VF8UF2-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i8>, ptr [[TMP1]], align 1
-; VF8UF2-NEXT:    [[TMP6:%.*]] = icmp eq <8 x i8> [[WIDE_LOAD1]], zeroinitializer
-; VF8UF2-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
-; VF8UF2-NEXT:    [[TMP3:%.*]] = call i1 @llvm.vector.reduce.or.v8i1(<8 x i1> [[TMP6]])
-; VF8UF2-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 16
-; VF8UF2-NEXT:    [[TMP5:%.*]] = or i1 [[TMP3]], [[TMP4]]
-; VF8UF2-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_SPLIT:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; VF8UF2-NEXT:    [[TMP0:%.*]] = getelementptr inbounds i8, ptr [[A]], i32 8
+; VF8UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[A]], align 1
+; VF8UF2-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i8>, ptr [[TMP0]], align 1
+; VF8UF2-NEXT:    [[TMP1:%.*]] = icmp eq <8 x i8> [[WIDE_LOAD]], zeroinitializer
+; VF8UF2-NEXT:    [[TMP2:%.*]] = icmp eq <8 x i8> [[WIDE_LOAD1]], zeroinitializer
+; VF8UF2-NEXT:    [[TMP3:%.*]] = or <8 x i1> [[TMP1]], [[TMP2]]
+; VF8UF2-NEXT:    [[TMP4:%.*]] = call i1 @llvm.vector.reduce.or.v8i1(<8 x i1> [[TMP3]])
+; VF8UF2-NEXT:    br label %[[MIDDLE_SPLIT:.*]]
 ; VF8UF2:       [[MIDDLE_SPLIT]]:
-; VF8UF2-NEXT:    br i1 [[TMP3]], label %[[VECTOR_EARLY_EXIT:.*]], label %[[MIDDLE_BLOCK:.*]]
+; VF8UF2-NEXT:    br i1 [[TMP4]], label %[[VECTOR_EARLY_EXIT:.*]], label %[[MIDDLE_BLOCK:.*]]
 ; VF8UF2:       [[MIDDLE_BLOCK]]:
-; VF8UF2-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; VF8UF2-NEXT:    br label %[[EXIT:.*]]
 ; VF8UF2:       [[VECTOR_EARLY_EXIT]]:
-; VF8UF2-NEXT:    [[TMP8:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v8i1(<8 x i1> [[TMP6]], i1 true)
-; VF8UF2-NEXT:    [[TMP7:%.*]] = add i64 [[INDEX]], [[TMP8]]
+; VF8UF2-NEXT:    [[TMP5:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v8i1(<8 x i1> [[TMP2]], i1 true)
+; VF8UF2-NEXT:    [[TMP7:%.*]] = add i64 8, [[TMP5]]
+; VF8UF2-NEXT:    [[TMP8:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v8i1(<8 x i1> [[TMP1]], i1 true)
+; VF8UF2-NEXT:    [[TMP9:%.*]] = add i64 0, [[TMP8]]
+; VF8UF2-NEXT:    [[TMP10:%.*]] = icmp ne i64 [[TMP8]], 8
+; VF8UF2-NEXT:    [[TMP11:%.*]] = select i1 [[TMP10]], i64 [[TMP9]], i64 [[TMP7]]
+; VF8UF2-NEXT:    [[TMP12:%.*]] = add i64 0, [[TMP11]]
 ; VF8UF2-NEXT:    br label %[[EXIT]]
 ; VF8UF2:       [[SCALAR_PH]]:
-; VF8UF2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 16, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; VF8UF2-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; VF8UF2-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; VF8UF2:       [[LOOP_HEADER]]:
 ; VF8UF2-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
@@ -223,9 +221,9 @@ define i64 @test_early_exit_max_tc_less_than_16_with_iv_used_outside(ptr derefer
 ; VF8UF2:       [[LOOP_LATCH]]:
 ; VF8UF2-NEXT:    [[IV_NEXT]] = add nsw i64 [[IV1]], 1
 ; VF8UF2-NEXT:    [[CMP:%.*]] = icmp eq i64 [[IV_NEXT]], 16
-; VF8UF2-NEXT:    br i1 [[CMP]], label %[[EXIT]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP5:![0-9]+]]
+; VF8UF2-NEXT:    br i1 [[CMP]], label %[[EXIT]], label %[[LOOP_HEADER]], !llvm.loop [[LOOP3:![0-9]+]]
 ; VF8UF2:       [[EXIT]]:
-; VF8UF2-NEXT:    [[RES:%.*]] = phi i64 [ [[IV1]], %[[LOOP_HEADER]] ], [ 1, %[[LOOP_LATCH]] ], [ 1, %[[MIDDLE_BLOCK]] ], [ [[TMP7]], %[[VECTOR_EARLY_EXIT]] ]
+; VF8UF2-NEXT:    [[RES:%.*]] = phi i64 [ [[IV1]], %[[LOOP_HEADER]] ], [ 1, %[[LOOP_LATCH]] ], [ 1, %[[MIDDLE_BLOCK]] ], [ [[TMP12]], %[[VECTOR_EARLY_EXIT]] ]
 ; VF8UF2-NEXT:    ret i64 [[RES]]
 ;
 ; VF16UF1-LABEL: define i64 @test_early_exit_max_tc_less_than_16_with_iv_used_outside(
@@ -235,21 +233,20 @@ define i64 @test_early_exit_max_tc_less_than_16_with_iv_used_outside(ptr derefer
 ; VF16UF1:       [[VECTOR_PH]]:
 ; VF16UF1-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; VF16UF1:       [[VECTOR_BODY]]:
-; VF16UF1-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[A]], i32 0
-; VF16UF1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; VF16UF1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[A]], align 1
 ; VF16UF1-NEXT:    [[TMP3:%.*]] = icmp eq <16 x i8> [[WIDE_LOAD]], zeroinitializer
 ; VF16UF1-NEXT:    [[TMP4:%.*]] = call i1 @llvm.vector.reduce.or.v16i1(<16 x i1> [[TMP3]])
 ; VF16UF1-NEXT:    br label %[[MIDDLE_SPLIT:.*]]
 ; VF16UF1:       [[MIDDLE_SPLIT]]:
 ; VF16UF1-NEXT:    br i1 [[TMP4]], label %[[VECTOR_EARLY_EXIT:.*]], label %[[MIDDLE_BLOCK:.*]]
 ; VF16UF1:       [[MIDDLE_BLOCK]]:
-; VF16UF1-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; VF16UF1-NEXT:    br label %[[EXIT:.*]]
 ; VF16UF1:       [[VECTOR_EARLY_EXIT]]:
 ; VF16UF1-NEXT:    [[FIRST_ACTIVE_LANE:%.*]] = call i64 @llvm.experimental.cttz.elts.i64.v16i1(<16 x i1> [[TMP3]], i1 true)
 ; VF16UF1-NEXT:    [[TMP5:%.*]] = add i64 0, [[FIRST_ACTIVE_LANE]]
 ; VF16UF1-NEXT:    br label %[[EXIT]]
 ; VF16UF1:       [[SCALAR_PH]]:
-; VF16UF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 16, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; VF16UF1-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; VF16UF1-NEXT:    br label %[[LOOP_HEADER:.*]]
 ; VF16UF1:       [[LOOP_HEADER]]:
 ; VF16UF1-NEXT:    [[IV1:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP_LATCH:.*]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/vector-loop-backedge-elimination-outside-iv-users.ll b/llvm/test/Transforms/LoopVectorize/vector-loop-backedge-elimination-outside-iv-users.ll
index 78f5cc7..5f1cee8 100644
--- a/llvm/test/Transforms/LoopVectorize/vector-loop-backedge-elimination-outside-iv-users.ll
+++ b/llvm/test/Transforms/LoopVectorize/vector-loop-backedge-elimination-outside-iv-users.ll
@@ -11,15 +11,14 @@ define i64 @remove_loop_region_int_iv_used_outside(ptr %dst) {
 ; CHECK:       [[VECTOR_PH]]:
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr ptr, ptr [[DST]], i32 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr ptr, ptr [[DST]], i32 8
-; CHECK-NEXT:    store <8 x ptr> zeroinitializer, ptr [[TMP1]], align 8
+; CHECK-NEXT:    store <8 x ptr> zeroinitializer, ptr [[DST]], align 8
 ; CHECK-NEXT:    store <8 x ptr> zeroinitializer, ptr [[TMP2]], align 8
 ; CHECK-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 16, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -56,15 +55,14 @@ define i64 @remove_loop_region_int_iv_inc_used_outside(ptr %dst) {
 ; CHECK:       [[VECTOR_PH]]:
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr ptr, ptr [[DST]], i32 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr ptr, ptr [[DST]], i32 8
-; CHECK-NEXT:    store <8 x ptr> zeroinitializer, ptr [[TMP1]], align 8
+; CHECK-NEXT:    store <8 x ptr> zeroinitializer, ptr [[DST]], align 8
 ; CHECK-NEXT:    store <8 x ptr> zeroinitializer, ptr [[TMP2]], align 8
 ; CHECK-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 16, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
@@ -102,17 +100,16 @@ define ptr @remove_loop_region_ptr_iv_used_outside(ptr %dst) {
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[DST]], i64 128
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr ptr, ptr [[DST]], i32 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr ptr, ptr [[DST]], i32 8
-; CHECK-NEXT:    store <8 x ptr> zeroinitializer, ptr [[TMP1]], align 8
+; CHECK-NEXT:    store <8 x ptr> zeroinitializer, ptr [[DST]], align 8
 ; CHECK-NEXT:    store <8 x ptr> zeroinitializer, ptr [[TMP2]], align 8
 ; CHECK-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
 ; CHECK-NEXT:    [[IND_ESCAPE:%.*]] = getelementptr i8, ptr [[TMP0]], i64 -8
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi ptr [ [[TMP0]], %[[MIDDLE_BLOCK]] ], [ [[DST]], %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i64 [ 16, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi ptr [ [[DST]], %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[PTR_IV:%.*]] = phi ptr [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[PTR_IV_NEXT:%.*]], %[[LOOP]] ]
@@ -152,16 +149,15 @@ define ptr @remove_loop_region_ptr_iv_inc_used_outside(ptr %dst) {
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr i8, ptr [[DST]], i64 128
 ; CHECK-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; CHECK:       [[VECTOR_BODY]]:
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr ptr, ptr [[DST]], i32 0
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr ptr, ptr [[DST]], i32 8
-; CHECK-NEXT:    store <8 x ptr> zeroinitializer, ptr [[TMP1]], align 8
+; CHECK-NEXT:    store <8 x ptr> zeroinitializer, ptr [[DST]], align 8
 ; CHECK-NEXT:    store <8 x ptr> zeroinitializer, ptr [[TMP2]], align 8
 ; CHECK-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi ptr [ [[TMP0]], %[[MIDDLE_BLOCK]] ], [ [[DST]], %[[ENTRY]] ]
-; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i64 [ 16, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi ptr [ [[DST]], %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i64 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[PTR_IV:%.*]] = phi ptr [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[PTR_IV_NEXT:%.*]], %[[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/vector-loop-backedge-elimination.ll b/llvm/test/Transforms/LoopVectorize/vector-loop-backedge-elimination.ll
index 001bf0c..b396e29 100644
--- a/llvm/test/Transforms/LoopVectorize/vector-loop-backedge-elimination.ll
+++ b/llvm/test/Transforms/LoopVectorize/vector-loop-backedge-elimination.ll
@@ -23,10 +23,9 @@ define void @test_tc_less_than_16(ptr %A, i64 %N) {
 ; VF8UF1:       [[VECTOR_BODY]]:
 ; VF8UF1-NEXT:    [[TMP2:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF8UF1-NEXT:    [[NEXT_GEP:%.*]] = getelementptr i8, ptr [[A]], i64 [[TMP2]]
-; VF8UF1-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[NEXT_GEP]], i32 0
-; VF8UF1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP3]], align 1
+; VF8UF1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[NEXT_GEP]], align 1
 ; VF8UF1-NEXT:    [[TMP4:%.*]] = add nsw <8 x i8> [[WIDE_LOAD]], splat (i8 10)
-; VF8UF1-NEXT:    store <8 x i8> [[TMP4]], ptr [[TMP3]], align 1
+; VF8UF1-NEXT:    store <8 x i8> [[TMP4]], ptr [[NEXT_GEP]], align 1
 ; VF8UF1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP2]], 8
 ; VF8UF1-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; VF8UF1-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -62,15 +61,13 @@ define void @test_tc_less_than_16(ptr %A, i64 %N) {
 ; VF8UF2-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[N_VEC]]
 ; VF8UF2-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; VF8UF2:       [[VECTOR_BODY]]:
-; VF8UF2-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[A]], i32 0
 ; VF8UF2-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[A]], i32 8
-; VF8UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP2]], align 1
+; VF8UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[A]], align 1
 ; VF8UF2-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i8>, ptr [[TMP3]], align 1
 ; VF8UF2-NEXT:    [[TMP4:%.*]] = add nsw <8 x i8> [[WIDE_LOAD]], splat (i8 10)
 ; VF8UF2-NEXT:    [[TMP5:%.*]] = add nsw <8 x i8> [[WIDE_LOAD1]], splat (i8 10)
-; VF8UF2-NEXT:    [[TMP6:%.*]] = getelementptr i8, ptr [[A]], i32 0
 ; VF8UF2-NEXT:    [[TMP7:%.*]] = getelementptr i8, ptr [[A]], i32 8
-; VF8UF2-NEXT:    store <8 x i8> [[TMP4]], ptr [[TMP6]], align 1
+; VF8UF2-NEXT:    store <8 x i8> [[TMP4]], ptr [[A]], align 1
 ; VF8UF2-NEXT:    store <8 x i8> [[TMP5]], ptr [[TMP7]], align 1
 ; VF8UF2-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; VF8UF2:       [[MIDDLE_BLOCK]]:
@@ -105,11 +102,9 @@ define void @test_tc_less_than_16(ptr %A, i64 %N) {
 ; VF16UF1-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[A]], i64 [[N_VEC]]
 ; VF16UF1-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; VF16UF1:       [[VECTOR_BODY]]:
-; VF16UF1-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[A]], i32 0
-; VF16UF1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP2]], align 1
+; VF16UF1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[A]], align 1
 ; VF16UF1-NEXT:    [[TMP3:%.*]] = add nsw <16 x i8> [[WIDE_LOAD]], splat (i8 10)
-; VF16UF1-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[A]], i32 0
-; VF16UF1-NEXT:    store <16 x i8> [[TMP3]], ptr [[TMP4]], align 1
+; VF16UF1-NEXT:    store <16 x i8> [[TMP3]], ptr [[A]], align 1
 ; VF16UF1-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; VF16UF1:       [[MIDDLE_BLOCK]]:
 ; VF16UF1-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[AND]], [[N_VEC]]
@@ -557,11 +552,9 @@ define void @remove_loop_region_outer_loop(i64 range(i64 8, 17) %N, ptr noalias
 ; VF8UF1:       [[VECTOR_BODY]]:
 ; VF8UF1-NEXT:    [[TMP0:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; VF8UF1-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[OUTER_IV]], i64 [[TMP0]]
-; VF8UF1-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP1]], i32 0
-; VF8UF1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP2]], align 1
+; VF8UF1-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP1]], align 1
 ; VF8UF1-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[DST]], i64 [[TMP0]]
-; VF8UF1-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[TMP3]], i32 0
-; VF8UF1-NEXT:    store <8 x i8> [[WIDE_LOAD]], ptr [[TMP4]], align 1
+; VF8UF1-NEXT:    store <8 x i8> [[WIDE_LOAD]], ptr [[TMP3]], align 1
 ; VF8UF1-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[TMP0]], 8
 ; VF8UF1-NEXT:    [[TMP5:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; VF8UF1-NEXT:    br i1 [[TMP5]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP5:![0-9]+]]
@@ -600,13 +593,11 @@ define void @remove_loop_region_outer_loop(i64 range(i64 8, 17) %N, ptr noalias
 ; VF8UF2-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; VF8UF2-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; VF8UF2:       [[VECTOR_BODY]]:
-; VF8UF2-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
 ; VF8UF2-NEXT:    [[TMP2:%.*]] = getelementptr i8, ptr [[TMP0]], i32 8
-; VF8UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP1]], align 1
+; VF8UF2-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i8>, ptr [[TMP0]], align 1
 ; VF8UF2-NEXT:    [[WIDE_LOAD1:%.*]] = load <8 x i8>, ptr [[TMP2]], align 1
-; VF8UF2-NEXT:    [[TMP4:%.*]] = getelementptr i8, ptr [[DST]], i32 0
 ; VF8UF2-NEXT:    [[TMP5:%.*]] = getelementptr i8, ptr [[DST]], i32 8
-; VF8UF2-NEXT:    store <8 x i8> [[WIDE_LOAD]], ptr [[TMP4]], align 1
+; VF8UF2-NEXT:    store <8 x i8> [[WIDE_LOAD]], ptr [[DST]], align 1
 ; VF8UF2-NEXT:    store <8 x i8> [[WIDE_LOAD1]], ptr [[TMP5]], align 1
 ; VF8UF2-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; VF8UF2:       [[MIDDLE_BLOCK]]:
@@ -636,7 +627,7 @@ define void @remove_loop_region_outer_loop(i64 range(i64 8, 17) %N, ptr noalias
 ; VF16UF1-NEXT:  [[ENTRY:.*]]:
 ; VF16UF1-NEXT:    br label %[[OUTER_HEADER:.*]]
 ; VF16UF1:       [[OUTER_HEADER]]:
-; VF16UF1-NEXT:    [[TMP0:%.*]] = phi ptr [ [[SRC]], %[[ENTRY]] ], [ [[OUTER_IV_NEXT:%.*]], %[[OUTER_LATCH:.*]] ]
+; VF16UF1-NEXT:    [[TMP1:%.*]] = phi ptr [ [[SRC]], %[[ENTRY]] ], [ [[OUTER_IV_NEXT:%.*]], %[[OUTER_LATCH:.*]] ]
 ; VF16UF1-NEXT:    [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 [[N]], 16
 ; VF16UF1-NEXT:    br i1 [[MIN_ITERS_CHECK]], label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
 ; VF16UF1:       [[VECTOR_PH]]:
@@ -644,10 +635,8 @@ define void @remove_loop_region_outer_loop(i64 range(i64 8, 17) %N, ptr noalias
 ; VF16UF1-NEXT:    [[N_VEC:%.*]] = sub i64 [[N]], [[N_MOD_VF]]
 ; VF16UF1-NEXT:    br label %[[VECTOR_BODY:.*]]
 ; VF16UF1:       [[VECTOR_BODY]]:
-; VF16UF1-NEXT:    [[TMP1:%.*]] = getelementptr i8, ptr [[TMP0]], i32 0
 ; VF16UF1-NEXT:    [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP1]], align 1
-; VF16UF1-NEXT:    [[TMP3:%.*]] = getelementptr i8, ptr [[DST]], i32 0
-; VF16UF1-NEXT:    store <16 x i8> [[WIDE_LOAD]], ptr [[TMP3]], align 1
+; VF16UF1-NEXT:    store <16 x i8> [[WIDE_LOAD]], ptr [[DST]], align 1
 ; VF16UF1-NEXT:    br label %[[MIDDLE_BLOCK:.*]]
 ; VF16UF1:       [[MIDDLE_BLOCK]]:
 ; VF16UF1-NEXT:    [[CMP_N:%.*]] = icmp eq i64 [[N]], [[N_VEC]]
@@ -657,7 +646,7 @@ define void @remove_loop_region_outer_loop(i64 range(i64 8, 17) %N, ptr noalias
 ; VF16UF1-NEXT:    br label %[[INNER:.*]]
 ; VF16UF1:       [[INNER]]:
 ; VF16UF1-NEXT:    [[INNER_IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[INNER]] ]
-; VF16UF1-NEXT:    [[GEP_SRC:%.*]] = getelementptr i8, ptr [[TMP0]], i64 [[INNER_IV]]
+; VF16UF1-NEXT:    [[GEP_SRC:%.*]] = getelementptr i8, ptr [[TMP1]], i64 [[INNER_IV]]
 ; VF16UF1-NEXT:    [[L:%.*]] = load i8, ptr [[GEP_SRC]], align 1
 ; VF16UF1-NEXT:    [[GEP_DST:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INNER_IV]]
 ; VF16UF1-NEXT:    store i8 [[L]], ptr [[GEP_DST]], align 1
@@ -665,7 +654,7 @@ define void @remove_loop_region_outer_loop(i64 range(i64 8, 17) %N, ptr noalias
 ; VF16UF1-NEXT:    [[C_1:%.*]] = icmp eq i64 [[IV_NEXT]], [[N]]
 ; VF16UF1-NEXT:    br i1 [[C_1]], label %[[OUTER_LATCH]], label %[[INNER]], !llvm.loop [[LOOP4:![0-9]+]]
 ; VF16UF1:       [[OUTER_LATCH]]:
-; VF16UF1-NEXT:    [[OUTER_IV_NEXT]] = getelementptr i8, ptr [[TMP0]], i64 1
+; VF16UF1-NEXT:    [[OUTER_IV_NEXT]] = getelementptr i8, ptr [[TMP1]], i64 1
 ; VF16UF1-NEXT:    [[C_2:%.*]] = call i1 @cond()
 ; VF16UF1-NEXT:    br i1 [[C_2]], label %[[OUTER_HEADER]], label %[[EXIT:.*]]
 ; VF16UF1:       [[EXIT]]:
diff --git a/llvm/test/Transforms/LoopVectorize/vectorize-force-tail-with-evl.ll b/llvm/test/Transforms/LoopVectorize/vectorize-force-tail-with-evl.ll
index 17979e5f..d3f7794 100644
--- a/llvm/test/Transforms/LoopVectorize/vectorize-force-tail-with-evl.ll
+++ b/llvm/test/Transforms/LoopVectorize/vectorize-force-tail-with-evl.ll
@@ -51,15 +51,12 @@ define void @foo(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %N) {
 ; NO-VP:       vector.body:
 ; NO-VP-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP5]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 4 x i32>, ptr [[TMP4]], align 4
 ; NO-VP-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; NO-VP-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i32>, ptr [[TMP7]], align 4
+; NO-VP-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 4 x i32>, ptr [[TMP6]], align 4
 ; NO-VP-NEXT:    [[TMP16:%.*]] = add nsw <vscale x 4 x i32> [[WIDE_LOAD1]], [[WIDE_LOAD]]
 ; NO-VP-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i32 0
-; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP16]], ptr [[TMP10]], align 4
+; NO-VP-NEXT:    store <vscale x 4 x i32> [[TMP16]], ptr [[TMP9]], align 4
 ; NO-VP-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP15]]
 ; NO-VP-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; NO-VP-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
@@ -98,15 +95,12 @@ define void @foo(ptr noalias %a, ptr noalias %b, ptr noalias %c, i64 %N) {
 ; NO-VP-DEF:       vector.body:
 ; NO-VP-DEF-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; NO-VP-DEF-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
-; NO-VP-DEF-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i32, ptr [[TMP4]], i32 0
-; NO-VP-DEF-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 1 x i32>, ptr [[TMP5]], align 4
+; NO-VP-DEF-NEXT:    [[WIDE_LOAD:%.*]] = load <vscale x 1 x i32>, ptr [[TMP4]], align 4
 ; NO-VP-DEF-NEXT:    [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[INDEX]]
-; NO-VP-DEF-NEXT:    [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
-; NO-VP-DEF-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 1 x i32>, ptr [[TMP7]], align 4
+; NO-VP-DEF-NEXT:    [[WIDE_LOAD1:%.*]] = load <vscale x 1 x i32>, ptr [[TMP6]], align 4
 ; NO-VP-DEF-NEXT:    [[TMP8:%.*]] = add nsw <vscale x 1 x i32> [[WIDE_LOAD1]], [[WIDE_LOAD]]
 ; NO-VP-DEF-NEXT:    [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
-; NO-VP-DEF-NEXT:    [[TMP10:%.*]] = getelementptr inbounds i32, ptr [[TMP9]], i32 0
-; NO-VP-DEF-NEXT:    store <vscale x 1 x i32> [[TMP8]], ptr [[TMP10]], align 4
+; NO-VP-DEF-NEXT:    store <vscale x 1 x i32> [[TMP8]], ptr [[TMP9]], align 4
 ; NO-VP-DEF-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP2]]
 ; NO-VP-DEF-NEXT:    [[TMP11:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; NO-VP-DEF-NEXT:    br i1 [[TMP11]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/version-stride-with-integer-casts.ll b/llvm/test/Transforms/LoopVectorize/version-stride-with-integer-casts.ll
index 8d854dd..2873947 100644
--- a/llvm/test/Transforms/LoopVectorize/version-stride-with-integer-casts.ll
+++ b/llvm/test/Transforms/LoopVectorize/version-stride-with-integer-casts.ll
@@ -10,7 +10,7 @@ define void @test_versioned_with_sext_use(i32 %offset, ptr %dst) {
 ; CHECK-NEXT:    [[OFFSET_EXT:%.*]] = sext i32 [[OFFSET]] to i64
 ; CHECK-NEXT:    br label [[OUTER_HEADER:%.*]]
 ; CHECK:       outer.header.loopexit:
-; CHECK-NEXT:    [[IV_2_NEXT_LCSSA:%.*]] = phi i64 [ [[IV_2_NEXT:%.*]], [[INNER_LOOP:%.*]] ], [ [[IND_END:%.*]], [[MIDDLE_BLOCK:%.*]] ]
+; CHECK-NEXT:    [[IV_2_NEXT_LCSSA:%.*]] = phi i64 [ [[IV_2_NEXT:%.*]], [[INNER_LOOP:%.*]] ]
 ; CHECK-NEXT:    br label [[OUTER_HEADER]]
 ; CHECK:       outer.header:
 ; CHECK-NEXT:    [[IV_1:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_2_NEXT_LCSSA]], [[OUTER_HEADER_LOOPEXIT:%.*]] ]
@@ -23,20 +23,19 @@ define void @test_versioned_with_sext_use(i32 %offset, ptr %dst) {
 ; CHECK-NEXT:    br i1 [[IDENT_CHECK]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    [[TMP0:%.*]] = mul i64 200, [[OFFSET_EXT]]
-; CHECK-NEXT:    [[IND_END]] = add i64 [[IV_1]], [[TMP0]]
+; CHECK-NEXT:    [[IND_END:%.*]] = add i64 [[IV_1]], [[TMP0]]
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = mul i64 [[INDEX]], [[OFFSET_EXT]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = add i64 [[IV_1]], [[TMP1]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[DST]], i64 [[TMP3]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP5]], align 8
+; CHECK-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP4]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 200
-; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[OUTER_HEADER_LOOPEXIT]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[IV_1]], [[INNER_LOOP_PREHEADER]] ], [ [[IV_1]], [[VECTOR_SCEVCHECK]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 200, [[MIDDLE_BLOCK]] ], [ 0, [[INNER_LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
@@ -83,7 +82,7 @@ define void @test_versioned_with_zext_use(i32 %offset, ptr %dst) {
 ; CHECK-NEXT:    [[OFFSET_EXT:%.*]] = zext i32 [[OFFSET]] to i64
 ; CHECK-NEXT:    br label [[OUTER_HEADER:%.*]]
 ; CHECK:       outer.header.loopexit:
-; CHECK-NEXT:    [[IV_2_NEXT_LCSSA:%.*]] = phi i64 [ [[IV_2_NEXT:%.*]], [[INNER_LOOP:%.*]] ], [ [[IND_END:%.*]], [[MIDDLE_BLOCK:%.*]] ]
+; CHECK-NEXT:    [[IV_2_NEXT_LCSSA:%.*]] = phi i64 [ [[IV_2_NEXT:%.*]], [[INNER_LOOP:%.*]] ]
 ; CHECK-NEXT:    br label [[OUTER_HEADER]]
 ; CHECK:       outer.header:
 ; CHECK-NEXT:    [[IV_1:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_2_NEXT_LCSSA]], [[OUTER_HEADER_LOOPEXIT:%.*]] ]
@@ -96,20 +95,19 @@ define void @test_versioned_with_zext_use(i32 %offset, ptr %dst) {
 ; CHECK-NEXT:    br i1 [[IDENT_CHECK]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    [[TMP0:%.*]] = mul i64 200, [[OFFSET_EXT]]
-; CHECK-NEXT:    [[IND_END]] = add i64 [[IV_1]], [[TMP0]]
+; CHECK-NEXT:    [[IND_END:%.*]] = add i64 [[IV_1]], [[TMP0]]
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP1:%.*]] = mul i64 [[INDEX]], [[OFFSET_EXT]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = add i64 [[IV_1]], [[TMP1]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr i32, ptr [[DST]], i64 [[TMP3]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr i32, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP5]], align 8
+; CHECK-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP4]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP7:%.*]] = icmp eq i64 [[INDEX_NEXT]], 200
-; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK-NEXT:    br i1 [[TMP7]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[OUTER_HEADER_LOOPEXIT]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[IV_1]], [[INNER_LOOP_PREHEADER]] ], [ [[IV_1]], [[VECTOR_SCEVCHECK]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 200, [[MIDDLE_BLOCK]] ], [ 0, [[INNER_LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
@@ -179,9 +177,9 @@ define void @versioned_sext_use_in_gep(i32 %scale, ptr %dst, i64 %scale.2) {
 ; CHECK-NEXT:    [[TMP18:%.*]] = icmp eq i64 [[INDEX_NEXT]], 256
 ; CHECK-NEXT:    br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 256, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -224,7 +222,7 @@ define void @test_versioned_with_different_uses(i32 %offset, ptr noalias %dst.1,
 ; CHECK-NEXT:    [[OFFSET_EXT:%.*]] = zext i32 [[OFFSET]] to i64
 ; CHECK-NEXT:    br label [[OUTER_HEADER:%.*]]
 ; CHECK:       outer.header.loopexit:
-; CHECK-NEXT:    [[IV_2_NEXT_LCSSA:%.*]] = phi i64 [ [[IV_2_NEXT:%.*]], [[INNER_LOOP:%.*]] ], [ [[IND_END:%.*]], [[MIDDLE_BLOCK:%.*]] ]
+; CHECK-NEXT:    [[IV_2_NEXT_LCSSA:%.*]] = phi i64 [ [[IV_2_NEXT:%.*]], [[INNER_LOOP:%.*]] ]
 ; CHECK-NEXT:    br label [[OUTER_HEADER]]
 ; CHECK:       outer.header:
 ; CHECK-NEXT:    [[IV_1:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[IV_2_NEXT_LCSSA]], [[OUTER_HEADER_LOOPEXIT:%.*]] ]
@@ -237,7 +235,7 @@ define void @test_versioned_with_different_uses(i32 %offset, ptr noalias %dst.1,
 ; CHECK-NEXT:    br i1 [[IDENT_CHECK]], label [[SCALAR_PH]], label [[VECTOR_PH:%.*]]
 ; CHECK:       vector.ph:
 ; CHECK-NEXT:    [[TMP0:%.*]] = mul i64 200, [[OFFSET_EXT]]
-; CHECK-NEXT:    [[IND_END]] = add i64 [[IV_1]], [[TMP0]]
+; CHECK-NEXT:    [[IND_END:%.*]] = add i64 [[IV_1]], [[TMP0]]
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
 ; CHECK:       vector.body:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
@@ -257,13 +255,12 @@ define void @test_versioned_with_different_uses(i32 %offset, ptr noalias %dst.1,
 ; CHECK-NEXT:    store i32 0, ptr [[TMP10]], align 8
 ; CHECK-NEXT:    store i32 0, ptr [[TMP11]], align 8
 ; CHECK-NEXT:    [[TMP12:%.*]] = getelementptr i32, ptr [[DST_2]], i64 [[TMP3]]
-; CHECK-NEXT:    [[TMP13:%.*]] = getelementptr i32, ptr [[TMP12]], i32 0
-; CHECK-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP13]], align 8
+; CHECK-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP12]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], 200
-; CHECK-NEXT:    br i1 [[TMP15]], label [[MIDDLE_BLOCK]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
+; CHECK-NEXT:    br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP8:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[OUTER_HEADER_LOOPEXIT]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ [[IND_END]], [[MIDDLE_BLOCK]] ], [ [[IV_1]], [[INNER_LOOP_PREHEADER]] ], [ [[IV_1]], [[VECTOR_SCEVCHECK]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 200, [[MIDDLE_BLOCK]] ], [ 0, [[INNER_LOOP_PREHEADER]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
@@ -352,14 +349,13 @@ define void @test_versioned_with_non_ex_use(i32 %offset, ptr noalias %dst.1, ptr
 ; CHECK-NEXT:    store i32 0, ptr [[TMP16]], align 8
 ; CHECK-NEXT:    store i32 0, ptr [[TMP18]], align 8
 ; CHECK-NEXT:    [[TMP20:%.*]] = getelementptr i32, ptr [[DST_2]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP21:%.*]] = getelementptr i32, ptr [[TMP20]], i32 0
-; CHECK-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP21]], align 8
+; CHECK-NEXT:    store <4 x i32> zeroinitializer, ptr [[TMP20]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i32> [[VEC_IND]], splat (i32 4)
 ; CHECK-NEXT:    [[TMP22:%.*]] = icmp eq i64 [[INDEX_NEXT]], 200
 ; CHECK-NEXT:    br i1 [[TMP22]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP10:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 false, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[SCALAR_PH]]
 ; CHECK:       scalar.ph:
 ; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 200, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
 ; CHECK-NEXT:    [[BC_RESUME_VAL1:%.*]] = phi i32 [ 200, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY]] ], [ 0, [[VECTOR_SCEVCHECK]] ]
@@ -376,7 +372,7 @@ define void @test_versioned_with_non_ex_use(i32 %offset, ptr noalias %dst.1, ptr
 ; CHECK-NEXT:    [[IV_2_NEXT]] = add i64 [[IV_2]], 1
 ; CHECK-NEXT:    [[IV_3_NEXT]] = add i32 [[IV_3]], 1
 ; CHECK-NEXT:    [[EC:%.*]] = icmp eq i32 [[IV_3]], 200
-; CHECK-NEXT:    br i1 [[EC]], label [[EXIT]], label [[LOOP]], !llvm.loop [[LOOP11:![0-9]+]]
+; CHECK-NEXT:    br i1 [[EC]], label [[EXIT:%.*]], label [[LOOP]], !llvm.loop [[LOOP11:![0-9]+]]
 ; CHECK:       exit:
 ; CHECK-NEXT:    ret void
 ;
@@ -427,8 +423,7 @@ define void @zext_of_i1_stride(i1 %g, ptr %dst) mustprogress {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = mul i64 [[INDEX]], [[G_64]]
 ; CHECK-NEXT:    [[TMP4:%.*]] = getelementptr inbounds i16, ptr [[DST]], i64 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP5:%.*]] = getelementptr inbounds i16, ptr [[TMP4]], i32 0
-; CHECK-NEXT:    store <4 x i16> splat (i16 1), ptr [[TMP5]], align 2
+; CHECK-NEXT:    store <4 x i16> splat (i16 1), ptr [[TMP4]], align 2
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP6:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
 ; CHECK-NEXT:    br i1 [[TMP6]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP12:![0-9]+]]
diff --git a/llvm/test/Transforms/LoopVectorize/vplan-iv-transforms.ll b/llvm/test/Transforms/LoopVectorize/vplan-iv-transforms.ll
index 49a7fb7..128594c 100644
--- a/llvm/test/Transforms/LoopVectorize/vplan-iv-transforms.ll
+++ b/llvm/test/Transforms/LoopVectorize/vplan-iv-transforms.ll
@@ -104,11 +104,9 @@ define void @iv_expand(ptr %p, i64 %n) {
 ; CHECK-NEXT:   EMIT-SCALAR vp<[[SCALAR_PHI:%.+]]> = phi [ ir<0>, ir-bb<vector.ph> ], [ vp<%index.next>, vector.body ]
 ; CHECK-NEXT:   WIDEN-PHI ir<%iv> = phi [ vp<[[INDUCTION]]>, ir-bb<vector.ph> ], [ vp<%vec.ind.next>, vector.body ]
 ; CHECK-NEXT:   CLONE ir<%q> = getelementptr ir<%p>, vp<[[SCALAR_PHI]]>
-; CHECK-NEXT:   vp<[[VEC_PTR_1:%.+]]> = vector-pointer ir<%q>
-; CHECK-NEXT:   WIDEN ir<%x> = load vp<[[VEC_PTR_1]]>
+; CHECK-NEXT:   WIDEN ir<%x> = load ir<%q>
 ; CHECK-NEXT:   WIDEN ir<%y> = add ir<%x>, ir<%iv>
-; CHECK-NEXT:   vp<[[VEC_PTR_2:%.+]]> = vector-pointer ir<%q>
-; CHECK-NEXT:   WIDEN store vp<[[VEC_PTR_2]]>, ir<%y>
+; CHECK-NEXT:   WIDEN store ir<%q>, ir<%y>
 ; CHECK-NEXT:   EMIT vp<%index.next> = add nuw vp<[[SCALAR_PHI]]>, ir<8>
 ; CHECK-NEXT:   EMIT vp<%vec.ind.next> = add ir<%iv>, vp<[[BROADCAST_INC]]>
 ; CHECK-NEXT:   EMIT branch-on-count vp<%index.next>, ir<%n.vec>
diff --git a/llvm/test/Transforms/LoopVectorize/vplan-predicate-switch.ll b/llvm/test/Transforms/LoopVectorize/vplan-predicate-switch.ll
index d441e41..46f9125 100644
--- a/llvm/test/Transforms/LoopVectorize/vplan-predicate-switch.ll
+++ b/llvm/test/Transforms/LoopVectorize/vplan-predicate-switch.ll
@@ -21,8 +21,7 @@ define void @switch4_default_common_dest_with_case(ptr %start, ptr %end) {
 ; CHECK-NEXT:   EMIT-SCALAR vp<[[CAN_IV:%.+]]> = phi [ ir<0>, ir-bb<vector.ph> ], [ vp<[[CAN_IV_NEXT:%.+]]>, default.2 ]
 ; CHECK-NEXT:   vp<[[STEPS:%.+]]> = SCALAR-STEPS vp<[[CAN_IV]]>, ir<1>, ir<2>
 ; CHECK-NEXT:   EMIT vp<[[PTR:%.+]]> = ptradd ir<%start>, vp<[[STEPS]]>
-; CHECK-NEXT:   vp<[[WIDE_PTR:%.+]]> = vector-pointer vp<[[PTR]]>
-; CHECK-NEXT:   WIDEN ir<%l> = load vp<[[WIDE_PTR]]>
+; CHECK-NEXT:   WIDEN ir<%l> = load vp<[[PTR]]>
 ; CHECK-NEXT:   EMIT vp<[[C1:%.+]]> = icmp eq ir<%l>, ir<-12>
 ; CHECK-NEXT:   EMIT vp<[[C2:%.+]]> = icmp eq ir<%l>, ir<13>
 ; CHECK-NEXT:   EMIT vp<[[OR_CASES:%.+]]> = or vp<[[C1]]>, vp<[[C2]]>
diff --git a/llvm/test/Transforms/LoopVectorize/vplan-printing-before-execute.ll b/llvm/test/Transforms/LoopVectorize/vplan-printing-before-execute.ll
index 91e70d2..d856387 100644
--- a/llvm/test/Transforms/LoopVectorize/vplan-printing-before-execute.ll
+++ b/llvm/test/Transforms/LoopVectorize/vplan-printing-before-execute.ll
@@ -77,15 +77,13 @@ define void @test_tc_less_than_16(ptr %A, i64 %N) {
 ; CHECK-NEXT: Successor(s): vector.body
 ; CHECK-EMPTY:
 ; CHECK-NEXT: vector.body:
-; CHECK-NEXT:   vp<[[VPTR1:%.]]> = vector-pointer ir<%A>
 ; CHECK-NEXT:   vp<[[VPTR2:%.]]> = vector-pointer ir<%A>, ir<1>
-; CHECK-NEXT:   WIDEN ir<%l> = load vp<[[VPTR1]]>
+; CHECK-NEXT:   WIDEN ir<%l> = load ir<%A>
 ; CHECK-NEXT:   WIDEN ir<%l>.1 = load vp<[[VPTR2]]>
 ; CHECK-NEXT:   WIDEN ir<%add> = add nsw ir<%l>, ir<10>
 ; CHECK-NEXT:   WIDEN ir<%add>.1 = add nsw ir<%l>.1, ir<10>
-; CHECK-NEXT:   vp<[[VPTR3:%.+]]> = vector-pointer ir<%A>
 ; CHECK-NEXT:   vp<[[VPTR4:%.+]]> = vector-pointer ir<%A>, ir<1>
-; CHECK-NEXT:   WIDEN store vp<[[VPTR3]]>, ir<%add>
+; CHECK-NEXT:   WIDEN store ir<%A>, ir<%add>
 ; CHECK-NEXT:   WIDEN store vp<[[VPTR4]]>, ir<%add>.1
 ; CHECK-NEXT: Successor(s): middle.block
 ; CHECK-EMPTY:
diff --git a/llvm/test/Transforms/LoopVectorize/widen-gep-all-indices-invariant.ll b/llvm/test/Transforms/LoopVectorize/widen-gep-all-indices-invariant.ll
index 7572a92..5d0d391 100644
--- a/llvm/test/Transforms/LoopVectorize/widen-gep-all-indices-invariant.ll
+++ b/llvm/test/Transforms/LoopVectorize/widen-gep-all-indices-invariant.ll
@@ -15,15 +15,14 @@ define void @pr63340(ptr %A, ptr %B) {
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i8
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds ptr, ptr [[B]], i8 [[OFFSET_IDX]]
-; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds ptr, ptr [[TMP1]], i32 0
-; CHECK-NEXT:    store <4 x ptr> [[DOTSPLAT]], ptr [[TMP2]], align 8
+; CHECK-NEXT:    store <4 x ptr> [[DOTSPLAT]], ptr [[TMP1]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 128
-; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK-NEXT:    [[TMP2:%.*]] = icmp eq i32 [[INDEX_NEXT]], 128
+; CHECK-NEXT:    br i1 [[TMP2]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ -128, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i8 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
 ; CHECK:       loop.header:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i8 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
@@ -74,15 +73,14 @@ define void @wide_gep_index_invariant(ptr noalias %dst, ptr noalias %src, i64 %n
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x ptr> [[BROADCAST_SPLATINSERT]], <4 x ptr> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr float, <4 x ptr> [[BROADCAST_SPLAT]], i64 [[N]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr ptr, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr ptr, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    store <4 x ptr> [[TMP1]], ptr [[TMP3]], align 8
+; CHECK-NEXT:    store <4 x ptr> [[TMP1]], ptr [[TMP2]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
-; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
-; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
+; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP4:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
@@ -128,16 +126,15 @@ define void @wide_gep_multiple_indices_some_invariant(ptr noalias %dst, ptr noal
 ; CHECK-NEXT:    [[BROADCAST_SPLAT:%.*]] = shufflevector <4 x ptr> [[BROADCAST_SPLATINSERT]], <4 x ptr> poison, <4 x i32> zeroinitializer
 ; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr [10 x float], <4 x ptr> [[BROADCAST_SPLAT]], i32 [[X]], <4 x i64> [[VEC_IND]]
 ; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr ptr, ptr [[DST]], i64 [[INDEX]]
-; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr ptr, ptr [[TMP2]], i32 0
-; CHECK-NEXT:    store <4 x ptr> [[TMP1]], ptr [[TMP3]], align 8
+; CHECK-NEXT:    store <4 x ptr> [[TMP1]], ptr [[TMP2]], align 8
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 4
 ; CHECK-NEXT:    [[VEC_IND_NEXT]] = add <4 x i64> [[VEC_IND]], splat (i64 4)
-; CHECK-NEXT:    [[TMP4:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
-; CHECK-NEXT:    br i1 [[TMP4]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
+; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i64 [[INDEX_NEXT]], 100
+; CHECK-NEXT:    br i1 [[TMP3]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP6:![0-9]+]]
 ; CHECK:       middle.block:
-; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK-NEXT:    br label [[EXIT:%.*]]
 ; CHECK:       scalar.ph:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 100, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ]
 ; CHECK-NEXT:    br label [[LOOP:%.*]]
 ; CHECK:       loop:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVectorize/widen-intrinsic.ll b/llvm/test/Transforms/LoopVectorize/widen-intrinsic.ll
index 072ff23..c23d2b3 100644
--- a/llvm/test/Transforms/LoopVectorize/widen-intrinsic.ll
+++ b/llvm/test/Transforms/LoopVectorize/widen-intrinsic.ll
@@ -12,17 +12,16 @@ define void @powi_only_first_lane_used_of_second_arg(ptr %p, i32 %pow) {
 ; CHECK:       [[VECTOR_BODY]]:
 ; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
 ; CHECK-NEXT:    [[TMP0:%.*]] = getelementptr float, ptr [[P]], i32 [[INDEX]]
-; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr float, ptr [[TMP0]], i32 0
-; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP1]], align 4
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <4 x float>, ptr [[TMP0]], align 4
 ; CHECK-NEXT:    [[TMP2:%.*]] = call <4 x float> @llvm.powi.v4f32.i32(<4 x float> [[WIDE_LOAD]], i32 [[POW]])
-; CHECK-NEXT:    store <4 x float> [[TMP2]], ptr [[TMP1]], align 4
+; CHECK-NEXT:    store <4 x float> [[TMP2]], ptr [[TMP0]], align 4
 ; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1024
 ; CHECK-NEXT:    br i1 [[TMP3]], label %[[MIDDLE_BLOCK:.*]], label %[[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
 ; CHECK:       [[MIDDLE_BLOCK]]:
-; CHECK-NEXT:    br i1 true, label %[[EXIT:.*]], label %[[SCALAR_PH]]
+; CHECK-NEXT:    br label %[[EXIT:.*]]
 ; CHECK:       [[SCALAR_PH]]:
-; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 1024, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
 ; CHECK-NEXT:    br label %[[LOOP:.*]]
 ; CHECK:       [[LOOP]]:
 ; CHECK-NEXT:    [[IV:%.*]] = phi i32 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
diff --git a/llvm/test/Transforms/LoopVersioning/invalidate-laa-after-versioning.ll b/llvm/test/Transforms/LoopVersioning/invalidate-laa-after-versioning.ll
index 8075314..4148c35 100644
--- a/llvm/test/Transforms/LoopVersioning/invalidate-laa-after-versioning.ll
+++ b/llvm/test/Transforms/LoopVersioning/invalidate-laa-after-versioning.ll
@@ -56,19 +56,14 @@ define void @test(ptr %arg, i64 %arg1) {
 ; CHECK-NEXT:    [[GEP_5:%.*]] = getelementptr inbounds double, ptr [[LCSSA_PTR_IV_1]], i64 1
 ; CHECK-NEXT:    br label [[INNER_2:%.*]]
 ; CHECK:       inner.2:
-; CHECK-NEXT:    [[INDVAR:%.*]] = phi i64 [ [[INDVAR_NEXT:%.*]], [[INNER_2]] ], [ 0, [[INNER_1_EXIT]] ]
 ; CHECK-NEXT:    [[PTR_IV_2:%.*]] = phi ptr [ [[GEP_5]], [[INNER_1_EXIT]] ], [ [[PTR_IV_2_NEXT:%.*]], [[INNER_2]] ]
 ; CHECK-NEXT:    [[PTR_IV_2_NEXT]] = getelementptr inbounds double, ptr [[PTR_IV_2]], i64 1
-; CHECK-NEXT:    [[INDVAR_NEXT]] = add i64 [[INDVAR]], 1
 ; CHECK-NEXT:    br i1 false, label [[INNER_3_LVER_CHECK:%.*]], label [[INNER_2]]
 ; CHECK:       inner.3.lver.check:
-; CHECK-NEXT:    [[INDVAR_LCSSA:%.*]] = phi i64 [ [[INDVAR]], [[INNER_2]] ]
 ; CHECK-NEXT:    [[LCSSA_PTR_IV_2:%.*]] = phi ptr [ [[PTR_IV_2]], [[INNER_2]] ]
 ; CHECK-NEXT:    [[GEP_6:%.*]] = getelementptr inbounds double, ptr [[PTR_PHI]], i64 1
 ; CHECK-NEXT:    [[GEP_7:%.*]] = getelementptr inbounds double, ptr [[LCSSA_PTR_IV_2]], i64 1
-; CHECK-NEXT:    [[TMP0:%.*]] = shl i64 [[INDVAR_LCSSA]], 3
-; CHECK-NEXT:    [[TMP1:%.*]] = add i64 [[TMP0]], 24
-; CHECK-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[LCSSA_PTR_IV_1]], i64 [[TMP1]]
+; CHECK-NEXT:    [[SCEVGEP:%.*]] = getelementptr i8, ptr [[LCSSA_PTR_IV_2]], i64 16
 ; CHECK-NEXT:    [[BOUND0:%.*]] = icmp ult ptr [[GEP_7]], [[GEP_1]]
 ; CHECK-NEXT:    [[BOUND1:%.*]] = icmp ult ptr [[PTR_PHI]], [[SCEVGEP]]
 ; CHECK-NEXT:    [[FOUND_CONFLICT:%.*]] = and i1 [[BOUND0]], [[BOUND1]]
@@ -90,10 +85,10 @@ define void @test(ptr %arg, i64 %arg1) {
 ; CHECK:       inner.3:
 ; CHECK-NEXT:    [[IV_2:%.*]] = phi i64 [ 0, [[INNER_3_PH]] ], [ [[IV_2_NEXT:%.*]], [[INNER_3]] ]
 ; CHECK-NEXT:    [[GEP_8:%.*]] = getelementptr inbounds double, ptr [[GEP_6]], i64 [[IV_2]]
-; CHECK-NEXT:    store double 0.000000e+00, ptr [[GEP_7]], align 8, !alias.scope !0, !noalias !3
-; CHECK-NEXT:    store double 0.000000e+00, ptr [[GEP_8]], align 8, !alias.scope !3
+; CHECK-NEXT:    store double 0.000000e+00, ptr [[GEP_7]], align 8, !alias.scope [[META0:![0-9]+]], !noalias [[META3:![0-9]+]]
+; CHECK-NEXT:    store double 0.000000e+00, ptr [[GEP_8]], align 8, !alias.scope [[META3]]
 ; CHECK-NEXT:    [[GEP_9:%.*]] = getelementptr double, ptr [[PTR_PHI]], i64 [[IV_2]]
-; CHECK-NEXT:    [[TMP18:%.*]] = load double, ptr [[GEP_9]], align 8, !alias.scope !3
+; CHECK-NEXT:    [[TMP18:%.*]] = load double, ptr [[GEP_9]], align 8, !alias.scope [[META3]]
 ; CHECK-NEXT:    [[IV_2_NEXT]] = add nuw nsw i64 [[IV_2]], 1
 ; CHECK-NEXT:    [[C_2:%.*]] = icmp eq i64 [[IV_2]], 1
 ; CHECK-NEXT:    br i1 [[C_2]], label [[OUTER_LATCH_LOOPEXIT3:%.*]], label [[INNER_3]]
diff --git a/llvm/test/Transforms/MemCpyOpt/stack-move.ll b/llvm/test/Transforms/MemCpyOpt/stack-move.ll
index 4a75c5e..31e255b 100644
--- a/llvm/test/Transforms/MemCpyOpt/stack-move.ll
+++ b/llvm/test/Transforms/MemCpyOpt/stack-move.ll
@@ -1649,20 +1649,13 @@ loop_exit:
   ret void
 }
 
-; Tests that failure because partial-sized lifetimes are counted as mod.
+; Tests that partial-sized lifetimes are not inhibiting the optimizer
 define void @partial_lifetime() {
 ; CHECK-LABEL: define void @partial_lifetime() {
-; CHECK-NEXT:    [[SRC:%.*]] = alloca [[STRUCT_FOO:%.*]], align 4
-; CHECK-NEXT:    [[DEST:%.*]] = alloca [[STRUCT_FOO]], align 4
-; CHECK-NEXT:    call void @llvm.lifetime.start.p0(i64 12, ptr captures(none) [[SRC]])
-; CHECK-NEXT:    call void @llvm.lifetime.start.p0(i64 3, ptr captures(none) [[DEST]])
-; CHECK-NEXT:    store [[STRUCT_FOO]] { i32 10, i32 20, i32 30 }, ptr [[SRC]], align 4
-; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @use_nocapture(ptr captures(none) [[SRC]])
-; CHECK-NEXT:    call void @llvm.memcpy.p0.p0.i64(ptr align 4 [[DEST]], ptr align 4 [[SRC]], i64 12, i1 false)
-; CHECK-NEXT:    call void @llvm.lifetime.end.p0(i64 3, ptr captures(none) [[SRC]])
+; CHECK-NEXT:    [[DEST:%.*]] = alloca [[STRUCT_FOO:%.*]], align 4
+; CHECK-NEXT:    store [[STRUCT_FOO]] { i32 10, i32 20, i32 30 }, ptr [[DEST]], align 4
+; CHECK-NEXT:    [[TMP1:%.*]] = call i32 @use_nocapture(ptr captures(none) [[DEST]])
 ; CHECK-NEXT:    [[TMP2:%.*]] = call i32 @use_nocapture(ptr captures(none) [[DEST]])
-; CHECK-NEXT:    call void @llvm.lifetime.end.p0(i64 12, ptr captures(none) [[SRC]])
-; CHECK-NEXT:    call void @llvm.lifetime.end.p0(i64 12, ptr captures(none) [[DEST]])
 ; CHECK-NEXT:    ret void
 ;
   %src = alloca %struct.Foo, align 4
diff --git a/llvm/test/Transforms/ObjCARC/apelim.ll b/llvm/test/Transforms/ObjCARC/apelim.ll
index 2ac5d15d0..01179f3 100644
--- a/llvm/test/Transforms/ObjCARC/apelim.ll
+++ b/llvm/test/Transforms/ObjCARC/apelim.ll
@@ -1,4 +1,4 @@
-; RUN: opt -S -passes=objc-arc-apelim < %s | FileCheck %s
+; RUN: opt -S -passes=objc-arc < %s | FileCheck %s
 ; rdar://10227311
 
 @llvm.global_ctors = appending global [2 x { i32, ptr, ptr }] [{ i32, ptr, ptr } { i32 65535, ptr @_GLOBAL__I_x, ptr null }, { i32, ptr, ptr } { i32 65535, ptr @_GLOBAL__I_y, ptr null }]
diff --git a/llvm/test/Transforms/ObjCARC/comdat-ipo.ll b/llvm/test/Transforms/ObjCARC/comdat-ipo.ll
index 3f91d3b..d43804c 100644
--- a/llvm/test/Transforms/ObjCARC/comdat-ipo.ll
+++ b/llvm/test/Transforms/ObjCARC/comdat-ipo.ll
@@ -1,4 +1,4 @@
-; RUN: opt -S -passes=objc-arc-apelim < %s | FileCheck %s
+; RUN: opt -S -passes=objc-arc < %s | FileCheck %s
 
 ; See PR26774
 
diff --git a/llvm/test/Transforms/PGOProfile/prof-verify.ll b/llvm/test/Transforms/PGOProfile/prof-verify.ll
index c83475a..3d984d8 100644
--- a/llvm/test/Transforms/PGOProfile/prof-verify.ll
+++ b/llvm/test/Transforms/PGOProfile/prof-verify.ll
@@ -3,6 +3,7 @@
 ; RUN: opt -passes=prof-inject %s -S -o - | FileCheck %s --check-prefix=INJECT
 ; RUN: not opt -passes=prof-verify %s -S -o - 2>&1 | FileCheck %s --check-prefix=VERIFY
 ; RUN: opt -passes=prof-inject,prof-verify %s --disable-output
+; RUN: opt -enable-profcheck %s -S -o - | FileCheck %s --check-prefix=INJECT
 
 define void @foo(i32 %i) {
   %c = icmp eq i32 %i, 0
diff --git a/llvm/test/Transforms/PhaseOrdering/AArch64/slpordering.ll b/llvm/test/Transforms/PhaseOrdering/AArch64/slpordering.ll
index 26573a3..a201983 100644
--- a/llvm/test/Transforms/PhaseOrdering/AArch64/slpordering.ll
+++ b/llvm/test/Transforms/PhaseOrdering/AArch64/slpordering.ll
@@ -80,29 +80,33 @@ define i32 @slpordering(ptr noundef %p1, i32 noundef %ip1, ptr noundef %p2, i32
 ; CHECK-NEXT:    [[TMP47:%.*]] = shufflevector <16 x i32> [[TMP43]], <16 x i32> poison, <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
 ; CHECK-NEXT:    [[TMP48:%.*]] = add nsw <16 x i32> [[TMP45]], [[TMP47]]
 ; CHECK-NEXT:    [[TMP49:%.*]] = sub nsw <16 x i32> [[TMP44]], [[TMP46]]
-; CHECK-NEXT:    [[TMP50:%.*]] = shufflevector <16 x i32> [[TMP48]], <16 x i32> [[TMP49]], <16 x i32> <i32 16, i32 0, i32 17, i32 1, i32 18, i32 2, i32 19, i32 3, i32 20, i32 4, i32 21, i32 5, i32 22, i32 6, i32 23, i32 7>
-; CHECK-NEXT:    [[TMP51:%.*]] = shufflevector <16 x i32> [[TMP48]], <16 x i32> [[TMP49]], <16 x i32> <i32 17, i32 1, i32 16, i32 0, i32 19, i32 3, i32 18, i32 2, i32 21, i32 5, i32 20, i32 4, i32 23, i32 7, i32 22, i32 6>
-; CHECK-NEXT:    [[TMP52:%.*]] = add nsw <16 x i32> [[TMP50]], [[TMP51]]
-; CHECK-NEXT:    [[TMP53:%.*]] = sub nsw <16 x i32> [[TMP50]], [[TMP51]]
-; CHECK-NEXT:    [[TMP54:%.*]] = shufflevector <16 x i32> [[TMP53]], <16 x i32> [[TMP52]], <16 x i32> <i32 0, i32 1, i32 18, i32 19, i32 4, i32 5, i32 22, i32 23, i32 8, i32 9, i32 26, i32 27, i32 12, i32 13, i32 30, i32 31>
-; CHECK-NEXT:    [[TMP55:%.*]] = shufflevector <16 x i32> [[TMP54]], <16 x i32> poison, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 12, i32 13, i32 14, i32 15, i32 8, i32 9, i32 10, i32 11>
-; CHECK-NEXT:    [[TMP56:%.*]] = sub nsw <16 x i32> [[TMP54]], [[TMP55]]
-; CHECK-NEXT:    [[TMP57:%.*]] = add nsw <16 x i32> [[TMP54]], [[TMP55]]
-; CHECK-NEXT:    [[TMP58:%.*]] = shufflevector <16 x i32> [[TMP56]], <16 x i32> [[TMP57]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP59:%.*]] = shufflevector <16 x i32> [[TMP56]], <16 x i32> [[TMP57]], <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP60:%.*]] = shufflevector <16 x i32> [[TMP56]], <16 x i32> [[TMP57]], <16 x i32> <i32 8, i32 9, i32 10, i32 11, i32 28, i32 29, i32 30, i32 31, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP61:%.*]] = shufflevector <16 x i32> [[TMP56]], <16 x i32> [[TMP57]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP62:%.*]] = add nsw <16 x i32> [[TMP59]], [[TMP61]]
-; CHECK-NEXT:    [[TMP63:%.*]] = sub nsw <16 x i32> [[TMP58]], [[TMP60]]
-; CHECK-NEXT:    [[TMP64:%.*]] = shufflevector <16 x i32> [[TMP62]], <16 x i32> [[TMP63]], <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>
-; CHECK-NEXT:    [[TMP65:%.*]] = lshr <16 x i32> [[TMP64]], splat (i32 15)
-; CHECK-NEXT:    [[TMP66:%.*]] = and <16 x i32> [[TMP65]], splat (i32 65537)
-; CHECK-NEXT:    [[TMP67:%.*]] = mul nuw <16 x i32> [[TMP66]], splat (i32 65535)
-; CHECK-NEXT:    [[TMP68:%.*]] = add <16 x i32> [[TMP67]], [[TMP64]]
-; CHECK-NEXT:    [[TMP69:%.*]] = xor <16 x i32> [[TMP68]], [[TMP67]]
-; CHECK-NEXT:    [[TMP70:%.*]] = tail call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[TMP69]])
-; CHECK-NEXT:    [[CONV118:%.*]] = and i32 [[TMP70]], 65535
-; CHECK-NEXT:    [[SHR:%.*]] = lshr i32 [[TMP70]], 16
+; CHECK-NEXT:    [[TMP50:%.*]] = shufflevector <16 x i32> [[TMP48]], <16 x i32> [[TMP49]], <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 16, i32 18, i32 20, i32 22, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP51:%.*]] = shufflevector <16 x i32> [[TMP48]], <16 x i32> [[TMP49]], <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 17, i32 19, i32 21, i32 23, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP52:%.*]] = shufflevector <16 x i32> [[TMP48]], <16 x i32> [[TMP49]], <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 17, i32 19, i32 21, i32 23, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP53:%.*]] = shufflevector <16 x i32> [[TMP48]], <16 x i32> [[TMP49]], <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 16, i32 18, i32 20, i32 22, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP54:%.*]] = add nsw <16 x i32> [[TMP51]], [[TMP53]]
+; CHECK-NEXT:    [[TMP55:%.*]] = sub nsw <16 x i32> [[TMP50]], [[TMP52]]
+; CHECK-NEXT:    [[TMP56:%.*]] = shufflevector <16 x i32> [[TMP54]], <16 x i32> [[TMP55]], <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 17, i32 19, i32 21, i32 23, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP57:%.*]] = shufflevector <16 x i32> [[TMP54]], <16 x i32> [[TMP55]], <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 16, i32 18, i32 20, i32 22, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP58:%.*]] = shufflevector <16 x i32> [[TMP54]], <16 x i32> [[TMP55]], <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 16, i32 18, i32 20, i32 22, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP59:%.*]] = shufflevector <16 x i32> [[TMP54]], <16 x i32> [[TMP55]], <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 17, i32 19, i32 21, i32 23, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP60:%.*]] = sub nsw <16 x i32> [[TMP57]], [[TMP59]]
+; CHECK-NEXT:    [[TMP61:%.*]] = add nsw <16 x i32> [[TMP56]], [[TMP58]]
+; CHECK-NEXT:    [[TMP62:%.*]] = shufflevector <16 x i32> [[TMP60]], <16 x i32> [[TMP61]], <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 16, i32 18, i32 20, i32 22, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP63:%.*]] = shufflevector <16 x i32> [[TMP60]], <16 x i32> [[TMP61]], <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 17, i32 19, i32 21, i32 23, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP64:%.*]] = shufflevector <16 x i32> [[TMP60]], <16 x i32> [[TMP61]], <16 x i32> <i32 1, i32 3, i32 5, i32 7, i32 17, i32 19, i32 21, i32 23, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP65:%.*]] = shufflevector <16 x i32> [[TMP60]], <16 x i32> [[TMP61]], <16 x i32> <i32 0, i32 2, i32 4, i32 6, i32 16, i32 18, i32 20, i32 22, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP66:%.*]] = add nsw <16 x i32> [[TMP63]], [[TMP65]]
+; CHECK-NEXT:    [[TMP67:%.*]] = sub nsw <16 x i32> [[TMP62]], [[TMP64]]
+; CHECK-NEXT:    [[TMP68:%.*]] = shufflevector <16 x i32> [[TMP66]], <16 x i32> [[TMP67]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23>
+; CHECK-NEXT:    [[TMP69:%.*]] = lshr <16 x i32> [[TMP68]], splat (i32 15)
+; CHECK-NEXT:    [[TMP70:%.*]] = and <16 x i32> [[TMP69]], splat (i32 65537)
+; CHECK-NEXT:    [[TMP71:%.*]] = mul nuw <16 x i32> [[TMP70]], splat (i32 65535)
+; CHECK-NEXT:    [[TMP72:%.*]] = add <16 x i32> [[TMP71]], [[TMP68]]
+; CHECK-NEXT:    [[TMP73:%.*]] = xor <16 x i32> [[TMP72]], [[TMP71]]
+; CHECK-NEXT:    [[TMP74:%.*]] = tail call i32 @llvm.vector.reduce.add.v16i32(<16 x i32> [[TMP73]])
+; CHECK-NEXT:    [[CONV118:%.*]] = and i32 [[TMP74]], 65535
+; CHECK-NEXT:    [[SHR:%.*]] = lshr i32 [[TMP74]], 16
 ; CHECK-NEXT:    [[RDD119:%.*]] = add nuw nsw i32 [[CONV118]], [[SHR]]
 ; CHECK-NEXT:    [[SHR120:%.*]] = lshr i32 [[RDD119]], 1
 ; CHECK-NEXT:    ret i32 [[SHR120]]
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/buildvector-schedule-for-subvector.ll b/llvm/test/Transforms/SLPVectorizer/X86/buildvector-schedule-for-subvector.ll
index 07fdc9d..7408ba1 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/buildvector-schedule-for-subvector.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/buildvector-schedule-for-subvector.ll
@@ -4,9 +4,6 @@
 define void @test() {
 ; CHECK-LABEL: define void @test() {
 ; CHECK-NEXT:  [[BB:.*:]]
-; CHECK-NEXT:    [[ADD:%.*]] = add i32 1, 0
-; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i32> <i32 0, i32 0, i32 0, i32 poison>, i32 [[ADD]], i32 3
-; CHECK-NEXT:    [[TMP1:%.*]] = icmp ult <4 x i32> [[TMP0]], zeroinitializer
 ; CHECK-NEXT:    [[ICMP:%.*]] = icmp samesign ult i32 0, 0
 ; CHECK-NEXT:    [[SELECT:%.*]] = select i1 [[ICMP]], i32 0, i32 0
 ; CHECK-NEXT:    [[ZEXT:%.*]] = zext i32 [[SELECT]] to i64
@@ -17,8 +14,7 @@ define void @test() {
 ; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <4 x i32> <i32 0, i32 0, i32 0, i32 poison>, i32 [[CALL]], i32 3
 ; CHECK-NEXT:    [[TMP3:%.*]] = icmp eq <4 x i32> [[TMP2]], zeroinitializer
 ; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i1> [[TMP3]], <4 x i1> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <4 x i1> [[TMP1]], <4 x i1> poison, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP6:%.*]] = shufflevector <8 x i1> [[TMP4]], <8 x i1> [[TMP5]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
+; CHECK-NEXT:    [[TMP5:%.*]] = shufflevector <8 x i1> [[TMP4]], <8 x i1> <i1 false, i1 false, i1 false, i1 false, i1 undef, i1 undef, i1 undef, i1 undef>, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 8, i32 9, i32 10, i32 11>
 ; CHECK-NEXT:    ret void
 ;
 bb:
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/extract-scalar-from-undef.ll b/llvm/test/Transforms/SLPVectorizer/X86/extract-scalar-from-undef.ll
index 514d5f9..7a1cf7b 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/extract-scalar-from-undef.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/extract-scalar-from-undef.ll
@@ -6,15 +6,13 @@ define i64 @foo(i32 %tmp7) {
 ; CHECK-NEXT:  bb:
 ; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <8 x i32> <i32 0, i32 0, i32 poison, i32 poison, i32 0, i32 poison, i32 poison, i32 poison>, i32 [[TMP8:%.*]], i32 3
 ; CHECK-NEXT:    [[TMP4:%.*]] = sub <8 x i32> [[TMP0]], <i32 0, i32 0, i32 poison, i32 0, i32 0, i32 poison, i32 0, i32 poison>
-; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 poison, i32 poison, i32 poison, i32 0>, <8 x i32> [[TMP4]], <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 14, i32 poison, i32 poison, i32 7>
-; CHECK-NEXT:    [[TMP13:%.*]] = insertelement <8 x i32> [[TMP2]], i32 0, i32 5
-; CHECK-NEXT:    [[TMP5:%.*]] = sub nsw <8 x i32> [[TMP13]], [[TMP4]]
-; CHECK-NEXT:    [[TMP6:%.*]] = add nsw <8 x i32> [[TMP13]], [[TMP4]]
+; CHECK-NEXT:    [[TMP5:%.*]] = sub nsw <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 poison, i32 0, i32 poison, i32 0>, [[TMP4]]
+; CHECK-NEXT:    [[TMP6:%.*]] = add nsw <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 poison, i32 0, i32 poison, i32 0>, [[TMP4]]
 ; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <8 x i32> [[TMP5]], <8 x i32> [[TMP6]], <8 x i32> <i32 8, i32 1, i32 2, i32 3, i32 12, i32 13, i32 6, i32 7>
-; CHECK-NEXT:    [[TMP8:%.*]] = add <8 x i32> zeroinitializer, [[TMP7]]
-; CHECK-NEXT:    [[TMP9:%.*]] = xor <8 x i32> [[TMP8]], zeroinitializer
-; CHECK-NEXT:    [[TMP10:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[TMP9]])
-; CHECK-NEXT:    [[OP_RDX:%.*]] = add i32 0, [[TMP10]]
+; CHECK-NEXT:    [[TMP9:%.*]] = add <8 x i32> zeroinitializer, [[TMP7]]
+; CHECK-NEXT:    [[TMP10:%.*]] = xor <8 x i32> [[TMP9]], zeroinitializer
+; CHECK-NEXT:    [[TMP8:%.*]] = call i32 @llvm.vector.reduce.add.v8i32(<8 x i32> [[TMP10]])
+; CHECK-NEXT:    [[OP_RDX:%.*]] = add i32 0, [[TMP8]]
 ; CHECK-NEXT:    [[TMP64:%.*]] = zext i32 [[OP_RDX]] to i64
 ; CHECK-NEXT:    ret i64 [[TMP64]]
 ;
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/full-match-with-poison-scalar.ll b/llvm/test/Transforms/SLPVectorizer/X86/full-match-with-poison-scalar.ll
index 15ba98f..5e3d471 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/full-match-with-poison-scalar.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/full-match-with-poison-scalar.ll
@@ -7,17 +7,10 @@ define i32 @test() {
 ; CHECK-NEXT:    br label %[[FUNC_135_EXIT_I:.*]]
 ; CHECK:       [[FUNC_135_EXIT_I]]:
 ; CHECK-NEXT:    [[G_228_PROMOTED166_I1105_I:%.*]] = phi i32 [ 0, %[[ENTRY]] ]
-; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <4 x i32> <i32 poison, i32 0, i32 poison, i32 poison>, i32 [[G_228_PROMOTED166_I1105_I]], i32 0
-; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <4 x i32> [[TMP0]], <4 x i32> poison, <4 x i32> <i32 0, i32 0, i32 0, i32 1>
-; CHECK-NEXT:    [[TMP2:%.*]] = add <4 x i32> zeroinitializer, [[TMP1]]
-; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> poison, <12 x i32> <i32 0, i32 0, i32 0, i32 1, i32 1, i32 1, i32 1, i32 2, i32 2, i32 2, i32 2, i32 3>
-; CHECK-NEXT:    [[TMP7:%.*]] = shufflevector <4 x i32> [[TMP0]], <4 x i32> poison, <16 x i32> <i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 0, i32 0, i32 0, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <16 x i32> poison, i32 [[G_228_PROMOTED166_I1105_I]], i32 0
-; CHECK-NEXT:    [[TMP9:%.*]] = shufflevector <16 x i32> [[TMP5]], <16 x i32> poison, <16 x i32> <i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 0, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP8:%.*]] = shufflevector <16 x i32> [[TMP7]], <16 x i32> [[TMP9]], <16 x i32> <i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 23, i32 8, i32 9, i32 10, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP10:%.*]] = shufflevector <12 x i32> [[TMP3]], <12 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP17:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP11:%.*]] = shufflevector <16 x i32> [[TMP17]], <16 x i32> [[TMP8]], <16 x i32> <i32 0, i32 0, i32 0, i32 1, i32 1, i32 1, i32 1, i32 23, i32 24, i32 25, i32 26, i32 2, i32 2, i32 2, i32 2, i32 3>
+; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <8 x i32> <i32 poison, i32 poison, i32 poison, i32 poison, i32 0, i32 poison, i32 poison, i32 poison>, i32 [[G_228_PROMOTED166_I1105_I]], i32 0
+; CHECK-NEXT:    [[TMP1:%.*]] = shufflevector <8 x i32> [[TMP0]], <8 x i32> poison, <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 4, i32 poison, i32 poison, i32 poison>
+; CHECK-NEXT:    [[TMP2:%.*]] = add <8 x i32> <i32 0, i32 0, i32 0, i32 0, i32 0, i32 poison, i32 poison, i32 poison>, [[TMP1]]
+; CHECK-NEXT:    [[TMP11:%.*]] = shufflevector <8 x i32> [[TMP2]], <8 x i32> poison, <16 x i32> <i32 0, i32 0, i32 0, i32 1, i32 1, i32 1, i32 1, i32 2, i32 2, i32 2, i32 2, i32 3, i32 3, i32 3, i32 3, i32 4>
 ; CHECK-NEXT:    [[TMP12:%.*]] = icmp ugt <16 x i32> [[TMP11]], zeroinitializer
 ; CHECK-NEXT:    [[TMP13:%.*]] = icmp ult <16 x i32> [[TMP11]], zeroinitializer
 ; CHECK-NEXT:    [[TMP14:%.*]] = shufflevector <16 x i1> [[TMP12]], <16 x i1> [[TMP13]], <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 31>
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/node-outside-used-only.ll b/llvm/test/Transforms/SLPVectorizer/X86/node-outside-used-only.ll
index 1c482e0..03d76ef 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/node-outside-used-only.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/node-outside-used-only.ll
@@ -4,11 +4,10 @@
 define i64 @test() {
 ; CHECK-LABEL: define i64 @test() {
 ; CHECK-NEXT:  [[BB:.*]]:
-; CHECK-NEXT:    [[TMP0:%.*]] = insertelement <2 x i32> <i32 0, i32 poison>, i32 0, i32 1
 ; CHECK-NEXT:    br label %[[BB1:.*]]
 ; CHECK:       [[BB1]]:
 ; CHECK-NEXT:    [[TMP1:%.*]] = phi <2 x i32> [ zeroinitializer, %[[BB]] ], [ [[TMP4:%.*]], %[[BB5:.*]] ]
-; CHECK-NEXT:    [[TMP2:%.*]] = or <2 x i32> [[TMP0]], [[TMP1]]
+; CHECK-NEXT:    [[TMP2:%.*]] = or <2 x i32> zeroinitializer, [[TMP1]]
 ; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> [[TMP2]], <2 x i32> <i32 0, i32 3>
 ; CHECK-NEXT:    [[TMP4]] = or <2 x i32> [[TMP3]], zeroinitializer
 ; CHECK-NEXT:    br label %[[BB5]]
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/non-schedulable-instructions-become-schedulable.ll b/llvm/test/Transforms/SLPVectorizer/X86/non-schedulable-instructions-become-schedulable.ll
index 652abef..6bb52e0 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/non-schedulable-instructions-become-schedulable.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/non-schedulable-instructions-become-schedulable.ll
@@ -7,19 +7,17 @@ define void @test() {
 ; CHECK-NEXT:  [[ENTRY:.*:]]
 ; CHECK-NEXT:    br label %[[BB1:.*]]
 ; CHECK:       [[IF_THEN_I_I:.*]]:
-; CHECK-NEXT:    br label %[[BB5:.*]]
+; CHECK-NEXT:    br label %[[BB3:.*]]
 ; CHECK:       [[BB1]]:
 ; CHECK-NEXT:    [[TMP0:%.*]] = zext i1 false to i64
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x i64> <i64 poison, i64 0>, i64 [[TMP0]], i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = add <2 x i64> zeroinitializer, [[TMP1]]
-; CHECK-NEXT:    [[TMP3:%.*]] = shufflevector <2 x i64> [[TMP2]], <2 x i64> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <4 x i64> <i64 0, i64 0, i64 poison, i64 poison>, <4 x i64> [[TMP3]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
-; CHECK-NEXT:    br i1 false, label %[[BB5]], label %[[BB2:.*]]
-; CHECK:       [[BB5]]:
-; CHECK-NEXT:    [[TMP6:%.*]] = phi <4 x i64> [ [[TMP4]], %[[BB1]] ], [ poison, %[[IF_THEN_I_I]] ]
+; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <4 x i64> <i64 0, i64 0, i64 poison, i64 0>, i64 [[TMP0]], i32 2
+; CHECK-NEXT:    [[TMP2:%.*]] = add <4 x i64> zeroinitializer, [[TMP1]]
+; CHECK-NEXT:    br i1 false, label %[[BB3]], label %[[BB2:.*]]
+; CHECK:       [[BB3]]:
+; CHECK-NEXT:    [[TMP4:%.*]] = phi <4 x i64> [ [[TMP2]], %[[BB1]] ], [ poison, %[[IF_THEN_I_I]] ]
 ; CHECK-NEXT:    br label %[[BB2]]
 ; CHECK:       [[BB2]]:
-; CHECK-NEXT:    [[TMP7:%.*]] = phi <4 x i64> [ [[TMP6]], %[[BB5]] ], [ [[TMP4]], %[[BB1]] ]
+; CHECK-NEXT:    [[TMP7:%.*]] = phi <4 x i64> [ [[TMP4]], %[[BB3]] ], [ [[TMP2]], %[[BB1]] ]
 ; CHECK-NEXT:    store <4 x i64> [[TMP7]], ptr getelementptr inbounds nuw (i8, ptr null, i64 40), align 8
 ; CHECK-NEXT:    ret void
 ;
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/pr47642.ll b/llvm/test/Transforms/SLPVectorizer/X86/pr47642.ll
index a4949bc..782aada 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/pr47642.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/pr47642.ll
@@ -6,14 +6,9 @@ target triple = "x86_64-unknown-linux-gnu"
 
 define <4 x i32> @foo(<4 x i32> %x, i32 %f) {
 ; CHECK-LABEL: @foo(
-; CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <4 x i32> undef, i32 [[F:%.*]], i32 0
-; CHECK-NEXT:    [[ADD:%.*]] = add nsw i32 [[F]], 1
-; CHECK-NEXT:    [[VECINIT1:%.*]] = insertelement <4 x i32> [[VECINIT]], i32 [[ADD]], i32 1
-; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x i32> poison, i32 [[F]], i32 0
-; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <2 x i32> [[TMP1]], <2 x i32> poison, <2 x i32> zeroinitializer
-; CHECK-NEXT:    [[TMP3:%.*]] = add nsw <2 x i32> [[TMP2]], <i32 2, i32 3>
-; CHECK-NEXT:    [[TMP4:%.*]] = shufflevector <2 x i32> [[TMP3]], <2 x i32> poison, <4 x i32> <i32 0, i32 1, i32 poison, i32 poison>
-; CHECK-NEXT:    [[VECINIT51:%.*]] = shufflevector <4 x i32> [[VECINIT1]], <4 x i32> [[TMP4]], <4 x i32> <i32 0, i32 1, i32 4, i32 5>
+; CHECK-NEXT:    [[VECINIT:%.*]] = insertelement <4 x i32> poison, i32 [[F:%.*]], i32 0
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <4 x i32> [[VECINIT]], <4 x i32> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[VECINIT51:%.*]] = add <4 x i32> [[TMP2]], <i32 0, i32 1, i32 2, i32 3>
 ; CHECK-NEXT:    ret <4 x i32> [[VECINIT51]]
 ;
   %vecinit = insertelement <4 x i32> undef, i32 %f, i32 0
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/remark_extract_broadcast.ll b/llvm/test/Transforms/SLPVectorizer/X86/remark_extract_broadcast.ll
index 049bb2e..c4d133a 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/remark_extract_broadcast.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/remark_extract_broadcast.ll
@@ -18,7 +18,7 @@ define void @fextr(ptr %ptr) {
 ; YAML-NEXT: Function:        fextr
 ; YAML-NEXT: Args:
 ; YAML-NEXT:   - String:          'Stores SLP vectorized with cost '
-; YAML-NEXT:   - Cost:            '-20'
+; YAML-NEXT:   - Cost:            '-21'
 ; YAML-NEXT:   - String:          ' and with tree size '
 ; YAML-NEXT:   - TreeSize:        '4'
 
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/revec-reduced-value-vectorized-later.ll b/llvm/test/Transforms/SLPVectorizer/X86/revec-reduced-value-vectorized-later.ll
index 8f6a53c..f7811ab 100644
--- a/llvm/test/Transforms/SLPVectorizer/X86/revec-reduced-value-vectorized-later.ll
+++ b/llvm/test/Transforms/SLPVectorizer/X86/revec-reduced-value-vectorized-later.ll
@@ -6,14 +6,6 @@ define <4 x i16> @test() {
 ; CHECK-NEXT:  [[ENTRY:.*:]]
 ; CHECK-NEXT:    [[TMP0:%.*]] = shufflevector <4 x i16> zeroinitializer, <4 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3, i32 0, i32 1, i32 2, i32 3>
 ; CHECK-NEXT:    [[TMP1:%.*]] = add <16 x i16> [[TMP0]], zeroinitializer
-; CHECK-NEXT:    [[TMP25:%.*]] = call i16 @llvm.vector.reduce.or.v4i16(<4 x i16> zeroinitializer)
-; CHECK-NEXT:    [[TMP26:%.*]] = insertelement <4 x i16> poison, i16 [[TMP25]], i64 0
-; CHECK-NEXT:    [[TMP28:%.*]] = call i16 @llvm.vector.reduce.or.v4i16(<4 x i16> zeroinitializer)
-; CHECK-NEXT:    [[TMP29:%.*]] = insertelement <4 x i16> [[TMP26]], i16 [[TMP28]], i64 1
-; CHECK-NEXT:    [[TMP31:%.*]] = call i16 @llvm.vector.reduce.or.v4i16(<4 x i16> zeroinitializer)
-; CHECK-NEXT:    [[TMP32:%.*]] = insertelement <4 x i16> [[TMP29]], i16 [[TMP31]], i64 2
-; CHECK-NEXT:    [[TMP34:%.*]] = call i16 @llvm.vector.reduce.or.v4i16(<4 x i16> zeroinitializer)
-; CHECK-NEXT:    [[TMP35:%.*]] = insertelement <4 x i16> [[TMP32]], i16 [[TMP34]], i64 3
 ; CHECK-NEXT:    [[RDX_OP:%.*]] = or <16 x i16> zeroinitializer, [[TMP1]]
 ; CHECK-NEXT:    [[TMP36:%.*]] = shufflevector <16 x i16> [[RDX_OP]], <16 x i16> poison, <4 x i32> <i32 0, i32 4, i32 8, i32 12>
 ; CHECK-NEXT:    [[TMP37:%.*]] = call i16 @llvm.vector.reduce.or.v4i16(<4 x i16> [[TMP36]])
@@ -28,8 +20,7 @@ define <4 x i16> @test() {
 ; CHECK-NEXT:    [[TMP46:%.*]] = call i16 @llvm.vector.reduce.or.v4i16(<4 x i16> [[TMP45]])
 ; CHECK-NEXT:    [[TMP47:%.*]] = insertelement <4 x i16> [[TMP44]], i16 [[TMP46]], i64 3
 ; CHECK-NEXT:    [[OP_RDX9:%.*]] = or <4 x i16> [[TMP47]], zeroinitializer
-; CHECK-NEXT:    [[OP_RDX11:%.*]] = or <4 x i16> [[OP_RDX9]], [[TMP35]]
-; CHECK-NEXT:    ret <4 x i16> [[OP_RDX11]]
+; CHECK-NEXT:    ret <4 x i16> [[OP_RDX9]]
 ;
 entry:
   %subi = add <4 x i16> zeroinitializer, zeroinitializer
diff --git a/llvm/test/Transforms/SLPVectorizer/X86/user-node-no-state.ll b/llvm/test/Transforms/SLPVectorizer/X86/user-node-no-state.ll
new file mode 100644
index 0000000..237f308
--- /dev/null
+++ b/llvm/test/Transforms/SLPVectorizer/X86/user-node-no-state.ll
@@ -0,0 +1,48 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux-gnu < %s | FileCheck %s
+
+@g = global [128 x i8] zeroinitializer, align 16
+
+define i64 @test() {
+; CHECK-LABEL: define i64 @test() {
+; CHECK-NEXT:  [[ENTRY:.*:]]
+; CHECK-NEXT:    [[TMP0:%.*]] = load i64, ptr @g, align 8
+; CHECK-NEXT:    br label %[[FUNC_154_EXIT_FUNC_146_EXIT_CRIT_EDGE_I:.*]]
+; CHECK:       [[FUNC_154_EXIT_FUNC_146_EXIT_CRIT_EDGE_I]]:
+; CHECK-NEXT:    [[TMP1:%.*]] = load i64, ptr getelementptr inbounds nuw (i8, ptr @g, i64 80), align 16
+; CHECK-NEXT:    [[TMP2:%.*]] = load i64, ptr getelementptr inbounds nuw (i8, ptr @g, i64 88), align 8
+; CHECK-NEXT:    [[TMP3:%.*]] = load i64, ptr getelementptr inbounds nuw (i8, ptr @g, i64 32), align 16
+; CHECK-NEXT:    [[TMP4:%.*]] = load i64, ptr @g, align 16
+; CHECK-NEXT:    [[TMP5:%.*]] = load i64, ptr getelementptr inbounds nuw (i8, ptr @g, i64 8), align 8
+; CHECK-NEXT:    [[TMP6:%.*]] = load i64, ptr @g, align 16
+; CHECK-NEXT:    [[TMP7:%.*]] = load i64, ptr getelementptr inbounds nuw (i8, ptr @g, i64 24), align 8
+; CHECK-NEXT:    [[TMP8:%.*]] = xor i64 [[TMP1]], [[TMP2]]
+; CHECK-NEXT:    [[TMP9:%.*]] = xor i64 [[TMP8]], [[TMP3]]
+; CHECK-NEXT:    [[TMP10:%.*]] = xor i64 [[TMP9]], [[TMP4]]
+; CHECK-NEXT:    [[TMP11:%.*]] = xor i64 [[TMP10]], [[TMP5]]
+; CHECK-NEXT:    [[TMP12:%.*]] = xor i64 [[TMP11]], [[TMP6]]
+; CHECK-NEXT:    [[TMP13:%.*]] = xor i64 [[TMP12]], [[TMP7]]
+; CHECK-NEXT:    [[TMP14:%.*]] = xor i64 [[TMP13]], [[TMP0]]
+; CHECK-NEXT:    ret i64 [[TMP14]]
+;
+entry:
+  %0 = load i64, ptr @g, align 8
+  br label %func_154.exit.func_146.exit_crit_edge.i
+
+func_154.exit.func_146.exit_crit_edge.i:
+  %1 = load i64, ptr getelementptr inbounds nuw (i8, ptr @g, i64 80), align 16
+  %2 = load i64, ptr getelementptr inbounds nuw (i8, ptr @g, i64 88), align 8
+  %3 = load i64, ptr getelementptr inbounds nuw (i8, ptr @g, i64 32), align 16
+  %4 = load i64, ptr @g, align 16
+  %5 = load i64, ptr getelementptr inbounds nuw (i8, ptr @g, i64 8), align 8
+  %6 = load i64, ptr @g, align 16
+  %7 = load i64, ptr getelementptr inbounds nuw (i8, ptr @g, i64 24), align 8
+  %8 = xor i64 %1, %2
+  %9 = xor i64 %8, %3
+  %10 = xor i64 %9, %4
+  %11 = xor i64 %10, %5
+  %12 = xor i64 %11, %6
+  %13 = xor i64 %12, %7
+  %14 = xor i64 %13, %0
+  ret i64 %14
+}
diff --git a/llvm/test/Transforms/SLPVectorizer/alternate-non-profitable.ll b/llvm/test/Transforms/SLPVectorizer/alternate-non-profitable.ll
index ad4daea..125c2dc 100644
--- a/llvm/test/Transforms/SLPVectorizer/alternate-non-profitable.ll
+++ b/llvm/test/Transforms/SLPVectorizer/alternate-non-profitable.ll
@@ -150,9 +150,9 @@ define <2 x i32> @replace_through_int_casts_ele0_only(i16 %inp, <2 x i16> %dead)
 define <2 x i8> @replace_through_binop_fail_cant_speculate(i8 %inp, <2 x i8> %d, <2 x i8> %any) {
 ; CHECK-LABEL: define <2 x i8> @replace_through_binop_fail_cant_speculate(
 ; CHECK-SAME: i8 [[INP:%.*]], <2 x i8> [[D:%.*]], <2 x i8> [[ANY:%.*]]) {
-; CHECK-NEXT:    [[ADD:%.*]] = add i8 [[INP]], 5
-; CHECK-NEXT:    [[V0:%.*]] = insertelement <2 x i8> poison, i8 [[INP]], i64 0
-; CHECK-NEXT:    [[V:%.*]] = insertelement <2 x i8> [[V0]], i8 [[ADD]], i64 1
+; CHECK-NEXT:    [[TMP3:%.*]] = insertelement <2 x i8> poison, i8 [[INP]], i32 0
+; CHECK-NEXT:    [[TMP2:%.*]] = shufflevector <2 x i8> [[TMP3]], <2 x i8> poison, <2 x i32> zeroinitializer
+; CHECK-NEXT:    [[V:%.*]] = add <2 x i8> [[TMP2]], <i8 0, i8 5>
 ; CHECK-NEXT:    [[DIV0:%.*]] = sdiv <2 x i8> splat (i8 -128), [[V]]
 ; CHECK-NEXT:    [[TMP1:%.*]] = xor i8 [[INP]], 123
 ; CHECK-NEXT:    [[R:%.*]] = insertelement <2 x i8> [[DIV0]], i8 [[TMP1]], i64 0
diff --git a/llvm/test/Transforms/StructurizeCFG/hoist-zerocost.ll b/llvm/test/Transforms/StructurizeCFG/hoist-zerocost.ll
new file mode 100644
index 0000000..10d4fa2
--- /dev/null
+++ b/llvm/test/Transforms/StructurizeCFG/hoist-zerocost.ll
@@ -0,0 +1,161 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 5
+; RUN: opt -S -passes=structurizecfg < %s | FileCheck %s
+
+
+%pair = type { i32, i32 }
+define void @test_if_then_else(ptr %ptr, i1 %cond) {
+; CHECK-LABEL: define void @test_if_then_else(
+; CHECK-SAME: ptr [[PTR:%.*]], i1 [[COND:%.*]]) {
+; CHECK-NEXT:  [[ENTRY:.*]]:
+; CHECK-NEXT:    [[COND_INV:%.*]] = xor i1 [[COND]], true
+; CHECK-NEXT:    [[LOAD_THEN:%.*]] = load [[PAIR:%.*]], ptr [[PTR]], align 4
+; CHECK-NEXT:    [[A_THEN:%.*]] = extractvalue [[PAIR]] [[LOAD_THEN]], 0
+; CHECK-NEXT:    br i1 [[COND_INV]], label %[[ELSE:.*]], label %[[FLOW:.*]]
+; CHECK:       [[FLOW]]:
+; CHECK-NEXT:    [[TMP0:%.*]] = phi i32 [ [[SUM_ELSE:%.*]], %[[ELSE]] ], [ [[A_THEN]], %[[ENTRY]] ]
+; CHECK-NEXT:    [[TMP1:%.*]] = phi i1 [ false, %[[ELSE]] ], [ true, %[[ENTRY]] ]
+; CHECK-NEXT:    br i1 [[TMP1]], label %[[THEN:.*]], label %[[MERGE:.*]]
+; CHECK:       [[THEN]]:
+; CHECK-NEXT:    br label %[[MERGE]]
+; CHECK:       [[ELSE]]:
+; CHECK-NEXT:    [[A_ELSE:%.*]] = extractvalue [[PAIR]] [[LOAD_THEN]], 0
+; CHECK-NEXT:    [[SUM_ELSE]] = add i32 [[A_ELSE]], 1
+; CHECK-NEXT:    br label %[[FLOW]]
+; CHECK:       [[MERGE]]:
+; CHECK-NEXT:    store i32 [[TMP0]], ptr [[PTR]], align 4
+; CHECK-NEXT:    ret void
+;
+entry:
+  %load_then = load %pair, ptr %ptr
+  br i1 %cond, label %then, label %else
+
+then:
+  %a_then = extractvalue %pair %load_then, 0
+  br label %merge
+
+else:
+  %a_else = extractvalue %pair %load_then, 0
+  %sum_else = add i32 %a_else, 1
+  br label %merge
+
+merge:
+  %phi = phi i32  [ %a_then, %then ], [ %sum_else, %else ]
+  store i32 %phi, ptr  %ptr
+  ret void
+}
+
+define void @test_if_else_then(ptr %ptr, i1 %cond) {
+; CHECK-LABEL: define void @test_if_else_then(
+; CHECK-SAME: ptr [[PTR:%.*]], i1 [[COND:%.*]]) {
+; CHECK-NEXT:  [[ENTRY:.*]]:
+; CHECK-NEXT:    [[COND_INV:%.*]] = xor i1 [[COND]], true
+; CHECK-NEXT:    [[LOAD_THEN:%.*]] = load [[PAIR:%.*]], ptr [[PTR]], align 4
+; CHECK-NEXT:    br i1 [[COND_INV]], label %[[THEN:.*]], label %[[FLOW:.*]]
+; CHECK:       [[THEN]]:
+; CHECK-NEXT:    [[A_THEN:%.*]] = extractvalue [[PAIR]] [[LOAD_THEN]], 0
+; CHECK-NEXT:    br label %[[FLOW]]
+; CHECK:       [[FLOW]]:
+; CHECK-NEXT:    [[TMP0:%.*]] = phi i32 [ [[A_THEN]], %[[THEN]] ], [ poison, %[[ENTRY]] ]
+; CHECK-NEXT:    [[TMP1:%.*]] = phi i1 [ false, %[[THEN]] ], [ true, %[[ENTRY]] ]
+; CHECK-NEXT:    br i1 [[TMP1]], label %[[ELSE:.*]], label %[[MERGE:.*]]
+; CHECK:       [[ELSE]]:
+; CHECK-NEXT:    [[A_ELSE:%.*]] = extractvalue [[PAIR]] [[LOAD_THEN]], 0
+; CHECK-NEXT:    [[SUM_ELSE:%.*]] = add i32 [[A_ELSE]], 1
+; CHECK-NEXT:    br label %[[MERGE]]
+; CHECK:       [[MERGE]]:
+; CHECK-NEXT:    [[PHI:%.*]] = phi i32 [ [[TMP0]], %[[FLOW]] ], [ [[SUM_ELSE]], %[[ELSE]] ]
+; CHECK-NEXT:    store i32 [[PHI]], ptr [[PTR]], align 4
+; CHECK-NEXT:    ret void
+;
+entry:
+  %load_then = load %pair, ptr %ptr
+  br i1 %cond, label %else, label %then
+
+then:
+  %a_then = extractvalue %pair %load_then, 0
+  br label %merge
+
+else:
+  %a_else = extractvalue %pair %load_then, 0
+  %sum_else = add i32 %a_else, 1
+  br label %merge
+
+merge:
+  %phi = phi i32  [ %a_then, %then ], [ %sum_else, %else ]
+  store i32 %phi, ptr  %ptr
+  ret void
+}
+
+define amdgpu_kernel void @test_loop_with_if( ptr %ptr, i1 %cond) #0 {
+; CHECK-LABEL: define amdgpu_kernel void @test_loop_with_if(
+; CHECK-SAME: ptr [[PTR:%.*]], i1 [[COND:%.*]]) {
+; CHECK-NEXT:  [[ENTRY:.*]]:
+; CHECK-NEXT:    [[COND_INV:%.*]] = xor i1 [[COND]], true
+; CHECK-NEXT:    [[I:%.*]] = tail call i32 @llvm.amdgcn.workitem.id.x()
+; CHECK-NEXT:    br label %[[LOOP:.*]]
+; CHECK:       [[LOOP]]:
+; CHECK-NEXT:    [[I3:%.*]] = phi i32 [ 0, %[[ENTRY]] ], [ [[I15:%.*]], %[[LATCH:.*]] ]
+; CHECK-NEXT:    [[LOAD:%.*]] = load [[PAIR:%.*]], ptr [[PTR]], align 4
+; CHECK-NEXT:    [[A_THEN:%.*]] = extractvalue [[PAIR]] [[LOAD]], 0
+; CHECK-NEXT:    br i1 [[COND]], label %[[IF:.*]], label %[[FLOW:.*]]
+; CHECK:       [[IF]]:
+; CHECK-NEXT:    [[I9:%.*]] = icmp sle i32 [[I3]], 10
+; CHECK-NEXT:    br label %[[FLOW]]
+; CHECK:       [[FLOW1:.*]]:
+; CHECK-NEXT:    [[TMP0:%.*]] = phi i32 [ [[Y:%.*]], %[[ELSE:.*]] ], [ [[A_THEN]], %[[FLOW]] ]
+; CHECK-NEXT:    [[TMP1:%.*]] = phi i1 [ false, %[[ELSE]] ], [ [[TMP2:%.*]], %[[FLOW]] ]
+; CHECK-NEXT:    br i1 [[TMP1]], label %[[THEN:.*]], label %[[LATCH]]
+; CHECK:       [[THEN]]:
+; CHECK-NEXT:    store i32 [[I]], ptr [[PTR]], align 4
+; CHECK-NEXT:    br label %[[LATCH]]
+; CHECK:       [[FLOW]]:
+; CHECK-NEXT:    [[TMP2]] = phi i1 [ true, %[[IF]] ], [ false, %[[LOOP]] ]
+; CHECK-NEXT:    [[TMP3:%.*]] = phi i1 [ [[I9]], %[[IF]] ], [ [[COND_INV]], %[[LOOP]] ]
+; CHECK-NEXT:    br i1 [[TMP3]], label %[[ELSE]], label %[[FLOW1]]
+; CHECK:       [[ELSE]]:
+; CHECK-NEXT:    [[I2:%.*]] = extractvalue [[PAIR]] [[LOAD]], 1
+; CHECK-NEXT:    [[A_ELSE:%.*]] = extractvalue [[PAIR]] [[LOAD]], 0
+; CHECK-NEXT:    [[Y]] = add i32 [[A_ELSE]], [[I2]]
+; CHECK-NEXT:    br label %[[FLOW1]]
+; CHECK:       [[LATCH]]:
+; CHECK-NEXT:    store i32 [[TMP0]], ptr [[PTR]], align 4
+; CHECK-NEXT:    [[I15]] = add nsw i32 [[TMP0]], 20
+; CHECK-NEXT:    [[I16:%.*]] = icmp sge i32 [[I15]], 255
+; CHECK-NEXT:    br i1 [[I16]], label %[[END:.*]], label %[[LOOP]]
+; CHECK:       [[END]]:
+; CHECK-NEXT:    ret void
+;
+entry:
+  %a = tail call i32 @llvm.amdgcn.workitem.id.x()
+  br label %loop
+
+loop:
+  %entry_phi = phi i32 [ 0, %entry ], [ %a15, %latch ]
+  %load = load %pair, ptr %ptr
+  br i1 %cond, label %if, label %else
+
+if:
+  %cmp = icmp sgt i32 %entry_phi, 10
+  br i1 %cmp, label %then, label %else
+
+then:
+  %a_then = extractvalue %pair %load, 0
+  store i32 %a, ptr %ptr, align 4
+  br label %latch
+
+else:
+  %a2 = extractvalue %pair %load, 1
+  %y = extractvalue %pair %load, 0
+  %a_else = add i32 %y, %a2
+  br label %latch
+
+latch:
+  %a_test = phi i32 [ %a_then, %then ], [ %a_else, %else ]
+  store i32 %a_test, ptr  %ptr
+  %a15 = add nsw i32 %a_test, 20
+  %a16 = icmp slt i32  %a15, 255
+  br i1 %a16, label %loop, label %end
+
+end:
+  ret void
+}
diff --git a/llvm/test/lit.site.cfg.py.in b/llvm/test/lit.site.cfg.py.in
index ee76beb..893e2cb 100644
--- a/llvm/test/lit.site.cfg.py.in
+++ b/llvm/test/lit.site.cfg.py.in
@@ -55,6 +55,7 @@ config.linked_bye_extension = @LLVM_BYE_LINK_INTO_TOOLS@
 config.linked_exampleirtransforms_extension = @LLVM_EXAMPLEIRTRANSFORMS_LINK_INTO_TOOLS@
 config.have_tf_aot = @LLVM_HAVE_TF_AOT@
 config.have_tflite = @LLVM_HAVE_TFLITE@
+config.enable_profcheck = @LLVM_ENABLE_PROFCHECK@
 config.llvm_inliner_model_autogenerated = @LLVM_INLINER_MODEL_AUTOGENERATED@
 config.llvm_raevict_model_autogenerated = @LLVM_RAEVICT_MODEL_AUTOGENERATED@
 config.expensive_checks = @LLVM_ENABLE_EXPENSIVE_CHECKS@
diff --git a/llvm/test/tools/llvm-readtapi/many-targets.test b/llvm/test/tools/llvm-readtapi/many-targets.test
new file mode 100644
index 0000000..efb44b5
--- /dev/null
+++ b/llvm/test/tools/llvm-readtapi/many-targets.test
@@ -0,0 +1,20 @@
+; RUN: rm -rf %t
+; RUN: split-file %s %t
+;
+; RUN: llvm-readtapi %t/many-targets.tbd
+;
+; Check that tbds containing symbols with many targets parse correctly (and in
+; particular parse without leaks).
+
+;--- many-targets.tbd
+--- !tapi-tbd
+tbd-version:     4
+targets:         [ x86_64-macos, x86_64-maccatalyst, arm64-macos, arm64-maccatalyst,
+                   arm64e-macos, arm64e-maccatalyst, arm64-ios, arm64e-ios ]
+install-name:    '/usr/lib/foo.dylib'
+current-version: 1
+exports:
+  - targets:         [ x86_64-macos, x86_64-maccatalyst, arm64-macos, arm64-maccatalyst,
+                       arm64e-macos, arm64e-maccatalyst, arm64-ios, arm64e-ios ]
+    symbols:         [ 'foo' ]
+...
diff --git a/llvm/tools/dsymutil/MachOUtils.cpp b/llvm/tools/dsymutil/MachOUtils.cpp
index be1934f..362a999 100644
--- a/llvm/tools/dsymutil/MachOUtils.cpp
+++ b/llvm/tools/dsymutil/MachOUtils.cpp
@@ -331,7 +331,7 @@ static bool createDwarfSegment(const MCAssembler& Asm,uint64_t VMAddr, uint64_t
                                  /* InitProt =*/3);
 
   for (unsigned int i = 0, n = Writer.getSectionOrder().size(); i != n; ++i) {
-    MCSection *Sec = Writer.getSectionOrder()[i];
+    auto *Sec = static_cast<MCSectionMachO *>(Writer.getSectionOrder()[i]);
     if (!Asm.getSectionFileSize(*Sec))
       continue;
 
diff --git a/llvm/tools/opt/NewPMDriver.cpp b/llvm/tools/opt/NewPMDriver.cpp
index 7d168a6..b9b8929 100644
--- a/llvm/tools/opt/NewPMDriver.cpp
+++ b/llvm/tools/opt/NewPMDriver.cpp
@@ -40,6 +40,7 @@
 #include "llvm/Transforms/Instrumentation/AddressSanitizer.h"
 #include "llvm/Transforms/Scalar/LoopPassManager.h"
 #include "llvm/Transforms/Utils/Debugify.h"
+#include "llvm/Transforms/Utils/ProfileVerify.h"
 
 using namespace llvm;
 using namespace opt_tool;
@@ -356,7 +357,7 @@ bool llvm::runPassPipeline(
     OutputKind OK, VerifierKind VK, bool ShouldPreserveAssemblyUseListOrder,
     bool ShouldPreserveBitcodeUseListOrder, bool EmitSummaryIndex,
     bool EmitModuleHash, bool EnableDebugify, bool VerifyDIPreserve,
-    bool UnifiedLTO) {
+    bool EnableProfcheck, bool UnifiedLTO) {
   auto FS = vfs::getRealFileSystem();
   std::optional<PGOOptions> P;
   switch (PGOKindFlag) {
@@ -487,7 +488,8 @@ bool llvm::runPassPipeline(
   if (VerifyDIPreserve)
     MPM.addPass(NewPMDebugifyPass(DebugifyMode::OriginalDebugInfo, "",
                                   &DebugInfoBeforePass));
-
+  if (EnableProfcheck)
+    MPM.addPass(createModuleToFunctionPassAdaptor(ProfileInjectorPass()));
   // Add passes according to the -passes options.
   if (!PassPipeline.empty()) {
     if (auto Err = PB.parsePassPipeline(MPM, PassPipeline)) {
@@ -504,6 +506,8 @@ bool llvm::runPassPipeline(
     MPM.addPass(NewPMCheckDebugifyPass(
         false, "", nullptr, DebugifyMode::OriginalDebugInfo,
         &DebugInfoBeforePass, VerifyDIPreserveExport));
+  if (EnableProfcheck)
+    MPM.addPass(createModuleToFunctionPassAdaptor(ProfileVerifierPass()));
 
   // Add any relevant output pass at the end of the pipeline.
   switch (OK) {
diff --git a/llvm/tools/opt/NewPMDriver.h b/llvm/tools/opt/NewPMDriver.h
index 2daae57..6c21d6c 100644
--- a/llvm/tools/opt/NewPMDriver.h
+++ b/llvm/tools/opt/NewPMDriver.h
@@ -75,7 +75,7 @@ bool runPassPipeline(
     bool ShouldPreserveAssemblyUseListOrder,
     bool ShouldPreserveBitcodeUseListOrder, bool EmitSummaryIndex,
     bool EmitModuleHash, bool EnableDebugify, bool VerifyDIPreserve,
-    bool UnifiedLTO = false);
+    bool EnableProfcheck, bool UnifiedLTO = false);
 } // namespace llvm
 
 #endif
diff --git a/llvm/tools/opt/optdriver.cpp b/llvm/tools/opt/optdriver.cpp
index 892b63b..4a3b058 100644
--- a/llvm/tools/opt/optdriver.cpp
+++ b/llvm/tools/opt/optdriver.cpp
@@ -217,6 +217,15 @@ static cl::opt<bool> VerifyDebugInfoPreserve(
     cl::desc("Start the pipeline with collecting and end it with checking of "
              "debug info preservation."));
 
+static cl::opt<bool> EnableProfileVerification(
+    "enable-profcheck",
+#if defined(LLVM_ENABLE_PROFCHECK)
+    cl::init(true),
+#else
+    cl::init(false),
+#endif
+    cl::desc("Start the pipeline with prof-inject and end it with prof-check"));
+
 static cl::opt<std::string> ClDataLayout("data-layout",
                                          cl::desc("data layout string to use"),
                                          cl::value_desc("layout-string"),
@@ -746,7 +755,8 @@ extern "C" int optMain(
                RemarksFile.get(), Pipeline, PluginList, PassBuilderCallbacks,
                OK, VK, PreserveAssemblyUseListOrder,
                PreserveBitcodeUseListOrder, EmitSummaryIndex, EmitModuleHash,
-               EnableDebugify, VerifyDebugInfoPreserve, UnifiedLTO)
+               EnableDebugify, VerifyDebugInfoPreserve,
+               EnableProfileVerification, UnifiedLTO)
                ? 0
                : 1;
   }
diff --git a/llvm/unittests/CodeGen/DroppedVariableStatsMIRTest.cpp b/llvm/unittests/CodeGen/DroppedVariableStatsMIRTest.cpp
index 36504f5..e72b4e4 100644
--- a/llvm/unittests/CodeGen/DroppedVariableStatsMIRTest.cpp
+++ b/llvm/unittests/CodeGen/DroppedVariableStatsMIRTest.cpp
@@ -39,8 +39,7 @@ createTargetMachine(std::string TargetStr, StringRef CPU, StringRef FS) {
     return nullptr;
   TargetOptions Options;
   return std::unique_ptr<TargetMachine>(
-      static_cast<TargetMachine *>(T->createTargetMachine(
-          TT, CPU, FS, Options, std::nullopt, std::nullopt)));
+      T->createTargetMachine(TT, CPU, FS, Options, std::nullopt, std::nullopt));
 }
 
 std::unique_ptr<Module> parseMIR(const TargetMachine &TM, StringRef MIRCode,
diff --git a/llvm/unittests/Support/DebugLogTest.cpp b/llvm/unittests/Support/DebugLogTest.cpp
index 22dffd8..c34d888 100644
--- a/llvm/unittests/Support/DebugLogTest.cpp
+++ b/llvm/unittests/Support/DebugLogTest.cpp
@@ -27,7 +27,7 @@ TEST(DebugLogTest, Basic) {
     std::string str;
     raw_string_ostream os(str);
     DEBUGLOG_WITH_STREAM_AND_TYPE(os, nullptr) << "NoType";
-    EXPECT_TRUE(StringRef(os.str()).starts_with('['));
+    EXPECT_FALSE(StringRef(os.str()).starts_with('['));
     EXPECT_TRUE(StringRef(os.str()).ends_with("NoType\n"));
   }
 
@@ -37,6 +37,7 @@ TEST(DebugLogTest, Basic) {
     raw_string_ostream os(str);
     DEBUGLOG_WITH_STREAM_AND_TYPE(os, "A") << "A";
     DEBUGLOG_WITH_STREAM_AND_TYPE(os, "B") << "B";
+    EXPECT_TRUE(StringRef(os.str()).starts_with('['));
     EXPECT_THAT(os.str(), AllOf(HasSubstr("A\n"), HasSubstr("B\n")));
   }
 
@@ -61,6 +62,28 @@ TEST(DebugLogTest, Basic) {
     EXPECT_THAT(count, Eq(1));
   }
 }
+
+TEST(DebugLogTest, StreamPrefix) {
+  llvm::DebugFlag = true;
+  static const char *DT[] = {"A", "B"};
+  setCurrentDebugTypes(DT, 2);
+
+  std::string str;
+  raw_string_ostream os(str);
+  std::string expected = "PrefixA 1\nPrefixA 2\nPrefixA \nPrefixB "
+                         "3\nPrefixB 4\nPrefixA 5";
+  {
+    llvm::impl::raw_ldbg_ostream ldbg_osB("PrefixB ", os);
+    llvm::impl::raw_ldbg_ostream ldbg_osA("PrefixA ", os);
+    ldbg_osA << "1\n2";
+    ldbg_osA << "\n\n";
+    ldbg_osB << "3\n4\n";
+    ldbg_osA << "5";
+    EXPECT_EQ(os.str(), expected);
+  }
+  // After destructors, there was a pending newline for stream B.
+  EXPECT_EQ(os.str(), expected + "\nPrefixB \n");
+}
 #else
 TEST(DebugLogTest, Basic) {
   // LDBG should be compiled out in NDEBUG, so just check it compiles and has
diff --git a/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp b/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp
index 8012d9e..94b74d2 100644
--- a/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp
+++ b/llvm/unittests/Transforms/Vectorize/VPlanTest.cpp
@@ -1075,7 +1075,7 @@ TEST_F(VPRecipeTest, CastVPBlendRecipeToVPUser) {
   Args.push_back(I1);
   Args.push_back(I2);
   Args.push_back(M2);
-  VPBlendRecipe Recipe(Phi, Args);
+  VPBlendRecipe Recipe(Phi, Args, {});
   EXPECT_TRUE(isa<VPUser>(&Recipe));
   VPRecipeBase *BaseR = &Recipe;
   EXPECT_TRUE(isa<VPUser>(BaseR));
diff --git a/llvm/unittests/Transforms/Vectorize/VPlanVerifierTest.cpp b/llvm/unittests/Transforms/Vectorize/VPlanVerifierTest.cpp
index 5394472..6214ea3 100644
--- a/llvm/unittests/Transforms/Vectorize/VPlanVerifierTest.cpp
+++ b/llvm/unittests/Transforms/Vectorize/VPlanVerifierTest.cpp
@@ -103,7 +103,7 @@ TEST_F(VPVerifierTest, VPBlendUseBeforeDefDifferentBB) {
   auto *CanIV = new VPCanonicalIVPHIRecipe(Zero, {});
   VPInstruction *BranchOnCond =
       new VPInstruction(VPInstruction::BranchOnCond, {CanIV});
-  auto *Blend = new VPBlendRecipe(Phi, {DefI});
+  auto *Blend = new VPBlendRecipe(Phi, {DefI}, {});
 
   VPBasicBlock *VPBB1 = Plan.getEntry();
   VPBasicBlock *VPBB2 = Plan.createVPBasicBlock("");
diff --git a/llvm/utils/TableGen/Basic/RuntimeLibcallsEmitter.cpp b/llvm/utils/TableGen/Basic/RuntimeLibcallsEmitter.cpp
index 7f90d6b..a280604 100644
--- a/llvm/utils/TableGen/Basic/RuntimeLibcallsEmitter.cpp
+++ b/llvm/utils/TableGen/Basic/RuntimeLibcallsEmitter.cpp
@@ -242,11 +242,10 @@ public:
     SmallVector<const Record *, 1024> AllRuntimeLibcallImpls(
         AllRuntimeLibcallImplsRaw);
 
-    // Sort by libcall impl name, not the enum name. This keeps the order
-    // suitable for using the name table for libcall recognition binary search.
-    llvm::sort(AllRuntimeLibcallImpls, [](const Record *A, const Record *B) {
-      return A->getValueAsString("LibCallFuncName") <
-             B->getValueAsString("LibCallFuncName");
+    // Sort by libcall impl name and secondarily by the enum name.
+    sort(AllRuntimeLibcallImpls, [](const Record *A, const Record *B) {
+      return std::pair(A->getValueAsString("LibCallFuncName"), A->getName()) <
+             std::pair(B->getValueAsString("LibCallFuncName"), B->getName());
     });
 
     RuntimeLibcallImplDefList.reserve(AllRuntimeLibcallImpls.size());
diff --git a/llvm/utils/TableGen/CompressInstEmitter.cpp b/llvm/utils/TableGen/CompressInstEmitter.cpp
index 331cea1..af119c3 100644
--- a/llvm/utils/TableGen/CompressInstEmitter.cpp
+++ b/llvm/utils/TableGen/CompressInstEmitter.cpp
@@ -86,16 +86,22 @@ namespace {
 class CompressInstEmitter {
   struct OpData {
     enum MapKind { Operand, Imm, Reg } Kind;
-    union {
+    // Info for an operand.
+    struct OpndInfo {
+      // Record from the Dag.
+      const Record *DagRec;
       // Operand number mapped to.
-      unsigned OpNo;
+      unsigned Idx;
+      // Tied operand index within the instruction.
+      int TiedOpIdx;
+    };
+    union {
+      OpndInfo OpInfo;
       // Integer immediate value.
       int64_t ImmVal;
       // Physical register.
       const Record *RegRec;
     };
-    // Tied operand index within the instruction.
-    int TiedOpIdx = -1;
   };
   struct ArgData {
     unsigned DAGOpNo;
@@ -273,6 +279,31 @@ void CompressInstEmitter::addDagOperandMapping(const Record *Rec,
                               "' in the corresponding instruction operand!");
 
         OperandMap[OpNo].Kind = OpData::Operand;
+        OperandMap[OpNo].OpInfo.DagRec = DI->getDef();
+        OperandMap[OpNo].OpInfo.TiedOpIdx = -1;
+
+        // Create a mapping between the operand name in the Dag (e.g. $rs1) and
+        // its index in the list of Dag operands and check that operands with
+        // the same name have the same type. For example in 'C_ADD $rs1, $rs2'
+        // we generate the mapping $rs1 --> 0, $rs2 ---> 1. If the operand
+        // appears twice in the same Dag (tied in the compressed instruction),
+        // we note the previous index in the TiedOpIdx field.
+        StringRef ArgName = Dag->getArgNameStr(DAGOpNo);
+        if (ArgName.empty())
+          continue;
+
+        if (IsSourceInst) {
+          auto It = Operands.find(ArgName);
+          if (It != Operands.end()) {
+            OperandMap[OpNo].OpInfo.TiedOpIdx = It->getValue().MIOpNo;
+            if (OperandMap[It->getValue().MIOpNo].OpInfo.DagRec != DI->getDef())
+              PrintFatalError(Rec->getLoc(),
+                              "Input Operand '" + ArgName +
+                                  "' has a mismatched tied operand!");
+          }
+        }
+
+        Operands[ArgName] = {DAGOpNo, OpNo};
       } else if (const auto *II = dyn_cast<IntInit>(Dag->getArg(DAGOpNo))) {
         // Validate that corresponding instruction operand expects an immediate.
         if (!OpndRec->isSubClassOf("Operand"))
@@ -292,30 +323,6 @@ void CompressInstEmitter::addDagOperandMapping(const Record *Rec,
       } else {
         llvm_unreachable("Unhandled CompressPat argument type!");
       }
-
-      // Create a mapping between the operand name in the Dag (e.g. $rs1) and
-      // its index in the list of Dag operands and check that operands with the
-      // same name have the same type. For example in 'C_ADD $rs1, $rs2' we
-      // generate the mapping $rs1 --> 0, $rs2 ---> 1. If the operand appears
-      // twice in the same Dag (tied in the compressed instruction), we note
-      // the previous index in the TiedOpIdx field.
-      StringRef ArgName = Dag->getArgNameStr(DAGOpNo);
-      if (ArgName.empty())
-        continue;
-
-      if (IsSourceInst) {
-        auto It = Operands.find(ArgName);
-        if (It != Operands.end()) {
-          OperandMap[OpNo].TiedOpIdx = It->getValue().MIOpNo;
-          if (!validateArgsTypes(Dag->getArg(It->getValue().DAGOpNo),
-                                 Dag->getArg(DAGOpNo)))
-            PrintFatalError(Rec->getLoc(),
-                            "Input Operand '" + ArgName +
-                                "' has a mismatched tied operand!");
-        }
-      }
-
-      Operands[ArgName] = {DAGOpNo, OpNo};
     }
   }
 
@@ -372,8 +379,9 @@ void CompressInstEmitter::createInstOperandMapping(
       if (DestOperandMap[OpNo].Kind == OpData::Operand)
         // No need to fill the SourceOperandMap here since it was mapped to
         // destination operand 'TiedInstOpIdx' in a previous iteration.
-        LLVM_DEBUG(dbgs() << "    " << DestOperandMap[OpNo].OpNo << " ====> "
-                          << OpNo << "  Dest operand tied with operand '"
+        LLVM_DEBUG(dbgs() << "    " << DestOperandMap[OpNo].OpInfo.Idx
+                          << " ====> " << OpNo
+                          << "  Dest operand tied with operand '"
                           << TiedInstOpIdx << "'\n");
       ++OpNo;
       continue;
@@ -398,8 +406,8 @@ void CompressInstEmitter::createInstOperandMapping(
              "Incorrect operand mapping detected!\n");
 
       unsigned SourceOpNo = SourceOp->getValue().MIOpNo;
-      DestOperandMap[OpNo].OpNo = SourceOpNo;
-      SourceOperandMap[SourceOpNo].OpNo = OpNo;
+      DestOperandMap[OpNo].OpInfo.Idx = SourceOpNo;
+      SourceOperandMap[SourceOpNo].OpInfo.Idx = OpNo;
       LLVM_DEBUG(dbgs() << "    " << SourceOpNo << " ====> " << OpNo << "\n");
     }
   }
@@ -640,7 +648,7 @@ void CompressInstEmitter::emitCompressInstEmitter(raw_ostream &OS,
   StringRef PrevOp;
   StringRef CurOp;
   CaseStream << "  switch (MI.getOpcode()) {\n";
-  CaseStream << "    default: return false;\n";
+  CaseStream << "  default: return false;\n";
 
   bool CompressOrCheck =
       EType == EmitterType::Compress || EType == EmitterType::CheckCompress;
@@ -653,7 +661,7 @@ void CompressInstEmitter::emitCompressInstEmitter(raw_ostream &OS,
                 .str()
           : "";
 
-  for (auto &CompressPat : CompressPatterns) {
+  for (const auto &CompressPat : CompressPatterns) {
     if (EType == EmitterType::Uncompress && CompressPat.IsCompressOnly)
       continue;
 
@@ -661,23 +669,25 @@ void CompressInstEmitter::emitCompressInstEmitter(raw_ostream &OS,
     std::string CodeString;
     raw_string_ostream CondStream(CondString);
     raw_string_ostream CodeStream(CodeString);
-    CodeGenInstruction &Source =
+    const CodeGenInstruction &Source =
         CompressOrCheck ? CompressPat.Source : CompressPat.Dest;
-    CodeGenInstruction &Dest =
+    const CodeGenInstruction &Dest =
         CompressOrCheck ? CompressPat.Dest : CompressPat.Source;
-    IndexedMap<OpData> SourceOperandMap = CompressOrCheck
-                                              ? CompressPat.SourceOperandMap
-                                              : CompressPat.DestOperandMap;
-    IndexedMap<OpData> &DestOperandMap = CompressOrCheck
-                                             ? CompressPat.DestOperandMap
-                                             : CompressPat.SourceOperandMap;
+    const IndexedMap<OpData> &SourceOperandMap =
+        CompressOrCheck ? CompressPat.SourceOperandMap
+                        : CompressPat.DestOperandMap;
+    const IndexedMap<OpData> &DestOperandMap =
+        CompressOrCheck ? CompressPat.DestOperandMap
+                        : CompressPat.SourceOperandMap;
 
     CurOp = Source.TheDef->getName();
     // Check current and previous opcode to decide to continue or end a case.
     if (CurOp != PrevOp) {
-      if (!PrevOp.empty())
-        CaseStream.indent(6) << "break;\n    } // case " + PrevOp + "\n";
-      CaseStream.indent(4) << "case " + TargetName + "::" + CurOp + ": {\n";
+      if (!PrevOp.empty()) {
+        CaseStream.indent(4) << "break;\n";
+        CaseStream.indent(2) << "} // case " + PrevOp + "\n";
+      }
+      CaseStream.indent(2) << "case " + TargetName + "::" + CurOp + ": {\n";
     }
 
     std::set<std::pair<bool, StringRef>> FeaturesSet;
@@ -719,21 +729,24 @@ void CompressInstEmitter::emitCompressInstEmitter(raw_ostream &OS,
     // Start Source Inst operands validation.
     unsigned OpNo = 0;
     for (const auto &SourceOperand : Source.Operands) {
-      if (SourceOperandMap[OpNo].TiedOpIdx != -1) {
-        if (Source.Operands[OpNo].Rec->isSubClassOf("RegisterClass"))
-          CondStream.indent(8)
-              << "(MI.getOperand(" << OpNo << ").isReg()) && (MI.getOperand("
-              << SourceOperandMap[OpNo].TiedOpIdx << ").isReg()) &&\n"
-              << indent(8) << "(MI.getOperand(" << OpNo
-              << ").getReg() ==  MI.getOperand("
-              << SourceOperandMap[OpNo].TiedOpIdx << ").getReg()) &&\n";
-        else
-          PrintFatalError("Unexpected tied operand types!");
-      }
       for (unsigned SubOp = 0; SubOp != SourceOperand.MINumOperands; ++SubOp) {
         // Check for fixed immediates\registers in the source instruction.
         switch (SourceOperandMap[OpNo].Kind) {
         case OpData::Operand:
+          if (SourceOperandMap[OpNo].OpInfo.TiedOpIdx != -1) {
+            if (Source.Operands[OpNo].Rec->isSubClassOf("RegisterClass"))
+              CondStream.indent(8) << "(MI.getOperand(" << OpNo
+                                   << ").isReg()) && (MI.getOperand("
+                                   << SourceOperandMap[OpNo].OpInfo.TiedOpIdx
+                                   << ").isReg()) &&\n"
+                                   << indent(8) << "(MI.getOperand(" << OpNo
+                                   << ").getReg() ==  MI.getOperand("
+                                   << SourceOperandMap[OpNo].OpInfo.TiedOpIdx
+                                   << ").getReg()) &&\n";
+            else
+              PrintFatalError("Unexpected tied operand types!");
+          }
+
           // We don't need to do anything for source instruction operand checks.
           break;
         case OpData::Imm:
@@ -771,7 +784,8 @@ void CompressInstEmitter::emitCompressInstEmitter(raw_ostream &OS,
 
         switch (DestOperandMap[OpNo].Kind) {
         case OpData::Operand: {
-          unsigned OpIdx = DestOperandMap[OpNo].OpNo;
+          unsigned OpIdx = DestOperandMap[OpNo].OpInfo.Idx;
+          DestRec = DestOperandMap[OpNo].OpInfo.DagRec;
           // Check that the operand in the Source instruction fits
           // the type for the Dest instruction.
           if (DestRec->isSubClassOf("RegisterClass") ||
@@ -781,7 +795,7 @@ void CompressInstEmitter::emitCompressInstEmitter(raw_ostream &OS,
                                  : DestRec->getValueAsDef("RegClass");
             // This is a register operand. Check the register class.
             // Don't check register class if this is a tied operand, it was done
-            // for the operand its tied to.
+            // for the operand it's tied to.
             if (DestOperand.getTiedRegister() == -1) {
               CondStream.indent(8) << "MI.getOperand(" << OpIdx << ").isReg()";
               if (EType == EmitterType::CheckCompress)
@@ -861,9 +875,10 @@ void CompressInstEmitter::emitCompressInstEmitter(raw_ostream &OS,
     mergeCondAndCode(CaseStream, CondString, CodeString);
     PrevOp = CurOp;
   }
-  Func << CaseString << "\n";
+  Func << CaseString;
+  Func.indent(4) << "break;\n";
   // Close brace for the last case.
-  Func.indent(4) << "} // case " << CurOp << "\n";
+  Func.indent(2) << "} // case " << CurOp << "\n";
   Func.indent(2) << "} // switch\n";
   Func.indent(2) << "return false;\n}\n";
 
diff --git a/llvm/utils/TableGen/InstrInfoEmitter.cpp b/llvm/utils/TableGen/InstrInfoEmitter.cpp
index f028fcd..fa38d01 100644
--- a/llvm/utils/TableGen/InstrInfoEmitter.cpp
+++ b/llvm/utils/TableGen/InstrInfoEmitter.cpp
@@ -248,9 +248,7 @@ void InstrInfoEmitter::emitOperandNameMappings(
   /// scan of the instructions below.
 
   // Map of operand names to their ID.
-  std::map<StringRef, unsigned> OperandNameToID;
-  // Map from operand name enum value -> ID.
-  std::vector<unsigned> OperandEnumToID;
+  MapVector<StringRef, unsigned> OperandNameToID;
 
   /// The keys of this map is a map which have OpName ID values as their keys
   /// and instruction operand indices as their values. The values of this map
@@ -278,16 +276,13 @@ void InstrInfoEmitter::emitOperandNameMappings(
   }
 
   const size_t NumOperandNames = OperandNameToID.size();
-  OperandEnumToID.reserve(NumOperandNames);
-  for (const auto &Op : OperandNameToID)
-    OperandEnumToID.push_back(Op.second);
 
   OS << "#ifdef GET_INSTRINFO_OPERAND_ENUM\n";
   OS << "#undef GET_INSTRINFO_OPERAND_ENUM\n";
   OS << "namespace llvm::" << Namespace << " {\n";
   OS << "enum class OpName {\n";
-  for (const auto &[I, Op] : enumerate(OperandNameToID))
-    OS << "  " << Op.first << " = " << I << ",\n";
+  for (const auto &[Op, I] : OperandNameToID)
+    OS << "  " << Op << " = " << I << ",\n";
   OS << "  NUM_OPERAND_NAMES = " << NumOperandNames << ",\n";
   OS << "}; // enum class OpName\n\n";
   OS << "LLVM_READONLY\n";
@@ -312,7 +307,7 @@ void InstrInfoEmitter::emitOperandNameMappings(
 
       // Emit a row of the OperandMap table.
       OS << "    {";
-      for (unsigned ID : OperandEnumToID) {
+      for (unsigned ID = 0; ID < NumOperandNames; ++ID) {
         auto Iter = OpList.find(ID);
         OS << (Iter != OpList.end() ? (int)Iter->second : -1) << ", ";
       }
diff --git a/llvm/utils/gn/build/write_vcsrevision.py b/llvm/utils/gn/build/write_vcsrevision.py
index afd6aae..3a627ee 100755
--- a/llvm/utils/gn/build/write_vcsrevision.py
+++ b/llvm/utils/gn/build/write_vcsrevision.py
@@ -6,22 +6,13 @@ import argparse
 import os
 import subprocess
 import sys
+import shutil
 
 
 THIS_DIR = os.path.abspath(os.path.dirname(__file__))
 LLVM_DIR = os.path.dirname(os.path.dirname(os.path.dirname(THIS_DIR)))
 
 
-def which(program):
-    # distutils.spawn.which() doesn't find .bat files,
-    # https://bugs.python.org/issue2200
-    for path in os.environ["PATH"].split(os.pathsep):
-        candidate = os.path.join(path, program)
-        if os.path.isfile(candidate) and os.access(candidate, os.X_OK):
-            return candidate
-    return None
-
-
 def main():
     parser = argparse.ArgumentParser(description=__doc__)
     parser.add_argument(
@@ -46,11 +37,11 @@ def main():
 
     vcsrevision_contents = ""
     if args.write_git_rev:
-        git, use_shell = which("git"), False
+        git, use_shell = shutil.which("git"), False
         if not git:
-            git = which("git.exe")
+            git = shutil.which("git.exe")
         if not git:
-            git, use_shell = which("git.bat"), True
+            git, use_shell = shutil.which("git.bat"), True
         git_dir = (
             subprocess.check_output(
                 [git, "rev-parse", "--git-dir"], cwd=LLVM_DIR, shell=use_shell
diff --git a/llvm/utils/gn/secondary/llvm/include/llvm/Config/BUILD.gn b/llvm/utils/gn/secondary/llvm/include/llvm/Config/BUILD.gn
index b8f3b4f..499ded9 100644
--- a/llvm/utils/gn/secondary/llvm/include/llvm/Config/BUILD.gn
+++ b/llvm/utils/gn/secondary/llvm/include/llvm/Config/BUILD.gn
@@ -296,6 +296,7 @@ write_cmake_config("llvm-config") {
   input = "llvm-config.h.cmake"
   output = "$target_gen_dir/llvm-config.h"
   values = [
+    "LLVM_ENABLE_PROFCHECK=",
     "LLVM_BUILD_LLVM_DYLIB=",
     "LLVM_BUILD_SHARED_LIBS=",
     "LLVM_ENABLE_LLVM_C_EXPORT_ANNOTATIONS=",
diff --git a/llvm/utils/gn/secondary/llvm/lib/MC/BUILD.gn b/llvm/utils/gn/secondary/llvm/lib/MC/BUILD.gn
index 5c96bd8..eac2cd4 100644
--- a/llvm/utils/gn/secondary/llvm/lib/MC/BUILD.gn
+++ b/llvm/utils/gn/secondary/llvm/lib/MC/BUILD.gn
@@ -57,13 +57,7 @@ static_library("MC") {
     "MCSPIRVStreamer.cpp",
     "MCSchedule.cpp",
     "MCSection.cpp",
-    "MCSectionCOFF.cpp",
-    "MCSectionDXContainer.cpp",
-    "MCSectionELF.cpp",
-    "MCSectionGOFF.cpp",
     "MCSectionMachO.cpp",
-    "MCSectionWasm.cpp",
-    "MCSectionXCOFF.cpp",
     "MCStreamer.cpp",
     "MCSubtargetInfo.cpp",
     "MCSymbol.cpp",
diff --git a/llvm/utils/gn/secondary/llvm/lib/Transforms/ObjCARC/BUILD.gn b/llvm/utils/gn/secondary/llvm/lib/Transforms/ObjCARC/BUILD.gn
index e7b2084..d4ad915 100644
--- a/llvm/utils/gn/secondary/llvm/lib/Transforms/ObjCARC/BUILD.gn
+++ b/llvm/utils/gn/secondary/llvm/lib/Transforms/ObjCARC/BUILD.gn
@@ -9,7 +9,6 @@ static_library("ObjCARC") {
   sources = [
     "DependencyAnalysis.cpp",
     "ObjCARC.cpp",
-    "ObjCARCAPElim.cpp",
     "ObjCARCContract.cpp",
     "ObjCARCExpand.cpp",
     "ObjCARCOpts.cpp",
diff --git a/llvm/utils/gn/secondary/llvm/test/BUILD.gn b/llvm/utils/gn/secondary/llvm/test/BUILD.gn
index 7ed0d3c..08cddc1 100644
--- a/llvm/utils/gn/secondary/llvm/test/BUILD.gn
+++ b/llvm/utils/gn/secondary/llvm/test/BUILD.gn
@@ -64,6 +64,7 @@ write_lit_config("lit_site_cfg") {
     "LLVM_APPEND_VC_REV=0",
     "LLVM_ENABLE_FFI=0",
     "LLVM_ENABLE_HTTPLIB=0",
+    "LLVM_ENABLE_PROFCHECK=0",
     "LLVM_EXPERIMENTAL_KEY_INSTRUCTIONS=0",
     "LLVM_FORCE_VC_REVISION=",
     "LLVM_HAS_LOGF128=0",
diff --git a/llvm/utils/lit/lit/formats/base.py b/llvm/utils/lit/lit/formats/base.py
index 27f7c7e..db5c1c3 100644
--- a/llvm/utils/lit/lit/formats/base.py
+++ b/llvm/utils/lit/lit/formats/base.py
@@ -34,8 +34,7 @@ class FileBasedTest(TestFormat):
         if filename.startswith(".") or filename in localConfig.excludes:
             return
 
-        base, ext = os.path.splitext(filename)
-        if ext in localConfig.suffixes:
+        if any(filename.endswith(suffix) for suffix in localConfig.suffixes):
             yield lit.Test.Test(testSuite, path_in_suite, localConfig)
 
     def getTestsInDirectory(self, testSuite, path_in_suite, litConfig, localConfig):
diff --git a/mlir/Maintainers.md b/mlir/Maintainers.md
new file mode 100644
index 0000000..7c852ef
--- /dev/null
+++ b/mlir/Maintainers.md
@@ -0,0 +1,63 @@
+# MLIR Maintainers
+
+This file is a list of the
+[maintainers](https://llvm.org/docs/DeveloperPolicy.html#maintainers) for MLIR.
+
+The following people are the active maintainers for the project. For the sake of
+simplicity, responsibility areas are subdivided into broad categories, which are
+further subdivided into individual components, such as dialects. Please reach
+out to them for code reviews, questions about their area of expertise, or other
+assistance.
+
+## Core
+
+Core components of MLIR, including core IR, analyses and rewriters, fundamental
+dialects, build system and language bindings.
+
+- Alex Zinenko \
+  ftynse@gmail.com (email),
+  [@ftynse](https://github.com/ftynse) (GitHub),
+  ftynse (Discourse)
+- Jacques Pienaar \
+  jpienaar@google.com (email),
+  [@jpienaar](https://github.com/jpienaar) (GitHub),
+  jpienaar (Discourse)
+- Mehdi Amini \
+  joker.eph@gmail.com (email),
+  [@joker-eph](https://github.com/joker-eph) (GitHub),
+  mehdi_amini (Discourse)
+
+## Egress
+
+MLIR components pertaining to egress flows from MLIR, in particular to LLVM IR.
+
+- Matthias Springer \
+  me@m-sp.org (email),
+  [@matthias-springer](https://github.com/matthias-springer) (GitHub),
+  matthias-springer (Discourse)
+- Andrzej Warzynski \
+  andrzej.warzynski@arm.com (email),
+  [@banach-space](https://github.com/banach-space) (GitHub),
+  banach-space (Discourse)
+- Tobias Gysi \
+  tobias.gysi@nextsilicon.com (email),
+  [@gysit](https://github.com/gysit) (GitHub),
+  gysit (Discourse)
+
+## Tensor Compiler
+
+MLIR components specific to construction of compilers for tensor algebra, in
+particular for machine learning compilers.
+
+- Renato Golin \
+  rengolin@gmail.com (email),
+  [@rengolin](https://github.com/rengolin) (GitHub),
+  rengolin (Discourse)
+- Jacques Pienaar \
+  jpienaar@google.com (email),
+  [@jpienaar](https://github.com/jpienaar) (GitHub),
+  jpienaar (Discourse)
+- Andrzej Warzynski \
+  andrzej.warzynski@arm.com (email),
+  [@banach-space](https://github.com/banach-space) (GitHub),
+  banach-space (Discourse)
diff --git a/mlir/docs/Dialects/Mesh.md b/mlir/docs/Dialects/Mesh.md
deleted file mode 100644
index 5eb6569..0000000
--- a/mlir/docs/Dialects/Mesh.md
+++ /dev/null
@@ -1,74 +0,0 @@
-# 'mesh' Dialect
-
-The `mesh` dialect contains a set of attributes, operations and interfaces that
-are useful for representing sharding and communication on a device mesh
-cluster.
-
-[TOC]
-
-## Collective Communication Operations
-There are a number of operations in the Mesh dialect to facilitate
-communication between devices in a mesh.
-It is assumed that the user is familiar with collective operations.
-[Wikipedia](https://en.wikipedia.org/wiki/Collective_operation) has a good
-explanation.
-The main addition is that the collectives in this dialect have mesh
-semantics.
-
-### Device groups
-The operation attributes `mesh` and `mesh_axes` specifies a list of device mesh
-axes that partition the devices into disjoint groups.
-The collective operation is performed between devices in the same group.
-Devices that have the same coordinates outside of axes `mesh_axes` are in the
-same group.
-A group is described by its multi-index along the axes outside of `mesh_axes`.
-For example if we have a device mesh of size `2x3x4x5` and the partition mesh
-axes list is `[0, 1]` then devices are partitioned into the groups
-`{ { (i, j, k, m) | 0<=i<2, 0<=j<3 } | 0<=k<4, 0<=m<5 }`.
-The device groups would be `{ (k, m) | 0<=k<4, 0<=m<5 }`.
-Devices (1, 0, 2, 3) and (1, 1, 2, 3) will be in the same group.
-Device (1, 0, 2, 4) will be in another group.
-Some collective operations like all-to-all and all-gather care about the
-order of devices.
-The order of device in a device group is induced by the order of axes in
-`mesh_axes`.
-The axes are ordered from outer to inner.
-If we have an axis list `[3, 1]` then device `(i, 1, k, 0)` will precede
-both devices `(i, 0, k, 1)` and `(i, 2, k, 0)`.
-
-### In-group Device
-Some operations like `broadcast`, `scatter` and `send` specify devices in each
-device-group.
-These devices are represented with their multi-index over the mesh axes that
-are not constant within a device group.
-These are the axes specified by `mesh_axes` attribute.
-
-For Example on a 3D mesh an operation with `mesh_axes = [0, 2]` would specify
-an in-group device with `(i, j)`. Then for each group with index `g` on the
-second axis, the in-group device would be `(i, g, j)`.
-### Purity
-Collectives that involve the whole device group to perform a single operation
-are pure. The exceptions are `send` and `recv`.
-
-There is an assumption that the execution is SPMD.
-Not only that each process runs the same program, but that at the point of
-execution of a collective operation, all processes are in a coherent state.
-All compiler transformations must be consistent.
-Collective operations in the IR that may correspond to the same runtime
-collective operation must be transformed in a consistent manner.
-For example if a collective operation is optimized out, than it must also
-not appear in any path of execution on any process.
-
-Having the operations as `Pure` implies that if an interpreter is to execute
-the IR containing the `mesh` collectives, all processes would execute the same
-line when they reach a pure collective operation.
-This requirement stems from the need to be compatible with general optimization
-passes like dead code and common sub-expression elimination.
-
-## Operations
-
-[include "Dialects/MeshOps.md"]
-
-## Attributes
-
-[include "Dialects/MeshAttrs.md"]
diff --git a/mlir/docs/Dialects/Shard.md b/mlir/docs/Dialects/Shard.md
new file mode 100644
index 0000000..eb6ff61
--- /dev/null
+++ b/mlir/docs/Dialects/Shard.md
@@ -0,0 +1,92 @@
+# 'shard' Dialect
+
+The 'shard' dialect defines a set of attributes, operations, and interfaces for
+working with tensor sharding and device communication.
+
+It’s inspired by [GSPMD](*General and Scalable Parallelization for ML Computation Graphs*).
+
+Originally, the dialect was called `mesh`, but it was renamed to better reflect
+what it actually does.
+
+[TOC]
+
+## Collective Communication Operations
+
+The 'shard' dialect includes several collective operations that help coordinate
+communication between devices arranged in a grid.
+
+If you’re not already familiar with collective operations, [this Wikipedia
+article](https://en.wikipedia.org/wiki/Collective_operation) is a good starting
+point.
+
+Unlike traditional collectives that are defined in terms of message-passing
+between explicit buffers on each process, the collectives in this dialect work
+at a higher level. They’re defined in terms of how data moves across the
+dimensions of a tensor, and the participating processes are inferred from how
+the tensor is sharded - not specified manually.
+
+### Device Groups
+
+Each collective operation runs within a group of devices. You define groups
+using the `grid` and `grid_axes` attributes, which describe how to slice the
+full device grid into smaller groups.
+
+Devices that have the same coordinates *outside* the listed `grid_axes` belong
+to the same group.
+
+Example: Say your device grid is shaped `2×3×4×5`, and you set
+`grid_axes = [0, 1]`. This splits the grid into groups by fixing axes 2 and 3. You’d get groups like:
+
+```
+{ { (i, j, k, m) | 0 ≤ i < 2, 0 ≤ j < 3 } | 0 ≤ k < 4, 0 ≤ m < 5 }
+```
+
+So the groups are identified by the coordinates `(k, m)`, and devices like
+`(1, 0, 2, 3)` and `(1, 1, 2, 3)` are in the same group. But `(1, 0, 2, 4)`
+is in a different group.
+
+For some collectives (like `all-to-all`), the order of devices in the group
+matters. The device order is based on the order of axes in `grid_axes`, from
+outermost to innermost.
+
+Example: If `grid_axes = [3, 1]`, then device `(i, 1, k, 0)` comes before
+`(i, 0, k, 1)` and `(i, 2, k, 0)`.
+
+### In-group Devices
+
+Some operations (like `broadcast`, `scatter`, and `send`) refer to a specific
+device within each group. These in-group devices are identified using their
+coordinates over the axes listed in `grid_axes`.
+
+Example: In a 3D grid with `grid_axes = [0, 2]`, an in-group device is specified
+as `(i, j)`. If a group is fixed at coordinate `g` on axis 1, then the full
+device index would be `(i, g, j)`.
+
+### Purity and Execution Model
+
+Collective operations involve all devices in a group (e.g. `all-gather`,
+`all-to-all`) and are considered pure. Operations like `send` and `recv` are not
+collective and are not pure.
+
+The execution model assumes SPMD (Single Program, Multiple Data):
+
+* Every process runs the same program.
+* At any collective operation, all processes are in sync.
+
+This means compiler optimizations must treat collective ops carefully. For
+example, if a collective is removed during optimization, it must be removed from
+*every* path and *every* process that would have participated - otherwise, you’ll
+get undefined behavior at runtime.
+
+Marking these ops as pure also helps with standard compiler passes like dead
+code elimination and common subexpression elimination. It ensures that when the
+program is executed, all devices hit the same line of code at the same time
+during collectives and so avoid dead-locks.
+
+## Operations
+
+[include "Dialects/ShardOps.md"]
+
+## Attributes
+
+[include "Dialects/ShardAttrs.md"]
diff --git a/mlir/docs/Interfaces.md b/mlir/docs/Interfaces.md
index bf590ac..7e1c5fe 100644
--- a/mlir/docs/Interfaces.md
+++ b/mlir/docs/Interfaces.md
@@ -563,7 +563,7 @@ def MyInterface : OpInterface<"MyInterface"> {
         template <typename ConcreteOp>
         struct Model : public Concept {
           Operation *create(OpBuilder &builder, Location loc) const override {
-            return builder.create<ConcreteOp>(loc);
+            return ConcreteOp::create(builder, loc);
           }
         }
       };
@@ -574,7 +574,7 @@ def MyInterface : OpInterface<"MyInterface"> {
     }],
       "Operation *", "create", (ins "OpBuilder &":$builder, "Location":$loc),
       /*methodBody=*/[{
-        return builder.create<ConcreteOp>(loc);
+        return ConcreteOp::create(builder, loc);
     }]>,
 
     InterfaceMethod<[{
diff --git a/mlir/docs/PDLL.md b/mlir/docs/PDLL.md
index 9839d1d..c6e352f 100644
--- a/mlir/docs/PDLL.md
+++ b/mlir/docs/PDLL.md
@@ -1483,7 +1483,7 @@ be defined by specifying a string code block after the rewrite declaration:
 
 ```pdll
 Rewrite BuildOp(value: Value) -> (foo: Op<my_dialect.foo>, bar: Op<my_dialect.bar>) [{
-  return {rewriter.create<my_dialect::FooOp>(value), rewriter.create<my_dialect::BarOp>()};
+  return {my_dialect::FooOp::create(rewriter, value), my_dialect::BarOp::create(rewriter)};
 }];
 
 Pattern {
@@ -1508,7 +1508,7 @@ translated into:
 
 ```c++
 std::tuple<my_dialect::FooOp, my_dialect::BarOp> BuildOp(Value value) {
-  return {rewriter.create<my_dialect::FooOp>(value), rewriter.create<my_dialect::BarOp>()};
+  return {my_dialect::FooOp::create(rewriter, value), my_dialect::BarOp::create(rewriter)};
 }
 ```
 
@@ -1530,7 +1530,7 @@ below describes the various result translation scenarios:
 
 ```pdll
 Rewrite createOp() [{
-  rewriter.create<my_dialect::FooOp>();
+  my_dialect::FooOp::create(rewriter);
 }];
 ```
 
@@ -1538,7 +1538,7 @@ In the case where a native `Rewrite` has no results, the native function returns
 
 ```c++
 void createOp(PatternRewriter &rewriter) {
-  rewriter.create<my_dialect::FooOp>();
+  my_dialect::FooOp::create(rewriter);
 }
 ```
 
@@ -1546,7 +1546,7 @@ void createOp(PatternRewriter &rewriter) {
 
 ```pdll
 Rewrite createOp() -> Op<my_dialect.foo> [{
-  return rewriter.create<my_dialect::FooOp>();
+  return my_dialect::FooOp::create(rewriter);
 }];
 ```
 
@@ -1555,7 +1555,7 @@ native type for that single result:
 
 ```c++
 my_dialect::FooOp createOp(PatternRewriter &rewriter) {
-  return rewriter.create<my_dialect::FooOp>();
+  return my_dialect::FooOp::create(rewriter);
 }
 ```
 
diff --git a/mlir/docs/Passes.md b/mlir/docs/Passes.md
index e9d22d1..9df3266 100644
--- a/mlir/docs/Passes.md
+++ b/mlir/docs/Passes.md
@@ -72,9 +72,9 @@ This document describes the available MLIR passes and their contracts.
 
 [include "MemRefPasses.md"]
 
-## 'mesh' Dialect Passes
+## 'shard' Dialect Passes
 
-[include "MeshPasses.md"]
+[include "ShardPasses.md"]
 
 ## 'ml\_program' Dialect Passes
 
diff --git a/mlir/docs/Tutorials/QuickstartRewrites.md b/mlir/docs/Tutorials/QuickstartRewrites.md
index 0c89065..cbb6f03 100644
--- a/mlir/docs/Tutorials/QuickstartRewrites.md
+++ b/mlir/docs/Tutorials/QuickstartRewrites.md
@@ -130,7 +130,7 @@ def : Pat<(TF_LeakyReluOp:$old_value, $arg, F32Attr:$a),
 ```c++
 static Value createTFLLeakyRelu(PatternRewriter &rewriter, Operation *op,
                                 Value operand, Attribute attr) {
-  return rewriter.create<mlir::TFL::LeakyReluOp>(
+  return mlir::TFL::LeakyReluOp::create(rewriter,
       op->getLoc(), operands[0].getType(), /*arg=*/operands[0],
       /*alpha=*/cast<FloatAttr>(attrs[0]));
 }
@@ -194,10 +194,10 @@ LogicalResult circt::MulOp::canonicalize(MulOp op, PatternRewriter &rewriter) {
   // mul(x, c) -> shl(x, log2(c)), where c is a power of two.
   if (inputs.size() == 2 && matchPattern(inputs.back(), m_RConstant(value)) &&
       value.isPowerOf2()) {
-    auto shift = rewriter.create<rtl::ConstantOp>(op.getLoc(), op.getType(),
+    auto shift = rtl::ConstantOp::create(rewriter, op.getLoc(), op.getType(),
                                                   value.exactLogBase2());
     auto shlOp =
-        rewriter.create<comb::ShlOp>(op.getLoc(), inputs[0], shift);
+        comb::ShlOp::create(rewriter, op.getLoc(), inputs[0], shift);
     rewriter.replaceOpWithNewOp<MulOp>(op, op.getType(),
                                        ArrayRef<Value>(shlOp));
     return success();
diff --git a/mlir/docs/Tutorials/Toy/Ch-2.md b/mlir/docs/Tutorials/Toy/Ch-2.md
index 039417c..81e4161 100644
--- a/mlir/docs/Tutorials/Toy/Ch-2.md
+++ b/mlir/docs/Tutorials/Toy/Ch-2.md
@@ -521,7 +521,7 @@ def ConstantOp : Toy_Op<"constant"> {
 
   // Add custom build methods for the constant operation. These methods populate
   // the `state` that MLIR uses to create operations, i.e. these are used when
-  // using `builder.create<ConstantOp>(...)`.
+  // using `ConstantOp::create(builder, ...)`.
   let builders = [
     // Build a constant with a given constant tensor value.
     OpBuilder<(ins "DenseElementsAttr":$value), [{
diff --git a/mlir/docs/Tutorials/Toy/Ch-4.md b/mlir/docs/Tutorials/Toy/Ch-4.md
index 1275d36..e9abe36 100644
--- a/mlir/docs/Tutorials/Toy/Ch-4.md
+++ b/mlir/docs/Tutorials/Toy/Ch-4.md
@@ -300,7 +300,7 @@ struct ToyInlinerInterface : public DialectInlinerInterface {
   Operation *materializeCallConversion(OpBuilder &builder, Value input,
                                        Type resultType,
                                        Location conversionLoc) const final {
-    return builder.create<CastOp>(conversionLoc, resultType, input);
+    return CastOp::create(builder, conversionLoc, resultType, input);
   }
 };
 ```
@@ -445,7 +445,7 @@ When processing an operation like described, we query if it registered the
 
 ```c++
   // Ask the operation to infer its output shapes.
-  LLVM_DEBUG(llvm::dbgs() << "Inferring shape for: " << *op << "\n");
+  LDBG() << "Inferring shape for: " << *op;
 
   /// We check if an operation has a particular interface by casting.
   if (ShapeInference shapeOp = dyn_cast<ShapeInference>(op)) {
diff --git a/mlir/docs/Tutorials/Toy/Ch-5.md b/mlir/docs/Tutorials/Toy/Ch-5.md
index d483cd8..17cd6bb 100644
--- a/mlir/docs/Tutorials/Toy/Ch-5.md
+++ b/mlir/docs/Tutorials/Toy/Ch-5.md
@@ -91,13 +91,11 @@ doesn't matter. See `ConversionTarget::getOpInfo` for the details.
 After the conversion target has been defined, we can define how to convert the
 *illegal* operations into *legal* ones. Similarly to the canonicalization
 framework introduced in [chapter 3](Ch-3.md), the
-[`DialectConversion` framework](../../DialectConversion.md) also uses
-[RewritePatterns](../QuickstartRewrites.md) to perform the conversion logic.
-These patterns may be the `RewritePatterns` seen before or a new type of pattern
-specific to the conversion framework `ConversionPattern`. `ConversionPatterns`
+[`DialectConversion` framework](../../DialectConversion.md) uses a special kind
+of `ConversionPattern` to perform the conversion logic. `ConversionPatterns`
 are different from traditional `RewritePatterns` in that they accept an
-additional `operands` parameter containing operands that have been
-remapped/replaced. This is used when dealing with type conversions, as the
+additional `operands` (or `adaptor`) parameter containing operands that have
+been remapped/replaced. This is used when dealing with type conversions, as the
 pattern will want to operate on values of the new type but match against the
 old. For our lowering, this invariant will be useful as it translates from the
 [TensorType](../../Dialects/Builtin.md/#rankedtensortype) currently being
@@ -106,38 +104,23 @@ look at a snippet of lowering the `toy.transpose` operation:
 
 ```c++
 /// Lower the `toy.transpose` operation to an affine loop nest.
-struct TransposeOpLowering : public mlir::ConversionPattern {
-  TransposeOpLowering(mlir::MLIRContext *ctx)
-      : mlir::ConversionPattern(TransposeOp::getOperationName(), 1, ctx) {}
-
-  /// Match and rewrite the given `toy.transpose` operation, with the given
-  /// operands that have been remapped from `tensor<...>` to `memref<...>`.
-  llvm::LogicalResult
-  matchAndRewrite(mlir::Operation *op, ArrayRef<mlir::Value> operands,
-                  mlir::ConversionPatternRewriter &rewriter) const final {
-    auto loc = op->getLoc();
+struct TransposeOpLowering : public OpConversionPattern<toy::TransposeOp> {
+  using OpConversionPattern<toy::TransposeOp>::OpConversionPattern;
 
-    // Call to a helper function that will lower the current operation to a set
-    // of affine loops. We provide a functor that operates on the remapped
-    // operands, as well as the loop induction variables for the inner most
-    // loop body.
-    lowerOpToLoops(
-        op, operands, rewriter,
-        [loc](mlir::PatternRewriter &rewriter,
-              ArrayRef<mlir::Value> memRefOperands,
-              ArrayRef<mlir::Value> loopIvs) {
-          // Generate an adaptor for the remapped operands of the TransposeOp.
-          // This allows for using the nice named accessors that are generated
-          // by the ODS. This adaptor is automatically provided by the ODS
-          // framework.
-          TransposeOpAdaptor transposeAdaptor(memRefOperands);
-          mlir::Value input = transposeAdaptor.input();
-
-          // Transpose the elements by generating a load from the reverse
-          // indices.
-          SmallVector<mlir::Value, 2> reverseIvs(llvm::reverse(loopIvs));
-          return rewriter.create<mlir::AffineLoadOp>(loc, input, reverseIvs);
-        });
+  LogicalResult
+  matchAndRewrite(toy::TransposeOp op, OpAdaptor adaptor,
+                  ConversionPatternRewriter &rewriter) const final {
+    auto loc = op->getLoc();
+    lowerOpToLoops(op, rewriter,
+                   [&](OpBuilder &builder, ValueRange loopIvs) {
+                     Value input = adaptor.getInput();
+
+                     // Transpose the elements by generating a load from the
+                     // reverse indices.
+                     SmallVector<Value, 2> reverseIvs(llvm::reverse(loopIvs));
+                     return affine::AffineLoadOp::create(builder, loc, input,
+                                                         reverseIvs);
+                   });
     return success();
   }
 };
diff --git a/mlir/docs/Tutorials/Toy/Ch-6.md b/mlir/docs/Tutorials/Toy/Ch-6.md
index e8a68b5..529de55 100644
--- a/mlir/docs/Tutorials/Toy/Ch-6.md
+++ b/mlir/docs/Tutorials/Toy/Ch-6.md
@@ -47,7 +47,7 @@ static FlatSymbolRefAttr getOrInsertPrintf(PatternRewriter &rewriter,
   // Insert the printf function into the body of the parent module.
   PatternRewriter::InsertionGuard insertGuard(rewriter);
   rewriter.setInsertionPointToStart(module.getBody());
-  rewriter.create<LLVM::LLVMFuncOp>(module.getLoc(), "printf", llvmFnType);
+  LLVM::LLVMFuncOp::create(rewriter, module.getLoc(), "printf", llvmFnType);
   return SymbolRefAttr::get("printf", context);
 }
 ```
diff --git a/mlir/docs/Tutorials/Toy/Ch-7.md b/mlir/docs/Tutorials/Toy/Ch-7.md
index dce3490..0f50c49 100644
--- a/mlir/docs/Tutorials/Toy/Ch-7.md
+++ b/mlir/docs/Tutorials/Toy/Ch-7.md
@@ -488,9 +488,9 @@ mlir::Operation *ToyDialect::materializeConstant(mlir::OpBuilder &builder,
                                                  mlir::Type type,
                                                  mlir::Location loc) {
   if (isa<StructType>(type))
-    return builder.create<StructConstantOp>(loc, type,
+    return StructConstantOp::create(builder, loc, type,
                                             cast<mlir::ArrayAttr>(value));
-  return builder.create<ConstantOp>(loc, type,
+  return ConstantOp::create(builder, loc, type,
                                     cast<mlir::DenseElementsAttr>(value));
 }
 ```
diff --git a/mlir/docs/Tutorials/transform/Ch3.md b/mlir/docs/Tutorials/transform/Ch3.md
index fa788d1..eeab770 100644
--- a/mlir/docs/Tutorials/transform/Ch3.md
+++ b/mlir/docs/Tutorials/transform/Ch3.md
@@ -139,7 +139,21 @@ void MyExtension::init() {
 ```
 
 This type is now directly available in the Transform dialect and can be used in operations.
+In the previous tablegen definition, the type of `$call` must be `Transform_ConcreteOp<“func.call”>`,
+By adding `CallOpInterfaceHandle` as an allowed type for `$call`, the corresponding handle
+is allowed to be to any op implementing the interface.
 
+```tablegen
+def ChangeCallTargetOp : ... {
+    let arguments = (ins
+    // Allow the handle to be to concrete `func.call` ops as well as any op implementing
+    // the `CallOpInterface`.
+    AnyTypeOf<[Transform_ConcreteOpType<"func.call">, CallOpInterfaceHandle]>:$call,
+    StrAttr:$new_target); 
+}
+```
+
+We can then add the following code to `sequence.mlir` and run it with the interpreter.
 
 ```mlir
   // Cast to our new type.
@@ -172,7 +186,7 @@ def CallToOp : Op<Transform_Dialect, "my.call_to_op",
   let results = (outs TransformHandleTypeInterface:$transformed);
 
   // Provide nice syntax.
-  let assemblyFormat = "$call attr-dict `:` functional-type(inputs, outputs)";
+  let assemblyFormat = "$call attr-dict `:` functional-type(operands, results)";
 
   // Declare the function implementing the interface for a single payload operation.
   let extraClassDeclaration = [{
diff --git a/mlir/examples/toy/Ch2/include/toy/Ops.td b/mlir/examples/toy/Ch2/include/toy/Ops.td
index ef65c9c..91bf83a 100644
--- a/mlir/examples/toy/Ch2/include/toy/Ops.td
+++ b/mlir/examples/toy/Ch2/include/toy/Ops.td
@@ -70,7 +70,7 @@ def ConstantOp : Toy_Op<"constant", [Pure]> {
 
   // Add custom build methods for the constant operation. These method populates
   // the `state` that MLIR uses to create operations, i.e. these are used when
-  // using `builder.create<ConstantOp>(...)`.
+  // using `ConstantOp::create(builder, ...)`.
   let builders = [
     // Build a constant with a given constant tensor value.
     OpBuilder<(ins "DenseElementsAttr":$value), [{
diff --git a/mlir/examples/toy/Ch2/mlir/MLIRGen.cpp b/mlir/examples/toy/Ch2/mlir/MLIRGen.cpp
index 96925be..39ae6a0 100644
--- a/mlir/examples/toy/Ch2/mlir/MLIRGen.cpp
+++ b/mlir/examples/toy/Ch2/mlir/MLIRGen.cpp
@@ -121,8 +121,8 @@ private:
     llvm::SmallVector<mlir::Type, 4> argTypes(proto.getArgs().size(),
                                               getType(VarType{}));
     auto funcType = builder.getFunctionType(argTypes, {});
-    return builder.create<mlir::toy::FuncOp>(location, proto.getName(),
-                                             funcType);
+    return mlir::toy::FuncOp::create(builder, location, proto.getName(),
+                                     funcType);
   }
 
   /// Emit a new function and add it to the MLIR module.
@@ -166,7 +166,7 @@ private:
     if (!entryBlock.empty())
       returnOp = dyn_cast<ReturnOp>(entryBlock.back());
     if (!returnOp) {
-      builder.create<ReturnOp>(loc(funcAST.getProto()->loc()));
+      ReturnOp::create(builder, loc(funcAST.getProto()->loc()));
     } else if (returnOp.hasOperand()) {
       // Otherwise, if this return operation has an operand then add a result to
       // the function.
@@ -202,9 +202,9 @@ private:
     // support '+' and '*'.
     switch (binop.getOp()) {
     case '+':
-      return builder.create<AddOp>(location, lhs, rhs);
+      return AddOp::create(builder, location, lhs, rhs);
     case '*':
-      return builder.create<MulOp>(location, lhs, rhs);
+      return MulOp::create(builder, location, lhs, rhs);
     }
 
     emitError(location, "invalid binary operator '") << binop.getOp() << "'";
@@ -235,8 +235,8 @@ private:
     }
 
     // Otherwise, this return operation has zero operands.
-    builder.create<ReturnOp>(location,
-                             expr ? ArrayRef(expr) : ArrayRef<mlir::Value>());
+    ReturnOp::create(builder, location,
+                     expr ? ArrayRef(expr) : ArrayRef<mlir::Value>());
     return mlir::success();
   }
 
@@ -280,7 +280,7 @@ private:
 
     // Build the MLIR op `toy.constant`. This invokes the `ConstantOp::build`
     // method.
-    return builder.create<ConstantOp>(loc(lit.loc()), type, dataAttribute);
+    return ConstantOp::create(builder, loc(lit.loc()), type, dataAttribute);
   }
 
   /// Recursive helper function to accumulate the data that compose an array
@@ -325,13 +325,13 @@ private:
                             "does not accept multiple arguments");
         return nullptr;
       }
-      return builder.create<TransposeOp>(location, operands[0]);
+      return TransposeOp::create(builder, location, operands[0]);
     }
 
     // Otherwise this is a call to a user-defined function. Calls to
     // user-defined functions are mapped to a custom call that takes the callee
     // name as an attribute.
-    return builder.create<GenericCallOp>(location, callee, operands);
+    return GenericCallOp::create(builder, location, callee, operands);
   }
 
   /// Emit a print expression. It emits specific operations for two builtins:
@@ -341,13 +341,13 @@ private:
     if (!arg)
       return mlir::failure();
 
-    builder.create<PrintOp>(loc(call.loc()), arg);
+    PrintOp::create(builder, loc(call.loc()), arg);
     return mlir::success();
   }
 
   /// Emit a constant for a single number (FIXME: semantic? broadcast?)
   mlir::Value mlirGen(NumberExprAST &num) {
-    return builder.create<ConstantOp>(loc(num.loc()), num.getValue());
+    return ConstantOp::create(builder, loc(num.loc()), num.getValue());
   }
 
   /// Dispatch codegen for the right expression subclass using RTTI.
@@ -391,8 +391,8 @@ private:
     // with specific shape, we emit a "reshape" operation. It will get
     // optimized out later as needed.
     if (!vardecl.getType().shape.empty()) {
-      value = builder.create<ReshapeOp>(loc(vardecl.loc()),
-                                        getType(vardecl.getType()), value);
+      value = ReshapeOp::create(builder, loc(vardecl.loc()),
+                                getType(vardecl.getType()), value);
     }
 
     // Register the value in the symbol table.
diff --git a/mlir/examples/toy/Ch3/include/toy/Ops.td b/mlir/examples/toy/Ch3/include/toy/Ops.td
index 4859804..027b076 100644
--- a/mlir/examples/toy/Ch3/include/toy/Ops.td
+++ b/mlir/examples/toy/Ch3/include/toy/Ops.td
@@ -69,7 +69,7 @@ def ConstantOp : Toy_Op<"constant", [Pure]> {
 
   // Add custom build methods for the constant operation. These method populates
   // the `state` that MLIR uses to create operations, i.e. these are used when
-  // using `builder.create<ConstantOp>(...)`.
+  // using `ConstantOp::create(builder, ...)`.
   let builders = [
     // Build a constant with a given constant tensor value.
     OpBuilder<(ins "DenseElementsAttr":$value), [{
diff --git a/mlir/examples/toy/Ch3/mlir/MLIRGen.cpp b/mlir/examples/toy/Ch3/mlir/MLIRGen.cpp
index c8cba82..0573af6 100644
--- a/mlir/examples/toy/Ch3/mlir/MLIRGen.cpp
+++ b/mlir/examples/toy/Ch3/mlir/MLIRGen.cpp
@@ -121,8 +121,8 @@ private:
     llvm::SmallVector<mlir::Type, 4> argTypes(proto.getArgs().size(),
                                               getType(VarType{}));
     auto funcType = builder.getFunctionType(argTypes, /*results=*/{});
-    return builder.create<mlir::toy::FuncOp>(location, proto.getName(),
-                                             funcType);
+    return mlir::toy::FuncOp::create(builder, location, proto.getName(),
+                                     funcType);
   }
 
   /// Emit a new function and add it to the MLIR module.
@@ -166,7 +166,7 @@ private:
     if (!entryBlock.empty())
       returnOp = dyn_cast<ReturnOp>(entryBlock.back());
     if (!returnOp) {
-      builder.create<ReturnOp>(loc(funcAST.getProto()->loc()));
+      ReturnOp::create(builder, loc(funcAST.getProto()->loc()));
     } else if (returnOp.hasOperand()) {
       // Otherwise, if this return operation has an operand then add a result to
       // the function.
@@ -202,9 +202,9 @@ private:
     // support '+' and '*'.
     switch (binop.getOp()) {
     case '+':
-      return builder.create<AddOp>(location, lhs, rhs);
+      return AddOp::create(builder, location, lhs, rhs);
     case '*':
-      return builder.create<MulOp>(location, lhs, rhs);
+      return MulOp::create(builder, location, lhs, rhs);
     }
 
     emitError(location, "invalid binary operator '") << binop.getOp() << "'";
@@ -235,8 +235,8 @@ private:
     }
 
     // Otherwise, this return operation has zero operands.
-    builder.create<ReturnOp>(location,
-                             expr ? ArrayRef(expr) : ArrayRef<mlir::Value>());
+    ReturnOp::create(builder, location,
+                     expr ? ArrayRef(expr) : ArrayRef<mlir::Value>());
     return mlir::success();
   }
 
@@ -280,7 +280,7 @@ private:
 
     // Build the MLIR op `toy.constant`. This invokes the `ConstantOp::build`
     // method.
-    return builder.create<ConstantOp>(loc(lit.loc()), type, dataAttribute);
+    return ConstantOp::create(builder, loc(lit.loc()), type, dataAttribute);
   }
 
   /// Recursive helper function to accumulate the data that compose an array
@@ -325,13 +325,13 @@ private:
                             "does not accept multiple arguments");
         return nullptr;
       }
-      return builder.create<TransposeOp>(location, operands[0]);
+      return TransposeOp::create(builder, location, operands[0]);
     }
 
     // Otherwise this is a call to a user-defined function. Calls to
     // user-defined functions are mapped to a custom call that takes the callee
     // name as an attribute.
-    return builder.create<GenericCallOp>(location, callee, operands);
+    return GenericCallOp::create(builder, location, callee, operands);
   }
 
   /// Emit a print expression. It emits specific operations for two builtins:
@@ -341,13 +341,13 @@ private:
     if (!arg)
       return mlir::failure();
 
-    builder.create<PrintOp>(loc(call.loc()), arg);
+    PrintOp::create(builder, loc(call.loc()), arg);
     return mlir::success();
   }
 
   /// Emit a constant for a single number (FIXME: semantic? broadcast?)
   mlir::Value mlirGen(NumberExprAST &num) {
-    return builder.create<ConstantOp>(loc(num.loc()), num.getValue());
+    return ConstantOp::create(builder, loc(num.loc()), num.getValue());
   }
 
   /// Dispatch codegen for the right expression subclass using RTTI.
@@ -391,8 +391,8 @@ private:
     // with specific shape, we emit a "reshape" operation. It will get
     // optimized out later as needed.
     if (!vardecl.getType().shape.empty()) {
-      value = builder.create<ReshapeOp>(loc(vardecl.loc()),
-                                        getType(vardecl.getType()), value);
+      value = ReshapeOp::create(builder, loc(vardecl.loc()),
+                                getType(vardecl.getType()), value);
     }
 
     // Register the value in the symbol table.
diff --git a/mlir/examples/toy/Ch4/include/toy/Ops.td b/mlir/examples/toy/Ch4/include/toy/Ops.td
index 0b32b1b..6c6b739 100644
--- a/mlir/examples/toy/Ch4/include/toy/Ops.td
+++ b/mlir/examples/toy/Ch4/include/toy/Ops.td
@@ -72,7 +72,7 @@ def ConstantOp : Toy_Op<"constant", [Pure]> {
 
   // Add custom build methods for the constant operation. These method populates
   // the `state` that MLIR uses to create operations, i.e. these are used when
-  // using `builder.create<ConstantOp>(...)`.
+  // using `ConstantOp::create(builder, ...)`.
   let builders = [
     // Build a constant with a given constant tensor value.
     OpBuilder<(ins "DenseElementsAttr":$value), [{
diff --git a/mlir/examples/toy/Ch4/mlir/Dialect.cpp b/mlir/examples/toy/Ch4/mlir/Dialect.cpp
index 076a75a..1e5e672 100644
--- a/mlir/examples/toy/Ch4/mlir/Dialect.cpp
+++ b/mlir/examples/toy/Ch4/mlir/Dialect.cpp
@@ -91,7 +91,7 @@ struct ToyInlinerInterface : public DialectInlinerInterface {
   Operation *materializeCallConversion(OpBuilder &builder, Value input,
                                        Type resultType,
                                        Location conversionLoc) const final {
-    return builder.create<CastOp>(conversionLoc, resultType, input);
+    return CastOp::create(builder, conversionLoc, resultType, input);
   }
 };
 
@@ -206,7 +206,8 @@ void ConstantOp::print(mlir::OpAsmPrinter &printer) {
 llvm::LogicalResult ConstantOp::verify() {
   // If the return type of the constant is not an unranked tensor, the shape
   // must match the shape of the attribute holding the data.
-  auto resultType = llvm::dyn_cast<mlir::RankedTensorType>(getResult().getType());
+  auto resultType =
+      llvm::dyn_cast<mlir::RankedTensorType>(getResult().getType());
   if (!resultType)
     return success();
 
@@ -395,7 +396,8 @@ llvm::LogicalResult ReturnOp::verify() {
   auto resultType = results.front();
 
   // Check that the result type of the function matches the operand type.
-  if (inputType == resultType || llvm::isa<mlir::UnrankedTensorType>(inputType) ||
+  if (inputType == resultType ||
+      llvm::isa<mlir::UnrankedTensorType>(inputType) ||
       llvm::isa<mlir::UnrankedTensorType>(resultType))
     return mlir::success();
 
diff --git a/mlir/examples/toy/Ch4/mlir/MLIRGen.cpp b/mlir/examples/toy/Ch4/mlir/MLIRGen.cpp
index 9371815..7d676f1 100644
--- a/mlir/examples/toy/Ch4/mlir/MLIRGen.cpp
+++ b/mlir/examples/toy/Ch4/mlir/MLIRGen.cpp
@@ -121,8 +121,8 @@ private:
     llvm::SmallVector<mlir::Type, 4> argTypes(proto.getArgs().size(),
                                               getType(VarType{}));
     auto funcType = builder.getFunctionType(argTypes, /*results=*/{});
-    return builder.create<mlir::toy::FuncOp>(location, proto.getName(),
-                                             funcType);
+    return mlir::toy::FuncOp::create(builder, location, proto.getName(),
+                                     funcType);
   }
 
   /// Emit a new function and add it to the MLIR module.
@@ -166,7 +166,7 @@ private:
     if (!entryBlock.empty())
       returnOp = dyn_cast<ReturnOp>(entryBlock.back());
     if (!returnOp) {
-      builder.create<ReturnOp>(loc(funcAST.getProto()->loc()));
+      ReturnOp::create(builder, loc(funcAST.getProto()->loc()));
     } else if (returnOp.hasOperand()) {
       // Otherwise, if this return operation has an operand then add a result to
       // the function.
@@ -206,9 +206,9 @@ private:
     // support '+' and '*'.
     switch (binop.getOp()) {
     case '+':
-      return builder.create<AddOp>(location, lhs, rhs);
+      return AddOp::create(builder, location, lhs, rhs);
     case '*':
-      return builder.create<MulOp>(location, lhs, rhs);
+      return MulOp::create(builder, location, lhs, rhs);
     }
 
     emitError(location, "invalid binary operator '") << binop.getOp() << "'";
@@ -239,8 +239,8 @@ private:
     }
 
     // Otherwise, this return operation has zero operands.
-    builder.create<ReturnOp>(location,
-                             expr ? ArrayRef(expr) : ArrayRef<mlir::Value>());
+    ReturnOp::create(builder, location,
+                     expr ? ArrayRef(expr) : ArrayRef<mlir::Value>());
     return mlir::success();
   }
 
@@ -284,7 +284,7 @@ private:
 
     // Build the MLIR op `toy.constant`. This invokes the `ConstantOp::build`
     // method.
-    return builder.create<ConstantOp>(loc(lit.loc()), type, dataAttribute);
+    return ConstantOp::create(builder, loc(lit.loc()), type, dataAttribute);
   }
 
   /// Recursive helper function to accumulate the data that compose an array
@@ -329,13 +329,13 @@ private:
                             "does not accept multiple arguments");
         return nullptr;
       }
-      return builder.create<TransposeOp>(location, operands[0]);
+      return TransposeOp::create(builder, location, operands[0]);
     }
 
     // Otherwise this is a call to a user-defined function. Calls to
     // user-defined functions are mapped to a custom call that takes the callee
     // name as an attribute.
-    return builder.create<GenericCallOp>(location, callee, operands);
+    return GenericCallOp::create(builder, location, callee, operands);
   }
 
   /// Emit a print expression. It emits specific operations for two builtins:
@@ -345,13 +345,13 @@ private:
     if (!arg)
       return mlir::failure();
 
-    builder.create<PrintOp>(loc(call.loc()), arg);
+    PrintOp::create(builder, loc(call.loc()), arg);
     return mlir::success();
   }
 
   /// Emit a constant for a single number (FIXME: semantic? broadcast?)
   mlir::Value mlirGen(NumberExprAST &num) {
-    return builder.create<ConstantOp>(loc(num.loc()), num.getValue());
+    return ConstantOp::create(builder, loc(num.loc()), num.getValue());
   }
 
   /// Dispatch codegen for the right expression subclass using RTTI.
@@ -395,8 +395,8 @@ private:
     // with specific shape, we emit a "reshape" operation. It will get
     // optimized out later as needed.
     if (!vardecl.getType().shape.empty()) {
-      value = builder.create<ReshapeOp>(loc(vardecl.loc()),
-                                        getType(vardecl.getType()), value);
+      value = ReshapeOp::create(builder, loc(vardecl.loc()),
+                                getType(vardecl.getType()), value);
     }
 
     // Register the value in the symbol table.
diff --git a/mlir/examples/toy/Ch4/mlir/ShapeInferencePass.cpp b/mlir/examples/toy/Ch4/mlir/ShapeInferencePass.cpp
index 2522abe..a552e1f0 100644
--- a/mlir/examples/toy/Ch4/mlir/ShapeInferencePass.cpp
+++ b/mlir/examples/toy/Ch4/mlir/ShapeInferencePass.cpp
@@ -23,7 +23,7 @@
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/Support/Casting.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 #include "llvm/Support/raw_ostream.h"
 #include <memory>
 
@@ -81,7 +81,7 @@ struct ShapeInferencePass
       opWorklist.erase(op);
 
       // Ask the operation to infer its output shapes.
-      LLVM_DEBUG(llvm::dbgs() << "Inferring shape for: " << *op << "\n");
+      LDBG() << "Inferring shape for: " << *op;
       if (auto shapeOp = dyn_cast<ShapeInference>(op)) {
         shapeOp.inferShapes();
       } else {
diff --git a/mlir/examples/toy/Ch5/include/toy/Ops.td b/mlir/examples/toy/Ch5/include/toy/Ops.td
index d11d18dc..6a136ec 100644
--- a/mlir/examples/toy/Ch5/include/toy/Ops.td
+++ b/mlir/examples/toy/Ch5/include/toy/Ops.td
@@ -72,7 +72,7 @@ def ConstantOp : Toy_Op<"constant", [Pure]> {
 
   // Add custom build methods for the constant operation. These method populates
   // the `state` that MLIR uses to create operations, i.e. these are used when
-  // using `builder.create<ConstantOp>(...)`.
+  // using `ConstantOp::create(builder, ...)`.
   let builders = [
     // Build a constant with a given constant tensor value.
     OpBuilder<(ins "DenseElementsAttr":$value), [{
diff --git a/mlir/examples/toy/Ch5/mlir/Dialect.cpp b/mlir/examples/toy/Ch5/mlir/Dialect.cpp
index fb7c742..69fb69f 100644
--- a/mlir/examples/toy/Ch5/mlir/Dialect.cpp
+++ b/mlir/examples/toy/Ch5/mlir/Dialect.cpp
@@ -91,7 +91,7 @@ struct ToyInlinerInterface : public DialectInlinerInterface {
   Operation *materializeCallConversion(OpBuilder &builder, Value input,
                                        Type resultType,
                                        Location conversionLoc) const final {
-    return builder.create<CastOp>(conversionLoc, resultType, input);
+    return CastOp::create(builder, conversionLoc, resultType, input);
   }
 };
 
@@ -206,7 +206,8 @@ void ConstantOp::print(mlir::OpAsmPrinter &printer) {
 llvm::LogicalResult ConstantOp::verify() {
   // If the return type of the constant is not an unranked tensor, the shape
   // must match the shape of the attribute holding the data.
-  auto resultType = llvm::dyn_cast<mlir::RankedTensorType>(getResult().getType());
+  auto resultType =
+      llvm::dyn_cast<mlir::RankedTensorType>(getResult().getType());
   if (!resultType)
     return success();
 
@@ -395,7 +396,8 @@ llvm::LogicalResult ReturnOp::verify() {
   auto resultType = results.front();
 
   // Check that the result type of the function matches the operand type.
-  if (inputType == resultType || llvm::isa<mlir::UnrankedTensorType>(inputType) ||
+  if (inputType == resultType ||
+      llvm::isa<mlir::UnrankedTensorType>(inputType) ||
       llvm::isa<mlir::UnrankedTensorType>(resultType))
     return mlir::success();
 
diff --git a/mlir/examples/toy/Ch5/mlir/LowerToAffineLoops.cpp b/mlir/examples/toy/Ch5/mlir/LowerToAffineLoops.cpp
index bf2bc43..2969d3a 100644
--- a/mlir/examples/toy/Ch5/mlir/LowerToAffineLoops.cpp
+++ b/mlir/examples/toy/Ch5/mlir/LowerToAffineLoops.cpp
@@ -44,7 +44,7 @@
 using namespace mlir;
 
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns
+// ToyToAffine Conversion Patterns
 //===----------------------------------------------------------------------===//
 
 /// Convert the given RankedTensorType into the corresponding MemRefType.
@@ -55,7 +55,7 @@ static MemRefType convertTensorToMemRef(RankedTensorType type) {
 /// Insert an allocation and deallocation for the given MemRefType.
 static Value insertAllocAndDealloc(MemRefType type, Location loc,
                                    PatternRewriter &rewriter) {
-  auto alloc = rewriter.create<memref::AllocOp>(loc, type);
+  auto alloc = memref::AllocOp::create(rewriter, loc, type);
 
   // Make sure to allocate at the beginning of the block.
   auto *parentBlock = alloc->getBlock();
@@ -63,21 +63,19 @@ static Value insertAllocAndDealloc(MemRefType type, Location loc,
 
   // Make sure to deallocate this alloc at the end of the block. This is fine
   // as toy functions have no control flow.
-  auto dealloc = rewriter.create<memref::DeallocOp>(loc, alloc);
+  auto dealloc = memref::DeallocOp::create(rewriter, loc, alloc);
   dealloc->moveBefore(&parentBlock->back());
   return alloc;
 }
 
 /// This defines the function type used to process an iteration of a lowered
-/// loop. It takes as input an OpBuilder, an range of memRefOperands
-/// corresponding to the operands of the input operation, and the range of loop
-/// induction variables for the iteration. It returns a value to store at the
-/// current index of the iteration.
-using LoopIterationFn = function_ref<Value(
-    OpBuilder &rewriter, ValueRange memRefOperands, ValueRange loopIvs)>;
-
-static void lowerOpToLoops(Operation *op, ValueRange operands,
-                           PatternRewriter &rewriter,
+/// loop. It takes as input an OpBuilder and the range of loop induction
+/// variables for the iteration. It returns a value to store at the current
+/// index of the iteration.
+using LoopIterationFn =
+    function_ref<Value(OpBuilder &rewriter, ValueRange loopIvs)>;
+
+static void lowerOpToLoops(Operation *op, PatternRewriter &rewriter,
                            LoopIterationFn processIteration) {
   auto tensorType = llvm::cast<RankedTensorType>((*op->result_type_begin()));
   auto loc = op->getLoc();
@@ -95,12 +93,12 @@ static void lowerOpToLoops(Operation *op, ValueRange operands,
   affine::buildAffineLoopNest(
       rewriter, loc, lowerBounds, tensorType.getShape(), steps,
       [&](OpBuilder &nestedBuilder, Location loc, ValueRange ivs) {
-        // Call the processing function with the rewriter, the memref operands,
-        // and the loop induction variables. This function will return the value
-        // to store at the current index.
-        Value valueToStore = processIteration(nestedBuilder, operands, ivs);
-        nestedBuilder.create<affine::AffineStoreOp>(loc, valueToStore, alloc,
-                                                    ivs);
+        // Call the processing function with the rewriter and the loop
+        // induction variables. This function will return the value to store at
+        // the current index.
+        Value valueToStore = processIteration(nestedBuilder, ivs);
+        affine::AffineStoreOp::create(nestedBuilder, loc, valueToStore, alloc,
+                                      ivs);
       });
 
   // Replace this operation with the generated alloc.
@@ -109,38 +107,30 @@ static void lowerOpToLoops(Operation *op, ValueRange operands,
 
 namespace {
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns: Binary operations
+// ToyToAffine Conversion Patterns: Binary operations
 //===----------------------------------------------------------------------===//
 
 template <typename BinaryOp, typename LoweredBinaryOp>
-struct BinaryOpLowering : public ConversionPattern {
-  BinaryOpLowering(MLIRContext *ctx)
-      : ConversionPattern(BinaryOp::getOperationName(), 1, ctx) {}
+struct BinaryOpLowering : public OpConversionPattern<BinaryOp> {
+  using OpConversionPattern<BinaryOp>::OpConversionPattern;
+  using OpAdaptor = typename OpConversionPattern<BinaryOp>::OpAdaptor;
 
   LogicalResult
-  matchAndRewrite(Operation *op, ArrayRef<Value> operands,
+  matchAndRewrite(BinaryOp op, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const final {
     auto loc = op->getLoc();
-    lowerOpToLoops(op, operands, rewriter,
-                   [loc](OpBuilder &builder, ValueRange memRefOperands,
-                         ValueRange loopIvs) {
-                     // Generate an adaptor for the remapped operands of the
-                     // BinaryOp. This allows for using the nice named accessors
-                     // that are generated by the ODS.
-                     typename BinaryOp::Adaptor binaryAdaptor(memRefOperands);
-
-                     // Generate loads for the element of 'lhs' and 'rhs' at the
-                     // inner loop.
-                     auto loadedLhs = builder.create<affine::AffineLoadOp>(
-                         loc, binaryAdaptor.getLhs(), loopIvs);
-                     auto loadedRhs = builder.create<affine::AffineLoadOp>(
-                         loc, binaryAdaptor.getRhs(), loopIvs);
-
-                     // Create the binary operation performed on the loaded
-                     // values.
-                     return builder.create<LoweredBinaryOp>(loc, loadedLhs,
-                                                            loadedRhs);
-                   });
+    lowerOpToLoops(op, rewriter, [&](OpBuilder &builder, ValueRange loopIvs) {
+      // Generate loads for the element of 'lhs' and 'rhs' at the
+      // inner loop.
+      auto loadedLhs =
+          affine::AffineLoadOp::create(builder, loc, adaptor.getLhs(), loopIvs);
+      auto loadedRhs =
+          affine::AffineLoadOp::create(builder, loc, adaptor.getRhs(), loopIvs);
+
+      // Create the binary operation performed on the loaded
+      // values.
+      return LoweredBinaryOp::create(builder, loc, loadedLhs, loadedRhs);
+    });
     return success();
   }
 };
@@ -148,14 +138,15 @@ using AddOpLowering = BinaryOpLowering<toy::AddOp, arith::AddFOp>;
 using MulOpLowering = BinaryOpLowering<toy::MulOp, arith::MulFOp>;
 
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns: Constant operations
+// ToyToAffine Conversion Patterns: Constant operations
 //===----------------------------------------------------------------------===//
 
-struct ConstantOpLowering : public OpRewritePattern<toy::ConstantOp> {
-  using OpRewritePattern<toy::ConstantOp>::OpRewritePattern;
+struct ConstantOpLowering : public OpConversionPattern<toy::ConstantOp> {
+  using OpConversionPattern<toy::ConstantOp>::OpConversionPattern;
 
-  LogicalResult matchAndRewrite(toy::ConstantOp op,
-                                PatternRewriter &rewriter) const final {
+  LogicalResult
+  matchAndRewrite(toy::ConstantOp op, OpAdaptor adaptor,
+                  ConversionPatternRewriter &rewriter) const final {
     DenseElementsAttr constantValue = op.getValue();
     Location loc = op.getLoc();
 
@@ -174,11 +165,11 @@ struct ConstantOpLowering : public OpRewritePattern<toy::ConstantOp> {
     if (!valueShape.empty()) {
       for (auto i : llvm::seq<int64_t>(0, *llvm::max_element(valueShape)))
         constantIndices.push_back(
-            rewriter.create<arith::ConstantIndexOp>(loc, i));
+            arith::ConstantIndexOp::create(rewriter, loc, i));
     } else {
       // This is the case of a tensor of rank 0.
       constantIndices.push_back(
-          rewriter.create<arith::ConstantIndexOp>(loc, 0));
+          arith::ConstantIndexOp::create(rewriter, loc, 0));
     }
 
     // The constant operation represents a multi-dimensional constant, so we
@@ -191,9 +182,9 @@ struct ConstantOpLowering : public OpRewritePattern<toy::ConstantOp> {
       // The last dimension is the base case of the recursion, at this point
       // we store the element at the given index.
       if (dimension == valueShape.size()) {
-        rewriter.create<affine::AffineStoreOp>(
-            loc, rewriter.create<arith::ConstantOp>(loc, *valueIt++), alloc,
-            llvm::ArrayRef(indices));
+        affine::AffineStoreOp::create(
+            rewriter, loc, arith::ConstantOp::create(rewriter, loc, *valueIt++),
+            alloc, llvm::ArrayRef(indices));
         return;
       }
 
@@ -216,7 +207,7 @@ struct ConstantOpLowering : public OpRewritePattern<toy::ConstantOp> {
 };
 
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns: Func operations
+// ToyToAffine Conversion Patterns: Func operations
 //===----------------------------------------------------------------------===//
 
 struct FuncOpLowering : public OpConversionPattern<toy::FuncOp> {
@@ -238,8 +229,8 @@ struct FuncOpLowering : public OpConversionPattern<toy::FuncOp> {
     }
 
     // Create a new non-toy function, with the same region.
-    auto func = rewriter.create<mlir::func::FuncOp>(op.getLoc(), op.getName(),
-                                                    op.getFunctionType());
+    auto func = mlir::func::FuncOp::create(rewriter, op.getLoc(), op.getName(),
+                                           op.getFunctionType());
     rewriter.inlineRegionBefore(op.getRegion(), func.getBody(), func.end());
     rewriter.eraseOp(op);
     return success();
@@ -247,7 +238,7 @@ struct FuncOpLowering : public OpConversionPattern<toy::FuncOp> {
 };
 
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns: Print operations
+// ToyToAffine Conversion Patterns: Print operations
 //===----------------------------------------------------------------------===//
 
 struct PrintOpLowering : public OpConversionPattern<toy::PrintOp> {
@@ -265,14 +256,15 @@ struct PrintOpLowering : public OpConversionPattern<toy::PrintOp> {
 };
 
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns: Return operations
+// ToyToAffine Conversion Patterns: Return operations
 //===----------------------------------------------------------------------===//
 
-struct ReturnOpLowering : public OpRewritePattern<toy::ReturnOp> {
-  using OpRewritePattern<toy::ReturnOp>::OpRewritePattern;
+struct ReturnOpLowering : public OpConversionPattern<toy::ReturnOp> {
+  using OpConversionPattern<toy::ReturnOp>::OpConversionPattern;
 
-  LogicalResult matchAndRewrite(toy::ReturnOp op,
-                                PatternRewriter &rewriter) const final {
+  LogicalResult
+  matchAndRewrite(toy::ReturnOp op, OpAdaptor adaptor,
+                  ConversionPatternRewriter &rewriter) const final {
     // During this lowering, we expect that all function calls have been
     // inlined.
     if (op.hasOperand())
@@ -285,32 +277,24 @@ struct ReturnOpLowering : public OpRewritePattern<toy::ReturnOp> {
 };
 
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns: Transpose operations
+// ToyToAffine Conversion Patterns: Transpose operations
 //===----------------------------------------------------------------------===//
 
-struct TransposeOpLowering : public ConversionPattern {
-  TransposeOpLowering(MLIRContext *ctx)
-      : ConversionPattern(toy::TransposeOp::getOperationName(), 1, ctx) {}
+struct TransposeOpLowering : public OpConversionPattern<toy::TransposeOp> {
+  using OpConversionPattern<toy::TransposeOp>::OpConversionPattern;
 
   LogicalResult
-  matchAndRewrite(Operation *op, ArrayRef<Value> operands,
+  matchAndRewrite(toy::TransposeOp op, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const final {
     auto loc = op->getLoc();
-    lowerOpToLoops(op, operands, rewriter,
-                   [loc](OpBuilder &builder, ValueRange memRefOperands,
-                         ValueRange loopIvs) {
-                     // Generate an adaptor for the remapped operands of the
-                     // TransposeOp. This allows for using the nice named
-                     // accessors that are generated by the ODS.
-                     toy::TransposeOpAdaptor transposeAdaptor(memRefOperands);
-                     Value input = transposeAdaptor.getInput();
-
-                     // Transpose the elements by generating a load from the
-                     // reverse indices.
-                     SmallVector<Value, 2> reverseIvs(llvm::reverse(loopIvs));
-                     return builder.create<affine::AffineLoadOp>(loc, input,
-                                                                 reverseIvs);
-                   });
+    lowerOpToLoops(op, rewriter, [&](OpBuilder &builder, ValueRange loopIvs) {
+      Value input = adaptor.getInput();
+
+      // Transpose the elements by generating a load from the
+      // reverse indices.
+      SmallVector<Value, 2> reverseIvs(llvm::reverse(loopIvs));
+      return affine::AffineLoadOp::create(builder, loc, input, reverseIvs);
+    });
     return success();
   }
 };
diff --git a/mlir/examples/toy/Ch5/mlir/MLIRGen.cpp b/mlir/examples/toy/Ch5/mlir/MLIRGen.cpp
index 9371815..7d676f1 100644
--- a/mlir/examples/toy/Ch5/mlir/MLIRGen.cpp
+++ b/mlir/examples/toy/Ch5/mlir/MLIRGen.cpp
@@ -121,8 +121,8 @@ private:
     llvm::SmallVector<mlir::Type, 4> argTypes(proto.getArgs().size(),
                                               getType(VarType{}));
     auto funcType = builder.getFunctionType(argTypes, /*results=*/{});
-    return builder.create<mlir::toy::FuncOp>(location, proto.getName(),
-                                             funcType);
+    return mlir::toy::FuncOp::create(builder, location, proto.getName(),
+                                     funcType);
   }
 
   /// Emit a new function and add it to the MLIR module.
@@ -166,7 +166,7 @@ private:
     if (!entryBlock.empty())
       returnOp = dyn_cast<ReturnOp>(entryBlock.back());
     if (!returnOp) {
-      builder.create<ReturnOp>(loc(funcAST.getProto()->loc()));
+      ReturnOp::create(builder, loc(funcAST.getProto()->loc()));
     } else if (returnOp.hasOperand()) {
       // Otherwise, if this return operation has an operand then add a result to
       // the function.
@@ -206,9 +206,9 @@ private:
     // support '+' and '*'.
     switch (binop.getOp()) {
     case '+':
-      return builder.create<AddOp>(location, lhs, rhs);
+      return AddOp::create(builder, location, lhs, rhs);
     case '*':
-      return builder.create<MulOp>(location, lhs, rhs);
+      return MulOp::create(builder, location, lhs, rhs);
     }
 
     emitError(location, "invalid binary operator '") << binop.getOp() << "'";
@@ -239,8 +239,8 @@ private:
     }
 
     // Otherwise, this return operation has zero operands.
-    builder.create<ReturnOp>(location,
-                             expr ? ArrayRef(expr) : ArrayRef<mlir::Value>());
+    ReturnOp::create(builder, location,
+                     expr ? ArrayRef(expr) : ArrayRef<mlir::Value>());
     return mlir::success();
   }
 
@@ -284,7 +284,7 @@ private:
 
     // Build the MLIR op `toy.constant`. This invokes the `ConstantOp::build`
     // method.
-    return builder.create<ConstantOp>(loc(lit.loc()), type, dataAttribute);
+    return ConstantOp::create(builder, loc(lit.loc()), type, dataAttribute);
   }
 
   /// Recursive helper function to accumulate the data that compose an array
@@ -329,13 +329,13 @@ private:
                             "does not accept multiple arguments");
         return nullptr;
       }
-      return builder.create<TransposeOp>(location, operands[0]);
+      return TransposeOp::create(builder, location, operands[0]);
     }
 
     // Otherwise this is a call to a user-defined function. Calls to
     // user-defined functions are mapped to a custom call that takes the callee
     // name as an attribute.
-    return builder.create<GenericCallOp>(location, callee, operands);
+    return GenericCallOp::create(builder, location, callee, operands);
   }
 
   /// Emit a print expression. It emits specific operations for two builtins:
@@ -345,13 +345,13 @@ private:
     if (!arg)
       return mlir::failure();
 
-    builder.create<PrintOp>(loc(call.loc()), arg);
+    PrintOp::create(builder, loc(call.loc()), arg);
     return mlir::success();
   }
 
   /// Emit a constant for a single number (FIXME: semantic? broadcast?)
   mlir::Value mlirGen(NumberExprAST &num) {
-    return builder.create<ConstantOp>(loc(num.loc()), num.getValue());
+    return ConstantOp::create(builder, loc(num.loc()), num.getValue());
   }
 
   /// Dispatch codegen for the right expression subclass using RTTI.
@@ -395,8 +395,8 @@ private:
     // with specific shape, we emit a "reshape" operation. It will get
     // optimized out later as needed.
     if (!vardecl.getType().shape.empty()) {
-      value = builder.create<ReshapeOp>(loc(vardecl.loc()),
-                                        getType(vardecl.getType()), value);
+      value = ReshapeOp::create(builder, loc(vardecl.loc()),
+                                getType(vardecl.getType()), value);
     }
 
     // Register the value in the symbol table.
diff --git a/mlir/examples/toy/Ch5/mlir/ShapeInferencePass.cpp b/mlir/examples/toy/Ch5/mlir/ShapeInferencePass.cpp
index 2522abe..a552e1f0 100644
--- a/mlir/examples/toy/Ch5/mlir/ShapeInferencePass.cpp
+++ b/mlir/examples/toy/Ch5/mlir/ShapeInferencePass.cpp
@@ -23,7 +23,7 @@
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/Support/Casting.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 #include "llvm/Support/raw_ostream.h"
 #include <memory>
 
@@ -81,7 +81,7 @@ struct ShapeInferencePass
       opWorklist.erase(op);
 
       // Ask the operation to infer its output shapes.
-      LLVM_DEBUG(llvm::dbgs() << "Inferring shape for: " << *op << "\n");
+      LDBG() << "Inferring shape for: " << *op;
       if (auto shapeOp = dyn_cast<ShapeInference>(op)) {
         shapeOp.inferShapes();
       } else {
diff --git a/mlir/examples/toy/Ch6/include/toy/Ops.td b/mlir/examples/toy/Ch6/include/toy/Ops.td
index 63950f4..897b36d 100644
--- a/mlir/examples/toy/Ch6/include/toy/Ops.td
+++ b/mlir/examples/toy/Ch6/include/toy/Ops.td
@@ -72,7 +72,7 @@ def ConstantOp : Toy_Op<"constant", [Pure]> {
 
   // Add custom build methods for the constant operation. These method populates
   // the `state` that MLIR uses to create operations, i.e. these are used when
-  // using `builder.create<ConstantOp>(...)`.
+  // using `ConstantOp::create(builder, ...)`.
   let builders = [
     // Build a constant with a given constant tensor value.
     OpBuilder<(ins "DenseElementsAttr":$value), [{
diff --git a/mlir/examples/toy/Ch6/mlir/Dialect.cpp b/mlir/examples/toy/Ch6/mlir/Dialect.cpp
index fb7c742..69fb69f 100644
--- a/mlir/examples/toy/Ch6/mlir/Dialect.cpp
+++ b/mlir/examples/toy/Ch6/mlir/Dialect.cpp
@@ -91,7 +91,7 @@ struct ToyInlinerInterface : public DialectInlinerInterface {
   Operation *materializeCallConversion(OpBuilder &builder, Value input,
                                        Type resultType,
                                        Location conversionLoc) const final {
-    return builder.create<CastOp>(conversionLoc, resultType, input);
+    return CastOp::create(builder, conversionLoc, resultType, input);
   }
 };
 
@@ -206,7 +206,8 @@ void ConstantOp::print(mlir::OpAsmPrinter &printer) {
 llvm::LogicalResult ConstantOp::verify() {
   // If the return type of the constant is not an unranked tensor, the shape
   // must match the shape of the attribute holding the data.
-  auto resultType = llvm::dyn_cast<mlir::RankedTensorType>(getResult().getType());
+  auto resultType =
+      llvm::dyn_cast<mlir::RankedTensorType>(getResult().getType());
   if (!resultType)
     return success();
 
@@ -395,7 +396,8 @@ llvm::LogicalResult ReturnOp::verify() {
   auto resultType = results.front();
 
   // Check that the result type of the function matches the operand type.
-  if (inputType == resultType || llvm::isa<mlir::UnrankedTensorType>(inputType) ||
+  if (inputType == resultType ||
+      llvm::isa<mlir::UnrankedTensorType>(inputType) ||
       llvm::isa<mlir::UnrankedTensorType>(resultType))
     return mlir::success();
 
diff --git a/mlir/examples/toy/Ch6/mlir/LowerToAffineLoops.cpp b/mlir/examples/toy/Ch6/mlir/LowerToAffineLoops.cpp
index bf2bc43..2969d3a 100644
--- a/mlir/examples/toy/Ch6/mlir/LowerToAffineLoops.cpp
+++ b/mlir/examples/toy/Ch6/mlir/LowerToAffineLoops.cpp
@@ -44,7 +44,7 @@
 using namespace mlir;
 
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns
+// ToyToAffine Conversion Patterns
 //===----------------------------------------------------------------------===//
 
 /// Convert the given RankedTensorType into the corresponding MemRefType.
@@ -55,7 +55,7 @@ static MemRefType convertTensorToMemRef(RankedTensorType type) {
 /// Insert an allocation and deallocation for the given MemRefType.
 static Value insertAllocAndDealloc(MemRefType type, Location loc,
                                    PatternRewriter &rewriter) {
-  auto alloc = rewriter.create<memref::AllocOp>(loc, type);
+  auto alloc = memref::AllocOp::create(rewriter, loc, type);
 
   // Make sure to allocate at the beginning of the block.
   auto *parentBlock = alloc->getBlock();
@@ -63,21 +63,19 @@ static Value insertAllocAndDealloc(MemRefType type, Location loc,
 
   // Make sure to deallocate this alloc at the end of the block. This is fine
   // as toy functions have no control flow.
-  auto dealloc = rewriter.create<memref::DeallocOp>(loc, alloc);
+  auto dealloc = memref::DeallocOp::create(rewriter, loc, alloc);
   dealloc->moveBefore(&parentBlock->back());
   return alloc;
 }
 
 /// This defines the function type used to process an iteration of a lowered
-/// loop. It takes as input an OpBuilder, an range of memRefOperands
-/// corresponding to the operands of the input operation, and the range of loop
-/// induction variables for the iteration. It returns a value to store at the
-/// current index of the iteration.
-using LoopIterationFn = function_ref<Value(
-    OpBuilder &rewriter, ValueRange memRefOperands, ValueRange loopIvs)>;
-
-static void lowerOpToLoops(Operation *op, ValueRange operands,
-                           PatternRewriter &rewriter,
+/// loop. It takes as input an OpBuilder and the range of loop induction
+/// variables for the iteration. It returns a value to store at the current
+/// index of the iteration.
+using LoopIterationFn =
+    function_ref<Value(OpBuilder &rewriter, ValueRange loopIvs)>;
+
+static void lowerOpToLoops(Operation *op, PatternRewriter &rewriter,
                            LoopIterationFn processIteration) {
   auto tensorType = llvm::cast<RankedTensorType>((*op->result_type_begin()));
   auto loc = op->getLoc();
@@ -95,12 +93,12 @@ static void lowerOpToLoops(Operation *op, ValueRange operands,
   affine::buildAffineLoopNest(
       rewriter, loc, lowerBounds, tensorType.getShape(), steps,
       [&](OpBuilder &nestedBuilder, Location loc, ValueRange ivs) {
-        // Call the processing function with the rewriter, the memref operands,
-        // and the loop induction variables. This function will return the value
-        // to store at the current index.
-        Value valueToStore = processIteration(nestedBuilder, operands, ivs);
-        nestedBuilder.create<affine::AffineStoreOp>(loc, valueToStore, alloc,
-                                                    ivs);
+        // Call the processing function with the rewriter and the loop
+        // induction variables. This function will return the value to store at
+        // the current index.
+        Value valueToStore = processIteration(nestedBuilder, ivs);
+        affine::AffineStoreOp::create(nestedBuilder, loc, valueToStore, alloc,
+                                      ivs);
       });
 
   // Replace this operation with the generated alloc.
@@ -109,38 +107,30 @@ static void lowerOpToLoops(Operation *op, ValueRange operands,
 
 namespace {
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns: Binary operations
+// ToyToAffine Conversion Patterns: Binary operations
 //===----------------------------------------------------------------------===//
 
 template <typename BinaryOp, typename LoweredBinaryOp>
-struct BinaryOpLowering : public ConversionPattern {
-  BinaryOpLowering(MLIRContext *ctx)
-      : ConversionPattern(BinaryOp::getOperationName(), 1, ctx) {}
+struct BinaryOpLowering : public OpConversionPattern<BinaryOp> {
+  using OpConversionPattern<BinaryOp>::OpConversionPattern;
+  using OpAdaptor = typename OpConversionPattern<BinaryOp>::OpAdaptor;
 
   LogicalResult
-  matchAndRewrite(Operation *op, ArrayRef<Value> operands,
+  matchAndRewrite(BinaryOp op, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const final {
     auto loc = op->getLoc();
-    lowerOpToLoops(op, operands, rewriter,
-                   [loc](OpBuilder &builder, ValueRange memRefOperands,
-                         ValueRange loopIvs) {
-                     // Generate an adaptor for the remapped operands of the
-                     // BinaryOp. This allows for using the nice named accessors
-                     // that are generated by the ODS.
-                     typename BinaryOp::Adaptor binaryAdaptor(memRefOperands);
-
-                     // Generate loads for the element of 'lhs' and 'rhs' at the
-                     // inner loop.
-                     auto loadedLhs = builder.create<affine::AffineLoadOp>(
-                         loc, binaryAdaptor.getLhs(), loopIvs);
-                     auto loadedRhs = builder.create<affine::AffineLoadOp>(
-                         loc, binaryAdaptor.getRhs(), loopIvs);
-
-                     // Create the binary operation performed on the loaded
-                     // values.
-                     return builder.create<LoweredBinaryOp>(loc, loadedLhs,
-                                                            loadedRhs);
-                   });
+    lowerOpToLoops(op, rewriter, [&](OpBuilder &builder, ValueRange loopIvs) {
+      // Generate loads for the element of 'lhs' and 'rhs' at the
+      // inner loop.
+      auto loadedLhs =
+          affine::AffineLoadOp::create(builder, loc, adaptor.getLhs(), loopIvs);
+      auto loadedRhs =
+          affine::AffineLoadOp::create(builder, loc, adaptor.getRhs(), loopIvs);
+
+      // Create the binary operation performed on the loaded
+      // values.
+      return LoweredBinaryOp::create(builder, loc, loadedLhs, loadedRhs);
+    });
     return success();
   }
 };
@@ -148,14 +138,15 @@ using AddOpLowering = BinaryOpLowering<toy::AddOp, arith::AddFOp>;
 using MulOpLowering = BinaryOpLowering<toy::MulOp, arith::MulFOp>;
 
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns: Constant operations
+// ToyToAffine Conversion Patterns: Constant operations
 //===----------------------------------------------------------------------===//
 
-struct ConstantOpLowering : public OpRewritePattern<toy::ConstantOp> {
-  using OpRewritePattern<toy::ConstantOp>::OpRewritePattern;
+struct ConstantOpLowering : public OpConversionPattern<toy::ConstantOp> {
+  using OpConversionPattern<toy::ConstantOp>::OpConversionPattern;
 
-  LogicalResult matchAndRewrite(toy::ConstantOp op,
-                                PatternRewriter &rewriter) const final {
+  LogicalResult
+  matchAndRewrite(toy::ConstantOp op, OpAdaptor adaptor,
+                  ConversionPatternRewriter &rewriter) const final {
     DenseElementsAttr constantValue = op.getValue();
     Location loc = op.getLoc();
 
@@ -174,11 +165,11 @@ struct ConstantOpLowering : public OpRewritePattern<toy::ConstantOp> {
     if (!valueShape.empty()) {
       for (auto i : llvm::seq<int64_t>(0, *llvm::max_element(valueShape)))
         constantIndices.push_back(
-            rewriter.create<arith::ConstantIndexOp>(loc, i));
+            arith::ConstantIndexOp::create(rewriter, loc, i));
     } else {
       // This is the case of a tensor of rank 0.
       constantIndices.push_back(
-          rewriter.create<arith::ConstantIndexOp>(loc, 0));
+          arith::ConstantIndexOp::create(rewriter, loc, 0));
     }
 
     // The constant operation represents a multi-dimensional constant, so we
@@ -191,9 +182,9 @@ struct ConstantOpLowering : public OpRewritePattern<toy::ConstantOp> {
       // The last dimension is the base case of the recursion, at this point
       // we store the element at the given index.
       if (dimension == valueShape.size()) {
-        rewriter.create<affine::AffineStoreOp>(
-            loc, rewriter.create<arith::ConstantOp>(loc, *valueIt++), alloc,
-            llvm::ArrayRef(indices));
+        affine::AffineStoreOp::create(
+            rewriter, loc, arith::ConstantOp::create(rewriter, loc, *valueIt++),
+            alloc, llvm::ArrayRef(indices));
         return;
       }
 
@@ -216,7 +207,7 @@ struct ConstantOpLowering : public OpRewritePattern<toy::ConstantOp> {
 };
 
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns: Func operations
+// ToyToAffine Conversion Patterns: Func operations
 //===----------------------------------------------------------------------===//
 
 struct FuncOpLowering : public OpConversionPattern<toy::FuncOp> {
@@ -238,8 +229,8 @@ struct FuncOpLowering : public OpConversionPattern<toy::FuncOp> {
     }
 
     // Create a new non-toy function, with the same region.
-    auto func = rewriter.create<mlir::func::FuncOp>(op.getLoc(), op.getName(),
-                                                    op.getFunctionType());
+    auto func = mlir::func::FuncOp::create(rewriter, op.getLoc(), op.getName(),
+                                           op.getFunctionType());
     rewriter.inlineRegionBefore(op.getRegion(), func.getBody(), func.end());
     rewriter.eraseOp(op);
     return success();
@@ -247,7 +238,7 @@ struct FuncOpLowering : public OpConversionPattern<toy::FuncOp> {
 };
 
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns: Print operations
+// ToyToAffine Conversion Patterns: Print operations
 //===----------------------------------------------------------------------===//
 
 struct PrintOpLowering : public OpConversionPattern<toy::PrintOp> {
@@ -265,14 +256,15 @@ struct PrintOpLowering : public OpConversionPattern<toy::PrintOp> {
 };
 
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns: Return operations
+// ToyToAffine Conversion Patterns: Return operations
 //===----------------------------------------------------------------------===//
 
-struct ReturnOpLowering : public OpRewritePattern<toy::ReturnOp> {
-  using OpRewritePattern<toy::ReturnOp>::OpRewritePattern;
+struct ReturnOpLowering : public OpConversionPattern<toy::ReturnOp> {
+  using OpConversionPattern<toy::ReturnOp>::OpConversionPattern;
 
-  LogicalResult matchAndRewrite(toy::ReturnOp op,
-                                PatternRewriter &rewriter) const final {
+  LogicalResult
+  matchAndRewrite(toy::ReturnOp op, OpAdaptor adaptor,
+                  ConversionPatternRewriter &rewriter) const final {
     // During this lowering, we expect that all function calls have been
     // inlined.
     if (op.hasOperand())
@@ -285,32 +277,24 @@ struct ReturnOpLowering : public OpRewritePattern<toy::ReturnOp> {
 };
 
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns: Transpose operations
+// ToyToAffine Conversion Patterns: Transpose operations
 //===----------------------------------------------------------------------===//
 
-struct TransposeOpLowering : public ConversionPattern {
-  TransposeOpLowering(MLIRContext *ctx)
-      : ConversionPattern(toy::TransposeOp::getOperationName(), 1, ctx) {}
+struct TransposeOpLowering : public OpConversionPattern<toy::TransposeOp> {
+  using OpConversionPattern<toy::TransposeOp>::OpConversionPattern;
 
   LogicalResult
-  matchAndRewrite(Operation *op, ArrayRef<Value> operands,
+  matchAndRewrite(toy::TransposeOp op, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const final {
     auto loc = op->getLoc();
-    lowerOpToLoops(op, operands, rewriter,
-                   [loc](OpBuilder &builder, ValueRange memRefOperands,
-                         ValueRange loopIvs) {
-                     // Generate an adaptor for the remapped operands of the
-                     // TransposeOp. This allows for using the nice named
-                     // accessors that are generated by the ODS.
-                     toy::TransposeOpAdaptor transposeAdaptor(memRefOperands);
-                     Value input = transposeAdaptor.getInput();
-
-                     // Transpose the elements by generating a load from the
-                     // reverse indices.
-                     SmallVector<Value, 2> reverseIvs(llvm::reverse(loopIvs));
-                     return builder.create<affine::AffineLoadOp>(loc, input,
-                                                                 reverseIvs);
-                   });
+    lowerOpToLoops(op, rewriter, [&](OpBuilder &builder, ValueRange loopIvs) {
+      Value input = adaptor.getInput();
+
+      // Transpose the elements by generating a load from the
+      // reverse indices.
+      SmallVector<Value, 2> reverseIvs(llvm::reverse(loopIvs));
+      return affine::AffineLoadOp::create(builder, loc, input, reverseIvs);
+    });
     return success();
   }
 };
diff --git a/mlir/examples/toy/Ch6/mlir/LowerToLLVM.cpp b/mlir/examples/toy/Ch6/mlir/LowerToLLVM.cpp
index 22f75e0..987dfa1 100644
--- a/mlir/examples/toy/Ch6/mlir/LowerToLLVM.cpp
+++ b/mlir/examples/toy/Ch6/mlir/LowerToLLVM.cpp
@@ -55,19 +55,18 @@
 using namespace mlir;
 
 //===----------------------------------------------------------------------===//
-// ToyToLLVM RewritePatterns
+// ToyToLLVM Conversion Patterns
 //===----------------------------------------------------------------------===//
 
 namespace {
 /// Lowers `toy.print` to a loop nest calling `printf` on each of the individual
 /// elements of the array.
-class PrintOpLowering : public ConversionPattern {
+class PrintOpLowering : public OpConversionPattern<toy::PrintOp> {
 public:
-  explicit PrintOpLowering(MLIRContext *context)
-      : ConversionPattern(toy::PrintOp::getOperationName(), 1, context) {}
+  using OpConversionPattern<toy::PrintOp>::OpConversionPattern;
 
   LogicalResult
-  matchAndRewrite(Operation *op, ArrayRef<Value> operands,
+  matchAndRewrite(toy::PrintOp op, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const override {
     auto *context = rewriter.getContext();
     auto memRefType = llvm::cast<MemRefType>((*op->operand_type_begin()));
@@ -86,12 +85,12 @@ public:
     // Create a loop for each of the dimensions within the shape.
     SmallVector<Value, 4> loopIvs;
     for (unsigned i = 0, e = memRefShape.size(); i != e; ++i) {
-      auto lowerBound = rewriter.create<arith::ConstantIndexOp>(loc, 0);
+      auto lowerBound = arith::ConstantIndexOp::create(rewriter, loc, 0);
       auto upperBound =
-          rewriter.create<arith::ConstantIndexOp>(loc, memRefShape[i]);
-      auto step = rewriter.create<arith::ConstantIndexOp>(loc, 1);
+          arith::ConstantIndexOp::create(rewriter, loc, memRefShape[i]);
+      auto step = arith::ConstantIndexOp::create(rewriter, loc, 1);
       auto loop =
-          rewriter.create<scf::ForOp>(loc, lowerBound, upperBound, step);
+          scf::ForOp::create(rewriter, loc, lowerBound, upperBound, step);
       for (Operation &nested : make_early_inc_range(*loop.getBody()))
         rewriter.eraseOp(&nested);
       loopIvs.push_back(loop.getInductionVar());
@@ -101,19 +100,17 @@ public:
 
       // Insert a newline after each of the inner dimensions of the shape.
       if (i != e - 1)
-        rewriter.create<LLVM::CallOp>(loc, getPrintfType(context), printfRef,
-                                      newLineCst);
-      rewriter.create<scf::YieldOp>(loc);
+        LLVM::CallOp::create(rewriter, loc, getPrintfType(context), printfRef,
+                             newLineCst);
+      scf::YieldOp::create(rewriter, loc);
       rewriter.setInsertionPointToStart(loop.getBody());
     }
 
     // Generate a call to printf for the current element of the loop.
-    auto printOp = cast<toy::PrintOp>(op);
     auto elementLoad =
-        rewriter.create<memref::LoadOp>(loc, printOp.getInput(), loopIvs);
-    rewriter.create<LLVM::CallOp>(
-        loc, getPrintfType(context), printfRef,
-        ArrayRef<Value>({formatSpecifierCst, elementLoad}));
+        memref::LoadOp::create(rewriter, loc, op.getInput(), loopIvs);
+    LLVM::CallOp::create(rewriter, loc, getPrintfType(context), printfRef,
+                         ArrayRef<Value>({formatSpecifierCst, elementLoad}));
 
     // Notify the rewriter that this operation has been removed.
     rewriter.eraseOp(op);
@@ -142,8 +139,8 @@ private:
     // Insert the printf function into the body of the parent module.
     PatternRewriter::InsertionGuard insertGuard(rewriter);
     rewriter.setInsertionPointToStart(module.getBody());
-    rewriter.create<LLVM::LLVMFuncOp>(module.getLoc(), "printf",
-                                      getPrintfType(context));
+    LLVM::LLVMFuncOp::create(rewriter, module.getLoc(), "printf",
+                             getPrintfType(context));
     return SymbolRefAttr::get(context, "printf");
   }
 
@@ -159,19 +156,19 @@ private:
       builder.setInsertionPointToStart(module.getBody());
       auto type = LLVM::LLVMArrayType::get(
           IntegerType::get(builder.getContext(), 8), value.size());
-      global = builder.create<LLVM::GlobalOp>(loc, type, /*isConstant=*/true,
-                                              LLVM::Linkage::Internal, name,
-                                              builder.getStringAttr(value),
-                                              /*alignment=*/0);
+      global = LLVM::GlobalOp::create(builder, loc, type, /*isConstant=*/true,
+                                      LLVM::Linkage::Internal, name,
+                                      builder.getStringAttr(value),
+                                      /*alignment=*/0);
     }
 
     // Get the pointer to the first character in the global string.
-    Value globalPtr = builder.create<LLVM::AddressOfOp>(loc, global);
-    Value cst0 = builder.create<LLVM::ConstantOp>(loc, builder.getI64Type(),
-                                                  builder.getIndexAttr(0));
-    return builder.create<LLVM::GEPOp>(
-        loc, LLVM::LLVMPointerType::get(builder.getContext()), global.getType(),
-        globalPtr, ArrayRef<Value>({cst0, cst0}));
+    Value globalPtr = LLVM::AddressOfOp::create(builder, loc, global);
+    Value cst0 = LLVM::ConstantOp::create(builder, loc, builder.getI64Type(),
+                                          builder.getIndexAttr(0));
+    return LLVM::GEPOp::create(
+        builder, loc, LLVM::LLVMPointerType::get(builder.getContext()),
+        global.getType(), globalPtr, ArrayRef<Value>({cst0, cst0}));
   }
 };
 } // namespace
diff --git a/mlir/examples/toy/Ch6/mlir/MLIRGen.cpp b/mlir/examples/toy/Ch6/mlir/MLIRGen.cpp
index 9371815..7d676f1 100644
--- a/mlir/examples/toy/Ch6/mlir/MLIRGen.cpp
+++ b/mlir/examples/toy/Ch6/mlir/MLIRGen.cpp
@@ -121,8 +121,8 @@ private:
     llvm::SmallVector<mlir::Type, 4> argTypes(proto.getArgs().size(),
                                               getType(VarType{}));
     auto funcType = builder.getFunctionType(argTypes, /*results=*/{});
-    return builder.create<mlir::toy::FuncOp>(location, proto.getName(),
-                                             funcType);
+    return mlir::toy::FuncOp::create(builder, location, proto.getName(),
+                                     funcType);
   }
 
   /// Emit a new function and add it to the MLIR module.
@@ -166,7 +166,7 @@ private:
     if (!entryBlock.empty())
       returnOp = dyn_cast<ReturnOp>(entryBlock.back());
     if (!returnOp) {
-      builder.create<ReturnOp>(loc(funcAST.getProto()->loc()));
+      ReturnOp::create(builder, loc(funcAST.getProto()->loc()));
     } else if (returnOp.hasOperand()) {
       // Otherwise, if this return operation has an operand then add a result to
       // the function.
@@ -206,9 +206,9 @@ private:
     // support '+' and '*'.
     switch (binop.getOp()) {
     case '+':
-      return builder.create<AddOp>(location, lhs, rhs);
+      return AddOp::create(builder, location, lhs, rhs);
     case '*':
-      return builder.create<MulOp>(location, lhs, rhs);
+      return MulOp::create(builder, location, lhs, rhs);
     }
 
     emitError(location, "invalid binary operator '") << binop.getOp() << "'";
@@ -239,8 +239,8 @@ private:
     }
 
     // Otherwise, this return operation has zero operands.
-    builder.create<ReturnOp>(location,
-                             expr ? ArrayRef(expr) : ArrayRef<mlir::Value>());
+    ReturnOp::create(builder, location,
+                     expr ? ArrayRef(expr) : ArrayRef<mlir::Value>());
     return mlir::success();
   }
 
@@ -284,7 +284,7 @@ private:
 
     // Build the MLIR op `toy.constant`. This invokes the `ConstantOp::build`
     // method.
-    return builder.create<ConstantOp>(loc(lit.loc()), type, dataAttribute);
+    return ConstantOp::create(builder, loc(lit.loc()), type, dataAttribute);
   }
 
   /// Recursive helper function to accumulate the data that compose an array
@@ -329,13 +329,13 @@ private:
                             "does not accept multiple arguments");
         return nullptr;
       }
-      return builder.create<TransposeOp>(location, operands[0]);
+      return TransposeOp::create(builder, location, operands[0]);
     }
 
     // Otherwise this is a call to a user-defined function. Calls to
     // user-defined functions are mapped to a custom call that takes the callee
     // name as an attribute.
-    return builder.create<GenericCallOp>(location, callee, operands);
+    return GenericCallOp::create(builder, location, callee, operands);
   }
 
   /// Emit a print expression. It emits specific operations for two builtins:
@@ -345,13 +345,13 @@ private:
     if (!arg)
       return mlir::failure();
 
-    builder.create<PrintOp>(loc(call.loc()), arg);
+    PrintOp::create(builder, loc(call.loc()), arg);
     return mlir::success();
   }
 
   /// Emit a constant for a single number (FIXME: semantic? broadcast?)
   mlir::Value mlirGen(NumberExprAST &num) {
-    return builder.create<ConstantOp>(loc(num.loc()), num.getValue());
+    return ConstantOp::create(builder, loc(num.loc()), num.getValue());
   }
 
   /// Dispatch codegen for the right expression subclass using RTTI.
@@ -395,8 +395,8 @@ private:
     // with specific shape, we emit a "reshape" operation. It will get
     // optimized out later as needed.
     if (!vardecl.getType().shape.empty()) {
-      value = builder.create<ReshapeOp>(loc(vardecl.loc()),
-                                        getType(vardecl.getType()), value);
+      value = ReshapeOp::create(builder, loc(vardecl.loc()),
+                                getType(vardecl.getType()), value);
     }
 
     // Register the value in the symbol table.
diff --git a/mlir/examples/toy/Ch6/mlir/ShapeInferencePass.cpp b/mlir/examples/toy/Ch6/mlir/ShapeInferencePass.cpp
index 2522abe..a552e1f0 100644
--- a/mlir/examples/toy/Ch6/mlir/ShapeInferencePass.cpp
+++ b/mlir/examples/toy/Ch6/mlir/ShapeInferencePass.cpp
@@ -23,7 +23,7 @@
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/Support/Casting.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 #include "llvm/Support/raw_ostream.h"
 #include <memory>
 
@@ -81,7 +81,7 @@ struct ShapeInferencePass
       opWorklist.erase(op);
 
       // Ask the operation to infer its output shapes.
-      LLVM_DEBUG(llvm::dbgs() << "Inferring shape for: " << *op << "\n");
+      LDBG() << "Inferring shape for: " << *op;
       if (auto shapeOp = dyn_cast<ShapeInference>(op)) {
         shapeOp.inferShapes();
       } else {
diff --git a/mlir/examples/toy/Ch7/include/toy/Ops.td b/mlir/examples/toy/Ch7/include/toy/Ops.td
index bdf8ad0b..9151396 100644
--- a/mlir/examples/toy/Ch7/include/toy/Ops.td
+++ b/mlir/examples/toy/Ch7/include/toy/Ops.td
@@ -93,7 +93,7 @@ def ConstantOp : Toy_Op<"constant",
 
   // Add custom build methods for the constant operation. These method populates
   // the `state` that MLIR uses to create operations, i.e. these are used when
-  // using `builder.create<ConstantOp>(...)`.
+  // using `ConstantOp::create(builder, ...)`.
   let builders = [
     // Build a constant with a given constant tensor value.
     OpBuilder<(ins "DenseElementsAttr":$value), [{
diff --git a/mlir/examples/toy/Ch7/mlir/Dialect.cpp b/mlir/examples/toy/Ch7/mlir/Dialect.cpp
index 52881db..4d2f063 100644
--- a/mlir/examples/toy/Ch7/mlir/Dialect.cpp
+++ b/mlir/examples/toy/Ch7/mlir/Dialect.cpp
@@ -97,7 +97,7 @@ struct ToyInlinerInterface : public DialectInlinerInterface {
   Operation *materializeCallConversion(OpBuilder &builder, Value input,
                                        Type resultType,
                                        Location conversionLoc) const final {
-    return builder.create<CastOp>(conversionLoc, resultType, input);
+    return CastOp::create(builder, conversionLoc, resultType, input);
   }
 };
 
@@ -429,7 +429,8 @@ llvm::LogicalResult ReturnOp::verify() {
   auto resultType = results.front();
 
   // Check that the result type of the function matches the operand type.
-  if (inputType == resultType || llvm::isa<mlir::UnrankedTensorType>(inputType) ||
+  if (inputType == resultType ||
+      llvm::isa<mlir::UnrankedTensorType>(inputType) ||
       llvm::isa<mlir::UnrankedTensorType>(resultType))
     return mlir::success();
 
@@ -657,8 +658,8 @@ mlir::Operation *ToyDialect::materializeConstant(mlir::OpBuilder &builder,
                                                  mlir::Type type,
                                                  mlir::Location loc) {
   if (llvm::isa<StructType>(type))
-    return builder.create<StructConstantOp>(loc, type,
-                                            llvm::cast<mlir::ArrayAttr>(value));
-  return builder.create<ConstantOp>(loc, type,
-                                    llvm::cast<mlir::DenseElementsAttr>(value));
+    return StructConstantOp::create(builder, loc, type,
+                                    llvm::cast<mlir::ArrayAttr>(value));
+  return ConstantOp::create(builder, loc, type,
+                            llvm::cast<mlir::DenseElementsAttr>(value));
 }
diff --git a/mlir/examples/toy/Ch7/mlir/LowerToAffineLoops.cpp b/mlir/examples/toy/Ch7/mlir/LowerToAffineLoops.cpp
index bf2bc43..cbe4236 100644
--- a/mlir/examples/toy/Ch7/mlir/LowerToAffineLoops.cpp
+++ b/mlir/examples/toy/Ch7/mlir/LowerToAffineLoops.cpp
@@ -44,7 +44,7 @@
 using namespace mlir;
 
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns
+// ToyToAffine Conversion Patterns
 //===----------------------------------------------------------------------===//
 
 /// Convert the given RankedTensorType into the corresponding MemRefType.
@@ -55,7 +55,7 @@ static MemRefType convertTensorToMemRef(RankedTensorType type) {
 /// Insert an allocation and deallocation for the given MemRefType.
 static Value insertAllocAndDealloc(MemRefType type, Location loc,
                                    PatternRewriter &rewriter) {
-  auto alloc = rewriter.create<memref::AllocOp>(loc, type);
+  auto alloc = memref::AllocOp::create(rewriter, loc, type);
 
   // Make sure to allocate at the beginning of the block.
   auto *parentBlock = alloc->getBlock();
@@ -63,21 +63,19 @@ static Value insertAllocAndDealloc(MemRefType type, Location loc,
 
   // Make sure to deallocate this alloc at the end of the block. This is fine
   // as toy functions have no control flow.
-  auto dealloc = rewriter.create<memref::DeallocOp>(loc, alloc);
+  auto dealloc = memref::DeallocOp::create(rewriter, loc, alloc);
   dealloc->moveBefore(&parentBlock->back());
   return alloc;
 }
 
 /// This defines the function type used to process an iteration of a lowered
-/// loop. It takes as input an OpBuilder, an range of memRefOperands
-/// corresponding to the operands of the input operation, and the range of loop
-/// induction variables for the iteration. It returns a value to store at the
-/// current index of the iteration.
-using LoopIterationFn = function_ref<Value(
-    OpBuilder &rewriter, ValueRange memRefOperands, ValueRange loopIvs)>;
-
-static void lowerOpToLoops(Operation *op, ValueRange operands,
-                           PatternRewriter &rewriter,
+/// loop. It takes as input an OpBuilder and the range of loop induction
+/// variables for the iteration. It returns a value to store at the current
+/// index of the iteration.
+using LoopIterationFn =
+    function_ref<Value(OpBuilder &rewriter, ValueRange loopIvs)>;
+
+static void lowerOpToLoops(Operation *op, PatternRewriter &rewriter,
                            LoopIterationFn processIteration) {
   auto tensorType = llvm::cast<RankedTensorType>((*op->result_type_begin()));
   auto loc = op->getLoc();
@@ -95,12 +93,12 @@ static void lowerOpToLoops(Operation *op, ValueRange operands,
   affine::buildAffineLoopNest(
       rewriter, loc, lowerBounds, tensorType.getShape(), steps,
       [&](OpBuilder &nestedBuilder, Location loc, ValueRange ivs) {
-        // Call the processing function with the rewriter, the memref operands,
+        // Call the processing function with the rewriter
         // and the loop induction variables. This function will return the value
         // to store at the current index.
-        Value valueToStore = processIteration(nestedBuilder, operands, ivs);
-        nestedBuilder.create<affine::AffineStoreOp>(loc, valueToStore, alloc,
-                                                    ivs);
+        Value valueToStore = processIteration(nestedBuilder, ivs);
+        affine::AffineStoreOp::create(nestedBuilder, loc, valueToStore, alloc,
+                                      ivs);
       });
 
   // Replace this operation with the generated alloc.
@@ -109,38 +107,30 @@ static void lowerOpToLoops(Operation *op, ValueRange operands,
 
 namespace {
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns: Binary operations
+// ToyToAffine Conversion Patterns: Binary operations
 //===----------------------------------------------------------------------===//
 
 template <typename BinaryOp, typename LoweredBinaryOp>
-struct BinaryOpLowering : public ConversionPattern {
-  BinaryOpLowering(MLIRContext *ctx)
-      : ConversionPattern(BinaryOp::getOperationName(), 1, ctx) {}
+struct BinaryOpLowering : public OpConversionPattern<BinaryOp> {
+  using OpConversionPattern<BinaryOp>::OpConversionPattern;
+  using OpAdaptor = typename OpConversionPattern<BinaryOp>::OpAdaptor;
 
   LogicalResult
-  matchAndRewrite(Operation *op, ArrayRef<Value> operands,
+  matchAndRewrite(BinaryOp op, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const final {
     auto loc = op->getLoc();
-    lowerOpToLoops(op, operands, rewriter,
-                   [loc](OpBuilder &builder, ValueRange memRefOperands,
-                         ValueRange loopIvs) {
-                     // Generate an adaptor for the remapped operands of the
-                     // BinaryOp. This allows for using the nice named accessors
-                     // that are generated by the ODS.
-                     typename BinaryOp::Adaptor binaryAdaptor(memRefOperands);
-
-                     // Generate loads for the element of 'lhs' and 'rhs' at the
-                     // inner loop.
-                     auto loadedLhs = builder.create<affine::AffineLoadOp>(
-                         loc, binaryAdaptor.getLhs(), loopIvs);
-                     auto loadedRhs = builder.create<affine::AffineLoadOp>(
-                         loc, binaryAdaptor.getRhs(), loopIvs);
-
-                     // Create the binary operation performed on the loaded
-                     // values.
-                     return builder.create<LoweredBinaryOp>(loc, loadedLhs,
-                                                            loadedRhs);
-                   });
+    lowerOpToLoops(op, rewriter, [&](OpBuilder &builder, ValueRange loopIvs) {
+      // Generate loads for the element of 'lhs' and 'rhs' at the
+      // inner loop.
+      auto loadedLhs =
+          affine::AffineLoadOp::create(builder, loc, adaptor.getLhs(), loopIvs);
+      auto loadedRhs =
+          affine::AffineLoadOp::create(builder, loc, adaptor.getRhs(), loopIvs);
+
+      // Create the binary operation performed on the loaded
+      // values.
+      return LoweredBinaryOp::create(builder, loc, loadedLhs, loadedRhs);
+    });
     return success();
   }
 };
@@ -148,14 +138,15 @@ using AddOpLowering = BinaryOpLowering<toy::AddOp, arith::AddFOp>;
 using MulOpLowering = BinaryOpLowering<toy::MulOp, arith::MulFOp>;
 
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns: Constant operations
+// ToyToAffine Conversion Patterns: Constant operations
 //===----------------------------------------------------------------------===//
 
-struct ConstantOpLowering : public OpRewritePattern<toy::ConstantOp> {
-  using OpRewritePattern<toy::ConstantOp>::OpRewritePattern;
+struct ConstantOpLowering : public OpConversionPattern<toy::ConstantOp> {
+  using OpConversionPattern<toy::ConstantOp>::OpConversionPattern;
 
-  LogicalResult matchAndRewrite(toy::ConstantOp op,
-                                PatternRewriter &rewriter) const final {
+  LogicalResult
+  matchAndRewrite(toy::ConstantOp op, OpAdaptor adaptor,
+                  ConversionPatternRewriter &rewriter) const final {
     DenseElementsAttr constantValue = op.getValue();
     Location loc = op.getLoc();
 
@@ -174,11 +165,11 @@ struct ConstantOpLowering : public OpRewritePattern<toy::ConstantOp> {
     if (!valueShape.empty()) {
       for (auto i : llvm::seq<int64_t>(0, *llvm::max_element(valueShape)))
         constantIndices.push_back(
-            rewriter.create<arith::ConstantIndexOp>(loc, i));
+            arith::ConstantIndexOp::create(rewriter, loc, i));
     } else {
       // This is the case of a tensor of rank 0.
       constantIndices.push_back(
-          rewriter.create<arith::ConstantIndexOp>(loc, 0));
+          arith::ConstantIndexOp::create(rewriter, loc, 0));
     }
 
     // The constant operation represents a multi-dimensional constant, so we
@@ -191,9 +182,9 @@ struct ConstantOpLowering : public OpRewritePattern<toy::ConstantOp> {
       // The last dimension is the base case of the recursion, at this point
       // we store the element at the given index.
       if (dimension == valueShape.size()) {
-        rewriter.create<affine::AffineStoreOp>(
-            loc, rewriter.create<arith::ConstantOp>(loc, *valueIt++), alloc,
-            llvm::ArrayRef(indices));
+        affine::AffineStoreOp::create(
+            rewriter, loc, arith::ConstantOp::create(rewriter, loc, *valueIt++),
+            alloc, llvm::ArrayRef(indices));
         return;
       }
 
@@ -216,7 +207,7 @@ struct ConstantOpLowering : public OpRewritePattern<toy::ConstantOp> {
 };
 
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns: Func operations
+// ToyToAffine Conversion Patterns: Func operations
 //===----------------------------------------------------------------------===//
 
 struct FuncOpLowering : public OpConversionPattern<toy::FuncOp> {
@@ -238,8 +229,8 @@ struct FuncOpLowering : public OpConversionPattern<toy::FuncOp> {
     }
 
     // Create a new non-toy function, with the same region.
-    auto func = rewriter.create<mlir::func::FuncOp>(op.getLoc(), op.getName(),
-                                                    op.getFunctionType());
+    auto func = mlir::func::FuncOp::create(rewriter, op.getLoc(), op.getName(),
+                                           op.getFunctionType());
     rewriter.inlineRegionBefore(op.getRegion(), func.getBody(), func.end());
     rewriter.eraseOp(op);
     return success();
@@ -247,7 +238,7 @@ struct FuncOpLowering : public OpConversionPattern<toy::FuncOp> {
 };
 
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns: Print operations
+// ToyToAffine Conversion Patterns: Print operations
 //===----------------------------------------------------------------------===//
 
 struct PrintOpLowering : public OpConversionPattern<toy::PrintOp> {
@@ -265,14 +256,15 @@ struct PrintOpLowering : public OpConversionPattern<toy::PrintOp> {
 };
 
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns: Return operations
+// ToyToAffine Conversion Patterns: Return operations
 //===----------------------------------------------------------------------===//
 
-struct ReturnOpLowering : public OpRewritePattern<toy::ReturnOp> {
-  using OpRewritePattern<toy::ReturnOp>::OpRewritePattern;
+struct ReturnOpLowering : public OpConversionPattern<toy::ReturnOp> {
+  using OpConversionPattern<toy::ReturnOp>::OpConversionPattern;
 
-  LogicalResult matchAndRewrite(toy::ReturnOp op,
-                                PatternRewriter &rewriter) const final {
+  LogicalResult
+  matchAndRewrite(toy::ReturnOp op, OpAdaptor adaptor,
+                  ConversionPatternRewriter &rewriter) const final {
     // During this lowering, we expect that all function calls have been
     // inlined.
     if (op.hasOperand())
@@ -285,32 +277,24 @@ struct ReturnOpLowering : public OpRewritePattern<toy::ReturnOp> {
 };
 
 //===----------------------------------------------------------------------===//
-// ToyToAffine RewritePatterns: Transpose operations
+// ToyToAffine Conversion Patterns: Transpose operations
 //===----------------------------------------------------------------------===//
 
-struct TransposeOpLowering : public ConversionPattern {
-  TransposeOpLowering(MLIRContext *ctx)
-      : ConversionPattern(toy::TransposeOp::getOperationName(), 1, ctx) {}
+struct TransposeOpLowering : public OpConversionPattern<toy::TransposeOp> {
+  using OpConversionPattern<toy::TransposeOp>::OpConversionPattern;
 
   LogicalResult
-  matchAndRewrite(Operation *op, ArrayRef<Value> operands,
+  matchAndRewrite(toy::TransposeOp op, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const final {
     auto loc = op->getLoc();
-    lowerOpToLoops(op, operands, rewriter,
-                   [loc](OpBuilder &builder, ValueRange memRefOperands,
-                         ValueRange loopIvs) {
-                     // Generate an adaptor for the remapped operands of the
-                     // TransposeOp. This allows for using the nice named
-                     // accessors that are generated by the ODS.
-                     toy::TransposeOpAdaptor transposeAdaptor(memRefOperands);
-                     Value input = transposeAdaptor.getInput();
-
-                     // Transpose the elements by generating a load from the
-                     // reverse indices.
-                     SmallVector<Value, 2> reverseIvs(llvm::reverse(loopIvs));
-                     return builder.create<affine::AffineLoadOp>(loc, input,
-                                                                 reverseIvs);
-                   });
+    lowerOpToLoops(op, rewriter, [&](OpBuilder &builder, ValueRange loopIvs) {
+      Value input = adaptor.getInput();
+
+      // Transpose the elements by generating a load from the
+      // reverse indices.
+      SmallVector<Value, 2> reverseIvs(llvm::reverse(loopIvs));
+      return affine::AffineLoadOp::create(builder, loc, input, reverseIvs);
+    });
     return success();
   }
 };
diff --git a/mlir/examples/toy/Ch7/mlir/LowerToLLVM.cpp b/mlir/examples/toy/Ch7/mlir/LowerToLLVM.cpp
index 54eeb27..8b48a8f 100644
--- a/mlir/examples/toy/Ch7/mlir/LowerToLLVM.cpp
+++ b/mlir/examples/toy/Ch7/mlir/LowerToLLVM.cpp
@@ -55,19 +55,18 @@
 using namespace mlir;
 
 //===----------------------------------------------------------------------===//
-// ToyToLLVM RewritePatterns
+// ToyToLLVM Conversion Patterns
 //===----------------------------------------------------------------------===//
 
 namespace {
 /// Lowers `toy.print` to a loop nest calling `printf` on each of the individual
 /// elements of the array.
-class PrintOpLowering : public ConversionPattern {
+class PrintOpLowering : public OpConversionPattern<toy::PrintOp> {
 public:
-  explicit PrintOpLowering(MLIRContext *context)
-      : ConversionPattern(toy::PrintOp::getOperationName(), 1, context) {}
+  using OpConversionPattern<toy::PrintOp>::OpConversionPattern;
 
   LogicalResult
-  matchAndRewrite(Operation *op, ArrayRef<Value> operands,
+  matchAndRewrite(toy::PrintOp op, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const override {
     auto *context = rewriter.getContext();
     auto memRefType = llvm::cast<MemRefType>((*op->operand_type_begin()));
@@ -86,12 +85,12 @@ public:
     // Create a loop for each of the dimensions within the shape.
     SmallVector<Value, 4> loopIvs;
     for (unsigned i = 0, e = memRefShape.size(); i != e; ++i) {
-      auto lowerBound = rewriter.create<arith::ConstantIndexOp>(loc, 0);
+      auto lowerBound = arith::ConstantIndexOp::create(rewriter, loc, 0);
       auto upperBound =
-          rewriter.create<arith::ConstantIndexOp>(loc, memRefShape[i]);
-      auto step = rewriter.create<arith::ConstantIndexOp>(loc, 1);
+          arith::ConstantIndexOp::create(rewriter, loc, memRefShape[i]);
+      auto step = arith::ConstantIndexOp::create(rewriter, loc, 1);
       auto loop =
-          rewriter.create<scf::ForOp>(loc, lowerBound, upperBound, step);
+          scf::ForOp::create(rewriter, loc, lowerBound, upperBound, step);
       for (Operation &nested : make_early_inc_range(*loop.getBody()))
         rewriter.eraseOp(&nested);
       loopIvs.push_back(loop.getInductionVar());
@@ -101,19 +100,17 @@ public:
 
       // Insert a newline after each of the inner dimensions of the shape.
       if (i != e - 1)
-        rewriter.create<LLVM::CallOp>(loc, getPrintfType(context), printfRef,
-                                      newLineCst);
-      rewriter.create<scf::YieldOp>(loc);
+        LLVM::CallOp::create(rewriter, loc, getPrintfType(context), printfRef,
+                             newLineCst);
+      scf::YieldOp::create(rewriter, loc);
       rewriter.setInsertionPointToStart(loop.getBody());
     }
 
     // Generate a call to printf for the current element of the loop.
-    auto printOp = cast<toy::PrintOp>(op);
     auto elementLoad =
-        rewriter.create<memref::LoadOp>(loc, printOp.getInput(), loopIvs);
-    rewriter.create<LLVM::CallOp>(
-        loc, getPrintfType(context), printfRef,
-        ArrayRef<Value>({formatSpecifierCst, elementLoad}));
+        memref::LoadOp::create(rewriter, loc, op.getInput(), loopIvs);
+    LLVM::CallOp::create(rewriter, loc, getPrintfType(context), printfRef,
+                         ArrayRef<Value>({formatSpecifierCst, elementLoad}));
 
     // Notify the rewriter that this operation has been removed.
     rewriter.eraseOp(op);
@@ -142,8 +139,8 @@ private:
     // Insert the printf function into the body of the parent module.
     PatternRewriter::InsertionGuard insertGuard(rewriter);
     rewriter.setInsertionPointToStart(module.getBody());
-    rewriter.create<LLVM::LLVMFuncOp>(module.getLoc(), "printf",
-                                      getPrintfType(context));
+    LLVM::LLVMFuncOp::create(rewriter, module.getLoc(), "printf",
+                             getPrintfType(context));
     return SymbolRefAttr::get(context, "printf");
   }
 
@@ -159,19 +156,19 @@ private:
       builder.setInsertionPointToStart(module.getBody());
       auto type = LLVM::LLVMArrayType::get(
           IntegerType::get(builder.getContext(), 8), value.size());
-      global = builder.create<LLVM::GlobalOp>(loc, type, /*isConstant=*/true,
-                                              LLVM::Linkage::Internal, name,
-                                              builder.getStringAttr(value),
-                                              /*alignment=*/0);
+      global = LLVM::GlobalOp::create(builder, loc, type, /*isConstant=*/true,
+                                      LLVM::Linkage::Internal, name,
+                                      builder.getStringAttr(value),
+                                      /*alignment=*/0);
     }
 
     // Get the pointer to the first character in the global string.
-    Value globalPtr = builder.create<LLVM::AddressOfOp>(loc, global);
-    Value cst0 = builder.create<LLVM::ConstantOp>(loc, builder.getI64Type(),
-                                                  builder.getIndexAttr(0));
-    return builder.create<LLVM::GEPOp>(
-        loc, LLVM::LLVMPointerType::get(builder.getContext()), global.getType(),
-        globalPtr, ArrayRef<Value>({cst0, cst0}));
+    Value globalPtr = LLVM::AddressOfOp::create(builder, loc, global);
+    Value cst0 = LLVM::ConstantOp::create(builder, loc, builder.getI64Type(),
+                                          builder.getIndexAttr(0));
+    return LLVM::GEPOp::create(
+        builder, loc, LLVM::LLVMPointerType::get(builder.getContext()),
+        global.getType(), globalPtr, ArrayRef<Value>({cst0, cst0}));
   }
 };
 } // namespace
diff --git a/mlir/examples/toy/Ch7/mlir/MLIRGen.cpp b/mlir/examples/toy/Ch7/mlir/MLIRGen.cpp
index 2490f17..75dbc91 100644
--- a/mlir/examples/toy/Ch7/mlir/MLIRGen.cpp
+++ b/mlir/examples/toy/Ch7/mlir/MLIRGen.cpp
@@ -183,8 +183,8 @@ private:
       argTypes.push_back(type);
     }
     auto funcType = builder.getFunctionType(argTypes, /*results=*/{});
-    return builder.create<mlir::toy::FuncOp>(location, proto.getName(),
-                                             funcType);
+    return mlir::toy::FuncOp::create(builder, location, proto.getName(),
+                                     funcType);
   }
 
   /// Emit a new function and add it to the MLIR module.
@@ -227,7 +227,7 @@ private:
     if (!entryBlock.empty())
       returnOp = dyn_cast<ReturnOp>(entryBlock.back());
     if (!returnOp) {
-      builder.create<ReturnOp>(loc(funcAST.getProto()->loc()));
+      ReturnOp::create(builder, loc(funcAST.getProto()->loc()));
     } else if (returnOp.hasOperand()) {
       // Otherwise, if this return operation has an operand then add a result to
       // the function.
@@ -333,7 +333,7 @@ private:
         emitError(location, "invalid access into struct expression");
         return nullptr;
       }
-      return builder.create<StructAccessOp>(location, lhs, *accessIndex);
+      return StructAccessOp::create(builder, location, lhs, *accessIndex);
     }
 
     // Otherwise, this is a normal binary op.
@@ -345,9 +345,9 @@ private:
     // support '+' and '*'.
     switch (binop.getOp()) {
     case '+':
-      return builder.create<AddOp>(location, lhs, rhs);
+      return AddOp::create(builder, location, lhs, rhs);
     case '*':
-      return builder.create<MulOp>(location, lhs, rhs);
+      return MulOp::create(builder, location, lhs, rhs);
     }
 
     emitError(location, "invalid binary operator '") << binop.getOp() << "'";
@@ -378,8 +378,8 @@ private:
     }
 
     // Otherwise, this return operation has zero operands.
-    builder.create<ReturnOp>(location,
-                             expr ? ArrayRef(expr) : ArrayRef<mlir::Value>());
+    ReturnOp::create(builder, location,
+                     expr ? ArrayRef(expr) : ArrayRef<mlir::Value>());
     return mlir::success();
   }
 
@@ -464,7 +464,7 @@ private:
 
     // Build the MLIR op `toy.constant`. This invokes the `ConstantOp::build`
     // method.
-    return builder.create<ConstantOp>(loc(lit.loc()), type, dataAttribute);
+    return ConstantOp::create(builder, loc(lit.loc()), type, dataAttribute);
   }
 
   /// Emit a struct literal. It will be emitted as an array of
@@ -477,7 +477,8 @@ private:
 
     // Build the MLIR op `toy.struct_constant`. This invokes the
     // `StructConstantOp::build` method.
-    return builder.create<StructConstantOp>(loc(lit.loc()), dataType, dataAttr);
+    return StructConstantOp::create(builder, loc(lit.loc()), dataType,
+                                    dataAttr);
   }
 
   /// Recursive helper function to accumulate the data that compose an array
@@ -522,7 +523,7 @@ private:
                             "does not accept multiple arguments");
         return nullptr;
       }
-      return builder.create<TransposeOp>(location, operands[0]);
+      return TransposeOp::create(builder, location, operands[0]);
     }
 
     // Otherwise this is a call to a user-defined function. Calls to
@@ -534,8 +535,9 @@ private:
       return nullptr;
     }
     mlir::toy::FuncOp calledFunc = calledFuncIt->second;
-    return builder.create<GenericCallOp>(
-        location, calledFunc.getFunctionType().getResult(0), callee, operands);
+    return GenericCallOp::create(builder, location,
+                                 calledFunc.getFunctionType().getResult(0),
+                                 callee, operands);
   }
 
   /// Emit a print expression. It emits specific operations for two builtins:
@@ -545,13 +547,13 @@ private:
     if (!arg)
       return mlir::failure();
 
-    builder.create<PrintOp>(loc(call.loc()), arg);
+    PrintOp::create(builder, loc(call.loc()), arg);
     return mlir::success();
   }
 
   /// Emit a constant for a single number (FIXME: semantic? broadcast?)
   mlir::Value mlirGen(NumberExprAST &num) {
-    return builder.create<ConstantOp>(loc(num.loc()), num.getValue());
+    return ConstantOp::create(builder, loc(num.loc()), num.getValue());
   }
 
   /// Dispatch codegen for the right expression subclass using RTTI.
@@ -613,8 +615,8 @@ private:
       // declared with specific shape, we emit a "reshape" operation. It will
       // get optimized out later as needed.
     } else if (!varType.shape.empty()) {
-      value = builder.create<ReshapeOp>(loc(vardecl.loc()),
-                                        getType(varType.shape), value);
+      value = ReshapeOp::create(builder, loc(vardecl.loc()),
+                                getType(varType.shape), value);
     }
 
     // Register the value in the symbol table.
diff --git a/mlir/examples/toy/Ch7/mlir/ShapeInferencePass.cpp b/mlir/examples/toy/Ch7/mlir/ShapeInferencePass.cpp
index 2522abe..a552e1f0 100644
--- a/mlir/examples/toy/Ch7/mlir/ShapeInferencePass.cpp
+++ b/mlir/examples/toy/Ch7/mlir/ShapeInferencePass.cpp
@@ -23,7 +23,7 @@
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/Support/Casting.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 #include "llvm/Support/raw_ostream.h"
 #include <memory>
 
@@ -81,7 +81,7 @@ struct ShapeInferencePass
       opWorklist.erase(op);
 
       // Ask the operation to infer its output shapes.
-      LLVM_DEBUG(llvm::dbgs() << "Inferring shape for: " << *op << "\n");
+      LDBG() << "Inferring shape for: " << *op;
       if (auto shapeOp = dyn_cast<ShapeInference>(op)) {
         shapeOp.inferShapes();
       } else {
diff --git a/mlir/examples/transform/Ch3/include/MyExtension.td b/mlir/examples/transform/Ch3/include/MyExtension.td
index 5a78186..49874a7 100644
--- a/mlir/examples/transform/Ch3/include/MyExtension.td
+++ b/mlir/examples/transform/Ch3/include/MyExtension.td
@@ -46,9 +46,9 @@ def ChangeCallTargetOp : Op<Transform_Dialect, "my.change_call_target",
   // We use a string attribute as the symbol may not exist in the transform IR so the 
   // verification may fail. 
   let arguments = (ins
-    // Specify the type constraint on the input accepting only `func.call` payload
-    // operations.
-    Transform_ConcreteOpType<"func.call">:$call,
+    // Allow the handle to be to concrete func.call ops as well as any op implementing
+    // the CallOpInterface.
+    AnyTypeOf<[Transform_ConcreteOpType<"func.call">, CallOpInterfaceHandle]>:$call,
     StrAttr:$new_target);
 
   // The results are empty as the transformation does not produce any new payload.
diff --git a/mlir/examples/transform/Ch4/lib/MyExtension.cpp b/mlir/examples/transform/Ch4/lib/MyExtension.cpp
index fa0ffc9..2159483 100644
--- a/mlir/examples/transform/Ch4/lib/MyExtension.cpp
+++ b/mlir/examples/transform/Ch4/lib/MyExtension.cpp
@@ -13,11 +13,9 @@
 
 #include "MyExtension.h"
 #include "mlir/Dialect/Transform/IR/TransformDialect.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 
-#define DEBUG_TYPE_MATCHER "transform-matcher"
-#define DBGS_MATCHER() (llvm::dbgs() << "[" DEBUG_TYPE_MATCHER "] ")
-#define DEBUG_MATCHER(x) DEBUG_WITH_TYPE(DEBUG_TYPE_MATCHER, x)
+#define DEBUG_TYPE "transform-matcher"
 
 #define GET_OP_CLASSES
 #include "MyExtension.cpp.inc"
@@ -124,9 +122,8 @@ mlir::transform::HasOperandSatisfyingOp::apply(
       // Report failure-to-match for debugging purposes and stop matching this
       // operand.
       assert(diag.isSilenceableFailure());
-      DEBUG_MATCHER(DBGS_MATCHER()
-                    << "failed to match operand #" << operand.getOperandNumber()
-                    << ": " << diag.getMessage());
+      LDBG() << "failed to match operand #" << operand.getOperandNumber()
+             << ": " << diag.getMessage();
       (void)diag.silence();
       matchSucceeded = false;
       break;
diff --git a/mlir/include/mlir/Conversion/Passes.h b/mlir/include/mlir/Conversion/Passes.h
index d93fbef..3dc48b2 100644
--- a/mlir/include/mlir/Conversion/Passes.h
+++ b/mlir/include/mlir/Conversion/Passes.h
@@ -52,7 +52,6 @@
 #include "mlir/Conversion/MemRefToEmitC/MemRefToEmitCPass.h"
 #include "mlir/Conversion/MemRefToLLVM/MemRefToLLVM.h"
 #include "mlir/Conversion/MemRefToSPIRV/MemRefToSPIRVPass.h"
-#include "mlir/Conversion/MeshToMPI/MeshToMPI.h"
 #include "mlir/Conversion/NVGPUToNVVM/NVGPUToNVVM.h"
 #include "mlir/Conversion/NVVMToLLVM/NVVMToLLVM.h"
 #include "mlir/Conversion/OpenACCToSCF/ConvertOpenACCToSCF.h"
@@ -66,6 +65,7 @@
 #include "mlir/Conversion/SCFToSPIRV/SCFToSPIRVPass.h"
 #include "mlir/Conversion/SPIRVToLLVM/SPIRVToLLVMPass.h"
 #include "mlir/Conversion/ShapeToStandard/ShapeToStandard.h"
+#include "mlir/Conversion/ShardToMPI/ShardToMPI.h"
 #include "mlir/Conversion/TensorToLinalg/TensorToLinalgPass.h"
 #include "mlir/Conversion/TensorToSPIRV/TensorToSPIRVPass.h"
 #include "mlir/Conversion/TosaToArith/TosaToArith.h"
diff --git a/mlir/include/mlir/Conversion/Passes.td b/mlir/include/mlir/Conversion/Passes.td
index 8183f35..eb18160 100644
--- a/mlir/include/mlir/Conversion/Passes.td
+++ b/mlir/include/mlir/Conversion/Passes.td
@@ -903,13 +903,13 @@ def ConvertMemRefToSPIRVPass : Pass<"convert-memref-to-spirv"> {
 }
 
 //===----------------------------------------------------------------------===//
-// MeshToMPI
+// ShardToMPI
 //===----------------------------------------------------------------------===//
 
-def ConvertMeshToMPIPass : Pass<"convert-mesh-to-mpi"> {
-  let summary = "Convert Mesh dialect to MPI dialect.";
+def ConvertShardToMPIPass : Pass<"convert-shard-to-mpi"> {
+  let summary = "Convert Shard dialect to MPI dialect.";
   let description = [{
-    This pass converts communication operations from the Mesh dialect to the
+    This pass converts communication operations from the Shard dialect to the
     MPI dialect.
     If it finds the DLTI attribute "MPI:comm_world-rank" on the module it will
     use that integer value instead of calling MPI_Comm_rank. This allows
diff --git a/mlir/include/mlir/Conversion/MeshToMPI/MeshToMPI.h b/mlir/include/mlir/Conversion/ShardToMPI/ShardToMPI.h
index bc64e7a..b1aa08c 100644
--- a/mlir/include/mlir/Conversion/MeshToMPI/MeshToMPI.h
+++ b/mlir/include/mlir/Conversion/ShardToMPI/ShardToMPI.h
@@ -1,4 +1,4 @@
-//===- MeshToMPI.h - Convert Mesh to MPI dialect ----------------*- C++ -*-===//
+//===- ShardToMPI.h - Convert Shard to MPI dialect --------------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef MLIR_CONVERSION_MESHTOMPI_MESHTOMPI_H
-#define MLIR_CONVERSION_MESHTOMPI_MESHTOMPI_H
+#ifndef MLIR_CONVERSION_SHARDTOMPI_SHARDTOMPI_H
+#define MLIR_CONVERSION_SHARDTOMPI_SHARDTOMPI_H
 
 #include "mlir/Pass/Pass.h"
 #include "mlir/Support/LLVM.h"
@@ -15,9 +15,9 @@
 namespace mlir {
 class Pass;
 
-#define GEN_PASS_DECL_CONVERTMESHTOMPIPASS
+#define GEN_PASS_DECL_CONVERTSHARDTOMPIPASS
 #include "mlir/Conversion/Passes.h.inc"
 
 } // namespace mlir
 
-#endif // MLIR_CONVERSION_MESHTOMPI_MESHTOMPI_H
+#endif // MLIR_CONVERSION_SHARDTOMPI_SHARDTOMPI_H
diff --git a/mlir/include/mlir/Dialect/CMakeLists.txt b/mlir/include/mlir/Dialect/CMakeLists.txt
index 56dc972..e27b167 100644
--- a/mlir/include/mlir/Dialect/CMakeLists.txt
+++ b/mlir/include/mlir/Dialect/CMakeLists.txt
@@ -19,7 +19,7 @@ add_subdirectory(Linalg)
 add_subdirectory(LLVMIR)
 add_subdirectory(Math)
 add_subdirectory(MemRef)
-add_subdirectory(Mesh)
+add_subdirectory(Shard)
 add_subdirectory(MLProgram)
 add_subdirectory(MPI)
 add_subdirectory(NVGPU)
diff --git a/mlir/include/mlir/Dialect/EmitC/Transforms/Passes.td b/mlir/include/mlir/Dialect/EmitC/Transforms/Passes.td
index 74c4913..1893c10 100644
--- a/mlir/include/mlir/Dialect/EmitC/Transforms/Passes.td
+++ b/mlir/include/mlir/Dialect/EmitC/Transforms/Passes.td
@@ -51,11 +51,6 @@ def WrapFuncInClassPass : Pass<"wrap-emitc-func-in-class"> {
     ```
   }];
   let dependentDialects = ["emitc::EmitCDialect"];
-  let options = [Option<
-      "namedAttribute", "named-attribute", "std::string",
-      /*default=*/"",
-      "Attribute key used to extract field names from function argument's "
-      "dictionary attributes">];
 }
 
 #endif // MLIR_DIALECT_EMITC_TRANSFORMS_PASSES
diff --git a/mlir/include/mlir/Dialect/EmitC/Transforms/Transforms.h b/mlir/include/mlir/Dialect/EmitC/Transforms/Transforms.h
index a4e8fe1..bdf6d09 100644
--- a/mlir/include/mlir/Dialect/EmitC/Transforms/Transforms.h
+++ b/mlir/include/mlir/Dialect/EmitC/Transforms/Transforms.h
@@ -29,8 +29,7 @@ ExpressionOp createExpression(Operation *op, OpBuilder &builder);
 void populateExpressionPatterns(RewritePatternSet &patterns);
 
 /// Populates 'patterns' with func-related patterns.
-void populateFuncPatterns(RewritePatternSet &patterns,
-                          StringRef namedAttribute);
+void populateFuncPatterns(RewritePatternSet &patterns);
 
 } // namespace emitc
 } // namespace mlir
diff --git a/mlir/include/mlir/Dialect/Func/Extensions/MeshShardingExtensions.h b/mlir/include/mlir/Dialect/Func/Extensions/ShardingExtensions.h
index 30d3033..e22b24b 100644
--- a/mlir/include/mlir/Dialect/Func/Extensions/MeshShardingExtensions.h
+++ b/mlir/include/mlir/Dialect/Func/Extensions/ShardingExtensions.h
@@ -1,4 +1,4 @@
-//===- MeshShardingExtensions.h - -----------------------------------------===//
+//===- ShardingExtensions.h - -----------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
diff --git a/mlir/include/mlir/Dialect/LLVMIR/LLVMDialect.h b/mlir/include/mlir/Dialect/LLVMIR/LLVMDialect.h
index e355bb8..f3bd5c0 100644
--- a/mlir/include/mlir/Dialect/LLVMIR/LLVMDialect.h
+++ b/mlir/include/mlir/Dialect/LLVMIR/LLVMDialect.h
@@ -32,11 +32,6 @@
 #include "mlir/Interfaces/ViewLikeInterface.h"
 #include "mlir/Support/ThreadLocalCache.h"
 #include "llvm/ADT/PointerEmbeddedInt.h"
-#include "llvm/IR/DerivedTypes.h"
-#include "llvm/IR/InstrTypes.h"
-#include "llvm/IR/LLVMContext.h"
-#include "llvm/IR/Module.h"
-#include "llvm/IR/Type.h"
 
 namespace llvm {
 class Type;
diff --git a/mlir/include/mlir/Dialect/LLVMIR/LLVMIntrinsicOps.td b/mlir/include/mlir/Dialect/LLVMIR/LLVMIntrinsicOps.td
index caba614..8c6f1ee 100644
--- a/mlir/include/mlir/Dialect/LLVMIR/LLVMIntrinsicOps.td
+++ b/mlir/include/mlir/Dialect/LLVMIR/LLVMIntrinsicOps.td
@@ -555,8 +555,6 @@ def LLVM_AssumeOp
 
   let builders = [
     OpBuilder<(ins "Value":$cond)>,
-    OpBuilder<(ins "Value":$cond,
-                   "ArrayRef<llvm::OperandBundleDefT<Value>>":$opBundles)>,
     OpBuilder<(ins "Value":$cond, "llvm::StringRef":$tag, "ValueRange":$args)>,
     OpBuilder<(ins "Value":$cond, "AssumeAlignTag":$tag, "Value":$ptr,
                    "Value":$align)>,
diff --git a/mlir/include/mlir/Dialect/Linalg/TransformOps/GPUHeuristics.h b/mlir/include/mlir/Dialect/Linalg/TransformOps/GPUHeuristics.h
index 5430fd9..c0c6085 100644
--- a/mlir/include/mlir/Dialect/Linalg/TransformOps/GPUHeuristics.h
+++ b/mlir/include/mlir/Dialect/Linalg/TransformOps/GPUHeuristics.h
@@ -119,6 +119,12 @@ public:
   Status status;
 };
 
+inline llvm::raw_ostream &operator<<(llvm::raw_ostream &os,
+                                     const CopyMappingInfo &info) {
+  info.print(os);
+  return os;
+}
+
 } // namespace gpu
 } // namespace transform
 } // namespace mlir
diff --git a/mlir/include/mlir/Dialect/Linalg/Transforms/MeshShardingInterfaceImpl.h b/mlir/include/mlir/Dialect/Linalg/Transforms/ShardingInterfaceImpl.h
index c57501e..dc21bc0 100644
--- a/mlir/include/mlir/Dialect/Linalg/Transforms/MeshShardingInterfaceImpl.h
+++ b/mlir/include/mlir/Dialect/Linalg/Transforms/ShardingInterfaceImpl.h
@@ -1,4 +1,4 @@
-//===- MeshShardingInterfaceImpl.h ----------------------------------------===//
+//===- ShardingInterfaceImpl.h ----------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -6,15 +6,15 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef MLIR_DIALECT_LINALG_MESHSHARDINGINTERFACEIMPL_H
-#define MLIR_DIALECT_LINALG_MESHSHARDINGINTERFACEIMPL_H
+#ifndef MLIR_DIALECT_LINALG_SHARDSHARDINGINTERFACEIMPL_H
+#define MLIR_DIALECT_LINALG_SHARDSHARDINGINTERFACEIMPL_H
 
 namespace mlir {
 class DialectRegistry;
 
 namespace linalg {
-void registerMeshShardingInterfaceExternalModels(DialectRegistry &registry);
+void registerShardingInterfaceExternalModels(DialectRegistry &registry);
 } // namespace linalg
 } // namespace mlir
 
-#endif // MLIR_DIALECT_LINALG_MESHSHARDINGINTERFACEIMPL_H
+#endif // MLIR_DIALECT_LINALG_SHARDSHARDINGINTERFACEIMPL_H
diff --git a/mlir/include/mlir/Dialect/Mesh/IR/CMakeLists.txt b/mlir/include/mlir/Dialect/Mesh/IR/CMakeLists.txt
deleted file mode 100644
index f26c628..0000000
--- a/mlir/include/mlir/Dialect/Mesh/IR/CMakeLists.txt
+++ /dev/null
@@ -1,25 +0,0 @@
-add_mlir_doc(MeshOps MeshOps Dialects/ -gen-op-doc -dialect=mesh)
-add_mlir_doc(MeshOps MeshAttrs Dialects/ -gen-attrdef-doc -dialect=mesh)
-
-set(LLVM_TARGET_DEFINITIONS MeshOps.td)
-mlir_tablegen(MeshDialect.cpp.inc -gen-dialect-defs -dialect=mesh)
-mlir_tablegen(MeshDialect.h.inc -gen-dialect-decls -dialect=mesh)
-
-set(LLVM_TARGET_DEFINITIONS MeshBase.td)
-mlir_tablegen(MeshAttributes.h.inc -gen-attrdef-decls)
-mlir_tablegen(MeshAttributes.cpp.inc -gen-attrdef-defs)
-
-set(LLVM_TARGET_DEFINITIONS MeshBase.td)
-mlir_tablegen(MeshEnums.h.inc -gen-enum-decls)
-mlir_tablegen(MeshEnums.cpp.inc -gen-enum-defs)
-
-set(LLVM_TARGET_DEFINITIONS MeshBase.td)
-mlir_tablegen(MeshTypes.h.inc -gen-typedef-decls)
-mlir_tablegen(MeshTypes.cpp.inc -gen-typedef-defs)
-
-set(LLVM_TARGET_DEFINITIONS MeshOps.td)
-mlir_tablegen(MeshOps.h.inc -gen-op-decls)
-mlir_tablegen(MeshOps.cpp.inc -gen-op-defs)
-
-add_public_tablegen_target(MLIRMeshIncGen)
-add_dependencies(mlir-headers MLIRMeshIncGen)
diff --git a/mlir/include/mlir/Dialect/Mesh/Transforms/CMakeLists.txt b/mlir/include/mlir/Dialect/Mesh/Transforms/CMakeLists.txt
deleted file mode 100644
index 8d76848..0000000
--- a/mlir/include/mlir/Dialect/Mesh/Transforms/CMakeLists.txt
+++ /dev/null
@@ -1,6 +0,0 @@
-set(LLVM_TARGET_DEFINITIONS Passes.td)
-mlir_tablegen(Passes.h.inc -gen-pass-decls -name Mesh)
-add_public_tablegen_target(MLIRMeshPassIncGen)
-add_dependencies(mlir-headers MLIRMeshPassIncGen)
-
-add_mlir_doc(Passes MeshPasses ./ -gen-pass-doc)
diff --git a/mlir/include/mlir/Dialect/Mesh/CMakeLists.txt b/mlir/include/mlir/Dialect/Shard/CMakeLists.txt
index fa8842f..fa8842f 100644
--- a/mlir/include/mlir/Dialect/Mesh/CMakeLists.txt
+++ b/mlir/include/mlir/Dialect/Shard/CMakeLists.txt
diff --git a/mlir/include/mlir/Dialect/Shard/IR/CMakeLists.txt b/mlir/include/mlir/Dialect/Shard/IR/CMakeLists.txt
new file mode 100644
index 0000000..a2495af
--- /dev/null
+++ b/mlir/include/mlir/Dialect/Shard/IR/CMakeLists.txt
@@ -0,0 +1,25 @@
+add_mlir_doc(ShardOps ShardOps Dialects/ -gen-op-doc -dialect=shard)
+add_mlir_doc(ShardOps ShardAttrs Dialects/ -gen-attrdef-doc -dialect=shard)
+
+set(LLVM_TARGET_DEFINITIONS ShardOps.td)
+mlir_tablegen(ShardDialect.cpp.inc -gen-dialect-defs -dialect=shard)
+mlir_tablegen(ShardDialect.h.inc -gen-dialect-decls -dialect=shard)
+
+set(LLVM_TARGET_DEFINITIONS ShardBase.td)
+mlir_tablegen(ShardAttributes.h.inc -gen-attrdef-decls)
+mlir_tablegen(ShardAttributes.cpp.inc -gen-attrdef-defs)
+
+set(LLVM_TARGET_DEFINITIONS ShardBase.td)
+mlir_tablegen(ShardEnums.h.inc -gen-enum-decls)
+mlir_tablegen(ShardEnums.cpp.inc -gen-enum-defs)
+
+set(LLVM_TARGET_DEFINITIONS ShardBase.td)
+mlir_tablegen(ShardTypes.h.inc -gen-typedef-decls)
+mlir_tablegen(ShardTypes.cpp.inc -gen-typedef-defs)
+
+set(LLVM_TARGET_DEFINITIONS ShardOps.td)
+mlir_tablegen(ShardOps.h.inc -gen-op-decls)
+mlir_tablegen(ShardOps.cpp.inc -gen-op-defs)
+
+add_public_tablegen_target(MLIRShardIncGen)
+add_dependencies(mlir-headers MLIRShardIncGen)
diff --git a/mlir/include/mlir/Dialect/Mesh/IR/MeshBase.td b/mlir/include/mlir/Dialect/Shard/IR/ShardBase.td
index 61403ac..41ae318 100644
--- a/mlir/include/mlir/Dialect/Mesh/IR/MeshBase.td
+++ b/mlir/include/mlir/Dialect/Shard/IR/ShardBase.td
@@ -1,4 +1,4 @@
-//===- MeshBase.td - Mesh Dialect --------------------------*- tablegen -*-===//
+//===- ShardBase.td - Shard Dialect ------------------------*- tablegen -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef MLIR_DIALECT_MESH_IR_MESHBASE_TD
-#define MLIR_DIALECT_MESH_IR_MESHBASE_TD
+#ifndef MLIR_DIALECT_SHARD_IR_SHARDBASE_TD
+#define MLIR_DIALECT_SHARD_IR_SHARDBASE_TD
 
 include "mlir/IR/OpBase.td"
 include "mlir/IR/AttrTypeBase.td"
@@ -16,15 +16,15 @@ include "mlir/IR/CommonAttrConstraints.td"
 include "mlir/IR/EnumAttr.td"
 
 //===----------------------------------------------------------------------===//
-// Mesh Dialect
+// Shard Dialect
 //===----------------------------------------------------------------------===//
 
-def Mesh_Dialect : Dialect {
-  let name = "mesh";
-  let cppNamespace = "::mlir::mesh";
+def Shard_Dialect : Dialect {
+  let name = "shard";
+  let cppNamespace = "::mlir::shard";
 
   let description = [{
-    See [Mesh dialect documentation](mlir/docs/Dialects/Mesh.md).
+    See [Shard dialect documentation](mlir/docs/Dialects/Shard.md).
   }];
 
   let dependentDialects = [
@@ -36,16 +36,16 @@ def Mesh_Dialect : Dialect {
   let hasConstantMaterializer = 1;
 }
 
-def Mesh_MeshAxis : I<16>;
-def Mesh_MeshAxesAttr : DenseArrayAttrBase<"DenseI16ArrayAttr", "int16_t", "i16">;
-def Mesh_ShardShapeAttr : DenseArrayAttrBase<"DenseI64ArrayAttr", "int64_t", "i64">;
+def Shard_GridAxis : I<16>;
+def Shard_GridAxesAttr : DenseArrayAttrBase<"DenseI16ArrayAttr", "int16_t", "i16">;
+def Shard_ShardShapeAttr : DenseArrayAttrBase<"DenseI64ArrayAttr", "int64_t", "i64">;
 
 //===----------------------------------------------------------------------===//
-// Mesh Enums.
+// Shard Enums.
 //===----------------------------------------------------------------------===//
 
-def Mesh_ReductionKind : I32EnumAttr<"ReductionKind",
-  "Reduction of an iterator/mesh dimension.", [
+def Shard_ReductionKind : I32EnumAttr<"ReductionKind",
+  "Reduction of an iterator/grid dimension.", [
   I32EnumAttrCase<"Sum", 1, "sum">,
   I32EnumAttrCase<"Max", 2, "max">,
   I32EnumAttrCase<"Min", 3, "min">,
@@ -58,31 +58,31 @@ def Mesh_ReductionKind : I32EnumAttr<"ReductionKind",
   I32EnumAttrCase<"Generic", 100, "generic">
 ]> {
   let genSpecializedAttr = 0;
-  let cppNamespace = "::mlir::mesh";
+  let cppNamespace = "::mlir::shard";
 }
 
-def Mesh_ReductionKindAttr : EnumAttr<Mesh_Dialect, Mesh_ReductionKind, "partial"> {
+def Shard_ReductionKindAttr : EnumAttr<Shard_Dialect, Shard_ReductionKind, "partial"> {
   let assemblyFormat = "$value";
 }
 
-class Mesh_Type<string name, string typeMnemonic, list<Trait> traits = [],
+class Shard_Type<string name, string typeMnemonic, list<Trait> traits = [],
                    string baseCppClass = "::mlir::Type">
-    : TypeDef<Mesh_Dialect, name, traits, baseCppClass> {
+    : TypeDef<Shard_Dialect, name, traits, baseCppClass> {
   let mnemonic = typeMnemonic;
 }
 
-def Mesh_Sharding : Mesh_Type<"Sharding", "sharding"> {
+def Shard_Sharding : Shard_Type<"Sharding", "sharding"> {
   let summary = "sharding definition";
   let assemblyFormat = "";
 }
 
 //===----------------------------------------------------------------------===//
-// Mesh Attribute
+// Shard Attribute
 //===----------------------------------------------------------------------===//
 
-def Mesh_MeshAxesArrayAttr : AttrDef<Mesh_Dialect, "MeshAxesArray"> {
+def Shard_GridAxesArrayAttr : AttrDef<Shard_Dialect, "GridAxesArray"> {
   let mnemonic = "axisarray";
-  let parameters = (ins ArrayRefParameter<"MeshAxesAttr">:$axes);
+  let parameters = (ins ArrayRefParameter<"GridAxesAttr">:$axes);
   let assemblyFormat = "`[` $axes `]`";
   let extraClassDeclaration = [{
     size_t size() const { return getAxes().size(); }
@@ -91,4 +91,4 @@ def Mesh_MeshAxesArrayAttr : AttrDef<Mesh_Dialect, "MeshAxesArray"> {
   }];
 }
 
-#endif // MLIR_DIALECT_MESH_IR_MESHBASE_TD
+#endif // MLIR_DIALECT_SHARD_IR_SHARDBASE_TD
diff --git a/mlir/include/mlir/Dialect/Mesh/IR/MeshDialect.h b/mlir/include/mlir/Dialect/Shard/IR/ShardDialect.h
index a30cf91..4113a66 100644
--- a/mlir/include/mlir/Dialect/Mesh/IR/MeshDialect.h
+++ b/mlir/include/mlir/Dialect/Shard/IR/ShardDialect.h
@@ -1,4 +1,4 @@
-//===- MeshOps.h - Mesh Dialect ---------------------------------*- C++ -*-===//
+//===- ShardOps.h - Shard Dialect -------------------------------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -6,11 +6,11 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef MLIR_DIALECT_MESH_IR_MESHDIALECT_H
-#define MLIR_DIALECT_MESH_IR_MESHDIALECT_H
+#ifndef MLIR_DIALECT_SHARD_IR_SHARDDIALECT_H
+#define MLIR_DIALECT_SHARD_IR_SHARDDIALECT_H
 
 #include "mlir/IR/Dialect.h"
 
-#include "mlir/Dialect/Mesh/IR/MeshDialect.h.inc"
+#include "mlir/Dialect/Shard/IR/ShardDialect.h.inc"
 
-#endif // MLIR_DIALECT_MESH_IR_MESHDIALECT_H
+#endif // MLIR_DIALECT_SHARD_IR_SHARDDIALECT_H
diff --git a/mlir/include/mlir/Dialect/Mesh/IR/MeshOps.h b/mlir/include/mlir/Dialect/Shard/IR/ShardOps.h
index 7cfe59d..457fe6f 100644
--- a/mlir/include/mlir/Dialect/Mesh/IR/MeshOps.h
+++ b/mlir/include/mlir/Dialect/Shard/IR/ShardOps.h
@@ -1,4 +1,4 @@
-//===- MeshOps.h - Mesh Dialect Operations ----------------------*- C++ -*-===//
+//===- ShardOps.h - Shard Dialect Operations --------------------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef MLIR_DIALECT_MESH_IR_MESHOPS_H
-#define MLIR_DIALECT_MESH_IR_MESHOPS_H
+#ifndef MLIR_DIALECT_SHARD_IR_SHARDOPS_H
+#define MLIR_DIALECT_SHARD_IR_SHARDOPS_H
 
 #include "mlir/Bytecode/BytecodeOpInterface.h"
 #include "mlir/Dialect/Utils/StructuredOpsUtils.h"
@@ -21,45 +21,45 @@
 #include "llvm/Support/MathExtras.h"
 
 namespace mlir {
-namespace mesh {
+namespace shard {
 
-using MeshAxis = int16_t;
-using MeshAxesAttr = DenseI16ArrayAttr;
+using GridAxis = int16_t;
+using GridAxesAttr = DenseI16ArrayAttr;
 using ShardShapeAttr = DenseI64ArrayAttr;
 using HaloSizePairAttr = DenseI64ArrayAttr;
 
-} // namespace mesh
+} // namespace shard
 } // namespace mlir
 
-#include "mlir/Dialect/Mesh/IR/MeshEnums.h.inc"
+#include "mlir/Dialect/Shard/IR/ShardEnums.h.inc"
 
 #define GET_ATTRDEF_CLASSES
-#include "mlir/Dialect/Mesh/IR/MeshAttributes.h.inc"
+#include "mlir/Dialect/Shard/IR/ShardAttributes.h.inc"
 
 namespace mlir {
-namespace mesh {
+namespace shard {
 
-class MeshSharding {
+class Sharding {
 private:
-  ::mlir::FlatSymbolRefAttr mesh;
-  SmallVector<MeshAxesAttr> split_axes;
+  ::mlir::FlatSymbolRefAttr grid;
+  SmallVector<GridAxesAttr> split_axes;
   SmallVector<int64_t> static_halo_sizes;
   SmallVector<int64_t> static_sharded_dims_offsets;
   SmallVector<Value> dynamic_halo_sizes;
   SmallVector<Value> dynamic_sharded_dims_offsets;
 
 public:
-  MeshSharding(::mlir::FlatSymbolRefAttr mesh_ = nullptr);
-  MeshSharding(Value rhs);
-  static MeshSharding get(::mlir::FlatSymbolRefAttr mesh_,
-                          ArrayRef<MeshAxesAttr> split_axes_,
-                          ArrayRef<int64_t> static_halo_sizes_ = {},
-                          ArrayRef<int64_t> static_sharded_dims_offsets_ = {},
-                          ArrayRef<Value> dynamic_halo_sizes_ = {},
-                          ArrayRef<Value> dynamic_sharded_dims_offsets_ = {});
-  ::mlir::FlatSymbolRefAttr getMeshAttr() const { return mesh; }
-  ::llvm::StringRef getMesh() const { return mesh ? mesh.getValue() : ""; }
-  ArrayRef<MeshAxesAttr> getSplitAxes() const { return split_axes; }
+  Sharding(::mlir::FlatSymbolRefAttr grid_ = nullptr);
+  Sharding(Value rhs);
+  static Sharding get(::mlir::FlatSymbolRefAttr grid_,
+                      ArrayRef<GridAxesAttr> split_axes_,
+                      ArrayRef<int64_t> static_halo_sizes_ = {},
+                      ArrayRef<int64_t> static_sharded_dims_offsets_ = {},
+                      ArrayRef<Value> dynamic_halo_sizes_ = {},
+                      ArrayRef<Value> dynamic_sharded_dims_offsets_ = {});
+  ::mlir::FlatSymbolRefAttr getGridAttr() const { return grid; }
+  ::llvm::StringRef getGrid() const { return grid ? grid.getValue() : ""; }
+  ArrayRef<GridAxesAttr> getSplitAxes() const { return split_axes; }
   ArrayRef<int64_t> getStaticHaloSizes() const { return static_halo_sizes; }
   ArrayRef<int64_t> getStaticShardedDimsOffsets() const {
     return static_sharded_dims_offsets;
@@ -68,28 +68,28 @@ public:
   ArrayRef<Value> getDynamicShardedDimsOffsets() const {
     return dynamic_sharded_dims_offsets;
   }
-  operator bool() const { return (!mesh) == false; }
+  operator bool() const { return (!grid) == false; }
   bool operator==(Value rhs) const;
   bool operator!=(Value rhs) const;
-  bool operator==(const MeshSharding &rhs) const;
-  bool operator!=(const MeshSharding &rhs) const;
-  bool equalSplitAxes(const MeshSharding &rhs) const;
-  bool equalHaloAndShardSizes(const MeshSharding &rhs) const;
-  bool equalHaloSizes(const MeshSharding &rhs) const;
-  bool equalShardSizes(const MeshSharding &rhs) const;
+  bool operator==(const Sharding &rhs) const;
+  bool operator!=(const Sharding &rhs) const;
+  bool equalSplitAxes(const Sharding &rhs) const;
+  bool equalHaloAndShardSizes(const Sharding &rhs) const;
+  bool equalHaloSizes(const Sharding &rhs) const;
+  bool equalShardSizes(const Sharding &rhs) const;
 };
 
-} // namespace mesh
+} // namespace shard
 } // namespace mlir
 
 #define GET_TYPEDEF_CLASSES
-#include "mlir/Dialect/Mesh/IR/MeshTypes.h.inc"
+#include "mlir/Dialect/Shard/IR/ShardTypes.h.inc"
 
 #define GET_OP_CLASSES
-#include "mlir/Dialect/Mesh/IR/MeshOps.h.inc"
+#include "mlir/Dialect/Shard/IR/ShardOps.h.inc"
 
 namespace mlir {
-namespace mesh {
+namespace shard {
 
 inline bool isReductionLoop(utils::IteratorType iType) {
   return iType == utils::IteratorType::reduction;
@@ -103,52 +103,52 @@ void removeTrailingEmptySubArray(SmallVector<SmallVector<T>> &array) {
 }
 
 // Is the same tensor replicated on all processes.
-inline bool isFullReplication(MeshSharding sharding) {
-  return llvm::all_of(sharding.getSplitAxes(), [](MeshAxesAttr axes) {
+inline bool isFullReplication(Sharding sharding) {
+  return llvm::all_of(sharding.getSplitAxes(), [](GridAxesAttr axes) {
     return axes.asArrayRef().empty();
   });
 }
 
-inline mesh::MeshOp
-getMeshOrNull(Operation *op, FlatSymbolRefAttr meshSymbol,
+inline shard::GridOp
+getGridOrNull(Operation *op, FlatSymbolRefAttr gridSymbol,
               SymbolTableCollection &symbolTableCollection) {
-  if (!meshSymbol)
+  if (!gridSymbol)
     return nullptr;
-  return symbolTableCollection.lookupNearestSymbolFrom<mesh::MeshOp>(
-      op, meshSymbol);
+  return symbolTableCollection.lookupNearestSymbolFrom<shard::GridOp>(
+      op, gridSymbol);
 }
 
-inline mesh::MeshOp getMesh(Operation *op, FlatSymbolRefAttr meshSymbol,
-                            SymbolTableCollection &symbolTableCollection) {
-  mesh::MeshOp meshOp = getMeshOrNull(op, meshSymbol, symbolTableCollection);
-  assert(meshOp);
-  return meshOp;
+inline shard::GridOp getGrid(Operation *op, FlatSymbolRefAttr gridSymbol,
+                             SymbolTableCollection &symbolTableCollection) {
+  shard::GridOp gridOp = getGridOrNull(op, gridSymbol, symbolTableCollection);
+  assert(gridOp);
+  return gridOp;
 }
 
-// Get the corresponding mesh op using the standard attribute nomenclature.
+// Get the corresponding grid op using the standard attribute nomenclature.
 template <typename Op>
-mesh::MeshOp getMesh(Op op, SymbolTableCollection &symbolTableCollection) {
-  return getMesh(op.getOperation(), op.getMeshAttr(), symbolTableCollection);
+shard::GridOp getGrid(Op op, SymbolTableCollection &symbolTableCollection) {
+  return getGrid(op.getOperation(), op.getGridAttr(), symbolTableCollection);
 }
 
 template <>
-inline mesh::MeshOp
-getMesh<ShardOp>(ShardOp op, SymbolTableCollection &symbolTableCollection) {
-  return getMesh(
+inline shard::GridOp
+getGrid<ShardOp>(ShardOp op, SymbolTableCollection &symbolTableCollection) {
+  return getGrid(
       op.getOperation(),
-      cast<ShardingOp>(op.getSharding().getDefiningOp()).getMeshAttr(),
+      cast<ShardingOp>(op.getSharding().getDefiningOp()).getGridAttr(),
       symbolTableCollection);
 }
 
 // Get the number of processes that participate in each group
-// induced by `meshAxes`.
-template <typename MeshAxesRange, typename MeshShapeRange>
-int64_t collectiveProcessGroupSize(MeshAxesRange &&meshAxes,
-                                   MeshShapeRange &&meshShape) {
+// induced by `gridAxes`.
+template <typename GridAxesRange, typename GridShapeRange>
+int64_t collectiveProcessGroupSize(GridAxesRange &&gridAxes,
+                                   GridShapeRange &&gridShape) {
   int64_t res = 1;
 
-  for (MeshAxis axis : meshAxes) {
-    auto axisSize = *(std::begin(meshShape) + axis);
+  for (GridAxis axis : gridAxes) {
+    auto axisSize = *(std::begin(gridShape) + axis);
     if (ShapedType::isDynamic(axisSize)) {
       return ShapedType::kDynamic;
     }
@@ -158,10 +158,10 @@ int64_t collectiveProcessGroupSize(MeshAxesRange &&meshAxes,
   return res;
 }
 
-template <typename MeshAxesRange>
-int64_t collectiveProcessGroupSize(MeshAxesRange &&meshAxes, MeshOp mesh) {
-  return collectiveProcessGroupSize(std::forward<MeshAxesRange>(meshAxes),
-                                    mesh.getShape());
+template <typename GridAxesRange>
+int64_t collectiveProcessGroupSize(GridAxesRange &&gridAxes, GridOp grid) {
+  return collectiveProcessGroupSize(std::forward<GridAxesRange>(gridAxes),
+                                    grid.getShape());
 }
 
 // Get the size of a sharded dimension.
@@ -182,27 +182,25 @@ inline int64_t gatherDimension(int64_t dimSize, int64_t shardCount) {
 }
 
 // Return the sharded shape `shape` according ot sharding `sharding`.
-// The shape for the tensor on each device in the mesh.
+// The shape for the tensor on each device in the grid.
 // Example:
-// On a 2x4x? mesh with split axes = [[0], [1], [2]] the shape ?x5x1 would
+// On a 2x4x? grid with split axes = [[0], [1], [2]] the shape ?x5x1 would
 // result in a shape for each shard of ?x2x?.
-ShapedType shardShapedType(ShapedType shape, MeshOp mesh,
-                           MeshSharding sharding);
+ShapedType shardShapedType(ShapedType shape, GridOp grid, Sharding sharding);
 
 // If ranked tensor type return its sharded counterpart.
 //
 // If not ranked tensor type return `type`.
 // `sharding` in that case must be null.
-Type shardType(Type type, MeshOp mesh, MeshSharding sharding);
+Type shardType(Type type, GridOp grid, Sharding sharding);
 
 // Insert shard op if there is not one that already has the same sharding.
 // Use newShardOp if it is not null. Otherwise create a new one.
 // May insert resharding if required.
 // Potentially updates newShardOp.
-void maybeInsertTargetShardingAnnotation(MeshSharding sharding, OpResult result,
+void maybeInsertTargetShardingAnnotation(Sharding sharding, OpResult result,
                                          OpBuilder &builder);
-void maybeInsertSourceShardingAnnotation(MeshSharding sharding,
-                                         OpOperand &operand,
+void maybeInsertSourceShardingAnnotation(Sharding sharding, OpOperand &operand,
                                          OpBuilder &builder);
 
 /// Converts a vector of OpFoldResults (ints) into vector of Values of the
@@ -210,7 +208,7 @@ void maybeInsertSourceShardingAnnotation(MeshSharding sharding,
 SmallVector<Value> getMixedAsValues(OpBuilder b, const Location &loc,
                                     llvm::ArrayRef<int64_t> statics,
                                     ValueRange dynamics, Type type = Type());
-} // namespace mesh
+} // namespace shard
 } // namespace mlir
 
-#endif // MLIR_DIALECT_MESH_IR_MESHOPS_H
+#endif // MLIR_DIALECT_SHARD_IR_SHARDOPS_H
diff --git a/mlir/include/mlir/Dialect/Mesh/IR/MeshOps.td b/mlir/include/mlir/Dialect/Shard/IR/ShardOps.td
index 1662885..29b384f 100644
--- a/mlir/include/mlir/Dialect/Mesh/IR/MeshOps.td
+++ b/mlir/include/mlir/Dialect/Shard/IR/ShardOps.td
@@ -1,4 +1,4 @@
-//===-- MeshOps.td - Mesh dialect operation definitions ----*- tablegen -*-===//
+//===-- ShardOps.td - Shard dialect operation definitions ----*- tablegen -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -6,10 +6,10 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef MLIR_DIALECT_MESH_IR_MESHOPS_TD
-#define MLIR_DIALECT_MESH_IR_MESHOPS_TD
+#ifndef MLIR_DIALECT_SHARD_IR_SHARDOPS_TD
+#define MLIR_DIALECT_SHARD_IR_SHARDOPS_TD
 
-include "mlir/Dialect/Mesh/IR/MeshBase.td"
+include "mlir/Dialect/Shard/IR/ShardBase.td"
 include "mlir/Dialect/Shape/IR/ShapeBase.td"
 include "mlir/Interfaces/DestinationStyleOpInterface.td"
 include "mlir/Interfaces/InferTypeOpInterface.td"
@@ -21,24 +21,24 @@ include "mlir/IR/OpAsmInterface.td"
 include "mlir/IR/SymbolInterfaces.td"
 
 //===----------------------------------------------------------------------===//
-// Mesh operations.
+// Shard operations.
 //===----------------------------------------------------------------------===//
 
-class Mesh_Op<string mnemonic, list<Trait> traits = []> :
-    Op<Mesh_Dialect, mnemonic, traits> {
+class Shard_Op<string mnemonic, list<Trait> traits = []> :
+    Op<Shard_Dialect, mnemonic, traits> {
 }
 
-def Mesh_MeshOp : Mesh_Op<"mesh", [Symbol, Pure]> {
-  let summary = "Description of a device/process mesh.";
+def Shard_GridOp : Shard_Op<"grid", [Symbol, Pure]> {
+  let summary = "Description of a device/process grid.";
   let description = [{
-    The mesh.mesh operation is a symbol operation that identifies a specific
-    mesh. The operation has three attributes:
+    The shard.grid operation is a symbol operation that identifies a specific
+    grid. The operation has three attributes:
 
-    1. `sym_name`: This attribute uniquely identifies the name of the mesh.
-    This name serves as a symbolic reference to the mesh throughout
+    1. `sym_name`: This attribute uniquely identifies the name of the grid.
+    This name serves as a symbolic reference to the grid throughout
     the MLIR module, allowing for consistent referencing and easier debugging.
 
-    2. `shape`: This attribute represents the shape of the device mesh.
+    2. `shape`: This attribute represents the shape of the device grid.
     It uses the same notation as a tensor shape. Also allowing for dynamic
     dimensions.
     This flexibility allows for dynamic device assignment or configurations
@@ -48,21 +48,21 @@ def Mesh_MeshOp : Mesh_Op<"mesh", [Symbol, Pure]> {
 
     Example:
     ```
-    // A device mesh with 3 axes, the total device number is 4 * 8 * 12
+    // A device grid with 3 axes, the total device number is 4 * 8 * 12
     // The dimension sizes are 4, 8, 12 
-    mesh.mesh @mesh0(shape = 4x8x12)
+    shard.grid @grid0(shape = 4x8x12)
 
-    // A device mesh with 2 axes, the total device number is unknown
+    // A device grid with 2 axes, the total device number is unknown
     // The first dimension size is 4 and the second is unknown
-    mesh.mesh @mesh1(shape = 4x?)
+    shard.grid @grid1(shape = 4x?)
 
-    // A device mesh with 2 axes, the total device number is unknown
+    // A device grid with 2 axes, the total device number is unknown
     // The first dimension size is unknown and the second is 4
-    mesh.mesh @mesh2(shape = ?x4)
+    shard.grid @grid2(shape = ?x4)
 
-    // A device mesh with 2 axes, the number of devices along both axes
+    // A device grid with 2 axes, the number of devices along both axes
     // is unknown
-    mesh.mesh @mesh3(shape = ?x?)
+    shard.grid @grid3(shape = ?x?)
     ```
   }];
   let arguments = (ins
@@ -79,15 +79,15 @@ def Mesh_MeshOp : Mesh_Op<"mesh", [Symbol, Pure]> {
   let hasVerifier = 1;
 }
 
-def Mesh_MeshShapeOp : Mesh_Op<"mesh_shape", [
+def Shard_GridShapeOp : Shard_Op<"grid_shape", [
     Pure,
     DeclareOpInterfaceMethods<SymbolUserOpInterface>,
     DeclareOpInterfaceMethods<OpAsmOpInterface, ["getAsmResultNames"]>
   ]> {
-  let summary = "Get the shape of the mesh.";
+  let summary = "Get the shape of the grid.";
   let arguments = (ins
-    FlatSymbolRefAttr:$mesh,
-    DefaultValuedAttr<Mesh_MeshAxesAttr, "{}">:$axes
+    FlatSymbolRefAttr:$grid,
+    DefaultValuedAttr<Shard_GridAxesAttr, "{}">:$axes
   );
 
   let results = (outs
@@ -95,46 +95,46 @@ def Mesh_MeshShapeOp : Mesh_Op<"mesh_shape", [
   );
 
   let assemblyFormat = [{
-    $mesh (`axes` `=` $axes^)?
+    $grid (`axes` `=` $axes^)?
     attr-dict `:` type($result)
   }];
 
   let builders = [
-    OpBuilder<(ins "::mlir::mesh::MeshOp":$mesh)>,
-    OpBuilder<(ins "::mlir::mesh::MeshOp":$mesh, "ArrayRef<MeshAxis>":$axes)>,
-    OpBuilder<(ins "StringRef":$mesh, "ArrayRef<MeshAxis>":$axes)>
+    OpBuilder<(ins "::mlir::shard::GridOp":$grid)>,
+    OpBuilder<(ins "::mlir::shard::GridOp":$grid, "ArrayRef<GridAxis>":$axes)>,
+    OpBuilder<(ins "StringRef":$grid, "ArrayRef<GridAxis>":$axes)>
   ];
 }
 
-def Mesh_ProcessMultiIndexOp : Mesh_Op<"process_multi_index", [
+def Shard_ProcessMultiIndexOp : Shard_Op<"process_multi_index", [
   Pure,
   DeclareOpInterfaceMethods<SymbolUserOpInterface>,
   DeclareOpInterfaceMethods<OpAsmOpInterface, ["getAsmResultNames"]>
 ]> {
-  let summary = "Get the multi index of current device along specified mesh axes.";
+  let summary = "Get the multi index of current device along specified grid axes.";
   let description = [{
     It is used in the SPMD format of IR.
-    The `axes` mush be non-negative and less than the total number of mesh axes.
+    The `axes` mush be non-negative and less than the total number of grid axes.
     If the axes are empty then get the index along all axes.
   }];
   let arguments = (ins
-    FlatSymbolRefAttr:$mesh,
-    DefaultValuedAttr<Mesh_MeshAxesAttr, "{}">:$axes
+    FlatSymbolRefAttr:$grid,
+    DefaultValuedAttr<Shard_GridAxesAttr, "{}">:$axes
   );
   let results = (outs
     Variadic<Index>:$result
   );
   let assemblyFormat = [{
-    `on` $mesh (`axes` `=` $axes^)?
+    `on` $grid (`axes` `=` $axes^)?
     attr-dict `:` type($result)
   }];
   let builders = [
-    OpBuilder<(ins "::mlir::mesh::MeshOp":$mesh)>,
-    OpBuilder<(ins "StringRef":$mesh, "ArrayRef<MeshAxis>":$axes)>
+    OpBuilder<(ins "::mlir::shard::GridOp":$grid)>,
+    OpBuilder<(ins "StringRef":$grid, "ArrayRef<GridAxis>":$axes)>
   ];
 }
 
-def Mesh_ProcessLinearIndexOp : Mesh_Op<"process_linear_index", [
+def Shard_ProcessLinearIndexOp : Shard_Op<"process_linear_index", [
   Pure,
   DeclareOpInterfaceMethods<SymbolUserOpInterface>,
   DeclareOpInterfaceMethods<OpAsmOpInterface, ["getAsmResultNames"]>
@@ -143,34 +143,34 @@ def Mesh_ProcessLinearIndexOp : Mesh_Op<"process_linear_index", [
   let description = [{
     Example:
     ```
-    %idx = mesh.process_linear_index on @mesh : index
+    %idx = shard.process_linear_index on @grid : index
     ```
-    if `@mesh` has shape `(10, 20, 30)`, a device with multi
+    if `@grid` has shape `(10, 20, 30)`, a device with multi
     index `(1, 2, 3)` will have linear index `3 + 30*2 + 20*30*1`.
   }];
-  let arguments = (ins FlatSymbolRefAttr:$mesh);
+  let arguments = (ins FlatSymbolRefAttr:$grid);
   let results = (outs Index:$result);
-  let assemblyFormat = "`on` $mesh attr-dict `:` type($result)";
+  let assemblyFormat = "`on` $grid attr-dict `:` type($result)";
   let builders = [
-    OpBuilder<(ins "::mlir::mesh::MeshOp":$mesh)>
+    OpBuilder<(ins "::mlir::shard::GridOp":$grid)>
   ];
 }
 
-def Mesh_NeighborsLinearIndicesOp : Mesh_Op<"neighbors_linear_indices", [
+def Shard_NeighborsLinearIndicesOp : Shard_Op<"neighbors_linear_indices", [
   Pure,
   DeclareOpInterfaceMethods<SymbolUserOpInterface>,
   DeclareOpInterfaceMethods<OpAsmOpInterface, ["getAsmResultNames"]>
 ]> {
   let summary =
-      "For given mesh index get the linear indices of the direct neighbor processes along the given split.";
+      "For given grid index get the linear indices of the direct neighbor processes along the given split.";
   let description = [{
     Example:
     ```
-    mesh.mesh @mesh0(shape = 10x20x30)
+    shard.grid @grid0(shape = 10x20x30)
     %c1 = arith.constant 1 : index
     %c2 = arith.constant 2 : index
     %c3 = arith.constant 3 : index
-    %idx = mesh.neighbors_linear_indices on @mesh[%c1, %c2, %c3] split_axes = [1] : index
+    %idx = shard.neighbors_linear_indices on @grid[%c1, %c2, %c3] split_axes = [1] : index
     ```
     The above returns two indices, `633` and `693`, which correspond to the
     index of the previous process `(1, 1, 3)`, and the next process 
@@ -179,12 +179,12 @@ def Mesh_NeighborsLinearIndicesOp : Mesh_Op<"neighbors_linear_indices", [
     A negative value is returned if there is no neighbor in the respective
     direction along the given `split_axes`.
   }];
-  let arguments = (ins FlatSymbolRefAttr:$mesh,
+  let arguments = (ins FlatSymbolRefAttr:$grid,
                        Variadic<Index>:$device,
-                       Mesh_MeshAxesAttr:$split_axes);
+                       Shard_GridAxesAttr:$split_axes);
   let results = (outs Index:$neighbor_down, Index:$neighbor_up);
   let assemblyFormat =  [{
-      `on` $mesh `[` $device `]`
+      `on` $grid `[` $device `]`
       `split_axes` `=` $split_axes
       attr-dict `:` type(results)
   }];
@@ -194,7 +194,7 @@ def Mesh_NeighborsLinearIndicesOp : Mesh_Op<"neighbors_linear_indices", [
 // Sharding operations.
 //===----------------------------------------------------------------------===//
 
-def Mesh_ShardingOp : Mesh_Op<"sharding", [
+def Shard_ShardingOp : Shard_Op<"sharding", [
     Pure,
     AttrSizedOperandSegments,
     DeclareOpInterfaceMethods<SymbolUserOpInterface>,
@@ -202,18 +202,18 @@ def Mesh_ShardingOp : Mesh_Op<"sharding", [
   ]> {
   let summary = "Define a sharding of a tensor.";
   let description = [{
-    The MeshSharding specifies how a tensor is sharded and distributed across the
-    process mesh. It is typically used in a `mesh.shard` operation.
+    The Sharding specifies how a tensor is sharded and distributed across the
+    process shard. It is typically used in a `shard.shard` operation.
     The operation has the following attributes and operands:
 
-    1. `mesh`: this attribute is a FlatSymbolRefAttr that refers to the device
-    mesh where the distributed tensor is placed. The symbol must resolve to a
-    `mesh.mesh` operation.
+    1. `grid`: this attribute is a FlatSymbolRefAttr that refers to the device
+    grid where the distributed tensor is placed. The symbol must resolve to a
+    `shard.grid` operation.
 
     2. `split_axes`: is an array composed of int64_t sub-arrays. The outer array's
     maximum size is the `rank` of the related tensor. For the i-th sub-array, if
     its value is [x, y], it indicates that the tensor's i-th dimension is splitted
-    along the x and y axes of the device mesh.
+    along the x and y axes of the device grid.
 
     3. [Optional] Sizes of halos to be added for each sharded tensor dimension.
     `halo_sizes` is provided as a flattened 1d array of i64s, 2 values for each
@@ -233,7 +233,7 @@ def Mesh_ShardingOp : Mesh_Op<"sharding", [
     
     Assuming a 3d-tensor of shape 32x32x32 with the first 2 dimensions being sharded,
     `sharded_dims_offsets` = [0, 24, 32, 0, 20, 32] means that the first device of
-    the device-mesh will get a shard of shape 24x20x32 and the second device will get
+    the device-grid will get a shard of shape 24x20x32 and the second device will get
     a shard of shape 8x12x32. `?` indicates dynamic shard dimensions.
     
     `halo_sizes` and `sharded_dims_offsets` are mutually exclusive.
@@ -241,101 +241,101 @@ def Mesh_ShardingOp : Mesh_Op<"sharding", [
     Examples:
 
     ```
-    mesh.mesh @mesh0(shape = 2x2x4)
-    mesh.mesh @mesh1d_4(shape = 4)
+    shard.grid @grid0(shape = 2x2x4)
+    shard.grid @grid1d_4(shape = 4)
 
-    // The tensor is fully replicated on @mesh0.
+    // The tensor is fully replicated on @grid0.
     // Currently, there must be at least one sub-array present in axes, even
     // if it's empty. Otherwise, a parsing error will occur.
-    %sharding0 = mesh.sharding @mesh0 split_axes = [[]]
+    %sharding0 = shard.sharding @grid0 split_axes = [[]]
 
-    // The tensor is sharded on the first dimension along axis 0 of @mesh0
-    %sharding1 = mesh.sharding @mesh0 split_axes = [[0]]
+    // The tensor is sharded on the first dimension along axis 0 of @grid0
+    %sharding1 = shard.sharding @grid0 split_axes = [[0]]
 
-    // Could be used for a mesh.shard op
-    %sharded0 = mesh.shard %arg0 to %sharding3 : tensor<4x8xf32>
+    // Could be used for a shard.shard op
+    %sharded0 = shard.shard %arg0 to %sharding3 : tensor<4x8xf32>
 
-    // The tensor is sharded on its first dimension along axis 0 of @mesh0 and
+    // The tensor is sharded on its first dimension along axis 0 of @grid0 and
     // and it has halo-sizes of 1 and 2 on the sharded dim.
-    %halo_sharding = mesh.sharding @mesh0 split_axes = [[0]] halo_sizes = [1, 2]
-    %sharded1 = mesh.shard %arg0 to %halo_sharding : tensor<4x8xf32>
+    %halo_sharding = shard.sharding @grid0 split_axes = [[0]] halo_sizes = [1, 2]
+    %sharded1 = shard.shard %arg0 to %halo_sharding : tensor<4x8xf32>
     
-    // The tensor is sharded on its second dimension along axis 0 of @mesh1d_4
+    // The tensor is sharded on its second dimension along axis 0 of @grid1d_4
     // and it has pre-defined shard sizes. The shards of the devices will have
     // the following shapes: [4x2, 4x3, 4x4, 4x5]
-    %sharding4 = mesh.sharding @mesh1d_4 split_axes = [[], [0]] sharded_dims_offsets = [0, 2, 5, 9, 14]
-    %sharded2 = mesh.shard %arg0 to %sharding4 : tensor<4x14xf32>
+    %sharding4 = shard.sharding @grid1d_4 split_axes = [[], [0]] sharded_dims_offsets = [0, 2, 5, 9, 14]
+    %sharded2 = shard.shard %arg0 to %sharding4 : tensor<4x14xf32>
     ```
   }];
     
   let arguments = (ins
-    FlatSymbolRefAttr:$mesh,
-    Mesh_MeshAxesArrayAttr:$split_axes,
+    FlatSymbolRefAttr:$grid,
+    Shard_GridAxesArrayAttr:$split_axes,
     DefaultValuedAttr<DenseI64ArrayAttr, "{}">:$static_sharded_dims_offsets,
     Variadic<I64>:$dynamic_sharded_dims_offsets,
     DefaultValuedAttr<DenseI64ArrayAttr, "{}">:$static_halo_sizes,
     Variadic<I64>:$dynamic_halo_sizes
   );
   let results = (outs
-    Mesh_Sharding:$result
+    Shard_Sharding:$result
   );
   let assemblyFormat = [{
-    $mesh
+    $grid
     `split_axes` `=` $split_axes
     (`halo_sizes` `=` custom<DynamicIndexList>($dynamic_halo_sizes, $static_halo_sizes)^)?
     (`sharded_dims_offsets` `=` custom<DynamicIndexList>($dynamic_sharded_dims_offsets, $static_sharded_dims_offsets)^)?
     attr-dict `:` type($result)
   }];
   let builders = [
-    OpBuilder<(ins "FlatSymbolRefAttr":$mesh,
-                   "ArrayRef<MeshAxesAttr>":$split_axes,
+    OpBuilder<(ins "FlatSymbolRefAttr":$grid,
+                   "ArrayRef<GridAxesAttr>":$split_axes,
                    CArg<"ArrayRef<int64_t>", "{}">:$static_halo_sizes,
                    CArg<"ArrayRef<int64_t>", "{}">:$static_sharded_dims_offsets)>,
-    OpBuilder<(ins "FlatSymbolRefAttr":$mesh,
-                   "ArrayRef<MeshAxesAttr>":$split_axes,
+    OpBuilder<(ins "FlatSymbolRefAttr":$grid,
+                   "ArrayRef<GridAxesAttr>":$split_axes,
                    "::mlir::ArrayRef<::mlir::OpFoldResult>":$halo_sizes,
                    "::mlir::ArrayRef<::mlir::OpFoldResult>":$sharded_dims_offsets)>,
-    OpBuilder<(ins "llvm::StringRef":$mesh,
-                   "ArrayRef<MeshAxesAttr>":$split_axes,
+    OpBuilder<(ins "llvm::StringRef":$grid,
+                   "ArrayRef<GridAxesAttr>":$split_axes,
                    CArg<"ArrayRef<int64_t>", "{}">:$static_halo_sizes,
                    CArg<"ArrayRef<int64_t>", "{}">:$static_sharded_dims_offsets
     )>,
-    OpBuilder<(ins "mlir::mesh::MeshSharding":$from)>
+    OpBuilder<(ins "mlir::shard::Sharding":$from)>
   ];
   let hasVerifier = 1;
   let hasCanonicalizer = 1;
 }
 
-def Mesh_GetShardingOp : Mesh_Op<"get_sharding", [Pure]> {
+def Shard_GetShardingOp : Shard_Op<"get_sharding", [Pure]> {
   let summary = "Get the sharding of the given tensor.";
   let description = [{
-    This operation returns the sharding of the given tensor as a MeshSharding.
+    This operation returns the sharding of the given tensor as a Sharding.
   }];
   let arguments = (ins
     AnyRankedTensor:$source
   );
   let results = (outs
-    Mesh_Sharding:$result
+    Shard_Sharding:$result
   );
   let assemblyFormat = [{
     $source attr-dict `:` type($source) `->` type($result)
   }];
 }
 
-def Mesh_ShardShapeOp : Mesh_Op<"shard_shape", [
+def Shard_ShardShapeOp : Shard_Op<"shard_shape", [
     Pure, AttrSizedOperandSegments,
     DeclareOpInterfaceMethods<OpAsmOpInterface, ["getAsmResultNames"]>
   ]> {
   let summary = "Get the shard shape for a given process/device.";
   let description = [{
-    The device/process id is a multi-index of the device/process in the mesh.
-    This operation might be used during spmdization when the shard shape depends
-    on (non-constant) values used in `mesh.sharding`.
+    The device/process id is a multi-index of the device/process in the shard.
+    This operation might be used during partition when the shard shape depends
+    on (non-constant) values used in `shard.sharding`.
   }];
   let arguments = (ins
     DenseI64ArrayAttr:$dims,
     Variadic<Index>:$dims_dynamic,
-    Mesh_Sharding:$sharding,
+    Shard_Sharding:$sharding,
     DenseI64ArrayAttr:$device,
     Variadic<Index>:$device_dynamic
   );
@@ -351,23 +351,23 @@ def Mesh_ShardShapeOp : Mesh_Op<"shard_shape", [
   ];
 }
 
-def Mesh_ShardOp : Mesh_Op<"shard", [
+def Shard_ShardOp : Shard_Op<"shard", [
     Pure,
     AllTypesMatch<["result", "src"]>,
     DeclareOpInterfaceMethods<OpAsmOpInterface, ["getAsmResultNames"]>
   ]> {
-  let summary = "Annotate on how a tensor is sharded across a mesh.";
+  let summary = "Annotate on how a tensor is sharded across a shard.";
   let description = [{
-    The mesh.shard operation is designed to specify and guide the sharding
-    behavior of a tensor value across a mesh topology. This operation has two
+    The shard.shard operation is designed to specify and guide the sharding
+    behavior of a tensor value across a grid topology. This operation has two
     operands and two optional attributes:
 
     1. `input`: This operand represents the tensor value that needs to be
     annotated for sharding.
 
-    2. `sharding`: This attribute is type of `MeshShardingType`, which is the core data
-    structure to represent distribution of a tensor on a mesh. it is typically defiend
-    by an `mesh.sharding` operation.
+    2. `sharding`: This attribute is type of `ShardingType`, which is the core data
+    structure to represent distribution of a tensor on a shard. it is typically defined
+    by an `shard.sharding` operation.
 
     3. `annotate_for_users`: A unit attribute addressing the scenario when a
     tensor's sharding annotation differs based on its context of use (either as
@@ -378,36 +378,36 @@ def Mesh_ShardOp : Mesh_Op<"shard", [
 
     Example:
     ```
-    func.func @only_result_annotated(%arg0 : tensor<4x8xf32>) -> () {
-      %sharding = mesh.sharding @mesh0 split_axes = [[0]] : !mesh.sharding
-      %0 = mesh.shard %arg0 to %sharding : tensor<4x8xf32>
+  func.func @only_result_annotated(%arg0 : tensor<4x8xf32>) -> () {
+      %sharding = shard.sharding @grid0 split_axes = [[0]] : !shard.sharding
+      %0 = shard.shard %arg0 to %sharding : tensor<4x8xf32>
       ...
     }
 
     func.func @only_operand_annotated(%arg0 : tensor<4x8xf32>) -> () {
-      %sharding = mesh.sharding @mesh0 split_axes = [[0]] : !mesh.sharding
-      %0 = mesh.shard %arg0 to %sharding annotate_for_users : tensor<4x8xf32>
+      %sharding = shard.sharding @grid0 split_axes = [[0]] : !shard.sharding
+      %0 = shard.shard %arg0 to %sharding annotate_for_users : tensor<4x8xf32>
       ...
     }
     
     func.func @two_operands_annotated(%arg0 : tensor<4x8xf32>, %arg1 : tensor<16x8xf32>) -> () {
-      %sharding = mesh.sharding @mesh0 split_axes = [[0]] : !mesh.sharding
-      %0 = mesh.shard %arg0 to %sharding annotate_for_users : tensor<4x8xf32>
-      %1 = mesh.shard %arg1 to %sharding annotate_for_users : tensor<16x8xf32>
+      %sharding = shard.sharding @grid0 split_axes = [[0]] : !shard.sharding
+      %0 = shard.shard %arg0 to %sharding annotate_for_users : tensor<4x8xf32>
+      %1 = shard.shard %arg1 to %sharding annotate_for_users : tensor<16x8xf32>
       ...
     }
 
-    // The first mesh.shard op applies to %arg0, the second mesh.shard op
-    // applies for the operand of op0, the third mesh.shard op applies for the
+    // The first shard.shard op applies to %arg0, the second shard.shard op
+    // applies for the operand of op0, the third shard.shard op applies for the
     // operand of op2
     func.func @both_result_and_multi_operands_annotated(
         %arg0 : tensor<4x8xf32>) -> () {
-      %sharding = mesh.sharding @mesh0 split_axes = [[0]] : !mesh.sharding
-      %0 = mesh.shard %arg0 to %sharding : tensor<4x8xf32>
-      %sharding1 = mesh.sharding @mesh0 split_axes = [[1]] : !mesh.sharding
-      %1 = mesh.shard %0 to %sharding1 annotate_for_users : tensor<4x8xf32>
-      %sharding2 = mesh.sharding @mesh0 split_axes = [[2]] : !mesh.sharding
-      %2 = mesh.shard %0 to %sharding2 annotate_for_users : tensor<4x8xf32>
+      %sharding = shard.sharding @grid0 split_axes = [[0]] : !shard.sharding
+      %0 = shard.shard %arg0 to %sharding : tensor<4x8xf32>
+      %sharding1 = shard.sharding @grid0 split_axes = [[1]] : !shard.sharding
+      %1 = shard.shard %0 to %sharding1 annotate_for_users : tensor<4x8xf32>
+      %sharding2 = shard.sharding @grid0 split_axes = [[2]] : !shard.sharding
+      %2 = shard.shard %0 to %sharding2 annotate_for_users : tensor<4x8xf32>
       "op0"(%1) : ...
       "op1"(%2) : ...
       ...
@@ -418,44 +418,44 @@ def Mesh_ShardOp : Mesh_Op<"shard", [
     ```
     func.func @annotate_on_same_result_with_different_sharding(
         %arg0 : tensor<4x8xf32>) -> () {
-      %sharding1 = mesh.sharding @mesh0 split_axes = [[0]] : !mesh.sharding
-      %sharding2 = mesh.sharding @mesh0 split_axes = [[1]] : !mesh.sharding
-      %0 = mesh.shard %arg0 to $sharding1 : tensor<4x8xf32>
-      %1 = mesh.shard %0 to sharding2 : tensor<4x8xf32>
+      %sharding1 = shard.sharding @grid0 split_axes = [[0]] : !shard.sharding
+      %sharding2 = shard.sharding @grid0 split_axes = [[1]] : !shard.sharding
+      %0 = shard.shard %arg0 to $sharding1 : tensor<4x8xf32>
+      %1 = shard.shard %0 to sharding2 : tensor<4x8xf32>
       ...
     }
 
     func.func @annotate_on_same_result_same_value_with_different_sharding(
         %arg0 : tensor<4x8xf32>) -> () {
-      %sharding1 = mesh.sharding @mesh0 split_axes = [[0]] : !mesh.sharding
-      %sharding2 = mesh.sharding @mesh0 split_axes = [[1]] : !mesh.sharding
-      %0 = mesh.shard %arg0 to %sharding1 : tensor<4x8xf32>
-      %1 = mesh.shard %arg0 to %sharding2 : tensor<4x8xf32>
+      %sharding1 = shard.sharding @grid0 split_axes = [[0]] : !shard.sharding
+      %sharding2 = shard.sharding @grid0 split_axes = [[1]] : !shard.sharding
+      %0 = shard.shard %arg0 to %sharding1 : tensor<4x8xf32>
+      %1 = shard.shard %arg0 to %sharding2 : tensor<4x8xf32>
       ...
     }
 
     func.func @annotate_on_same_operand_with_different_sharding(
         %arg0 : tensor<4x8xf32>) -> () {
-      %sharding1 = mesh.sharding @mesh0 split_axes = [[0]] : !mesh.sharding
-      %sharding2 = mesh.sharding @mesh0 split_axes = [[1]] : !mesh.sharding
-      %0 = mesh.shard %arg0 to %sharding1 annotate_for_users : tensor<4x8xf32>
-      %1 = mesh.shard %0 to %sharding2 annotate_for_users : tensor<4x8xf32>
+      %sharding1 = shard.sharding @grid0 split_axes = [[0]] : !shard.sharding
+      %sharding2 = shard.sharding @grid0 split_axes = [[1]] : !shard.sharding
+      %0 = shard.shard %arg0 to %sharding1 annotate_for_users : tensor<4x8xf32>
+      %1 = shard.shard %0 to %sharding2 annotate_for_users : tensor<4x8xf32>
       ...
     }
 
     func.func @result_annotated_after_operand(
         %arg0 : tensor<4x8xf32>) -> () {
-      %sharding1 = mesh.sharding @mesh0 split_axes = [[0]] : !mesh.sharding
-      %sharding2 = mesh.sharding @mesh0 split_axes = [[1]] : !mesh.sharding
-      %0 = mesh.shard %arg0 to %sharding1 annotate_for_users : tensor<4x8xf32>
-      %1 = mesh.shard %0 to %sharding2 : tensor<4x8xf32>
+      %sharding1 = shard.sharding @grid0 split_axes = [[0]] : !shard.sharding
+      %sharding2 = shard.sharding @grid0 split_axes = [[1]] : !shard.sharding
+      %0 = shard.shard %arg0 to %sharding1 annotate_for_users : tensor<4x8xf32>
+      %1 = shard.shard %0 to %sharding2 : tensor<4x8xf32>
       ...
     }
     ```
   }];
   let arguments = (ins
     AnyRankedTensor:$src,
-    Mesh_Sharding:$sharding,
+    Shard_Sharding:$sharding,
     UnitAttr:$annotate_for_users
   );
   let results = (outs
@@ -473,34 +473,34 @@ def Mesh_ShardOp : Mesh_Op<"shard", [
 // collective communication ops
 //===----------------------------------------------------------------------===//
 
-class Mesh_CollectiveCommunicationOpBase<
+class Shard_CollectiveCommunicationOpBase<
     string mnemonic, list<Trait> traits = []> :
-    Mesh_Op<mnemonic,
+    Shard_Op<mnemonic,
       !listconcat(traits,
         [
           DeclareOpInterfaceMethods<SymbolUserOpInterface>,
           DeclareOpInterfaceMethods<OpAsmOpInterface, ["getAsmResultNames"]>
         ])> {
   dag commonArgs = (ins
-    FlatSymbolRefAttr:$mesh,
-    DefaultValuedAttr<Mesh_MeshAxesAttr, "{}">:$mesh_axes
+    FlatSymbolRefAttr:$grid,
+    DefaultValuedAttr<Shard_GridAxesAttr, "{}">:$grid_axes
   );
 }
 
-def Mesh_AllGatherOp : Mesh_CollectiveCommunicationOpBase<"all_gather", [
+def Shard_AllGatherOp : Shard_CollectiveCommunicationOpBase<"all_gather", [
     Pure,
     SameOperandsAndResultElementType,
     SameOperandsAndResultRank,
   ]> {
-  let summary = "All-gather over a device mesh.";
+  let summary = "All-gather over a device grid.";
   let description = [{
     Gathers along the `gather_axis` tensor axis.
 
     Example:
     ```mlir
-    mesh.mesh @mesh0(shape = 2x2)
+    shard.grid @grid0(shape = 2x2)
     ...
-    %1 = mesh.all_gather %0 on @mesh0 mesh_axes = [1] gather_axis = 1
+    %1 = shard.all_gather %0 on @grid0 grid_axes = [1] gather_axis = 1
       : tensor<2x2xi8> -> tensor<2x4xi8>
     ```
     Input:
@@ -535,16 +535,16 @@ def Mesh_AllGatherOp : Mesh_CollectiveCommunicationOpBase<"all_gather", [
     AnyNon0RankedTensor:$result
   );
   let assemblyFormat = [{
-    $input `on` $mesh (`mesh_axes` `=` $mesh_axes^)? `gather_axis` `=` $gather_axis
+    $input `on` $grid (`grid_axes` `=` $grid_axes^)? `gather_axis` `=` $gather_axis
     attr-dict `:` type($input) `->` type($result)
   }];
   let hasCanonicalizer = 1;
 }
 
-def Mesh_AllReduceOp : Mesh_CollectiveCommunicationOpBase<"all_reduce", [
+def Shard_AllReduceOp : Shard_CollectiveCommunicationOpBase<"all_reduce", [
     Pure,
     SameOperandsAndResultShape]> {
-  let summary = "All-reduce over a device mesh.";
+  let summary = "All-reduce over a device grid.";
   let description = [{
     The accumulation element type is specified by the result type and
     it does not need to match the input element type.
@@ -556,34 +556,34 @@ def Mesh_AllReduceOp : Mesh_CollectiveCommunicationOpBase<"all_reduce", [
 
     Example:
     ```
-    %1 = mesh.all_reduce %0 on @mesh0 mesh_axes = [1, 0] reduction = <max>
+    %1 = shard.all_reduce %0 on @grid0 grid_axes = [1, 0] reduction = <max>
       : tensor<3x4xf32> -> tensor<3x4xf64>
     ```
   }];
   let arguments = !con(commonArgs, (ins
     AnyTypeOf<[AnyMemRef, AnyRankedTensor]>:$input,
-    DefaultValuedAttr<Mesh_ReductionKindAttr, "::mlir::mesh::ReductionKind::Sum">:$reduction
+    DefaultValuedAttr<Shard_ReductionKindAttr, "::mlir::shard::ReductionKind::Sum">:$reduction
   ));
   let results = (outs
     AnyTypeOf<[AnyMemRef, AnyRankedTensor]>:$result
   );
   let assemblyFormat = [{
-    $input `on` $mesh (`mesh_axes` `=` $mesh_axes^)? (`reduction` `=` $reduction^)?
+    $input `on` $grid (`grid_axes` `=` $grid_axes^)? (`reduction` `=` $reduction^)?
     attr-dict `:` type($input) `->` type($result)
   }];
   let hasCanonicalizer = 1;
     let builders = [
-    OpBuilder<(ins "Value":$input, "StringRef":$mesh,
-      "ArrayRef<MeshAxis>":$meshAxes, "ReductionKind":$reduction)>
+    OpBuilder<(ins "Value":$input, "StringRef":$grid,
+      "ArrayRef<GridAxis>":$gridAxes, "ReductionKind":$reduction)>
   ];
 }
 
-def Mesh_AllSliceOp : Mesh_CollectiveCommunicationOpBase<"all_slice", [
+def Shard_AllSliceOp : Shard_CollectiveCommunicationOpBase<"all_slice", [
     Pure,
     SameOperandsAndResultElementType,
     SameOperandsAndResultRank
   ]> {
-  let summary = "All-slice over a device mesh. This is the inverse of all-gather.";
+  let summary = "All-slice over a device grid. This is the inverse of all-gather.";
   let description = [{
     Slice along the `slice_axis` tensor axis.
     This operation can be thought of as the inverse of all-gather.
@@ -593,9 +593,9 @@ def Mesh_AllSliceOp : Mesh_CollectiveCommunicationOpBase<"all_slice", [
 
     Example:
     ```mlir
-    mesh.mesh @mesh0(shape = 2x2)
+    shard.grid @grid0(shape = 2x2)
     ...
-    %1 = mesh.all_slice %0 on @mesh0 mesh_axes = [1] slice_axis = 1
+    %1 = shard.all_slice %0 on @grid0 grid_axes = [1] slice_axis = 1
       : tensor<2x4xi8> -> tensor<2x2xi8>
     ```
     Input:
@@ -630,30 +630,30 @@ def Mesh_AllSliceOp : Mesh_CollectiveCommunicationOpBase<"all_slice", [
     AnyNon0RankedTensor:$result
   );
   let assemblyFormat = [{
-    $input `on` $mesh (`mesh_axes` `=` $mesh_axes^)? `slice_axis` `=` $slice_axis
+    $input `on` $grid (`grid_axes` `=` $grid_axes^)? `slice_axis` `=` $slice_axis
     attr-dict `:` type($input) `->` type($result)
   }];
   let hasCanonicalizer = 1;
   let builders = [
-    OpBuilder<(ins "Value":$input, "MeshOp":$mesh, "ArrayRef<MeshAxis>":$meshAxes, "int64_t":$sliceAxis)>,
-    OpBuilder<(ins "Type":$result_type, "Value":$input, "StringRef":$mesh, "ArrayRef<MeshAxis>":$meshAxes, "int64_t":$sliceAxis)>
+    OpBuilder<(ins "Value":$input, "GridOp":$grid, "ArrayRef<GridAxis>":$gridAxes, "int64_t":$sliceAxis)>,
+    OpBuilder<(ins "Type":$result_type, "Value":$input, "StringRef":$grid, "ArrayRef<GridAxis>":$gridAxes, "int64_t":$sliceAxis)>
   ];
 }
 
-def Mesh_AllToAllOp : Mesh_CollectiveCommunicationOpBase<"all_to_all", [
+def Shard_AllToAllOp : Shard_CollectiveCommunicationOpBase<"all_to_all", [
     Pure,
     SameOperandsAndResultElementType,
     SameOperandsAndResultRank]> {
-  let summary = "All-to-all over a device mesh.";
+  let summary = "All-to-all over a device grid.";
   let description = [{
     Performs an all-to-all on tensor pieces split along `split_axis`.
     The resulting pieces are concatenated along `concat_axis` on ech device.
 
     Example:
     ```
-    mesh.mesh @mesh0(shape = 3)
+    shard.grid @grid0(shape = 3)
     ...
-    %1 = mesh.all_to_all %0 on @mesh0 mesh_axes = [0]
+    %1 = shard.all_to_all %0 on @grid0 grid_axes = [0]
       split_axis = 0 concat_axis = 0
       : tensor<3x2xi8> -> tensor<3x2xi8>
     ```
@@ -687,7 +687,7 @@ def Mesh_AllToAllOp : Mesh_CollectiveCommunicationOpBase<"all_to_all", [
     AnyNon0RankedTensor:$result
   );
   let assemblyFormat = [{
-    $input `on` $mesh (`mesh_axes` `=` $mesh_axes^)?
+    $input `on` $grid (`grid_axes` `=` $grid_axes^)?
     `split_axis` `=` $split_axis
     `concat_axis` `=` $concat_axis
     attr-dict `:` type($input) `->` type($result)
@@ -695,24 +695,24 @@ def Mesh_AllToAllOp : Mesh_CollectiveCommunicationOpBase<"all_to_all", [
   let hasCanonicalizer = 1;
 }
 
-def Mesh_BroadcastOp : Mesh_CollectiveCommunicationOpBase<"broadcast", [
+def Shard_BroadcastOp : Shard_CollectiveCommunicationOpBase<"broadcast", [
     Pure,
     AllShapesMatch<["input", "result"]>,
     AllElementTypesMatch<["input", "result"]>
   ]> {
-  let summary = "Broadcast over a device mesh.";
+  let summary = "Broadcast over a device grid.";
   let description = [{
     Broadcast the tensor on `root` to all devices in each respective group.
-    The operation broadcasts along mesh axes `mesh_axes`.
+    The operation broadcasts along grid axes `grid_axes`.
     The `root` device specifies the in-group multi-index that is broadcast to
     all other devices in the group.
     
     Example:
     ```
-    mesh.mesh @mesh0(shape = 2x2)
+    shard.grid @grid0(shape = 2x2)
 
-    %1 = mesh.broadcast %0 on @mesh0
-      mesh_axes = [0]
+    %1 = shard.broadcast %0 on @grid0
+      grid_axes = [0]
       root = [0]
       : (tensor<2xi8>) -> tensor<2xi8>
     ```
@@ -744,31 +744,31 @@ def Mesh_BroadcastOp : Mesh_CollectiveCommunicationOpBase<"broadcast", [
     AnyRankedTensor:$result
   );
   let assemblyFormat = [{
-    $input `on` $mesh (`mesh_axes` `=` $mesh_axes^)?
+    $input `on` $grid (`grid_axes` `=` $grid_axes^)?
     `root` `=` custom<DynamicIndexList>($root_dynamic, $root)
     attr-dict `:` functional-type(operands, results)
   }];
   let hasCanonicalizer = 1;
 }
 
-def Mesh_GatherOp : Mesh_CollectiveCommunicationOpBase<"gather", [
+def Shard_GatherOp : Shard_CollectiveCommunicationOpBase<"gather", [
     Pure,
     AllRanksMatch<["input", "result"]>,
     AllElementTypesMatch<["input", "result"]>
   ]> {
-  let summary = "Gather over a device mesh.";
+  let summary = "Gather over a device grid.";
   let description = [{
     Gathers on device `root` along the `gather_axis` tensor axis.
-    `root` specifies the coordinates of a device along `mesh_axes`.
+    `root` specifies the coordinates of a device along `grid_axes`.
     It uniquely identifies the root device for each device group.
     The result tensor on non-root devices is undefined.
     Using it will result in undefined behavior.
 
     Example:
     ```mlir
-    mesh.mesh @mesh0(shape = 2x2)
+    shard.grid @grid0(shape = 2x2)
     ...
-    %1 = mesh.gather %0 on @mesh0 mesh_axes = [1]
+    %1 = shard.gather %0 on @grid0 grid_axes = [1]
       gather_axis = 1 root = [1]
       : (tensor<2x2xi8>) -> tensor<2x4xi8>
     ```
@@ -807,7 +807,7 @@ def Mesh_GatherOp : Mesh_CollectiveCommunicationOpBase<"gather", [
     AnyNon0RankedTensor:$result
   );
   let assemblyFormat = [{
-    $input `on` $mesh (`mesh_axes` `=` $mesh_axes^)?
+    $input `on` $grid (`grid_axes` `=` $grid_axes^)?
     `gather_axis` `=` $gather_axis
     `root` `=` custom<DynamicIndexList>($root_dynamic, $root)
     attr-dict `:` functional-type(operands, results)
@@ -815,11 +815,11 @@ def Mesh_GatherOp : Mesh_CollectiveCommunicationOpBase<"gather", [
   let hasCanonicalizer = 1;
 }
 
-def Mesh_RecvOp : Mesh_CollectiveCommunicationOpBase<"recv", [
+def Shard_RecvOp : Shard_CollectiveCommunicationOpBase<"recv", [
     AllShapesMatch<["input", "result"]>,
     AllElementTypesMatch<["input", "result"]>
   ]> {
-  let summary = "Send over a device mesh.";
+  let summary = "Send over a device grid.";
   let description = [{
     Receive from a device within a device group.
   }];
@@ -832,21 +832,21 @@ def Mesh_RecvOp : Mesh_CollectiveCommunicationOpBase<"recv", [
     AnyRankedTensor:$result
   );
   let assemblyFormat = [{
-    $input `on` $mesh (`mesh_axes` `=` $mesh_axes^)?
+    $input `on` $grid (`grid_axes` `=` $grid_axes^)?
     (`source` `=` custom<DynamicIndexList>($source_dynamic, $source)^)?
     attr-dict `:` functional-type(operands, results)
   }];
   let hasCanonicalizer = 1;
 }
 
-def Mesh_ReduceOp : Mesh_CollectiveCommunicationOpBase<"reduce", [
+def Shard_ReduceOp : Shard_CollectiveCommunicationOpBase<"reduce", [
     Pure,
     AllShapesMatch<["input", "result"]>
   ]> {
-  let summary = "Reduce over a device mesh.";
+  let summary = "Reduce over a device grid.";
   let description = [{
     Reduces on device `root` within each device group.
-    `root` specifies the coordinates of a device along `mesh_axes`.
+    `root` specifies the coordinates of a device along `grid_axes`.
     It uniquely identifies the root device within its device group.
     The accumulation element type is specified by the result type and
     it does not need to match the input element type.
@@ -858,14 +858,14 @@ def Mesh_ReduceOp : Mesh_CollectiveCommunicationOpBase<"reduce", [
 
     Example:
     ```
-    %1 = mesh.reduce %0 on @mesh0 mesh_axes = [1, 0]
+    %1 = shard.reduce %0 on @grid0 grid_axes = [1, 0]
       reduction = <max> root = [2, 3]
       : (tensor<3x4xf32>) -> tensor<3x4xf64>
     ```
   }];
   let arguments = !con(commonArgs, (ins
     AnyRankedTensor:$input,
-    DefaultValuedAttr<Mesh_ReductionKindAttr, "::mlir::mesh::ReductionKind::Sum">:$reduction,
+    DefaultValuedAttr<Shard_ReductionKindAttr, "::mlir::shard::ReductionKind::Sum">:$reduction,
     DenseI64ArrayAttr:$root,
     Variadic<Index>:$root_dynamic
   ));
@@ -873,7 +873,7 @@ def Mesh_ReduceOp : Mesh_CollectiveCommunicationOpBase<"reduce", [
     AnyRankedTensor:$result
   );
   let assemblyFormat = [{
-    $input `on` $mesh (`mesh_axes` `=` $mesh_axes^)?
+    $input `on` $grid (`grid_axes` `=` $grid_axes^)?
     (`reduction` `=` $reduction^)?
     `root` `=` custom<DynamicIndexList>($root_dynamic, $root)
     attr-dict `:` functional-type(operands, results)
@@ -881,19 +881,19 @@ def Mesh_ReduceOp : Mesh_CollectiveCommunicationOpBase<"reduce", [
   let hasCanonicalizer = 1;
 }
 
-def Mesh_ReduceScatterOp : Mesh_CollectiveCommunicationOpBase<"reduce_scatter", [
+def Shard_ReduceScatterOp : Shard_CollectiveCommunicationOpBase<"reduce_scatter", [
     Pure,
     SameOperandsAndResultRank]> {
-  let summary = "Reduce-scatter over a device mesh.";
+  let summary = "Reduce-scatter over a device grid.";
   let description = [{
     After the reduction, the result is scattered within each device group.
     The tensor is split along `scatter_axis` and the pieces distributed
     across the device group.
     Example:
     ```
-    mesh.mesh @mesh0(shape = 2x2)
+    shard.grid @grid0(shape = 2x2)
     ...
-    %1 = mesh.reduce_scatter %0 on @mesh0 mesh_axes = [1]
+    %1 = shard.reduce_scatter %0 on @grid0 grid_axes = [1]
       reduction = <max> scatter_axis = 0
       : tensor<3x4xf32> -> tensor<1x4xf64>
     ```
@@ -928,14 +928,14 @@ def Mesh_ReduceScatterOp : Mesh_CollectiveCommunicationOpBase<"reduce_scatter",
   }];
   let arguments = !con(commonArgs, (ins
     AnyNon0RankedTensor:$input,
-    DefaultValuedAttr<Mesh_ReductionKindAttr, "::mlir::mesh::ReductionKind::Sum">:$reduction,
+    DefaultValuedAttr<Shard_ReductionKindAttr, "::mlir::shard::ReductionKind::Sum">:$reduction,
     IndexAttr:$scatter_axis
   ));
   let results = (outs
     AnyRankedTensor:$result
   );
   let assemblyFormat = [{
-    $input `on` $mesh (`mesh_axes` `=` $mesh_axes^)?
+    $input `on` $grid (`grid_axes` `=` $grid_axes^)?
     (`reduction` `=` $reduction^)?
     `scatter_axis` `=` $scatter_axis
     attr-dict `:` type($input) `->` type($result)
@@ -943,20 +943,20 @@ def Mesh_ReduceScatterOp : Mesh_CollectiveCommunicationOpBase<"reduce_scatter",
   let hasCanonicalizer = 1;
 }
 
-def Mesh_ScatterOp : Mesh_CollectiveCommunicationOpBase<"scatter", [
+def Shard_ScatterOp : Shard_CollectiveCommunicationOpBase<"scatter", [
     Pure,
     AllRanksMatch<["input", "result"]>,
     AllElementTypesMatch<["input", "result"]>
   ]> {
-  let summary = "Scatter over a device mesh.";
+  let summary = "Scatter over a device grid.";
   let description = [{
     For each device group split the input tensor on the `root` device along
     axis `scatter_axis` and scatter the parts across the group devices.
 
     Example:
     ```
-    mesh.mesh @mesh0(shape = 2x2)
-    %1 = mesh.scatter %0 on @mesh0 mesh_axes = [0]
+    shard.grid @grid0(shape = 2x2)
+    %1 = shard.scatter %0 on @grid0 grid_axes = [0]
       scatter_axis = 0
       root = [1]
       : (tensor<2x2xi8>) -> tensor<1x2xi8>
@@ -1004,7 +1004,7 @@ def Mesh_ScatterOp : Mesh_CollectiveCommunicationOpBase<"scatter", [
     AnyRankedTensor:$result
   );
   let assemblyFormat = [{
-    $input `on` $mesh (`mesh_axes` `=` $mesh_axes^)?
+    $input `on` $grid (`grid_axes` `=` $grid_axes^)?
     `scatter_axis` `=` $scatter_axis
     `root` `=` custom<DynamicIndexList>($root_dynamic, $root)
     attr-dict `:` functional-type(operands, results)
@@ -1012,11 +1012,11 @@ def Mesh_ScatterOp : Mesh_CollectiveCommunicationOpBase<"scatter", [
   let hasCanonicalizer = 1;
 }
 
-def Mesh_SendOp : Mesh_CollectiveCommunicationOpBase<"send", [
+def Shard_SendOp : Shard_CollectiveCommunicationOpBase<"send", [
     AllShapesMatch<["input", "result"]>,
     AllElementTypesMatch<["input", "result"]>
   ]> {
-  let summary = "Send over a device mesh.";
+  let summary = "Send over a device grid.";
   let description = [{
     Send from one device to another within a device group.
   }];
@@ -1029,38 +1029,38 @@ def Mesh_SendOp : Mesh_CollectiveCommunicationOpBase<"send", [
     AnyRankedTensor:$result
   );
   let assemblyFormat = [{
-    $input `on` $mesh (`mesh_axes` `=` $mesh_axes^)?
+    $input `on` $grid (`grid_axes` `=` $grid_axes^)?
     `destination` `=` custom<DynamicIndexList>($destination_dynamic, $destination)
     attr-dict `:` functional-type(operands, results)
   }];
   let hasCanonicalizer = 1;
 }
 
-def Mesh_ShiftOp : Mesh_CollectiveCommunicationOpBase<"shift", [
+def Shard_ShiftOp : Shard_CollectiveCommunicationOpBase<"shift", [
     Pure,
     SameOperandsAndResultElementType,
     SameOperandsAndResultShape
   ]> {
-  let summary = "Shift over a device mesh.";
+  let summary = "Shift over a device grid.";
   let description = [{
-    Within each device group shift along mesh axis `shift_axis` by an offset
+    Within each device group shift along grid axis `shift_axis` by an offset
     `offset`.
     The result on devices that do not have a corresponding source is undefined.
-    `shift_axis` must be one of `mesh_axes`.
+    `shift_axis` must be one of `grid_axes`.
     If the `rotate` attribute is present,
     instead of a shift a rotation is done.
 
     Example:
     ```
-    mesh.mesh @mesh0(shape = 2x4)
-    %1 = mesh.shift on @mesh0 mesh_axes = [1]
+    shard.grid @grid0(shape = 2x4)
+    %1 = shard.shift on @grid0 grid_axes = [1]
       shift_axis = 1 offset = 2 rotate
       : tensor<2xi8> -> tensor<2xi8>
     ```
 
     Input:
     ```
-    mesh axis 1
+    grid axis 1
     ----------->
 
     +----+----+----+----+
@@ -1089,7 +1089,7 @@ def Mesh_ShiftOp : Mesh_CollectiveCommunicationOpBase<"shift", [
     AnyRankedTensor:$result
   );
   let assemblyFormat = [{
-    $input `on` $mesh (`mesh_axes` `=` $mesh_axes^)?
+    $input `on` $grid (`grid_axes` `=` $grid_axes^)?
     `shift_axis` `=` $shift_axis
     `offset` `=` $offset
     (`rotate` $rotate^)?
@@ -1098,7 +1098,7 @@ def Mesh_ShiftOp : Mesh_CollectiveCommunicationOpBase<"shift", [
   let hasCanonicalizer = 1;
 }
 
-def Mesh_UpdateHaloOp : Mesh_Op<"update_halo", [
+def Shard_UpdateHaloOp : Shard_Op<"update_halo", [
   Pure,
   DestinationStyleOpInterface,
   TypesMatchWith<
@@ -1120,14 +1120,14 @@ def Mesh_UpdateHaloOp : Mesh_Op<"update_halo", [
     `destination_halo_sizes/static_destination_halo_sizes` in source shard
     and destination/result shard.
 
-    `split_axes` specifies for each tensor axis along which mesh axes its halo
+    `split_axes` specifies for each tensor axis along which grid axes its halo
     data is updated.
 
   }];
   let arguments = (ins
     AnyTypeOf<[AnyNon0RankedMemRef, AnyNon0RankedTensor]>:$destination,
-    FlatSymbolRefAttr:$mesh,
-    Mesh_MeshAxesArrayAttr:$split_axes,
+    FlatSymbolRefAttr:$grid,
+    Shard_GridAxesArrayAttr:$split_axes,
     Variadic<I64>:$halo_sizes,
     DefaultValuedAttr<DenseI64ArrayAttr, "{}">:$static_halo_sizes
   );
@@ -1136,7 +1136,7 @@ def Mesh_UpdateHaloOp : Mesh_Op<"update_halo", [
   );
   let assemblyFormat = [{
     $destination
-    `on` $mesh
+    `on` $grid
     `split_axes` `=` $split_axes
     (`halo_sizes` `=` custom<DynamicIndexList>($halo_sizes, $static_halo_sizes)^)?
     attr-dict `:` type($result)
@@ -1145,4 +1145,4 @@ def Mesh_UpdateHaloOp : Mesh_Op<"update_halo", [
     MutableOperandRange getDpsInitsMutable() { return getDestinationMutable(); }
   }];
 }
-#endif // MLIR_DIALECT_MESH_IR_MESHOPS_TD
+#endif // MLIR_DIALECT_SHARD_IR_SHARDOPS_TD
diff --git a/mlir/include/mlir/Dialect/Mesh/Interfaces/CMakeLists.txt b/mlir/include/mlir/Dialect/Shard/Interfaces/CMakeLists.txt
index b3a44f3..b3a44f3 100644
--- a/mlir/include/mlir/Dialect/Mesh/Interfaces/CMakeLists.txt
+++ b/mlir/include/mlir/Dialect/Shard/Interfaces/CMakeLists.txt
diff --git a/mlir/include/mlir/Dialect/Mesh/Interfaces/ShardingInterface.h b/mlir/include/mlir/Dialect/Shard/Interfaces/ShardingInterface.h
index 14aad7f..46f6ed4 100644
--- a/mlir/include/mlir/Dialect/Mesh/Interfaces/ShardingInterface.h
+++ b/mlir/include/mlir/Dialect/Shard/Interfaces/ShardingInterface.h
@@ -6,10 +6,10 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef MLIR_DIALECT_MESH_INTERFACES_SHARDINGINTERFACE_H_
-#define MLIR_DIALECT_MESH_INTERFACES_SHARDINGINTERFACE_H_
+#ifndef MLIR_DIALECT_SHARD_INTERFACES_SHARDINGINTERFACE_H_
+#define MLIR_DIALECT_SHARD_INTERFACES_SHARDINGINTERFACE_H_
 
-#include "mlir/Dialect/Mesh/IR/MeshOps.h"
+#include "mlir/Dialect/Shard/IR/ShardOps.h"
 #include "mlir/Dialect/Utils/StructuredOpsUtils.h"
 #include "mlir/IR/Value.h"
 #include "mlir/Support/LLVM.h"
@@ -20,24 +20,24 @@ class Operation;
 class IRMapping;
 class SymbolTableCollection;
 
-namespace mesh {
+namespace shard {
 
-using ShardingArray = SmallVector<SmallVector<MeshAxis>>;
-using ShardingArrayRef = ArrayRef<SmallVector<MeshAxis>>;
+using ShardingArray = SmallVector<SmallVector<GridAxis>>;
+using ShardingArrayRef = ArrayRef<SmallVector<GridAxis>>;
 
 struct ShardingOption {
   // An array of int array. The sub-array at the i-th position signifies the
-  // mesh axes the i-th loop will be sharded on.
+  // grid axes the i-th loop will be sharded on.
   ShardingArray shardingArray = {};
-  FlatSymbolRefAttr mesh = nullptr;
+  FlatSymbolRefAttr grid = nullptr;
   // `empty` being true indicates that no sharding information can be inferred
   // at present. Note that it is different from the case where an operation is
   // not sharded.
   bool empty = false;
   ShardingOption() = default;
-  ShardingOption(ShardingArray shardingArray, FlatSymbolRefAttr mesh)
-      : shardingArray(std::move(shardingArray)), mesh(mesh) {
-    assert(this->mesh);
+  ShardingOption(ShardingArray shardingArray, FlatSymbolRefAttr grid)
+      : shardingArray(std::move(shardingArray)), grid(grid) {
+    assert(this->grid);
   }
   static ShardingOption makeEmpty() {
     auto res = ShardingOption();
@@ -46,21 +46,21 @@ struct ShardingOption {
   }
 };
 
-// This method retrieves the 'MeshSharding' from a given operation
+// This method retrieves the 'Sharding' from a given operation
 // result and includes the 'annotate_for_users' information.
-FailureOr<std::pair<bool, MeshSharding>> getMeshSharding(OpResult result);
+FailureOr<std::pair<bool, Sharding>> getSharding(OpResult result);
 
-// This method retrieves the 'MeshSharding' from a given operation
+// This method retrieves the 'Sharding' from a given operation
 // operand and includes the 'annotate_for_users' information.
-FailureOr<std::pair<bool, MeshSharding>> getMeshSharding(OpOperand &opOperand);
+FailureOr<std::pair<bool, Sharding>> getSharding(OpOperand &opOperand);
 
 namespace detail {
 
 FailureOr<ShardingOption>
-defaultGetShardingOption(Operation *op, ArrayRef<MeshSharding> operandShardings,
-                         ArrayRef<MeshSharding> resultShardings);
+defaultGetShardingOption(Operation *op, ArrayRef<Sharding> operandShardings,
+                         ArrayRef<Sharding> resultShardings);
 
-FailureOr<std::vector<MeshSharding>>
+FailureOr<std::vector<Sharding>>
 defaultGetShardingAnnotations(Operation *op,
                               const ShardingOption &shardingOption);
 
@@ -71,18 +71,18 @@ defaultAddShardingAnnotations(Operation *op, OpBuilder &b,
 } // namespace detail
 
 // Assumes full replication on all ranked tensor arguments and results.
-void spmdizeFullyReplicatedOperation(Operation &op,
-                                     ArrayRef<Value> spmdizedOperands,
-                                     ArrayRef<MeshSharding> operandShardings,
-                                     ArrayRef<MeshSharding> resultShardings,
-                                     IRMapping &spmdizationMap,
-                                     SymbolTableCollection &symbolTable,
-                                     OpBuilder &builder);
-
-} // namespace mesh
+void partitionFullyReplicatedOperation(Operation &op,
+                                       ArrayRef<Value> partitionedOperands,
+                                       ArrayRef<Sharding> operandShardings,
+                                       ArrayRef<Sharding> resultShardings,
+                                       IRMapping &partitionMap,
+                                       SymbolTableCollection &symbolTable,
+                                       OpBuilder &builder);
+
+} // namespace shard
 } // namespace mlir
 
 /// Include the ODS generated interface header files.
-#include "mlir/Dialect/Mesh/Interfaces/ShardingInterface.h.inc"
+#include "mlir/Dialect/Shard/Interfaces/ShardingInterface.h.inc"
 
-#endif // MLIR_DIALECT_MESH_INTERFACES_SHARDINGINTERFACE_H_
+#endif // MLIR_DIALECT_SHARD_INTERFACES_SHARDINGINTERFACE_H_
diff --git a/mlir/include/mlir/Dialect/Mesh/Interfaces/ShardingInterface.td b/mlir/include/mlir/Dialect/Shard/Interfaces/ShardingInterface.td
index a70d2c3..8f5332b 100644
--- a/mlir/include/mlir/Dialect/Mesh/Interfaces/ShardingInterface.td
+++ b/mlir/include/mlir/Dialect/Shard/Interfaces/ShardingInterface.td
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef MLIR_DIALECT_MESH_INTERFACES_SHARDINGINTERFACE_TD
-#define MLIR_DIALECT_MESH_INTERFACES_SHARDINGINTERFACE_TD
+#ifndef MLIR_DIALECT_SHARD_INTERFACES_SHARDINGINTERFACE_TD
+#define MLIR_DIALECT_SHARD_INTERFACES_SHARDINGINTERFACE_TD
 
 include "mlir/IR/OpBase.td"
 
@@ -16,7 +16,7 @@ def ShardingInterface : OpInterface<"ShardingInterface"> {
         Interface for allowing operations to expose information needed to
         shard them.
     }];
-    let cppNamespace = "::mlir::mesh";
+    let cppNamespace = "::mlir::shard";
 
     let methods = [
       InterfaceMethod<
@@ -84,8 +84,8 @@ def ShardingInterface : OpInterface<"ShardingInterface"> {
         /*retTy=*/"FailureOr<ShardingOption>",
         /*methodName=*/"getShardingOption",
         /*args=*/(ins
-          "ArrayRef<MeshSharding>": $operandShardings,
-          "ArrayRef<MeshSharding>": $resultShardings
+          "ArrayRef<Sharding>": $operandShardings,
+          "ArrayRef<Sharding>": $resultShardings
         ),
         /*methodBody=*/"",
         /*defaultImplementation=*/[{
@@ -100,7 +100,7 @@ def ShardingInterface : OpInterface<"ShardingInterface"> {
           This is what shardings the operands and results need to have in order
           to shard the op according to shardingOption.
         }],
-        /*retTy=*/"FailureOr<std::vector<MeshSharding>>",
+        /*retTy=*/"FailureOr<std::vector<Sharding>>",
         /*methodName=*/"getShardingAnnotations",
         /*args=*/(ins
           "const ShardingOption &":$shardingOption
@@ -113,7 +113,7 @@ def ShardingInterface : OpInterface<"ShardingInterface"> {
       >,
       InterfaceMethod<
         /*desc=*/[{
-          Based on a given ShardingOption, this method adds `mesh.shard`
+          Based on a given ShardingOption, this method adds `shard.shard`
           operations for the operands and results that previously lacked
           sharding annotations.
         }],
@@ -132,21 +132,21 @@ def ShardingInterface : OpInterface<"ShardingInterface"> {
       InterfaceMethod<
         /*desc=*/[{
           Convert self to SPMD form.
-          This method is used during the spmdization pass of a program fully
+          This method is used during the partition pass of a program fully
           annotated with shardings.
 
-          The spmdization algorithm would read the surrounding sharding
+          The partition algorithm would read the surrounding sharding
           annotations from the IR for each argument/result and prepare
           `operandShardings` and `resultShardings`.
           Values that are not ranked tensors do not have sharding annotations.
-          In this case their corresponding MeshSharding is null.
+          In this case their corresponding Sharding is null.
 
-          For convenience it will also prepare `spmdizedOperands`, although
-          they can be retrieved from the `spmdizationMap`.
+          For convenience it will also prepare `partitionedOperands`, although
+          they can be retrieved from the `partitionMap`.
 
-          The `spmdizationMap` contains a mapping from unsharded to
-          sharded/spmdized values that are constructed during the spmdization
-          pass. The interface implementation must populate `spmdizationMap`
+          The `partitionMap` contains a mapping from unsharded to
+          sharded/partitioned values that are constructed during the partition
+          pass. The interface implementation must populate `partitionMap`
           with the mapping for this op's results.
 
           `builder` is set to insert new operations in the appropriate point.
@@ -158,20 +158,20 @@ def ShardingInterface : OpInterface<"ShardingInterface"> {
           This assumes that all sharding annotations are for full replication.
         }],
         /*retTy=*/"LogicalResult",
-        /*methodName=*/"spmdize",
+        /*methodName=*/"partition",
         /*args=*/(ins
-          "ArrayRef<Value>": $spmdizedOperands,
-          "ArrayRef<MeshSharding>": $operandShardings,
-          "ArrayRef<MeshSharding>": $resultShardings,
-          "IRMapping&": $spmdizationMap,
+          "ArrayRef<Value>": $partitionedOperands,
+          "ArrayRef<Sharding>": $operandShardings,
+          "ArrayRef<Sharding>": $resultShardings,
+          "IRMapping&": $partitionMap,
           "SymbolTableCollection &": $symbolTableCollection,
           "OpBuilder &":$builder
         ),
         /*methodBody=*/"",
         /*defaultImplementation=*/[{
-          spmdizeFullyReplicatedOperation(
-            *$_op.getOperation(), spmdizedOperands, operandShardings,
-              resultShardings, spmdizationMap, symbolTableCollection, builder);
+          partitionFullyReplicatedOperation(
+            *$_op.getOperation(), partitionedOperands, operandShardings,
+              resultShardings, partitionMap, symbolTableCollection, builder);
           return success();
         }]>
     ];
@@ -184,4 +184,4 @@ def ShardingInterface : OpInterface<"ShardingInterface"> {
 }
 
 
-#endif // MLIR_DIALECT_MESH_INTERFACES_SHARDINGINTERFACE_TD
+#endif // MLIR_DIALECT_SHARD_INTERFACES_SHARDINGINTERFACE_TD
diff --git a/mlir/include/mlir/Dialect/Mesh/Interfaces/ShardingInterfaceImpl.h b/mlir/include/mlir/Dialect/Shard/Interfaces/ShardingInterfaceImpl.h
index 2af8b2b..d34ba79 100644
--- a/mlir/include/mlir/Dialect/Mesh/Interfaces/ShardingInterfaceImpl.h
+++ b/mlir/include/mlir/Dialect/Shard/Interfaces/ShardingInterfaceImpl.h
@@ -6,11 +6,11 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef MLIR_DIALECT_MESH_INTERFACES_SHARDINGINTERFACEIMPL_H_
-#define MLIR_DIALECT_MESH_INTERFACES_SHARDINGINTERFACEIMPL_H_
+#ifndef MLIR_DIALECT_SHARD_INTERFACES_SHARDINGINTERFACEIMPL_H_
+#define MLIR_DIALECT_SHARD_INTERFACES_SHARDINGINTERFACEIMPL_H_
 
-#include "mlir/Dialect/Mesh/IR/MeshOps.h"
-#include "mlir/Dialect/Mesh/Interfaces/ShardingInterface.h"
+#include "mlir/Dialect/Shard/IR/ShardOps.h"
+#include "mlir/Dialect/Shard/Interfaces/ShardingInterface.h"
 #include "mlir/IR/AffineMap.h"
 #include "mlir/IR/Value.h"
 
@@ -20,35 +20,34 @@ class Operation;
 class IRMapping;
 class SymbolTableCollection;
 
-namespace mesh {
+namespace shard {
 
-// Retrieve the mesh axes corresponding to each operation loop iterator based
+// Retrieve the grid axes corresponding to each operation loop iterator based
 // on the provided shardings for the op's operands and results.
 // Assumes that the indexingMaps are projected permutations.
-ShardingArray getMeshAxisAssignmentForLoopIterators(
-    ArrayRef<MeshSharding> operandShardings,
-    ArrayRef<MeshSharding> resultShardings,
+ShardingArray getGridAxisAssignmentForLoopIterators(
+    ArrayRef<Sharding> operandShardings, ArrayRef<Sharding> resultShardings,
     ArrayRef<utils::IteratorType> loopIteratorTypes,
     ArrayRef<AffineMap> indexingMaps);
 
 bool isAtLeastOneReductionIteratorSharded(
     ArrayRef<utils::IteratorType> loopIteratorTypes,
-    ArrayRef<SmallVector<MeshAxis>> meshAxisAssignmentForLoopIterators);
+    ArrayRef<SmallVector<GridAxis>> gridAxisAssignmentForLoopIterators);
 
-// Get the set of mesh axes that correspond to reduction loop iterators.
-SmallVector<MeshAxis> getReductionMeshAxes(
+// Get the set of grid axes that correspond to reduction loop iterators.
+SmallVector<GridAxis> getReductionGridAxes(
     ArrayRef<utils::IteratorType> loopIteratorTypes,
-    ArrayRef<SmallVector<MeshAxis>> meshAxisAssignmentForLoopIterators);
+    ArrayRef<SmallVector<GridAxis>> gridAxisAssignmentForLoopIterators);
 
 // Inserts a clone of the operation that has all ranked tensor
 // arguments/results sharded.
-void spmdizeTriviallyShardableOperation(Operation &op,
-                                        ArrayRef<Value> spmdizedOperands,
-                                        ArrayRef<MeshSharding> operandShardings,
-                                        ArrayRef<MeshSharding> resultShardings,
-                                        IRMapping &spmdizationMap,
-                                        SymbolTableCollection &symbolTable,
-                                        OpBuilder &builder);
+void partitionTriviallyShardableOperation(Operation &op,
+                                          ArrayRef<Value> partitionedOperands,
+                                          ArrayRef<Sharding> operandShardings,
+                                          ArrayRef<Sharding> resultShardings,
+                                          IRMapping &partitionMap,
+                                          SymbolTableCollection &symbolTable,
+                                          OpBuilder &builder);
 
 // All ranked tensor argument and result dimensions have
 // independent parallel loop iterators.
@@ -73,15 +72,15 @@ struct IndependentParallelIteratorDomainShardingInterface
     return SmallVector<AffineMap>();
   }
 
-  LogicalResult spmdize(Operation *op, ArrayRef<Value> spmdizedOperands,
-                        ArrayRef<MeshSharding> operandShardings,
-                        ArrayRef<MeshSharding> resultShardings,
-                        IRMapping &spmdizationMap,
-                        SymbolTableCollection &symbolTable,
-                        OpBuilder &builder) const {
-    spmdizeTriviallyShardableOperation(*op, spmdizedOperands, operandShardings,
-                                       resultShardings, spmdizationMap,
-                                       symbolTable, builder);
+  LogicalResult partition(Operation *op, ArrayRef<Value> partitionedOperands,
+                          ArrayRef<Sharding> operandShardings,
+                          ArrayRef<Sharding> resultShardings,
+                          IRMapping &partitionMap,
+                          SymbolTableCollection &symbolTable,
+                          OpBuilder &builder) const {
+    partitionTriviallyShardableOperation(*op, partitionedOperands,
+                                         operandShardings, resultShardings,
+                                         partitionMap, symbolTable, builder);
     return success();
   }
 
@@ -129,20 +128,20 @@ struct ElementwiseShardingInterface
     return maps;
   }
 
-  LogicalResult spmdize(Operation *op, ArrayRef<Value> spmdizedOperands,
-                        ArrayRef<MeshSharding> operandShardings,
-                        ArrayRef<MeshSharding> resultShardings,
-                        IRMapping &spmdizationMap,
-                        SymbolTableCollection &symbolTable,
-                        OpBuilder &builder) const {
-    spmdizeTriviallyShardableOperation(*op, spmdizedOperands, operandShardings,
-                                       resultShardings, spmdizationMap,
-                                       symbolTable, builder);
+  LogicalResult partition(Operation *op, ArrayRef<Value> partitionedOperands,
+                          ArrayRef<Sharding> operandShardings,
+                          ArrayRef<Sharding> resultShardings,
+                          IRMapping &partitionMap,
+                          SymbolTableCollection &symbolTable,
+                          OpBuilder &builder) const {
+    partitionTriviallyShardableOperation(*op, partitionedOperands,
+                                         operandShardings, resultShardings,
+                                         partitionMap, symbolTable, builder);
     return success();
   }
 };
 
-} // namespace mesh
+} // namespace shard
 } // namespace mlir
 
-#endif // MLIR_DIALECT_MESH_INTERFACES_SHARDINGINTERFACEIMPL_H_
+#endif // MLIR_DIALECT_SHARD_INTERFACES_SHARDINGINTERFACEIMPL_H_
diff --git a/mlir/include/mlir/Dialect/Shard/Transforms/CMakeLists.txt b/mlir/include/mlir/Dialect/Shard/Transforms/CMakeLists.txt
new file mode 100644
index 0000000..9e2c8d0
--- /dev/null
+++ b/mlir/include/mlir/Dialect/Shard/Transforms/CMakeLists.txt
@@ -0,0 +1,6 @@
+set(LLVM_TARGET_DEFINITIONS Passes.td)
+mlir_tablegen(Passes.h.inc -gen-pass-decls -name Shard)
+add_public_tablegen_target(MLIRShardPassIncGen)
+add_dependencies(mlir-headers MLIRShardPassIncGen)
+
+add_mlir_doc(Passes ShardPasses ./ -gen-pass-doc)
diff --git a/mlir/include/mlir/Dialect/Mesh/Transforms/Spmdization.h b/mlir/include/mlir/Dialect/Shard/Transforms/Partition.h
index 2f6de3e..3790376 100644
--- a/mlir/include/mlir/Dialect/Mesh/Transforms/Spmdization.h
+++ b/mlir/include/mlir/Dialect/Shard/Transforms/Partition.h
@@ -1,4 +1,4 @@
-//===- Simplifications.h - Mesh Simplifications -----------------*- C++ -*-===//
+//===- Simplifications.h - Shard Simplifications ----------------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -6,35 +6,35 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef MLIR_DIALECT_MESH_TRANSFORMS_SPMDIZATION_H
-#define MLIR_DIALECT_MESH_TRANSFORMS_SPMDIZATION_H
+#ifndef MLIR_DIALECT_SHARD_TRANSFORMS_PARTITION_H
+#define MLIR_DIALECT_SHARD_TRANSFORMS_PARTITION_H
 
-#include "mlir/Dialect/Mesh/IR/MeshOps.h"
+#include "mlir/Dialect/Shard/IR/ShardOps.h"
 #include "mlir/IR/DialectRegistry.h"
 
 namespace mlir {
-namespace mesh {
+namespace shard {
 
-// Insert resharding spmdization of the value `sourceShardValue`
+// Insert resharding partition of the value `sourceShardValue`
 // from sharding `source` to sharding `target`.
 // `sourceShardValue` is the already sharded value according to `source`.
 //
 // Example
 //
 // ```mlir
-//   mesh.mesh @mesh_1d(shape = 2)
+//   shard.grid @grid_1d(shape = 2)
 //   ...
-//   %1 = mesh.shard %0 to <@mesh_1d, [[0]]> : tensor<2xi8>
-//   %2 = mesh.shard %1 to <@mesh_1d, [[]]> annotate_for_users: tensor<2xi8>
+//   %1 = shard.shard %0 to <@grid_1d, [[0]]> : tensor<2xi8>
+//   %2 = shard.shard %1 to <@grid_1d, [[]]> annotate_for_users: tensor<2xi8>
 // ```
 //
 // Will result in
 //
 // ```mlir
-//   %1 = mesh.all_gather %0 on @mesh_1d mesh_axes = [0] gather_axis = 0 :
+//   %1 = shard.all_gather %0 on @grid_1d grid_axes = [0] gather_axis = 0 :
 //     tensor<1xi8> -> tensor<2xi8>
 // ```
-TypedValue<ShapedType> reshard(OpBuilder &builder, MeshOp mesh, ShardOp source,
+TypedValue<ShapedType> reshard(OpBuilder &builder, GridOp grid, ShardOp source,
                                ShardOp target,
                                TypedValue<ShapedType> sourceShardValue);
 TypedValue<ShapedType> reshard(OpBuilder &builder, ShardOp source,
@@ -44,7 +44,7 @@ TypedValue<ShapedType> reshard(OpBuilder &builder, ShardOp source,
 
 void reshardingRegisterDependentDialects(DialectRegistry &registry);
 
-} // namespace mesh
+} // namespace shard
 } // namespace mlir
 
-#endif // MLIR_DIALECT_MESH_TRANSFORMS_SPMDIZATION_H
+#endif // MLIR_DIALECT_SHARD_TRANSFORMS_PARTITION_H
diff --git a/mlir/include/mlir/Dialect/Mesh/Transforms/Passes.h b/mlir/include/mlir/Dialect/Shard/Transforms/Passes.h
index a2424d4..88bb460 100644
--- a/mlir/include/mlir/Dialect/Mesh/Transforms/Passes.h
+++ b/mlir/include/mlir/Dialect/Shard/Transforms/Passes.h
@@ -1,4 +1,4 @@
-//===- Passes.h - Mesh Passes -----------------------------------*- C++ -*-===//
+//===- Passes.h - Shard Passes ----------------------------------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -6,8 +6,8 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef MLIR_DIALECT_MESH_TRANSFORMS_PASSES_H
-#define MLIR_DIALECT_MESH_TRANSFORMS_PASSES_H
+#ifndef MLIR_DIALECT_SHARD_TRANSFORMS_PASSES_H
+#define MLIR_DIALECT_SHARD_TRANSFORMS_PASSES_H
 
 #include "mlir/Pass/Pass.h"
 
@@ -17,7 +17,7 @@ namespace func {
 class FuncOp;
 }
 
-namespace mesh {
+namespace shard {
 
 /// This enum controls the traversal order for the sharding propagation.
 enum class TraversalOrder {
@@ -36,16 +36,16 @@ enum class TraversalOrder {
 //===----------------------------------------------------------------------===//
 
 #define GEN_PASS_DECL
-#include "mlir/Dialect/Mesh/Transforms/Passes.h.inc"
+#include "mlir/Dialect/Shard/Transforms/Passes.h.inc"
 
 //===----------------------------------------------------------------------===//
 // Registration
 //===----------------------------------------------------------------------===//
 
 #define GEN_PASS_REGISTRATION
-#include "mlir/Dialect/Mesh/Transforms/Passes.h.inc"
+#include "mlir/Dialect/Shard/Transforms/Passes.h.inc"
 
-} // namespace mesh
+} // namespace shard
 } // namespace mlir
 
-#endif // MLIR_DIALECT_MESH_TRANSFORMS_PASSES_H
+#endif // MLIR_DIALECT_SHARD_TRANSFORMS_PASSES_H
diff --git a/mlir/include/mlir/Dialect/Mesh/Transforms/Passes.td b/mlir/include/mlir/Dialect/Shard/Transforms/Passes.td
index 11ec7e7..bbc6a19 100644
--- a/mlir/include/mlir/Dialect/Mesh/Transforms/Passes.td
+++ b/mlir/include/mlir/Dialect/Shard/Transforms/Passes.td
@@ -1,4 +1,4 @@
-//===-- Passes.td - Mesh transformation definition file ----*- tablegen -*-===//
+//===-- Passes.td - Shard transformation definition file ---*- tablegen -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -7,8 +7,8 @@
 //===----------------------------------------------------------------------===//
 
 
-#ifndef MLIR_DIALECT_MESH_TRANSFORMS_PASSES_TD
-#define MLIR_DIALECT_MESH_TRANSFORMS_PASSES_TD
+#ifndef MLIR_DIALECT_SHARD_TRANSFORMS_PASSES_TD
+#define MLIR_DIALECT_SHARD_TRANSFORMS_PASSES_TD
 
 include "mlir/Pass/PassBase.td"
 
@@ -20,31 +20,31 @@ def ShardingPropagation : InterfacePass<"sharding-propagation", "mlir::FunctionO
   let summary = "sharding propagation";
   let description = [{
     Propagates sharding information throughout the graph. After this pass, each
-    of the operations' operands and results is annotated with a `mesh.shard`
+    of the operations' operands and results is annotated with a `shard.shard`
     operation, and the operations themselves are added with sharding option
     attributes.
   }];
   let options = [
     Option<"traversal", "traversal",
-           "mlir::mesh::TraversalOrder", /*default=*/"mlir::mesh::TraversalOrder::BackwardForward",
+           "mlir::shard::TraversalOrder", /*default=*/"mlir::shard::TraversalOrder::BackwardForward",
            "Traversal order to use for sharding propagation:",
             [{::llvm::cl::values(
-              clEnumValN(mlir::mesh::TraversalOrder::Forward, "forward",
+              clEnumValN(mlir::shard::TraversalOrder::Forward, "forward",
               "Forward only traversal."),
-              clEnumValN(mlir::mesh::TraversalOrder::Backward, "backward",
+              clEnumValN(mlir::shard::TraversalOrder::Backward, "backward",
               "backward only traversal."),
-              clEnumValN(mlir::mesh::TraversalOrder::ForwardBackward, "forward-backward",
+              clEnumValN(mlir::shard::TraversalOrder::ForwardBackward, "forward-backward",
               "forward-backward traversal."),
-              clEnumValN(mlir::mesh::TraversalOrder::BackwardForward, "backward-forward",
+              clEnumValN(mlir::shard::TraversalOrder::BackwardForward, "backward-forward",
               "backward-forward traversal.")
             )}]>,
   ];
   let dependentDialects = [
-    "mesh::MeshDialect"
+    "shard::ShardDialect"
   ];
 }
 
-def Spmdization : InterfacePass<"mesh-spmdization", "mlir::FunctionOpInterface"> {
+def Partition : InterfacePass<"shard-partition", "mlir::FunctionOpInterface"> {
   let summary = "Partition a function into SPMD form.";
   let description = [{
     This pass fits in right after a pass that annotates the function with
@@ -52,15 +52,15 @@ def Spmdization : InterfacePass<"mesh-spmdization", "mlir::FunctionOpInterface">
     It operates on a fully annotated IR.
 
     A fully annotated IR required that all ranked tensor operands, results and
-    block arguments are annotated with the `mesh.shard` operation.
+    block arguments are annotated with the `shard.shard` operation.
   
     All direct descendant operations in the function must implement the
     `ShardingInterface` interface or all their ranked tensor operands and
     results must have full replication sharding.
 
     The input IR must have sharding annotations such that each operation
-    that implements `ShardingInterface` can handle during spmdization with
-    its `spmdize` method.
+    that implements `ShardingInterface` can handle during partition with
+    its `partition` method.
     This can be achieved with the `ShardingPropagation` pass.
 
     If the function has multiple terminating blocks,
@@ -70,36 +70,36 @@ def Spmdization : InterfacePass<"mesh-spmdization", "mlir::FunctionOpInterface">
 
     Example:
     ```mlir
-    mesh.mesh @mesh_1d(shape = 2)
+    shard.grid @grid_1d(shape = 2)
 
     func.func @f(
       %arg0: tensor<2xi8>
     ) -> tensor<2xi8> {
-      %0 = mesh.shard %arg0 to <@mesh_1d, [[0]]> : tensor<2xi8>
-      %1 = mesh.shard %0 to <@mesh_1d, [[0]]> annotate_for_users: tensor<2xi8>
+      %0 = shard.shard %arg0 to <@grid_1d, [[0]]> : tensor<2xi8>
+      %1 = shard.shard %0 to <@grid_1d, [[0]]> annotate_for_users: tensor<2xi8>
       %2 = tosa.abs %1 : (tensor<2xi8>) -> tensor<2xi8>
-      %3 = mesh.shard %2 to <@mesh_1d, [[0]]> : tensor<2xi8>
-      %4 = mesh.shard %3 to <@mesh_1d, [[]]> annotate_for_users: tensor<2xi8>
+      %3 = shard.shard %2 to <@grid_1d, [[0]]> : tensor<2xi8>
+      %4 = shard.shard %3 to <@grid_1d, [[]]> annotate_for_users: tensor<2xi8>
       return %4 : tensor<2xi8>
     }
     ```
-    Spmdizing the above would result in 
+    Partitioning the above would result in 
     * Performing the element-wise `abs` operation on each device.
     * Resharding to full replication with an all-gather.
 
     ```mlir
-    mesh.mesh @mesh_1d(shape = 2)
+    shard.grid @grid_1d(shape = 2)
   
     func.func @f(%arg0: tensor<1xi8>) -> tensor<2xi8> {
       %0 = tosa.abs %arg0 : (tensor<1xi8>) -> tensor<1xi8>
-      %1 = mesh.all_gather %0 on @mesh_1d mesh_axes = [0] gather_axis = 0 : tensor<1xi8> -> tensor<2xi8>
+      %1 = shard.all_gather %0 on @grid_1d grid_axes = [0] gather_axis = 0 : tensor<1xi8> -> tensor<2xi8>
       return %1 : tensor<2xi8>
     }
     ```
   }];
   let dependentDialects = [
-    "mesh::MeshDialect"
+    "shard::ShardDialect"
   ];
 }
 
-#endif // MLIR_DIALECT_MESH_TRANSFORMS_PASSES_TD
+#endif // MLIR_DIALECT_SHARD_TRANSFORMS_PASSES_TD
diff --git a/mlir/include/mlir/Dialect/Mesh/Transforms/ReshardingSpmdizationDoc.md b/mlir/include/mlir/Dialect/Shard/Transforms/ReshardingPartitionDoc.md
index 6368931..cf5ae12 100644
--- a/mlir/include/mlir/Dialect/Mesh/Transforms/ReshardingSpmdizationDoc.md
+++ b/mlir/include/mlir/Dialect/Shard/Transforms/ReshardingPartitionDoc.md
@@ -1,6 +1,6 @@
-# Resharding Spmdization Examples
+# Resharding Partition Examples
 
-Reshard `2x3` tensor from sharding `[[0, 1]]` to sharding `[[0, 1]]` on a `2x3` mesh.
+Reshard `2x3` tensor from sharding `[[0, 1]]` to sharding `[[0, 1]]` on a `2x3` shard.
 
 unsharded `2x3` tensor
 ```
@@ -8,16 +8,16 @@ unsharded `2x3` tensor
 21 22 23
 ```
 
-sharded on a `2x3` mesh
+sharded on a `2x3` grid
 
 sharding = `[[0, 1]]`
 
-mesh contents:
+grid contents:
 
 ```
-mesh axis 1
+grid axis 1
 ----------->
-+----+----+----+ mesh axis 0 |
++----+----+----+ grid axis 0 |
 | 11 | 12 | 13 |             |
 +----+----+----+             |
 | 21 | 22 | 23 |             |
@@ -27,9 +27,9 @@ mesh axis 1
 Transform into
 sharding = `[[1, 0]]`
 ```
-mesh axis 1
+grid axis 1
 ----------->
-+----+----+----+ mesh axis 0 |
++----+----+----+ grid axis 0 |
 | 11 | 13 | 22 |             |
 +----+----+----+             |
 | 12 | 21 | 23 |             |
@@ -40,7 +40,7 @@ Swap contents on devices that have the same linear index in the 2 shardings.
 
 --------------------------------------------------------------
 
-Reshard `2x3` tensor from sharding `[[0, 1]]` to sharding `[[1]]` on a `2x3` mesh.
+Reshard `2x3` tensor from sharding `[[0, 1]]` to sharding `[[1]]` on a `2x3` shard.
 
 unsharded `2x3` tensor
 ```
@@ -48,15 +48,15 @@ unsharded `2x3` tensor
 21 22 23
 ```
 
-sharded on a `2x3` mesh
+sharded on a `2x3` grid
 
 sharding = `[[0, 1]]`
 
-mesh contents:
+grid contents:
 ```
-mesh axis 1
+grid axis 1
 ----------->
-+----+----+----+ mesh axis 0 |
++----+----+----+ grid axis 0 |
 | 11 | 12 | 13 |             |
 +----+----+----+             |
 | 21 | 22 | 23 |             |
@@ -66,9 +66,9 @@ mesh axis 1
 Transform into
 sharding = `[[1]]`
 ```
-mesh axis 1
+grid axis 1
 ----------->
-+----+----+----+ mesh axis 0 |
++----+----+----+ grid axis 0 |
 | 11 | 12 | 13 |             |
 | 21 | 22 | 23 |             |
 +----+----+----+             |
@@ -77,11 +77,11 @@ mesh axis 1
 +----+----+----+             ↓
 ```
 Algorithm:
-All-gather along mesh axis 0.
+All-gather along grid axis 0.
 
 --------------------------------------------------------------
 
-Reshard `4x6` tensor from sharding `[[], [0, 1]]` to sharding `[[], [0]]` on a `2x3` mesh.
+Reshard `4x6` tensor from sharding `[[], [0, 1]]` to sharding `[[], [0]]` on a `2x3` shard.
 
 unsharded `4x6` tensor
 ```
@@ -89,15 +89,15 @@ unsharded `4x6` tensor
 21 22 23 24 25 26
 ```
 
-sharded on a `2x3` mesh
+sharded on a `2x3` grid
 
 sharding = `[[], [0, 1]]`
 
-mesh contents:
+grid contents:
 ```
-mesh axis 1
+grid axis 1
 ----------->
-+----+----+----+ mesh axis 0 |
++----+----+----+ grid axis 0 |
 | 11 | 12 | 13 |             |
 | 21 | 22 | 23 |             |
 +----+----+----+             |
@@ -108,9 +108,9 @@ mesh axis 1
 Transform into
 sharding = `[[], [0]]`
 ```
-mesh axis 1
+grid axis 1
 ----------->
-+----------+----------+ mesh axis 0 |
++----------+----------+ grid axis 0 |
 | 11 12 13 | 11 12 13 |             |
 | 21 22 23 | 21 22 23 |             |
 +----------+----------+             |
@@ -119,11 +119,11 @@ mesh axis 1
 +----------+----------+             ↓
 ```
 Algorithm:
-All-gather along mesh axis 1.
+All-gather along grid axis 1.
 
 --------------------------------------------------------------
 
-Reshard `4x8` tensor from sharding `[[0], [1, 2]]` to sharding `[[0], [2]]` on a `2x2x2` mesh.
+Reshard `4x8` tensor from sharding `[[0], [1, 2]]` to sharding `[[0], [2]]` on a `2x2x2` shard.
 
 unsharded `4x8` tensor
 ```
@@ -132,15 +132,15 @@ unsharded `4x8` tensor
 31 32 33 34 35 36 37 38
 41 42 43 44 45 46 47 48
 ```
-sharded on a `2x2x2` mesh
+sharded on a `2x2x2` grid
 
 sharding = `[[0], [1, 2]]`
 
-mesh contents:
+grid contents:
 ```
-mesh axis 2
+grid axis 2
 ----------->
-+-------+-------+ mesh axis 1 | mesh axis 0 |
++-------+-------+ grid axis 1 | grid axis 0 |
 | 11 12 | 13 14 |             |             |
 | 21 22 | 23 24 |             |             |
 +-------+-------+             |             |
@@ -158,9 +158,9 @@ mesh axis 2
 Transform into
 sharding = `[[0], [2]]`
 ```
-mesh axis 2
+grid axis 2
 ----------->
-+-------------+-------------+ mesh axis 1 | mesh axis 0 |
++-------------+-------------+ grid axis 1 | grid axis 0 |
 | 11 12 13 14 | 15 16 17 18 |             |             |
 | 21 22 23 24 | 25 26 27 28 |             |             |
 +-------------+-------------+             |             |
@@ -177,13 +177,13 @@ mesh axis 2
 ```
 Algorithm:
 
-Can't be done with just an all-gather along mesh axis 1.
+Can't be done with just an all-gather along grid axis 1.
 Can be handled by multiple resharding transformations
 `[[0], [1, 2]] -> [[0], [2, 1]] -> [[0], [2]]`
 
 --------------------------------------------------------------
 
-Reshard `6x6` tensor from sharding `[[0], [1]]` to sharding `[[1], [0]]` on a `2x3` mesh.
+Reshard `6x6` tensor from sharding `[[0], [1]]` to sharding `[[1], [0]]` on a `2x3` shard.
 
 unsharded `6x6` tensor
 ```
@@ -194,13 +194,13 @@ unsharded `6x6` tensor
 51 52 53 54 55 56
 61 62 63 64 65 66
 ```
-sharded on a `2x3` mesh
+sharded on a `2x3` grid
 
 sharding = `[[0], [1]]`
 ```
-mesh axis 1
+grid axis 1
 ----------->
-+-------+-------+-------+ mesh axis 0 |
++-------+-------+-------+ grid axis 0 |
 | 11 12 | 13 14 | 15 16 |             |
 | 21 22 | 23 24 | 25 26 |             |
 | 31 32 | 33 34 | 35 36 |             |
@@ -213,9 +213,9 @@ mesh axis 1
 transform to
 sharding = `[[1], [0]]`
 ```
-mesh axis 1
+grid axis 1
 ----------->
-+----------+----------+----------+ mesh axis 0 |
++----------+----------+----------+ grid axis 0 |
 | 11 12 13 | 31 32 33 | 51 52 53 |             |
 | 21 22 23 | 41 42 43 | 61 62 63 |             |
 +----------+----------+----------+             |
@@ -223,9 +223,9 @@ mesh axis 1
 | 24 25 26 | 44 45 46 | 64 65 66 |             |
 +----------+----------+----------+             ↓
 
-mesh axis 0
+grid axis 0
 ----------->
-+----------+----------+ mesh axis 1 |
++----------+----------+ grid axis 1 |
 | 11 12 13 | 14 15 16 |             |
 | 21 22 23 | 24 25 26 |             |
 +----------+----------+             |
@@ -240,7 +240,7 @@ Algorithm: TODO
 
 --------------------------------------------------------------
 
-Reshard `6x6` tensor from sharding `[[0], [1]]` to sharding `[[1], [0]]` on a `2x6` mesh.
+Reshard `6x6` tensor from sharding `[[0], [1]]` to sharding `[[1], [0]]` on a `2x6` shard.
 
 unsharded 6x6 tensor
 ```
@@ -251,13 +251,13 @@ unsharded 6x6 tensor
 51 52 53 54 55 56
 61 62 63 64 65 66
 ```
-shard on `2x6` mesh
+shard on `2x6` grid
 
 sharding = `[[0], [1]]`
 ```
-mesh axis 1
+grid axis 1
 ----------->
-+----+----+----+----+----+----+ mesh axis 0 |
++----+----+----+----+----+----+ grid axis 0 |
 | 11 | 12 | 13 ‖ 14 | 15 | 16 |             |
 | 21 | 22 | 23 ‖ 24 | 23 | 26 |             |
 | 31 | 32 | 33 ‖ 34 | 35 | 36 |             |
@@ -270,9 +270,9 @@ mesh axis 1
 transform to
 sharding = `[[1], [0]]`
 ```
-mesh axis 0
+grid axis 0
 ----------->
-+----------+----------+ mesh axis 1 |
++----------+----------+ grid axis 1 |
 | 11 12 13 | 14 15 16 |             |
 +----------+----------+             |
 | 21 22 23 | 24 25 26 |             |
@@ -290,9 +290,9 @@ Algorithm: TODO
 
 --------------------------------------------------------------
 
-Reshard KxL tensor from `[[0], [1]]` to `[[1], [0]]` on `MxN` mesh.
+Reshard KxL tensor from `[[0], [1]]` to `[[1], [0]]` on `MxN` shard.
 
-`M x N` mesh.
+`M x N` shard.
 `K x L` tensor `t`.
 `d(m, n)` the tensor on device `(m, n)`.
 
@@ -433,9 +433,9 @@ TODO
 
 --------------------------------------------------------------
 
-Reshard `KxL` tensor from sharding `[[0], [1]]` to sharding `[[1], [0]]` on a `2x3` mesh.
+Reshard `KxL` tensor from sharding `[[0], [1]]` to sharding `[[1], [0]]` on a `2x3` shard.
 
-Device placement on a `2x3` mesh
+Device placement on a `2x3` grid
 ```
 11 12 13  <- devices
 21 22 23
@@ -512,7 +512,7 @@ TODO
 
 --------------------------------------------------------------
 
-Reshard `6x6` tensor from sharding `[[0], []]` to sharding `[[], [0]]` on a `3` mesh.
+Reshard `6x6` tensor from sharding `[[0], []]` to sharding `[[], [0]]` on a `3` shard.
 
 unsharded `6x6` tensor
 ```
@@ -523,11 +523,11 @@ unsharded `6x6` tensor
 51 52 53 54 55 56
 61 62 63 64 65 66
 ```
-sharded on a `3` mesh
+sharded on a `3` grid
 
 sharding = `[[0], []]`
 ```
-+-------------------+ mesh axis 0 |
++-------------------+ grid axis 0 |
 | 11 12 13 14 15 16 |             |
 | 21 22 23 24 25 26 |             |
 +-------------------+             |
@@ -541,7 +541,7 @@ sharding = `[[0], []]`
 transform to
 sharding = `[[], [0]]`
 ```
-mesh axis 0
+grid axis 0
 ----------->
 +-------+-------+-------+
 | 11 12 | 13 14 | 15 16 |
@@ -554,11 +554,11 @@ mesh axis 0
 ```
 Algorithm:
 ```mlir
-%1 = all_to_all %0 on @mesh mesh_axes = [0] split_axis = 1 concat_axis = 0 : tensor<2x6xi8> -> tensor<6x2xi8>
+%1 = all_to_all %0 on @grid grid_axes = [0] split_axis = 1 concat_axis = 0 : tensor<2x6xi8> -> tensor<6x2xi8>
 ```
 --------------------------------------------------------------
 
-Reshard `4x4` tensor from sharding `[[0], [1, 2]]` to sharding `[[0, 1], [2]]` on a `2x2x2` mesh.
+Reshard `4x4` tensor from sharding `[[0], [1, 2]]` to sharding `[[0, 1], [2]]` on a `2x2x2` shard.
 
 unsharded `4x4` tensor
 ```
@@ -567,13 +567,13 @@ unsharded `4x4` tensor
 31 32 33 34
 41 42 43 44
 ```
-sharded on a `2x2x2` mesh
+sharded on a `2x2x2` grid
 
 sharding = `[[0], [1, 2]]`
 ```
-mesh axis 2
+grid axis 2
 ----------->
-+----+----+ mesh axis 1 | mesh axis 0 |
++----+----+ grid axis 1 | grid axis 0 |
 | 11 | 12 |             |             |
 | 21 | 22 |             |             |
 +----+----+             |             |
@@ -591,9 +591,9 @@ mesh axis 2
 transform to
 sharding = `[[0, 1], [2]]`
 ```
-mesh axis 2
+grid axis 2
 ----------->
-+-------+-------+ mesh axis 1 | mesh axis 0 |
++-------+-------+ grid axis 1 | grid axis 0 |
 | 11 12 | 13 41 |             |             |
 +-------+-------+             |             |
 | 21 22 | 23 24 |             |             |
@@ -606,7 +606,7 @@ mesh axis 2
 ```
 Algorithm:
 ```mlir
-%1 = all_to_all %0 on @mesh mesh_axes = [2] split_axis = 1 concat_axis = 0 : tensor<2x1xi8> -> tensor<1x2xi8>
+%1 = all_to_all %0 on @grid grid_axes = [2] split_axis = 1 concat_axis = 0 : tensor<2x1xi8> -> tensor<1x2xi8>
 ```
 is not enough.
 
@@ -639,15 +639,15 @@ Basis:
     [[0]] -> [[]]
     [[0, 1]] -> [[1]]
     ```
-    All-gather along mesh axis 0.
+    All-gather along grid axis 0.
 
-*   Swap mesh axes order when assigned to the same tensor axis.
+*   Swap grid axes order when assigned to the same tensor axis.
     ```
     [[0, 1]] -> [[1, 0]]
     ```
     Swap contents on devices with the same linear index.
 
-*   Move mesh axis to different tensor dimension.
+*   Move grid axis to different tensor dimension.
     ```
     [[0], []] -> [[], [0]]
     ```
@@ -661,9 +661,9 @@ Example decomposition of
 ```
 into
 ```
-[[0], [1]] -> all-gather along mesh axis 1    ->
-[[0], []]  -> all-to-all along mesh axis 0    ->
-[[], [0]]  -> extract slice along mesh axis 1 ->
+[[0], [1]] -> all-gather along grid axis 1    ->
+[[0], []]  -> all-to-all along grid axis 0    ->
+[[], [0]]  -> extract slice along grid axis 1 ->
 [[1], [0]]
 ```
 
@@ -675,9 +675,9 @@ Example decomposition of
 ```
 into
 ```
-[[3, 2], [], [0, 1]] -> all-to-all along mesh axis 1 ->
-[[3, 2], [1], [0]]   -> all-to-all along mesh axis 2 ->
-[[3], [1, 2], [0]]   -> all-gather along mesh axis 3 ->
-[[], [1, 2], [0]]    -> all-to-all along mesh axis 0 ->
+[[3, 2], [], [0, 1]] -> all-to-all along grid axis 1 ->
+[[3, 2], [1], [0]]   -> all-to-all along grid axis 2 ->
+[[3], [1, 2], [0]]   -> all-gather along grid axis 3 ->
+[[], [1, 2], [0]]    -> all-to-all along grid axis 0 ->
 [[0], [1, 2], []]
 ```
diff --git a/mlir/include/mlir/Dialect/Mesh/Transforms/Simplifications.h b/mlir/include/mlir/Dialect/Shard/Transforms/Simplifications.h
index 243dbf0..452d4f6 100644
--- a/mlir/include/mlir/Dialect/Mesh/Transforms/Simplifications.h
+++ b/mlir/include/mlir/Dialect/Shard/Transforms/Simplifications.h
@@ -1,4 +1,4 @@
-//===- Simplifications.h - Mesh Simplifications -----------------*- C++ -*-===//
+//===- Simplifications.h - Shard Simplifications ----------------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -6,10 +6,10 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef MLIR_DIALECT_MESH_TRANSFORMS_SIMPLIFICATIONS_H
-#define MLIR_DIALECT_MESH_TRANSFORMS_SIMPLIFICATIONS_H
+#ifndef MLIR_DIALECT_SHARD_TRANSFORMS_SIMPLIFICATIONS_H
+#define MLIR_DIALECT_SHARD_TRANSFORMS_SIMPLIFICATIONS_H
 
-#include "mlir/Dialect/Mesh/IR/MeshOps.h"
+#include "mlir/Dialect/Shard/IR/ShardOps.h"
 #include "mlir/IR/PatternMatch.h"
 #include "mlir/Transforms/EndomorphismSimplification.h"
 #include "llvm/Support/Casting.h"
@@ -22,7 +22,7 @@ namespace mlir {
 
 class SymbolTableCollection;
 
-namespace mesh {
+namespace shard {
 
 // If we have an algebraic op like "+" and a summing all-reduce,
 // `all_reduce_sum(x) + all_reduce_sum(y)` will be transformed to
@@ -112,7 +112,7 @@ void populateSimplificationPatterns(
 void populateFoldingPatterns(RewritePatternSet &patterns,
                              SymbolTableCollection &symbolTableCollection);
 
-} // namespace mesh
+} // namespace shard
 } // namespace mlir
 
-#endif // MLIR_DIALECT_MESH_TRANSFORMS_SIMPLIFICATIONS_H
+#endif // MLIR_DIALECT_SHARD_TRANSFORMS_SIMPLIFICATIONS_H
diff --git a/mlir/include/mlir/Dialect/Mesh/Transforms/Transforms.h b/mlir/include/mlir/Dialect/Shard/Transforms/Transforms.h
index f46c0db..57d65e6 100644
--- a/mlir/include/mlir/Dialect/Mesh/Transforms/Transforms.h
+++ b/mlir/include/mlir/Dialect/Shard/Transforms/Transforms.h
@@ -1,4 +1,4 @@
-//===- Transforms.h - Mesh Transforms ---------------------------*- C++ -*-===//
+//===- Transforms.h - Shard Transforms --------------------------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -6,10 +6,10 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef MLIR_DIALECT_MESH_TRANSFORMS_TRANSFORMS_H
-#define MLIR_DIALECT_MESH_TRANSFORMS_TRANSFORMS_H
+#ifndef MLIR_DIALECT_SHARD_TRANSFORMS_TRANSFORMS_H
+#define MLIR_DIALECT_SHARD_TRANSFORMS_TRANSFORMS_H
 
-#include "mlir/Dialect/Mesh/IR/MeshOps.h"
+#include "mlir/Dialect/Shard/IR/ShardOps.h"
 #include "mlir/IR/BuiltinTypes.h"
 #include "mlir/IR/Value.h"
 #include "mlir/Support/LLVM.h"
@@ -20,7 +20,7 @@ class RewritePatternSet;
 class SymbolTableCollection;
 class DialectRegistry;
 class ImplicitLocOpBuilder;
-namespace mesh {
+namespace shard {
 
 void populateProcessMultiIndexOpLoweringPatterns(
     RewritePatternSet &patterns, SymbolTableCollection &symbolTableCollection);
@@ -35,20 +35,20 @@ void populateAllOpLoweringPatterns(
 void registerAllOpLoweringDialects(DialectRegistry &registry);
 
 TypedValue<IndexType>
-createCollectiveProcessGroupSize(MeshOp mesh, ArrayRef<MeshAxis> axes,
+createCollectiveProcessGroupSize(GridOp grid, ArrayRef<GridAxis> axes,
                                  ImplicitLocOpBuilder &builder);
 
-// Get process linear index along the given mesh axes.
-TypedValue<IndexType> createProcessLinearIndex(StringRef mesh,
-                                               ArrayRef<MeshAxis> meshAxes,
+// Get process linear index along the given grid axes.
+TypedValue<IndexType> createProcessLinearIndex(StringRef grid,
+                                               ArrayRef<GridAxis> gridAxes,
                                                ImplicitLocOpBuilder &builder);
-// Get process linear index from a multi-index along the given mesh axes .
+// Get process linear index from a multi-index along the given grid axes .
 TypedValue<IndexType>
-createProcessLinearIndex(StringRef mesh, ValueRange processInGroupMultiIndex,
-                         ArrayRef<MeshAxis> meshAxes,
+createProcessLinearIndex(StringRef grid, ValueRange processInGroupMultiIndex,
+                         ArrayRef<GridAxis> gridAxes,
                          ImplicitLocOpBuilder &builder);
 
-} // namespace mesh
+} // namespace shard
 } // namespace mlir
 
-#endif // MLIR_DIALECT_MESH_TRANSFORMS_TRANSFORMS_H
+#endif // MLIR_DIALECT_SHARD_TRANSFORMS_TRANSFORMS_H
diff --git a/mlir/include/mlir/Dialect/Tensor/Extensions/MeshShardingExtensions.h b/mlir/include/mlir/Dialect/Tensor/Extensions/ShardingExtensions.h
index cfac485b..895e7e5 100644
--- a/mlir/include/mlir/Dialect/Tensor/Extensions/MeshShardingExtensions.h
+++ b/mlir/include/mlir/Dialect/Tensor/Extensions/ShardingExtensions.h
@@ -1,4 +1,4 @@
-//===- MeshShardingExtensions.h - -------------------------------*- C++ -*-===//
+//===- ShardingExtensions.h - -------------------------------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
diff --git a/mlir/include/mlir/Dialect/Tosa/IR/TosaTypesBase.td b/mlir/include/mlir/Dialect/Tosa/IR/TosaTypesBase.td
index 349e8ed..754640d 100644
--- a/mlir/include/mlir/Dialect/Tosa/IR/TosaTypesBase.td
+++ b/mlir/include/mlir/Dialect/Tosa/IR/TosaTypesBase.td
@@ -151,7 +151,7 @@ class Tosa_TensorOfOrNone<list<Type> allowedTypes, string description = ""> :
 def Tosa_Rank0Tensor : TosaTensorRankOf<[Tosa_AnyNumber], [0]>;
 
 def Tosa_ScalarTensor : AnyTypeOf<[TosaUnrankedTensorOf<[Tosa_Int, AnyFloat]>, TosaScalarTensorOf<[Tosa_AnyNumber], [1]>]>;
-def Tosa_ScalarInt8Tensor : TosaScalarTensorOf<[Tosa_Int8], [1]>;
+def Tosa_ScalarInt8Tensor : AnyTypeOf<[TosaUnrankedTensorOf<[Tosa_Int8]>, TosaScalarTensorOf<[Tosa_Int8], [1]>]>;
 def Tosa_ScalarIntOrFloatTensor : AnyTypeOf<[TosaUnrankedTensorOf<[Tosa_Int, AnyFloat]>, TosaScalarTensorOf<[Tosa_Int, AnyFloat], [1]>]>;
 
 // We include unranked tensors as a supported type for all possible tosa
diff --git a/mlir/include/mlir/IR/Operation.h b/mlir/include/mlir/IR/Operation.h
index fa8a487..edc8ab4 100644
--- a/mlir/include/mlir/IR/Operation.h
+++ b/mlir/include/mlir/IR/Operation.h
@@ -1102,6 +1102,29 @@ inline raw_ostream &operator<<(raw_ostream &os, const Operation &op) {
   return os;
 }
 
+/// A wrapper class that allows for printing an operation with a set of flags,
+/// useful to act as a "stream modifier" to customize printing an operation
+/// with a stream using the operator<< overload, e.g.:
+///   llvm::dbgs() << OpWithFlags(op, OpPrintingFlags().skipRegions());
+class OpWithFlags {
+public:
+  OpWithFlags(Operation *op, OpPrintingFlags flags = {})
+      : op(op), theFlags(flags) {}
+  OpPrintingFlags &flags() { return theFlags; }
+  const OpPrintingFlags &flags() const { return theFlags; }
+
+private:
+  Operation *op;
+  OpPrintingFlags theFlags;
+  friend raw_ostream &operator<<(raw_ostream &os, const OpWithFlags &op);
+};
+
+inline raw_ostream &operator<<(raw_ostream &os,
+                               const OpWithFlags &opWithFlags) {
+  opWithFlags.op->print(os, opWithFlags.flags());
+  return os;
+}
+
 } // namespace mlir
 
 namespace llvm {
diff --git a/mlir/include/mlir/IR/PatternMatch.h b/mlir/include/mlir/IR/PatternMatch.h
index b3608b4..b5a93a0 100644
--- a/mlir/include/mlir/IR/PatternMatch.h
+++ b/mlir/include/mlir/IR/PatternMatch.h
@@ -525,6 +525,11 @@ public:
   }
 
   /// This method erases an operation that is known to have no uses.
+  ///
+  /// If the current insertion point is before the erased operation, it is
+  /// adjusted to the following operation (or the end of the block). If the
+  /// current insertion point is within the erased operation, the insertion
+  /// point is left in an invalid state.
   virtual void eraseOp(Operation *op);
 
   /// This method erases all operations in a block.
@@ -539,6 +544,9 @@ public:
   /// somewhere in the middle (or beginning) of the dest block, the source block
   /// must have no successors. Otherwise, the resulting IR would have
   /// unreachable operations.
+  ///
+  /// If the insertion point is within the source block, it is adjusted to the
+  /// destination block.
   virtual void inlineBlockBefore(Block *source, Block *dest,
                                  Block::iterator before,
                                  ValueRange argValues = {});
@@ -549,6 +557,9 @@ public:
   ///
   /// The source block must have no successors. Otherwise, the resulting IR
   /// would have unreachable operations.
+  ///
+  /// If the insertion point is within the source block, it is adjusted to the
+  /// destination block.
   void inlineBlockBefore(Block *source, Operation *op,
                          ValueRange argValues = {});
 
@@ -558,6 +569,9 @@ public:
   ///
   /// The dest block must have no successors. Otherwise, the resulting IR would
   /// have unreachable operation.
+  ///
+  /// If the insertion point is within the source block, it is adjusted to the
+  /// destination block.
   void mergeBlocks(Block *source, Block *dest, ValueRange argValues = {});
 
   /// Split the operations starting at "before" (inclusive) out of the given
diff --git a/mlir/include/mlir/InitAllDialects.h b/mlir/include/mlir/InitAllDialects.h
index c6fcf1a..856170e 100644
--- a/mlir/include/mlir/InitAllDialects.h
+++ b/mlir/include/mlir/InitAllDialects.h
@@ -60,7 +60,6 @@
 #include "mlir/Dialect/MemRef/Transforms/AllocationOpInterfaceImpl.h"
 #include "mlir/Dialect/MemRef/Transforms/BufferViewFlowOpInterfaceImpl.h"
 #include "mlir/Dialect/MemRef/Transforms/RuntimeOpVerification.h"
-#include "mlir/Dialect/Mesh/IR/MeshDialect.h"
 #include "mlir/Dialect/NVGPU/IR/NVGPUDialect.h"
 #include "mlir/Dialect/OpenACC/OpenACC.h"
 #include "mlir/Dialect/OpenMP/OpenMPDialect.h"
@@ -77,6 +76,7 @@
 #include "mlir/Dialect/SPIRV/IR/SPIRVDialect.h"
 #include "mlir/Dialect/Shape/IR/Shape.h"
 #include "mlir/Dialect/Shape/Transforms/BufferizableOpInterfaceImpl.h"
+#include "mlir/Dialect/Shard/IR/ShardDialect.h"
 #include "mlir/Dialect/SparseTensor/IR/SparseTensor.h"
 #include "mlir/Dialect/SparseTensor/Transforms/BufferizableOpInterfaceImpl.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
@@ -131,7 +131,7 @@ inline void registerAllDialects(DialectRegistry &registry) {
                   LLVM::LLVMDialect,
                   math::MathDialect,
                   memref::MemRefDialect,
-                  mesh::MeshDialect,
+                  shard::ShardDialect,
                   ml_program::MLProgramDialect,
                   mpi::MPIDialect,
                   nvgpu::NVGPUDialect,
diff --git a/mlir/include/mlir/InitAllPasses.h b/mlir/include/mlir/InitAllPasses.h
index dd8b292..002ff61 100644
--- a/mlir/include/mlir/InitAllPasses.h
+++ b/mlir/include/mlir/InitAllPasses.h
@@ -32,13 +32,13 @@
 #include "mlir/Dialect/MLProgram/Transforms/Passes.h"
 #include "mlir/Dialect/Math/Transforms/Passes.h"
 #include "mlir/Dialect/MemRef/Transforms/Passes.h"
-#include "mlir/Dialect/Mesh/Transforms/Passes.h"
 #include "mlir/Dialect/NVGPU/Transforms/Passes.h"
 #include "mlir/Dialect/OpenACC/Transforms/Passes.h"
 #include "mlir/Dialect/Quant/Transforms/Passes.h"
 #include "mlir/Dialect/SCF/Transforms/Passes.h"
 #include "mlir/Dialect/SPIRV/Transforms/Passes.h"
 #include "mlir/Dialect/Shape/Transforms/Passes.h"
+#include "mlir/Dialect/Shard/Transforms/Passes.h"
 #include "mlir/Dialect/SparseTensor/Pipelines/Passes.h"
 #include "mlir/Dialect/SparseTensor/Transforms/Passes.h"
 #include "mlir/Dialect/Tensor/Transforms/Passes.h"
@@ -81,7 +81,7 @@ inline void registerAllPasses() {
   LLVM::registerLLVMPasses();
   math::registerMathPasses();
   memref::registerMemRefPasses();
-  mesh::registerMeshPasses();
+  shard::registerShardPasses();
   ml_program::registerMLProgramPasses();
   quant::registerQuantPasses();
   registerSCFPasses();
diff --git a/mlir/include/mlir/Target/LLVMIR/ModuleImport.h b/mlir/include/mlir/Target/LLVMIR/ModuleImport.h
index c484072..17ef8e4 100644
--- a/mlir/include/mlir/Target/LLVMIR/ModuleImport.h
+++ b/mlir/include/mlir/Target/LLVMIR/ModuleImport.h
@@ -19,6 +19,7 @@
 #include "mlir/Target/LLVMIR/Import.h"
 #include "mlir/Target/LLVMIR/LLVMImportInterface.h"
 #include "mlir/Target/LLVMIR/TypeFromLLVM.h"
+#include "llvm/IR/Module.h"
 
 namespace llvm {
 class BasicBlock;
diff --git a/mlir/lib/Analysis/DataFlow/ConstantPropagationAnalysis.cpp b/mlir/lib/Analysis/DataFlow/ConstantPropagationAnalysis.cpp
index 51fa773..fb5649e 100644
--- a/mlir/lib/Analysis/DataFlow/ConstantPropagationAnalysis.cpp
+++ b/mlir/lib/Analysis/DataFlow/ConstantPropagationAnalysis.cpp
@@ -16,6 +16,7 @@
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/Support/Casting.h"
 #include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 #include <cassert>
 
 #define DEBUG_TYPE "constant-propagation"
@@ -46,7 +47,7 @@ void ConstantValue::print(raw_ostream &os) const {
 LogicalResult SparseConstantPropagation::visitOperation(
     Operation *op, ArrayRef<const Lattice<ConstantValue> *> operands,
     ArrayRef<Lattice<ConstantValue> *> results) {
-  LLVM_DEBUG(llvm::dbgs() << "SCP: Visiting operation: " << *op << "\n");
+  LDBG() << "SCP: Visiting operation: " << *op;
 
   // Don't try to simulate the results of a region operation as we can't
   // guarantee that folding will be out-of-place. We don't allow in-place
@@ -98,12 +99,11 @@ LogicalResult SparseConstantPropagation::visitOperation(
     // Merge in the result of the fold, either a constant or a value.
     OpFoldResult foldResult = std::get<1>(it);
     if (Attribute attr = llvm::dyn_cast_if_present<Attribute>(foldResult)) {
-      LLVM_DEBUG(llvm::dbgs() << "Folded to constant: " << attr << "\n");
+      LDBG() << "Folded to constant: " << attr;
       propagateIfChanged(lattice,
                          lattice->join(ConstantValue(attr, op->getDialect())));
     } else {
-      LLVM_DEBUG(llvm::dbgs()
-                 << "Folded to value: " << cast<Value>(foldResult) << "\n");
+      LDBG() << "Folded to value: " << cast<Value>(foldResult);
       AbstractSparseForwardDataFlowAnalysis::join(
           lattice, *getLatticeElement(cast<Value>(foldResult)));
     }
diff --git a/mlir/lib/Analysis/DataFlow/DeadCodeAnalysis.cpp b/mlir/lib/Analysis/DataFlow/DeadCodeAnalysis.cpp
index 1abdfcb..10874fd 100644
--- a/mlir/lib/Analysis/DataFlow/DeadCodeAnalysis.cpp
+++ b/mlir/lib/Analysis/DataFlow/DeadCodeAnalysis.cpp
@@ -23,12 +23,11 @@
 #include "mlir/Support/LLVM.h"
 #include "llvm/Support/Casting.h"
 #include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 #include <cassert>
 #include <optional>
 
 #define DEBUG_TYPE "dead-code-analysis"
-#define DBGS() (llvm::dbgs() << '[' << DEBUG_TYPE << "] ")
-#define LDBG(X) LLVM_DEBUG(DBGS() << X << "\n")
 
 using namespace mlir;
 using namespace mlir::dataflow;
@@ -127,7 +126,8 @@ DeadCodeAnalysis::DeadCodeAnalysis(DataFlowSolver &solver)
 }
 
 LogicalResult DeadCodeAnalysis::initialize(Operation *top) {
-  LDBG("Initializing DeadCodeAnalysis for top-level op: " << top->getName());
+  LDBG() << "Initializing DeadCodeAnalysis for top-level op: "
+         << top->getName();
   // Mark the top-level blocks as executable.
   for (Region &region : top->getRegions()) {
     if (region.empty())
@@ -135,7 +135,7 @@ LogicalResult DeadCodeAnalysis::initialize(Operation *top) {
     auto *state =
         getOrCreate<Executable>(getProgramPointBefore(&region.front()));
     propagateIfChanged(state, state->setToLive());
-    LDBG("Marked entry block live for region in op: " << top->getName());
+    LDBG() << "Marked entry block live for region in op: " << top->getName();
   }
 
   // Mark as overdefined the predecessors of symbol callables with potentially
@@ -146,18 +146,18 @@ LogicalResult DeadCodeAnalysis::initialize(Operation *top) {
 }
 
 void DeadCodeAnalysis::initializeSymbolCallables(Operation *top) {
-  LDBG("[init] Entering initializeSymbolCallables for top-level op: "
-       << top->getName());
+  LDBG() << "[init] Entering initializeSymbolCallables for top-level op: "
+         << top->getName();
   analysisScope = top;
   auto walkFn = [&](Operation *symTable, bool allUsesVisible) {
-    LDBG("[init] Processing symbol table op: " << symTable->getName());
+    LDBG() << "[init] Processing symbol table op: " << symTable->getName();
     Region &symbolTableRegion = symTable->getRegion(0);
     Block *symbolTableBlock = &symbolTableRegion.front();
 
     bool foundSymbolCallable = false;
     for (auto callable : symbolTableBlock->getOps<CallableOpInterface>()) {
-      LDBG("[init] Found CallableOpInterface: "
-           << callable.getOperation()->getName());
+      LDBG() << "[init] Found CallableOpInterface: "
+             << callable.getOperation()->getName();
       Region *callableRegion = callable.getCallableRegion();
       if (!callableRegion)
         continue;
@@ -171,8 +171,8 @@ void DeadCodeAnalysis::initializeSymbolCallables(Operation *top) {
         auto *state =
             getOrCreate<PredecessorState>(getProgramPointAfter(callable));
         propagateIfChanged(state, state->setHasUnknownPredecessors());
-        LDBG("[init] Marked callable as having unknown predecessors: "
-             << callable.getOperation()->getName());
+        LDBG() << "[init] Marked callable as having unknown predecessors: "
+               << callable.getOperation()->getName();
       }
       foundSymbolCallable = true;
     }
@@ -187,15 +187,15 @@ void DeadCodeAnalysis::initializeSymbolCallables(Operation *top) {
     if (!uses) {
       // If we couldn't gather the symbol uses, conservatively assume that
       // we can't track information for any nested symbols.
-      LDBG("[init] Could not gather symbol uses, conservatively marking "
-           "all nested callables as having unknown predecessors");
+      LDBG() << "[init] Could not gather symbol uses, conservatively marking "
+                "all nested callables as having unknown predecessors";
       return top->walk([&](CallableOpInterface callable) {
         auto *state =
             getOrCreate<PredecessorState>(getProgramPointAfter(callable));
         propagateIfChanged(state, state->setHasUnknownPredecessors());
-        LDBG("[init] Marked nested callable as "
-             "having unknown predecessors: "
-             << callable.getOperation()->getName());
+        LDBG() << "[init] Marked nested callable as "
+                  "having unknown predecessors: "
+               << callable.getOperation()->getName();
       });
     }
 
@@ -209,15 +209,15 @@ void DeadCodeAnalysis::initializeSymbolCallables(Operation *top) {
         continue;
       auto *state = getOrCreate<PredecessorState>(getProgramPointAfter(symbol));
       propagateIfChanged(state, state->setHasUnknownPredecessors());
-      LDBG("[init] Found non-call use for symbol, "
-           "marked as having unknown predecessors: "
-           << symbol->getName());
+      LDBG() << "[init] Found non-call use for symbol, "
+                "marked as having unknown predecessors: "
+             << symbol->getName();
     }
   };
   SymbolTable::walkSymbolTables(top, /*allSymUsesVisible=*/!top->getBlock(),
                                 walkFn);
-  LDBG("[init] Finished initializeSymbolCallables for top-level op: "
-       << top->getName());
+  LDBG() << "[init] Finished initializeSymbolCallables for top-level op: "
+         << top->getName();
 }
 
 /// Returns true if the operation is a returning terminator in region
@@ -229,14 +229,14 @@ static bool isRegionOrCallableReturn(Operation *op) {
 }
 
 LogicalResult DeadCodeAnalysis::initializeRecursively(Operation *op) {
-  LDBG("[init] Entering initializeRecursively for op: " << op->getName()
-                                                        << " at " << op);
+  LDBG() << "[init] Entering initializeRecursively for op: " << op->getName()
+         << " at " << op;
   // Initialize the analysis by visiting every op with control-flow semantics.
   if (op->getNumRegions() || op->getNumSuccessors() ||
       isRegionOrCallableReturn(op) || isa<CallOpInterface>(op)) {
-    LDBG("[init] Visiting op with control-flow semantics: " << *op);
-    // When the liveness of the parent block changes, make sure to re-invoke the
-    // analysis on the op.
+    LDBG() << "[init] Visiting op with control-flow semantics: " << *op;
+    // When the liveness of the parent block changes, make sure to
+    // re-invoke the analysis on the op.
     if (op->getBlock())
       getOrCreate<Executable>(getProgramPointBefore(op->getBlock()))
           ->blockContentSubscribe(this);
@@ -246,21 +246,21 @@ LogicalResult DeadCodeAnalysis::initializeRecursively(Operation *op) {
   }
   // Recurse on nested operations.
   for (Region &region : op->getRegions()) {
-    LDBG("[init] Recursing into region of op: " << op->getName());
+    LDBG() << "[init] Recursing into region of op: " << op->getName();
     for (Operation &nestedOp : region.getOps()) {
-      LDBG("[init] Recursing into nested op: " << nestedOp.getName() << " at "
-                                               << &nestedOp);
+      LDBG() << "[init] Recursing into nested op: " << nestedOp.getName()
+             << " at " << &nestedOp;
       if (failed(initializeRecursively(&nestedOp)))
         return failure();
     }
   }
-  LDBG("[init] Finished initializeRecursively for op: " << op->getName()
-                                                        << " at " << op);
+  LDBG() << "[init] Finished initializeRecursively for op: " << op->getName()
+         << " at " << op;
   return success();
 }
 
 void DeadCodeAnalysis::markEdgeLive(Block *from, Block *to) {
-  LDBG("Marking edge live from block " << from << " to block " << to);
+  LDBG() << "Marking edge live from block " << from << " to block " << to;
   auto *state = getOrCreate<Executable>(getProgramPointBefore(to));
   propagateIfChanged(state, state->setToLive());
   auto *edgeState =
@@ -269,35 +269,35 @@ void DeadCodeAnalysis::markEdgeLive(Block *from, Block *to) {
 }
 
 void DeadCodeAnalysis::markEntryBlocksLive(Operation *op) {
-  LDBG("Marking entry blocks live for op: " << op->getName());
+  LDBG() << "Marking entry blocks live for op: " << op->getName();
   for (Region &region : op->getRegions()) {
     if (region.empty())
       continue;
     auto *state =
         getOrCreate<Executable>(getProgramPointBefore(&region.front()));
     propagateIfChanged(state, state->setToLive());
-    LDBG("Marked entry block live for region in op: " << op->getName());
+    LDBG() << "Marked entry block live for region in op: " << op->getName();
   }
 }
 
 LogicalResult DeadCodeAnalysis::visit(ProgramPoint *point) {
-  LDBG("Visiting program point: " << point << " " << *point);
+  LDBG() << "Visiting program point: " << point << " " << *point;
   if (point->isBlockStart())
     return success();
   Operation *op = point->getPrevOp();
-  LDBG("Visiting operation: " << *op);
+  LDBG() << "Visiting operation: " << *op;
 
   // If the parent block is not executable, there is nothing to do.
   if (op->getBlock() != nullptr &&
       !getOrCreate<Executable>(getProgramPointBefore(op->getBlock()))
            ->isLive()) {
-    LDBG("Parent block not live, skipping op: " << *op);
+    LDBG() << "Parent block not live, skipping op: " << *op;
     return success();
   }
 
   // We have a live call op. Add this as a live predecessor of the callee.
   if (auto call = dyn_cast<CallOpInterface>(op)) {
-    LDBG("Visiting call operation: " << *op);
+    LDBG() << "Visiting call operation: " << *op;
     visitCallOperation(call);
   }
 
@@ -305,12 +305,12 @@ LogicalResult DeadCodeAnalysis::visit(ProgramPoint *point) {
   if (op->getNumRegions()) {
     // Check if we can reason about the region control-flow.
     if (auto branch = dyn_cast<RegionBranchOpInterface>(op)) {
-      LDBG("Visiting region branch operation: " << *op);
+      LDBG() << "Visiting region branch operation: " << *op;
       visitRegionBranchOperation(branch);
 
       // Check if this is a callable operation.
     } else if (auto callable = dyn_cast<CallableOpInterface>(op)) {
-      LDBG("Visiting callable operation: " << *op);
+      LDBG() << "Visiting callable operation: " << *op;
       const auto *callsites = getOrCreateFor<PredecessorState>(
           getProgramPointAfter(op), getProgramPointAfter(callable));
 
@@ -322,19 +322,19 @@ LogicalResult DeadCodeAnalysis::visit(ProgramPoint *point) {
 
       // Otherwise, conservatively mark all entry blocks as executable.
     } else {
-      LDBG("Marking all entry blocks live for op: " << *op);
+      LDBG() << "Marking all entry blocks live for op: " << *op;
       markEntryBlocksLive(op);
     }
   }
 
   if (isRegionOrCallableReturn(op)) {
     if (auto branch = dyn_cast<RegionBranchOpInterface>(op->getParentOp())) {
-      LDBG("Visiting region terminator: " << *op);
+      LDBG() << "Visiting region terminator: " << *op;
       // Visit the exiting terminator of a region.
       visitRegionTerminator(op, branch);
     } else if (auto callable =
                    dyn_cast<CallableOpInterface>(op->getParentOp())) {
-      LDBG("Visiting callable terminator: " << *op);
+      LDBG() << "Visiting callable terminator: " << *op;
       // Visit the exiting terminator of a callable.
       visitCallableTerminator(op, callable);
     }
@@ -343,12 +343,12 @@ LogicalResult DeadCodeAnalysis::visit(ProgramPoint *point) {
   if (op->getNumSuccessors()) {
     // Check if we can reason about the control-flow.
     if (auto branch = dyn_cast<BranchOpInterface>(op)) {
-      LDBG("Visiting branch operation: " << *op);
+      LDBG() << "Visiting branch operation: " << *op;
       visitBranchOperation(branch);
 
       // Otherwise, conservatively mark all successors as exectuable.
     } else {
-      LDBG("Marking all successors live for op: " << *op);
+      LDBG() << "Marking all successors live for op: " << *op;
       for (Block *successor : op->getSuccessors())
         markEdgeLive(op->getBlock(), successor);
     }
@@ -358,7 +358,7 @@ LogicalResult DeadCodeAnalysis::visit(ProgramPoint *point) {
 }
 
 void DeadCodeAnalysis::visitCallOperation(CallOpInterface call) {
-  LDBG("visitCallOperation: " << call.getOperation()->getName());
+  LDBG() << "visitCallOperation: " << call.getOperation()->getName();
   Operation *callableOp = call.resolveCallableInTable(&symbolTable);
 
   // A call to a externally-defined callable has unknown predecessors.
@@ -381,15 +381,15 @@ void DeadCodeAnalysis::visitCallOperation(CallOpInterface call) {
     auto *callsites =
         getOrCreate<PredecessorState>(getProgramPointAfter(callableOp));
     propagateIfChanged(callsites, callsites->join(call));
-    LDBG("Added callsite as predecessor for callable: "
-         << callableOp->getName());
+    LDBG() << "Added callsite as predecessor for callable: "
+           << callableOp->getName();
   } else {
     // Mark this call op's predecessors as overdefined.
     auto *predecessors =
         getOrCreate<PredecessorState>(getProgramPointAfter(call));
     propagateIfChanged(predecessors, predecessors->setHasUnknownPredecessors());
-    LDBG("Marked call op's predecessors as unknown for: "
-         << call.getOperation()->getName());
+    LDBG() << "Marked call op's predecessors as unknown for: "
+           << call.getOperation()->getName();
   }
 }
 
@@ -421,7 +421,7 @@ DeadCodeAnalysis::getOperandValues(Operation *op) {
 }
 
 void DeadCodeAnalysis::visitBranchOperation(BranchOpInterface branch) {
-  LDBG("visitBranchOperation: " << branch.getOperation()->getName());
+  LDBG() << "visitBranchOperation: " << branch.getOperation()->getName();
   // Try to deduce a single successor for the branch.
   std::optional<SmallVector<Attribute>> operands = getOperandValues(branch);
   if (!operands)
@@ -429,18 +429,18 @@ void DeadCodeAnalysis::visitBranchOperation(BranchOpInterface branch) {
 
   if (Block *successor = branch.getSuccessorForOperands(*operands)) {
     markEdgeLive(branch->getBlock(), successor);
-    LDBG("Branch has single successor: " << successor);
+    LDBG() << "Branch has single successor: " << successor;
   } else {
     // Otherwise, mark all successors as executable and outgoing edges.
     for (Block *successor : branch->getSuccessors())
       markEdgeLive(branch->getBlock(), successor);
-    LDBG("Branch has multiple/all successors live");
+    LDBG() << "Branch has multiple/all successors live";
   }
 }
 
 void DeadCodeAnalysis::visitRegionBranchOperation(
     RegionBranchOpInterface branch) {
-  LDBG("visitRegionBranchOperation: " << branch.getOperation()->getName());
+  LDBG() << "visitRegionBranchOperation: " << branch.getOperation()->getName();
   // Try to deduce which regions are executable.
   std::optional<SmallVector<Attribute>> operands = getOperandValues(branch);
   if (!operands)
@@ -457,19 +457,19 @@ void DeadCodeAnalysis::visitRegionBranchOperation(
     // Mark the entry block as executable.
     auto *state = getOrCreate<Executable>(point);
     propagateIfChanged(state, state->setToLive());
-    LDBG("Marked region successor live: " << point);
+    LDBG() << "Marked region successor live: " << point;
     // Add the parent op as a predecessor.
     auto *predecessors = getOrCreate<PredecessorState>(point);
     propagateIfChanged(
         predecessors,
         predecessors->join(branch, successor.getSuccessorInputs()));
-    LDBG("Added region branch as predecessor for successor: " << point);
+    LDBG() << "Added region branch as predecessor for successor: " << point;
   }
 }
 
 void DeadCodeAnalysis::visitRegionTerminator(Operation *op,
                                              RegionBranchOpInterface branch) {
-  LDBG("visitRegionTerminator: " << *op);
+  LDBG() << "visitRegionTerminator: " << *op;
   std::optional<SmallVector<Attribute>> operands = getOperandValues(op);
   if (!operands)
     return;
@@ -488,7 +488,7 @@ void DeadCodeAnalysis::visitRegionTerminator(Operation *op,
       auto *state =
           getOrCreate<Executable>(getProgramPointBefore(&region->front()));
       propagateIfChanged(state, state->setToLive());
-      LDBG("Marked region entry block live for region: " << region);
+      LDBG() << "Marked region entry block live for region: " << region;
       predecessors = getOrCreate<PredecessorState>(
           getProgramPointBefore(&region->front()));
     } else {
@@ -498,14 +498,14 @@ void DeadCodeAnalysis::visitRegionTerminator(Operation *op,
     }
     propagateIfChanged(predecessors,
                        predecessors->join(op, successor.getSuccessorInputs()));
-    LDBG("Added region terminator as predecessor for successor: "
-         << (successor.getSuccessor() ? "region entry" : "parent op"));
+    LDBG() << "Added region terminator as predecessor for successor: "
+           << (successor.getSuccessor() ? "region entry" : "parent op");
   }
 }
 
 void DeadCodeAnalysis::visitCallableTerminator(Operation *op,
                                                CallableOpInterface callable) {
-  LDBG("visitCallableTerminator: " << *op);
+  LDBG() << "visitCallableTerminator: " << *op;
   // Add as predecessors to all callsites this return op.
   auto *callsites = getOrCreateFor<PredecessorState>(
       getProgramPointAfter(op), getProgramPointAfter(callable));
@@ -516,15 +516,15 @@ void DeadCodeAnalysis::visitCallableTerminator(Operation *op,
         getOrCreate<PredecessorState>(getProgramPointAfter(predecessor));
     if (canResolve) {
       propagateIfChanged(predecessors, predecessors->join(op));
-      LDBG("Added callable terminator as predecessor for callsite: "
-           << predecessor->getName());
+      LDBG() << "Added callable terminator as predecessor for callsite: "
+             << predecessor->getName();
     } else {
       // If the terminator is not a return-like, then conservatively assume we
       // can't resolve the predecessor.
       propagateIfChanged(predecessors,
                          predecessors->setHasUnknownPredecessors());
-      LDBG("Could not resolve callable terminator for callsite: "
-           << predecessor->getName());
+      LDBG() << "Could not resolve callable terminator for callsite: "
+             << predecessor->getName();
     }
   }
 }
diff --git a/mlir/lib/Analysis/DataFlow/LivenessAnalysis.cpp b/mlir/lib/Analysis/DataFlow/LivenessAnalysis.cpp
index 6a12fe3..509f520 100644
--- a/mlir/lib/Analysis/DataFlow/LivenessAnalysis.cpp
+++ b/mlir/lib/Analysis/DataFlow/LivenessAnalysis.cpp
@@ -10,7 +10,7 @@
 #include <cassert>
 #include <mlir/Analysis/DataFlow/LivenessAnalysis.h>
 
-#include <llvm/Support/Debug.h>
+#include <llvm/Support/DebugLog.h>
 #include <mlir/Analysis/DataFlow/SparseAnalysis.h>
 #include <mlir/Analysis/DataFlow/Utils.h>
 #include <mlir/Analysis/DataFlowFramework.h>
@@ -21,8 +21,6 @@
 #include <mlir/Support/LLVM.h>
 
 #define DEBUG_TYPE "liveness-analysis"
-#define DBGS() (llvm::dbgs() << '[' << DEBUG_TYPE << "] ")
-#define LDBG(X) LLVM_DEBUG(DBGS() << X << "\n")
 
 using namespace mlir;
 using namespace mlir::dataflow;
@@ -81,16 +79,15 @@ ChangeResult Liveness::meet(const AbstractSparseLattice &other) {
 LogicalResult
 LivenessAnalysis::visitOperation(Operation *op, ArrayRef<Liveness *> operands,
                                  ArrayRef<const Liveness *> results) {
-  LLVM_DEBUG(DBGS() << "[visitOperation] Enter: ";
-             op->print(llvm::dbgs(), OpPrintingFlags().skipRegions());
-             llvm::dbgs() << "\n");
+  LDBG() << "[visitOperation] Enter: "
+         << OpWithFlags(op, OpPrintingFlags().skipRegions());
   // This marks values of type (1.a) and (4) liveness as "live".
   if (!isMemoryEffectFree(op) || op->hasTrait<OpTrait::ReturnLike>()) {
-    LDBG("[visitOperation] Operation has memory effects or is "
-         "return-like, marking operands live");
+    LDBG() << "[visitOperation] Operation has memory effects or is "
+              "return-like, marking operands live";
     for (auto *operand : operands) {
-      LDBG(" [visitOperation] Marking operand live: "
-           << operand << " (" << operand->isLive << ")");
+      LDBG() << " [visitOperation] Marking operand live: " << operand << " ("
+             << operand->isLive << ")";
       propagateIfChanged(operand, operand->markLive());
     }
   }
@@ -99,28 +96,28 @@ LivenessAnalysis::visitOperation(Operation *op, ArrayRef<Liveness *> operands,
   bool foundLiveResult = false;
   for (const Liveness *r : results) {
     if (r->isLive && !foundLiveResult) {
-      LDBG("[visitOperation] Found live result, "
-           "meeting all operands with result: "
-           << r);
+      LDBG() << "[visitOperation] Found live result, "
+                "meeting all operands with result: "
+             << r;
       // It is assumed that each operand is used to compute each result of an
       // op. Thus, if at least one result is live, each operand is live.
       for (Liveness *operand : operands) {
-        LDBG(" [visitOperation] Meeting operand: " << operand
-                                                   << " with result: " << r);
+        LDBG() << " [visitOperation] Meeting operand: " << operand
+               << " with result: " << r;
         meet(operand, *r);
       }
       foundLiveResult = true;
     }
-    LDBG("[visitOperation] Adding dependency for result: " << r << " after op: "
-                                                           << *op);
+    LDBG() << "[visitOperation] Adding dependency for result: " << r
+           << " after op: " << *op;
     addDependency(const_cast<Liveness *>(r), getProgramPointAfter(op));
   }
   return success();
 }
 
 void LivenessAnalysis::visitBranchOperand(OpOperand &operand) {
-  LDBG("Visiting branch operand: " << operand.get()
-                                   << " in op: " << *operand.getOwner());
+  LDBG() << "Visiting branch operand: " << operand.get()
+         << " in op: " << *operand.getOwner();
   // We know (at the moment) and assume (for the future) that `operand` is a
   // non-forwarded branch operand of a `RegionBranchOpInterface`,
   // `BranchOpInterface`, `RegionBranchTerminatorOpInterface` or return-like op.
@@ -152,9 +149,9 @@ void LivenessAnalysis::visitBranchOperand(OpOperand &operand) {
       for (Value result : op->getResults()) {
         if (getLatticeElement(result)->isLive) {
           mayLive = true;
-          LDBG("[visitBranchOperand] Non-forwarded branch "
-               "operand may be live due to live result: "
-               << result);
+          LDBG() << "[visitBranchOperand] Non-forwarded branch "
+                    "operand may be live due to live result: "
+                 << result;
           break;
         }
       }
@@ -174,8 +171,8 @@ void LivenessAnalysis::visitBranchOperand(OpOperand &operand) {
     // Therefore, we conservatively consider the non-forwarded operand of the
     // branch operation may live.
     mayLive = true;
-    LDBG("[visitBranchOperand] Non-forwarded branch operand may "
-         "be live due to branch op interface");
+    LDBG() << "[visitBranchOperand] Non-forwarded branch operand may "
+              "be live due to branch op interface";
   } else {
     Operation *parentOp = op->getParentOp();
     assert(isa<RegionBranchOpInterface>(parentOp) &&
@@ -191,9 +188,9 @@ void LivenessAnalysis::visitBranchOperand(OpOperand &operand) {
       for (Value result : parentOp->getResults()) {
         if (getLatticeElement(result)->isLive) {
           mayLive = true;
-          LDBG("[visitBranchOperand] Non-forwarded branch "
-               "operand may be live due to parent live result: "
-               << result);
+          LDBG() << "[visitBranchOperand] Non-forwarded branch "
+                    "operand may be live due to parent live result: "
+                 << result;
           break;
         }
       }
@@ -214,9 +211,9 @@ void LivenessAnalysis::visitBranchOperand(OpOperand &operand) {
     for (Operation &nestedOp : *block) {
       if (!isMemoryEffectFree(&nestedOp)) {
         mayLive = true;
-        LDBG("Non-forwarded branch operand may be "
-             "live due to memory effect in block: "
-             << block);
+        LDBG() << "Non-forwarded branch operand may be "
+                  "live due to memory effect in block: "
+               << block;
         break;
       }
     }
@@ -224,7 +221,7 @@ void LivenessAnalysis::visitBranchOperand(OpOperand &operand) {
 
   if (mayLive) {
     Liveness *operandLiveness = getLatticeElement(operand.get());
-    LDBG("Marking branch operand live: " << operand.get());
+    LDBG() << "Marking branch operand live: " << operand.get();
     propagateIfChanged(operandLiveness, operandLiveness->markLive());
   }
 
@@ -236,7 +233,7 @@ void LivenessAnalysis::visitBranchOperand(OpOperand &operand) {
   SmallVector<const Liveness *, 4> resultsLiveness;
   for (const Value result : op->getResults())
     resultsLiveness.push_back(getLatticeElement(result));
-  LDBG("Visiting operation for non-forwarded branch operand: " << *op);
+  LDBG() << "Visiting operation for non-forwarded branch operand: " << *op;
   (void)visitOperation(op, operandLiveness, resultsLiveness);
 
   // We also visit the parent op with the parent's results and this operand if
@@ -249,14 +246,14 @@ void LivenessAnalysis::visitBranchOperand(OpOperand &operand) {
   SmallVector<const Liveness *, 4> parentResultsLiveness;
   for (const Value parentResult : parentOp->getResults())
     parentResultsLiveness.push_back(getLatticeElement(parentResult));
-  LDBG("Visiting parent operation for non-forwarded branch operand: "
-       << *parentOp);
+  LDBG() << "Visiting parent operation for non-forwarded branch operand: "
+         << *parentOp;
   (void)visitOperation(parentOp, operandLiveness, parentResultsLiveness);
 }
 
 void LivenessAnalysis::visitCallOperand(OpOperand &operand) {
-  LDBG("Visiting call operand: " << operand.get()
-                                 << " in op: " << *operand.getOwner());
+  LDBG() << "Visiting call operand: " << operand.get()
+         << " in op: " << *operand.getOwner();
   // We know (at the moment) and assume (for the future) that `operand` is a
   // non-forwarded call operand of an op implementing `CallOpInterface`.
   assert(isa<CallOpInterface>(operand.getOwner()) &&
@@ -269,18 +266,18 @@ void LivenessAnalysis::visitCallOperand(OpOperand &operand) {
   // This marks values of type (1.c) liveness as "live". A non-forwarded
   // call operand is live.
   Liveness *operandLiveness = getLatticeElement(operand.get());
-  LDBG("Marking call operand live: " << operand.get());
+  LDBG() << "Marking call operand live: " << operand.get();
   propagateIfChanged(operandLiveness, operandLiveness->markLive());
 }
 
 void LivenessAnalysis::setToExitState(Liveness *lattice) {
-  LDBG("setToExitState for lattice: " << lattice);
+  LDBG() << "setToExitState for lattice: " << lattice;
   if (lattice->isLive) {
-    LDBG("Lattice already live, nothing to do");
+    LDBG() << "Lattice already live, nothing to do";
     return;
   }
   // This marks values of type (2) liveness as "live".
-  LDBG("Marking lattice live due to exit state");
+  LDBG() << "Marking lattice live due to exit state";
   (void)lattice->markLive();
   propagateIfChanged(lattice, ChangeResult::Change);
 }
@@ -290,14 +287,14 @@ void LivenessAnalysis::setToExitState(Liveness *lattice) {
 //===----------------------------------------------------------------------===//
 
 RunLivenessAnalysis::RunLivenessAnalysis(Operation *op) {
-  LDBG("Constructing RunLivenessAnalysis for op: " << op->getName());
+  LDBG() << "Constructing RunLivenessAnalysis for op: " << op->getName();
   SymbolTableCollection symbolTable;
 
   loadBaselineAnalyses(solver);
   solver.load<LivenessAnalysis>(symbolTable);
-  LDBG("Initializing and running solver");
+  LDBG() << "Initializing and running solver";
   (void)solver.initializeAndRun(op);
-  LDBG("Dumping liveness state for op");
+  LDBG() << "RunLivenessAnalysis initialized for op: " << op->getName();
 }
 
 const Liveness *RunLivenessAnalysis::getLiveness(Value val) {
diff --git a/mlir/lib/Analysis/DataFlowFramework.cpp b/mlir/lib/Analysis/DataFlowFramework.cpp
index 176d53e..16f7033 100644
--- a/mlir/lib/Analysis/DataFlowFramework.cpp
+++ b/mlir/lib/Analysis/DataFlowFramework.cpp
@@ -14,7 +14,7 @@
 #include "llvm/ADT/iterator.h"
 #include "llvm/Config/abi-breaking.h"
 #include "llvm/Support/Casting.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 #include "llvm/Support/raw_ostream.h"
 
 #define DEBUG_TYPE "dataflow"
@@ -44,9 +44,8 @@ void AnalysisState::addDependency(ProgramPoint *dependent,
   (void)inserted;
   DATAFLOW_DEBUG({
     if (inserted) {
-      llvm::dbgs() << "Creating dependency between " << debugName << " of "
-                   << anchor << "\nand " << debugName << " on " << dependent
-                   << "\n";
+      LDBG() << "Creating dependency between " << debugName << " of " << anchor
+             << "\nand " << debugName << " on " << dependent;
     }
   });
 }
@@ -116,8 +115,7 @@ LogicalResult DataFlowSolver::initializeAndRun(Operation *top) {
 
   // Initialize the analyses.
   for (DataFlowAnalysis &analysis : llvm::make_pointee_range(childAnalyses)) {
-    DATAFLOW_DEBUG(llvm::dbgs()
-                   << "Priming analysis: " << analysis.debugName << "\n");
+    DATAFLOW_DEBUG(LDBG() << "Priming analysis: " << analysis.debugName);
     if (failed(analysis.initialize(top)))
       return failure();
   }
@@ -129,8 +127,8 @@ LogicalResult DataFlowSolver::initializeAndRun(Operation *top) {
     auto [point, analysis] = worklist.front();
     worklist.pop();
 
-    DATAFLOW_DEBUG(llvm::dbgs() << "Invoking '" << analysis->debugName
-                                << "' on: " << point << "\n");
+    DATAFLOW_DEBUG(LDBG() << "Invoking '" << analysis->debugName
+                          << "' on: " << point);
     if (failed(analysis->visit(point)))
       return failure();
   }
@@ -143,9 +141,9 @@ void DataFlowSolver::propagateIfChanged(AnalysisState *state,
   assert(isRunning &&
          "DataFlowSolver is not running, should not use propagateIfChanged");
   if (changed == ChangeResult::Change) {
-    DATAFLOW_DEBUG(llvm::dbgs() << "Propagating update to " << state->debugName
-                                << " of " << state->anchor << "\n"
-                                << "Value: " << *state << "\n");
+    DATAFLOW_DEBUG(LDBG() << "Propagating update to " << state->debugName
+                          << " of " << state->anchor << "\n"
+                          << "Value: " << *state);
     state->onUpdate(this);
   }
 }
diff --git a/mlir/lib/Bindings/Python/IRAttributes.cpp b/mlir/lib/Bindings/Python/IRAttributes.cpp
index 8f79caf..db84ee1 100644
--- a/mlir/lib/Bindings/Python/IRAttributes.cpp
+++ b/mlir/lib/Bindings/Python/IRAttributes.cpp
@@ -16,8 +16,8 @@
 #include "NanobindUtils.h"
 #include "mlir-c/BuiltinAttributes.h"
 #include "mlir-c/BuiltinTypes.h"
-#include "mlir/Bindings/Python/NanobindAdaptors.h"
 #include "mlir/Bindings/Python/Nanobind.h"
+#include "mlir/Bindings/Python/NanobindAdaptors.h"
 #include "llvm/ADT/ScopeExit.h"
 #include "llvm/Support/raw_ostream.h"
 
@@ -1428,6 +1428,12 @@ public:
   }
 };
 
+// Check if the python version is less than 3.13. Py_IsFinalizing is a part
+// of stable ABI since 3.13 and before it was available as _Py_IsFinalizing.
+#if PY_VERSION_HEX < 0x030d0000
+#define Py_IsFinalizing _Py_IsFinalizing
+#endif
+
 class PyDenseResourceElementsAttribute
     : public PyConcreteAttribute<PyDenseResourceElementsAttribute> {
 public:
@@ -1474,8 +1480,9 @@ public:
     // The userData is a Py_buffer* that the deleter owns.
     auto deleter = [](void *userData, const void *data, size_t size,
                       size_t align) {
-      if (!Py_IsInitialized())
-        Py_Initialize();
+      if (Py_IsFinalizing())
+        return;
+      assert(Py_IsInitialized() && "expected interpreter to be initialized");
       Py_buffer *ownedView = static_cast<Py_buffer *>(userData);
       nb::gil_scoped_acquire gil;
       PyBuffer_Release(ownedView);
diff --git a/mlir/lib/Conversion/AMDGPUToROCDL/AMDGPUToROCDL.cpp b/mlir/lib/Conversion/AMDGPUToROCDL/AMDGPUToROCDL.cpp
index bc0d9bf..64720bf 100644
--- a/mlir/lib/Conversion/AMDGPUToROCDL/AMDGPUToROCDL.cpp
+++ b/mlir/lib/Conversion/AMDGPUToROCDL/AMDGPUToROCDL.cpp
@@ -232,8 +232,8 @@ struct FatRawBufferCastLowering
     Value result = MemRefDescriptor::poison(
         rewriter, loc,
         getTypeConverter()->convertType(op.getResult().getType()));
-    result = LLVM::InsertValueOp::create(rewriter, loc, result, fatPtr,
-                                         kAllocatedPtrPosInMemRefDescriptor);
+    SmallVector<int64_t> pos{kAllocatedPtrPosInMemRefDescriptor};
+    result = LLVM::InsertValueOp::create(rewriter, loc, result, fatPtr, pos);
     result = LLVM::InsertValueOp::create(rewriter, loc, result, fatPtr,
                                          kAlignedPtrPosInMemRefDescriptor);
     result = LLVM::InsertValueOp::create(rewriter, loc, result, offset,
@@ -481,16 +481,16 @@ struct MemoryCounterWaitOpLowering
     if (chipset.majorVersion >= 12) {
       Location loc = op.getLoc();
       if (std::optional<int> ds = adaptor.getDs())
-        rewriter.create<ROCDL::WaitDscntOp>(loc, *ds);
+        ROCDL::WaitDscntOp::create(rewriter, loc, *ds);
 
       if (std::optional<int> load = adaptor.getLoad())
-        rewriter.create<ROCDL::WaitLoadcntOp>(loc, *load);
+        ROCDL::WaitLoadcntOp::create(rewriter, loc, *load);
 
       if (std::optional<int> store = adaptor.getStore())
-        rewriter.create<ROCDL::WaitStorecntOp>(loc, *store);
+        ROCDL::WaitStorecntOp::create(rewriter, loc, *store);
 
       if (std::optional<int> exp = adaptor.getExp())
-        rewriter.create<ROCDL::WaitExpcntOp>(loc, *exp);
+        ROCDL::WaitExpcntOp::create(rewriter, loc, *exp);
 
       rewriter.eraseOp(op);
       return success();
diff --git a/mlir/lib/Conversion/ArithToEmitC/ArithToEmitC.cpp b/mlir/lib/Conversion/ArithToEmitC/ArithToEmitC.cpp
index 59b3fe2..515fe5c 100644
--- a/mlir/lib/Conversion/ArithToEmitC/ArithToEmitC.cpp
+++ b/mlir/lib/Conversion/ArithToEmitC/ArithToEmitC.cpp
@@ -402,8 +402,8 @@ public:
     Value actualOp = adaptValueType(adaptor.getIn(), rewriter, castSrcType);
 
     // Actual cast (may change bitwidth)
-    auto cast = rewriter.template create<emitc::CastOp>(op.getLoc(),
-                                                        castDestType, actualOp);
+    auto cast =
+        emitc::CastOp::create(rewriter, op.getLoc(), castDestType, actualOp);
 
     // Cast to the expected output type
     auto result = adaptValueType(cast, rewriter, opReturnType);
@@ -507,8 +507,8 @@ public:
     Value lhs = adaptValueType(adaptor.getLhs(), rewriter, arithmeticType);
     Value rhs = adaptValueType(adaptor.getRhs(), rewriter, arithmeticType);
 
-    Value arithmeticResult = rewriter.template create<EmitCOp>(
-        op.getLoc(), arithmeticType, lhs, rhs);
+    Value arithmeticResult =
+        EmitCOp::create(rewriter, op.getLoc(), arithmeticType, lhs, rhs);
 
     Value result = adaptValueType(arithmeticResult, rewriter, type);
 
@@ -547,8 +547,8 @@ public:
     Value lhs = adaptValueType(adaptor.getLhs(), rewriter, arithmeticType);
     Value rhs = adaptValueType(adaptor.getRhs(), rewriter, arithmeticType);
 
-    Value arithmeticResult = rewriter.template create<EmitCOp>(
-        op.getLoc(), arithmeticType, lhs, rhs);
+    Value arithmeticResult =
+        EmitCOp::create(rewriter, op.getLoc(), arithmeticType, lhs, rhs);
 
     Value result = adaptValueType(arithmeticResult, rewriter, type);
 
@@ -748,8 +748,8 @@ public:
     }
     Value fpCastOperand = adaptor.getIn();
     if (actualOperandType != operandType) {
-      fpCastOperand = rewriter.template create<emitc::CastOp>(
-          castOp.getLoc(), actualOperandType, fpCastOperand);
+      fpCastOperand = emitc::CastOp::create(rewriter, castOp.getLoc(),
+                                            actualOperandType, fpCastOperand);
     }
     rewriter.replaceOpWithNewOp<emitc::CastOp>(castOp, dstType, fpCastOperand);
 
diff --git a/mlir/lib/Conversion/BufferizationToMemRef/BufferizationToMemRef.cpp b/mlir/lib/Conversion/BufferizationToMemRef/BufferizationToMemRef.cpp
index 30a7170..3edcbb8 100644
--- a/mlir/lib/Conversion/BufferizationToMemRef/BufferizationToMemRef.cpp
+++ b/mlir/lib/Conversion/BufferizationToMemRef/BufferizationToMemRef.cpp
@@ -68,9 +68,8 @@ struct CloneOpConversion : public OpConversionPattern<bufferization::CloneOp> {
 
         scf::YieldOp::create(rewriter, loc, acc);
       };
-      auto size = rewriter
-                      .create<scf::ForOp>(loc, zero, rank, one, ValueRange(one),
-                                          loopBody)
+      auto size = scf::ForOp::create(rewriter, loc, zero, rank, one,
+                                     ValueRange(one), loopBody)
                       .getResult(0);
 
       MemRefType memrefType = MemRefType::get({ShapedType::kDynamic},
diff --git a/mlir/lib/Conversion/CMakeLists.txt b/mlir/lib/Conversion/CMakeLists.txt
index f84375b..785cb82 100644
--- a/mlir/lib/Conversion/CMakeLists.txt
+++ b/mlir/lib/Conversion/CMakeLists.txt
@@ -43,7 +43,7 @@ add_subdirectory(MathToSPIRV)
 add_subdirectory(MemRefToEmitC)
 add_subdirectory(MemRefToLLVM)
 add_subdirectory(MemRefToSPIRV)
-add_subdirectory(MeshToMPI)
+add_subdirectory(ShardToMPI)
 add_subdirectory(MPIToLLVM)
 add_subdirectory(NVGPUToNVVM)
 add_subdirectory(NVVMToLLVM)
diff --git a/mlir/lib/Conversion/ControlFlowToSCF/ControlFlowToSCF.cpp b/mlir/lib/Conversion/ControlFlowToSCF/ControlFlowToSCF.cpp
index c8311eb..5ac838c 100644
--- a/mlir/lib/Conversion/ControlFlowToSCF/ControlFlowToSCF.cpp
+++ b/mlir/lib/Conversion/ControlFlowToSCF/ControlFlowToSCF.cpp
@@ -144,12 +144,11 @@ ControlFlowToSCFTransformation::createUnreachableTerminator(Location loc,
     return emitError(loc, "Cannot create unreachable terminator for '")
            << parentOp->getName() << "'";
 
-  return builder
-      .create<func::ReturnOp>(
-          loc, llvm::map_to_vector(funcOp.getResultTypes(),
-                                   [&](Type type) {
-                                     return getUndefValue(loc, builder, type);
-                                   }))
+  return func::ReturnOp::create(
+             builder, loc,
+             llvm::map_to_vector(
+                 funcOp.getResultTypes(),
+                 [&](Type type) { return getUndefValue(loc, builder, type); }))
       .getOperation();
 }
 
diff --git a/mlir/lib/Conversion/GPUCommon/GPUToLLVMConversion.cpp b/mlir/lib/Conversion/GPUCommon/GPUToLLVMConversion.cpp
index 63eb6c58..3cfbd89 100644
--- a/mlir/lib/Conversion/GPUCommon/GPUToLLVMConversion.cpp
+++ b/mlir/lib/Conversion/GPUCommon/GPUToLLVMConversion.cpp
@@ -579,8 +579,8 @@ LLVM::CallOp FunctionCallBuilder::create(Location loc, OpBuilder &builder,
   auto function = [&] {
     if (auto function = module.lookupSymbol<LLVM::LLVMFuncOp>(functionName))
       return function;
-    return OpBuilder::atBlockEnd(module.getBody())
-        .create<LLVM::LLVMFuncOp>(loc, functionName, functionType);
+    auto builder = OpBuilder::atBlockEnd(module.getBody());
+    return LLVM::LLVMFuncOp::create(builder, loc, functionName, functionType);
   }();
   return LLVM::CallOp::create(builder, loc, function, arguments);
 }
diff --git a/mlir/lib/Conversion/GPUToSPIRV/GPUToSPIRV.cpp b/mlir/lib/Conversion/GPUToSPIRV/GPUToSPIRV.cpp
index a19194e..75e6563 100644
--- a/mlir/lib/Conversion/GPUToSPIRV/GPUToSPIRV.cpp
+++ b/mlir/lib/Conversion/GPUToSPIRV/GPUToSPIRV.cpp
@@ -559,8 +559,8 @@ static Value createGroupReduceOpImpl(OpBuilder &builder, Location loc,
         builder, loc, builder.getI32Type(),
         builder.getIntegerAttr(builder.getI32Type(), *clusterSize));
 
-  return builder
-      .create<NonUniformOp>(loc, type, scope, groupOp, arg, clusterSizeValue)
+  return NonUniformOp::create(builder, loc, type, scope, groupOp, arg,
+                              clusterSizeValue)
       .getResult();
 }
 
diff --git a/mlir/lib/Conversion/LLVMCommon/Pattern.cpp b/mlir/lib/Conversion/LLVMCommon/Pattern.cpp
index ecd5b63..2568044 100644
--- a/mlir/lib/Conversion/LLVMCommon/Pattern.cpp
+++ b/mlir/lib/Conversion/LLVMCommon/Pattern.cpp
@@ -272,14 +272,13 @@ LogicalResult ConvertToLLVMPattern::copyUnrankedDescriptors(
 
     // Allocate memory, copy, and free the source if necessary.
     Value memory =
-        toDynamic
-            ? builder
-                  .create<LLVM::CallOp>(loc, mallocFunc.value(), allocationSize)
-                  .getResult()
-            : LLVM::AllocaOp::create(builder, loc, getPtrType(),
-                                     IntegerType::get(getContext(), 8),
-                                     allocationSize,
-                                     /*alignment=*/0);
+        toDynamic ? LLVM::CallOp::create(builder, loc, mallocFunc.value(),
+                                         allocationSize)
+                        .getResult()
+                  : LLVM::AllocaOp::create(builder, loc, getPtrType(),
+                                           IntegerType::get(getContext(), 8),
+                                           allocationSize,
+                                           /*alignment=*/0);
     Value source = desc.memRefDescPtr(builder, loc);
     LLVM::MemcpyOp::create(builder, loc, memory, source, allocationSize, false);
     if (!toDynamic)
diff --git a/mlir/lib/Conversion/MPIToLLVM/MPIToLLVM.cpp b/mlir/lib/Conversion/MPIToLLVM/MPIToLLVM.cpp
index 5b68eb8..e5496e5 100644
--- a/mlir/lib/Conversion/MPIToLLVM/MPIToLLVM.cpp
+++ b/mlir/lib/Conversion/MPIToLLVM/MPIToLLVM.cpp
@@ -35,7 +35,7 @@ static Op getOrDefineGlobal(ModuleOp &moduleOp, const Location loc,
   if (!(ret = moduleOp.lookupSymbol<Op>(name))) {
     ConversionPatternRewriter::InsertionGuard guard(rewriter);
     rewriter.setInsertionPointToStart(moduleOp.getBody());
-    ret = rewriter.template create<Op>(loc, std::forward<Args>(args)...);
+    ret = Op::create(rewriter, loc, std::forward<Args>(args)...);
   }
   return ret;
 }
diff --git a/mlir/lib/Conversion/MathToFuncs/MathToFuncs.cpp b/mlir/lib/Conversion/MathToFuncs/MathToFuncs.cpp
index b09afd9..855c582 100644
--- a/mlir/lib/Conversion/MathToFuncs/MathToFuncs.cpp
+++ b/mlir/lib/Conversion/MathToFuncs/MathToFuncs.cpp
@@ -698,7 +698,8 @@ static func::FuncOp createCtlzFunc(ModuleOp *module, Type elementType) {
   scf::IfOp ifOp =
       scf::IfOp::create(builder, elementType, inputEqZero,
                         /*addThenBlock=*/true, /*addElseBlock=*/true);
-  ifOp.getThenBodyBuilder().create<scf::YieldOp>(loc, bitWidthValue);
+  auto thenBuilder = ifOp.getThenBodyBuilder();
+  scf::YieldOp::create(thenBuilder, loc, bitWidthValue);
 
   auto elseBuilder =
       ImplicitLocOpBuilder::atBlockEnd(loc, &ifOp.getElseRegion().front());
diff --git a/mlir/lib/Conversion/MemRefToLLVM/MemRefToLLVM.cpp b/mlir/lib/Conversion/MemRefToLLVM/MemRefToLLVM.cpp
index 53a1912..6ba5bfe4 100644
--- a/mlir/lib/Conversion/MemRefToLLVM/MemRefToLLVM.cpp
+++ b/mlir/lib/Conversion/MemRefToLLVM/MemRefToLLVM.cpp
@@ -575,8 +575,8 @@ private:
     Value sizePtr = LLVM::GEPOp::create(rewriter, loc, indexPtrTy,
                                         getTypeConverter()->getIndexType(),
                                         offsetPtr, idxPlusOne);
-    return rewriter
-        .create<LLVM::LoadOp>(loc, getTypeConverter()->getIndexType(), sizePtr)
+    return LLVM::LoadOp::create(rewriter, loc,
+                                getTypeConverter()->getIndexType(), sizePtr)
         .getResult();
   }
 
diff --git a/mlir/lib/Conversion/SCFToControlFlow/SCFToControlFlow.cpp b/mlir/lib/Conversion/SCFToControlFlow/SCFToControlFlow.cpp
index 240491a..807be7e 100644
--- a/mlir/lib/Conversion/SCFToControlFlow/SCFToControlFlow.cpp
+++ b/mlir/lib/Conversion/SCFToControlFlow/SCFToControlFlow.cpp
@@ -582,6 +582,7 @@ LogicalResult WhileLowering::matchAndRewrite(WhileOp whileOp,
   // block. This should be reconsidered if we allow break/continue in SCF.
   rewriter.setInsertionPointToEnd(before);
   auto condOp = cast<ConditionOp>(before->getTerminator());
+  SmallVector<Value> args = llvm::to_vector(condOp.getArgs());
   rewriter.replaceOpWithNewOp<cf::CondBranchOp>(condOp, condOp.getCondition(),
                                                 after, condOp.getArgs(),
                                                 continuation, ValueRange());
@@ -593,7 +594,7 @@ LogicalResult WhileLowering::matchAndRewrite(WhileOp whileOp,
 
   // Replace the op with values "yielded" from the "before" region, which are
   // visible by dominance.
-  rewriter.replaceOp(whileOp, condOp.getArgs());
+  rewriter.replaceOp(whileOp, args);
 
   return success();
 }
diff --git a/mlir/lib/Conversion/SPIRVToLLVM/SPIRVToLLVM.cpp b/mlir/lib/Conversion/SPIRVToLLVM/SPIRVToLLVM.cpp
index aae3271..9b61540 100644
--- a/mlir/lib/Conversion/SPIRVToLLVM/SPIRVToLLVM.cpp
+++ b/mlir/lib/Conversion/SPIRVToLLVM/SPIRVToLLVM.cpp
@@ -1493,11 +1493,11 @@ public:
     Value extended;
     if (op2TypeWidth < dstTypeWidth) {
       if (isUnsignedIntegerOrVector(op2Type)) {
-        extended = rewriter.template create<LLVM::ZExtOp>(
-            loc, dstType, adaptor.getOperand2());
+        extended =
+            LLVM::ZExtOp::create(rewriter, loc, dstType, adaptor.getOperand2());
       } else {
-        extended = rewriter.template create<LLVM::SExtOp>(
-            loc, dstType, adaptor.getOperand2());
+        extended =
+            LLVM::SExtOp::create(rewriter, loc, dstType, adaptor.getOperand2());
       }
     } else if (op2TypeWidth == dstTypeWidth) {
       extended = adaptor.getOperand2();
@@ -1505,8 +1505,8 @@ public:
       return failure();
     }
 
-    Value result = rewriter.template create<LLVMOp>(
-        loc, dstType, adaptor.getOperand1(), extended);
+    Value result =
+        LLVMOp::create(rewriter, loc, dstType, adaptor.getOperand1(), extended);
     rewriter.replaceOp(op, result);
     return success();
   }
diff --git a/mlir/lib/Conversion/MeshToMPI/CMakeLists.txt b/mlir/lib/Conversion/ShardToMPI/CMakeLists.txt
index 15560aa..564f36f 100644
--- a/mlir/lib/Conversion/MeshToMPI/CMakeLists.txt
+++ b/mlir/lib/Conversion/ShardToMPI/CMakeLists.txt
@@ -1,8 +1,8 @@
-add_mlir_conversion_library(MLIRMeshToMPI
-  MeshToMPI.cpp
+add_mlir_conversion_library(MLIRShardToMPI
+  ShardToMPI.cpp
 
   ADDITIONAL_HEADER_DIRS
-  ${MLIR_MAIN_INCLUDE_DIR}/mlir/Conversion/MeshToMPI
+  ${MLIR_MAIN_INCLUDE_DIR}/mlir/Conversion/ShardToMPI
 
   DEPENDS
   MLIRConversionPassIncGen
@@ -17,7 +17,7 @@ add_mlir_conversion_library(MLIRMeshToMPI
   MLIRLinalgTransforms
   MLIRMemRefDialect
   MLIRPass
-  MLIRMeshDialect
+  MLIRShardDialect
   MLIRMPIDialect
   MLIRTransforms
   )
diff --git a/mlir/lib/Conversion/MeshToMPI/MeshToMPI.cpp b/mlir/lib/Conversion/ShardToMPI/ShardToMPI.cpp
index 63b1fda..fd40e7c 100644
--- a/mlir/lib/Conversion/MeshToMPI/MeshToMPI.cpp
+++ b/mlir/lib/Conversion/ShardToMPI/ShardToMPI.cpp
@@ -1,4 +1,4 @@
-//===- MeshToMPI.cpp - Mesh to MPI  dialect conversion -----------------===//
+//===- ShardToMPI.cpp - Shard to MPI  dialect conversion -----------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -6,11 +6,11 @@
 //
 //===----------------------------------------------------------------------===//
 //
-// This file implements a translation of Mesh communication ops tp MPI ops.
+// This file implements a translation of Shard communication ops to MPI ops.
 //
 //===----------------------------------------------------------------------===//
 
-#include "mlir/Conversion/MeshToMPI/MeshToMPI.h"
+#include "mlir/Conversion/ShardToMPI/ShardToMPI.h"
 
 #include "mlir/Dialect/Affine/IR/AffineOps.h"
 #include "mlir/Dialect/Arith/IR/Arith.h"
@@ -20,11 +20,11 @@
 #include "mlir/Dialect/Linalg/IR/Linalg.h"
 #include "mlir/Dialect/MPI/IR/MPI.h"
 #include "mlir/Dialect/MemRef/IR/MemRef.h"
-#include "mlir/Dialect/Mesh/IR/MeshDialect.h"
-#include "mlir/Dialect/Mesh/IR/MeshOps.h"
-#include "mlir/Dialect/Mesh/Transforms/Simplifications.h"
-#include "mlir/Dialect/Mesh/Transforms/Transforms.h"
 #include "mlir/Dialect/SCF/IR/SCF.h"
+#include "mlir/Dialect/Shard/IR/ShardDialect.h"
+#include "mlir/Dialect/Shard/IR/ShardOps.h"
+#include "mlir/Dialect/Shard/Transforms/Simplifications.h"
+#include "mlir/Dialect/Shard/Transforms/Transforms.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/Dialect/Utils/StaticValueUtils.h"
 #include "mlir/IR/Builders.h"
@@ -35,16 +35,16 @@
 #include "mlir/Transforms/DialectConversion.h"
 #include "mlir/Transforms/GreedyPatternRewriteDriver.h"
 
-#define DEBUG_TYPE "mesh-to-mpi"
+#define DEBUG_TYPE "shard-to-mpi"
 #define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE "]: ")
 
 namespace mlir {
-#define GEN_PASS_DEF_CONVERTMESHTOMPIPASS
+#define GEN_PASS_DEF_CONVERTSHARDTOMPIPASS
 #include "mlir/Conversion/Passes.h.inc"
 } // namespace mlir
 
 using namespace mlir;
-using namespace mesh;
+using namespace shard;
 
 namespace {
 /// Converts a vector of OpFoldResults (ints) into vector of Values of the
@@ -177,9 +177,8 @@ struct ConvertShardingOp : public OpConversionPattern<ShardingOp> {
     auto type = RankedTensorType::get({nSplits, 2}, i64);
     Value resHaloSizes =
         haloSizes.empty()
-            ? rewriter
-                  .create<tensor::EmptyOp>(loc, std::array<int64_t, 2>{0, 0},
-                                           i64)
+            ? tensor::EmptyOp::create(rewriter, loc,
+                                      std::array<int64_t, 2>{0, 0}, i64)
                   .getResult()
             : tensor::FromElementsOp::create(rewriter, loc, type, haloSizes)
                   .getResult();
@@ -188,18 +187,18 @@ struct ConvertShardingOp : public OpConversionPattern<ShardingOp> {
     // maxSplitSize+1}. Store the offsets in the tensor but set trailing
     // elements for smaller split-groups to -1. Computing the max size of the
     // split groups needs using collectiveProcessGroupSize (which needs the
-    // MeshOp)
+    // GridOp)
     Value resOffsets;
     if (adaptor.getStaticShardedDimsOffsets().empty()) {
       resOffsets = tensor::EmptyOp::create(rewriter, loc,
                                            std::array<int64_t, 2>{0, 0}, i64);
     } else {
       SymbolTableCollection symbolTableCollection;
-      auto meshOp = getMesh(op, symbolTableCollection);
+      auto gridOp = getGrid(op, symbolTableCollection);
       int64_t maxSplitSize = 0;
       for (auto axes : splitAxes) {
         int64_t splitSize =
-            collectiveProcessGroupSize(axes.asArrayRef(), meshOp.getShape());
+            collectiveProcessGroupSize(axes.asArrayRef(), gridOp.getShape());
         assert(splitSize != ShapedType::kDynamic);
         maxSplitSize = std::max<int64_t>(maxSplitSize, splitSize);
       }
@@ -218,7 +217,7 @@ struct ConvertShardingOp : public OpConversionPattern<ShardingOp> {
       int64_t curr = 0;
       for (auto [i, axes] : llvm::enumerate(splitAxes)) {
         int64_t splitSize =
-            collectiveProcessGroupSize(axes.asArrayRef(), meshOp.getShape());
+            collectiveProcessGroupSize(axes.asArrayRef(), gridOp.getShape());
         assert(splitSize != ShapedType::kDynamic && splitSize < maxSplitSize);
         ++splitSize; // add one for the total size
         ArrayRef<Value> values(&offsets[curr], splitSize);
@@ -264,20 +263,20 @@ struct ConvertProcessMultiIndexOp
 
     SymbolTableCollection symbolTableCollection;
     Location loc = op.getLoc();
-    auto meshOp = getMesh(op, symbolTableCollection);
-    // For now we only support static mesh shapes
-    if (ShapedType::isDynamicShape(meshOp.getShape()))
+    auto gridOp = getGrid(op, symbolTableCollection);
+    // For now we only support static grid shapes
+    if (ShapedType::isDynamicShape(gridOp.getShape()))
       return failure();
 
     SmallVector<Value> dims;
     llvm::transform(
-        meshOp.getShape(), std::back_inserter(dims), [&](int64_t i) {
+        gridOp.getShape(), std::back_inserter(dims), [&](int64_t i) {
           return arith::ConstantIndexOp::create(rewriter, loc, i).getResult();
         });
-    Value rank = ProcessLinearIndexOp::create(rewriter, op.getLoc(), meshOp);
+    Value rank = ProcessLinearIndexOp::create(rewriter, op.getLoc(), gridOp);
     auto mIdx = linearToMultiIndex(loc, rewriter, rank, dims);
 
-    // optionally extract subset of mesh axes
+    // optionally extract subset of grid axes
     auto axes = adaptor.getAxes();
     if (!axes.empty()) {
       SmallVector<Value> subIndex;
@@ -306,13 +305,11 @@ public:
     auto ctx = op.getContext();
     Value commWorld =
         mpi::CommWorldOp::create(rewriter, loc, mpi::CommType::get(ctx));
-    auto rank =
-        rewriter
-            .create<mpi::CommRankOp>(
-                loc,
-                TypeRange{mpi::RetvalType::get(ctx), rewriter.getI32Type()},
-                commWorld)
-            .getRank();
+    auto rank = mpi::CommRankOp::create(
+                    rewriter, loc,
+                    TypeRange{mpi::RetvalType::get(ctx), rewriter.getI32Type()},
+                    commWorld)
+                    .getRank();
     rewriter.replaceOpWithNewOp<arith::IndexCastOp>(op, rewriter.getIndexType(),
                                                     rank);
     return success();
@@ -338,12 +335,12 @@ struct ConvertNeighborsLinearIndicesOp
 
     Location loc = op.getLoc();
     SymbolTableCollection symbolTableCollection;
-    auto meshOp = getMesh(op, symbolTableCollection);
+    auto gridOp = getGrid(op, symbolTableCollection);
     auto mIdx = adaptor.getDevice();
     auto orgIdx = mIdx[axes[0]];
     SmallVector<Value> dims;
     llvm::transform(
-        meshOp.getShape(), std::back_inserter(dims), [&](int64_t i) {
+        gridOp.getShape(), std::back_inserter(dims), [&](int64_t i) {
           return arith::ConstantIndexOp::create(rewriter, loc, i).getResult();
         });
     Value dimSz = dims[axes[0]];
@@ -394,14 +391,14 @@ struct ConvertShardShapeOp : public OpConversionPattern<ShardShapeOp> {
     auto sharding = op.getSharding().getDefiningOp<ShardingOp>();
     if (!sharding) {
       return op->emitError()
-             << "Expected SharingOp as defining op for sharding"
+             << "Expected ShardingOp as defining op for sharding"
              << " but found " << adaptor.getSharding()[0].getDefiningOp();
     }
 
     // Compute the sharded shape by applying the sharding to the input shape.
     // If shardedDimsOffsets is not defined in the sharding, the shard shape is
     // computed by dividing the dimension size by the number of shards in that
-    // dimension (which is given by the size of the mesh axes provided in
+    // dimension (which is given by the size of the grid axes provided in
     // split-axes). Odd elements get distributed to trailing shards. If a
     // shardedDimsOffsets is provided, the shard shape is computed by
     // subtracting the offset of the current shard from the offset of the next
@@ -431,11 +428,11 @@ struct ConvertShardShapeOp : public OpConversionPattern<ShardShapeOp> {
     SmallVector<Value> multiIdx =
         getMixedAsValues(rewriter, loc, adaptor.getDevice(), dynDevice, index);
 
-    // Get the MeshOp, the mesh shape is needed to compute the sharded shape.
+    // Get the GridOp, the grid shape is needed to compute the sharded shape.
     SymbolTableCollection symbolTableCollection;
-    auto meshOp = getMesh(sharding, symbolTableCollection);
-    // For now we only support static mesh shapes
-    if (ShapedType::isDynamicShape(meshOp.getShape()))
+    auto gridOp = getGrid(sharding, symbolTableCollection);
+    // For now we only support static grid shapes
+    if (ShapedType::isDynamicShape(gridOp.getShape()))
       return failure();
 
     auto splitAxes = sharding.getSplitAxes().getAxes();
@@ -455,7 +452,7 @@ struct ConvertShardShapeOp : public OpConversionPattern<ShardShapeOp> {
             tmp);
     }
 
-    // With static mesh shape the sizes of the split axes are known.
+    // With static grid shape the sizes of the split axes are known.
     // Hence the start/pos for each split axes in shardDimsOffsets can be
     // computed statically.
     int64_t pos = 0;
@@ -475,10 +472,10 @@ struct ConvertShardShapeOp : public OpConversionPattern<ShardShapeOp> {
         // Create a value from the static position in shardDimsOffsets.
         Value posVal = arith::ConstantOp::create(rewriter, loc,
                                                  rewriter.getIndexAttr(pos));
-        // Get the index of the local shard in the mesh axis.
+        // Get the index of the local shard in the grid axis.
         Value idx = multiIdx[axes[0]];
         auto numShards =
-            collectiveProcessGroupSize(axes.asArrayRef(), meshOp.getShape());
+            collectiveProcessGroupSize(axes.asArrayRef(), gridOp.getShape());
         if (shardedDimsOffs) {
           // If sharded dims offsets are provided, use them to compute the
           // sharded shape.
@@ -556,13 +553,13 @@ struct ConvertAllReduceOp : public OpConversionPattern<AllReduceOp> {
   matchAndRewrite(AllReduceOp op, OpAdaptor adaptor,
                   ConversionPatternRewriter &rewriter) const override {
     SymbolTableCollection symbolTableCollection;
-    auto mesh = adaptor.getMesh();
-    mlir::mesh::MeshOp meshOp = getMesh(op, symbolTableCollection);
-    if (!meshOp)
-      return op->emitError() << "No mesh found for AllReduceOp";
-    if (ShapedType::isDynamicShape(meshOp.getShape()))
+    auto grid = adaptor.getGrid();
+    mlir::shard::GridOp gridOp = getGrid(op, symbolTableCollection);
+    if (!gridOp)
+      return op->emitError() << "No grid found for AllReduceOp";
+    if (ShapedType::isDynamicShape(gridOp.getShape()))
       return op->emitError()
-             << "Dynamic mesh shape not supported in AllReduceOp";
+             << "Dynamic grid shape not supported in AllReduceOp";
 
     ImplicitLocOpBuilder iBuilder(op.getLoc(), rewriter);
     Value input = adaptor.getInput();
@@ -592,27 +589,27 @@ struct ConvertAllReduceOp : public OpConversionPattern<AllReduceOp> {
     linalg::CopyOp::create(iBuilder, input, buffer);
 
     // Get an MPI_Comm_split for the AllReduce operation.
-    // The color is the linear index of the process in the mesh along the
-    // non-reduced axes. The key is the linear index of the process in the mesh
+    // The color is the linear index of the process in the grid along the
+    // non-reduced axes. The key is the linear index of the process in the grid
     // along the reduced axes.
-    SmallVector<Type> indexResultTypes(meshOp.getShape().size(),
+    SmallVector<Type> indexResultTypes(gridOp.getShape().size(),
                                        iBuilder.getIndexType());
     SmallVector<Value> myMultiIndex =
-        ProcessMultiIndexOp::create(iBuilder, indexResultTypes, mesh)
+        ProcessMultiIndexOp::create(iBuilder, indexResultTypes, grid)
             .getResult();
     Value zero = arith::ConstantIndexOp::create(iBuilder, 0);
     SmallVector<Value> multiKey(myMultiIndex.size(), zero);
 
-    auto redAxes = adaptor.getMeshAxes();
+    auto redAxes = adaptor.getGridAxes();
     for (auto axis : redAxes) {
       multiKey[axis] = myMultiIndex[axis];
       myMultiIndex[axis] = zero;
     }
 
     Value color =
-        createProcessLinearIndex(mesh, myMultiIndex, redAxes, iBuilder);
+        createProcessLinearIndex(grid, myMultiIndex, redAxes, iBuilder);
     color = arith::IndexCastOp::create(iBuilder, iBuilder.getI32Type(), color);
-    Value key = createProcessLinearIndex(mesh, multiKey, redAxes, iBuilder);
+    Value key = createProcessLinearIndex(grid, multiKey, redAxes, iBuilder);
     key = arith::IndexCastOp::create(iBuilder, iBuilder.getI32Type(), key);
 
     // Finally split the communicator
@@ -698,15 +695,14 @@ struct ConvertUpdateHaloOp : public OpConversionPattern<UpdateHaloOp> {
     }
     auto rank = cast<ShapedType>(array.getType()).getRank();
     auto opSplitAxes = adaptor.getSplitAxes().getAxes();
-    auto mesh = adaptor.getMesh();
-    auto meshOp = getMesh(op, symbolTableCollection);
+    auto grid = adaptor.getGrid();
+    auto gridOp = getGrid(op, symbolTableCollection);
     // subviews need Index values
     for (auto &sz : haloSizes) {
       if (auto value = dyn_cast<Value>(sz))
-        sz =
-            rewriter
-                .create<arith::IndexCastOp>(loc, rewriter.getIndexType(), value)
-                .getResult();
+        sz = arith::IndexCastOp::create(rewriter, loc, rewriter.getIndexType(),
+                                        value)
+                 .getResult();
     }
 
     // most of the offset/size/stride data is the same for all dims
@@ -745,10 +741,10 @@ struct ConvertUpdateHaloOp : public OpConversionPattern<UpdateHaloOp> {
     auto zeroAttr = rewriter.getI32IntegerAttr(0); // for detecting v<0
     auto zero = arith::ConstantOp::create(rewriter, loc, zeroAttr);
 
-    SmallVector<Type> indexResultTypes(meshOp.getShape().size(),
+    SmallVector<Type> indexResultTypes(gridOp.getShape().size(),
                                        rewriter.getIndexType());
     auto myMultiIndex =
-        ProcessMultiIndexOp::create(rewriter, loc, indexResultTypes, mesh)
+        ProcessMultiIndexOp::create(rewriter, loc, indexResultTypes, grid)
             .getResult();
     // traverse all split axes from high to low dim
     for (ssize_t dim = opSplitAxes.size() - 1; dim >= 0; --dim) {
@@ -758,9 +754,8 @@ struct ConvertUpdateHaloOp : public OpConversionPattern<UpdateHaloOp> {
       assert(currHaloDim >= 0 && (size_t)currHaloDim < haloSizes.size() / 2);
       // Get the linearized ids of the neighbors (down and up) for the
       // given split
-      auto tmp = rewriter
-                     .create<NeighborsLinearIndicesOp>(loc, mesh, myMultiIndex,
-                                                       splitAxes)
+      auto tmp = NeighborsLinearIndicesOp::create(rewriter, loc, grid,
+                                                  myMultiIndex, splitAxes)
                      .getResults();
       // MPI operates on i32...
       Value neighbourIDs[2] = {
@@ -791,7 +786,7 @@ struct ConvertUpdateHaloOp : public OpConversionPattern<UpdateHaloOp> {
         dimSizes[dim] = upperHalo ? haloSizes[currHaloDim * 2 + 1]
                                   : haloSizes[currHaloDim * 2];
         // Check if we need to send and/or receive
-        // Processes on the mesh borders have only one neighbor
+        // Processes on the grid borders have only one neighbor
         auto to = upperHalo ? neighbourIDs[0] : neighbourIDs[1];
         auto from = upperHalo ? neighbourIDs[1] : neighbourIDs[0];
         auto hasFrom = arith::CmpIOp::create(
@@ -869,8 +864,8 @@ struct ConvertUpdateHaloOp : public OpConversionPattern<UpdateHaloOp> {
   }
 };
 
-struct ConvertMeshToMPIPass
-    : public impl::ConvertMeshToMPIPassBase<ConvertMeshToMPIPass> {
+struct ConvertShardToMPIPass
+    : public impl::ConvertShardToMPIPassBase<ConvertShardToMPIPass> {
   using Base::Base;
 
   /// Run the dialect converter on the module.
@@ -879,12 +874,12 @@ struct ConvertMeshToMPIPass
     RewritePatternSet patterns(ctxt);
     ConversionTarget target(getContext());
 
-    // Define a type converter to convert mesh::ShardingType,
+    // Define a type converter to convert shard::ShardingType,
     // mostly for use in return operations.
     TypeConverter typeConverter;
     typeConverter.addConversion([](Type type) { return type; });
 
-    // convert mesh::ShardingType to a tuple of RankedTensorTypes
+    // convert shard::ShardingType to a tuple of RankedTensorTypes
     typeConverter.addConversion(
         [](ShardingType type,
            SmallVectorImpl<Type> &results) -> std::optional<LogicalResult> {
@@ -920,10 +915,10 @@ struct ConvertMeshToMPIPass
           return results;
         });
 
-    // No mesh dialect should left after conversion...
-    target.addIllegalDialect<mesh::MeshDialect>();
-    // ...except the global MeshOp. MeshShapeOp which will get folded later.
-    target.addLegalOp<mesh::MeshOp, mesh::MeshShapeOp>();
+    // No shard dialect should left after conversion...
+    target.addIllegalDialect<shard::ShardDialect>();
+    // ...except the global GridOp. GridShapeOp which will get folded later.
+    target.addLegalOp<shard::GridOp, shard::GridShapeOp>();
     // Allow all the stuff that our patterns will convert to
     target.addLegalDialect<
         BuiltinDialect, mpi::MPIDialect, scf::SCFDialect, arith::ArithDialect,
@@ -951,7 +946,7 @@ struct ConvertMeshToMPIPass
     // Folding patterns cannot be mixed with conversion patterns -> extra pass.
     patterns.clear();
     SymbolTableCollection symbolTableCollection;
-    mlir::mesh::populateFoldingPatterns(patterns, symbolTableCollection);
+    mlir::shard::populateFoldingPatterns(patterns, symbolTableCollection);
     (void)applyPatternsGreedily(getOperation(), std::move(patterns));
   }
 };
diff --git a/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp b/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
index 5c7c027..0e3de06 100644
--- a/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
+++ b/mlir/lib/Conversion/TosaToLinalg/TosaToLinalg.cpp
@@ -569,10 +569,9 @@ static Value createLinalgBodyCalculationForElementwiseOp(
     // to UIToFP.
     if (srcTy.isUnsignedInteger() && isa<FloatType>(dstTy)) {
       auto unrealizedCast =
-          rewriter
-              .create<UnrealizedConversionCastOp>(
-                  loc, rewriter.getIntegerType(srcTy.getIntOrFloatBitWidth()),
-                  args[0])
+          UnrealizedConversionCastOp::create(
+              rewriter, loc,
+              rewriter.getIntegerType(srcTy.getIntOrFloatBitWidth()), args[0])
               .getResult(0);
       return arith::UIToFPOp::create(rewriter, loc, resultTypes[0],
                                      unrealizedCast);
@@ -868,14 +867,13 @@ static Value broadcastDynamicDimension(PatternRewriter &rewriter, Location loc,
 
     // Emit 'linalg.generic' op
     auto resultTensor =
-        opBuilder
-            .create<linalg::GenericOp>(
-                loc, outputTensor.getType(), operand, outputTensor, affineMaps,
-                getNParallelLoopsAttrs(rank),
-                [&](OpBuilder &opBuilder, Location loc, ValueRange blockArgs) {
-                  // Emit 'linalg.yield' op
-                  linalg::YieldOp::create(opBuilder, loc, blockArgs.front());
-                })
+        linalg::GenericOp::create(
+            opBuilder, loc, outputTensor.getType(), operand, outputTensor,
+            affineMaps, getNParallelLoopsAttrs(rank),
+            [&](OpBuilder &opBuilder, Location loc, ValueRange blockArgs) {
+              // Emit 'linalg.yield' op
+              linalg::YieldOp::create(opBuilder, loc, blockArgs.front());
+            })
             .getResult(0);
 
     // Cast to original operand type if necessary
@@ -1155,11 +1153,9 @@ static LogicalResult reduceMatchAndRewriteHelper(OpTy op, uint64_t axis,
   inputs.push_back(input);
 
   // First fill the output buffer with the init value.
-  auto emptyTensor =
-      rewriter
-          .create<tensor::EmptyOp>(loc, reduceShape, resultTy.getElementType(),
-                                   dynDims)
-          .getResult();
+  auto emptyTensor = tensor::EmptyOp::create(rewriter, loc, reduceShape,
+                                             resultTy.getElementType(), dynDims)
+                         .getResult();
 
   auto fillValueAttr = createInitialValueForReduceOp(op, elementTy, rewriter);
   if (!fillValueAttr)
@@ -1167,10 +1163,10 @@ static LogicalResult reduceMatchAndRewriteHelper(OpTy op, uint64_t axis,
         op, "No initial value found for reduction operation");
 
   auto fillValue = arith::ConstantOp::create(rewriter, loc, fillValueAttr);
-  auto filledTensor = rewriter
-                          .create<linalg::FillOp>(loc, ValueRange{fillValue},
-                                                  ValueRange{emptyTensor})
-                          .result();
+  auto filledTensor =
+      linalg::FillOp::create(rewriter, loc, ValueRange{fillValue},
+                             ValueRange{emptyTensor})
+          .result();
   outputs.push_back(filledTensor);
 
   bool isNanIgnoreMode = false;
@@ -1186,14 +1182,12 @@ static LogicalResult reduceMatchAndRewriteHelper(OpTy op, uint64_t axis,
       auto trueAttr = rewriter.getBoolAttr(true);
       auto trueValue = arith::ConstantOp::create(rewriter, loc, trueAttr);
       auto emptyBoolTensor =
-          rewriter
-              .create<tensor::EmptyOp>(loc, reduceShape, trueValue.getType(),
-                                       dynDims)
+          tensor::EmptyOp::create(rewriter, loc, reduceShape,
+                                  trueValue.getType(), dynDims)
               .getResult();
       auto allResultsNaNTensor =
-          rewriter
-              .create<linalg::FillOp>(loc, ValueRange{trueValue},
-                                      ValueRange{emptyBoolTensor})
+          linalg::FillOp::create(rewriter, loc, ValueRange{trueValue},
+                                 ValueRange{emptyBoolTensor})
               .result();
       // Note that because the linalg::ReduceOp has two variadic arguments
       // (inputs and outputs) and it has the SameVariadicOperandSize trait we
@@ -1261,22 +1255,19 @@ static LogicalResult reduceMatchAndRewriteHelper(OpTy op, uint64_t axis,
         APFloat::getNaN(cast<FloatType>(elementTy).getFloatSemantics(), false));
     auto nanValue = arith::ConstantOp::create(rewriter, loc, nanValueAttr);
     auto emptyNanTensor =
-        rewriter
-            .create<tensor::EmptyOp>(loc, reduceShape,
-                                     resultTy.getElementType(), dynDims)
+        tensor::EmptyOp::create(rewriter, loc, reduceShape,
+                                resultTy.getElementType(), dynDims)
             .getResult();
     auto nanFilledTensor =
-        rewriter
-            .create<linalg::FillOp>(loc, ValueRange{nanValue},
-                                    ValueRange{emptyNanTensor})
+        linalg::FillOp::create(rewriter, loc, ValueRange{nanValue},
+                               ValueRange{emptyNanTensor})
             .result();
 
     // Create an empty tensor, non need to fill this since it will be
     // overwritten by the select.
     auto finalEmptyTensor =
-        rewriter
-            .create<tensor::EmptyOp>(loc, reduceShape,
-                                     resultTy.getElementType(), dynDims)
+        tensor::EmptyOp::create(rewriter, loc, reduceShape,
+                                resultTy.getElementType(), dynDims)
             .getResult();
 
     // Do a selection between the tensors akin to:
@@ -1503,12 +1494,11 @@ public:
           Value shift = shiftConstant ? shiftConstant : blockArgs[shiftArg];
 
           if (valueTy.isUnsignedInteger()) {
-            value = nestedBuilder
-                        .create<UnrealizedConversionCastOp>(
-                            nestedLoc,
-                            nestedBuilder.getIntegerType(
-                                valueTy.getIntOrFloatBitWidth()),
-                            value)
+            value = UnrealizedConversionCastOp::create(
+                        nestedBuilder, nestedLoc,
+                        nestedBuilder.getIntegerType(
+                            valueTy.getIntOrFloatBitWidth()),
+                        value)
                         .getResult(0);
           }
           if (valueTy.getIntOrFloatBitWidth() < 32) {
@@ -1557,9 +1547,8 @@ public:
           }
 
           if (outIntType.isUnsignedInteger()) {
-            value = nestedBuilder
-                        .create<UnrealizedConversionCastOp>(nestedLoc,
-                                                            outIntType, value)
+            value = UnrealizedConversionCastOp::create(nestedBuilder, nestedLoc,
+                                                       outIntType, value)
                         .getResult(0);
           }
           linalg::YieldOp::create(nestedBuilder, loc, value);
@@ -2095,10 +2084,9 @@ public:
     Value axisDimSize = tensor::DimOp::create(rewriter, loc, input, axis);
 
     // First fill the output buffer with the init value.
-    auto emptyTensor = rewriter
-                           .create<tensor::EmptyOp>(loc, inputTy.getShape(),
-                                                    inputTy.getElementType(),
-                                                    ArrayRef<Value>({dynDims}))
+    auto emptyTensor = tensor::EmptyOp::create(
+                           rewriter, loc, inputTy.getShape(),
+                           inputTy.getElementType(), ArrayRef<Value>({dynDims}))
                            .getResult();
     SmallVector<AffineMap, 2> affineMaps = {
         rewriter.getMultiDimIdentityMap(resultTy.getRank())};
@@ -2241,23 +2229,22 @@ public:
     }
 
     // First fill the output buffer for the index.
-    auto emptyTensorIdx = rewriter
-                              .create<tensor::EmptyOp>(loc, resultTy.getShape(),
-                                                       outElementTy, dynDims)
-                              .getResult();
+    auto emptyTensorIdx =
+        tensor::EmptyOp::create(rewriter, loc, resultTy.getShape(),
+                                outElementTy, dynDims)
+            .getResult();
     auto fillValueIdx = arith::ConstantOp::create(
         rewriter, loc, rewriter.getIntegerAttr(outElementTy, 0));
     auto filledTensorIdx =
-        rewriter
-            .create<linalg::FillOp>(loc, ValueRange{fillValueIdx},
-                                    ValueRange{emptyTensorIdx})
+        linalg::FillOp::create(rewriter, loc, ValueRange{fillValueIdx},
+                               ValueRange{emptyTensorIdx})
             .result();
 
     // Second fill the output buffer for the running max.
-    auto emptyTensorMax = rewriter
-                              .create<tensor::EmptyOp>(loc, resultTy.getShape(),
-                                                       inElementTy, dynDims)
-                              .getResult();
+    auto emptyTensorMax =
+        tensor::EmptyOp::create(rewriter, loc, resultTy.getShape(), inElementTy,
+                                dynDims)
+            .getResult();
     auto fillValueMaxAttr =
         createInitialValueForReduceOp(argmaxOp, inElementTy, rewriter);
 
@@ -2268,9 +2255,8 @@ public:
     auto fillValueMax =
         arith::ConstantOp::create(rewriter, loc, fillValueMaxAttr);
     auto filledTensorMax =
-        rewriter
-            .create<linalg::FillOp>(loc, ValueRange{fillValueMax},
-                                    ValueRange{emptyTensorMax})
+        linalg::FillOp::create(rewriter, loc, ValueRange{fillValueMax},
+                               ValueRange{emptyTensorMax})
             .result();
 
     // We need to reduce along the arg-max axis, with parallel operations along
@@ -2371,9 +2357,8 @@ public:
 
     auto loc = op.getLoc();
     auto emptyTensor =
-        rewriter
-            .create<tensor::EmptyOp>(loc, resultTy.getShape(), resultElementTy,
-                                     dynamicDims)
+        tensor::EmptyOp::create(rewriter, loc, resultTy.getShape(),
+                                resultElementTy, dynamicDims)
             .getResult();
 
     SmallVector<AffineMap, 2> affineMaps = {
@@ -2448,10 +2433,10 @@ public:
       }
     }
 
-    auto emptyTensor = rewriter
-                           .create<tensor::EmptyOp>(loc, resultTy.getShape(),
-                                                    resultElementTy, dynDims)
-                           .getResult();
+    auto emptyTensor =
+        tensor::EmptyOp::create(rewriter, loc, resultTy.getShape(),
+                                resultElementTy, dynDims)
+            .getResult();
 
     SmallVector<AffineMap, 2> affineMaps = {
         rewriter.getMultiDimIdentityMap(resultTy.getRank()),
@@ -2585,10 +2570,10 @@ struct RFFT2dConverter final : public OpRewritePattern<RFFT2dOp> {
         tensor::EmptyOp::create(rewriter, loc, type, dynamicSizes);
     auto fillValueAttr = rewriter.getZeroAttr(type.getElementType());
     auto fillValue = arith::ConstantOp::create(rewriter, loc, fillValueAttr);
-    auto filledTensor = rewriter
-                            .create<linalg::FillOp>(loc, ValueRange{fillValue},
-                                                    ValueRange{emptyTensor})
-                            .result();
+    auto filledTensor =
+        linalg::FillOp::create(rewriter, loc, ValueRange{fillValue},
+                               ValueRange{emptyTensor})
+            .result();
     return filledTensor;
   }
 
diff --git a/mlir/lib/Conversion/TosaToLinalg/TosaToLinalgNamed.cpp b/mlir/lib/Conversion/TosaToLinalg/TosaToLinalgNamed.cpp
index 3a20524..da1fb20 100644
--- a/mlir/lib/Conversion/TosaToLinalg/TosaToLinalgNamed.cpp
+++ b/mlir/lib/Conversion/TosaToLinalg/TosaToLinalgNamed.cpp
@@ -64,19 +64,20 @@ linalgIntBroadcastExtSIAdd(PatternRewriter &rewriter, Location loc, Value bias,
                            Value conv, Value result,
                            ArrayRef<AffineMap> indexingMaps) {
   ShapedType resultTy = cast<ShapedType>(conv.getType());
-  return rewriter
-      .create<linalg::GenericOp>(
-          loc, resultTy, ValueRange({bias, conv}), result, indexingMaps,
-          getNParallelLoopsAttrs(resultTy.getRank()),
-          [](OpBuilder &builder, Location loc, ValueRange args) {
-            Value biasVal = args[0];
-            Type resType = args[1].getType();
-            if (resType != biasVal.getType()) {
-              biasVal = arith::ExtSIOp::create(builder, loc, resType, biasVal);
-            }
-            Value added = arith::AddIOp::create(builder, loc, biasVal, args[1]);
-            linalg::YieldOp::create(builder, loc, added);
-          })
+  return linalg::GenericOp::create(
+             rewriter, loc, resultTy, ValueRange({bias, conv}), result,
+             indexingMaps, getNParallelLoopsAttrs(resultTy.getRank()),
+             [](OpBuilder &builder, Location loc, ValueRange args) {
+               Value biasVal = args[0];
+               Type resType = args[1].getType();
+               if (resType != biasVal.getType()) {
+                 biasVal =
+                     arith::ExtSIOp::create(builder, loc, resType, biasVal);
+               }
+               Value added =
+                   arith::AddIOp::create(builder, loc, biasVal, args[1]);
+               linalg::YieldOp::create(builder, loc, added);
+             })
       .getResult(0);
 }
 
@@ -124,23 +125,23 @@ static mlir::Value linalgBroadcastAndMaybeExt(PatternRewriter &rewriter,
   indexingMaps.push_back(rewriter.getMultiDimIdentityMap(resultRank));
 
   // Build the broadcast-like operation as a linalg.generic.
-  return rewriter
-      .create<linalg::GenericOp>(
-          loc, resultTy, ValueRange({source}), result, indexingMaps,
-          getNParallelLoopsAttrs(resultTy.getRank()),
-          [&resultTy](OpBuilder &builder, Location loc, ValueRange args) {
-            Value biasVal = args[0];
-            Type resType = args[1].getType();
-            if (resType != biasVal.getType()) {
-              biasVal =
-                  resultTy.getElementType().isFloat()
-                      ? arith::ExtFOp::create(builder, loc, resType, biasVal)
-                            .getResult()
-                      : arith::ExtSIOp::create(builder, loc, resType, biasVal)
-                            .getResult();
-            }
-            linalg::YieldOp::create(builder, loc, biasVal);
-          })
+  return linalg::GenericOp::create(
+             rewriter, loc, resultTy, ValueRange({source}), result,
+             indexingMaps, getNParallelLoopsAttrs(resultTy.getRank()),
+             [&resultTy](OpBuilder &builder, Location loc, ValueRange args) {
+               Value biasVal = args[0];
+               Type resType = args[1].getType();
+               if (resType != biasVal.getType()) {
+                 biasVal =
+                     resultTy.getElementType().isFloat()
+                         ? arith::ExtFOp::create(builder, loc, resType, biasVal)
+                               .getResult()
+                         : arith::ExtSIOp::create(builder, loc, resType,
+                                                  biasVal)
+                               .getResult();
+               }
+               linalg::YieldOp::create(builder, loc, biasVal);
+             })
       .getResult(0);
 }
 
@@ -397,21 +398,19 @@ public:
       auto iZpVal = arith::ConstantOp::create(rewriter, loc, iZp);
       auto kZpVal = arith::ConstantOp::create(rewriter, loc, kZp);
 
-      Value conv =
-          rewriter
-              .create<LinalgConvQOp>(
-                  loc, resultTy, ValueRange{input, weight, iZpVal, kZpVal},
-                  ValueRange{broadcastBias}, strideAttr, dilationAttr)
-              ->getResult(0);
+      Value conv = LinalgConvQOp::create(
+                       rewriter, loc, resultTy,
+                       ValueRange{input, weight, iZpVal, kZpVal},
+                       ValueRange{broadcastBias}, strideAttr, dilationAttr)
+                       ->getResult(0);
 
       rewriter.replaceOp(op, conv);
       return success();
     }
 
-    Value conv = rewriter
-                     .create<LinalgConvOp>(
-                         loc, accTy, ValueRange{input, weight},
-                         ValueRange{broadcastBias}, strideAttr, dilationAttr)
+    Value conv = LinalgConvOp::create(
+                     rewriter, loc, accTy, ValueRange{input, weight},
+                     ValueRange{broadcastBias}, strideAttr, dilationAttr)
                      ->getResult(0);
 
     // We may need to truncate back to the result type if the accumulator was
@@ -529,9 +528,8 @@ public:
     Value emptyTensor = tensor::EmptyOp::create(
         rewriter, loc, linalgConvTy.getShape(), accETy, filteredDims);
     Value zero = arith::ConstantOp::create(rewriter, loc, resultZeroAttr);
-    Value zeroTensor = rewriter
-                           .create<linalg::FillOp>(loc, ValueRange{zero},
-                                                   ValueRange{emptyTensor})
+    Value zeroTensor = linalg::FillOp::create(rewriter, loc, ValueRange{zero},
+                                              ValueRange{emptyTensor})
                            .result();
 
     Value biasEmptyTensor = tensor::EmptyOp::create(
@@ -544,10 +542,9 @@ public:
     indexingMaps.push_back(rewriter.getMultiDimIdentityMap(resultRank));
 
     if (hasNullZps) {
-      Value conv = rewriter
-                       .create<linalg::DepthwiseConv2DNhwcHwcmOp>(
-                           loc, linalgConvTy, ValueRange{input, weight},
-                           ValueRange{zeroTensor}, strideAttr, dilationAttr)
+      Value conv = linalg::DepthwiseConv2DNhwcHwcmOp::create(
+                       rewriter, loc, linalgConvTy, ValueRange{input, weight},
+                       ValueRange{zeroTensor}, strideAttr, dilationAttr)
                        .getResult(0);
 
       // We may need to truncate back to the result type if the accumulator was
@@ -565,22 +562,20 @@ public:
           rewriter, loc, resultTy, conv, reassociationMap);
 
       Value result =
-          rewriter
-              .create<linalg::GenericOp>(
-                  loc, resultTy, ValueRange({bias, convReshape}),
-                  biasEmptyTensor, indexingMaps,
-                  getNParallelLoopsAttrs(resultRank),
-                  [&](OpBuilder &nestedBuilder, Location nestedLoc,
-                      ValueRange args) {
-                    Value added;
-                    if (llvm::isa<FloatType>(inputETy))
-                      added = arith::AddFOp::create(nestedBuilder, loc, args[0],
-                                                    args[1]);
-                    else
-                      added = arith::AddIOp::create(nestedBuilder, loc, args[0],
-                                                    args[1]);
-                    linalg::YieldOp::create(nestedBuilder, nestedLoc, added);
-                  })
+          linalg::GenericOp::create(
+              rewriter, loc, resultTy, ValueRange({bias, convReshape}),
+              biasEmptyTensor, indexingMaps, getNParallelLoopsAttrs(resultRank),
+              [&](OpBuilder &nestedBuilder, Location nestedLoc,
+                  ValueRange args) {
+                Value added;
+                if (llvm::isa<FloatType>(inputETy))
+                  added = arith::AddFOp::create(nestedBuilder, loc, args[0],
+                                                args[1]);
+                else
+                  added = arith::AddIOp::create(nestedBuilder, loc, args[0],
+                                                args[1]);
+                linalg::YieldOp::create(nestedBuilder, nestedLoc, added);
+              })
               .getResult(0);
       rewriter.replaceOp(op, result);
     } else {
@@ -588,12 +583,11 @@ public:
       IntegerAttr wZp = rewriter.getI32IntegerAttr(weightZpVal);
       auto iZpVal = arith::ConstantOp::create(rewriter, loc, iZp);
       auto kZpVal = arith::ConstantOp::create(rewriter, loc, wZp);
-      Value conv =
-          rewriter
-              .create<linalg::DepthwiseConv2DNhwcHwcmQOp>(
-                  loc, linalgConvTy, ValueRange{input, weight, iZpVal, kZpVal},
-                  ValueRange{zeroTensor}, strideAttr, dilationAttr)
-              .getResult(0);
+      Value conv = linalg::DepthwiseConv2DNhwcHwcmQOp::create(
+                       rewriter, loc, linalgConvTy,
+                       ValueRange{input, weight, iZpVal, kZpVal},
+                       ValueRange{zeroTensor}, strideAttr, dilationAttr)
+                       .getResult(0);
       SmallVector<ReassociationExprs, 4> reassociationMap;
       createDepthwiseConvCollapseMap(resultRank, reassociationMap, rewriter);
       Value convReshape = tensor::CollapseShapeOp::create(
@@ -639,9 +633,8 @@ public:
     auto emptyTensor =
         tensor::EmptyOp::create(rewriter, loc, outputTy.getShape(),
                                 outputTy.getElementType(), filteredDims);
-    Value zeroTensor = rewriter
-                           .create<linalg::FillOp>(loc, ValueRange{zero},
-                                                   ValueRange{emptyTensor})
+    Value zeroTensor = linalg::FillOp::create(rewriter, loc, ValueRange{zero},
+                                              ValueRange{emptyTensor})
                            .result();
 
     FailureOr<int64_t> maybeAZp = op.getAZeroPoint();
@@ -910,20 +903,18 @@ public:
         rewriter, loc, accTy.getShape(), accETy, dynamicDims);
 
     Value filledEmptyTensor =
-        rewriter
-            .create<linalg::FillOp>(loc, ValueRange{initialValue},
-                                    ValueRange{poolEmptyTensor})
+        linalg::FillOp::create(rewriter, loc, ValueRange{initialValue},
+                               ValueRange{poolEmptyTensor})
             .result();
 
     Value fakeWindowDims =
         tensor::EmptyOp::create(rewriter, loc, kernel, accETy);
 
     // Sum across the pooled region.
-    Value poolingOp = rewriter
-                          .create<linalg::PoolingNhwcSumOp>(
-                              loc, ArrayRef<Type>{accTy},
-                              ValueRange{paddedInput, fakeWindowDims},
-                              filledEmptyTensor, strideAttr, dilationAttr)
+    Value poolingOp = linalg::PoolingNhwcSumOp::create(
+                          rewriter, loc, ArrayRef<Type>{accTy},
+                          ValueRange{paddedInput, fakeWindowDims},
+                          filledEmptyTensor, strideAttr, dilationAttr)
                           .getResult(0);
 
     // Normalize the summed value by the number of elements grouped in each
@@ -1050,10 +1041,9 @@ public:
             Value shift = arith::AddIOp::create(rewriter, loc, k8, thirty8);
 
             auto scaled =
-                rewriter
-                    .create<tosa::ApplyScaleOp>(
-                        loc, rewriter.getI32Type(), poolVal, multiplier, shift,
-                        rewriter.getStringAttr("SINGLE_ROUND"))
+                tosa::ApplyScaleOp::create(
+                    rewriter, loc, rewriter.getI32Type(), poolVal, multiplier,
+                    shift, rewriter.getStringAttr("SINGLE_ROUND"))
                     .getResult();
 
             // If we have quantization information we need to apply output
diff --git a/mlir/lib/Conversion/VectorToGPU/VectorToGPU.cpp b/mlir/lib/Conversion/VectorToGPU/VectorToGPU.cpp
index 77aab85..a425eff 100644
--- a/mlir/lib/Conversion/VectorToGPU/VectorToGPU.cpp
+++ b/mlir/lib/Conversion/VectorToGPU/VectorToGPU.cpp
@@ -482,14 +482,12 @@ struct CombineTransferReadOpTranspose final
         permutationMap.compose(transferReadOp.getPermutationMap());
 
     auto loc = op.getLoc();
-    Value result =
-        rewriter
-            .create<vector::TransferReadOp>(
-                loc, resultType, transferReadOp.getBase(),
-                transferReadOp.getIndices(), AffineMapAttr::get(newMap),
-                transferReadOp.getPadding(), transferReadOp.getMask(),
-                transferReadOp.getInBoundsAttr())
-            .getResult();
+    Value result = vector::TransferReadOp::create(
+                       rewriter, loc, resultType, transferReadOp.getBase(),
+                       transferReadOp.getIndices(), AffineMapAttr::get(newMap),
+                       transferReadOp.getPadding(), transferReadOp.getMask(),
+                       transferReadOp.getInBoundsAttr())
+                       .getResult();
 
     // Fuse through the integer extend op.
     if (extOp) {
diff --git a/mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVM.cpp b/mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVM.cpp
index 9cd491c..4307bc6 100644
--- a/mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVM.cpp
+++ b/mlir/lib/Conversion/VectorToLLVM/ConvertVectorToLLVM.cpp
@@ -29,7 +29,9 @@
 #include "mlir/Target/LLVMIR/TypeToLLVM.h"
 #include "mlir/Transforms/DialectConversion.h"
 #include "llvm/ADT/APFloat.h"
+#include "llvm/IR/LLVMContext.h"
 #include "llvm/Support/Casting.h"
+
 #include <optional>
 
 using namespace mlir;
diff --git a/mlir/lib/Dialect/AMX/IR/AMXDialect.cpp b/mlir/lib/Dialect/AMX/IR/AMXDialect.cpp
index 748ff1e..6f3110c 100644
--- a/mlir/lib/Dialect/AMX/IR/AMXDialect.cpp
+++ b/mlir/lib/Dialect/AMX/IR/AMXDialect.cpp
@@ -96,9 +96,8 @@ static Value getStride(Location loc, MemRefType mType, Value base,
     MemRefDescriptor memrefDescriptor(base);
     auto attr = rewriter.getI64IntegerAttr(bytes);
     Value scale = LLVM::ConstantOp::create(rewriter, loc, llvmInt64Type, attr);
-    return rewriter
-        .create<LLVM::MulOp>(loc, llvmInt64Type, scale,
-                             memrefDescriptor.stride(rewriter, loc, preLast))
+    return LLVM::MulOp::create(rewriter, loc, llvmInt64Type, scale,
+                               memrefDescriptor.stride(rewriter, loc, preLast))
         .getResult();
   }
   // Use direct constant for static stride.
diff --git a/mlir/lib/Dialect/Affine/Transforms/DecomposeAffineOps.cpp b/mlir/lib/Dialect/Affine/Transforms/DecomposeAffineOps.cpp
index 3c00b32..6265f46 100644
--- a/mlir/lib/Dialect/Affine/Transforms/DecomposeAffineOps.cpp
+++ b/mlir/lib/Dialect/Affine/Transforms/DecomposeAffineOps.cpp
@@ -15,13 +15,13 @@
 #include "mlir/Dialect/Affine/Transforms/Transforms.h"
 #include "mlir/IR/PatternMatch.h"
 #include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
+#include "llvm/Support/InterleavedRange.h"
 
 using namespace mlir;
 using namespace mlir::affine;
 
 #define DEBUG_TYPE "decompose-affine-ops"
-#define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE "]: ")
-#define DBGSNL() (llvm::dbgs() << "\n")
 
 /// Count the number of loops surrounding `operand` such that operand could be
 /// hoisted above.
@@ -115,7 +115,7 @@ FailureOr<AffineApplyOp> mlir::affine::decompose(RewriterBase &rewriter,
     return rewriter.notifyMatchFailure(
         op, "only add or mul binary expr can be reassociated");
 
-  LLVM_DEBUG(DBGS() << "Start decomposeIntoFinerGrainedOps: " << op << "\n");
+  LDBG() << "Start decomposeIntoFinerGrainedOps: " << op;
 
   // 2. Iteratively extract the RHS subexpressions while the top-level binary
   // expr kind remains the same.
@@ -125,11 +125,11 @@ FailureOr<AffineApplyOp> mlir::affine::decompose(RewriterBase &rewriter,
     auto currentBinExpr = dyn_cast<AffineBinaryOpExpr>(remainingExp);
     if (!currentBinExpr || currentBinExpr.getKind() != binExpr.getKind()) {
       subExpressions.push_back(remainingExp);
-      LLVM_DEBUG(DBGS() << "--terminal: " << subExpressions.back() << "\n");
+      LDBG() << "--terminal: " << subExpressions.back();
       break;
     }
     subExpressions.push_back(currentBinExpr.getRHS());
-    LLVM_DEBUG(DBGS() << "--subExpr: " << subExpressions.back() << "\n");
+    LDBG() << "--subExpr: " << subExpressions.back();
     remainingExp = currentBinExpr.getLHS();
   }
 
@@ -146,9 +146,7 @@ FailureOr<AffineApplyOp> mlir::affine::decompose(RewriterBase &rewriter,
   llvm::stable_sort(subExpressions, [&](AffineExpr e1, AffineExpr e2) {
     return getMaxSymbol(e1) < getMaxSymbol(e2);
   });
-  LLVM_DEBUG(
-      llvm::interleaveComma(subExpressions, DBGS() << "--sorted subexprs: ");
-      llvm::dbgs() << "\n");
+  LDBG() << "--sorted subexprs: " << llvm::interleaved(subExpressions);
 
   // 4. Merge sorted subExpressions iteratively, thus achieving reassociation.
   auto s0 = getAffineSymbolExpr(0, ctx);
@@ -162,7 +160,7 @@ FailureOr<AffineApplyOp> mlir::affine::decompose(RewriterBase &rewriter,
     Value tmp = createSubApply(rewriter, op, subExpressions[i]);
     current = AffineApplyOp::create(rewriter, op.getLoc(), binMap,
                                     ValueRange{current, tmp});
-    LLVM_DEBUG(DBGS() << "--reassociate into: " << current << "\n");
+    LDBG() << "--reassociate into: " << current;
   }
 
   // 5. Replace original op.
diff --git a/mlir/lib/Dialect/Affine/Transforms/SimplifyAffineMinMax.cpp b/mlir/lib/Dialect/Affine/Transforms/SimplifyAffineMinMax.cpp
index 8493b60..2521512 100644
--- a/mlir/lib/Dialect/Affine/Transforms/SimplifyAffineMinMax.cpp
+++ b/mlir/lib/Dialect/Affine/Transforms/SimplifyAffineMinMax.cpp
@@ -19,11 +19,10 @@
 #include "mlir/Interfaces/ValueBoundsOpInterface.h"
 #include "mlir/Transforms/GreedyPatternRewriteDriver.h"
 #include "llvm/ADT/IntEqClasses.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 #include "llvm/Support/InterleavedRange.h"
 
 #define DEBUG_TYPE "affine-min-max"
-#define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE << "]: ")
 
 using namespace mlir;
 using namespace mlir::affine;
@@ -39,7 +38,7 @@ static bool simplifyAffineMinMaxOp(RewriterBase &rewriter, AffineOp affineOp) {
   ValueRange operands = affineOp.getOperands();
   static constexpr bool isMin = std::is_same_v<AffineOp, AffineMinOp>;
 
-  LLVM_DEBUG({ DBGS() << "analyzing value: `" << affineOp << "`\n"; });
+  LDBG() << "analyzing value: `" << affineOp;
 
   // Create a `Variable` list with values corresponding to each of the results
   // in the affine affineMap.
@@ -48,12 +47,9 @@ static bool simplifyAffineMinMaxOp(RewriterBase &rewriter, AffineOp affineOp) {
       [&](unsigned i) {
         return Variable(affineMap.getSliceMap(i, 1), operands);
       });
-  LLVM_DEBUG({
-    DBGS() << "- constructed variables are: "
-           << llvm::interleaved_array(llvm::map_range(
-                  variables, [](const Variable &v) { return v.getMap(); }))
-           << "`\n";
-  });
+  LDBG() << "- constructed variables are: "
+         << llvm::interleaved_array(llvm::map_range(
+                variables, [](const Variable &v) { return v.getMap(); }));
 
   // Get the comparison operation.
   ComparisonOperator cmpOp =
@@ -72,10 +68,8 @@ static bool simplifyAffineMinMaxOp(RewriterBase &rewriter, AffineOp affineOp) {
     // Initialize the bound.
     Variable *bound = &v;
 
-    LLVM_DEBUG({
-      DBGS() << "- inspecting variable: #" << i << ", with map: `" << v.getMap()
-             << "`\n";
-    });
+    LDBG() << "- inspecting variable: #" << i << ", with map: `" << v.getMap()
+           << "`\n";
 
     // Check against the other variables.
     for (size_t j = i + 1; j < variables.size(); ++j) {
@@ -87,10 +81,8 @@ static bool simplifyAffineMinMaxOp(RewriterBase &rewriter, AffineOp affineOp) {
       // Get the bound of the equivalence class or itself.
       Variable *nv = bounds.lookup_or(jEqClass, &variables[j]);
 
-      LLVM_DEBUG({
-        DBGS() << "- comparing with variable: #" << jEqClass
-               << ", with map: " << nv->getMap() << "\n";
-      });
+      LDBG() << "- comparing with variable: #" << jEqClass
+             << ", with map: " << nv->getMap();
 
       // Compare the variables.
       FailureOr<bool> cmpResult =
@@ -98,18 +90,14 @@ static bool simplifyAffineMinMaxOp(RewriterBase &rewriter, AffineOp affineOp) {
 
       // The variables cannot be compared.
       if (failed(cmpResult)) {
-        LLVM_DEBUG({
-          DBGS() << "-- classes: #" << i << ", #" << jEqClass
-                 << " cannot be merged\n";
-        });
+        LDBG() << "-- classes: #" << i << ", #" << jEqClass
+               << " cannot be merged";
         continue;
       }
 
       // Join the equivalent classes and update the bound if necessary.
-      LLVM_DEBUG({
-        DBGS() << "-- merging classes: #" << i << ", #" << jEqClass
-               << ", is cmp(lhs, rhs): " << *cmpResult << "`\n";
-      });
+      LDBG() << "-- merging classes: #" << i << ", #" << jEqClass
+             << ", is cmp(lhs, rhs): " << *cmpResult << "`";
       if (*cmpResult) {
         boundedClasses.join(eqClass, jEqClass);
       } else {
@@ -124,8 +112,7 @@ static bool simplifyAffineMinMaxOp(RewriterBase &rewriter, AffineOp affineOp) {
 
   // Return if there's no simplification.
   if (bounds.size() >= affineMap.getNumResults()) {
-    LLVM_DEBUG(
-        { DBGS() << "- the affine operation couldn't get simplified\n"; });
+    LDBG() << "- the affine operation couldn't get simplified";
     return false;
   }
 
@@ -135,13 +122,11 @@ static bool simplifyAffineMinMaxOp(RewriterBase &rewriter, AffineOp affineOp) {
   for (auto [k, bound] : bounds)
     results.push_back(bound->getMap().getResult(0));
 
-  LLVM_DEBUG({
-    DBGS() << "- starting from map: " << affineMap << "\n";
-    DBGS() << "- creating new map with: \n";
-    DBGS() << "--- dims: " << affineMap.getNumDims() << "\n";
-    DBGS() << "--- syms: " << affineMap.getNumSymbols() << "\n";
-    DBGS() << "--- res: " << llvm::interleaved_array(results) << "\n";
-  });
+  LDBG() << "- starting from map: " << affineMap;
+  LDBG() << "- creating new map with:";
+  LDBG() << "--- dims: " << affineMap.getNumDims();
+  LDBG() << "--- syms: " << affineMap.getNumSymbols();
+  LDBG() << "--- res: " << llvm::interleaved_array(results);
 
   affineMap =
       AffineMap::get(0, affineMap.getNumSymbols() + affineMap.getNumDims(),
@@ -149,7 +134,7 @@ static bool simplifyAffineMinMaxOp(RewriterBase &rewriter, AffineOp affineOp) {
 
   // Update the affine op.
   rewriter.modifyOpInPlace(affineOp, [&]() { affineOp.setMap(affineMap); });
-  LLVM_DEBUG({ DBGS() << "- simplified affine op: `" << affineOp << "`\n"; });
+  LDBG() << "- simplified affine op: `" << affineOp << "`";
   return true;
 }
 
diff --git a/mlir/lib/Dialect/Arith/Transforms/CMakeLists.txt b/mlir/lib/Dialect/Arith/Transforms/CMakeLists.txt
index f96bda6..93682a9 100644
--- a/mlir/lib/Dialect/Arith/Transforms/CMakeLists.txt
+++ b/mlir/lib/Dialect/Arith/Transforms/CMakeLists.txt
@@ -27,7 +27,7 @@ add_mlir_dialect_library(MLIRArithTransforms
   MLIRInferIntRangeInterface
   MLIRIR
   MLIRMemRefDialect
-  MLIRMeshDialect
+  MLIRShardDialect
   MLIRPass
   MLIRShardingInterface
   MLIRTensorDialect
diff --git a/mlir/lib/Dialect/Arith/Transforms/ShardingInterfaceImpl.cpp b/mlir/lib/Dialect/Arith/Transforms/ShardingInterfaceImpl.cpp
index dd6efe6..3e34246 100644
--- a/mlir/lib/Dialect/Arith/Transforms/ShardingInterfaceImpl.cpp
+++ b/mlir/lib/Dialect/Arith/Transforms/ShardingInterfaceImpl.cpp
@@ -6,22 +6,22 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "mlir/Dialect/Mesh/Interfaces/ShardingInterfaceImpl.h"
+#include "mlir/Dialect/Shard/Interfaces/ShardingInterfaceImpl.h"
 #include "mlir/Dialect/Arith/IR/Arith.h"
 #include "mlir/Dialect/Arith/Transforms/ShardingInterfaceImpl.h"
-#include "mlir/Dialect/Mesh/Interfaces/ShardingInterface.h"
+#include "mlir/Dialect/Shard/Interfaces/ShardingInterface.h"
 #include "mlir/IR/DialectRegistry.h"
 
 using namespace mlir;
 using namespace mlir::arith;
-using namespace mlir::mesh;
+using namespace mlir::shard;
 
 namespace {
 
 // Sharding of arith.constant
 // RankedTensor constants can be sharded like any other tensor.
 //   %cst = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
-//   %sharding = mesh.sharding @mesh4x4 split_axes = [[0]] : !mesh.sharding
+//   %sharding = shard.sharding @grid4x4 split_axes = [[0]] : !shard.sharding
 // Scalar constants are always replicated and need no sharding annotation.
 
 struct ConstantShardingInterface
@@ -48,8 +48,8 @@ struct ConstantShardingInterface
   // Otherwise mirror result sharding if it is a tensor constant.
   // Otherwise return replication option.
   FailureOr<ShardingOption>
-  getShardingOption(Operation *op, ArrayRef<MeshSharding> operandShardings,
-                    ArrayRef<MeshSharding> resultShardings) const {
+  getShardingOption(Operation *op, ArrayRef<Sharding> operandShardings,
+                    ArrayRef<Sharding> resultShardings) const {
     assert(resultShardings.size() == 1 &&
            "Expecting exactly one result sharding for arith.constant");
     auto resultSharding = resultShardings[0];
@@ -61,17 +61,17 @@ struct ConstantShardingInterface
       for (auto [i, axes] : llvm::enumerate(resultSharding.getSplitAxes())) {
         axesArray[i].append(axes.asArrayRef().begin(), axes.asArrayRef().end());
       }
-      return ShardingOption(axesArray, resultSharding.getMeshAttr());
+      return ShardingOption(axesArray, resultSharding.getGridAttr());
     }
-    return ShardingOption({}, resultSharding.getMeshAttr());
+    return ShardingOption({}, resultSharding.getGridAttr());
   }
 
-  LogicalResult spmdize(Operation *op, ArrayRef<Value> spmdizedOperands,
-                        ArrayRef<MeshSharding> operandShardings,
-                        ArrayRef<MeshSharding> resultShardings,
-                        IRMapping &spmdizationMap,
-                        SymbolTableCollection &symbolTable,
-                        OpBuilder &builder) const {
+  LogicalResult partition(Operation *op, ArrayRef<Value> partitiondOperands,
+                          ArrayRef<Sharding> operandShardings,
+                          ArrayRef<Sharding> resultShardings,
+                          IRMapping &partitionMap,
+                          SymbolTableCollection &symbolTable,
+                          OpBuilder &builder) const {
     auto cOp = cast<ConstantOp>(op);
     if (auto value = dyn_cast<DenseIntOrFPElementsAttr>(cOp.getValue())) {
       if (!value.isSplat() || !resultShardings[0]) {
@@ -80,15 +80,15 @@ struct ConstantShardingInterface
       }
       auto sharding = resultShardings[0];
       auto newType = cast<RankedTensorType>(shardType(
-          cOp.getType(), getMesh(op, sharding.getMeshAttr(), symbolTable),
+          cOp.getType(), getGrid(op, sharding.getGridAttr(), symbolTable),
           sharding));
       auto newValue = value.resizeSplat(newType);
       auto newOp = ConstantOp::create(builder, op->getLoc(), newType, newValue);
-      spmdizationMap.map(op->getResult(0), newOp.getResult());
-      spmdizationMap.map(op, newOp.getOperation());
+      partitionMap.map(op->getResult(0), newOp.getResult());
+      partitionMap.map(op, newOp.getOperation());
     } else {
       // `clone` will populate the mapping of old to new results.
-      (void)builder.clone(*op, spmdizationMap);
+      (void)builder.clone(*op, partitionMap);
     }
     return success();
   }
diff --git a/mlir/lib/Dialect/ArmNeon/Transforms/LowerContractToNeonPatterns.cpp b/mlir/lib/Dialect/ArmNeon/Transforms/LowerContractToNeonPatterns.cpp
index 45b896d..1aa8064 100644
--- a/mlir/lib/Dialect/ArmNeon/Transforms/LowerContractToNeonPatterns.cpp
+++ b/mlir/lib/Dialect/ArmNeon/Transforms/LowerContractToNeonPatterns.cpp
@@ -145,8 +145,8 @@ protected:
       return rewriter.createOrFold<arm_neon::UsmmlaOp>(loc, acc.getType(), acc,
                                                        lhs, rhs);
     case MMLA::Bfloat:
-      return rewriter.create<arm_neon::BfmmlaOp>(loc, acc.getType(), acc, lhs,
-                                                 rhs);
+      return arm_neon::BfmmlaOp::create(rewriter, loc, acc.getType(), acc, lhs,
+                                        rhs);
     case MMLA::Nop:
       llvm_unreachable("Uninitialized operation type");
     }
@@ -226,8 +226,9 @@ public:
 
     // Initial accumulator for the final result. This is the un-tiled result if
     // tiling is done.
-    Value result = rewriter.create<arith::ConstantOp>(
-        loc, op.getResultType(), rewriter.getZeroAttr(op.getResultType()));
+    Value result =
+        arith::ConstantOp::create(rewriter, loc, op.getResultType(),
+                                  rewriter.getZeroAttr(op.getResultType()));
 
     SmallVector<int64_t, 3> loopOrder = {0, 1};
     if (iterationBounds.size() == 3)
@@ -263,8 +264,9 @@ public:
       if (dimM == 1) {
         auto expandRowVector = [&](Value tiledOperand,
                                    VectorType expandedTypeType) {
-          auto emptyOperand = rewriter.create<arith::ConstantOp>(
-              loc, expandedTypeType, rewriter.getZeroAttr(expandedTypeType));
+          auto emptyOperand =
+              arith::ConstantOp::create(rewriter, loc, expandedTypeType,
+                                        rewriter.getZeroAttr(expandedTypeType));
           SmallVector<int64_t> offsets(
               cast<ShapedType>(emptyOperand.getType()).getRank(), 0);
           SmallVector<int64_t> strides(
@@ -280,8 +282,8 @@ public:
       // using the instruction for unsigned by signed multiplication with
       // reversed operands.
       if (swapOperands)
-        tiledAcc = rewriter.create<vector::TransposeOp>(
-            loc, tiledAcc, ArrayRef<int64_t>({1, 0}));
+        tiledAcc = vector::TransposeOp::create(rewriter, loc, tiledAcc,
+                                               ArrayRef<int64_t>({1, 0}));
 
       // Collapse tiled operands to 1D vectors required by the ArmNeon ops
       auto collapsedLhs = rewriter.createOrFold<vector::ShapeCastOp>(
@@ -309,8 +311,8 @@ public:
       // Because of the reversed operands the result is obtained transposed.
       // Transpose it back,
       if (swapOperands)
-        tiledRes = rewriter.create<vector::TransposeOp>(
-            loc, tiledRes, ArrayRef<int64_t>({1, 0}));
+        tiledRes = vector::TransposeOp::create(rewriter, loc, tiledRes,
+                                               ArrayRef<int64_t>({1, 0}));
 
       // With vecmat, only one row of tiled ACC can be inserted into the final
       // result
diff --git a/mlir/lib/Dialect/ArmSVE/Transforms/LowerContractToSVEPatterns.cpp b/mlir/lib/Dialect/ArmSVE/Transforms/LowerContractToSVEPatterns.cpp
index fcfeb9c..35b0bd1 100644
--- a/mlir/lib/Dialect/ArmSVE/Transforms/LowerContractToSVEPatterns.cpp
+++ b/mlir/lib/Dialect/ArmSVE/Transforms/LowerContractToSVEPatterns.cpp
@@ -214,13 +214,13 @@ Value VectorContractRewriter::createMMLA(PatternRewriter &rewriter,
 
   switch (mmlaOp) {
   case MMLA::SignedInt:
-    return rewriter.create<arm_sve::SmmlaOp>(loc, resTy, acc, lhs, rhs);
+    return arm_sve::SmmlaOp::create(rewriter, loc, resTy, acc, lhs, rhs);
   case MMLA::UnsignedInt:
-    return rewriter.create<arm_sve::UmmlaOp>(loc, resTy, acc, lhs, rhs);
+    return arm_sve::UmmlaOp::create(rewriter, loc, resTy, acc, lhs, rhs);
   case MMLA::MixedInt:
-    return rewriter.create<arm_sve::UsmmlaOp>(loc, resTy, acc, lhs, rhs);
+    return arm_sve::UsmmlaOp::create(rewriter, loc, resTy, acc, lhs, rhs);
   case MMLA::Bfloat:
-    return rewriter.create<arm_sve::BfmmlaOp>(loc, resTy, acc, lhs, rhs);
+    return arm_sve::BfmmlaOp::create(rewriter, loc, resTy, acc, lhs, rhs);
   default:
     llvm_unreachable("Uninitialized operation kind");
   }
@@ -316,62 +316,63 @@ Value VectorContractRewriter::lower(vector::ContractionOp op,
   for (int64_t i = 0; i < M; i += 2) {
     // Extract two consecutive rows of the LHS tile.
     auto r0 =
-        rewriter.create<vector::ExtractOp>(loc, lhs, ArrayRef<int64_t>{i});
+        vector::ExtractOp::create(rewriter, loc, lhs, ArrayRef<int64_t>{i});
     auto r1 =
-        rewriter.create<vector::ExtractOp>(loc, lhs, ArrayRef<int64_t>{i + 1});
+        vector::ExtractOp::create(rewriter, loc, lhs, ArrayRef<int64_t>{i + 1});
     // Concatenate to obtain a 2 x K x <input-type> flattened sub-tile.
     SmallVector<int64_t> shuffleIdx(2 * K);
     std::iota(shuffleIdx.begin(), shuffleIdx.end(), 0);
-    auto t = rewriter.create<vector::ShuffleOp>(loc, r0, r1, shuffleIdx);
+    auto t = vector::ShuffleOp::create(rewriter, loc, r0, r1, shuffleIdx);
     // Turn it into a scalable vector.
-    auto s = rewriter.create<vector::ScalableInsertOp>(
-        loc, t, rewriter.create<ub::PoisonOp>(loc, flatLhsType), 0);
+    auto s = vector::ScalableInsertOp::create(
+        rewriter, loc, t, ub::PoisonOp::create(rewriter, loc, flatLhsType), 0);
     // Replicate the sub-tile VSCALE times to fill the entire vector.
-    auto r = rewriter.create<arm_sve::DupQLaneOp>(loc, s, 0);
+    auto r = arm_sve::DupQLaneOp::create(rewriter, loc, s, 0);
     lhsTile.push_back(r);
   }
 
   // "Flatten" the RHS tile from <[N]xK> to <[N*K]>.
-  auto rhs = rewriter.create<vector::ShapeCastOp>(this->rhs.getLoc(),
-                                                  flatRhsTileType, this->rhs);
+  auto rhs = vector::ShapeCastOp::create(rewriter, this->rhs.getLoc(),
+                                         flatRhsTileType, this->rhs);
 
   // Extract the RHS sub-tiles with logical shape <Kx[2]>.
   SmallVector<Value> rhsTile;
   for (int64_t j = 0; j < N; j += 2)
-    rhsTile.push_back(rewriter.create<vector::ScalableExtractOp>(
-        loc, flatRhsType, rhs, j * K));
+    rhsTile.push_back(vector::ScalableExtractOp::create(
+        rewriter, loc, flatRhsType, rhs, j * K));
 
   // Extract and pack the ACC sub-tiles.
   SmallVector<Value> accTile;
   for (int64_t i = 0; i < M; i += 2) {
     // Extract two consecutive rows of the accumulator tile.
-    auto r0 = rewriter.create<vector::ExtractOp>(loc, op.getAcc(),
-                                                 ArrayRef<int64_t>{i});
-    auto r1 = rewriter.create<vector::ExtractOp>(loc, op.getAcc(),
-                                                 ArrayRef<int64_t>{i + 1});
+    auto r0 = vector::ExtractOp::create(rewriter, loc, op.getAcc(),
+                                        ArrayRef<int64_t>{i});
+    auto r1 = vector::ExtractOp::create(rewriter, loc, op.getAcc(),
+                                        ArrayRef<int64_t>{i + 1});
     Value accTileVec;
     if (swapOperands) {
       // We are performing the operation with swapped LHS and RHS we need to
       // transpose each individual 2x2 tile of the accumulator and (later) the
       // final result.
-      accTileVec = rewriter.create<vector::InterleaveOp>(loc, r0, r1);
+      accTileVec = vector::InterleaveOp::create(rewriter, loc, r0, r1);
     } else {
       // Bitcast accumulator rows to double-width integer elements, so
       // subsequent interleave/deinterleave work on pairs of elements.
-      auto r0I64 = rewriter.create<vector::BitCastOp>(loc, accRow64Ty, r0);
-      auto r1I64 = rewriter.create<vector::BitCastOp>(loc, accRow64Ty, r1);
+      auto r0I64 = vector::BitCastOp::create(rewriter, loc, accRow64Ty, r0);
+      auto r1I64 = vector::BitCastOp::create(rewriter, loc, accRow64Ty, r1);
 
       // Interleave the rows, effectively flattening each 2x2 tile into 4
       // consecutive elements.
-      auto intrI64 = rewriter.create<vector::InterleaveOp>(loc, r0I64, r1I64);
+      auto intrI64 = vector::InterleaveOp::create(rewriter, loc, r0I64, r1I64);
 
       // Bitcast back to original element type.
-      accTileVec = rewriter.create<vector::BitCastOp>(loc, accRowX2Ty, intrI64);
+      accTileVec =
+          vector::BitCastOp::create(rewriter, loc, accRowX2Ty, intrI64);
     }
     // Extract ACC sub-tiles.
     for (int64_t j = 0; j < N; j += 2)
-      accTile.push_back(rewriter.create<vector::ScalableExtractOp>(
-          loc, flatAccType, accTileVec, j * 2));
+      accTile.push_back(vector::ScalableExtractOp::create(
+          rewriter, loc, flatAccType, accTileVec, j * 2));
   }
 
   // Emit sub-tile matrix multiplications.
@@ -384,13 +385,13 @@ Value VectorContractRewriter::lower(vector::ContractionOp op,
     }
 
   // Unpack the OUT sub-tiles and insert into the result.
-  Value result = rewriter.create<ub::PoisonOp>(loc, op.getResultType());
+  Value result = ub::PoisonOp::create(rewriter, loc, op.getResultType());
   for (int64_t i = 0; i < M / 2; ++i) {
     // Collect a number of sub-tiles in a row.
-    Value row = rewriter.create<ub::PoisonOp>(loc, accRowX2Ty);
+    Value row = ub::PoisonOp::create(rewriter, loc, accRowX2Ty);
     for (int64_t j = 0; j < N / 2; ++j)
-      row = rewriter.create<vector::ScalableInsertOp>(
-          loc, outTile[i * N / 2 + j], row, j * 4);
+      row = vector::ScalableInsertOp::create(
+          rewriter, loc, outTile[i * N / 2 + j], row, j * 4);
 
     // Unpack the row to obtain two rows of the output. If we have the out
     // sub-tiles transposed we obtain two consecutive output rows by
@@ -398,22 +399,22 @@ Value VectorContractRewriter::lower(vector::ContractionOp op,
     // Otherwise, the interleave is by pairs.
     Value out0, out1;
     if (swapOperands) {
-      auto tmp = rewriter.create<vector::DeinterleaveOp>(loc, row);
+      auto tmp = vector::DeinterleaveOp::create(rewriter, loc, row);
       out0 = tmp.getRes1();
       out1 = tmp.getRes2();
     } else {
       // Deinterleave by pairs.
-      auto row64 = rewriter.create<vector::BitCastOp>(loc, accRowX264Ty, row);
-      auto deintr64 = rewriter.create<vector::DeinterleaveOp>(loc, row64);
+      auto row64 = vector::BitCastOp::create(rewriter, loc, accRowX264Ty, row);
+      auto deintr64 = vector::DeinterleaveOp::create(rewriter, loc, row64);
 
       // Bitcast back into original element type and insert into the result.
-      out0 =
-          rewriter.create<vector::BitCastOp>(loc, accRowTy, deintr64.getRes1());
-      out1 =
-          rewriter.create<vector::BitCastOp>(loc, accRowTy, deintr64.getRes2());
+      out0 = vector::BitCastOp::create(rewriter, loc, accRowTy,
+                                       deintr64.getRes1());
+      out1 = vector::BitCastOp::create(rewriter, loc, accRowTy,
+                                       deintr64.getRes2());
     }
-    result = rewriter.create<vector::InsertOp>(loc, out0, result, i * 2);
-    result = rewriter.create<vector::InsertOp>(loc, out1, result, i * 2 + 1);
+    result = vector::InsertOp::create(rewriter, loc, out0, result, i * 2);
+    result = vector::InsertOp::create(rewriter, loc, out1, result, i * 2 + 1);
   }
 
   return result;
diff --git a/mlir/lib/Dialect/Bufferization/IR/BufferizableOpInterface.cpp b/mlir/lib/Dialect/Bufferization/IR/BufferizableOpInterface.cpp
index 825f63e..f7b0b87 100644
--- a/mlir/lib/Dialect/Bufferization/IR/BufferizableOpInterface.cpp
+++ b/mlir/lib/Dialect/Bufferization/IR/BufferizableOpInterface.cpp
@@ -18,7 +18,6 @@
 #include "mlir/IR/Value.h"
 #include "mlir/Interfaces/ControlFlowInterfaces.h"
 #include "llvm/ADT/ScopeExit.h"
-#include "llvm/Support/Debug.h"
 
 //===----------------------------------------------------------------------===//
 // BufferizableOpInterface
@@ -35,8 +34,6 @@ namespace bufferization {
 MLIR_DEFINE_EXPLICIT_TYPE_ID(mlir::bufferization::AnalysisState)
 
 #define DEBUG_TYPE "bufferizable-op-interface"
-#define DBGS() (llvm::dbgs() << '[' << DEBUG_TYPE << "] ")
-#define LDBG(X) LLVM_DEBUG(DBGS() << (X))
 
 using namespace mlir;
 using namespace bufferization;
@@ -691,8 +688,8 @@ FailureOr<Value> bufferization::getBuffer(RewriterBase &rewriter, Value value,
   if (failed(bufferType))
     return failure();
   ensureToBufferOpIsValid(value, *bufferType);
-  return rewriter
-      .create<bufferization::ToBufferOp>(value.getLoc(), *bufferType, value)
+  return bufferization::ToBufferOp::create(rewriter, value.getLoc(),
+                                           *bufferType, value)
       .getResult();
 }
 
@@ -775,9 +772,8 @@ FailureOr<Value> BufferizationOptions::createAlloc(OpBuilder &b, Location loc,
 
   // Default bufferallocation via AllocOp.
   if (bufferAlignment != 0)
-    return b
-        .create<memref::AllocOp>(loc, type, dynShape,
-                                 b.getI64IntegerAttr(bufferAlignment))
+    return memref::AllocOp::create(b, loc, type, dynShape,
+                                   b.getI64IntegerAttr(bufferAlignment))
         .getResult();
   return memref::AllocOp::create(b, loc, type, dynShape).getResult();
 }
diff --git a/mlir/lib/Dialect/Bufferization/Transforms/LowerDeallocations.cpp b/mlir/lib/Dialect/Bufferization/Transforms/LowerDeallocations.cpp
index f0d65b0..e9ad13f 100644
--- a/mlir/lib/Dialect/Bufferization/Transforms/LowerDeallocations.cpp
+++ b/mlir/lib/Dialect/Bufferization/Transforms/LowerDeallocations.cpp
@@ -482,10 +482,10 @@ func::FuncOp mlir::bufferization::buildDeallocationLibraryFunction(
 
         // Build the first for loop that computes aliasing with retained
         // memrefs.
-        Value noRetainAlias =
-            builder
-                .create<scf::ForOp>(
-                    loc, c0, toRetainSize, c1, trueValue,
+        Value
+            noRetainAlias =
+                scf::ForOp::create(
+                    builder, loc, c0, toRetainSize, c1, trueValue,
                     [&](OpBuilder &builder, Location loc, Value i,
                         ValueRange iterArgs) {
                       Value retainValue = memref::LoadOp::create(
@@ -512,14 +512,14 @@ func::FuncOp mlir::bufferization::buildDeallocationLibraryFunction(
                           builder, loc, iterArgs[0], doesntAlias);
                       scf::YieldOp::create(builder, loc, yieldValue);
                     })
-                .getResult(0);
+                    .getResult(0);
 
         // Build the second for loop that adds aliasing with previously
         // deallocated memrefs.
-        Value noAlias =
-            builder
-                .create<scf::ForOp>(
-                    loc, c0, outerIter, c1, noRetainAlias,
+        Value
+            noAlias =
+                scf::ForOp::create(
+                    builder, loc, c0, outerIter, c1, noRetainAlias,
                     [&](OpBuilder &builder, Location loc, Value i,
                         ValueRange iterArgs) {
                       Value prevDeallocValue = memref::LoadOp::create(
@@ -531,7 +531,7 @@ func::FuncOp mlir::bufferization::buildDeallocationLibraryFunction(
                           builder, loc, iterArgs[0], doesntAlias);
                       scf::YieldOp::create(builder, loc, yieldValue);
                     })
-                .getResult(0);
+                    .getResult(0);
 
         Value shouldDealoc = arith::AndIOp::create(builder, loc, noAlias, cond);
         memref::StoreOp::create(builder, loc, shouldDealoc, deallocCondsMemref,
diff --git a/mlir/lib/Dialect/Bufferization/Transforms/OptimizeAllocationLiveness.cpp b/mlir/lib/Dialect/Bufferization/Transforms/OptimizeAllocationLiveness.cpp
index 605a487..b8ddee6 100644
--- a/mlir/lib/Dialect/Bufferization/Transforms/OptimizeAllocationLiveness.cpp
+++ b/mlir/lib/Dialect/Bufferization/Transforms/OptimizeAllocationLiveness.cpp
@@ -18,11 +18,9 @@
 #include "mlir/IR/Operation.h"
 #include "mlir/IR/Value.h"
 #include "mlir/Interfaces/SideEffectInterfaces.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 
 #define DEBUG_TYPE "optimize-allocation-liveness"
-#define DBGS() (llvm::dbgs() << '[' << DEBUG_TYPE << "] ")
-#define LDBG(X) LLVM_DEBUG(DBGS() << X << "\n")
 
 namespace mlir {
 namespace bufferization {
@@ -65,8 +63,8 @@ Operation *findUserWithFreeSideEffect(Value value) {
       for (const auto &effect : effects) {
         if (isa<MemoryEffects::Free>(effect.getEffect())) {
           if (freeOpUser) {
-            LDBG("Multiple users with free effect found: " << *freeOpUser
-                                                           << " and " << *user);
+            LDBG() << "Multiple users with free effect found: " << *freeOpUser
+                   << " and " << *user;
             return nullptr;
           }
           freeOpUser = user;
@@ -121,7 +119,7 @@ public:
         return WalkResult::advance();
 
       auto allocOp = memEffectOp;
-      LDBG("Checking alloc op: " << allocOp);
+      LDBG() << "Checking alloc op: " << allocOp;
 
       SmallVector<OpResult> allocationResults = collectAllocations(allocOp);
       // Multiple allocations from a single op are not considered here yet.
@@ -129,7 +127,7 @@ public:
         return WalkResult::advance();
 
       OpResult allocResult = allocationResults[0];
-      LDBG("On allocation result: " << allocResult);
+      LDBG() << "On allocation result: " << allocResult;
 
       auto *deallocOp = findUserWithFreeSideEffect(allocResult);
       if (!deallocOp || (deallocOp->getBlock() != allocOp->getBlock())) {
@@ -159,12 +157,12 @@ public:
       if (lastUser == nullptr) {
         return WalkResult::advance();
       }
-      LDBG("Last user found: " << *lastUser);
+      LDBG() << "Last user found: " << *lastUser;
       assert(lastUser->getBlock() == allocOp->getBlock());
       assert(lastUser->getBlock() == deallocOp->getBlock());
       // Move the dealloc op after the last user.
       deallocOp->moveAfter(lastUser);
-      LDBG("Moved dealloc op after: " << *lastUser);
+      LDBG() << "Moved dealloc op after: " << *lastUser;
 
       return WalkResult::advance();
     });
diff --git a/mlir/lib/Dialect/Bufferization/Transforms/OwnershipBasedBufferDeallocation.cpp b/mlir/lib/Dialect/Bufferization/Transforms/OwnershipBasedBufferDeallocation.cpp
index 64c178d..725fa24 100644
--- a/mlir/lib/Dialect/Bufferization/Transforms/OwnershipBasedBufferDeallocation.cpp
+++ b/mlir/lib/Dialect/Bufferization/Transforms/OwnershipBasedBufferDeallocation.cpp
@@ -750,17 +750,16 @@ Value BufferDeallocation::materializeMemrefWithGuaranteedOwnership(
 
   // Insert a runtime check and only clone if we still don't have ownership at
   // runtime.
-  Value maybeClone = builder
-                         .create<scf::IfOp>(
-                             memref.getLoc(), condition,
-                             [&](OpBuilder &builder, Location loc) {
-                               scf::YieldOp::create(builder, loc, newMemref);
-                             },
-                             [&](OpBuilder &builder, Location loc) {
-                               Value clone = bufferization::CloneOp::create(
-                                   builder, loc, newMemref);
-                               scf::YieldOp::create(builder, loc, clone);
-                             })
+  Value maybeClone = scf::IfOp::create(
+                         builder, memref.getLoc(), condition,
+                         [&](OpBuilder &builder, Location loc) {
+                           scf::YieldOp::create(builder, loc, newMemref);
+                         },
+                         [&](OpBuilder &builder, Location loc) {
+                           Value clone = bufferization::CloneOp::create(
+                               builder, loc, newMemref);
+                           scf::YieldOp::create(builder, loc, clone);
+                         })
                          .getResult(0);
   Value trueVal = buildBoolValue(builder, memref.getLoc(), true);
   state.updateOwnership(maybeClone, trueVal);
diff --git a/mlir/lib/Dialect/CMakeLists.txt b/mlir/lib/Dialect/CMakeLists.txt
index 3cc52ebc..053ee95 100644
--- a/mlir/lib/Dialect/CMakeLists.txt
+++ b/mlir/lib/Dialect/CMakeLists.txt
@@ -19,7 +19,7 @@ add_subdirectory(Linalg)
 add_subdirectory(LLVMIR)
 add_subdirectory(Math)
 add_subdirectory(MemRef)
-add_subdirectory(Mesh)
+add_subdirectory(Shard)
 add_subdirectory(MLProgram)
 add_subdirectory(MPI)
 add_subdirectory(NVGPU)
diff --git a/mlir/lib/Dialect/EmitC/Transforms/WrapFuncInClass.cpp b/mlir/lib/Dialect/EmitC/Transforms/WrapFuncInClass.cpp
index 612e809..fa05ad8 100644
--- a/mlir/lib/Dialect/EmitC/Transforms/WrapFuncInClass.cpp
+++ b/mlir/lib/Dialect/EmitC/Transforms/WrapFuncInClass.cpp
@@ -31,7 +31,7 @@ struct WrapFuncInClassPass
     Operation *rootOp = getOperation();
 
     RewritePatternSet patterns(&getContext());
-    populateFuncPatterns(patterns, namedAttribute);
+    populateFuncPatterns(patterns);
 
     walkAndApplyPatterns(rootOp, std::move(patterns));
   }
@@ -43,8 +43,8 @@ struct WrapFuncInClassPass
 
 class WrapFuncInClass : public OpRewritePattern<emitc::FuncOp> {
 public:
-  WrapFuncInClass(MLIRContext *context, StringRef attrName)
-      : OpRewritePattern<emitc::FuncOp>(context), attributeName(attrName) {}
+  WrapFuncInClass(MLIRContext *context)
+      : OpRewritePattern<emitc::FuncOp>(context) {}
 
   LogicalResult matchAndRewrite(emitc::FuncOp funcOp,
                                 PatternRewriter &rewriter) const override {
@@ -101,12 +101,8 @@ public:
     rewriter.replaceOp(funcOp, newClassOp);
     return success();
   }
-
-private:
-  StringRef attributeName;
 };
 
-void mlir::emitc::populateFuncPatterns(RewritePatternSet &patterns,
-                                       StringRef namedAttribute) {
-  patterns.add<WrapFuncInClass>(patterns.getContext(), namedAttribute);
+void mlir::emitc::populateFuncPatterns(RewritePatternSet &patterns) {
+  patterns.add<WrapFuncInClass>(patterns.getContext());
 }
diff --git a/mlir/lib/Dialect/Func/Extensions/AllExtensions.cpp b/mlir/lib/Dialect/Func/Extensions/AllExtensions.cpp
index eb6b59b..1b18ef2 100644
--- a/mlir/lib/Dialect/Func/Extensions/AllExtensions.cpp
+++ b/mlir/lib/Dialect/Func/Extensions/AllExtensions.cpp
@@ -8,7 +8,7 @@
 
 #include "mlir/Dialect/Func/Extensions/AllExtensions.h"
 #include "mlir/Dialect/Func/Extensions/InlinerExtension.h"
-#include "mlir/Dialect/Func/Extensions/MeshShardingExtensions.h"
+#include "mlir/Dialect/Func/Extensions/ShardingExtensions.h"
 
 using namespace mlir;
 
diff --git a/mlir/lib/Dialect/Func/Extensions/CMakeLists.txt b/mlir/lib/Dialect/Func/Extensions/CMakeLists.txt
index 47363f4..87ef51e 100644
--- a/mlir/lib/Dialect/Func/Extensions/CMakeLists.txt
+++ b/mlir/lib/Dialect/Func/Extensions/CMakeLists.txt
@@ -1,7 +1,7 @@
 set(LLVM_OPTIONAL_SOURCES
   AllExtensions.cpp
   InlinerExtension.cpp
-  MeshShardingExtensions.cpp
+  ShardingExtensions.cpp
   )
 
 add_mlir_extension_library(MLIRFuncInlinerExtension
@@ -17,8 +17,8 @@ add_mlir_extension_library(MLIRFuncInlinerExtension
   MLIRFuncDialect
   )
 
-add_mlir_extension_library(MLIRFuncMeshShardingExtensions
-  MeshShardingExtensions.cpp
+add_mlir_extension_library(MLIRFuncShardingExtensions
+  ShardingExtensions.cpp
 
   ADDITIONAL_HEADER_DIRS
   ${MLIR_MAIN_INCLUDE_DIR}/mlir/Dialect/Func/Extensions
@@ -38,5 +38,5 @@ add_mlir_extension_library(MLIRFuncAllExtensions
 
   LINK_LIBS PUBLIC
   MLIRFuncInlinerExtension
-  MLIRFuncMeshShardingExtensions
+  MLIRFuncShardingExtensions
   )
diff --git a/mlir/lib/Dialect/Func/Extensions/MeshShardingExtensions.cpp b/mlir/lib/Dialect/Func/Extensions/ShardingExtensions.cpp
index da508cc..dfd1348 100644
--- a/mlir/lib/Dialect/Func/Extensions/MeshShardingExtensions.cpp
+++ b/mlir/lib/Dialect/Func/Extensions/ShardingExtensions.cpp
@@ -1,4 +1,4 @@
-//===- MeshShardingExtensions.cpp - ---------------------------------------===//
+//===- ShardingExtensions.cpp - ---------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -6,9 +6,9 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "mlir/Dialect/Func/Extensions/MeshShardingExtensions.h"
+#include "mlir/Dialect/Func/Extensions/ShardingExtensions.h"
 #include "mlir/Dialect/Func/IR/FuncOps.h"
-#include "mlir/Dialect/Mesh/Interfaces/ShardingInterfaceImpl.h"
+#include "mlir/Dialect/Shard/Interfaces/ShardingInterfaceImpl.h"
 #include "mlir/IR/MLIRContext.h"
 
 namespace mlir::func {
@@ -16,7 +16,7 @@ namespace mlir::func {
 void registerShardingInterfaceExternalModels(DialectRegistry &registry) {
   registry.addExtension(+[](MLIRContext *ctx, FuncDialect *dialect) {
     ReturnOp::attachInterface<
-        mesh::IndependentParallelIteratorDomainShardingInterface<ReturnOp>>(
+        shard::IndependentParallelIteratorDomainShardingInterface<ReturnOp>>(
         *ctx);
   });
 }
diff --git a/mlir/lib/Dialect/GPU/TransformOps/GPUTransformOps.cpp b/mlir/lib/Dialect/GPU/TransformOps/GPUTransformOps.cpp
index 1d8279c..21cb2f6 100644
--- a/mlir/lib/Dialect/GPU/TransformOps/GPUTransformOps.cpp
+++ b/mlir/lib/Dialect/GPU/TransformOps/GPUTransformOps.cpp
@@ -39,7 +39,7 @@
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/TypeSwitch.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/InterleavedRange.h"
 #include "llvm/Support/LogicalResult.h"
@@ -51,11 +51,6 @@ using namespace mlir::transform;
 using namespace mlir::transform::gpu;
 
 #define DEBUG_TYPE "gpu-transforms"
-#define DEBUG_TYPE_ALIAS "gpu-transforms-alias"
-
-#define DBGS() (llvm::dbgs() << '[' << DEBUG_TYPE << "] ")
-#define LDBG(X) LLVM_DEBUG(DBGS() << X << "\n")
-#define DBGS_ALIAS() (llvm::dbgs() << '[' << DEBUG_TYPE_ALIAS << "] ")
 
 //===----------------------------------------------------------------------===//
 // Apply...ConversionPatternsOp
@@ -471,7 +466,7 @@ static DiagnosedSilenceableFailure rewriteOneForallCommonImpl(
     RewriterBase &rewriter, std::optional<TransformOpInterface> transformOp,
     scf::ForallOp forallOp, ArrayRef<int64_t> availableMappingSizes,
     ForallRewriteResult &result, const GpuIdBuilder &gpuIdBuilder) {
-  LDBG("--start rewriteOneForallCommonImpl");
+  LDBG() << "--start rewriteOneForallCommonImpl";
 
   // Step 1. Complete the mapping to a full mapping (with 1s) if necessary.
   auto numParallelIterations =
@@ -506,14 +501,14 @@ static DiagnosedSilenceableFailure rewriteOneForallCommonImpl(
     // Otherwise, we have a new insertion without a size -> use size 1.
     tmpMappingSizes.push_back(1);
   }
-  LDBG("----tmpMappingSizes extracted from scf.forall op: "
-       << llvm::interleaved(tmpMappingSizes));
+  LDBG() << "----tmpMappingSizes extracted from scf.forall op: "
+         << llvm::interleaved(tmpMappingSizes);
 
   // Step 2. sort the values by the corresponding DeviceMappingAttrInterface.
   SmallVector<int64_t> forallMappingSizes = getValuesSortedByKey(
       forallMappingAttrs.getArrayRef(), tmpMappingSizes, comparator);
-  LDBG("----forallMappingSizes: " << llvm::interleaved(forallMappingSizes));
-  LDBG("----forallMappingAttrs: " << llvm::interleaved(forallMappingAttrs));
+  LDBG() << "----forallMappingSizes: " << llvm::interleaved(forallMappingSizes);
+  LDBG() << "----forallMappingAttrs: " << llvm::interleaved(forallMappingAttrs);
 
   // Step 3. Generate the mappingIdOps using the provided generator.
   Location loc = forallOp.getLoc();
@@ -522,24 +517,24 @@ static DiagnosedSilenceableFailure rewriteOneForallCommonImpl(
   SmallVector<int64_t> originalBasis(availableMappingSizes);
   bool originalBasisWasProvided = !originalBasis.empty();
   if (!originalBasisWasProvided) {
-    LDBG("----originalBasis was not provided, deriving it and there will be no "
-         "predication");
+    LDBG() << "----originalBasis was not provided, deriving it and there will "
+              "be no "
+              "predication";
     originalBasis = forallMappingSizes;
     while (originalBasis.size() < 3)
       originalBasis.push_back(1);
   } else {
-    LDBG("----originalBasis was provided, using it, there will be predication");
+    LDBG() << "----originalBasis was provided, using it, there will be "
+              "predication";
   }
-  LLVM_DEBUG(
-      llvm::interleaveComma(originalBasis, DBGS() << "------originalBasis: ");
-      llvm::dbgs() << "\n");
+  LDBG() << "------originalBasis: " << llvm::interleaved(originalBasis);
 
   IdBuilderResult builderResult =
       gpuIdBuilder.idBuilder(rewriter, loc, forallMappingSizes, originalBasis);
   if (!builderResult.errorMsg.empty())
     return definiteFailureHelper(transformOp, forallOp, builderResult.errorMsg);
 
-  LLVM_DEBUG(DBGS() << builderResult);
+  LDBG() << builderResult;
 
   // Step 4. Map the induction variables to the mappingIdOps, this may involve
   // a permutation.
@@ -550,7 +545,7 @@ static DiagnosedSilenceableFailure rewriteOneForallCommonImpl(
            forallMappingAttrs.getArrayRef().take_front(forallOp.getRank()))) {
     auto mappingAttr = cast<DeviceMappingAttrInterface>(dim);
     Value peIdOp = mappingIdOps[mappingAttr.getRelativeIndex()];
-    LDBG("----map: " << iv << " to " << peIdOp);
+    LDBG() << "----map: " << iv << " to " << peIdOp;
     bvm.map(iv, peIdOp);
   }
 
@@ -596,9 +591,9 @@ static DiagnosedSilenceableFailure rewriteOneForallCommonImpl(
   // Step 8. Erase old op.
   rewriter.eraseOp(forallOp);
 
-  LDBG("----result forallMappingSizes: "
-       << llvm::interleaved(forallMappingSizes));
-  LDBG("----result mappingIdOps: " << llvm::interleaved(mappingIdOps));
+  LDBG() << "----result forallMappingSizes: "
+         << llvm::interleaved(forallMappingSizes);
+  LDBG() << "----result mappingIdOps: " << llvm::interleaved(mappingIdOps);
 
   result = ForallRewriteResult{forallMappingSizes, mappingIdOps};
   return DiagnosedSilenceableFailure::success();
@@ -612,7 +607,7 @@ DiagnosedSilenceableFailure mlir::transform::gpu::mapForallToBlocksImpl(
     RewriterBase &rewriter, TransformOpInterface transformOp,
     scf::ForallOp forallOp, SmallVectorImpl<int64_t> &gridDims,
     const GpuIdBuilder &gpuIdBuilder) {
-  LDBG("Start mapForallToBlocksImpl");
+  LDBG() << "Start mapForallToBlocksImpl";
 
   {
     // GPU-specific verifications. There is no better place to anchor
@@ -893,7 +888,7 @@ DiagnosedSilenceableFailure mlir::transform::gpu::mapNestedForallToThreadsImpl(
     RewriterBase &rewriter, std::optional<TransformOpInterface> transformOp,
     Operation *target, ArrayRef<int64_t> blockDims, int64_t warpSize,
     bool syncAfterDistribute) {
-  LDBG("Start mapNestedForallToThreadsImpl");
+  LDBG() << "Start mapNestedForallToThreadsImpl";
   if (blockDims.size() != 3) {
     return definiteFailureHelper(transformOp, target,
                                  "requires size-3 thread mapping");
diff --git a/mlir/lib/Dialect/GPU/TransformOps/Utils.cpp b/mlir/lib/Dialect/GPU/TransformOps/Utils.cpp
index 2fba09b..05bd917 100644
--- a/mlir/lib/Dialect/GPU/TransformOps/Utils.cpp
+++ b/mlir/lib/Dialect/GPU/TransformOps/Utils.cpp
@@ -27,7 +27,8 @@
 #include "mlir/Support/LLVM.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallVector.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
+#include "llvm/Support/InterleavedRange.h"
 
 using namespace mlir;
 using namespace mlir::gpu;
@@ -36,10 +37,6 @@ using namespace mlir::transform::gpu;
 
 #define DEBUG_TYPE "gpu-transforms"
 
-#define DBGS() (llvm::dbgs() << '[' << DEBUG_TYPE << "] ")
-#define LDBG(X) LLVM_DEBUG(DBGS() << X << "\n")
-#define DBGS_ALIAS() (llvm::dbgs() << '[' << DEBUG_TYPE_ALIAS << "] ")
-
 /// Build predicates to filter execution by only the activeIds. Along each
 /// dimension, 3 cases appear:
 ///   1. activeMappingSize > availableMappingSize: this is an unsupported case
@@ -54,15 +51,9 @@ buildPredicates(RewriterBase &rewriter, Location loc, ArrayRef<Value> activeIds,
                 ArrayRef<int64_t> activeMappingSizes,
                 ArrayRef<int64_t> availableMappingSizes,
                 std::string &errorMsg) {
-  // clang-format off
-  LLVM_DEBUG(
-    llvm::interleaveComma(
-      activeMappingSizes, DBGS() << "----activeMappingSizes: ");
-    DBGS() << "\n";
-    llvm::interleaveComma(
-      availableMappingSizes, DBGS() << "----availableMappingSizes: ");
-    DBGS() << "\n";);
-  // clang-format on
+  LDBG() << "----activeMappingSizes: " << llvm::interleaved(activeMappingSizes);
+  LDBG() << "----availableMappingSizes: "
+         << llvm::interleaved(availableMappingSizes);
 
   SmallVector<Value> predicateOps;
   for (auto [activeId, activeMappingSize, availableMappingSize] :
@@ -88,10 +79,8 @@ buildPredicates(RewriterBase &rewriter, Location loc, ArrayRef<Value> activeIds,
 template <typename ThreadOrBlockIdOp>
 static Value buildLinearId(RewriterBase &rewriter, Location loc,
                            ArrayRef<OpFoldResult> originalBasisOfr) {
-  LLVM_DEBUG(llvm::interleaveComma(
-                 originalBasisOfr,
-                 DBGS() << "----buildLinearId with originalBasisOfr:  ");
-             llvm::dbgs() << "\n");
+  LDBG() << "----buildLinearId with originalBasisOfr:  "
+         << llvm::interleaved(originalBasisOfr);
   assert(originalBasisOfr.size() == 3 && "expected 3 sizes");
   IndexType indexType = rewriter.getIndexType();
   AffineExpr tx, ty, tz, bdx, bdy;
@@ -157,7 +146,7 @@ commonLinearIdBuilderFn(int64_t multiplicity = 1,
           mask.createLogicalLinearMappingId(rewriter, scaledLinearIdI64);
       scaledLinearId = arith::IndexCastUIOp::create(
           rewriter, loc, rewriter.getIndexType(), logicalLinearIdI64);
-      LDBG("------adjusting linearId with mask: " << scaledLinearId);
+      LDBG() << "------adjusting linearId with mask: " << scaledLinearId;
     }
 
     // 3. Compute remapped indices.
@@ -179,7 +168,7 @@ commonLinearIdBuilderFn(int64_t multiplicity = 1,
     if (mask) {
       Value isActiveIdPredicate =
           mask.createIsActiveIdPredicate(rewriter, scaledLinearIdI64);
-      LDBG("------adjusting predicate with mask: " << isActiveIdPredicate);
+      LDBG() << "------adjusting predicate with mask: " << isActiveIdPredicate;
       predicateOps.push_back(isActiveIdPredicate);
     } else {
       // 4.b. Otherwise, handle predicates using physicalLinearId.
diff --git a/mlir/lib/Dialect/GPU/Transforms/ShuffleRewriter.cpp b/mlir/lib/Dialect/GPU/Transforms/ShuffleRewriter.cpp
index d88f4d5..8e05436 100644
--- a/mlir/lib/Dialect/GPU/Transforms/ShuffleRewriter.cpp
+++ b/mlir/lib/Dialect/GPU/Transforms/ShuffleRewriter.cpp
@@ -60,14 +60,12 @@ struct GpuShuffleRewriter : public OpRewritePattern<gpu::ShuffleOp> {
 
     // Shuffle the values.
     ValueRange loRes =
-        rewriter
-            .create<gpu::ShuffleOp>(op.getLoc(), lo, op.getOffset(),
-                                    op.getWidth(), op.getMode())
+        gpu::ShuffleOp::create(rewriter, op.getLoc(), lo, op.getOffset(),
+                               op.getWidth(), op.getMode())
             .getResults();
     ValueRange hiRes =
-        rewriter
-            .create<gpu::ShuffleOp>(op.getLoc(), hi, op.getOffset(),
-                                    op.getWidth(), op.getMode())
+        gpu::ShuffleOp::create(rewriter, op.getLoc(), hi, op.getOffset(),
+                               op.getWidth(), op.getMode())
             .getResults();
 
     // Convert lo back to i64.
diff --git a/mlir/lib/Dialect/GPU/Transforms/SubgroupReduceLowering.cpp b/mlir/lib/Dialect/GPU/Transforms/SubgroupReduceLowering.cpp
index b9e2dd5..b45fdf3 100644
--- a/mlir/lib/Dialect/GPU/Transforms/SubgroupReduceLowering.cpp
+++ b/mlir/lib/Dialect/GPU/Transforms/SubgroupReduceLowering.cpp
@@ -197,10 +197,9 @@ Value createSubgroupShuffleReduction(OpBuilder &builder, Location loc,
   // Parallel reduction using butterfly shuffles.
   for (unsigned i = ci.clusterStride; i < ci.clusterStride * ci.clusterSize;
        i <<= 1) {
-    Value shuffled = builder
-                         .create<gpu::ShuffleOp>(loc, packFn(laneVal), i,
-                                                 /*width=*/ci.subgroupSize,
-                                                 /*mode=*/gpu::ShuffleMode::XOR)
+    Value shuffled = gpu::ShuffleOp::create(builder, loc, packFn(laneVal), i,
+                                            /*width=*/ci.subgroupSize,
+                                            /*mode=*/gpu::ShuffleMode::XOR)
                          .getShuffleResult();
     laneVal = vector::makeArithReduction(builder, loc,
                                          gpu::convertReductionKind(mode),
diff --git a/mlir/lib/Dialect/LLVMIR/CMakeLists.txt b/mlir/lib/Dialect/LLVMIR/CMakeLists.txt
index d987b72..ff55f17 100644
--- a/mlir/lib/Dialect/LLVMIR/CMakeLists.txt
+++ b/mlir/lib/Dialect/LLVMIR/CMakeLists.txt
@@ -21,10 +21,7 @@ add_mlir_dialect_library(MLIRLLVMDialect
   intrinsics_gen
 
   LINK_COMPONENTS
-  AsmParser
   BinaryFormat
-  BitReader
-  BitWriter
   Core
 
   LINK_LIBS PUBLIC
diff --git a/mlir/lib/Dialect/LLVMIR/IR/LLVMDialect.cpp b/mlir/lib/Dialect/LLVMIR/IR/LLVMDialect.cpp
index d42ce96..422039f 100644
--- a/mlir/lib/Dialect/LLVMIR/IR/LLVMDialect.cpp
+++ b/mlir/lib/Dialect/LLVMIR/IR/LLVMDialect.cpp
@@ -26,8 +26,7 @@
 
 #include "llvm/ADT/APFloat.h"
 #include "llvm/ADT/TypeSwitch.h"
-#include "llvm/IR/Function.h"
-#include "llvm/IR/Type.h"
+#include "llvm/IR/DataLayout.h"
 #include "llvm/Support/Error.h"
 
 #include <numeric>
@@ -4064,28 +4063,9 @@ void LLVM::AssumeOp::build(OpBuilder &builder, OperationState &state,
 }
 
 void LLVM::AssumeOp::build(OpBuilder &builder, OperationState &state,
-                           Value cond,
-                           ArrayRef<llvm::OperandBundleDefT<Value>> opBundles) {
-  SmallVector<ValueRange> opBundleOperands;
-  SmallVector<Attribute> opBundleTags;
-  opBundleOperands.reserve(opBundles.size());
-  opBundleTags.reserve(opBundles.size());
-
-  for (const llvm::OperandBundleDefT<Value> &bundle : opBundles) {
-    opBundleOperands.emplace_back(bundle.inputs());
-    opBundleTags.push_back(
-        StringAttr::get(builder.getContext(), bundle.getTag()));
-  }
-
-  auto opBundleTagsAttr = ArrayAttr::get(builder.getContext(), opBundleTags);
-  return build(builder, state, cond, opBundleOperands, opBundleTagsAttr);
-}
-
-void LLVM::AssumeOp::build(OpBuilder &builder, OperationState &state,
                            Value cond, llvm::StringRef tag, ValueRange args) {
-  llvm::OperandBundleDefT<Value> opBundle(
-      tag.str(), SmallVector<Value>(args.begin(), args.end()));
-  return build(builder, state, cond, opBundle);
+  return build(builder, state, cond, ArrayRef<ValueRange>(args),
+               builder.getStrArrayAttr(tag));
 }
 
 void LLVM::AssumeOp::build(OpBuilder &builder, OperationState &state,
diff --git a/mlir/lib/Dialect/LLVMIR/IR/NVVMDialect.cpp b/mlir/lib/Dialect/LLVMIR/IR/NVVMDialect.cpp
index 6e29b12..cffe310 100644
--- a/mlir/lib/Dialect/LLVMIR/IR/NVVMDialect.cpp
+++ b/mlir/lib/Dialect/LLVMIR/IR/NVVMDialect.cpp
@@ -30,15 +30,8 @@
 #include "mlir/IR/Types.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/TypeSwitch.h"
-#include "llvm/AsmParser/Parser.h"
-#include "llvm/IR/Attributes.h"
-#include "llvm/IR/Function.h"
-#include "llvm/IR/IRBuilder.h"
-#include "llvm/IR/IntrinsicsNVPTX.h"
-#include "llvm/IR/Type.h"
 #include "llvm/Support/Casting.h"
 #include "llvm/Support/FormatVariadic.h"
-#include "llvm/Support/SourceMgr.h"
 #include "llvm/Support/raw_ostream.h"
 #include <cassert>
 #include <optional>
diff --git a/mlir/lib/Dialect/LLVMIR/IR/ROCDLDialect.cpp b/mlir/lib/Dialect/LLVMIR/IR/ROCDLDialect.cpp
index 1a9ccf5..17371ec 100644
--- a/mlir/lib/Dialect/LLVMIR/IR/ROCDLDialect.cpp
+++ b/mlir/lib/Dialect/LLVMIR/IR/ROCDLDialect.cpp
@@ -24,7 +24,6 @@
 #include "mlir/IR/MLIRContext.h"
 #include "mlir/IR/Operation.h"
 #include "llvm/ADT/TypeSwitch.h"
-#include "llvm/IR/Type.h"
 
 using namespace mlir;
 using namespace ROCDL;
diff --git a/mlir/lib/Dialect/LLVMIR/IR/VCIXDialect.cpp b/mlir/lib/Dialect/LLVMIR/IR/VCIXDialect.cpp
index bd9d3528..1d4a0af 100644
--- a/mlir/lib/Dialect/LLVMIR/IR/VCIXDialect.cpp
+++ b/mlir/lib/Dialect/LLVMIR/IR/VCIXDialect.cpp
@@ -20,11 +20,6 @@
 #include "mlir/IR/MLIRContext.h"
 #include "mlir/IR/Operation.h"
 #include "llvm/ADT/TypeSwitch.h"
-#include "llvm/AsmParser/Parser.h"
-#include "llvm/IR/Attributes.h"
-#include "llvm/IR/Function.h"
-#include "llvm/IR/Type.h"
-#include "llvm/Support/SourceMgr.h"
 
 using namespace mlir;
 using namespace vcix;
diff --git a/mlir/lib/Dialect/Linalg/IR/LinalgDialect.cpp b/mlir/lib/Dialect/Linalg/IR/LinalgDialect.cpp
index b6e168e..7f6ecab 100644
--- a/mlir/lib/Dialect/Linalg/IR/LinalgDialect.cpp
+++ b/mlir/lib/Dialect/Linalg/IR/LinalgDialect.cpp
@@ -15,7 +15,7 @@
 #include "mlir/Dialect/Linalg/IR/Linalg.h"
 #include "mlir/Dialect/Math/IR/Math.h"
 #include "mlir/Dialect/MemRef/IR/MemRef.h"
-#include "mlir/Dialect/Mesh/Interfaces/ShardingInterface.h"
+#include "mlir/Dialect/Shard/Interfaces/ShardingInterface.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/IR/DialectImplementation.h"
 #include "mlir/Interfaces/SubsetOpInterface.h"
@@ -119,8 +119,8 @@ void mlir::linalg::LinalgDialect::initialize() {
 
   addInterfaces<LinalgInlinerInterface>();
 
-  declarePromisedInterface<mesh::ShardingInterface, GenericOp>();
-  declarePromisedInterfaces<mesh::ShardingInterface,
+  declarePromisedInterface<shard::ShardingInterface, GenericOp>();
+  declarePromisedInterfaces<shard::ShardingInterface,
 #define GET_OP_LIST
 #include "mlir/Dialect/Linalg/IR/LinalgStructuredOps.cpp.inc"
                             >();
diff --git a/mlir/lib/Dialect/Linalg/IR/LinalgOps.cpp b/mlir/lib/Dialect/Linalg/IR/LinalgOps.cpp
index 4fee81a..27b6617 100644
--- a/mlir/lib/Dialect/Linalg/IR/LinalgOps.cpp
+++ b/mlir/lib/Dialect/Linalg/IR/LinalgOps.cpp
@@ -791,9 +791,8 @@ struct FoldFillWithPad final : public OpRewritePattern<tensor::PadOp> {
         tensor::EmptyOp::create(rewriter, padOp.getLoc(), reifiedShape.front(),
                                 padOp.getResultType().getElementType());
     Value replacement =
-        rewriter
-            .create<FillOp>(fillOp.getLoc(), ValueRange{padValue},
-                            ValueRange{emptyTensor})
+        FillOp::create(rewriter, fillOp.getLoc(), ValueRange{padValue},
+                       ValueRange{emptyTensor})
             .getResult(0);
     if (replacement.getType() != padOp.getResultType()) {
       replacement = tensor::CastOp::create(rewriter, fillOp.getLoc(),
@@ -2154,9 +2153,8 @@ struct SwapTransposeWithBroadcast : OpRewritePattern<linalg::TransposeOp> {
 
     // Create broadcast(transpose(input)).
     Value transposeResult =
-        rewriter
-            .create<TransposeOp>(loc, broadcastOp.getInput(), transposeInit,
-                                 resultPerms)
+        TransposeOp::create(rewriter, loc, broadcastOp.getInput(),
+                            transposeInit, resultPerms)
             ->getResult(0);
     rewriter.replaceOpWithNewOp<BroadcastOp>(
         transposeOp, transposeResult, transposeOp.getInit(), resultDimensions);
diff --git a/mlir/lib/Dialect/Linalg/TransformOps/GPUHeuristics.cpp b/mlir/lib/Dialect/Linalg/TransformOps/GPUHeuristics.cpp
index ce1b1b9..5c8c2de 100644
--- a/mlir/lib/Dialect/Linalg/TransformOps/GPUHeuristics.cpp
+++ b/mlir/lib/Dialect/Linalg/TransformOps/GPUHeuristics.cpp
@@ -12,6 +12,7 @@
 #include "llvm/ADT/ArrayRef.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 #include "llvm/Support/InterleavedRange.h"
 #include "llvm/Support/MathExtras.h"
 #include "llvm/Support/raw_ostream.h"
@@ -21,8 +22,6 @@
 using namespace mlir;
 
 #define DEBUG_TYPE "linalg-transforms"
-#define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE "]: ")
-#define LDBG(X) LLVM_DEBUG(DBGS() << X << "\n")
 
 static Attribute linearId0(MLIRContext *ctx) {
   return gpu::GPUThreadMappingAttr::get(ctx, gpu::MappingId::LinearDim0);
@@ -43,9 +42,8 @@ transform::gpu::CopyMappingInfo::CopyMappingInfo(MLIRContext *ctx,
   assert(!copySizes.empty() && copySizes.size() <= 3 &&
          "only 1,2,3-D copies are supported for now");
 
-  LDBG("START CopyMappingInfo, favorPredication: " << favorPredication);
-  LLVM_DEBUG(DBGS() << "--copy shape: " << llvm::interleaved(copySizes)
-                    << "\n");
+  LDBG() << "START CopyMappingInfo, favorPredication: " << favorPredication;
+  LDBG() << "--copy shape: " << llvm::interleaved(copySizes);
 
   // Greedily find the largest vector size that can be used to copy the most
   // minor dimension: we are in the business of filling kMaxVectorLoadBitWidth
@@ -53,20 +51,19 @@ transform::gpu::CopyMappingInfo::CopyMappingInfo(MLIRContext *ctx,
   int64_t desiredVectorSize = CopyMappingInfo::maxContiguousElementsToTransfer(
       desiredBitAlignment, copySizes.back(), elementalBitwidth);
 
-  LDBG("--greedily determined vectorSize: "
-       << desiredVectorSize << " elements of " << elementalBitwidth
-       << "b each -> " << (desiredVectorSize * elementalBitwidth)
-       << "b total out of a max of " << kMaxVectorLoadBitWidth << "b");
+  LDBG() << "--greedily determined vectorSize: " << desiredVectorSize
+         << " elements of " << elementalBitwidth << "b each -> "
+         << (desiredVectorSize * elementalBitwidth)
+         << "b total out of a max of " << kMaxVectorLoadBitWidth << "b";
 
   status = inferNumThreads(totalNumThreads, copySizes, desiredVectorSize,
                            favorPredication);
   if (status == Status::Invalid)
     return;
 
-  LLVM_DEBUG(DBGS() << "--copy: " << llvm::interleaved(copySizes) << "\n"
-                    << "--numThreads: " << llvm::interleaved(this->numThreads)
-                    << "\n"
-                    << "--vectorSize: " << this->vectorSize << "\n");
+  LDBG() << "--copy: " << llvm::interleaved(copySizes) << "\n"
+         << "--numThreads: " << llvm::interleaved(this->numThreads) << "\n"
+         << "--vectorSize: " << this->vectorSize;
   assert(this->numThreads.size() == copySizes.size() &&
          "compute copy mapping expected same number of threads and copy sizes");
 
@@ -84,7 +81,7 @@ transform::gpu::CopyMappingInfo::CopyMappingInfo(MLIRContext *ctx,
   this->threadMapping =
       llvm::to_vector(ArrayRef(allThreadMappings)
                           .take_back(this->smallestBoundingTileSizes.size()));
-  LLVM_DEBUG(this->print(DBGS()); llvm::dbgs() << "\n");
+  LDBG() << *this;
 }
 
 int64_t transform::gpu::CopyMappingInfo::maxContiguousElementsToTransfer(
@@ -140,7 +137,7 @@ static SmallVector<int64_t> maximizeNumThreads(ArrayRef<int64_t> sizes,
          "currentIndex out of bounds");
   std::string indent(2 * currentIndex, '-');
   if (static_cast<size_t>(currentIndex) == sizes.size() - 1) {
-    LDBG(indent << "mandated globalBest: " << sizes[currentIndex]);
+    LDBG() << indent << "mandated globalBest: " << sizes[currentIndex];
     return SmallVector<int64_t>{sizes[currentIndex]};
   }
 
@@ -149,16 +146,16 @@ static SmallVector<int64_t> maximizeNumThreads(ArrayRef<int64_t> sizes,
   SmallVector<int64_t> factors = getFactors(s);
   SmallVector<int64_t> localThreadsPerDim;
   localThreadsPerDim.reserve(sizes.size());
-  LDBG(indent << "maximizeNumThreads in " << s
-              << " with limit: " << maxNumThreads);
+  LDBG() << indent << "maximizeNumThreads in " << s
+         << " with limit: " << maxNumThreads;
   for (auto factor : factors) {
     auto nestedThreadsPerDim =
         maximizeNumThreads(sizes, currentIndex + 1, maxNumThreads / factor);
     int64_t localBest = factor * product(nestedThreadsPerDim);
     if (localBest > best && localBest <= maxNumThreads) {
-      LDBG(indent << "new localBest: " << localBest);
-      LDBG(indent << "nestedThreadsPerDim: "
-                  << llvm::interleaved(nestedThreadsPerDim));
+      LDBG() << indent << "new localBest: " << localBest;
+      LDBG() << indent << "nestedThreadsPerDim: "
+             << llvm::interleaved(nestedThreadsPerDim);
       localThreadsPerDim.clear();
       localThreadsPerDim.push_back(factor);
       llvm::append_range(localThreadsPerDim, nestedThreadsPerDim);
@@ -166,8 +163,8 @@ static SmallVector<int64_t> maximizeNumThreads(ArrayRef<int64_t> sizes,
     }
   }
 
-  LDBG(indent << "found globalBest: " << best);
-  LDBG(indent << "numThreads: " << llvm::interleaved(localThreadsPerDim));
+  LDBG() << indent << "found globalBest: " << best;
+  LDBG() << indent << "numThreads: " << llvm::interleaved(localThreadsPerDim);
   return localThreadsPerDim;
 }
 
@@ -192,8 +189,8 @@ transform::gpu::CopyMappingInfo::inferNumThreads(int64_t totalNumThreads,
       if (status == Status::Success || status == Status::Invalid)
         return status;
 
-      LDBG("requires predication, try reducing vector size to "
-           << (localVectorSize / 2));
+      LDBG() << "requires predication, try reducing vector size to "
+             << (localVectorSize / 2);
     }
   }
 
@@ -210,8 +207,8 @@ transform::gpu::CopyMappingInfo::inferNumThreadsImpl(
   assert(sizes.back() % desiredVectorSize == 0 &&
          "most-minor size not divisible by actualVectorSize");
 
-  LDBG("inferNumThreadsImpl with totalNumThreads: "
-       << totalNumThreads << " and vectorSize: " << desiredVectorSize);
+  LDBG() << "inferNumThreadsImpl with totalNumThreads: " << totalNumThreads
+         << " and vectorSize: " << desiredVectorSize;
 
   // Scale the most minor size to account for the chosen vector size and
   // maximize the number of threads without exceeding the total number of
@@ -219,22 +216,22 @@ transform::gpu::CopyMappingInfo::inferNumThreadsImpl(
   SmallVector<int64_t> scaledSizes(sizes);
   scaledSizes.back() /= desiredVectorSize;
   if (scaledSizes.back() > totalNumThreads) {
-    LDBG("--Too few threads given the required vector size -> FAIL");
+    LDBG() << "--Too few threads given the required vector size -> FAIL";
     return Status::Invalid;
   }
   SmallVector<int64_t> inferredNumThreads =
       maximizeNumThreads(scaledSizes, 0, totalNumThreads);
 
-  LDBG("inferred numThreads: " << llvm::interleaved(inferredNumThreads));
-  LDBG("computed actualVectorSize: " << desiredVectorSize);
+  LDBG() << "inferred numThreads: " << llvm::interleaved(inferredNumThreads);
+  LDBG() << "computed actualVectorSize: " << desiredVectorSize;
 
   // Corner case: we cannot use more threads than available. If the dimension of
   // the copy is so bad it is because higher-level tiling did not do its job, we
   // do not try to recover from it here.
   int64_t totalNumThreadsUsed = product(inferredNumThreads);
-  LDBG("--totalNumThreadsUsed: " << totalNumThreadsUsed);
+  LDBG() << "--totalNumThreadsUsed: " << totalNumThreadsUsed;
   if (totalNumThreadsUsed == 0 || totalNumThreadsUsed > totalNumThreads) {
-    LDBG("--Too few threads given the required vector size -> FAIL");
+    LDBG() << "--Too few threads given the required vector size -> FAIL";
     return Status::Invalid;
   }
 
diff --git a/mlir/lib/Dialect/Linalg/TransformOps/LinalgMatchOps.cpp b/mlir/lib/Dialect/Linalg/TransformOps/LinalgMatchOps.cpp
index 2fe72a3..d4a3e5f 100644
--- a/mlir/lib/Dialect/Linalg/TransformOps/LinalgMatchOps.cpp
+++ b/mlir/lib/Dialect/Linalg/TransformOps/LinalgMatchOps.cpp
@@ -15,14 +15,13 @@
 #include "mlir/Dialect/Transform/IR/TransformTypes.h"
 #include "mlir/Dialect/Transform/Interfaces/MatchInterfaces.h"
 #include "mlir/IR/BuiltinAttributes.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 #include "llvm/Support/FormatVariadic.h"
 #include "llvm/Support/InterleavedRange.h"
 
 using namespace mlir;
 
 #define DEBUG_TYPE "linalg-transforms"
-#define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE "]: ")
 
 //===----------------------------------------------------------------------===//
 // StructuredMatchOp
@@ -39,7 +38,7 @@ DiagnosedSilenceableFailure transform::MatchStructuredOp::matchOperation(
       return emitSilenceableError() << "expected a Linalg op";
     }
     // If errors are suppressed, succeed and set all results to empty lists.
-    LLVM_DEBUG(DBGS() << "optional nested matcher expected a Linalg op");
+    LDBG() << "optional nested matcher expected a Linalg op";
     results.setRemainingToEmpty(cast<TransformOpInterface>(getOperation()));
     return DiagnosedSilenceableFailure::success();
   }
@@ -75,8 +74,7 @@ DiagnosedSilenceableFailure transform::MatchStructuredOp::matchOperation(
     // When they are defined in this block, we additionally check if we have
     // already applied the operation that defines them. If not, the
     // corresponding results will be set to empty lists.
-    LLVM_DEBUG(DBGS() << "optional nested matcher failed: " << diag.getMessage()
-                      << "\n");
+    LDBG() << "optional nested matcher failed: " << diag.getMessage();
     (void)diag.silence();
     SmallVector<OpOperand *> undefinedOperands;
     for (OpOperand &terminatorOperand :
diff --git a/mlir/lib/Dialect/Linalg/TransformOps/LinalgTransformOps.cpp b/mlir/lib/Dialect/Linalg/TransformOps/LinalgTransformOps.cpp
index 9f523e9d..bdfc8d0 100644
--- a/mlir/lib/Dialect/Linalg/TransformOps/LinalgTransformOps.cpp
+++ b/mlir/lib/Dialect/Linalg/TransformOps/LinalgTransformOps.cpp
@@ -40,7 +40,7 @@
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/ScopeExit.h"
 #include "llvm/ADT/TypeSwitch.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 #include "llvm/Support/LogicalResult.h"
 #include <type_traits>
 
@@ -49,9 +49,6 @@ using namespace mlir::linalg;
 using namespace mlir::transform;
 
 #define DEBUG_TYPE "linalg-transforms"
-#define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE "]: ")
-#define DBGSNL() (llvm::dbgs() << "\n")
-#define LDBG(X) LLVM_DEBUG(DBGS() << (X) << "\n")
 
 /// Attempts to apply the pattern specified as template argument to the given
 /// operation. The pattern is expected to have a `returningMatchAndRewrite`
@@ -773,7 +770,7 @@ static bool sameOrEquivalentIterArg(Value src, Value dst) {
 static std::tuple<SmallVector<Operation *>, Operation *>
 tileAndFuseFirstExtractUse(RewriterBase &rewriter, Diagnostic &diag,
                            Operation *producerOp, Operation *containingOp) {
-  LLVM_DEBUG(DBGS() << "Try to fuse a direct extract use\n");
+  LDBG() << "Try to fuse a direct extract use";
   auto tileableProducer = dyn_cast<TilingInterface>(producerOp);
   if (!tileableProducer) {
     diag.attachNote(producerOp->getLoc())
@@ -838,7 +835,7 @@ tileAndFuseFirstExtractUse(RewriterBase &rewriter, Diagnostic &diag,
   // Tile the producer.
   int64_t resultNumber =
       cast<OpResult>(sliceOpToTile.getSource()).getResultNumber();
-  LLVM_DEBUG(DBGS() << "resultNumber: " << resultNumber << "\n");
+  LDBG() << "resultNumber: " << resultNumber;
 
   SmallVector<OpFoldResult> offsets = sliceOpToTile.getMixedOffsets();
   SmallVector<OpFoldResult> sizes = sliceOpToTile.getMixedSizes();
@@ -855,7 +852,7 @@ tileAndFuseFirstExtractUse(RewriterBase &rewriter, Diagnostic &diag,
 
 #ifndef NDEBUG
   for (auto *tiledOp : tileAndFuseResult->tiledOps) {
-    LLVM_DEBUG(DBGS() << "tiledProducer: " << *tiledOp << "\n");
+    LDBG() << "tiledProducer: " << *tiledOp;
   }
 #endif
 
@@ -894,7 +891,7 @@ static SmallVector<Operation *>
 tileAndFuseFirstExtractUseThroughContainingOpBlockArgument(
     RewriterBase &rewriter, Diagnostic &diag, Operation *producerOp,
     Operation *containingOp) {
-  LLVM_DEBUG(DBGS() << "Try to fuse an extract use through block argument\n");
+  LDBG() << "Try to fuse an extract use through block argument";
 
   auto tileableProducer = dyn_cast<TilingInterface>(producerOp);
   if (!tileableProducer) {
@@ -947,7 +944,7 @@ tileAndFuseFirstExtractUseThroughContainingOpBlockArgument(
   // Replace the use in the tileableProducer before tiling: clone, replace and
   // then tile.
   int64_t resultNumber = cast<OpResult>(pUse->get()).getResultNumber();
-  LLVM_DEBUG(DBGS() << "resultNumber: " << resultNumber << "\n");
+  LDBG() << "resultNumber: " << resultNumber;
 
   // Gather destination tensors.
   SmallVector<Value> destinationTensors;
@@ -996,7 +993,7 @@ tileAndFuseFirstExtractUseThroughContainingOpBlockArgument(
 static Operation *cloneAndFuseFirstUse(RewriterBase &rewriter, Diagnostic &diag,
                                        Operation *producerOp,
                                        Operation *containingOp) {
-  LLVM_DEBUG(DBGS() << "Try to fuse an use by cloning\n");
+  LDBG() << "Try to fuse an use by cloning";
 
   // Gather all uses inside the containing op.
   SmallVector<OpOperand *> uses;
@@ -1030,7 +1027,7 @@ static Operation *cloneAndFuseFirstUse(RewriterBase &rewriter, Diagnostic &diag,
   assert(!isa<tensor::ParallelInsertSliceOp>(use->getOwner()) &&
          "Parallel insert slice is not a valid clone destination");
   unsigned resultNumber = cast<OpResult>(use->get()).getResultNumber();
-  LLVM_DEBUG(DBGS() << "resultNumber: " << resultNumber << "\n");
+  LDBG() << "resultNumber: " << resultNumber;
 
   OpBuilder::InsertionGuard guard(rewriter);
   rewriter.setInsertionPoint(use->getOwner());
@@ -1113,7 +1110,7 @@ transform::FuseIntoContainingOp::apply(transform::TransformRewriter &rewriter,
     auto [tiledOps, newContainingOp] =
         tileAndFuseFirstExtractUse(rewriter, diag, producerOp, containingOp);
     if (!tiledOps.empty()) {
-      LLVM_DEBUG(DBGS() << "\nFused a direct extract use\n" << *containingOp);
+      LDBG() << "\nFused a direct extract use\n" << *containingOp;
       fusedOps.append(tiledOps);
       if (newContainingOp) {
         // Update handles associated with the containing op so we don't need to
@@ -1139,8 +1136,8 @@ transform::FuseIntoContainingOp::apply(transform::TransformRewriter &rewriter,
         tileAndFuseFirstExtractUseThroughContainingOpBlockArgument(
             rewriter, diag, producerOp, containingOp);
     if (!tiledContainingOpOperand.empty()) {
-      LLVM_DEBUG(DBGS() << "\nFused an extract use through block argument\n"
-                        << *containingOp);
+      LDBG() << "\nFused an extract use through block argument\n"
+             << *containingOp;
       fusedOps.append(tiledContainingOpOperand);
       continue;
     }
@@ -1148,7 +1145,7 @@ transform::FuseIntoContainingOp::apply(transform::TransformRewriter &rewriter,
     Operation *cloned =
         cloneAndFuseFirstUse(rewriter, diag, producerOp, containingOp);
     if (cloned) {
-      LLVM_DEBUG(DBGS() << "\nFused an use by cloning\n" << *containingOp);
+      LDBG() << "\nFused an use by cloning\n" << *containingOp;
       fusedOps.push_back(cloned);
       continue;
     }
@@ -4136,9 +4133,8 @@ DiagnosedSilenceableFailure doit(RewriterBase &rewriter, OpTy target,
   Value extracted = tensor::ExtractSliceOp::create(
       rewriter, target.getLoc(), target.getDest(), target.getMixedOffsets(),
       target.getMixedSizes(), target.getMixedStrides());
-  Value copied = rewriter
-                     .create<linalg::CopyOp>(target.getLoc(),
-                                             target.getSource(), extracted)
+  Value copied = linalg::CopyOp::create(rewriter, target.getLoc(),
+                                        target.getSource(), extracted)
                      .getResult(0);
   // Reset the insertion point.
   rewriter.setInsertionPoint(target);
diff --git a/mlir/lib/Dialect/Linalg/Transforms/AllInterfaces.cpp b/mlir/lib/Dialect/Linalg/Transforms/AllInterfaces.cpp
index 281d9f2..ba94ad7 100644
--- a/mlir/lib/Dialect/Linalg/Transforms/AllInterfaces.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/AllInterfaces.cpp
@@ -10,14 +10,14 @@
 
 #include "mlir/Dialect/Linalg/IR/ValueBoundsOpInterfaceImpl.h"
 #include "mlir/Dialect/Linalg/Transforms/BufferizableOpInterfaceImpl.h"
-#include "mlir/Dialect/Linalg/Transforms/MeshShardingInterfaceImpl.h"
+#include "mlir/Dialect/Linalg/Transforms/ShardingInterfaceImpl.h"
 #include "mlir/Dialect/Linalg/Transforms/SubsetInsertionOpInterfaceImpl.h"
 #include "mlir/Dialect/Linalg/Transforms/TilingInterfaceImpl.h"
 
 void mlir::linalg::registerAllDialectInterfaceImplementations(
     DialectRegistry &registry) {
   registerBufferizableOpInterfaceExternalModels(registry);
-  registerMeshShardingInterfaceExternalModels(registry);
+  registerShardingInterfaceExternalModels(registry);
   registerSubsetOpInterfaceExternalModels(registry);
   registerTilingInterfaceExternalModels(registry);
   registerValueBoundsOpInterfaceExternalModels(registry);
diff --git a/mlir/lib/Dialect/Linalg/Transforms/CMakeLists.txt b/mlir/lib/Dialect/Linalg/Transforms/CMakeLists.txt
index 69e6fda..70f846e 100644
--- a/mlir/lib/Dialect/Linalg/Transforms/CMakeLists.txt
+++ b/mlir/lib/Dialect/Linalg/Transforms/CMakeLists.txt
@@ -24,7 +24,7 @@ add_mlir_dialect_library(MLIRLinalgTransforms
   Interchange.cpp
   Loops.cpp
   TransposeMatmul.cpp
-  MeshShardingInterfaceImpl.cpp
+  ShardingInterfaceImpl.cpp
   NamedOpConversions.cpp
   BlockPackMatmul.cpp
   PackAndUnpackPatterns.cpp
@@ -68,7 +68,7 @@ add_mlir_dialect_library(MLIRLinalgTransforms
   MLIRIR
   MLIRMemRefDialect
   MLIRMemRefTransforms
-  MLIRMeshTransforms
+  MLIRShardTransforms
   MLIRLinalgDialect
   MLIRLinalgUtils
   MLIRSCFDialect
diff --git a/mlir/lib/Dialect/Linalg/Transforms/DataLayoutPropagation.cpp b/mlir/lib/Dialect/Linalg/Transforms/DataLayoutPropagation.cpp
index 91a297f..0a9c176 100644
--- a/mlir/lib/Dialect/Linalg/Transforms/DataLayoutPropagation.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/DataLayoutPropagation.cpp
@@ -1143,10 +1143,9 @@ pushDownUnPackOpThroughGenericOp(RewriterBase &rewriter, GenericOp genericOp,
 
   // Insert an unPackOp right after the packed generic.
   Value unPackOpRes =
-      rewriter
-          .create<linalg::UnPackOp>(genericOp.getLoc(), newResult,
-                                    destPack.getSource(), innerDimsPos,
-                                    mixedTiles, outerDimsPerm)
+      linalg::UnPackOp::create(rewriter, genericOp.getLoc(), newResult,
+                               destPack.getSource(), innerDimsPos, mixedTiles,
+                               outerDimsPerm)
           .getResult();
 
   return std::make_tuple(newGenericOp, unPackOpRes);
diff --git a/mlir/lib/Dialect/Linalg/Transforms/DropUnitDims.cpp b/mlir/lib/Dialect/Linalg/Transforms/DropUnitDims.cpp
index 745a40db..7f9ba1b 100644
--- a/mlir/lib/Dialect/Linalg/Transforms/DropUnitDims.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/DropUnitDims.cpp
@@ -267,8 +267,8 @@ expandValue(RewriterBase &rewriter, Location loc, Value result, Value origDest,
   assert(rankReductionStrategy ==
              ControlDropUnitDims::RankReductionStrategy::ReassociativeReshape &&
          "unknown rank reduction strategy");
-  return rewriter
-      .create<tensor::ExpandShapeOp>(loc, origResultType, result, reassociation)
+  return tensor::ExpandShapeOp::create(rewriter, loc, origResultType, result,
+                                       reassociation)
       .getResult();
 }
 
diff --git a/mlir/lib/Dialect/Linalg/Transforms/ElementwiseOpFusion.cpp b/mlir/lib/Dialect/Linalg/Transforms/ElementwiseOpFusion.cpp
index 4a66b8b..3bd763e 100644
--- a/mlir/lib/Dialect/Linalg/Transforms/ElementwiseOpFusion.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/ElementwiseOpFusion.cpp
@@ -1572,12 +1572,12 @@ static Value getCollapsedOpOperand(Location loc, LinalgOp op,
 
   // Insert a reshape to collapse the dimensions.
   if (isa<MemRefType>(operand.getType())) {
-    return builder
-        .create<memref::CollapseShapeOp>(loc, operand, operandReassociation)
+    return memref::CollapseShapeOp::create(builder, loc, operand,
+                                           operandReassociation)
         .getResult();
   }
-  return builder
-      .create<tensor::CollapseShapeOp>(loc, operand, operandReassociation)
+  return tensor::CollapseShapeOp::create(builder, loc, operand,
+                                         operandReassociation)
       .getResult();
 }
 
diff --git a/mlir/lib/Dialect/Linalg/Transforms/PackAndUnpackPatterns.cpp b/mlir/lib/Dialect/Linalg/Transforms/PackAndUnpackPatterns.cpp
index a45a4e3..277e50b 100644
--- a/mlir/lib/Dialect/Linalg/Transforms/PackAndUnpackPatterns.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/PackAndUnpackPatterns.cpp
@@ -81,9 +81,8 @@ struct SimplifyPackToExpandShape : public OpRewritePattern<PackOp> {
                ArrayRef<ReassociationIndices> reassociation) const {
     if (operand.getType() == newOperandType)
       return operand;
-    return rewriter
-        .create<tensor::ExpandShapeOp>(loc, newOperandType, operand,
-                                       reassociation)
+    return tensor::ExpandShapeOp::create(rewriter, loc, newOperandType, operand,
+                                         reassociation)
         .getResult();
   }
 
diff --git a/mlir/lib/Dialect/Linalg/Transforms/Padding.cpp b/mlir/lib/Dialect/Linalg/Transforms/Padding.cpp
index b5c5aea..dd84379 100644
--- a/mlir/lib/Dialect/Linalg/Transforms/Padding.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/Padding.cpp
@@ -333,17 +333,16 @@ linalg::rewriteAsPaddedOp(RewriterBase &rewriter, LinalgOp opToPad,
   for (auto it :
        llvm::zip(paddedSubtensorResults, opToPad.getDpsInitsMutable())) {
     if (options.copyBackOp == LinalgPaddingOptions::CopyBackOp::LinalgCopy) {
-      replacements.push_back(rewriter
-                                 .create<linalg::CopyOp>(loc, std::get<0>(it),
-                                                         std::get<1>(it).get())
+      replacements.push_back(linalg::CopyOp::create(rewriter, loc,
+                                                    std::get<0>(it),
+                                                    std::get<1>(it).get())
                                  .getResult(0));
     } else if (options.copyBackOp ==
                LinalgPaddingOptions::CopyBackOp::
                    BufferizationMaterializeInDestination) {
       replacements.push_back(
-          rewriter
-              .create<bufferization::MaterializeInDestinationOp>(
-                  loc, std::get<0>(it), std::get<1>(it).get())
+          bufferization::MaterializeInDestinationOp::create(
+              rewriter, loc, std::get<0>(it), std::get<1>(it).get())
               ->getResult(0));
     } else {
       llvm_unreachable("unsupported copy back op");
diff --git a/mlir/lib/Dialect/Linalg/Transforms/MeshShardingInterfaceImpl.cpp b/mlir/lib/Dialect/Linalg/Transforms/ShardingInterfaceImpl.cpp
index 83d12e3..f277c5f 100644
--- a/mlir/lib/Dialect/Linalg/Transforms/MeshShardingInterfaceImpl.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/ShardingInterfaceImpl.cpp
@@ -1,4 +1,4 @@
-//===- MeshShardingInterfaceImpl.cpp --------------------------------------===//
+//===- ShardingInterfaceImpl.cpp --------------------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -6,18 +6,18 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "mlir/Dialect/Linalg/Transforms/MeshShardingInterfaceImpl.h"
+#include "mlir/Dialect/Linalg/Transforms/ShardingInterfaceImpl.h"
 
 #include "mlir/Analysis/SliceAnalysis.h"
 #include "mlir/Dialect/Affine/IR/AffineOps.h"
 #include "mlir/Dialect/Arith/IR/Arith.h"
 #include "mlir/Dialect/Linalg/IR/Linalg.h"
 #include "mlir/Dialect/Linalg/IR/LinalgInterfaces.h"
-#include "mlir/Dialect/Mesh/IR/MeshOps.h"
-#include "mlir/Dialect/Mesh/Interfaces/ShardingInterface.h"
-#include "mlir/Dialect/Mesh/Interfaces/ShardingInterfaceImpl.h"
-#include "mlir/Dialect/Mesh/Transforms/Transforms.h"
 #include "mlir/Dialect/SCF/IR/SCF.h"
+#include "mlir/Dialect/Shard/IR/ShardOps.h"
+#include "mlir/Dialect/Shard/Interfaces/ShardingInterface.h"
+#include "mlir/Dialect/Shard/Interfaces/ShardingInterfaceImpl.h"
+#include "mlir/Dialect/Shard/Transforms/Transforms.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/Dialect/Utils/StructuredOpsUtils.h"
 #include "mlir/IR/AffineExpr.h"
@@ -36,13 +36,13 @@
 
 namespace mlir::linalg {
 
-using MeshAxis = mesh::MeshAxis;
-using ReductionKind = mesh::ReductionKind;
-using MeshSharding = mesh::MeshSharding;
-using ShardingArray = mesh::ShardingArray;
-using MeshOp = mesh::MeshOp;
+using GridAxis = shard::GridAxis;
+using ReductionKind = shard::ReductionKind;
+using Sharding = shard::Sharding;
+using ShardingArray = shard::ShardingArray;
+using GridOp = shard::GridOp;
 
-// Returns the corresponding mesh reduction kind for the given arith op.
+// Returns the corresponding grid reduction kind for the given arith op.
 static ReductionKind getReductionKind(Operation *op) {
   return llvm::TypeSwitch<Operation *, ReductionKind>(op)
       // Floating-point operations.
@@ -97,18 +97,18 @@ static ReductionKind getReductionKindOfLinalgOp(LinalgOp op) {
   return getReductionKind(reductionOp.value());
 }
 
-static MeshOp getMesh(Operation *op, ArrayRef<MeshSharding> operandShardings,
-                      ArrayRef<MeshSharding> resultShardings,
+static GridOp getGrid(Operation *op, ArrayRef<Sharding> operandShardings,
+                      ArrayRef<Sharding> resultShardings,
                       SymbolTableCollection &symbolTable) {
-  for (const MeshSharding &sharding : operandShardings) {
+  for (const Sharding &sharding : operandShardings) {
     if (sharding) {
-      return mesh::getMesh(op, sharding.getMeshAttr(), symbolTable);
+      return shard::getGrid(op, sharding.getGridAttr(), symbolTable);
     }
   }
 
-  for (const MeshSharding &sharding : resultShardings) {
+  for (const Sharding &sharding : resultShardings) {
     if (sharding) {
-      return mesh::getMesh(op, sharding.getMeshAttr(), symbolTable);
+      return shard::getGrid(op, sharding.getGridAttr(), symbolTable);
     }
   }
 
@@ -117,29 +117,29 @@ static MeshOp getMesh(Operation *op, ArrayRef<MeshSharding> operandShardings,
 }
 
 // Choose the operand based on the current process index along the reduction
-// mesh axes.
+// grid axes.
 // We need to use the initial value only once to avoid including it in the
 // reduction multiple times.
 // In each process group only the leading process with linear index 0 would use
 // the original operand.
 // The other processes would use the reduction operation neutral tensor.
 static Value createDestinationPassingStyleInitOperand(
-    LinalgOp op, int operandNumber, Value spmdizedOperand,
-    ArrayRef<MeshAxis> reductionMeshAxes, MeshOp meshOp,
+    LinalgOp op, int operandNumber, Value partitionedOperand,
+    ArrayRef<GridAxis> reductionGridAxes, GridOp gridOp,
     ImplicitLocOpBuilder &builder) {
-  Value processLinearIndexInReductionGroup = mesh::createProcessLinearIndex(
-      meshOp.getSymName(), reductionMeshAxes, builder);
+  Value processLinearIndexInReductionGroup = shard::createProcessLinearIndex(
+      gridOp.getSymName(), reductionGridAxes, builder);
   Value zero = arith::ConstantIndexOp::create(builder, 0);
   Value isLeadProcess = arith::CmpIOp::create(
       builder, builder.getI1Type(), arith::CmpIPredicate::eq,
       processLinearIndexInReductionGroup, zero);
-  scf::IfOp ifOp = scf::IfOp::create(builder, spmdizedOperand.getType(),
+  scf::IfOp ifOp = scf::IfOp::create(builder, partitionedOperand.getType(),
                                      isLeadProcess, true, true);
   // Then block.
   {
     OpBuilder::InsertionGuard insertionGuard(builder);
     builder.setInsertionPointToEnd(&ifOp.getThenRegion().front());
-    scf::YieldOp::create(builder, spmdizedOperand);
+    scf::YieldOp::create(builder, partitionedOperand);
   }
 
   // Else block.
@@ -147,7 +147,7 @@ static Value createDestinationPassingStyleInitOperand(
     OpBuilder::InsertionGuard insertionGuard(builder);
     builder.setInsertionPointToEnd(&ifOp.getElseRegion().front());
     SmallVector<OpFoldResult> shape =
-        tensor::getMixedSizes(builder, builder.getLoc(), spmdizedOperand);
+        tensor::getMixedSizes(builder, builder.getLoc(), partitionedOperand);
 
     SmallVector<Operation *> combinerOps;
     matchReduction(op.getRegionOutputArgs(), operandNumber, combinerOps);
@@ -167,73 +167,72 @@ static Value createDestinationPassingStyleInitOperand(
   return ifOp.getResult(0);
 }
 
-// Create the DPS init operands for the spmdized Linalg op.
-// Return all the new spmdized operands.
+// Create the DPS init operands for the partitioned Linalg op.
+// Return all the new partitioned operands.
 static SmallVector<Value> createDestinationPassingStyleInitOperands(
-    LinalgOp op, MeshOp meshOp, ArrayRef<Value> spmdizedOperands,
-    ArrayRef<MeshAxis> reductionMeshAxes, IRMapping &spmdizationMap,
+    LinalgOp op, GridOp gridOp, ArrayRef<Value> partitionedOperands,
+    ArrayRef<GridAxis> reductionGridAxes, IRMapping &partitionMap,
     ImplicitLocOpBuilder &builder) {
   // TODO: add support for multiple destination passing style initial value
   // operands.
   assert(op.getNumDpsInits() == 1 && "Multiple initial values not supported.");
-  SmallVector<Value> newOperands = llvm::to_vector(spmdizedOperands);
+  SmallVector<Value> newOperands = llvm::to_vector(partitionedOperands);
   auto operandIdx = op.getDpsInitOperand(0)->getOperandNumber();
-  Value spmdizedInitOperand =
-      spmdizationMap.lookup(op->getOperands()[operandIdx]);
+  Value partitionedInitOperand =
+      partitionMap.lookup(op->getOperands()[operandIdx]);
   newOperands[operandIdx] = createDestinationPassingStyleInitOperand(
-      op, 0, spmdizedInitOperand, reductionMeshAxes, meshOp, builder);
+      op, 0, partitionedInitOperand, reductionGridAxes, gridOp, builder);
   return newOperands;
 }
 
 static void createAllReduceForResultsWithoutPartialShardings(
-    LinalgOp unshardedOp, ArrayRef<MeshAxis> opReductionMeshAxes,
-    ArrayRef<MeshSharding> resultShardings, IRMapping &spmdizationMap,
+    LinalgOp unshardedOp, ArrayRef<GridAxis> opReductionGridAxes,
+    ArrayRef<Sharding> resultShardings, IRMapping &partitionMap,
     ImplicitLocOpBuilder &builder) {
   ReductionKind reductionKind = getReductionKindOfLinalgOp(unshardedOp);
   for (auto [unshardedLinalgOpResult, resultSharding] :
        llvm::zip_equal(unshardedOp->getResults(), resultShardings)) {
-    Value spmdizedLinalgOpResult =
-        spmdizationMap.lookup(unshardedLinalgOpResult);
-    Value reducedValue = mesh::AllReduceOp::create(
-        builder, spmdizedLinalgOpResult, resultSharding.getMesh(),
-        opReductionMeshAxes, reductionKind);
-    spmdizationMap.map(unshardedLinalgOpResult, reducedValue);
+    Value partitionedLinalgOpResult =
+        partitionMap.lookup(unshardedLinalgOpResult);
+    Value reducedValue = shard::AllReduceOp::create(
+        builder, partitionedLinalgOpResult, resultSharding.getGrid(),
+        opReductionGridAxes, reductionKind);
+    partitionMap.map(unshardedLinalgOpResult, reducedValue);
   }
 }
 
-static void spmdizeLinalgOpWithShardedReduction(
-    LinalgOp op, ArrayRef<Value> spmdizedOperands,
-    ArrayRef<MeshSharding> operandShardings,
-    ArrayRef<MeshSharding> resultShardings,
+static void partitionLinalgOpWithShardedReduction(
+    LinalgOp op, ArrayRef<Value> partitionedOperands,
+    ArrayRef<Sharding> operandShardings, ArrayRef<Sharding> resultShardings,
     ArrayRef<utils::IteratorType> loopIteratorTypes,
-    ArrayRef<SmallVector<MeshAxis>> meshAxisAssignmentForLoopIterators,
-    IRMapping &spmdizationMap, SymbolTableCollection &symbolTable,
+    ArrayRef<SmallVector<GridAxis>> gridAxisAssignmentForLoopIterators,
+    IRMapping &partitionMap, SymbolTableCollection &symbolTable,
     ImplicitLocOpBuilder &builder) {
-  MeshOp mesh = getMesh(op, operandShardings, resultShardings, symbolTable);
-  SmallVector<MeshAxis> reductionMeshAxes = mesh::getReductionMeshAxes(
-      loopIteratorTypes, meshAxisAssignmentForLoopIterators);
-  SmallVector<Value> spmdizedLinalgOpOperands =
-      createDestinationPassingStyleInitOperands(op, mesh, spmdizedOperands,
-                                                reductionMeshAxes,
-                                                spmdizationMap, builder);
-  // We must not change the operand mappings of the original spmdizationMap as
-  // they are the mappings for the whole spmdization blob and may be used by
+  GridOp grid = getGrid(op, operandShardings, resultShardings, symbolTable);
+  SmallVector<GridAxis> reductionGridAxes = shard::getReductionGridAxes(
+      loopIteratorTypes, gridAxisAssignmentForLoopIterators);
+  SmallVector<Value> partitionedLinalgOpOperands =
+      createDestinationPassingStyleInitOperands(op, grid, partitionedOperands,
+                                                reductionGridAxes, partitionMap,
+                                                builder);
+  // We must not change the operand mappings of the original partitionMap as
+  // they are the mappings for the whole partition blob and may be used by
   // others.
-  IRMapping internalSpmdizationMap;
-  for (auto [unshardedOperand, spmdizedOperand] :
-       llvm::zip_equal(op->getOperands(), spmdizedLinalgOpOperands)) {
-    internalSpmdizationMap.map(unshardedOperand, spmdizedOperand);
+  IRMapping internalPartitionMap;
+  for (auto [unshardedOperand, partitionedOperand] :
+       llvm::zip_equal(op->getOperands(), partitionedLinalgOpOperands)) {
+    internalPartitionMap.map(unshardedOperand, partitionedOperand);
   }
-  spmdizeTriviallyShardableOperation(
-      *op, spmdizedLinalgOpOperands, operandShardings, resultShardings,
-      internalSpmdizationMap, symbolTable, builder);
+  partitionTriviallyShardableOperation(
+      *op, partitionedLinalgOpOperands, operandShardings, resultShardings,
+      internalPartitionMap, symbolTable, builder);
   for (Value result : op->getResults()) {
-    spmdizationMap.map(result, internalSpmdizationMap.lookup(result));
+    partitionMap.map(result, internalPartitionMap.lookup(result));
   }
 
   // Handle partial shardings.
   createAllReduceForResultsWithoutPartialShardings(
-      op, reductionMeshAxes, resultShardings, spmdizationMap, builder);
+      op, reductionGridAxes, resultShardings, partitionMap, builder);
 }
 
 namespace {
@@ -243,7 +242,7 @@ namespace {
 // permutations.
 template <typename Op>
 struct StructuredOpShardingInterface
-    : public mesh::ShardingInterface::ExternalModel<
+    : public shard::ShardingInterface::ExternalModel<
           StructuredOpShardingInterface<Op>, Op> {
   SmallVector<utils::IteratorType> getLoopIteratorTypes(Operation *op) const {
     return llvm::cast<LinalgOp>(op).getIteratorTypesArray();
@@ -272,16 +271,16 @@ struct StructuredOpShardingInterface
         [](unsigned count, utils::IteratorType iter) {
           return count + (iter == utils::IteratorType::reduction);
         });
-    mesh::ReductionKind reductionKind = getReductionKindOfLinalgOp(linalgOp);
+    shard::ReductionKind reductionKind = getReductionKindOfLinalgOp(linalgOp);
     return SmallVector<ReductionKind>(reductionItersCount, reductionKind);
   }
 
-  LogicalResult spmdize(Operation *op, ArrayRef<Value> spmdizedOperands,
-                        ArrayRef<MeshSharding> operandShardings,
-                        ArrayRef<MeshSharding> resultShardings,
-                        IRMapping &spmdizationMap,
-                        SymbolTableCollection &symbolTable,
-                        OpBuilder &builder) const {
+  LogicalResult partition(Operation *op, ArrayRef<Value> partitionedOperands,
+                          ArrayRef<Sharding> operandShardings,
+                          ArrayRef<Sharding> resultShardings,
+                          IRMapping &partitionMap,
+                          SymbolTableCollection &symbolTable,
+                          OpBuilder &builder) const {
     LinalgOp linalgOp = llvm::cast<LinalgOp>(op);
 
     SmallVector<AffineMap> indexingMaps = linalgOp.getIndexingMapsArray();
@@ -297,20 +296,20 @@ struct StructuredOpShardingInterface
 
     SmallVector<utils::IteratorType> loopIteratorTypes =
         linalgOp.getIteratorTypesArray();
-    ShardingArray meshAxisAssignmentForLoopIterators =
-        getMeshAxisAssignmentForLoopIterators(operandShardings, resultShardings,
+    ShardingArray gridAxisAssignmentForLoopIterators =
+        getGridAxisAssignmentForLoopIterators(operandShardings, resultShardings,
                                               loopIteratorTypes, indexingMaps);
-    if (mesh::isAtLeastOneReductionIteratorSharded(
-            loopIteratorTypes, meshAxisAssignmentForLoopIterators)) {
+    if (shard::isAtLeastOneReductionIteratorSharded(
+            loopIteratorTypes, gridAxisAssignmentForLoopIterators)) {
       ImplicitLocOpBuilder implicitLocBuilder(op->getLoc(), builder);
-      spmdizeLinalgOpWithShardedReduction(
-          linalgOp, spmdizedOperands, operandShardings, resultShardings,
-          loopIteratorTypes, meshAxisAssignmentForLoopIterators, spmdizationMap,
+      partitionLinalgOpWithShardedReduction(
+          linalgOp, partitionedOperands, operandShardings, resultShardings,
+          loopIteratorTypes, gridAxisAssignmentForLoopIterators, partitionMap,
           symbolTable, implicitLocBuilder);
     } else {
-      spmdizeTriviallyShardableOperation(*op, spmdizedOperands,
-                                         operandShardings, resultShardings,
-                                         spmdizationMap, symbolTable, builder);
+      partitionTriviallyShardableOperation(*op, partitionedOperands,
+                                           operandShardings, resultShardings,
+                                           partitionMap, symbolTable, builder);
     }
 
     return success();
@@ -330,7 +329,7 @@ static void registerAll(MLIRContext *ctx) {
   (registerOne<OpTypes>(ctx), ...);
 }
 
-void registerMeshShardingInterfaceExternalModels(DialectRegistry &registry) {
+void registerShardingInterfaceExternalModels(DialectRegistry &registry) {
   registry.addExtension(+[](MLIRContext *ctx, LinalgDialect *dialect) {
     DialectRegistry registry;
     registry.insert<affine::AffineDialect, arith::ArithDialect, scf::SCFDialect,
diff --git a/mlir/lib/Dialect/Linalg/Transforms/Transforms.cpp b/mlir/lib/Dialect/Linalg/Transforms/Transforms.cpp
index 1f1e617..bb725f2 100644
--- a/mlir/lib/Dialect/Linalg/Transforms/Transforms.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/Transforms.cpp
@@ -947,9 +947,9 @@ DecomposePadOpPattern::matchAndRewrite(tensor::PadOp padOp,
   auto getIdxValue = [&](OpFoldResult ofr) {
     if (auto val = llvm::dyn_cast_if_present<Value>(ofr))
       return val;
-    return rewriter
-        .create<arith::ConstantIndexOp>(
-            padOp.getLoc(), cast<IntegerAttr>(cast<Attribute>(ofr)).getInt())
+    return arith::ConstantIndexOp::create(
+               rewriter, padOp.getLoc(),
+               cast<IntegerAttr>(cast<Attribute>(ofr)).getInt())
         .getResult();
   };
 
diff --git a/mlir/lib/Dialect/Linalg/Transforms/TransposeConv2D.cpp b/mlir/lib/Dialect/Linalg/Transforms/TransposeConv2D.cpp
index 99fb8c7..35453e2 100644
--- a/mlir/lib/Dialect/Linalg/Transforms/TransposeConv2D.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/TransposeConv2D.cpp
@@ -70,9 +70,8 @@ FailureOr<Operation *> transposeConv2DHelper(RewriterBase &rewriter,
     input = tensor::EmptyOp::create(rewriter, loc, newFilterShape, elementTy)
                 .getResult();
   } else {
-    input = rewriter
-                .create<memref::AllocOp>(
-                    loc, MemRefType::get(newFilterShape, elementTy))
+    input = memref::AllocOp::create(rewriter, loc,
+                                    MemRefType::get(newFilterShape, elementTy))
                 .getResult();
   }
 
diff --git a/mlir/lib/Dialect/Linalg/Transforms/Vectorization.cpp b/mlir/lib/Dialect/Linalg/Transforms/Vectorization.cpp
index 78c6bd1..793eec7 100644
--- a/mlir/lib/Dialect/Linalg/Transforms/Vectorization.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/Vectorization.cpp
@@ -38,7 +38,8 @@
 #include "llvm/ADT/Sequence.h"
 #include "llvm/ADT/SmallVector.h"
 #include "llvm/ADT/TypeSwitch.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
+#include "llvm/Support/InterleavedRange.h"
 #include "llvm/Support/MathExtras.h"
 #include "llvm/Support/raw_ostream.h"
 #include <optional>
@@ -48,9 +49,6 @@ using namespace mlir::linalg;
 
 #define DEBUG_TYPE "linalg-vectorization"
 
-#define DBGS() (llvm::dbgs() << '[' << DEBUG_TYPE << "] ")
-#define LDBG(X) LLVM_DEBUG(DBGS() << X << "\n")
-
 /// Try to vectorize `convOp` as a convolution.
 static FailureOr<Operation *>
 vectorizeConvolution(RewriterBase &rewriter, LinalgOp convOp,
@@ -403,12 +401,8 @@ LogicalResult VectorizationState::initState(RewriterBase &rewriter,
     scalableVecDims.append(linalgOp.getNumLoops(), false);
   }
 
-  LDBG("Canonical vector shape: ");
-  LLVM_DEBUG(llvm::interleaveComma(canonicalVecShape, llvm::dbgs()));
-  LLVM_DEBUG(llvm::dbgs() << "\n");
-  LDBG("Scalable vector dims: ");
-  LLVM_DEBUG(llvm::interleaveComma(scalableVecDims, llvm::dbgs()));
-  LLVM_DEBUG(llvm::dbgs() << "\n");
+  LDBG() << "Canonical vector shape: " << llvm::interleaved(canonicalVecShape);
+  LDBG() << "Scalable vector dims: " << llvm::interleaved(scalableVecDims);
 
   if (ShapedType::isDynamicShape(canonicalVecShape))
     return failure();
@@ -452,14 +446,14 @@ Value VectorizationState::getOrCreateMaskFor(
                       : AffineMap::getMultiDimIdentityMap(
                             linalgOp.getNumLoops(), rewriter.getContext());
 
-  LDBG("Masking map: " << maskingMap << "\n");
+  LDBG() << "Masking map: " << maskingMap;
 
   // Return the active mask for the masking map of this operation if it was
   // already created.
   auto activeMaskIt = activeMaskCache.find(maskingMap);
   if (activeMaskIt != activeMaskCache.end()) {
     Value mask = activeMaskIt->second;
-    LDBG("Reusing mask: " << mask << "\n");
+    LDBG() << "Reusing mask: " << mask;
     return mask;
   }
 
@@ -474,12 +468,10 @@ Value VectorizationState::getOrCreateMaskFor(
   auto maskType = getCanonicalVecType(rewriter.getI1Type(), maskingMap);
   auto maskShape = maskType.getShape();
 
-  LDBG("Mask shape: ");
-  LLVM_DEBUG(llvm::interleaveComma(maskShape, llvm::dbgs()));
-  LLVM_DEBUG(llvm::dbgs() << "\n");
+  LDBG() << "Mask shape: " << llvm::interleaved(maskShape);
 
   if (permutedStaticSizes == maskShape) {
-    LDBG("Masking is not needed for masking map: " << maskingMap << "\n");
+    LDBG() << "Masking is not needed for masking map: " << maskingMap;
     activeMaskCache[maskingMap] = Value();
     return Value();
   }
@@ -494,8 +486,9 @@ Value VectorizationState::getOrCreateMaskFor(
                                   ? true
                                   : std::get<0>(it) == std::get<1>(it);
                      })) {
-      LDBG("Dynamic + static dimensions match vector sizes, masking is not "
-           "required.\n");
+      LDBG()
+          << "Dynamic + static dimensions match vector sizes, masking is not "
+             "required.";
       activeMaskCache[maskingMap] = Value();
       return Value();
     }
@@ -510,7 +503,7 @@ Value VectorizationState::getOrCreateMaskFor(
   // Create the mask based on the dimension values.
   Value mask = vector::CreateMaskOp::create(rewriter, linalgOp.getLoc(),
                                             maskType, upperBounds);
-  LDBG("Creating new mask: " << mask << "\n");
+  LDBG() << "Creating new mask: " << mask;
   activeMaskCache[maskingMap] = mask;
   return mask;
 }
@@ -519,7 +512,7 @@ Operation *
 VectorizationState::maskOperation(RewriterBase &rewriter, Operation *opToMask,
                                   LinalgOp linalgOp,
                                   std::optional<AffineMap> maybeIndexingMap) {
-  LDBG("Trying to mask: " << *opToMask << "\n");
+  LDBG() << "Trying to mask: " << *opToMask;
 
   std::optional<AffineMap> maybeMaskingMap = std::nullopt;
   if (maybeIndexingMap)
@@ -530,7 +523,7 @@ VectorizationState::maskOperation(RewriterBase &rewriter, Operation *opToMask,
       getOrCreateMaskFor(rewriter, opToMask, linalgOp, maybeMaskingMap);
 
   if (!mask) {
-    LDBG("No mask required\n");
+    LDBG() << "No mask required";
     return opToMask;
   }
 
@@ -544,7 +537,7 @@ VectorizationState::maskOperation(RewriterBase &rewriter, Operation *opToMask,
     rewriter.replaceAllUsesExcept(resVal, maskOp.getResult(resIdx),
                                   maskOpTerminator);
 
-  LDBG("Masked operation: " << *maskOp << "\n");
+  LDBG() << "Masked operation: " << *maskOp;
   return maskOp;
 }
 
@@ -748,7 +741,7 @@ static Value buildVectorWrite(RewriterBase &rewriter, Value value,
     maskedWriteOp.setInBoundsAttr(rewriter.getBoolArrayAttr(inBounds));
   }
 
-  LDBG("vectorized op: " << *write << "\n");
+  LDBG() << "vectorized op: " << *write;
   if (!write->getResults().empty())
     return write->getResult(0);
   return Value();
@@ -1090,7 +1083,7 @@ getTensorExtractMemoryAccessPattern(tensor::ExtractOp extractOp,
   }
 
   if (!leadingIdxsLoopInvariant) {
-    LDBG("Found gather load: " << extractOp);
+    LDBG() << "Found gather load: " << extractOp;
     return VectorMemoryAccessKind::Gather;
   }
 
@@ -1104,7 +1097,7 @@ getTensorExtractMemoryAccessPattern(tensor::ExtractOp extractOp,
   // If the trailing index is loop invariant then this is a scalar load.
   if (leadingIdxsLoopInvariant &&
       isLoopInvariantIdx(linalgOp, extractOpTrailingIdx, resType)) {
-    LDBG("Found scalar broadcast load: " << extractOp);
+    LDBG() << "Found scalar broadcast load: " << extractOp;
 
     return VectorMemoryAccessKind::ScalarBroadcast;
   }
@@ -1122,12 +1115,12 @@ getTensorExtractMemoryAccessPattern(tensor::ExtractOp extractOp,
   isContiguousLoad &= (foundIndexOp && isRowVector);
 
   if (isContiguousLoad) {
-    LDBG("Found contigous load: " << extractOp);
+    LDBG() << "Found contigous load: " << extractOp;
     return VectorMemoryAccessKind::Contiguous;
   }
 
   // 4. Fallback case - gather load.
-  LDBG("Found gather load: " << extractOp);
+  LDBG() << "Found gather load: " << extractOp;
   return VectorMemoryAccessKind::Gather;
 }
 
@@ -1171,7 +1164,7 @@ vectorizeTensorExtract(RewriterBase &rewriter, VectorizationState &state,
         maskConstantOp, passThruConstantOp);
     gatherOp = state.maskOperation(rewriter, gatherOp, linalgOp);
 
-    LDBG("Vectorised as gather load: " << extractOp << "\n");
+    LDBG() << "Vectorised as gather load: " << extractOp;
     return VectorizationHookResult{VectorizationHookStatus::NewOp, gatherOp};
   }
 
@@ -1235,7 +1228,7 @@ vectorizeTensorExtract(RewriterBase &rewriter, VectorizationState &state,
     auto *maskedReadOp =
         mlir::vector::maskOperation(rewriter, transferReadOp, allTrue);
 
-    LDBG("Vectorised as scalar broadcast load: " << extractOp << "\n");
+    LDBG() << "Vectorised as scalar broadcast load: " << extractOp;
     return VectorizationHookResult{VectorizationHookStatus::NewOp,
                                    maskedReadOp};
   }
@@ -1262,7 +1255,7 @@ vectorizeTensorExtract(RewriterBase &rewriter, VectorizationState &state,
       rewriter, loc, resultType, extractOp.getTensor(), transferReadIdxs,
       /*padding=*/std::nullopt, permutationMap, inBounds);
 
-  LDBG("Vectorised as contiguous load: " << extractOp);
+  LDBG() << "Vectorised as contiguous load: " << extractOp;
   return VectorizationHookResult{VectorizationHookStatus::NewOp,
                                  transferReadOp};
 }
@@ -1310,7 +1303,7 @@ static VectorizationHookResult
 vectorizeOneOp(RewriterBase &rewriter, VectorizationState &state,
                LinalgOp linalgOp, Operation *op, const IRMapping &bvm,
                ArrayRef<CustomVectorizationHook> customVectorizationHooks) {
-  LDBG("vectorize op " << *op << "\n");
+  LDBG() << "vectorize op " << *op;
 
   // 1. Try to apply any CustomVectorizationHook.
   if (!customVectorizationHooks.empty()) {
@@ -1425,7 +1418,7 @@ static LogicalResult
 vectorizeAsLinalgGeneric(RewriterBase &rewriter, VectorizationState &state,
                          LinalgOp linalgOp,
                          SmallVectorImpl<Value> &newResults) {
-  LDBG("Vectorizing operation as linalg generic\n");
+  LDBG() << "Vectorizing operation as linalg generic/n";
   Block *block = linalgOp.getBlock();
 
   // 2. Values defined above the region can only be broadcast for now. Make them
@@ -1490,8 +1483,8 @@ vectorizeAsLinalgGeneric(RewriterBase &rewriter, VectorizationState &state,
       readValue = vector::ExtractOp::create(rewriter, loc, readValue,
                                             ArrayRef<int64_t>());
 
-    LDBG("New vectorized bbarg(" << bbarg.getArgNumber() << "): " << readValue
-                                 << "\n");
+    LDBG() << "New vectorized bbarg(" << bbarg.getArgNumber()
+           << "): " << readValue;
     bvm.map(bbarg, readValue);
     bvm.map(opOperand->get(), readValue);
   }
@@ -1523,13 +1516,13 @@ vectorizeAsLinalgGeneric(RewriterBase &rewriter, VectorizationState &state,
     VectorizationHookResult result =
         vectorizeOneOp(rewriter, state, linalgOp, &op, bvm, hooks);
     if (result.status == VectorizationHookStatus::Failure) {
-      LDBG("failed to vectorize: " << op << "\n");
+      LDBG() << "failed to vectorize: " << op;
       return failure();
     }
     if (result.status == VectorizationHookStatus::NewOp) {
       Operation *maybeMaskedOp =
           state.maskOperation(rewriter, result.newOp, linalgOp);
-      LDBG("New vector op: " << *maybeMaskedOp << "\n");
+      LDBG() << "New vector op: " << *maybeMaskedOp;
       bvm.map(op.getResults(), maybeMaskedOp->getResults());
     }
   }
@@ -1920,14 +1913,6 @@ vectorizeAsTensorUnpackOp(RewriterBase &rewriter, linalg::UnPackOp unpackOp,
   readVectorSizes.append(sourceShape.begin() + vectorSizes.size(),
                          sourceShape.end());
 
-  ReifiedRankedShapedTypeDims reifiedRetShapes;
-  LogicalResult status =
-      cast<ReifyRankedShapedTypeOpInterface>(unpackOp.getOperation())
-          .reifyResultShapes(rewriter, reifiedRetShapes);
-  if (status.failed()) {
-    LDBG("Unable to reify result shapes of " << unpackOp);
-    return failure();
-  }
   Location loc = unpackOp->getLoc();
 
   auto padValue = arith::ConstantOp::create(
@@ -2010,7 +1995,7 @@ vectorizeAsTensorPadOp(RewriterBase &rewriter, tensor::PadOp padOp,
 // ops that may not commute (e.g. linear reduction + non-linear instructions).
 static LogicalResult reductionPreconditions(LinalgOp op) {
   if (llvm::none_of(op.getIteratorTypesArray(), isReductionIterator)) {
-    LDBG("reduction precondition failed: no reduction iterator\n");
+    LDBG() << "reduction precondition failed: no reduction iterator";
     return failure();
   }
   for (OpOperand &opOperand : op.getDpsInitsMutable()) {
@@ -2020,7 +2005,7 @@ static LogicalResult reductionPreconditions(LinalgOp op) {
 
     Operation *reduceOp = matchLinalgReduction(&opOperand);
     if (!reduceOp || !getCombinerOpKind(reduceOp)) {
-      LDBG("reduction precondition failed: reduction detection failed\n");
+      LDBG() << "reduction precondition failed: reduction detection failed";
       return failure();
     }
   }
@@ -2031,13 +2016,13 @@ static LogicalResult
 vectorizeDynamicConvOpPrecondition(linalg::LinalgOp conv,
                                    bool flatten1DDepthwiseConv) {
   if (flatten1DDepthwiseConv) {
-    LDBG("Vectorization of flattened convs with dynamic shapes is not "
-         "supported\n");
+    LDBG() << "Vectorization of flattened convs with dynamic shapes is not "
+              "supported";
     return failure();
   }
 
   if (!isa<linalg::DepthwiseConv1DNwcWcOp>(conv)) {
-    LDBG("Not a 1D depth-wise WC conv, dynamic shapes are not supported\n");
+    LDBG() << "Not a 1D depth-wise WC conv, dynamic shapes are not supported";
     return failure();
   }
 
@@ -2047,8 +2032,8 @@ vectorizeDynamicConvOpPrecondition(linalg::LinalgOp conv,
   ArrayRef<int64_t> lhsShape = cast<ShapedType>(lhs.getType()).getShape();
   auto shapeWithoutCh = lhsShape.drop_back(1);
   if (ShapedType::isDynamicShape(shapeWithoutCh)) {
-    LDBG("Dynamically-shaped op vectorization precondition failed: only "
-         "channel dim can be dynamic\n");
+    LDBG() << "Dynamically-shaped op vectorization precondition failed: only "
+              "channel dim can be dynamic";
     return failure();
   }
 
@@ -2071,7 +2056,7 @@ vectorizeDynamicLinalgOpPrecondition(linalg::LinalgOp op,
           op.getOperation()))
     return failure();
 
-  LDBG("Dynamically-shaped op meets vectorization pre-conditions\n");
+  LDBG() << "Dynamically-shaped op meets vectorization pre-conditions";
   return success();
 }
 
@@ -2083,7 +2068,7 @@ vectorizeUnPackOpPrecondition(linalg::UnPackOp unpackOp,
   if (llvm::any_of(unpackOp.getInnerTiles(), [](OpFoldResult res) {
         return !getConstantIntValue(res).has_value();
       })) {
-    LDBG("Inner-tiles must be constant: " << unpackOp << "\n");
+    LDBG() << "Inner-tiles must be constant: " << unpackOp;
     return failure();
   }
   ArrayRef<int64_t> resultShape = unpackOp.getDestType().getShape();
@@ -2123,7 +2108,7 @@ vectorizeInsertSliceOpPrecondition(tensor::InsertSliceOp sliceOp,
       !sourceType.hasStaticShape() && inputVectorSizes.empty();
 
   if (!padValue && isOutOfBoundsRead) {
-    LDBG("Failed to get a pad value for out-of-bounds read access\n");
+    LDBG() << "Failed to get a pad value for out-of-bounds read access";
     return failure();
   }
   return success();
@@ -2153,7 +2138,7 @@ vectorizeAsLinalgContraction(RewriterBase &rewriter, VectorizationState &state,
   Operation *reduceOp = matchLinalgReduction(outOperand);
   auto maybeKind = getCombinerOpKind(reduceOp);
   if (!maybeKind) {
-    LDBG("Failed to determine contraction combining kind.\n");
+    LDBG() << "Failed to determine contraction combining kind.";
     return failure();
   }
 
@@ -2163,7 +2148,7 @@ vectorizeAsLinalgContraction(RewriterBase &rewriter, VectorizationState &state,
   AffineMap lhsMap = linalgOp.getIndexingMapsArray()[0];
   AffineMap rhsMap = linalgOp.getIndexingMapsArray()[1];
   if (getUnusedDimsBitVector({lhsMap, rhsMap}).any()) {
-    LDBG("Contractions with broadcasts are not supported.\n");
+    LDBG() << "Contractions with broadcasts are not supported.";
     return failure();
   }
 
@@ -2198,8 +2183,8 @@ vectorizeAsLinalgContraction(RewriterBase &rewriter, VectorizationState &state,
   }
 
   // Create contraction.
-  Operation *contractOp = rewriter.create<vector::ContractionOp>(
-      loc, /*lhs=*/vecOperands[0],
+  Operation *contractOp = vector::ContractionOp::create(
+      rewriter, loc, /*lhs=*/vecOperands[0],
       /*rhs=*/vecOperands[1], /*acc=*/vecOperands[2],
       linalgOp.getIndexingMaps(), rewriter.getArrayAttr(iterAttrs), *maybeKind);
   contractOp = state.maskOperation(rewriter, contractOp, linalgOp);
@@ -2355,7 +2340,7 @@ static LogicalResult vectorizeLinalgOpPrecondition(
 
   if (linalgOp.hasDynamicShape() && failed(vectorizeDynamicLinalgOpPrecondition(
                                         linalgOp, flatten1DDepthwiseConv))) {
-    LDBG("Dynamically-shaped op failed vectorization pre-conditions\n");
+    LDBG() << "Dynamically-shaped op failed vectorization pre-conditions";
     return failure();
   }
 
@@ -2397,11 +2382,11 @@ static LogicalResult vectorizeLinalgOpPrecondition(
   // all indexing maps are projected permutations. For convs and stencils the
   // logic will need to evolve.
   if (!allIndexingsAreProjectedPermutation(linalgOp)) {
-    LDBG("precondition failed: not projected permutations\n");
+    LDBG() << "precondition failed: not projected permutations";
     return failure();
   }
   if (failed(reductionPreconditions(linalgOp))) {
-    LDBG("precondition failed: reduction preconditions\n");
+    LDBG() << "precondition failed: reduction preconditions";
     return failure();
   }
   return success();
@@ -2413,7 +2398,7 @@ vectorizePackOpPrecondition(linalg::PackOp packOp,
   auto padValue = packOp.getPaddingValue();
   Attribute cstAttr;
   if (padValue && !matchPattern(padValue, m_Constant(&cstAttr))) {
-    LDBG("pad value is not constant: " << packOp << "\n");
+    LDBG() << "pad value is not constant: " << packOp;
     return failure();
   }
   ArrayRef<int64_t> resultTensorShape = packOp.getDestType().getShape();
@@ -2433,7 +2418,7 @@ vectorizePackOpPrecondition(linalg::PackOp packOp,
   if (llvm::any_of(packOp.getInnerTiles(), [](OpFoldResult v) {
         return !getConstantIntValue(v).has_value();
       })) {
-    LDBG("inner_tiles must be constant: " << packOp << "\n");
+    LDBG() << "inner_tiles must be constant: " << packOp;
     return failure();
   }
 
@@ -2445,7 +2430,7 @@ vectorizePadOpPrecondition(tensor::PadOp padOp,
                            ArrayRef<int64_t> inputVectorSizes) {
   auto padValue = padOp.getConstantPaddingValue();
   if (!padValue) {
-    LDBG("pad value is not constant: " << padOp << "\n");
+    LDBG() << "pad value is not constant: " << padOp;
     return failure();
   }
 
@@ -2472,7 +2457,7 @@ vectorizePadOpPrecondition(tensor::PadOp padOp,
         return (!pad.has_value() || pad.value() != 0) &&
                resultTensorShape[pos] != 1;
       })) {
-    LDBG("low pad must all be zero for all non unit dims: " << padOp << "\n");
+    LDBG() << "low pad must all be zero for all non unit dims: " << padOp;
     return failure();
   }
 
@@ -2541,13 +2526,14 @@ vectorizeScalableVectorPrecondition(Operation *op,
   case utils::IteratorType::reduction: {
     // Check 3. above is met.
     if (iterators.size() != inputVectorSizes.size()) {
-      LDBG("Non-trailing reduction dim requested for scalable "
-           "vectorization\n");
+      LDBG() << "Non-trailing reduction dim requested for scalable "
+                "vectorization";
       return failure();
     }
     if (isa<linalg::MatmulOp>(op) || isa<linalg::MatmulTransposeAOp>(op)) {
-      LDBG("Scalable vectorization of the reduction dim in Matmul-like ops "
-           "is not supported\n");
+      LDBG()
+          << "Scalable vectorization of the reduction dim in Matmul-like ops "
+             "is not supported";
       return failure();
     }
     break;
@@ -2555,8 +2541,8 @@ vectorizeScalableVectorPrecondition(Operation *op,
   case utils::IteratorType::parallel: {
     // Check 1. and 2. above are met.
     if (seenNonUnitParallel) {
-      LDBG("Inner parallel dim not requested for scalable "
-           "vectorization\n");
+      LDBG() << "Inner parallel dim not requested for scalable "
+                "vectorization";
       return failure();
     }
     break;
@@ -2572,8 +2558,9 @@ vectorizeScalableVectorPrecondition(Operation *op,
     //    * iterators = [..., parallel, reduction]
     //    * scalable flags = [..., true, true]
     if (iterators.back() == utils::IteratorType::reduction) {
-      LDBG("Higher dim than the trailing reduction dim requested for scalable "
-           "vectorization\n");
+      LDBG() << "Higher dim than the trailing reduction dim requested for "
+                "scalable "
+                "vectorizatio";
       return failure();
     }
     scalableFlags.pop_back();
@@ -2656,18 +2643,15 @@ FailureOr<VectorizationResult> mlir::linalg::vectorize(
     ArrayRef<bool> inputScalableVecDims, bool vectorizeNDExtract,
     bool flatten1DDepthwiseConv, bool assumeDynamicDimsMatchVecSizes,
     bool createNamedContraction) {
-  LDBG("Attempting to vectorize:\n" << *op << "\n");
-  LDBG("Input vector sizes: ");
-  LLVM_DEBUG(llvm::interleaveComma(inputVectorSizes, llvm::dbgs()));
-  LLVM_DEBUG(llvm::dbgs() << "\n");
-  LDBG("Input scalable vector dims: ");
-  LLVM_DEBUG(llvm::interleaveComma(inputScalableVecDims, llvm::dbgs()));
-  LLVM_DEBUG(llvm::dbgs() << "\n");
+  LDBG() << "Attempting to vectorize: " << *op;
+  LDBG() << "Input vector sizes: " << llvm::interleaved(inputVectorSizes);
+  LDBG() << "Input scalable vector dims: "
+         << llvm::interleaved(inputScalableVecDims);
 
   if (failed(vectorizeOpPrecondition(op, inputVectorSizes, inputScalableVecDims,
                                      vectorizeNDExtract,
                                      flatten1DDepthwiseConv))) {
-    LDBG("Vectorization pre-conditions failed\n");
+    LDBG() << "Vectorization pre-conditions failed";
     return failure();
   }
 
@@ -2677,7 +2661,7 @@ FailureOr<VectorizationResult> mlir::linalg::vectorize(
     if (failed(state.initState(rewriter, linalgOp, inputVectorSizes,
                                inputScalableVecDims,
                                assumeDynamicDimsMatchVecSizes))) {
-      LDBG("Vectorization state couldn't be initialized\n");
+      LDBG() << "Vectorization state couldn't be initialized";
       return failure();
     }
   }
@@ -2698,7 +2682,7 @@ FailureOr<VectorizationResult> mlir::linalg::vectorize(
                 return success();
               }
 
-              LDBG("Unsupported convolution can't be vectorized.\n");
+              LDBG() << "Unsupported convolution can't be vectorized.";
               return failure();
             }
 
@@ -2707,8 +2691,9 @@ FailureOr<VectorizationResult> mlir::linalg::vectorize(
               return vectorizeAsLinalgContraction(rewriter, state, linalgOp,
                                                   results);
 
-            LDBG("Vectorize generic by broadcasting to the canonical vector "
-                 "shape\n");
+            LDBG()
+                << "Vectorize generic by broadcasting to the canonical vector "
+                   "shape";
 
             // Pre-process before proceeding.
             convertAffineApply(rewriter, linalgOp);
@@ -2739,7 +2724,7 @@ FailureOr<VectorizationResult> mlir::linalg::vectorize(
           .Default([](auto) { return failure(); });
 
   if (failed(vectorizeResult)) {
-    LDBG("Vectorization failed\n");
+    LDBG() << "Vectorization failed";
     return failure();
   }
 
@@ -3244,8 +3229,8 @@ static bool mayExistInterleavedUses(Operation *firstOp, Operation *secondOp,
                                     ValueRange values) {
   if (firstOp->getBlock() != secondOp->getBlock() ||
       !firstOp->isBeforeInBlock(secondOp)) {
-    LDBG("interleavedUses precondition failed, firstOp: "
-         << *firstOp << ", second op: " << *secondOp << "\n");
+    LDBG() << "interleavedUses precondition failed, firstOp: " << *firstOp
+           << ", second op: " << *secondOp;
     return true;
   }
   for (auto v : values) {
@@ -3257,8 +3242,8 @@ static bool mayExistInterleavedUses(Operation *firstOp, Operation *secondOp,
       if (owner->getBlock() == firstOp->getBlock() &&
           (owner->isBeforeInBlock(firstOp) || secondOp->isBeforeInBlock(owner)))
         continue;
-      LDBG(" found interleaved op " << *owner << ", firstOp: " << *firstOp
-                                    << ", second op: " << *secondOp << "\n");
+      LDBG() << " found interleaved op " << *owner << ", firstOp: " << *firstOp
+             << ", second op: " << *secondOp;
       return true;
     }
   }
@@ -3721,8 +3706,8 @@ struct Conv1DGenerator
     }
     }
 
-    return rewriter
-        .create<vector::TransferWriteOp>(loc, res, resShaped, resPadding)
+    return vector::TransferWriteOp::create(rewriter, loc, res, resShaped,
+                                           resPadding)
         .getOperation();
   }
 
diff --git a/mlir/lib/Dialect/Linalg/Transforms/WinogradConv2D.cpp b/mlir/lib/Dialect/Linalg/Transforms/WinogradConv2D.cpp
index 669fefc..b80b27f 100644
--- a/mlir/lib/Dialect/Linalg/Transforms/WinogradConv2D.cpp
+++ b/mlir/lib/Dialect/Linalg/Transforms/WinogradConv2D.cpp
@@ -398,10 +398,9 @@ Value filterTransform(RewriterBase &rewriter, Location loc, Value filter,
 
       retRows = GMatrix.rows;
       auto matmulType = RankedTensorType::get({retRows, filterW}, elementType);
-      auto empty =
-          builder
-              .create<tensor::EmptyOp>(loc, matmulType.getShape(), elementType)
-              .getResult();
+      auto empty = tensor::EmptyOp::create(builder, loc, matmulType.getShape(),
+                                           elementType)
+                       .getResult();
       auto init =
           linalg::FillOp::create(builder, loc, zero, empty).getResult(0);
 
@@ -422,10 +421,9 @@ Value filterTransform(RewriterBase &rewriter, Location loc, Value filter,
 
       auto matmulType =
           RankedTensorType::get({retRows, GTMatrix.cols}, elementType);
-      auto empty =
-          builder
-              .create<tensor::EmptyOp>(loc, matmulType.getShape(), elementType)
-              .getResult();
+      auto empty = tensor::EmptyOp::create(builder, loc, matmulType.getShape(),
+                                           elementType)
+                       .getResult();
       auto init =
           linalg::FillOp::create(builder, loc, zero, empty).getResult(0);
 
@@ -547,10 +545,9 @@ Value inputTransform(RewriterBase &rewriter, Location loc, Value input,
 
       retRows = BTMatrix.rows;
       auto matmulType = RankedTensorType::get({retRows, alphaW}, elementType);
-      auto empty =
-          builder
-              .create<tensor::EmptyOp>(loc, matmulType.getShape(), elementType)
-              .getResult();
+      auto empty = tensor::EmptyOp::create(builder, loc, matmulType.getShape(),
+                                           elementType)
+                       .getResult();
       auto init =
           linalg::FillOp::create(builder, loc, zero, empty).getResult(0);
 
@@ -572,10 +569,9 @@ Value inputTransform(RewriterBase &rewriter, Location loc, Value input,
 
       retCols = BMatrix.cols;
       auto matmulType = RankedTensorType::get({retRows, retCols}, elementType);
-      auto empty =
-          builder
-              .create<tensor::EmptyOp>(loc, matmulType.getShape(), elementType)
-              .getResult();
+      auto empty = tensor::EmptyOp::create(builder, loc, matmulType.getShape(),
+                                           elementType)
+                       .getResult();
       auto init =
           linalg::FillOp::create(builder, loc, zero, empty).getResult(0);
       Value B =
@@ -661,9 +657,8 @@ static Value matrixMultiply(RewriterBase &rewriter, Location loc,
       {inputShape[0] * inputShape[1],
        inputShape[2] * inputShape[3] * inputShape[4], filterShape[3]},
       outputElementType);
-  Value empty = rewriter
-                    .create<tensor::EmptyOp>(loc, matmulType.getShape(),
-                                             outputElementType)
+  Value empty = tensor::EmptyOp::create(rewriter, loc, matmulType.getShape(),
+                                        outputElementType)
                     .getResult();
   Value zero = arith::ConstantOp::create(
       rewriter, loc, rewriter.getZeroAttr(outputElementType));
@@ -782,9 +777,8 @@ Value outputTransform(RewriterBase &rewriter, Location loc, Value value,
       auto matmulType = RankedTensorType::get({retRows, valueW}, elementType);
       Value init = outInitVal;
       if (rightTransform || scalarFactor != 1) {
-        auto empty = builder
-                         .create<tensor::EmptyOp>(loc, matmulType.getShape(),
-                                                  elementType)
+        auto empty = tensor::EmptyOp::create(builder, loc,
+                                             matmulType.getShape(), elementType)
                          .getResult();
         init = linalg::FillOp::create(builder, loc, zero, empty).getResult(0);
       }
@@ -802,9 +796,8 @@ Value outputTransform(RewriterBase &rewriter, Location loc, Value value,
           RankedTensorType::get({retRows, AMatrix.cols}, elementType);
       Value init = outInitVal;
       if (scalarFactor != 1) {
-        auto empty = builder
-                         .create<tensor::EmptyOp>(loc, matmulType.getShape(),
-                                                  elementType)
+        auto empty = tensor::EmptyOp::create(builder, loc,
+                                             matmulType.getShape(), elementType)
                          .getResult();
         init = linalg::FillOp::create(builder, loc, zero, empty).getResult(0);
       }
@@ -827,23 +820,21 @@ Value outputTransform(RewriterBase &rewriter, Location loc, Value value,
           AffineMap::get(2, 0, context), identityAffineMap, identityAffineMap};
 
       matmulRetValue =
-          rewriter
-              .create<linalg::GenericOp>(
-                  loc, matmulType,
-                  ValueRange{scalarFactorValue, matmulRetValue},
-                  ValueRange{outInitVal}, affineMaps,
-                  llvm::ArrayRef<utils::IteratorType>{
-                      utils::IteratorType::parallel,
-                      utils::IteratorType::parallel},
-                  [&](OpBuilder &nestedBuilder, Location nestedLoc,
-                      ValueRange args) {
-                    auto mulf = arith::MulFOp::create(nestedBuilder, nestedLoc,
-                                                      args[0], args[1]);
-                    auto addf = arith::AddFOp::create(
-                        nestedBuilder, nestedLoc, mulf.getResult(), args[2]);
-                    linalg::YieldOp::create(nestedBuilder, nestedLoc,
-                                            addf.getResult());
-                  })
+          linalg::GenericOp::create(
+              rewriter, loc, matmulType,
+              ValueRange{scalarFactorValue, matmulRetValue},
+              ValueRange{outInitVal}, affineMaps,
+              llvm::ArrayRef<utils::IteratorType>{
+                  utils::IteratorType::parallel, utils::IteratorType::parallel},
+              [&](OpBuilder &nestedBuilder, Location nestedLoc,
+                  ValueRange args) {
+                auto mulf = arith::MulFOp::create(nestedBuilder, nestedLoc,
+                                                  args[0], args[1]);
+                auto addf = arith::AddFOp::create(nestedBuilder, nestedLoc,
+                                                  mulf.getResult(), args[2]);
+                linalg::YieldOp::create(nestedBuilder, nestedLoc,
+                                        addf.getResult());
+              })
               .getResult(0);
     }
 
diff --git a/mlir/lib/Dialect/MemRef/Transforms/ComposeSubView.cpp b/mlir/lib/Dialect/MemRef/Transforms/ComposeSubView.cpp
index 106c3b4..cce80db 100644
--- a/mlir/lib/Dialect/MemRef/Transforms/ComposeSubView.cpp
+++ b/mlir/lib/Dialect/MemRef/Transforms/ComposeSubView.cpp
@@ -80,10 +80,6 @@ struct ComposeSubViewOpPattern : public OpRewritePattern<memref::SubViewOp> {
     for (auto &&[opOffset, sourceOffset, sourceStride, opSize] :
          llvm::zip(op.getMixedOffsets(), sourceOp.getMixedOffsets(),
                    sourceOp.getMixedStrides(), op.getMixedSizes())) {
-      // We only support static sizes.
-      if (isa<Value>(opSize)) {
-        return failure();
-      }
       sizes.push_back(opSize);
       Attribute opOffsetAttr = llvm::dyn_cast_if_present<Attribute>(opOffset),
                 sourceOffsetAttr =
diff --git a/mlir/lib/Dialect/MemRef/Transforms/IndependenceTransforms.cpp b/mlir/lib/Dialect/MemRef/Transforms/IndependenceTransforms.cpp
index 66c1aa6..d5e2b97 100644
--- a/mlir/lib/Dialect/MemRef/Transforms/IndependenceTransforms.cpp
+++ b/mlir/lib/Dialect/MemRef/Transforms/IndependenceTransforms.cpp
@@ -56,9 +56,8 @@ FailureOr<Value> memref::buildIndependentOp(OpBuilder &b,
   // Create a memref::SubViewOp.
   SmallVector<OpFoldResult> offsets(newSizes.size(), b.getIndexAttr(0));
   SmallVector<OpFoldResult> strides(newSizes.size(), b.getIndexAttr(1));
-  return b
-      .create<SubViewOp>(loc, newAllocaOp, offsets, allocaOp.getMixedSizes(),
-                         strides)
+  return SubViewOp::create(b, loc, newAllocaOp, offsets,
+                           allocaOp.getMixedSizes(), strides)
       .getResult();
 }
 
diff --git a/mlir/lib/Dialect/MemRef/Transforms/RuntimeOpVerification.cpp b/mlir/lib/Dialect/MemRef/Transforms/RuntimeOpVerification.cpp
index 1f03e9a..d3a77c0 100644
--- a/mlir/lib/Dialect/MemRef/Transforms/RuntimeOpVerification.cpp
+++ b/mlir/lib/Dialect/MemRef/Transforms/RuntimeOpVerification.cpp
@@ -185,9 +185,8 @@ struct CopyOpInterface
                             int64_t dim) -> Value {
         return type.isDynamicDim(dim)
                    ? DimOp::create(builder, loc, memRef, dim).getResult()
-                   : builder
-                         .create<arith::ConstantIndexOp>(loc,
-                                                         type.getDimSize(dim))
+                   : arith::ConstantIndexOp::create(builder, loc,
+                                                    type.getDimSize(dim))
                          .getResult();
       };
       Value sourceDim = getDimSize(copyOp.getSource(), rankedSourceType, i);
diff --git a/mlir/lib/Dialect/MemRef/Utils/MemRefUtils.cpp b/mlir/lib/Dialect/MemRef/Utils/MemRefUtils.cpp
index 97fe3cb..5af46a4 100644
--- a/mlir/lib/Dialect/MemRef/Utils/MemRefUtils.cpp
+++ b/mlir/lib/Dialect/MemRef/Utils/MemRefUtils.cpp
@@ -237,8 +237,8 @@ LogicalResult resolveSourceIndicesExpandShape(
         llvm::map_to_vector(group, [&](int64_t d) { return destShape[d]; });
     SmallVector<Value> groupIndices =
         llvm::map_to_vector(group, [&](int64_t d) { return indices[d]; });
-    Value collapsedIndex = rewriter.create<affine::AffineLinearizeIndexOp>(
-        loc, groupIndices, groupBasis, /*disjoint=*/startsInbounds);
+    Value collapsedIndex = affine::AffineLinearizeIndexOp::create(
+        rewriter, loc, groupIndices, groupBasis, /*disjoint=*/startsInbounds);
     sourceIndices.push_back(collapsedIndex);
   }
   return success();
@@ -250,8 +250,8 @@ resolveSourceIndicesCollapseShape(Location loc, PatternRewriter &rewriter,
                                   ValueRange indices,
                                   SmallVectorImpl<Value> &sourceIndices) {
   // Note: collapse_shape requires a strided memref, we can do this.
-  auto metadata = rewriter.create<memref::ExtractStridedMetadataOp>(
-      loc, collapseShapeOp.getSrc());
+  auto metadata = memref::ExtractStridedMetadataOp::create(
+      rewriter, loc, collapseShapeOp.getSrc());
   SmallVector<OpFoldResult> sourceSizes = metadata.getConstifiedMixedSizes();
   for (auto [index, group] :
        llvm::zip(indices, collapseShapeOp.getReassociationIndices())) {
@@ -265,8 +265,8 @@ resolveSourceIndicesCollapseShape(Location loc, PatternRewriter &rewriter,
 
     SmallVector<OpFoldResult> basis =
         llvm::map_to_vector(group, [&](int64_t d) { return sourceSizes[d]; });
-    auto delinearize = rewriter.create<affine::AffineDelinearizeIndexOp>(
-        loc, index, basis, /*hasOuterBound=*/true);
+    auto delinearize = affine::AffineDelinearizeIndexOp::create(
+        rewriter, loc, index, basis, /*hasOuterBound=*/true);
     llvm::append_range(sourceIndices, delinearize.getResults());
   }
   if (collapseShapeOp.getReassociationIndices().empty()) {
diff --git a/mlir/lib/Dialect/NVGPU/TransformOps/NVGPUTransformOps.cpp b/mlir/lib/Dialect/NVGPU/TransformOps/NVGPUTransformOps.cpp
index f5f0bfa..bc3e8b2 100644
--- a/mlir/lib/Dialect/NVGPU/TransformOps/NVGPUTransformOps.cpp
+++ b/mlir/lib/Dialect/NVGPU/TransformOps/NVGPUTransformOps.cpp
@@ -38,9 +38,6 @@ using namespace mlir::NVVM;
 using namespace mlir::transform;
 
 #define DEBUG_TYPE "nvgpu-transforms"
-#define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE "]: ")
-#define DBGSNL() (llvm::dbgs() << "\n")
-#define LDBG(X) LLVM_DEBUG(DBGS() << (X) << "\n")
 
 //===----------------------------------------------------------------------===//
 // Apply...ConversionPatternsOp
diff --git a/mlir/lib/Dialect/Quant/Transforms/LowerQuantOps.cpp b/mlir/lib/Dialect/Quant/Transforms/LowerQuantOps.cpp
index 58cd160..9e37bc5 100644
--- a/mlir/lib/Dialect/Quant/Transforms/LowerQuantOps.cpp
+++ b/mlir/lib/Dialect/Quant/Transforms/LowerQuantOps.cpp
@@ -148,16 +148,14 @@ flattenUnrankedTensorAroundAxis(OpBuilder &builder, Location loc, Value input,
   auto axisValue = arith::ConstantIndexOp::create(builder, loc, axis);
   auto axisNextValue = arith::ConstantIndexOp::create(builder, loc, axis + 1);
   auto shapeLeft =
-      builder
-          .create<shape::SplitAtOp>(loc, TypeRange{shapeType, shapeType},
-                                    inputShape, axisValue)
+      shape::SplitAtOp::create(builder, loc, TypeRange{shapeType, shapeType},
+                               inputShape, axisValue)
           .getResult(0);
   auto sizeLeft =
       shape::NumElementsOp::create(builder, loc, indexType, shapeLeft);
   auto shapeRight =
-      builder
-          .create<shape::SplitAtOp>(loc, TypeRange{shapeType, shapeType},
-                                    inputShape, axisNextValue)
+      shape::SplitAtOp::create(builder, loc, TypeRange{shapeType, shapeType},
+                               inputShape, axisNextValue)
           .getResult(1);
   auto sizeRight =
       shape::NumElementsOp::create(builder, loc, indexType, shapeRight);
@@ -557,25 +555,24 @@ Value convertPerChannelRanked(OpBuilder &builder, Location loc, Operation *op,
   SmallVector<AffineMap> indexingMaps{
       builder.getMultiDimIdentityMap(inputRank), channelAxisAffineMap,
       channelAxisAffineMap, builder.getMultiDimIdentityMap(inputRank)};
-  auto result = builder
-                    .create<linalg::GenericOp>(
-                        loc,
-                        init.getType(),                        // resultType
-                        ValueRange{input, scales, zeroPoints}, // inputs
-                        ValueRange{init},                      // outputs
-                        indexingMaps, iteratorTypes,
-                        [&](OpBuilder &builder, Location loc, ValueRange args) {
-                          assert(args.size() == 4);
-                          auto input = args[0];
-                          auto scale = args[1];
-                          auto zeroPoint = args[2];
-
-                          auto result =
-                              convertRanked(builder, loc, op, input, {}, scale,
-                                            zeroPoint, quantizedType);
-
-                          linalg::YieldOp::create(builder, loc, result);
-                        })
+  auto result = linalg::GenericOp::create(
+                    builder, loc,
+                    init.getType(),                        // resultType
+                    ValueRange{input, scales, zeroPoints}, // inputs
+                    ValueRange{init},                      // outputs
+                    indexingMaps, iteratorTypes,
+                    [&](OpBuilder &builder, Location loc, ValueRange args) {
+                      assert(args.size() == 4);
+                      auto input = args[0];
+                      auto scale = args[1];
+                      auto zeroPoint = args[2];
+
+                      auto result =
+                          convertRanked(builder, loc, op, input, {}, scale,
+                                        zeroPoint, quantizedType);
+
+                      linalg::YieldOp::create(builder, loc, result);
+                    })
                     .getResult(0);
 
   return result;
@@ -660,25 +657,24 @@ Value convertSubChannel(OpBuilder &builder, Location loc, Operation *op,
   SmallVector<AffineMap> indexingMaps{
       builder.getMultiDimIdentityMap(inputRank), affineMap, affineMap,
       builder.getMultiDimIdentityMap(inputRank)};
-  auto result = builder
-                    .create<linalg::GenericOp>(
-                        loc,
-                        init.getType(),                        // resultType
-                        ValueRange{input, scales, zeroPoints}, // inputs
-                        ValueRange{init},                      // outputs
-                        indexingMaps, iteratorTypes,
-                        [&](OpBuilder &builder, Location loc, ValueRange args) {
-                          assert(args.size() == 4);
-                          auto input = args[0];
-                          auto scale = args[1];
-                          auto zeroPoint = args[2];
-
-                          auto result =
-                              convertRanked(builder, loc, op, input, {}, scale,
-                                            zeroPoint, quantizedType);
-
-                          linalg::YieldOp::create(builder, loc, result);
-                        })
+  auto result = linalg::GenericOp::create(
+                    builder, loc,
+                    init.getType(),                        // resultType
+                    ValueRange{input, scales, zeroPoints}, // inputs
+                    ValueRange{init},                      // outputs
+                    indexingMaps, iteratorTypes,
+                    [&](OpBuilder &builder, Location loc, ValueRange args) {
+                      assert(args.size() == 4);
+                      auto input = args[0];
+                      auto scale = args[1];
+                      auto zeroPoint = args[2];
+
+                      auto result =
+                          convertRanked(builder, loc, op, input, {}, scale,
+                                        zeroPoint, quantizedType);
+
+                      linalg::YieldOp::create(builder, loc, result);
+                    })
                     .getResult(0);
 
   return result;
diff --git a/mlir/lib/Dialect/SCF/Transforms/BufferizableOpInterfaceImpl.cpp b/mlir/lib/Dialect/SCF/Transforms/BufferizableOpInterfaceImpl.cpp
index 64c4d60..f8799c5 100644
--- a/mlir/lib/Dialect/SCF/Transforms/BufferizableOpInterfaceImpl.cpp
+++ b/mlir/lib/Dialect/SCF/Transforms/BufferizableOpInterfaceImpl.cpp
@@ -497,10 +497,10 @@ getBbArgReplacements(RewriterBase &rewriter, Block::BlockArgListType bbArgs,
     size_t idx = it.index();
     Value val = it.value();
     if (tensorIndices.contains(idx)) {
-      result.push_back(rewriter
-                           .create<bufferization::ToTensorOp>(
-                               val.getLoc(), oldBbArgs[idx].getType(), val)
-                           .getResult());
+      result.push_back(
+          bufferization::ToTensorOp::create(rewriter, val.getLoc(),
+                                            oldBbArgs[idx].getType(), val)
+              .getResult());
     } else {
       result.push_back(val);
     }
diff --git a/mlir/lib/Dialect/SCF/Transforms/LoopPipelining.cpp b/mlir/lib/Dialect/SCF/Transforms/LoopPipelining.cpp
index 5982856..1130538 100644
--- a/mlir/lib/Dialect/SCF/Transforms/LoopPipelining.cpp
+++ b/mlir/lib/Dialect/SCF/Transforms/LoopPipelining.cpp
@@ -19,12 +19,10 @@
 #include "mlir/IR/PatternMatch.h"
 #include "mlir/Transforms/RegionUtils.h"
 #include "llvm/ADT/MapVector.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 #include "llvm/Support/MathExtras.h"
 
 #define DEBUG_TYPE "scf-loop-pipelining"
-#define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE "]: ")
-#define LDBG(X) LLVM_DEBUG(DBGS() << X << "\n")
 
 using namespace mlir;
 using namespace mlir::scf;
@@ -100,7 +98,7 @@ public:
 
 bool LoopPipelinerInternal::initializeLoopInfo(
     ForOp op, const PipeliningOption &options) {
-  LDBG("Start initializeLoopInfo");
+  LDBG() << "Start initializeLoopInfo";
   forOp = op;
   ub = forOp.getUpperBound();
   lb = forOp.getLowerBound();
@@ -109,7 +107,7 @@ bool LoopPipelinerInternal::initializeLoopInfo(
   std::vector<std::pair<Operation *, unsigned>> schedule;
   options.getScheduleFn(forOp, schedule);
   if (schedule.empty()) {
-    LDBG("--empty schedule -> BAIL");
+    LDBG() << "--empty schedule -> BAIL";
     return false;
   }
 
@@ -126,7 +124,7 @@ bool LoopPipelinerInternal::initializeLoopInfo(
   auto stepCst = getConstantIntValue(step);
   if (!upperBoundCst || !lowerBoundCst || !stepCst) {
     if (!options.supportDynamicLoops) {
-      LDBG("--dynamic loop not supported -> BAIL");
+      LDBG() << "--dynamic loop not supported -> BAIL";
       return false;
     }
   } else {
@@ -134,21 +132,21 @@ bool LoopPipelinerInternal::initializeLoopInfo(
     int64_t lbImm = lowerBoundCst.value();
     int64_t stepImm = stepCst.value();
     if (stepImm <= 0) {
-      LDBG("--invalid loop step -> BAIL");
+      LDBG() << "--invalid loop step -> BAIL";
       return false;
     }
     int64_t numIteration = llvm::divideCeilSigned(ubImm - lbImm, stepImm);
     if (numIteration >= maxStage) {
       dynamicLoop = false;
     } else if (!options.supportDynamicLoops) {
-      LDBG("--fewer loop iterations than pipeline stages -> BAIL");
+      LDBG() << "--fewer loop iterations than pipeline stages -> BAIL";
       return false;
     }
   }
   peelEpilogue = options.peelEpilogue;
   predicateFn = options.predicateFn;
   if ((!peelEpilogue || dynamicLoop) && predicateFn == nullptr) {
-    LDBG("--no epilogue or predicate set -> BAIL");
+    LDBG() << "--no epilogue or predicate set -> BAIL";
     return false;
   }
 
@@ -156,13 +154,13 @@ bool LoopPipelinerInternal::initializeLoopInfo(
   for (Operation &op : forOp.getBody()->without_terminator()) {
     if (!stages.contains(&op)) {
       op.emitOpError("not assigned a pipeline stage");
-      LDBG("--op not assigned a pipeline stage: " << op << " -> BAIL");
+      LDBG() << "--op not assigned a pipeline stage: " << op << " -> BAIL";
       return false;
     }
   }
 
   if (!verifySchedule()) {
-    LDBG("--invalid schedule: " << op << " -> BAIL");
+    LDBG() << "--invalid schedule: " << op << " -> BAIL";
     return false;
   }
 
@@ -173,15 +171,16 @@ bool LoopPipelinerInternal::initializeLoopInfo(
     (void)stageNum;
     if (op == forOp.getBody()->getTerminator()) {
       op->emitError("terminator should not be assigned a stage");
-      LDBG("--terminator should not be assigned stage: " << *op << " -> BAIL");
+      LDBG() << "--terminator should not be assigned stage: " << *op
+             << " -> BAIL";
       return false;
     }
     if (op->getBlock() != forOp.getBody()) {
       op->emitOpError("the owning Block of all operations assigned a stage "
                       "should be the loop body block");
-      LDBG("--the owning Block of all operations assigned a stage "
-           "should be the loop body block: "
-           << *op << " -> BAIL");
+      LDBG() << "--the owning Block of all operations assigned a stage "
+                "should be the loop body block: "
+             << *op << " -> BAIL";
       return false;
     }
   }
@@ -196,8 +195,8 @@ bool LoopPipelinerInternal::initializeLoopInfo(
                      return !def ||
                             (!stages.contains(def) && forOp->isAncestor(def));
                    })) {
-    LDBG("--only support loop carried dependency with a distance of 1 or "
-         "defined outside of the loop -> BAIL");
+    LDBG() << "--only support loop carried dependency with a distance of 1 or "
+              "defined outside of the loop -> BAIL";
     return false;
   }
   annotateFn = options.annotateFn;
diff --git a/mlir/lib/Dialect/SCF/Utils/Utils.cpp b/mlir/lib/Dialect/SCF/Utils/Utils.cpp
index 4025ec6..5731795 100644
--- a/mlir/lib/Dialect/SCF/Utils/Utils.cpp
+++ b/mlir/lib/Dialect/SCF/Utils/Utils.cpp
@@ -24,14 +24,12 @@
 #include "mlir/Transforms/RegionUtils.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/ADT/SmallVector.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 #include <cstdint>
 
 using namespace mlir;
 
 #define DEBUG_TYPE "scf-utils"
-#define DBGS() (llvm::dbgs() << '[' << DEBUG_TYPE << "] ")
-#define LDBG(X) LLVM_DEBUG(DBGS() << X << "\n")
 
 SmallVector<scf::ForOp> mlir::replaceLoopNestWithNewYields(
     RewriterBase &rewriter, MutableArrayRef<scf::ForOp> loopNest,
@@ -525,13 +523,13 @@ LogicalResult mlir::loopUnrollJamByFactor(scf::ForOp forOp,
   // If any control operand of any inner loop of `forOp` is defined within
   // `forOp`, no unroll jam.
   if (!areInnerBoundsInvariant(forOp)) {
-    LDBG("failed to unroll and jam: inner bounds are not invariant");
+    LDBG() << "failed to unroll and jam: inner bounds are not invariant";
     return failure();
   }
 
   // Currently, for operations with results are not supported.
   if (forOp->getNumResults() > 0) {
-    LDBG("failed to unroll and jam: unsupported loop with results");
+    LDBG() << "failed to unroll and jam: unsupported loop with results";
     return failure();
   }
 
@@ -540,16 +538,17 @@ LogicalResult mlir::loopUnrollJamByFactor(scf::ForOp forOp,
   std::optional<uint64_t> tripCount = getConstantTripCount(forOp);
   if (!tripCount.has_value()) {
     // If the trip count is dynamic, do not unroll & jam.
-    LDBG("failed to unroll and jam: trip count could not be determined");
+    LDBG() << "failed to unroll and jam: trip count could not be determined";
     return failure();
   }
   if (unrollJamFactor > *tripCount) {
-    LDBG("unroll and jam factor is greater than trip count, set factor to trip "
-         "count");
+    LDBG() << "unroll and jam factor is greater than trip count, set factor to "
+              "trip "
+              "count";
     unrollJamFactor = *tripCount;
   } else if (*tripCount % unrollJamFactor != 0) {
-    LDBG("failed to unroll and jam: unsupported trip count that is not a "
-         "multiple of unroll jam factor");
+    LDBG() << "failed to unroll and jam: unsupported trip count that is not a "
+              "multiple of unroll jam factor";
     return failure();
   }
 
@@ -828,9 +827,8 @@ static Value getProductOfIntsOrIndexes(RewriterBase &rewriter, Location loc,
       productOf = v;
   }
   if (!productOf) {
-    productOf = rewriter
-                    .create<arith::ConstantOp>(
-                        loc, rewriter.getOneAttr(getType(values.front())))
+    productOf = arith::ConstantOp::create(
+                    rewriter, loc, rewriter.getOneAttr(getType(values.front())))
                     .getResult();
   }
   return productOf.value();
diff --git a/mlir/lib/Dialect/SPIRV/IR/SPIRVDialect.cpp b/mlir/lib/Dialect/SPIRV/IR/SPIRVDialect.cpp
index f2f7f70..9bee200 100644
--- a/mlir/lib/Dialect/SPIRV/IR/SPIRVDialect.cpp
+++ b/mlir/lib/Dialect/SPIRV/IR/SPIRVDialect.cpp
@@ -92,11 +92,13 @@ struct SPIRVInlinerInterface : public DialectInlinerInterface {
   /// as necessary.
   void handleTerminator(Operation *op, Block *newDest) const final {
     if (auto returnOp = dyn_cast<spirv::ReturnOp>(op)) {
-      OpBuilder(op).create<spirv::BranchOp>(op->getLoc(), newDest);
+      auto builder = OpBuilder(op);
+      spirv::BranchOp::create(builder, op->getLoc(), newDest);
       op->erase();
     } else if (auto retValOp = dyn_cast<spirv::ReturnValueOp>(op)) {
-      OpBuilder(op).create<spirv::BranchOp>(retValOp->getLoc(), newDest,
-                                            retValOp->getOperands());
+      auto builder = OpBuilder(op);
+      spirv::BranchOp::create(builder, retValOp->getLoc(), newDest,
+                              retValOp->getOperands());
       op->erase();
     }
   }
diff --git a/mlir/lib/Dialect/Shape/IR/Shape.cpp b/mlir/lib/Dialect/Shape/IR/Shape.cpp
index e24f0f8..5ba8289 100644
--- a/mlir/lib/Dialect/Shape/IR/Shape.cpp
+++ b/mlir/lib/Dialect/Shape/IR/Shape.cpp
@@ -1702,9 +1702,8 @@ struct ShapeOfOpToConstShapeOp : public OpRewritePattern<shape::ShapeOfOp> {
       return failure();
     Location loc = op.getLoc();
     Value constShape =
-        rewriter
-            .create<ConstShapeOp>(loc,
-                                  rewriter.getIndexTensorAttr(type.getShape()))
+        ConstShapeOp::create(rewriter, loc,
+                             rewriter.getIndexTensorAttr(type.getShape()))
             .getResult();
     if (constShape.getType() != op.getResult().getType())
       constShape = tensor::CastOp::create(rewriter, loc,
diff --git a/mlir/lib/Dialect/Mesh/CMakeLists.txt b/mlir/lib/Dialect/Shard/CMakeLists.txt
index fa8842f..fa8842f 100644
--- a/mlir/lib/Dialect/Mesh/CMakeLists.txt
+++ b/mlir/lib/Dialect/Shard/CMakeLists.txt
diff --git a/mlir/lib/Dialect/Mesh/IR/CMakeLists.txt b/mlir/lib/Dialect/Shard/IR/CMakeLists.txt
index 3fea4d6..70c604988 100644
--- a/mlir/lib/Dialect/Mesh/IR/CMakeLists.txt
+++ b/mlir/lib/Dialect/Shard/IR/CMakeLists.txt
@@ -1,11 +1,11 @@
-add_mlir_dialect_library(MLIRMeshDialect
-  MeshOps.cpp
+add_mlir_dialect_library(MLIRShardDialect
+  ShardOps.cpp
 
   ADDITIONAL_HEADER_DIRS
-  ${MLIR_MAIN_INCLUDE_DIR}/mlir/Dialect/Mesh
+  ${MLIR_MAIN_INCLUDE_DIR}/mlir/Dialect/Shard
 
   DEPENDS
-  MLIRMeshIncGen
+  MLIRShardIncGen
 
   LINK_LIBS PUBLIC
   MLIRArithDialect
diff --git a/mlir/lib/Dialect/Mesh/IR/MeshOps.cpp b/mlir/lib/Dialect/Shard/IR/ShardOps.cpp
index 01426e4..e5a3b5d 100644
--- a/mlir/lib/Dialect/Mesh/IR/MeshOps.cpp
+++ b/mlir/lib/Dialect/Shard/IR/ShardOps.cpp
@@ -1,4 +1,4 @@
-//===- MeshOps.cpp - Mesh Dialect Operations ------------------------------===//
+//===- ShardOps.cpp - Shard Dialect Operations ----------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -6,10 +6,10 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "mlir/Dialect/Mesh/IR/MeshOps.h"
+#include "mlir/Dialect/Shard/IR/ShardOps.h"
 
 #include "mlir/Dialect/Arith/IR/Arith.h"
-#include "mlir/Dialect/Mesh/IR/MeshDialect.h"
+#include "mlir/Dialect/Shard/IR/ShardDialect.h"
 #include "mlir/Dialect/Utils/StaticValueUtils.h"
 #include "mlir/IR/Attributes.h"
 #include "mlir/IR/BuiltinAttributes.h"
@@ -37,13 +37,13 @@
 #include <optional>
 #include <utility>
 
-#define DEBUG_TYPE "mesh-ops"
+#define DEBUG_TYPE "shard-ops"
 #define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE << "]: ")
 
 using namespace mlir;
-using namespace mlir::mesh;
+using namespace mlir::shard;
 
-#include "mlir/Dialect/Mesh/IR/MeshDialect.cpp.inc"
+#include "mlir/Dialect/Shard/IR/ShardDialect.cpp.inc"
 
 namespace {
 
@@ -74,11 +74,10 @@ static DimensionSize operator*(DimensionSize lhs, DimensionSize rhs) {
   return lhs.value() * rhs.value();
 }
 
-SmallVector<Value> mlir::mesh::getMixedAsValues(OpBuilder b,
-                                                const Location &loc,
-                                                llvm::ArrayRef<int64_t> statics,
-                                                ValueRange dynamics,
-                                                Type type) {
+SmallVector<Value>
+mlir::shard::getMixedAsValues(OpBuilder b, const Location &loc,
+                              llvm::ArrayRef<int64_t> statics,
+                              ValueRange dynamics, Type type) {
   SmallVector<Value> values;
   auto dyn = dynamics.begin();
   Type i64 = b.getI64Type();
@@ -102,7 +101,7 @@ SmallVector<Value> mlir::mesh::getMixedAsValues(OpBuilder b,
 //===----------------------------------------------------------------------===//
 
 namespace {
-struct MeshInlinerInterface : public DialectInlinerInterface {
+struct ShardInlinerinterface : public DialectInlinerInterface {
   using DialectInlinerInterface::DialectInlinerInterface;
   // Currently no restrictions are encoded for inlining.
   bool isLegalToInline(Operation *, Operation *, bool) const final {
@@ -118,44 +117,45 @@ struct MeshInlinerInterface : public DialectInlinerInterface {
 } // namespace
 
 //===----------------------------------------------------------------------===//
-// Mesh dialect
+// Shard dialect
 //===----------------------------------------------------------------------===//
 
-void MeshDialect::initialize() {
+void ShardDialect::initialize() {
   addOperations<
 #define GET_OP_LIST
-#include "mlir/Dialect/Mesh/IR/MeshOps.cpp.inc"
+#include "mlir/Dialect/Shard/IR/ShardOps.cpp.inc"
       >();
   addAttributes<
 #define GET_ATTRDEF_LIST
-#include "mlir/Dialect/Mesh/IR/MeshAttributes.cpp.inc"
+#include "mlir/Dialect/Shard/IR/ShardAttributes.cpp.inc"
       >();
   addTypes<
 #define GET_TYPEDEF_LIST
-#include "mlir/Dialect/Mesh/IR/MeshTypes.cpp.inc"
+#include "mlir/Dialect/Shard/IR/ShardTypes.cpp.inc"
       >();
-  addInterface<MeshInlinerInterface>();
+  addInterface<ShardInlinerinterface>();
 }
 
-Operation *MeshDialect::materializeConstant(OpBuilder &builder, Attribute value,
-                                            Type type, Location loc) {
+Operation *ShardDialect::materializeConstant(OpBuilder &builder,
+                                             Attribute value, Type type,
+                                             Location loc) {
   return arith::ConstantOp::materialize(builder, value, type, loc);
 }
 
 //===----------------------------------------------------------------------===//
-// Mesh utilities
+// Shard utilities
 //===----------------------------------------------------------------------===//
 
-static FailureOr<MeshOp> getMeshAndVerify(Operation *op,
-                                          FlatSymbolRefAttr meshSymbol,
+static FailureOr<GridOp> getGridAndVerify(Operation *op,
+                                          FlatSymbolRefAttr gridSymbol,
                                           SymbolTableCollection &symbolTable) {
-  mesh::MeshOp mesh = getMeshOrNull(op, meshSymbol, symbolTable);
-  if (!mesh) {
-    return op->emitError() << "Undefined required mesh symbol \""
-                           << meshSymbol.getValue() << "\".";
+  shard::GridOp grid = getGridOrNull(op, gridSymbol, symbolTable);
+  if (!grid) {
+    return op->emitError() << "Undefined required grid symbol \""
+                           << gridSymbol.getValue() << "\".";
   }
 
-  return mesh;
+  return grid;
 }
 
 template <typename It>
@@ -175,20 +175,20 @@ bool isUnique(It begin, It end) {
   return true;
 }
 
-static LogicalResult verifyMeshAxes(Location loc, ArrayRef<MeshAxis> axes,
-                                    MeshOp mesh) {
-  SmallVector<MeshAxis> sorted = llvm::to_vector(axes);
+static LogicalResult verifyGridAxes(Location loc, ArrayRef<GridAxis> axes,
+                                    GridOp grid) {
+  SmallVector<GridAxis> sorted = llvm::to_vector(axes);
   llvm::sort(sorted);
   if (!isUnique(sorted.begin(), sorted.end())) {
-    return emitError(loc) << "Mesh axes contains duplicate elements.";
+    return emitError(loc) << "Grid axes contains duplicate elements.";
   }
 
-  MeshAxis rank = mesh.getRank();
+  GridAxis rank = grid.getRank();
   for (auto axis : axes) {
     if (axis >= rank || axis < 0) {
       return emitError(loc)
-             << "0-based mesh axis index " << axis
-             << " is out of bounds. The referenced mesh \"" << mesh.getSymName()
+             << "0-based grid axis index " << axis
+             << " is out of bounds. The referenced grid \"" << grid.getSymName()
              << "\" is of rank " << rank << ".";
     }
   }
@@ -197,22 +197,22 @@ static LogicalResult verifyMeshAxes(Location loc, ArrayRef<MeshAxis> axes,
 }
 
 template <typename Op>
-static FailureOr<MeshOp>
-getMeshAndVerifyAxes(Op op, SymbolTableCollection &symbolTable) {
-  auto mesh =
-      ::getMeshAndVerify(op.getOperation(), op.getMeshAttr(), symbolTable);
-  if (failed(mesh)) {
+static FailureOr<GridOp>
+getGridAndVerifyAxes(Op op, SymbolTableCollection &symbolTable) {
+  auto grid =
+      ::getGridAndVerify(op.getOperation(), op.getGridAttr(), symbolTable);
+  if (failed(grid)) {
     return failure();
   }
-  if (failed(verifyMeshAxes(op.getLoc(), op.getMeshAxes(), mesh.value()))) {
+  if (failed(verifyGridAxes(op.getLoc(), op.getGridAxes(), grid.value()))) {
     return failure();
   }
-  return mesh;
+  return grid;
 }
 
-template <typename InShape, typename MeshShape, typename SplitAxes,
+template <typename InShape, typename GridShape, typename SplitAxes,
           typename OutShape>
-static void shardShape(const InShape &inShape, const MeshShape &meshShape,
+static void shardShape(const InShape &inShape, const GridShape &gridShape,
                        const SplitAxes &splitAxes, OutShape &outShape,
                        ArrayRef<int64_t> shardedDimsOffsets = {},
                        ArrayRef<int64_t> haloSizes = {}) {
@@ -226,7 +226,7 @@ static void shardShape(const InShape &inShape, const MeshShape &meshShape,
             llvm::adl_begin(outShape));
 
   if (!shardedDimsOffsets.empty()) {
-    auto isDynShape = ShapedType::isDynamicShape(meshShape);
+    auto isDynShape = ShapedType::isDynamicShape(gridShape);
     uint64_t pos = 1;
     for (auto [tensorAxis, innerSplitAxes] : llvm::enumerate(splitAxes)) {
       if (!innerSplitAxes.empty()) {
@@ -238,7 +238,7 @@ static void shardShape(const InShape &inShape, const MeshShape &meshShape,
           // non-uniform offs in shardedDimsOffsets.
           uint64_t numShards = 0;
           for (auto i : innerSplitAxes.asArrayRef()) {
-            numShards += meshShape[i];
+            numShards += gridShape[i];
           }
           for (size_t i = 1; i < numShards; ++i) {
             if (shardedDimsOffsets[pos + i] - shardedDimsOffsets[pos + i - 1] !=
@@ -256,7 +256,7 @@ static void shardShape(const InShape &inShape, const MeshShape &meshShape,
     for (auto [tensorAxis, innerSplitAxes] : llvm::enumerate(splitAxes)) {
       outShape[tensorAxis] = shardDimension(
           inShape[tensorAxis],
-          collectiveProcessGroupSize(innerSplitAxes.asArrayRef(), meshShape));
+          collectiveProcessGroupSize(innerSplitAxes.asArrayRef(), gridShape));
     }
 
     if (!haloSizes.empty()) {
@@ -279,25 +279,25 @@ static void shardShape(const InShape &inShape, const MeshShape &meshShape,
   }
 }
 
-ShapedType mesh::shardShapedType(ShapedType shape, MeshOp mesh,
-                                 MeshSharding sharding) {
+ShapedType shard::shardShapedType(ShapedType shape, GridOp grid,
+                                  Sharding sharding) {
   using Dim = std::decay_t<decltype(shape.getDimSize(0))>;
   SmallVector<Dim> resShapeArr(shape.getShape().size());
-  shardShape(shape.getShape(), mesh.getShape(), sharding.getSplitAxes(),
+  shardShape(shape.getShape(), grid.getShape(), sharding.getSplitAxes(),
              resShapeArr, sharding.getStaticShardedDimsOffsets(),
              sharding.getStaticHaloSizes());
   return shape.clone(resShapeArr);
 }
 
-Type mesh::shardType(Type type, MeshOp mesh, MeshSharding sharding) {
+Type shard::shardType(Type type, GridOp grid, Sharding sharding) {
   RankedTensorType rankedTensorType = dyn_cast<RankedTensorType>(type);
   if (rankedTensorType && !rankedTensorType.getShape().empty()) {
-    return shardShapedType(rankedTensorType, mesh, sharding);
+    return shardShapedType(rankedTensorType, grid, sharding);
   }
   return type;
 }
 
-static void maybeInsertTargetShardingAnnotationImpl(MeshSharding sharding,
+static void maybeInsertTargetShardingAnnotationImpl(Sharding sharding,
                                                     Value &operandValue,
                                                     Operation *operandOp,
                                                     OpBuilder &builder,
@@ -336,9 +336,9 @@ static void maybeInsertTargetShardingAnnotationImpl(MeshSharding sharding,
   newShardOp.getResult().replaceAllUsesExcept(newShardOp2, newShardOp2);
 }
 
-void mlir::mesh::maybeInsertTargetShardingAnnotation(MeshSharding sharding,
-                                                     OpResult result,
-                                                     OpBuilder &builder) {
+void mlir::shard::maybeInsertTargetShardingAnnotation(Sharding sharding,
+                                                      OpResult result,
+                                                      OpBuilder &builder) {
   ShardOp newShardOp;
   SmallVector<std::pair<Value, Operation *>> uses;
   for (auto &use : result.getUses()) {
@@ -350,9 +350,9 @@ void mlir::mesh::maybeInsertTargetShardingAnnotation(MeshSharding sharding,
   }
 }
 
-void mlir::mesh::maybeInsertSourceShardingAnnotation(MeshSharding sharding,
-                                                     OpOperand &operand,
-                                                     OpBuilder &builder) {
+void mlir::shard::maybeInsertSourceShardingAnnotation(Sharding sharding,
+                                                      OpOperand &operand,
+                                                      OpBuilder &builder) {
   OpBuilder::InsertionGuard insertionGuard(builder);
   Value operandValue = operand.get();
   Operation *operandSrcOp = operandValue.getDefiningOp();
@@ -404,18 +404,18 @@ void mlir::mesh::maybeInsertSourceShardingAnnotation(MeshSharding sharding,
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.mesh op
+// shard.grid op
 //===----------------------------------------------------------------------===//
 
-LogicalResult MeshOp::verify() {
+LogicalResult GridOp::verify() {
   int64_t rank = getRank();
 
   if (rank <= 0)
-    return emitOpError("rank of mesh is expected to be a positive integer");
+    return emitOpError("rank of grid is expected to be a positive integer");
 
   for (int64_t dimSize : getShape()) {
     if (dimSize < 0 && ShapedType::isStatic(dimSize))
-      return emitOpError("dimension size of a mesh is expected to be "
+      return emitOpError("dimension size of a grid is expected to be "
                          "non-negative or dynamic");
   }
 
@@ -423,21 +423,21 @@ LogicalResult MeshOp::verify() {
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.mesh_shape op
+// shard.grid_shape op
 //===----------------------------------------------------------------------===//
 
 LogicalResult
-MeshShapeOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
-  auto mesh = ::getMeshAndVerify(getOperation(), getMeshAttr(), symbolTable);
-  if (failed(mesh)) {
+GridShapeOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
+  auto grid = ::getGridAndVerify(getOperation(), getGridAttr(), symbolTable);
+  if (failed(grid)) {
     return failure();
   }
-  if (failed(verifyMeshAxes(getLoc(), getAxes(), mesh.value()))) {
+  if (failed(verifyGridAxes(getLoc(), getAxes(), grid.value()))) {
     return failure();
   }
 
   size_t expectedResultsCount =
-      getAxes().empty() ? mesh->getRank() : getAxes().size();
+      getAxes().empty() ? grid->getRank() : getAxes().size();
   if (getResult().size() != expectedResultsCount) {
     return emitError() << "Unexpected number of results " << getResult().size()
                        << ". Expected " << expectedResultsCount << ".";
@@ -446,53 +446,53 @@ MeshShapeOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
   return success();
 }
 
-void MeshShapeOp::build(OpBuilder &odsBuilder, OperationState &odsState,
-                        MeshOp mesh) {
-  build(odsBuilder, odsState, mesh, SmallVector<MeshAxis>());
+void GridShapeOp::build(OpBuilder &odsBuilder, OperationState &odsState,
+                        GridOp grid) {
+  build(odsBuilder, odsState, grid, SmallVector<GridAxis>());
 }
 
-void MeshShapeOp::build(OpBuilder &odsBuilder, OperationState &odsState,
-                        MeshOp mesh, ArrayRef<MeshAxis> axes) {
+void GridShapeOp::build(OpBuilder &odsBuilder, OperationState &odsState,
+                        GridOp grid, ArrayRef<GridAxis> axes) {
   build(odsBuilder, odsState,
-        SmallVector<Type>(axes.empty() ? mesh.getRank() : axes.size(),
+        SmallVector<Type>(axes.empty() ? grid.getRank() : axes.size(),
                           odsBuilder.getIndexType()),
-        mesh.getSymName(), MeshAxesAttr::get(odsBuilder.getContext(), axes));
+        grid.getSymName(), GridAxesAttr::get(odsBuilder.getContext(), axes));
 }
 
-void MeshShapeOp::build(OpBuilder &odsBuilder, OperationState &odsState,
-                        StringRef mesh, ArrayRef<MeshAxis> axes) {
+void GridShapeOp::build(OpBuilder &odsBuilder, OperationState &odsState,
+                        StringRef grid, ArrayRef<GridAxis> axes) {
   assert(!axes.empty());
   build(odsBuilder, odsState,
-        SmallVector<Type>(axes.size(), odsBuilder.getIndexType()), mesh,
-        MeshAxesAttr::get(odsBuilder.getContext(), axes));
+        SmallVector<Type>(axes.size(), odsBuilder.getIndexType()), grid,
+        GridAxesAttr::get(odsBuilder.getContext(), axes));
 }
 
-void MeshShapeOp::getAsmResultNames(
+void GridShapeOp::getAsmResultNames(
     function_ref<void(Value, StringRef)> setNameFn) {
-  setNameFn(getResults()[0], "mesh_shape");
+  setNameFn(getResults()[0], "grid_shape");
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.sharding
+// shard.sharding
 //===----------------------------------------------------------------------===//
 
 void ShardingOp::build(::mlir::OpBuilder &b, ::mlir::OperationState &odsState,
-                       FlatSymbolRefAttr mesh,
-                       ArrayRef<MeshAxesAttr> split_axes,
+                       FlatSymbolRefAttr grid,
+                       ArrayRef<GridAxesAttr> split_axes,
                        ArrayRef<int64_t> static_halos,
                        ArrayRef<int64_t> static_offsets) {
   return build(
-      b, odsState, mesh, MeshAxesArrayAttr::get(b.getContext(), split_axes),
+      b, odsState, grid, GridAxesArrayAttr::get(b.getContext(), split_axes),
       ::mlir::DenseI64ArrayAttr::get(b.getContext(), static_halos), {},
       ::mlir::DenseI64ArrayAttr::get(b.getContext(), static_offsets), {});
 }
 
 void ShardingOp::build(::mlir::OpBuilder &b, ::mlir::OperationState &odsState,
-                       llvm::StringRef mesh, ArrayRef<MeshAxesAttr> split_axes,
+                       llvm::StringRef grid, ArrayRef<GridAxesAttr> split_axes,
                        ArrayRef<int64_t> static_halos,
                        ArrayRef<int64_t> static_offsets) {
-  return build(b, odsState, FlatSymbolRefAttr::get(b.getContext(), mesh),
-               MeshAxesArrayAttr::get(b.getContext(), split_axes),
+  return build(b, odsState, FlatSymbolRefAttr::get(b.getContext(), grid),
+               GridAxesArrayAttr::get(b.getContext(), split_axes),
                ::mlir::DenseI64ArrayAttr::get(b.getContext(), static_halos), {},
                ::mlir::DenseI64ArrayAttr::get(b.getContext(), static_offsets),
                {});
@@ -500,7 +500,7 @@ void ShardingOp::build(::mlir::OpBuilder &b, ::mlir::OperationState &odsState,
 
 void ShardingOp::build(
     ::mlir::OpBuilder &b, ::mlir::OperationState &odsState,
-    FlatSymbolRefAttr mesh, ArrayRef<MeshAxesAttr> split_axes,
+    FlatSymbolRefAttr grid, ArrayRef<GridAxesAttr> split_axes,
     ::mlir::ArrayRef<::mlir::OpFoldResult> halo_sizes,
     ::mlir::ArrayRef<::mlir::OpFoldResult> sharded_dims_offsets) {
   mlir::SmallVector<int64_t> staticHalos, staticDims;
@@ -508,16 +508,16 @@ void ShardingOp::build(
   dispatchIndexOpFoldResults(halo_sizes, dynamicHalos, staticHalos);
   dispatchIndexOpFoldResults(sharded_dims_offsets, dynamicDims, staticDims);
   return build(
-      b, odsState, mesh, MeshAxesArrayAttr::get(b.getContext(), split_axes),
+      b, odsState, grid, GridAxesArrayAttr::get(b.getContext(), split_axes),
       ::mlir::DenseI64ArrayAttr::get(b.getContext(), staticHalos), dynamicHalos,
       ::mlir::DenseI64ArrayAttr::get(b.getContext(), staticDims), dynamicDims);
 }
 
 void ShardingOp::build(::mlir::OpBuilder &b, ::mlir::OperationState &odsState,
-                       mlir::mesh::MeshSharding from) {
+                       mlir::shard::Sharding from) {
 
-  build(b, odsState, ShardingType::get(b.getContext()), from.getMeshAttr(),
-        MeshAxesArrayAttr::get(b.getContext(), from.getSplitAxes()),
+  build(b, odsState, ShardingType::get(b.getContext()), from.getGridAttr(),
+        GridAxesArrayAttr::get(b.getContext(), from.getSplitAxes()),
         from.getStaticShardedDimsOffsets().empty()
             ? DenseI64ArrayAttr()
             : b.getDenseI64ArrayAttr(from.getStaticShardedDimsOffsets()),
@@ -529,21 +529,21 @@ void ShardingOp::build(::mlir::OpBuilder &b, ::mlir::OperationState &odsState,
 }
 
 LogicalResult ShardingOp::verify() {
-  llvm::SmallSet<MeshAxis, 4> visitedAxes;
+  llvm::SmallSet<GridAxis, 4> visitedAxes;
 
-  auto checkMeshAxis = [&](ArrayRef<MeshAxis> axesArray) -> LogicalResult {
-    for (MeshAxis axis : axesArray) {
+  auto checkGridAxis = [&](ArrayRef<GridAxis> axesArray) -> LogicalResult {
+    for (GridAxis axis : axesArray) {
       if (axis < 0)
-        return emitError() << "mesh axis is expected to be non-negative";
+        return emitError() << "grid axis is expected to be non-negative";
       if (!visitedAxes.insert(axis).second)
-        return emitError() << "mesh axis duplicated";
+        return emitError() << "grid axis duplicated";
     }
     return success();
   };
 
   for (auto subAxes : getSplitAxes().getAxes()) {
-    ArrayRef<MeshAxis> subAxesArray = subAxes.asArrayRef();
-    if (failed(checkMeshAxis(subAxesArray)))
+    ArrayRef<GridAxis> subAxesArray = subAxes.asArrayRef();
+    if (failed(checkGridAxis(subAxesArray)))
       return failure();
   }
 
@@ -572,26 +572,26 @@ void ShardingOp::getAsmResultNames(
 }
 
 LogicalResult ShardingOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
-  auto mesh = ::getMeshAndVerify(getOperation(), getMeshAttr(), symbolTable);
-  if (failed(mesh)) {
+  auto grid = ::getGridAndVerify(getOperation(), getGridAttr(), symbolTable);
+  if (failed(grid)) {
     return failure();
   }
-  if (mlir::ShapedType::isDynamicShape(mesh->getShape()) &&
+  if (mlir::ShapedType::isDynamicShape(grid->getShape()) &&
       getStaticShardedDimsOffsets().size() > 0) {
     return emitError() << "sharded dims offsets are not allowed for "
-                          "devices meshes with dynamic shape.";
+                          "device grids with dynamic shape.";
   }
 
   auto shardedDimsOffsets = getStaticShardedDimsOffsets();
   if (!shardedDimsOffsets.empty()) {
-    auto meshShape = mesh.value().getShape();
-    assert(ShapedType::isStaticShape(meshShape));
+    auto gridShape = grid.value().getShape();
+    assert(ShapedType::isStaticShape(gridShape));
     uint64_t pos = 0;
     for (auto [tensorAxis, innerSplitAxes] : llvm::enumerate(getSplitAxes())) {
       if (!innerSplitAxes.empty()) {
         int64_t numShards = 0, off = 0;
         for (auto i : innerSplitAxes.asArrayRef()) {
-          numShards += meshShape[i];
+          numShards += gridShape[i];
         }
         for (int64_t i = 0; i <= numShards; ++i) {
           if (shardedDimsOffsets.size() <= pos + i) {
@@ -684,11 +684,11 @@ void ShardingOp::getCanonicalizationPatterns(mlir::RewritePatternSet &results,
 }
 
 //===----------------------------------------------------------------------===//
-// MeshSharding
+// Sharding
 //===----------------------------------------------------------------------===//
 
-bool MeshSharding::equalSplitAxes(const MeshSharding &rhs) const {
-  if (getMesh() != rhs.getMesh()) {
+bool Sharding::equalSplitAxes(const Sharding &rhs) const {
+  if (getGrid() != rhs.getGrid()) {
     return false;
   }
 
@@ -701,16 +701,16 @@ bool MeshSharding::equalSplitAxes(const MeshSharding &rhs) const {
   }
 
   return llvm::all_of(llvm::drop_begin(getSplitAxes(), minSize),
-                      std::mem_fn(&MeshAxesAttr::empty)) &&
+                      std::mem_fn(&GridAxesAttr::empty)) &&
          llvm::all_of(llvm::drop_begin(rhs.getSplitAxes(), minSize),
-                      std::mem_fn(&MeshAxesAttr::empty));
+                      std::mem_fn(&GridAxesAttr::empty));
 }
 
-bool MeshSharding::equalHaloAndShardSizes(const MeshSharding &rhs) const {
+bool Sharding::equalHaloAndShardSizes(const Sharding &rhs) const {
   return equalShardSizes(rhs) && equalHaloSizes(rhs);
 }
 
-bool MeshSharding::equalShardSizes(const MeshSharding &rhs) const {
+bool Sharding::equalShardSizes(const Sharding &rhs) const {
   if (rhs.getStaticShardedDimsOffsets().size() !=
           getStaticShardedDimsOffsets().size() ||
       !llvm::equal(getStaticShardedDimsOffsets(),
@@ -726,7 +726,7 @@ bool MeshSharding::equalShardSizes(const MeshSharding &rhs) const {
   return true;
 }
 
-bool MeshSharding::equalHaloSizes(const MeshSharding &rhs) const {
+bool Sharding::equalHaloSizes(const Sharding &rhs) const {
   if (rhs.getStaticHaloSizes().size() != getStaticHaloSizes().size() ||
       !llvm::equal(getStaticHaloSizes(), rhs.getStaticHaloSizes())) {
     return false;
@@ -738,45 +738,43 @@ bool MeshSharding::equalHaloSizes(const MeshSharding &rhs) const {
   return true;
 }
 
-bool MeshSharding::operator==(Value rhs) const {
+bool Sharding::operator==(Value rhs) const {
   return equalSplitAxes(rhs) && equalHaloAndShardSizes(rhs);
 }
 
-bool MeshSharding::operator!=(Value rhs) const { return !(*this == rhs); }
+bool Sharding::operator!=(Value rhs) const { return !(*this == rhs); }
 
-bool MeshSharding::operator==(const MeshSharding &rhs) const {
+bool Sharding::operator==(const Sharding &rhs) const {
   return equalSplitAxes(rhs) && equalHaloAndShardSizes(rhs);
 }
 
-bool MeshSharding::operator!=(const MeshSharding &rhs) const {
-  return !(*this == rhs);
-}
+bool Sharding::operator!=(const Sharding &rhs) const { return !(*this == rhs); }
 
-MeshSharding::MeshSharding(::mlir::FlatSymbolRefAttr mesh_) : mesh(mesh_) {}
+Sharding::Sharding(::mlir::FlatSymbolRefAttr grid_) : grid(grid_) {}
 
-MeshSharding::MeshSharding(Value rhs) {
+Sharding::Sharding(Value rhs) {
   auto shardingOp = rhs.getDefiningOp<ShardingOp>();
   assert(shardingOp && "expected sharding op");
   auto splitAxes = shardingOp.getSplitAxes().getAxes();
   // If splitAxes are empty, use "empty" constructor.
   if (splitAxes.empty()) {
-    *this = MeshSharding(shardingOp.getMeshAttr());
+    *this = Sharding(shardingOp.getGridAttr());
     return;
   }
   *this =
-      get(shardingOp.getMeshAttr(), splitAxes, shardingOp.getStaticHaloSizes(),
+      get(shardingOp.getGridAttr(), splitAxes, shardingOp.getStaticHaloSizes(),
           shardingOp.getStaticShardedDimsOffsets(),
           SmallVector<Value>(shardingOp.getDynamicHaloSizes()),
           SmallVector<Value>(shardingOp.getDynamicShardedDimsOffsets()));
 }
 
-MeshSharding MeshSharding::get(::mlir::FlatSymbolRefAttr mesh_,
-                               ArrayRef<MeshAxesAttr> split_axes_,
-                               ArrayRef<int64_t> static_halo_sizes_,
-                               ArrayRef<int64_t> static_sharded_dims_offsets_,
-                               ArrayRef<Value> dynamic_halo_sizes_,
-                               ArrayRef<Value> dynamic_sharded_dims_offsets_) {
-  MeshSharding res(mesh_);
+Sharding Sharding::get(::mlir::FlatSymbolRefAttr grid_,
+                       ArrayRef<GridAxesAttr> split_axes_,
+                       ArrayRef<int64_t> static_halo_sizes_,
+                       ArrayRef<int64_t> static_sharded_dims_offsets_,
+                       ArrayRef<Value> dynamic_halo_sizes_,
+                       ArrayRef<Value> dynamic_sharded_dims_offsets_) {
+  Sharding res(grid_);
   if (split_axes_.empty()) {
     return res;
   }
@@ -784,7 +782,7 @@ MeshSharding MeshSharding::get(::mlir::FlatSymbolRefAttr mesh_,
   res.split_axes.resize(split_axes_.size());
   for (auto [i, axis] : llvm::enumerate(split_axes_)) {
     res.split_axes[i] =
-        MeshAxesAttr::get(mesh_.getContext(), axis.asArrayRef());
+        GridAxesAttr::get(grid_.getContext(), axis.asArrayRef());
   }
 
   auto clone = [](const auto src, auto &dst) {
@@ -801,7 +799,7 @@ MeshSharding MeshSharding::get(::mlir::FlatSymbolRefAttr mesh_,
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.shard_shape
+// shard.shard_shape
 //===----------------------------------------------------------------------===//
 
 void ShardShapeOp::getAsmResultNames(
@@ -820,7 +818,7 @@ void ShardShapeOp::build(::mlir::OpBuilder &odsBuilder,
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.shard op
+// shard.shard op
 //===----------------------------------------------------------------------===//
 
 void ShardOp::getAsmResultNames(
@@ -850,10 +848,10 @@ public:
         if (!otherOp || !otherOp->isBeforeInBlock(op)) {
           return failure();
         }
-        // Create a MeshSharding object for the current and the other ShardOp
+        // Create a Sharding object for the current and the other ShardOp
         // If the two are equal replace current op with the other op.
-        MeshSharding currentSharding(op.getSharding());
-        MeshSharding otherSharding(otherOp.getSharding());
+        Sharding currentSharding(op.getSharding());
+        Sharding otherSharding(otherOp.getSharding());
         if (currentSharding == otherSharding) {
           b.replaceAllUsesWith(op.getResult(), otherOp.getResult());
           b.eraseOp(op.getOperation());
@@ -876,21 +874,21 @@ void ShardOp::getCanonicalizationPatterns(mlir::RewritePatternSet &results,
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.process_multi_index op
+// shard.process_multi_index op
 //===----------------------------------------------------------------------===//
 
 LogicalResult
 ProcessMultiIndexOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
-  auto mesh = ::getMeshAndVerify(getOperation(), getMeshAttr(), symbolTable);
-  if (failed(mesh)) {
+  auto grid = ::getGridAndVerify(getOperation(), getGridAttr(), symbolTable);
+  if (failed(grid)) {
     return failure();
   }
-  if (failed(verifyMeshAxes(getLoc(), getAxes(), mesh.value()))) {
+  if (failed(verifyGridAxes(getLoc(), getAxes(), grid.value()))) {
     return failure();
   }
 
   size_t expectedResultsCount =
-      getAxes().empty() ? mesh->getRank() : getAxes().size();
+      getAxes().empty() ? grid->getRank() : getAxes().size();
   if (getResult().size() != expectedResultsCount) {
     return emitError() << "Unexpected number of results " << getResult().size()
                        << ". Expected " << expectedResultsCount << ".";
@@ -900,17 +898,17 @@ ProcessMultiIndexOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
 }
 
 void ProcessMultiIndexOp::build(OpBuilder &odsBuilder, OperationState &odsState,
-                                MeshOp mesh) {
+                                GridOp grid) {
   build(odsBuilder, odsState,
-        SmallVector<Type>(mesh.getRank(), odsBuilder.getIndexType()),
-        mesh.getSymName(), ArrayRef<MeshAxis>());
+        SmallVector<Type>(grid.getRank(), odsBuilder.getIndexType()),
+        grid.getSymName(), ArrayRef<GridAxis>());
 }
 
 void ProcessMultiIndexOp::build(OpBuilder &odsBuilder, OperationState &odsState,
-                                StringRef mesh, ArrayRef<MeshAxis> axes) {
+                                StringRef grid, ArrayRef<GridAxis> axes) {
   build(odsBuilder, odsState,
-        SmallVector<Type>(axes.size(), odsBuilder.getIndexType()), mesh,
-        MeshAxesAttr::get(odsBuilder.getContext(), axes));
+        SmallVector<Type>(axes.size(), odsBuilder.getIndexType()), grid,
+        GridAxesAttr::get(odsBuilder.getContext(), axes));
 }
 
 void ProcessMultiIndexOp::getAsmResultNames(
@@ -919,21 +917,21 @@ void ProcessMultiIndexOp::getAsmResultNames(
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.process_linear_index op
+// shard.process_linear_index op
 //===----------------------------------------------------------------------===//
 
 LogicalResult
 ProcessLinearIndexOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
-  auto mesh = ::getMeshAndVerify(getOperation(), getMeshAttr(), symbolTable);
-  if (failed(mesh)) {
+  auto grid = ::getGridAndVerify(getOperation(), getGridAttr(), symbolTable);
+  if (failed(grid)) {
     return failure();
   }
   return success();
 }
 
 void ProcessLinearIndexOp::build(OpBuilder &odsBuilder,
-                                 OperationState &odsState, MeshOp mesh) {
-  build(odsBuilder, odsState, mesh.getSymName());
+                                 OperationState &odsState, GridOp grid) {
+  build(odsBuilder, odsState, grid.getSymName());
 }
 
 void ProcessLinearIndexOp::getAsmResultNames(
@@ -942,13 +940,13 @@ void ProcessLinearIndexOp::getAsmResultNames(
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.neighbors_linear_indices op
+// shard.neighbors_linear_indices op
 //===----------------------------------------------------------------------===//
 
 LogicalResult
 NeighborsLinearIndicesOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
-  auto mesh = ::getMeshAndVerify(getOperation(), getMeshAttr(), symbolTable);
-  if (failed(mesh)) {
+  auto grid = ::getGridAndVerify(getOperation(), getGridAttr(), symbolTable);
+  if (failed(grid)) {
     return failure();
   }
   return success();
@@ -967,12 +965,12 @@ void NeighborsLinearIndicesOp::getAsmResultNames(
 namespace {
 
 template <typename Op>
-struct EmptyMeshAxesCanonicalizationPattern : OpRewritePattern<Op> {
+struct EmptyGridAxesCanonicalizationPattern : OpRewritePattern<Op> {
   using OpRewritePattern<Op>::OpRewritePattern;
   LogicalResult matchAndRewrite(Op op,
                                 PatternRewriter &rewriter) const override {
-    auto meshAxes = op.getMeshAxes();
-    if (!meshAxes.empty()) {
+    auto gridAxes = op.getGridAxes();
+    if (!gridAxes.empty()) {
       return failure();
     }
     if (op.getInput().getType() != op.getResult().getType()) {
@@ -990,24 +988,24 @@ struct EmptyMeshAxesCanonicalizationPattern : OpRewritePattern<Op> {
 static LogicalResult verifyInGroupDevice(Location loc, StringRef deviceName,
                                          ArrayRef<int64_t> device,
                                          Operation::operand_range deviceDynamic,
-                                         ArrayRef<MeshAxis> meshAxes,
-                                         ArrayRef<int64_t> meshShape) {
-  if (device.size() != meshAxes.size()) {
+                                         ArrayRef<GridAxis> gridAxes,
+                                         ArrayRef<int64_t> gridShape) {
+  if (device.size() != gridAxes.size()) {
     return emitError(loc) << "In-group device \"" << deviceName
                           << "\" has unexpected multi-index size "
-                          << device.size() << ". Expected " << meshAxes.size()
+                          << device.size() << ". Expected " << gridAxes.size()
                           << ".";
   }
 
   for (size_t i = 0; i < device.size(); ++i) {
     if (ShapedType::isStatic(device[i]) &&
-        ShapedType::isStatic(meshShape[meshAxes[i]]) &&
-        meshShape[meshAxes[i]] <= device[i]) {
+        ShapedType::isStatic(gridShape[gridAxes[i]]) &&
+        gridShape[gridAxes[i]] <= device[i]) {
       return emitError(loc)
              << "Out of bounds coordinate " << i << " for in-group device \""
              << deviceName << "\"."
              << " Got " << device[i] << ", but expected value in the range [0, "
-             << (meshShape[meshAxes[i]] - 1) << "].";
+             << (gridShape[gridAxes[i]] - 1) << "].";
     }
   }
   return success();
@@ -1043,7 +1041,7 @@ static LogicalResult verifyDimensionCompatibility(Location loc,
 
 static LogicalResult verifyGatherOperandAndResultShape(
     Value operand, Value result, int64_t gatherAxis,
-    ArrayRef<MeshAxis> meshAxes, ArrayRef<int64_t> meshShape) {
+    ArrayRef<GridAxis> gridAxes, ArrayRef<int64_t> gridShape) {
   auto resultRank = cast<ShapedType>(result.getType()).getRank();
   if (gatherAxis < 0 || gatherAxis >= resultRank) {
     return emitError(result.getLoc())
@@ -1054,7 +1052,7 @@ static LogicalResult verifyGatherOperandAndResultShape(
   ShapedType operandType = cast<ShapedType>(operand.getType());
   ShapedType resultType = cast<ShapedType>(result.getType());
   auto deviceGroupSize =
-      DimensionSize(collectiveProcessGroupSize(meshAxes, meshShape));
+      DimensionSize(collectiveProcessGroupSize(gridAxes, gridShape));
   for (int64_t axis = 0; axis < operandType.getRank(); ++axis) {
     auto operandDimSize = DimensionSize(operandType.getDimSize(axis));
     auto resultDimSize = DimensionSize(resultType.getDimSize(axis));
@@ -1070,7 +1068,7 @@ static LogicalResult verifyGatherOperandAndResultShape(
 
 static LogicalResult verifyAllToAllOperandAndResultShape(
     Value operand, Value result, int64_t splitAxis, int64_t concatAxis,
-    ArrayRef<MeshAxis> meshAxes, ArrayRef<int64_t> meshShape) {
+    ArrayRef<GridAxis> gridAxes, ArrayRef<int64_t> gridShape) {
   ShapedType operandType = cast<ShapedType>(operand.getType());
   ShapedType resultType = cast<ShapedType>(result.getType());
   for (int64_t axis = 0; axis < operandType.getRank(); ++axis) {
@@ -1088,7 +1086,7 @@ static LogicalResult verifyAllToAllOperandAndResultShape(
   }
 
   auto deviceGroupSize =
-      DimensionSize(collectiveProcessGroupSize(meshAxes, meshShape));
+      DimensionSize(collectiveProcessGroupSize(gridAxes, gridShape));
   auto operandConcatDimSize = DimensionSize(operandType.getDimSize(concatAxis));
   auto operandSplitDimSize = DimensionSize(operandType.getDimSize(splitAxis));
   DimensionSize expectedResultConcatDimSize =
@@ -1115,7 +1113,7 @@ static LogicalResult verifyAllToAllOperandAndResultShape(
 
 static LogicalResult verifyScatterOrSliceOperandAndResultShape(
     Value operand, Value result, int64_t tensorAxis,
-    ArrayRef<MeshAxis> meshAxes, ArrayRef<int64_t> meshShape) {
+    ArrayRef<GridAxis> gridAxes, ArrayRef<int64_t> gridShape) {
   ShapedType operandType = cast<ShapedType>(operand.getType());
   ShapedType resultType = cast<ShapedType>(result.getType());
   for (int64_t axis = 0; axis < operandType.getRank(); ++axis) {
@@ -1129,7 +1127,7 @@ static LogicalResult verifyScatterOrSliceOperandAndResultShape(
   }
 
   auto deviceGroupSize =
-      DimensionSize(collectiveProcessGroupSize(meshAxes, meshShape));
+      DimensionSize(collectiveProcessGroupSize(gridAxes, gridShape));
   auto operandScatterDimSize =
       DimensionSize(operandType.getDimSize(tensorAxis));
   if (!operandScatterDimSize.isDynamic() && !deviceGroupSize.isDynamic() &&
@@ -1151,8 +1149,8 @@ static LogicalResult verifyScatterOrSliceOperandAndResultShape(
   return success();
 }
 
-static RankedTensorType sliceResultType(Type operandType, MeshOp mesh,
-                                        ArrayRef<MeshAxis> meshAxes,
+static RankedTensorType sliceResultType(Type operandType, GridOp grid,
+                                        ArrayRef<GridAxis> gridAxes,
                                         int64_t sliceAxis) {
   RankedTensorType operandRankedTensorType =
       cast<RankedTensorType>(operandType);
@@ -1163,29 +1161,29 @@ static RankedTensorType sliceResultType(Type operandType, MeshOp mesh,
 
   resultShape[sliceAxis] =
       operandSliceAxisSize /
-      DimensionSize(collectiveProcessGroupSize(meshAxes, mesh));
+      DimensionSize(collectiveProcessGroupSize(gridAxes, grid));
   return operandRankedTensorType.clone(resultShape);
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.all_gather op
+// shard.all_gather op
 //===----------------------------------------------------------------------===//
 
 LogicalResult
 AllGatherOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
-  auto mesh = getMeshAndVerifyAxes(*this, symbolTable);
-  if (failed(mesh)) {
+  auto grid = getGridAndVerifyAxes(*this, symbolTable);
+  if (failed(grid)) {
     return failure();
   }
   auto gatherAxis = getGatherAxis().getSExtValue();
   return verifyGatherOperandAndResultShape(getOperand(), getResult(),
-                                           gatherAxis, getMeshAxes(),
-                                           mesh.value().getShape());
+                                           gatherAxis, getGridAxes(),
+                                           grid.value().getShape());
 }
 
 void AllGatherOp::getCanonicalizationPatterns(RewritePatternSet &patterns,
                                               MLIRContext *context) {
-  patterns.add<EmptyMeshAxesCanonicalizationPattern<AllGatherOp>>(context);
+  patterns.add<EmptyGridAxesCanonicalizationPattern<AllGatherOp>>(context);
 }
 
 void AllGatherOp::getAsmResultNames(
@@ -1194,23 +1192,23 @@ void AllGatherOp::getAsmResultNames(
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.all_reduce op
+// shard.all_reduce op
 //===----------------------------------------------------------------------===//
 
 LogicalResult
 AllReduceOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
-  return getMeshAndVerifyAxes(*this, symbolTable);
+  return getGridAndVerifyAxes(*this, symbolTable);
 }
 
 void AllReduceOp::getCanonicalizationPatterns(RewritePatternSet &patterns,
                                               MLIRContext *context) {
-  patterns.add<EmptyMeshAxesCanonicalizationPattern<AllReduceOp>>(context);
+  patterns.add<EmptyGridAxesCanonicalizationPattern<AllReduceOp>>(context);
 }
 
 void AllReduceOp::build(OpBuilder &odsBuilder, OperationState &odsState,
-                        Value input, StringRef mesh,
-                        ArrayRef<MeshAxis> meshAxes, ReductionKind reduction) {
-  build(odsBuilder, odsState, input.getType(), mesh, meshAxes, input,
+                        Value input, StringRef grid,
+                        ArrayRef<GridAxis> gridAxes, ReductionKind reduction) {
+  build(odsBuilder, odsState, input.getType(), grid, gridAxes, input,
         reduction);
 }
 
@@ -1220,36 +1218,36 @@ void AllReduceOp::getAsmResultNames(
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.all_slice op
+// shard.all_slice op
 //===----------------------------------------------------------------------===//
 
 LogicalResult AllSliceOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
-  auto mesh = getMeshAndVerifyAxes(*this, symbolTable);
-  if (failed(mesh)) {
+  auto grid = getGridAndVerifyAxes(*this, symbolTable);
+  if (failed(grid)) {
     return failure();
   }
   return verifyScatterOrSliceOperandAndResultShape(
-      getOperand(), getResult(), getSliceAxis().getSExtValue(), getMeshAxes(),
-      mesh.value().getShape());
+      getOperand(), getResult(), getSliceAxis().getSExtValue(), getGridAxes(),
+      grid.value().getShape());
 }
 
 void AllSliceOp::getCanonicalizationPatterns(RewritePatternSet &patterns,
                                              MLIRContext *context) {
-  patterns.add<EmptyMeshAxesCanonicalizationPattern<AllSliceOp>>(context);
+  patterns.add<EmptyGridAxesCanonicalizationPattern<AllSliceOp>>(context);
 }
 
 void AllSliceOp::build(OpBuilder &odsBuilder, OperationState &odsState,
-                       Value input, MeshOp mesh, ArrayRef<MeshAxis> meshAxes,
+                       Value input, GridOp grid, ArrayRef<GridAxis> gridAxes,
                        int64_t sliceAxis) {
-  Type resultType = sliceResultType(input.getType(), mesh, meshAxes, sliceAxis);
-  build(odsBuilder, odsState, resultType, input, mesh.getSymName(), meshAxes,
+  Type resultType = sliceResultType(input.getType(), grid, gridAxes, sliceAxis);
+  build(odsBuilder, odsState, resultType, input, grid.getSymName(), gridAxes,
         sliceAxis);
 }
 
 void AllSliceOp::build(OpBuilder &odsBuilder, OperationState &odsState,
-                       Type resultType, Value input, StringRef mesh,
-                       ArrayRef<MeshAxis> meshAxes, int64_t sliceAxis) {
-  build(odsBuilder, odsState, resultType, mesh, meshAxes, input,
+                       Type resultType, Value input, StringRef grid,
+                       ArrayRef<GridAxis> gridAxes, int64_t sliceAxis) {
+  build(odsBuilder, odsState, resultType, grid, gridAxes, input,
         APInt(sizeof(sliceAxis) * CHAR_BIT, sliceAxis));
 }
 
@@ -1259,23 +1257,23 @@ void AllSliceOp::getAsmResultNames(
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.all_to_all op
+// shard.all_to_all op
 //===----------------------------------------------------------------------===//
 
 LogicalResult AllToAllOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
-  auto mesh = getMeshAndVerifyAxes(*this, symbolTable);
-  if (failed(mesh)) {
+  auto grid = getGridAndVerifyAxes(*this, symbolTable);
+  if (failed(grid)) {
     return failure();
   }
 
   return verifyAllToAllOperandAndResultShape(
       getOperand(), getResult(), getSplitAxis().getSExtValue(),
-      getConcatAxis().getSExtValue(), getMeshAxes(), mesh.value().getShape());
+      getConcatAxis().getSExtValue(), getGridAxes(), grid.value().getShape());
 }
 
 void AllToAllOp::getCanonicalizationPatterns(RewritePatternSet &patterns,
                                              MLIRContext *context) {
-  patterns.add<EmptyMeshAxesCanonicalizationPattern<AllToAllOp>>(context);
+  patterns.add<EmptyGridAxesCanonicalizationPattern<AllToAllOp>>(context);
 }
 
 void AllToAllOp::getAsmResultNames(
@@ -1284,18 +1282,18 @@ void AllToAllOp::getAsmResultNames(
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.broadcast op
+// shard.broadcast op
 //===----------------------------------------------------------------------===//
 
 LogicalResult
 BroadcastOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
-  auto mesh = getMeshAndVerifyAxes(*this, symbolTable);
-  if (failed(mesh)) {
+  auto grid = getGridAndVerifyAxes(*this, symbolTable);
+  if (failed(grid)) {
     return failure();
   }
   if (failed(verifyInGroupDevice(getLoc(), getRootAttrName(), getRoot(),
-                                 getRootDynamic(), getMeshAxes(),
-                                 mesh.value().getShape()))) {
+                                 getRootDynamic(), getGridAxes(),
+                                 grid.value().getShape()))) {
     return failure();
   }
 
@@ -1304,7 +1302,7 @@ BroadcastOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
 
 void BroadcastOp::getCanonicalizationPatterns(RewritePatternSet &patterns,
                                               MLIRContext *context) {
-  patterns.add<EmptyMeshAxesCanonicalizationPattern<BroadcastOp>>(context);
+  patterns.add<EmptyGridAxesCanonicalizationPattern<BroadcastOp>>(context);
 }
 
 void BroadcastOp::getAsmResultNames(
@@ -1313,29 +1311,29 @@ void BroadcastOp::getAsmResultNames(
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.gather op
+// shard.gather op
 //===----------------------------------------------------------------------===//
 
 LogicalResult GatherOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
-  auto mesh = getMeshAndVerifyAxes(*this, symbolTable);
-  if (failed(mesh)) {
+  auto grid = getGridAndVerifyAxes(*this, symbolTable);
+  if (failed(grid)) {
     return failure();
   }
   if (failed(verifyInGroupDevice(getLoc(), getRootAttrName(), getRoot(),
-                                 getRootDynamic(), getMeshAxes(),
-                                 mesh.value().getShape()))) {
+                                 getRootDynamic(), getGridAxes(),
+                                 grid.value().getShape()))) {
     return failure();
   }
 
   auto gatherAxis = getGatherAxis().getSExtValue();
   return verifyGatherOperandAndResultShape(getInput(), getResult(), gatherAxis,
-                                           getMeshAxes(),
-                                           mesh.value().getShape());
+                                           getGridAxes(),
+                                           grid.value().getShape());
 }
 
 void GatherOp::getCanonicalizationPatterns(RewritePatternSet &patterns,
                                            MLIRContext *context) {
-  patterns.add<EmptyMeshAxesCanonicalizationPattern<GatherOp>>(context);
+  patterns.add<EmptyGridAxesCanonicalizationPattern<GatherOp>>(context);
 }
 
 void GatherOp::getAsmResultNames(
@@ -1344,18 +1342,18 @@ void GatherOp::getAsmResultNames(
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.recv op
+// shard.recv op
 //===----------------------------------------------------------------------===//
 
 LogicalResult RecvOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
-  auto mesh = getMeshAndVerifyAxes(*this, symbolTable);
-  if (failed(mesh)) {
+  auto grid = getGridAndVerifyAxes(*this, symbolTable);
+  if (failed(grid)) {
     return failure();
   }
   if (getSource() &&
       failed(verifyInGroupDevice(getLoc(), getSourceAttrName(),
                                  getSource().value(), getSourceDynamic(),
-                                 getMeshAxes(), mesh.value().getShape()))) {
+                                 getGridAxes(), grid.value().getShape()))) {
     return failure();
   }
   return success();
@@ -1363,7 +1361,7 @@ LogicalResult RecvOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
 
 void RecvOp::getCanonicalizationPatterns(RewritePatternSet &patterns,
                                          MLIRContext *context) {
-  patterns.add<EmptyMeshAxesCanonicalizationPattern<RecvOp>>(context);
+  patterns.add<EmptyGridAxesCanonicalizationPattern<RecvOp>>(context);
 }
 
 void RecvOp::getAsmResultNames(function_ref<void(Value, StringRef)> setNameFn) {
@@ -1371,17 +1369,17 @@ void RecvOp::getAsmResultNames(function_ref<void(Value, StringRef)> setNameFn) {
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.reduce op
+// shard.reduce op
 //===----------------------------------------------------------------------===//
 
 LogicalResult ReduceOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
-  auto mesh = getMeshAndVerifyAxes(*this, symbolTable);
-  if (failed(mesh)) {
+  auto grid = getGridAndVerifyAxes(*this, symbolTable);
+  if (failed(grid)) {
     return failure();
   }
   if (failed(verifyInGroupDevice(getLoc(), getRootAttrName(), getRoot(),
-                                 getRootDynamic(), getMeshAxes(),
-                                 mesh.value().getShape()))) {
+                                 getRootDynamic(), getGridAxes(),
+                                 grid.value().getShape()))) {
     return failure();
   }
 
@@ -1390,7 +1388,7 @@ LogicalResult ReduceOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
 
 void ReduceOp::getCanonicalizationPatterns(RewritePatternSet &patterns,
                                            MLIRContext *context) {
-  patterns.add<EmptyMeshAxesCanonicalizationPattern<ReduceOp>>(context);
+  patterns.add<EmptyGridAxesCanonicalizationPattern<ReduceOp>>(context);
 }
 
 void ReduceOp::getAsmResultNames(
@@ -1399,24 +1397,24 @@ void ReduceOp::getAsmResultNames(
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.reduce_scatter op
+// shard.reduce_scatter op
 //===----------------------------------------------------------------------===//
 
 LogicalResult
 ReduceScatterOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
-  auto mesh = getMeshAndVerifyAxes(*this, symbolTable);
-  if (failed(mesh)) {
+  auto grid = getGridAndVerifyAxes(*this, symbolTable);
+  if (failed(grid)) {
     return failure();
   }
 
   return verifyScatterOrSliceOperandAndResultShape(
-      getOperand(), getResult(), getScatterAxis().getSExtValue(), getMeshAxes(),
-      mesh.value().getShape());
+      getOperand(), getResult(), getScatterAxis().getSExtValue(), getGridAxes(),
+      grid.value().getShape());
 }
 
 void ReduceScatterOp::getCanonicalizationPatterns(RewritePatternSet &patterns,
                                                   MLIRContext *context) {
-  patterns.add<EmptyMeshAxesCanonicalizationPattern<ReduceScatterOp>>(context);
+  patterns.add<EmptyGridAxesCanonicalizationPattern<ReduceScatterOp>>(context);
 }
 
 void ReduceScatterOp::getAsmResultNames(
@@ -1425,29 +1423,29 @@ void ReduceScatterOp::getAsmResultNames(
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.scatter op
+// shard.scatter op
 //===----------------------------------------------------------------------===//
 
 LogicalResult ScatterOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
-  auto mesh = getMeshAndVerifyAxes(*this, symbolTable);
-  if (failed(mesh)) {
+  auto grid = getGridAndVerifyAxes(*this, symbolTable);
+  if (failed(grid)) {
     return failure();
   }
   if (failed(verifyInGroupDevice(getLoc(), getRootAttrName(), getRoot(),
-                                 getRootDynamic(), getMeshAxes(),
-                                 mesh.value().getShape()))) {
+                                 getRootDynamic(), getGridAxes(),
+                                 grid.value().getShape()))) {
     return failure();
   }
 
   auto scatterAxis = getScatterAxis().getSExtValue();
   return verifyScatterOrSliceOperandAndResultShape(getInput(), getResult(),
-                                                   scatterAxis, getMeshAxes(),
-                                                   mesh.value().getShape());
+                                                   scatterAxis, getGridAxes(),
+                                                   grid.value().getShape());
 }
 
 void ScatterOp::getCanonicalizationPatterns(RewritePatternSet &patterns,
                                             MLIRContext *context) {
-  patterns.add<EmptyMeshAxesCanonicalizationPattern<ScatterOp>>(context);
+  patterns.add<EmptyGridAxesCanonicalizationPattern<ScatterOp>>(context);
 }
 
 void ScatterOp::getAsmResultNames(
@@ -1456,17 +1454,17 @@ void ScatterOp::getAsmResultNames(
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.send op
+// shard.send op
 //===----------------------------------------------------------------------===//
 
 LogicalResult SendOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
-  auto mesh = getMeshAndVerifyAxes(*this, symbolTable);
-  if (failed(mesh)) {
+  auto grid = getGridAndVerifyAxes(*this, symbolTable);
+  if (failed(grid)) {
     return failure();
   }
   if (failed(verifyInGroupDevice(getLoc(), getDestinationAttrName(),
                                  getDestination(), getDestinationDynamic(),
-                                 getMeshAxes(), mesh.value().getShape()))) {
+                                 getGridAxes(), grid.value().getShape()))) {
     return failure();
   }
   return success();
@@ -1474,7 +1472,7 @@ LogicalResult SendOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
 
 void SendOp::getCanonicalizationPatterns(RewritePatternSet &patterns,
                                          MLIRContext *context) {
-  patterns.add<EmptyMeshAxesCanonicalizationPattern<SendOp>>(context);
+  patterns.add<EmptyGridAxesCanonicalizationPattern<SendOp>>(context);
 }
 
 void SendOp::getAsmResultNames(function_ref<void(Value, StringRef)> setNameFn) {
@@ -1482,20 +1480,20 @@ void SendOp::getAsmResultNames(function_ref<void(Value, StringRef)> setNameFn) {
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.shift op
+// shard.shift op
 //===----------------------------------------------------------------------===//
 
 LogicalResult ShiftOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
-  auto mesh = getMeshAndVerifyAxes(*this, symbolTable);
-  if (failed(mesh)) {
+  auto grid = getGridAndVerifyAxes(*this, symbolTable);
+  if (failed(grid)) {
     return failure();
   }
 
-  auto meshAxes = getMeshAxes();
+  auto gridAxes = getGridAxes();
   auto shiftAxis = getShiftAxis().getZExtValue();
-  if (!llvm::is_contained(meshAxes, shiftAxis)) {
+  if (!llvm::is_contained(gridAxes, shiftAxis)) {
     return emitError() << "Invalid shift axis " << shiftAxis
-                       << ". It must be one of the grouping mesh axes.";
+                       << ". It must be one of the grouping grid axes.";
   }
 
   return success();
@@ -1504,7 +1502,7 @@ LogicalResult ShiftOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
 void ShiftOp::getCanonicalizationPatterns(RewritePatternSet &patterns,
                                           MLIRContext *context) {
   // TODO: remove op when offset is 0 or if it is a rotate with and
-  // offset % shift_axis_mesh_dim_size == 0.
+  // offset % shift_axis_grid_dim_size == 0.
 }
 
 void ShiftOp::getAsmResultNames(
@@ -1513,13 +1511,13 @@ void ShiftOp::getAsmResultNames(
 }
 
 //===----------------------------------------------------------------------===//
-// mesh.update_halo op
+// shard.update_halo op
 //===----------------------------------------------------------------------===//
 
 LogicalResult
 UpdateHaloOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
-  auto mesh = getMeshAndVerify(getOperation(), getMeshAttr(), symbolTable);
-  if (failed(mesh)) {
+  auto grid = getGridAndVerify(getOperation(), getGridAttr(), symbolTable);
+  if (failed(grid)) {
     return failure();
   }
 
@@ -1531,12 +1529,12 @@ UpdateHaloOp::verifySymbolUses(SymbolTableCollection &symbolTable) {
 //===----------------------------------------------------------------------===//
 
 #define GET_OP_CLASSES
-#include "mlir/Dialect/Mesh/IR/MeshOps.cpp.inc"
+#include "mlir/Dialect/Shard/IR/ShardOps.cpp.inc"
 
 #define GET_ATTRDEF_CLASSES
-#include "mlir/Dialect/Mesh/IR/MeshAttributes.cpp.inc"
+#include "mlir/Dialect/Shard/IR/ShardAttributes.cpp.inc"
 
 #define GET_TYPEDEF_CLASSES
-#include "mlir/Dialect/Mesh/IR/MeshTypes.cpp.inc"
+#include "mlir/Dialect/Shard/IR/ShardTypes.cpp.inc"
 
-#include "mlir/Dialect/Mesh/IR/MeshEnums.cpp.inc"
+#include "mlir/Dialect/Shard/IR/ShardEnums.cpp.inc"
diff --git a/mlir/lib/Dialect/Mesh/Interfaces/CMakeLists.txt b/mlir/lib/Dialect/Shard/Interfaces/CMakeLists.txt
index afe76b5..01e8e56 100644
--- a/mlir/lib/Dialect/Mesh/Interfaces/CMakeLists.txt
+++ b/mlir/lib/Dialect/Shard/Interfaces/CMakeLists.txt
@@ -2,7 +2,7 @@ add_mlir_library(MLIRShardingInterface
   ShardingInterface.cpp
 
   ADDITIONAL_HEADER_DIRS
-  ${MLIR_MAIN_INCLUDE_DIR}/mlir/Dialect/Mesh
+  ${MLIR_MAIN_INCLUDE_DIR}/mlir/Dialect/Shard
 
   DEPENDS
   MLIRShardingInterfaceIncGen
@@ -10,7 +10,7 @@ add_mlir_library(MLIRShardingInterface
   LINK_LIBS PUBLIC
   MLIRDialectUtils
   MLIRIR
-  MLIRMeshDialect
+  MLIRShardDialect
   MLIRTensorDialect
   MLIRSupport
 )
diff --git a/mlir/lib/Dialect/Mesh/Interfaces/ShardingInterface.cpp b/mlir/lib/Dialect/Shard/Interfaces/ShardingInterface.cpp
index 6b3d49e..d4e7618 100644
--- a/mlir/lib/Dialect/Mesh/Interfaces/ShardingInterface.cpp
+++ b/mlir/lib/Dialect/Shard/Interfaces/ShardingInterface.cpp
@@ -6,10 +6,10 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "mlir/Dialect/Mesh/Interfaces/ShardingInterface.h"
-#include "mlir/Dialect/Mesh/Interfaces/ShardingInterfaceImpl.h"
+#include "mlir/Dialect/Shard/Interfaces/ShardingInterface.h"
+#include "mlir/Dialect/Shard/Interfaces/ShardingInterfaceImpl.h"
 
-#include "mlir/Dialect/Mesh/IR/MeshOps.h"
+#include "mlir/Dialect/Shard/IR/ShardOps.h"
 #include "mlir/IR/AffineMap.h"
 #include "mlir/IR/IRMapping.h"
 #include "mlir/Support/LLVM.h"
@@ -24,9 +24,9 @@
 #define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE << "]: ")
 
 using namespace mlir;
-using namespace mlir::mesh;
+using namespace mlir::shard;
 
-#include "mlir/Dialect/Mesh/Interfaces/ShardingInterface.cpp.inc"
+#include "mlir/Dialect/Shard/Interfaces/ShardingInterface.cpp.inc"
 
 //===----------------------------------------------------------------------===//
 // common util functions
@@ -93,40 +93,39 @@ checkOperandAffineExpr(AffineExpr expr, unsigned numDims) {
 }
 
 template <typename T>
-SmallVector<MeshAxesAttr>
+SmallVector<GridAxesAttr>
 fromArrayOfVector(MLIRContext *ctxt, const SmallVector<SmallVector<T>> &vec) {
-  SmallVector<MeshAxesAttr> res;
+  SmallVector<GridAxesAttr> res;
   for (const auto &v : vec) {
-    res.emplace_back(MeshAxesAttr::get(ctxt, v));
+    res.emplace_back(GridAxesAttr::get(ctxt, v));
   }
   return res;
 }
 
 //===----------------------------------------------------------------------===//
-// mesh::getMeshSharding
+// shard::getSharding
 //===----------------------------------------------------------------------===//
 
-FailureOr<std::pair<bool, MeshSharding>>
-mesh::getMeshSharding(OpResult result) {
+FailureOr<std::pair<bool, Sharding>> shard::getSharding(OpResult result) {
   Value val = cast<Value>(result);
   bool anyShardedForDef = llvm::any_of(val.getUsers(), [](Operation *user) {
-    auto shardOp = llvm::dyn_cast<mesh::ShardOp>(user);
+    auto shardOp = llvm::dyn_cast<shard::ShardOp>(user);
     if (!shardOp)
       return false;
     return !shardOp.getAnnotateForUsers();
   });
 
   if (anyShardedForDef) {
-    // expected to have exact one use if it has a use of `mesh.shard` without
+    // expected to have exact one use if it has a use of `shard.shard` without
     // unit attr annotate_for_users
     if (!val.hasOneUse())
       return failure();
-    auto shardOp = llvm::cast<mesh::ShardOp>(*val.getUsers().begin());
-    return std::make_pair(false, MeshSharding(shardOp.getSharding()));
+    auto shardOp = llvm::cast<shard::ShardOp>(*val.getUsers().begin());
+    return std::make_pair(false, Sharding(shardOp.getSharding()));
   }
 
   bool anyShardedForUsers = llvm::any_of(val.getUsers(), [](Operation *user) {
-    auto shardOp = llvm::dyn_cast<mesh::ShardOp>(user);
+    auto shardOp = llvm::dyn_cast<shard::ShardOp>(user);
     if (!shardOp)
       return false;
     return shardOp.getAnnotateForUsers();
@@ -138,24 +137,23 @@ mesh::getMeshSharding(OpResult result) {
       if (shardOp)
         shardOps.push_back(shardOp);
     }
-    MeshSharding shardForDef = shardOps[0].getSharding();
+    Sharding shardForDef = shardOps[0].getSharding();
     for (size_t i = 1; i < shardOps.size(); ++i) {
-      // TODO: Deduce a reasonable mesh sharding attr for def when they are
+      // TODO: Deduce a reasonable grid sharding attr for def when they are
       // different
       assert(shardForDef == shardOps[i].getSharding() &&
-             "only support all shard ops have the same mesh sharding attr");
+             "only support all shard ops have the same grid sharding attr");
     }
     return std::make_pair(true, shardForDef);
   }
   return failure();
 }
 
-FailureOr<std::pair<bool, MeshSharding>>
-mesh::getMeshSharding(OpOperand &opOperand) {
+FailureOr<std::pair<bool, Sharding>> shard::getSharding(OpOperand &opOperand) {
   Value val = opOperand.get();
   if (ShardOp shardOp = val.getDefiningOp<ShardOp>())
     return std::make_pair(shardOp.getAnnotateForUsers(),
-                          MeshSharding(shardOp.getSharding()));
+                          Sharding(shardOp.getSharding()));
 
   return failure();
 }
@@ -164,7 +162,7 @@ mesh::getMeshSharding(OpOperand &opOperand) {
 // ShardingInterface::verifyShardingInterfaceImpl
 //===----------------------------------------------------------------------===//
 
-LogicalResult mesh::ShardingInterface::verifyShardingInterfaceImpl() {
+LogicalResult shard::ShardingInterface::verifyShardingInterfaceImpl() {
   Operation *op = getOperation();
 
   // check operands and results type
@@ -201,7 +199,7 @@ LogicalResult mesh::ShardingInterface::verifyShardingInterfaceImpl() {
 // ShardingInterface::printLoopTypesAndIndexingMaps
 //===----------------------------------------------------------------------===//
 
-void mesh::ShardingInterface::printLoopTypesAndIndexingMaps(raw_ostream &os) {
+void shard::ShardingInterface::printLoopTypesAndIndexingMaps(raw_ostream &os) {
   os << "print loop types and indexing maps for: \n";
   getOperation()->print(os);
   os << "\n";
@@ -222,15 +220,15 @@ void mesh::ShardingInterface::printLoopTypesAndIndexingMaps(raw_ostream &os) {
 
 namespace {
 
-// Update the given `shardingOption` according to `meshAxes` and `loopIdx`
+// Update the given `shardingOption` according to `gridAxes` and `loopIdx`
 static LogicalResult fillShardingOption(Operation *op,
                                         ShardingOption &shardingOption,
-                                        FlatSymbolRefAttr mesh,
-                                        ArrayRef<MeshAxis> meshAxes,
+                                        FlatSymbolRefAttr grid,
+                                        ArrayRef<GridAxis> gridAxes,
                                         unsigned loopIdx) {
-  if ((shardingOption.mesh && mesh && shardingOption.mesh != mesh) ||
+  if ((shardingOption.grid && grid && shardingOption.grid != grid) ||
       (!shardingOption.shardingArray[loopIdx].empty() &&
-       shardingOption.shardingArray[loopIdx] != meshAxes)) {
+       shardingOption.shardingArray[loopIdx] != gridAxes)) {
     LLVM_DEBUG(DBGS() << "sharding option conflicts on loop iterator "
                       << loopIdx << "\n");
     return failure();
@@ -239,28 +237,28 @@ static LogicalResult fillShardingOption(Operation *op,
     if (i == loopIdx)
       continue;
 
-    for (MeshAxis axis : meshAxes) {
+    for (GridAxis axis : gridAxes) {
       if (llvm::is_contained(shardingOption.shardingArray[i], axis)) {
-        LLVM_DEBUG(DBGS() << "sharding option conflicts because mesh axes "
+        LLVM_DEBUG(DBGS() << "sharding option conflicts because grid axes "
                           << axis << " duplicate");
         return failure();
       }
     }
   }
-  if (mesh)
-    shardingOption.mesh = mesh;
+  if (grid)
+    shardingOption.grid = grid;
   if (shardingOption.shardingArray[loopIdx].empty())
-    shardingOption.shardingArray[loopIdx].append(meshAxes.begin(),
-                                                 meshAxes.end());
+    shardingOption.shardingArray[loopIdx].append(gridAxes.begin(),
+                                                 gridAxes.end());
   return success();
 }
 
 } // namespace
 
 FailureOr<ShardingOption>
-mesh::detail::defaultGetShardingOption(Operation *op,
-                                       ArrayRef<MeshSharding> operandShardings,
-                                       ArrayRef<MeshSharding> resultShardings) {
+shard::detail::defaultGetShardingOption(Operation *op,
+                                        ArrayRef<Sharding> operandShardings,
+                                        ArrayRef<Sharding> resultShardings) {
   ShardingInterface shardingOp = llvm::cast<ShardingInterface>(op);
   ShardingOption shardingOption;
 
@@ -276,25 +274,25 @@ mesh::detail::defaultGetShardingOption(Operation *op,
 
   // 1. Fill sharding option based on op results
   for (auto shardingIt : llvm::enumerate(resultShardings)) {
-    MeshSharding shardAttr = shardingIt.value();
+    Sharding shardAttr = shardingIt.value();
     if (!shardAttr)
       continue;
     AffineMap map = maps[numOperands + shardingIt.index()];
     anyShardingInResultsOrOperands = true;
     if (shardAttr.getSplitAxes().empty() || map.getResults().empty()) {
-      shardingOption.mesh = shardAttr.getMeshAttr();
+      shardingOption.grid = shardAttr.getGridAttr();
     } else {
       // Handle the split axes: calculate the corresponding loop index for each
       // split axes sub-array, and then store the sub-array to
       // shardingOption[index]
       for (auto it : llvm::zip(map.getResults(), shardAttr.getSplitAxes())) {
         AffineExpr expr = std::get<0>(it);
-        ArrayRef<MeshAxis> axes = std::get<1>(it).asArrayRef();
+        ArrayRef<GridAxis> axes = std::get<1>(it).asArrayRef();
         auto dim = cast<AffineDimExpr>(expr);
         unsigned index = dim.getPosition();
         visitedLoopIndices.insert(index);
         if (failed(fillShardingOption(op, shardingOption,
-                                      shardAttr.getMeshAttr(), axes, index)))
+                                      shardAttr.getGridAttr(), axes, index)))
           return failure();
       }
     }
@@ -302,7 +300,7 @@ mesh::detail::defaultGetShardingOption(Operation *op,
 
   // 2. Fill sharding option based on operands
   for (auto shardingIt : llvm::enumerate(operandShardings)) {
-    MeshSharding shardAttr = shardingIt.value();
+    Sharding shardAttr = shardingIt.value();
     if (!shardAttr)
       continue;
 
@@ -316,7 +314,7 @@ mesh::detail::defaultGetShardingOption(Operation *op,
     // then the operands with multiple loop indices.
     for (auto it : llvm::zip(map.getResults(), shardAttr.getSplitAxes())) {
       AffineExpr expr = std::get<0>(it);
-      ArrayRef<MeshAxis> axes = std::get<1>(it).asArrayRef();
+      ArrayRef<GridAxis> axes = std::get<1>(it).asArrayRef();
       FailureOr<llvm::SmallSet<unsigned, 2>> loopIndices =
           checkOperandAffineExpr(expr, numDims);
       if (failed(loopIndices))
@@ -329,7 +327,7 @@ mesh::detail::defaultGetShardingOption(Operation *op,
         unsigned loopIdx = *loopIndices->begin();
         visitedLoopIndices.insert(loopIdx);
         if (failed(fillShardingOption(op, shardingOption,
-                                      shardAttr.getMeshAttr(), axes, loopIdx)))
+                                      shardAttr.getGridAttr(), axes, loopIdx)))
           return failure();
       }
       // If multiple loop indices correspond to a dimension of an operand, it is
@@ -361,11 +359,11 @@ mesh::detail::defaultGetShardingOption(Operation *op,
 }
 
 // Get the sharding attributed for the given result and sharding option.
-MeshSharding getSharding(OpResult result, const ShardingOption &shardingOption,
-                         AffineMap map,
-                         ArrayRef<utils::IteratorType> loopTypes) {
+static Sharding getSharding(OpResult result,
+                            const ShardingOption &shardingOption, AffineMap map,
+                            ArrayRef<utils::IteratorType> loopTypes) {
   auto resultType = cast<RankedTensorType>(result.getType());
-  SmallVector<SmallVector<MeshAxis>> splitAxes(resultType.getRank());
+  SmallVector<SmallVector<GridAxis>> splitAxes(resultType.getRank());
 
   // process the split axes
   for (auto it : llvm::enumerate(map.getResults())) {
@@ -379,25 +377,25 @@ MeshSharding getSharding(OpResult result, const ShardingOption &shardingOption,
   }
 
   removeTrailingEmptySubArray(splitAxes);
-  return MeshSharding::get(shardingOption.mesh,
-                           fromArrayOfVector(result.getContext(), splitAxes));
+  return Sharding::get(shardingOption.grid,
+                       fromArrayOfVector(result.getContext(), splitAxes));
 }
 
-static FailureOr<MeshSharding> getSharding(OpOperand &opOperand,
-                                           const ShardingOption &shardingOption,
-                                           AffineMap map) {
+static FailureOr<Sharding> getSharding(OpOperand &opOperand,
+                                       const ShardingOption &shardingOption,
+                                       AffineMap map) {
   Value operandValue = opOperand.get();
   auto operandType = dyn_cast<RankedTensorType>(operandValue.getType());
   if (!operandType) {
     if (operandValue.getType().isIntOrIndexOrFloat())
-      return MeshSharding();
+      return Sharding();
     return failure();
   }
   // 0d tensors cannot be sharded and must get replicated
   if (operandType.getRank() == 0) {
-    return MeshSharding(shardingOption.mesh);
+    return Sharding(shardingOption.grid);
   }
-  SmallVector<SmallVector<MeshAxis>> splitAxes(operandType.getRank());
+  SmallVector<SmallVector<GridAxis>> splitAxes(operandType.getRank());
   unsigned numDims = map.getNumDims();
   for (auto it : llvm::enumerate(map.getResults())) {
     int64_t idx = it.index();
@@ -422,15 +420,14 @@ static FailureOr<MeshSharding> getSharding(OpOperand &opOperand,
   }
 
   removeTrailingEmptySubArray(splitAxes);
-  return MeshSharding::get(
-      shardingOption.mesh,
+  return Sharding::get(
+      shardingOption.grid,
       fromArrayOfVector(opOperand.get().getContext(), splitAxes));
 }
 
-FailureOr<std::vector<MeshSharding>>
-mesh::detail::defaultGetShardingAnnotations(
+FailureOr<std::vector<Sharding>> shard::detail::defaultGetShardingAnnotations(
     Operation *op, const ShardingOption &shardingOption) {
-  std::vector<MeshSharding> res;
+  std::vector<Sharding> res;
 
   ShardingInterface shardingOp = llvm::cast<ShardingInterface>(op);
   SmallVector<utils::IteratorType> loopTypes =
@@ -439,7 +436,7 @@ mesh::detail::defaultGetShardingAnnotations(
   unsigned numOperands = op->getNumOperands();
 
   for (OpOperand &opOperand : op->getOpOperands()) {
-    FailureOr<MeshSharding> shardingAttr = getSharding(
+    FailureOr<Sharding> shardingAttr = ::getSharding(
         opOperand, shardingOption, maps[opOperand.getOperandNumber()]);
     if (failed(shardingAttr))
       return failure();
@@ -447,9 +444,9 @@ mesh::detail::defaultGetShardingAnnotations(
   }
 
   for (OpResult result : op->getResults()) {
-    res.push_back(getSharding(result, shardingOption,
-                              maps[numOperands + result.getResultNumber()],
-                              loopTypes));
+    res.push_back(::getSharding(result, shardingOption,
+                                maps[numOperands + result.getResultNumber()],
+                                loopTypes));
   }
 
   return res;
@@ -459,26 +456,25 @@ mesh::detail::defaultGetShardingAnnotations(
 // detail::defaultAddShardingAnnotations
 //===----------------------------------------------------------------------===//
 
-// To add a `mesh.shard` op for the given result, based on the details provided
+// To add a `shard.shard` op for the given result, based on the details provided
 // in `shardingOption`, `map`, and `loopTypes`.
 static LogicalResult addShardOp(OpBuilder &b, OpResult result,
                                 const ShardingOption &shardingOption,
                                 AffineMap map,
                                 ArrayRef<utils::IteratorType> loopTypes) {
-  MeshSharding sharding = getSharding(result, shardingOption, map, loopTypes);
+  Sharding sharding = getSharding(result, shardingOption, map, loopTypes);
   maybeInsertTargetShardingAnnotation(sharding, result, b);
 
   return success();
 }
 
-// To add a `mesh.shard` op for the given operand, based on the details provided
-// in `shardingOption`, `map`, and `loopTypes`.
+// To add a `shard.shard` op for the given operand, based on the details
+// provided in `shardingOption`, `map`, and `loopTypes`.
 static LogicalResult addShardOp(OpBuilder &b, OpOperand &opOperand,
                                 const ShardingOption &shardingOption,
                                 AffineMap map) {
 
-  FailureOr<MeshSharding> sharding =
-      getSharding(opOperand, shardingOption, map);
+  FailureOr<Sharding> sharding = getSharding(opOperand, shardingOption, map);
   if (failed(sharding)) {
     return failure();
   }
@@ -488,9 +484,9 @@ static LogicalResult addShardOp(OpBuilder &b, OpOperand &opOperand,
   return success();
 }
 
-LogicalResult mesh::detail::defaultAddShardingAnnotations(
+LogicalResult shard::detail::defaultAddShardingAnnotations(
     Operation *op, OpBuilder &b, const ShardingOption &shardingOption) {
-  assert(!shardingOption.empty && shardingOption.mesh);
+  assert(!shardingOption.empty && shardingOption.grid);
 
   ShardingInterface shardingOp = llvm::cast<ShardingInterface>(op);
   SmallVector<utils::IteratorType> loopTypes =
@@ -498,7 +494,7 @@ LogicalResult mesh::detail::defaultAddShardingAnnotations(
   SmallVector<AffineMap> maps = shardingOp.getIndexingMaps();
   unsigned numOperands = op->getNumOperands();
 
-  // 1. add mesh.shard ops for all op results
+  // 1. add shard.shard ops for all op results
   for (OpResult result : op->getResults()) {
     if (failed(addShardOp(b, result, shardingOption,
                           maps[numOperands + result.getResultNumber()],
@@ -506,7 +502,7 @@ LogicalResult mesh::detail::defaultAddShardingAnnotations(
       return failure();
   }
 
-  // 2. add mesh.shard ops for all operands
+  // 2. add shard.shard ops for all operands
   for (OpOperand &opOperand : op->getOpOperands()) {
     if (failed(addShardOp(b, opOperand, shardingOption,
                           maps[opOperand.getOperandNumber()])))
@@ -517,9 +513,8 @@ LogicalResult mesh::detail::defaultAddShardingAnnotations(
 }
 
 #ifndef NDEBUG
-static bool
-isValueCompatibleWithFullReplicationSharding(Value value,
-                                             MeshSharding sharding) {
+static bool isValueCompatibleWithFullReplicationSharding(Value value,
+                                                         Sharding sharding) {
   if (isa<RankedTensorType>(value.getType())) {
     return isFullReplication(sharding);
   }
@@ -527,60 +522,59 @@ isValueCompatibleWithFullReplicationSharding(Value value,
   return !sharding;
 }
 
-template <typename ValueRange, typename MeshShardingRage>
+template <typename ValueRange, typename ShardingRage>
 static bool
 areValuesCompatibleWithFullReplicationShardings(ValueRange &&values,
-                                                MeshShardingRage &&shardings) {
+                                                ShardingRage &&shardings) {
   if (std::size(values) != std::size(shardings)) {
     return false;
   }
-  return llvm::all_of(
-      llvm::zip_equal(std::forward<ValueRange>(values),
-                      std::forward<MeshShardingRage>(shardings)),
-      [](auto valueAndSharding) {
-        return isValueCompatibleWithFullReplicationSharding(
-            std::get<0>(valueAndSharding), std::get<1>(valueAndSharding));
-      });
+  return llvm::all_of(llvm::zip_equal(std::forward<ValueRange>(values),
+                                      std::forward<ShardingRage>(shardings)),
+                      [](auto valueAndSharding) {
+                        return isValueCompatibleWithFullReplicationSharding(
+                            std::get<0>(valueAndSharding),
+                            std::get<1>(valueAndSharding));
+                      });
 }
 #endif // NDEBUG
 
-void mesh::spmdizeFullyReplicatedOperation(
-    Operation &op, ArrayRef<Value> spmdizedOperands,
-    ArrayRef<MeshSharding> operandShardings,
-    ArrayRef<MeshSharding> resultShardings, IRMapping &spmdizationMap,
-    SymbolTableCollection &symbolTable, OpBuilder &builder) {
-  assert(spmdizedOperands.size() == operandShardings.size());
+void shard::partitionFullyReplicatedOperation(
+    Operation &op, ArrayRef<Value> partitionedOperands,
+    ArrayRef<Sharding> operandShardings, ArrayRef<Sharding> resultShardings,
+    IRMapping &partitionMap, SymbolTableCollection &symbolTable,
+    OpBuilder &builder) {
+  assert(partitionedOperands.size() == operandShardings.size());
   assert(areValuesCompatibleWithFullReplicationShardings(op.getOperands(),
                                                          operandShardings));
   assert(areValuesCompatibleWithFullReplicationShardings(op.getResults(),
                                                          resultShardings));
   // `clone` will populate the mapping of old to new results.
-  builder.clone(op, spmdizationMap);
+  builder.clone(op, partitionMap);
 }
 
-static void updateMeshAxisAssignmentForLoopIterators(
-    ArrayRef<MeshAxis> meshAxesAssignmentForTensorAxis, AffineExpr indexingExpr,
-    SmallVector<std::optional<SmallVector<MeshAxis>>>
-        &meshAxesAssignmentForLoopIterators) {
+static void updateGridAxisAssignmentForLoopIterators(
+    ArrayRef<GridAxis> gridAxesAssignmentForTensorAxis, AffineExpr indexingExpr,
+    SmallVector<std::optional<SmallVector<GridAxis>>>
+        &gridAxesAssignmentForLoopIterators) {
   AffineDimExpr affineDimExpr = cast<AffineDimExpr>(indexingExpr);
   unsigned loopIteratorIdx = affineDimExpr.getPosition();
-  if (meshAxesAssignmentForLoopIterators[loopIteratorIdx]) {
-    assert(llvm::equal(meshAxesAssignmentForTensorAxis,
-                       *meshAxesAssignmentForLoopIterators[loopIteratorIdx]));
+  if (gridAxesAssignmentForLoopIterators[loopIteratorIdx]) {
+    assert(llvm::equal(gridAxesAssignmentForTensorAxis,
+                       *gridAxesAssignmentForLoopIterators[loopIteratorIdx]));
   } else {
-    meshAxesAssignmentForLoopIterators[loopIteratorIdx] =
-        llvm::to_vector(meshAxesAssignmentForTensorAxis);
+    gridAxesAssignmentForLoopIterators[loopIteratorIdx] =
+        llvm::to_vector(gridAxesAssignmentForTensorAxis);
   }
 }
 
-ShardingArray mesh::getMeshAxisAssignmentForLoopIterators(
-    ArrayRef<MeshSharding> operandShardings,
-    ArrayRef<MeshSharding> resultShardings,
+ShardingArray shard::getGridAxisAssignmentForLoopIterators(
+    ArrayRef<Sharding> operandShardings, ArrayRef<Sharding> resultShardings,
     ArrayRef<utils::IteratorType> loopIteratorTypes,
     ArrayRef<AffineMap> indexingMaps) {
-  SmallVector<std::optional<SmallVector<MeshAxis>>>
-      meshAxisAssignmentForLoopIterators(loopIteratorTypes.size());
-  std::vector<MeshSharding> operatorAndResultShardings;
+  SmallVector<std::optional<SmallVector<GridAxis>>>
+      gridAxisAssignmentForLoopIterators(loopIteratorTypes.size());
+  std::vector<Sharding> operatorAndResultShardings;
   operatorAndResultShardings.reserve(operandShardings.size() +
                                      resultShardings.size());
   llvm::append_range(operatorAndResultShardings, operandShardings);
@@ -589,69 +583,69 @@ ShardingArray mesh::getMeshAxisAssignmentForLoopIterators(
     if (!sharding) {
       continue;
     }
-    for (auto [meshAxesAssignmentForTensorAxis, indexingExpr] :
+    for (auto [gridAxesAssignmentForTensorAxis, indexingExpr] :
          llvm::zip(sharding.getSplitAxes(), affineMap.getResults())) {
-      updateMeshAxisAssignmentForLoopIterators(
-          meshAxesAssignmentForTensorAxis.asArrayRef(), indexingExpr,
-          meshAxisAssignmentForLoopIterators);
+      updateGridAxisAssignmentForLoopIterators(
+          gridAxesAssignmentForTensorAxis.asArrayRef(), indexingExpr,
+          gridAxisAssignmentForLoopIterators);
     }
     // Missing trailing split axes means replication on those tensor dimensions.
     for (unsigned i = sharding.getSplitAxes().size();
          i < affineMap.getNumResults(); ++i) {
-      updateMeshAxisAssignmentForLoopIterators(
-          {}, affineMap.getResults()[i], meshAxisAssignmentForLoopIterators);
+      updateGridAxisAssignmentForLoopIterators(
+          {}, affineMap.getResults()[i], gridAxisAssignmentForLoopIterators);
     }
   }
 
   ShardingArray res;
-  llvm::transform(meshAxisAssignmentForLoopIterators, std::back_inserter(res),
-                  [](std::optional<SmallVector<MeshAxis>> &axes) {
+  llvm::transform(gridAxisAssignmentForLoopIterators, std::back_inserter(res),
+                  [](std::optional<SmallVector<GridAxis>> &axes) {
                     if (!axes) {
-                      return SmallVector<MeshAxis>();
+                      return SmallVector<GridAxis>();
                     };
                     return std::move(*axes);
                   });
   return res;
 }
 
-bool mesh::isAtLeastOneReductionIteratorSharded(
+bool shard::isAtLeastOneReductionIteratorSharded(
     ArrayRef<utils::IteratorType> loopIteratorTypes,
-    ArrayRef<SmallVector<MeshAxis>> meshAxisAssignmentForLoopIterators) {
-  for (auto [loopIteratorType, meshAxisAssignment] :
-       llvm::zip_equal(loopIteratorTypes, meshAxisAssignmentForLoopIterators)) {
+    ArrayRef<SmallVector<GridAxis>> gridAxisAssignmentForLoopIterators) {
+  for (auto [loopIteratorType, gridAxisAssignment] :
+       llvm::zip_equal(loopIteratorTypes, gridAxisAssignmentForLoopIterators)) {
     if (loopIteratorType == utils::IteratorType::reduction &&
-        !meshAxisAssignment.empty()) {
+        !gridAxisAssignment.empty()) {
       return true;
     }
   }
   return false;
 }
 
-SmallVector<MeshAxis> mesh::getReductionMeshAxes(
+SmallVector<GridAxis> shard::getReductionGridAxes(
     ArrayRef<utils::IteratorType> loopIteratorTypes,
-    ArrayRef<SmallVector<MeshAxis>> meshAxisAssignmentForLoopIterators) {
-  SmallVector<MeshAxis> meshAxes;
-  for (auto [loopIteratorType, meshAxisAssignment] :
-       llvm::zip_equal(loopIteratorTypes, meshAxisAssignmentForLoopIterators)) {
+    ArrayRef<SmallVector<GridAxis>> gridAxisAssignmentForLoopIterators) {
+  SmallVector<GridAxis> gridAxes;
+  for (auto [loopIteratorType, gridAxisAssignment] :
+       llvm::zip_equal(loopIteratorTypes, gridAxisAssignmentForLoopIterators)) {
     if (loopIteratorType == utils::IteratorType::reduction) {
-      llvm::append_range(meshAxes, meshAxisAssignment);
+      llvm::append_range(gridAxes, gridAxisAssignment);
     }
   }
-  return meshAxes;
+  return gridAxes;
 }
 
-void mesh::spmdizeTriviallyShardableOperation(
-    Operation &op, ArrayRef<Value> spmdizedOperands,
-    ArrayRef<MeshSharding> operandShardings,
-    ArrayRef<MeshSharding> resultShardings, IRMapping &spmdizationMap,
-    SymbolTableCollection &symbolTable, OpBuilder &builder) {
+void shard::partitionTriviallyShardableOperation(
+    Operation &op, ArrayRef<Value> partitionedOperands,
+    ArrayRef<Sharding> operandShardings, ArrayRef<Sharding> resultShardings,
+    IRMapping &partitionMap, SymbolTableCollection &symbolTable,
+    OpBuilder &builder) {
   // `clone` will populate the mapping of old to new results.
-  Operation *newOp = builder.clone(op, spmdizationMap);
+  Operation *newOp = builder.clone(op, partitionMap);
   // Set the result types to the sharded counterparts.
   for (auto [oldResult, newResult, sharding] :
        llvm::zip_equal(op.getResults(), newOp->getResults(), resultShardings)) {
     newResult.setType(shardType(
         newResult.getType(),
-        getMeshOrNull(&op, sharding.getMeshAttr(), symbolTable), sharding));
+        getGridOrNull(&op, sharding.getGridAttr(), symbolTable), sharding));
   }
 }
diff --git a/mlir/lib/Dialect/Mesh/Transforms/CMakeLists.txt b/mlir/lib/Dialect/Shard/Transforms/CMakeLists.txt
index 381bc9a..a884764 100644
--- a/mlir/lib/Dialect/Mesh/Transforms/CMakeLists.txt
+++ b/mlir/lib/Dialect/Shard/Transforms/CMakeLists.txt
@@ -1,14 +1,14 @@
-add_mlir_dialect_library(MLIRMeshTransforms
+add_mlir_dialect_library(MLIRShardTransforms
   Simplifications.cpp
   ShardingPropagation.cpp
-  Spmdization.cpp
+  Partition.cpp
   Transforms.cpp
 
   ADDITIONAL_HEADER_DIRS
-  ${MLIR_MAIN_INCLUDE_DIR}/mlir/Dialect/Mesh
+  ${MLIR_MAIN_INCLUDE_DIR}/mlir/Dialect/Shard
 
   DEPENDS
-  MLIRMeshPassIncGen
+  MLIRShardPassIncGen
   MLIRShardingInterface
 
   LINK_LIBS PUBLIC
@@ -21,7 +21,7 @@ add_mlir_dialect_library(MLIRMeshTransforms
   MLIRFuncDialect
   MLIRFunctionInterfaces
   MLIRIR
-  MLIRMeshDialect
+  MLIRShardDialect
   MLIRPass
   MLIRSupport
   MLIRTensorDialect
diff --git a/mlir/lib/Dialect/Mesh/Transforms/Spmdization.cpp b/mlir/lib/Dialect/Shard/Transforms/Partition.cpp
index 5dd744d..3e3d476 100644
--- a/mlir/lib/Dialect/Mesh/Transforms/Spmdization.cpp
+++ b/mlir/lib/Dialect/Shard/Transforms/Partition.cpp
@@ -1,4 +1,4 @@
-//===- Spmdization.cpp --------------------------------------------- C++ --===//
+//===- Partition.cpp --------------------------------------------- C++ --===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -6,11 +6,11 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "mlir/Dialect/Mesh/Transforms/Spmdization.h"
+#include "mlir/Dialect/Shard/Transforms/Partition.h"
 
-#include "mlir/Dialect/Mesh/IR/MeshDialect.h"
-#include "mlir/Dialect/Mesh/IR/MeshOps.h"
-#include "mlir/Dialect/Mesh/Interfaces/ShardingInterface.h"
+#include "mlir/Dialect/Shard/IR/ShardDialect.h"
+#include "mlir/Dialect/Shard/IR/ShardOps.h"
+#include "mlir/Dialect/Shard/Interfaces/ShardingInterface.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/IR/Builders.h"
 #include "mlir/IR/BuiltinAttributes.h"
@@ -33,7 +33,7 @@
 #include <optional>
 #include <tuple>
 
-namespace mlir::mesh {
+namespace mlir::shard {
 
 template <typename SourceAxes, typename TargetAxes>
 static bool arePartialAxesCompatible(const SourceAxes &sourceAxes,
@@ -43,52 +43,49 @@ static bool arePartialAxesCompatible(const SourceAxes &sourceAxes,
   });
 }
 
-static MeshSharding targetShardingInSplitLastAxis(MLIRContext *ctx,
-                                                  MeshSharding sourceSharding,
-                                                  int64_t splitTensorAxis,
-                                                  MeshAxis splitMeshAxis) {
-  SmallVector<MeshAxesAttr> targetShardingSplitAxes =
+static Sharding targetShardingInSplitLastAxis(MLIRContext *ctx,
+                                              Sharding sourceSharding,
+                                              int64_t splitTensorAxis,
+                                              GridAxis splitGridAxis) {
+  SmallVector<GridAxesAttr> targetShardingSplitAxes =
       llvm::to_vector(sourceSharding.getSplitAxes());
   while (static_cast<int64_t>(targetShardingSplitAxes.size()) <=
          splitTensorAxis) {
-    targetShardingSplitAxes.push_back(MeshAxesAttr::get(ctx, {}));
+    targetShardingSplitAxes.push_back(GridAxesAttr::get(ctx, {}));
   }
   auto targetSplitAxes =
       llvm::to_vector(targetShardingSplitAxes[splitTensorAxis].asArrayRef());
-  targetSplitAxes.push_back(splitMeshAxis);
+  targetSplitAxes.push_back(splitGridAxis);
   targetShardingSplitAxes[splitTensorAxis] =
-      MeshAxesAttr::get(ctx, targetSplitAxes);
-  return MeshSharding::get(sourceSharding.getMeshAttr(),
-                           targetShardingSplitAxes);
+      GridAxesAttr::get(ctx, targetSplitAxes);
+  return Sharding::get(sourceSharding.getGridAttr(), targetShardingSplitAxes);
 }
 
-// Split a replicated tensor along a mesh axis.
+// Split a replicated tensor along a grid axis.
 // E.g. [[0, 1]] -> [[0, 1, 2]].
-// Returns the spmdized target value with its sharding.
-static std::tuple<TypedValue<ShapedType>, MeshSharding>
+// Returns the partitioned target value with its sharding.
+static std::tuple<TypedValue<ShapedType>, Sharding>
 splitLastAxisInResharding(ImplicitLocOpBuilder &builder,
-                          MeshSharding sourceSharding,
-                          TypedValue<ShapedType> sourceShard, MeshOp mesh,
-                          int64_t splitTensorAxis, MeshAxis splitMeshAxis) {
+                          Sharding sourceSharding,
+                          TypedValue<ShapedType> sourceShard, GridOp grid,
+                          int64_t splitTensorAxis, GridAxis splitGridAxis) {
   TypedValue<ShapedType> targetShard = cast<TypedValue<ShapedType>>(
-      builder
-          .create<AllSliceOp>(sourceShard, mesh,
-                              ArrayRef<MeshAxis>(splitMeshAxis),
-                              splitTensorAxis)
+      AllSliceOp::create(builder, sourceShard, grid,
+                         ArrayRef<GridAxis>(splitGridAxis), splitTensorAxis)
           .getResult());
-  MeshSharding targetSharding = targetShardingInSplitLastAxis(
-      builder.getContext(), sourceSharding, splitTensorAxis, splitMeshAxis);
+  Sharding targetSharding = targetShardingInSplitLastAxis(
+      builder.getContext(), sourceSharding, splitTensorAxis, splitGridAxis);
   return {targetShard, targetSharding};
 }
 
 // Detect if the resharding is of type e.g.
 // [[0, 1]] -> [[0, 1, 2]].
-// If detected, returns the corresponding tensor axis mesh axis pair.
+// If detected, returns the corresponding tensor axis grid axis pair.
 // Does not detect insertions like
 // [[0, 1]] -> [[0, 2, 1]].
-static std::optional<std::tuple<int64_t, MeshAxis>>
-detectSplitLastAxisInResharding(MeshSharding sourceSharding,
-                                MeshSharding targetSharding) {
+static std::optional<std::tuple<int64_t, GridAxis>>
+detectSplitLastAxisInResharding(Sharding sourceSharding,
+                                Sharding targetSharding) {
   for (size_t tensorAxis = 0; tensorAxis < targetSharding.getSplitAxes().size();
        ++tensorAxis) {
     if (sourceSharding.getSplitAxes().size() > tensorAxis) {
@@ -118,16 +115,15 @@ detectSplitLastAxisInResharding(MeshSharding sourceSharding,
   return std::nullopt;
 }
 
-static std::optional<std::tuple<TypedValue<ShapedType>, MeshSharding>>
-trySplitLastAxisInResharding(ImplicitLocOpBuilder &builder, MeshOp mesh,
-                             MeshSharding sourceSharding,
-                             MeshSharding targetSharding,
+static std::optional<std::tuple<TypedValue<ShapedType>, Sharding>>
+trySplitLastAxisInResharding(ImplicitLocOpBuilder &builder, GridOp grid,
+                             Sharding sourceSharding, Sharding targetSharding,
                              TypedValue<ShapedType> sourceShard) {
   if (auto detectRes =
           detectSplitLastAxisInResharding(sourceSharding, targetSharding)) {
-    auto [tensorAxis, meshAxis] = detectRes.value();
-    return splitLastAxisInResharding(builder, sourceSharding, sourceShard, mesh,
-                                     tensorAxis, meshAxis);
+    auto [tensorAxis, gridAxis] = detectRes.value();
+    return splitLastAxisInResharding(builder, sourceSharding, sourceShard, grid,
+                                     tensorAxis, gridAxis);
   }
 
   return std::nullopt;
@@ -135,10 +131,10 @@ trySplitLastAxisInResharding(ImplicitLocOpBuilder &builder, MeshOp mesh,
 
 // Detect if the resharding is of type e.g.
 // [[0, 1, 2]] -> [[0, 1]].
-// If detected, returns the corresponding tensor axis mesh axis pair.
-static std::optional<std::tuple<int64_t, MeshAxis>>
-detectUnsplitLastAxisInResharding(MeshSharding sourceSharding,
-                                  MeshSharding targetSharding) {
+// If detected, returns the corresponding tensor axis grid axis pair.
+static std::optional<std::tuple<int64_t, GridAxis>>
+detectUnsplitLastAxisInResharding(Sharding sourceSharding,
+                                  Sharding targetSharding) {
   for (size_t tensorAxis = 0; tensorAxis < sourceSharding.getSplitAxes().size();
        ++tensorAxis) {
     if (targetSharding.getSplitAxes().size() > tensorAxis) {
@@ -165,10 +161,10 @@ detectUnsplitLastAxisInResharding(MeshSharding sourceSharding,
   return std::nullopt;
 }
 
-static MeshSharding targetShardingInUnsplitLastAxis(MLIRContext *ctx,
-                                                    MeshSharding sourceSharding,
-                                                    int64_t splitTensorAxis) {
-  SmallVector<MeshAxesAttr> targetShardingSplitAxes =
+static Sharding targetShardingInUnsplitLastAxis(MLIRContext *ctx,
+                                                Sharding sourceSharding,
+                                                int64_t splitTensorAxis) {
+  SmallVector<GridAxesAttr> targetShardingSplitAxes =
       llvm::to_vector(sourceSharding.getSplitAxes());
   assert(static_cast<int64_t>(targetShardingSplitAxes.size()) >
          splitTensorAxis);
@@ -177,9 +173,8 @@ static MeshSharding targetShardingInUnsplitLastAxis(MLIRContext *ctx,
 
   targetSplitAxes.pop_back();
   targetShardingSplitAxes[splitTensorAxis] =
-      MeshAxesAttr::get(ctx, targetSplitAxes);
-  return MeshSharding::get(sourceSharding.getMeshAttr(),
-                           targetShardingSplitAxes);
+      GridAxesAttr::get(ctx, targetSplitAxes);
+  return Sharding::get(sourceSharding.getGridAttr(), targetShardingSplitAxes);
 }
 
 static ShapedType allGatherResultShapeInUnsplitLastAxis(
@@ -190,45 +185,42 @@ static ShapedType allGatherResultShapeInUnsplitLastAxis(
   return sourceShape.cloneWith(targetShape, sourceShape.getElementType());
 }
 
-static std::tuple<TypedValue<ShapedType>, MeshSharding>
-unsplitLastAxisInResharding(ImplicitLocOpBuilder &builder,
-                            MeshSharding sourceSharding,
-                            ShapedType sourceUnshardedShape,
-                            TypedValue<ShapedType> sourceShard, MeshOp mesh,
-                            int64_t splitTensorAxis, MeshAxis splitMeshAxis) {
+static std::tuple<TypedValue<ShapedType>, Sharding> unsplitLastAxisInResharding(
+    ImplicitLocOpBuilder &builder, Sharding sourceSharding,
+    ShapedType sourceUnshardedShape, TypedValue<ShapedType> sourceShard,
+    GridOp grid, int64_t splitTensorAxis, GridAxis splitGridAxis) {
   MLIRContext *ctx = builder.getContext();
   builder.setInsertionPointAfterValue(sourceShard);
 
-  MeshSharding targetSharding =
+  Sharding targetSharding =
       targetShardingInUnsplitLastAxis(ctx, sourceSharding, splitTensorAxis);
   ShapedType allGatherResultShape = allGatherResultShapeInUnsplitLastAxis(
-      sourceShard.getType(), mesh.getShape()[splitMeshAxis], splitTensorAxis);
+      sourceShard.getType(), grid.getShape()[splitGridAxis], splitTensorAxis);
   Value allGatherResult = AllGatherOp::create(
       builder,
       RankedTensorType::get(allGatherResultShape.getShape(),
                             allGatherResultShape.getElementType()),
-      mesh.getSymName(), SmallVector<MeshAxis>({splitMeshAxis}), sourceShard,
+      grid.getSymName(), SmallVector<GridAxis>({splitGridAxis}), sourceShard,
       APInt(64, splitTensorAxis));
   ShapedType targetShape =
-      shardShapedType(sourceUnshardedShape, mesh, targetSharding);
+      shardShapedType(sourceUnshardedShape, grid, targetSharding);
   TypedValue<ShapedType> targetShard = cast<TypedValue<ShapedType>>(
       tensor::CastOp::create(builder, targetShape, allGatherResult)
           .getResult());
   return {targetShard, targetSharding};
 }
 
-static std::optional<std::tuple<TypedValue<ShapedType>, MeshSharding>>
-tryUnsplitLastAxisInResharding(ImplicitLocOpBuilder &builder, MeshOp mesh,
-                               MeshSharding sourceSharding,
-                               MeshSharding targetSharding,
+static std::optional<std::tuple<TypedValue<ShapedType>, Sharding>>
+tryUnsplitLastAxisInResharding(ImplicitLocOpBuilder &builder, GridOp grid,
+                               Sharding sourceSharding, Sharding targetSharding,
                                ShapedType sourceUnshardedShape,
                                TypedValue<ShapedType> sourceShard) {
   if (auto detectRes =
           detectUnsplitLastAxisInResharding(sourceSharding, targetSharding)) {
-    auto [tensorAxis, meshAxis] = detectRes.value();
+    auto [tensorAxis, gridAxis] = detectRes.value();
     return unsplitLastAxisInResharding(builder, sourceSharding,
-                                       sourceUnshardedShape, sourceShard, mesh,
-                                       tensorAxis, meshAxis);
+                                       sourceUnshardedShape, sourceShard, grid,
+                                       tensorAxis, gridAxis);
   }
 
   return std::nullopt;
@@ -238,10 +230,10 @@ tryUnsplitLastAxisInResharding(ImplicitLocOpBuilder &builder, MeshOp mesh,
 // [[0, 1], [2]] -> [[0], [1, 2]].
 // Only moving the last axis counts.
 // If detected, returns the corresponding (source_tensor_axis,
-// target_tensor_axis, mesh_axis) tuple.
-static std::optional<std::tuple<int64_t, int64_t, MeshAxis>>
-detectMoveLastSplitAxisInResharding(MeshSharding sourceSharding,
-                                    MeshSharding targetSharding) {
+// target_tensor_axis, grid_axis) tuple.
+static std::optional<std::tuple<int64_t, int64_t, GridAxis>>
+detectMoveLastSplitAxisInResharding(Sharding sourceSharding,
+                                    Sharding targetSharding) {
   for (size_t sourceTensorAxis = 0;
        sourceTensorAxis < sourceSharding.getSplitAxes().size();
        ++sourceTensorAxis) {
@@ -281,33 +273,32 @@ detectMoveLastSplitAxisInResharding(MeshSharding sourceSharding,
   return std::nullopt;
 }
 
-static MeshSharding targetShardingInMoveLastAxis(MLIRContext *ctx,
-                                                 MeshSharding sourceSharding,
-                                                 int64_t sourceTensorAxis,
-                                                 int64_t targetTensorAxis) {
-  SmallVector<MeshAxesAttr> targetShardingSplitAxes =
+static Sharding targetShardingInMoveLastAxis(MLIRContext *ctx,
+                                             Sharding sourceSharding,
+                                             int64_t sourceTensorAxis,
+                                             int64_t targetTensorAxis) {
+  SmallVector<GridAxesAttr> targetShardingSplitAxes =
       llvm::to_vector(sourceSharding.getSplitAxes());
   while (static_cast<int64_t>(targetShardingSplitAxes.size()) <=
          targetTensorAxis) {
-    targetShardingSplitAxes.push_back(MeshAxesAttr::get(ctx, {}));
+    targetShardingSplitAxes.push_back(GridAxesAttr::get(ctx, {}));
   }
 
   auto sourceSplitAxes =
       llvm::to_vector(targetShardingSplitAxes[sourceTensorAxis].asArrayRef());
   assert(!sourceSplitAxes.empty());
-  auto meshAxis = sourceSplitAxes.back();
+  auto gridAxis = sourceSplitAxes.back();
   sourceSplitAxes.pop_back();
   targetShardingSplitAxes[sourceTensorAxis] =
-      MeshAxesAttr::get(ctx, sourceSplitAxes);
+      GridAxesAttr::get(ctx, sourceSplitAxes);
 
   auto targetSplitAxes =
       llvm::to_vector(targetShardingSplitAxes[targetTensorAxis].asArrayRef());
-  targetSplitAxes.push_back(meshAxis);
+  targetSplitAxes.push_back(gridAxis);
   targetShardingSplitAxes[targetTensorAxis] =
-      MeshAxesAttr::get(ctx, targetSplitAxes);
+      GridAxesAttr::get(ctx, targetSplitAxes);
 
-  return MeshSharding::get(sourceSharding.getMeshAttr(),
-                           targetShardingSplitAxes);
+  return Sharding::get(sourceSharding.getGridAttr(), targetShardingSplitAxes);
 }
 
 static ShapedType allToAllResultShapeInMoveLastAxis(ShapedType sourceShape,
@@ -322,46 +313,46 @@ static ShapedType allToAllResultShapeInMoveLastAxis(ShapedType sourceShape,
   return sourceShape.cloneWith(targetShape, sourceShape.getElementType());
 }
 
-static std::tuple<TypedValue<ShapedType>, MeshSharding>
-moveLastSplitAxisInResharding(ImplicitLocOpBuilder &builder, MeshOp mesh,
-                              MeshSharding sourceSharding,
+static std::tuple<TypedValue<ShapedType>, Sharding>
+moveLastSplitAxisInResharding(ImplicitLocOpBuilder &builder, GridOp grid,
+                              Sharding sourceSharding,
                               ShapedType sourceUnshardedShape,
                               TypedValue<ShapedType> sourceShard,
                               int64_t sourceTensorAxis,
-                              int64_t targetTensorAxis, MeshAxis meshAxis) {
+                              int64_t targetTensorAxis, GridAxis gridAxis) {
   MLIRContext *ctx = builder.getContext();
   builder.setInsertionPointAfterValue(sourceShard);
 
-  MeshSharding targetSharding = targetShardingInMoveLastAxis(
+  Sharding targetSharding = targetShardingInMoveLastAxis(
       ctx, sourceSharding, sourceTensorAxis, targetTensorAxis);
   ShapedType allToAllResultShape = allToAllResultShapeInMoveLastAxis(
-      sourceShard.getType(), mesh.getShape()[meshAxis], sourceTensorAxis,
+      sourceShard.getType(), grid.getShape()[gridAxis], sourceTensorAxis,
       targetTensorAxis);
   Value allToAllResult = AllToAllOp::create(
       builder,
       RankedTensorType::get(allToAllResultShape.getShape(),
                             allToAllResultShape.getElementType()),
-      mesh.getSymName(), SmallVector<MeshAxis>({meshAxis}), sourceShard,
+      grid.getSymName(), SmallVector<GridAxis>({gridAxis}), sourceShard,
       APInt(64, targetTensorAxis), APInt(64, sourceTensorAxis));
   ShapedType targetShape =
-      shardShapedType(sourceUnshardedShape, mesh, targetSharding);
+      shardShapedType(sourceUnshardedShape, grid, targetSharding);
   TypedValue<ShapedType> targetShard = cast<TypedValue<ShapedType>>(
       tensor::CastOp::create(builder, targetShape, allToAllResult).getResult());
   return {targetShard, targetSharding};
 }
 
-static std::optional<std::tuple<TypedValue<ShapedType>, MeshSharding>>
-tryMoveLastSplitAxisInResharding(ImplicitLocOpBuilder &builder, MeshOp mesh,
-                                 MeshSharding sourceSharding,
-                                 MeshSharding targetSharding,
+static std::optional<std::tuple<TypedValue<ShapedType>, Sharding>>
+tryMoveLastSplitAxisInResharding(ImplicitLocOpBuilder &builder, GridOp grid,
+                                 Sharding sourceSharding,
+                                 Sharding targetSharding,
                                  ShapedType sourceUnshardedShape,
                                  TypedValue<ShapedType> sourceShard) {
   if (auto detectRes =
           detectMoveLastSplitAxisInResharding(sourceSharding, targetSharding)) {
-    auto [sourceTensorAxis, targetTensorAxis, meshAxis] = detectRes.value();
+    auto [sourceTensorAxis, targetTensorAxis, gridAxis] = detectRes.value();
     return moveLastSplitAxisInResharding(
-        builder, mesh, sourceSharding, sourceUnshardedShape, sourceShard,
-        sourceTensorAxis, targetTensorAxis, meshAxis);
+        builder, grid, sourceSharding, sourceUnshardedShape, sourceShard,
+        sourceTensorAxis, targetTensorAxis, gridAxis);
   }
 
   return std::nullopt;
@@ -371,10 +362,9 @@ tryMoveLastSplitAxisInResharding(ImplicitLocOpBuilder &builder, MeshOp mesh,
 // needed. A changed halo sizes requires copying the "core" of the source tensor
 // into the "core" of the destination tensor followed by an update halo
 // operation.
-static std::optional<std::tuple<TypedValue<ShapedType>, MeshSharding>>
-tryUpdateHaloInResharding(ImplicitLocOpBuilder &builder, MeshOp mesh,
-                          MeshSharding sourceSharding,
-                          MeshSharding targetSharding,
+static std::optional<std::tuple<TypedValue<ShapedType>, Sharding>>
+tryUpdateHaloInResharding(ImplicitLocOpBuilder &builder, GridOp grid,
+                          Sharding sourceSharding, Sharding targetSharding,
                           ShapedType sourceUnshardedShape,
                           TypedValue<ShapedType> sourceShard) {
   // Currently handles only cases where halo sizes differ but everything else
@@ -392,7 +382,7 @@ tryUpdateHaloInResharding(ImplicitLocOpBuilder &builder, MeshOp mesh,
   assert(((srcHaloSizes.empty() || ShapedType::isStaticShape(srcHaloSizes)) &&
           ShapedType::isStaticShape(tgtHaloSizes) &&
           sourceShard.getType().hasStaticShape()) &&
-         "dynamic shapes/halos are not supported yet for mesh-spmdization");
+         "dynamic shapes/halos are not supported yet for shard-partition");
   auto rank = sourceShard.getType().getRank();
   auto splitAxes = sourceSharding.getSplitAxes();
   SmallVector<int64_t> srcCoreOffs(rank, 0), tgtCoreOffs(rank, 0),
@@ -428,56 +418,55 @@ tryUpdateHaloInResharding(ImplicitLocOpBuilder &builder, MeshOp mesh,
 
   // Finally update the halo.
   auto updateHaloResult =
-      builder
-          .create<UpdateHaloOp>(
-              sourceShard.getLoc(),
-              RankedTensorType::get(outShape,
-                                    sourceShard.getType().getElementType()),
-              initOprnd, mesh.getSymName(),
-              MeshAxesArrayAttr::get(builder.getContext(),
-                                     sourceSharding.getSplitAxes()),
-              targetSharding.getDynamicHaloSizes(),
-              targetSharding.getStaticHaloSizes())
+      UpdateHaloOp::create(
+          builder, sourceShard.getLoc(),
+          RankedTensorType::get(outShape,
+                                sourceShard.getType().getElementType()),
+          initOprnd, grid.getSymName(),
+          GridAxesArrayAttr::get(builder.getContext(),
+                                 sourceSharding.getSplitAxes()),
+          targetSharding.getDynamicHaloSizes(),
+          targetSharding.getStaticHaloSizes())
           .getResult();
   return std::make_tuple(cast<TypedValue<ShapedType>>(updateHaloResult),
                          targetSharding);
 }
 
-// Handles only resharding on a 1D mesh.
+// Handles only resharding on a 1D shard.
 // Currently the sharded tensor axes must be exactly divisible by the single
-// mesh axis size.
+// grid axis size.
 static TypedValue<ShapedType>
-reshardOn1DMesh(ImplicitLocOpBuilder &builder, MeshOp mesh,
-                MeshSharding sourceSharding, MeshSharding targetSharding,
+reshardOn1DGrid(ImplicitLocOpBuilder &builder, GridOp grid,
+                Sharding sourceSharding, Sharding targetSharding,
                 TypedValue<ShapedType> sourceUnshardedValue,
                 TypedValue<ShapedType> sourceShard) {
   assert(sourceShard.getType() ==
-         shardShapedType(sourceUnshardedValue.getType(), mesh, sourceSharding));
+         shardShapedType(sourceUnshardedValue.getType(), grid, sourceSharding));
   [[maybe_unused]] ShapedType targetShardType =
-      shardShapedType(sourceUnshardedValue.getType(), mesh, targetSharding);
+      shardShapedType(sourceUnshardedValue.getType(), grid, targetSharding);
   assert(sourceShard.getType().getRank() == targetShardType.getRank());
-  assert(mesh.getRank() == 1 && "Only 1D meshes are currently supported.");
+  assert(grid.getRank() == 1 && "Only 1D grides are currently supported.");
 
   if (sourceSharding == targetSharding) {
     return sourceShard;
   }
 
   TypedValue<ShapedType> targetShard;
-  MeshSharding actualTargetSharding;
+  Sharding actualTargetSharding;
   if (sourceSharding.getStaticShardedDimsOffsets().empty() &&
       targetSharding.getStaticShardedDimsOffsets().empty() &&
       sourceSharding.getStaticHaloSizes().empty() &&
       targetSharding.getStaticHaloSizes().empty()) {
     if (auto tryRes = tryMoveLastSplitAxisInResharding(
-            builder, mesh, sourceSharding, targetSharding,
+            builder, grid, sourceSharding, targetSharding,
             sourceUnshardedValue.getType(), sourceShard)) {
       std::tie(targetShard, actualTargetSharding) = tryRes.value();
     } else if (auto tryRes =
-                   trySplitLastAxisInResharding(builder, mesh, sourceSharding,
+                   trySplitLastAxisInResharding(builder, grid, sourceSharding,
                                                 targetSharding, sourceShard)) {
       std::tie(targetShard, actualTargetSharding) = tryRes.value();
     } else if (auto tryRes = tryUnsplitLastAxisInResharding(
-                   builder, mesh, sourceSharding, targetSharding,
+                   builder, grid, sourceSharding, targetSharding,
                    sourceUnshardedValue.getType(), sourceShard)) {
       std::tie(targetShard, actualTargetSharding) = tryRes.value();
     }
@@ -488,9 +477,8 @@ reshardOn1DMesh(ImplicitLocOpBuilder &builder, MeshOp mesh,
   return targetShard;
 }
 
-TypedValue<ShapedType> reshard(ImplicitLocOpBuilder &builder, MeshOp mesh,
-                               MeshSharding sourceSharding,
-                               MeshSharding targetSharding,
+TypedValue<ShapedType> reshard(ImplicitLocOpBuilder &builder, GridOp grid,
+                               Sharding sourceSharding, Sharding targetSharding,
                                TypedValue<ShapedType> sourceUnshardedValue,
                                TypedValue<ShapedType> sourceShard) {
   // If source and destination sharding are the same, no need to do anything.
@@ -500,28 +488,28 @@ TypedValue<ShapedType> reshard(ImplicitLocOpBuilder &builder, MeshOp mesh,
   }
 
   // Tries to handle the case where the resharding is needed because the halo
-  // sizes are different. Supports arbitrary mesh dimensionality.
+  // sizes are different. Supports arbitrary grid dimensionality.
   if (auto tryRes = tryUpdateHaloInResharding(
-          builder, mesh, sourceSharding, targetSharding,
+          builder, grid, sourceSharding, targetSharding,
           sourceUnshardedValue.getType(), sourceShard)) {
     return std::get<0>(tryRes.value()); // targetShard
   }
 
-  // Resort to handling only 1D meshes since the general case is complicated if
+  // Resort to handling only 1D grids since the general case is complicated if
   // it needs to be communication efficient in terms of minimizing the data
   // transfered between devices.
-  return reshardOn1DMesh(builder, mesh, sourceSharding, targetSharding,
+  return reshardOn1DGrid(builder, grid, sourceSharding, targetSharding,
                          sourceUnshardedValue, sourceShard);
 }
 
-TypedValue<ShapedType> reshard(OpBuilder &builder, MeshOp mesh, ShardOp source,
+TypedValue<ShapedType> reshard(OpBuilder &builder, GridOp grid, ShardOp source,
                                ShardOp target,
                                TypedValue<ShapedType> sourceShardValue) {
   assert(source.getResult() == target.getSrc());
   auto sourceSharding = source.getSharding();
   auto targetSharding = target.getSharding();
   ImplicitLocOpBuilder implicitLocOpBuilder(target->getLoc(), builder);
-  return reshard(implicitLocOpBuilder, mesh, sourceSharding, targetSharding,
+  return reshard(implicitLocOpBuilder, grid, sourceSharding, targetSharding,
                  cast<TypedValue<ShapedType>>(source.getSrc()),
                  sourceShardValue);
 }
@@ -530,21 +518,21 @@ TypedValue<ShapedType> reshard(OpBuilder &builder, ShardOp source,
                                ShardOp target,
                                TypedValue<ShapedType> sourceShardValue,
                                SymbolTableCollection &symbolTableCollection) {
-  MeshOp srcMesh = getMesh(source, symbolTableCollection);
-  assert(srcMesh && srcMesh == getMesh(target, symbolTableCollection));
-  return reshard(builder, srcMesh, source, target, sourceShardValue);
+  GridOp srcGrid = getGrid(source, symbolTableCollection);
+  assert(srcGrid && srcGrid == getGrid(target, symbolTableCollection));
+  return reshard(builder, srcGrid, source, target, sourceShardValue);
 }
 
 void reshardingRegisterDependentDialects(DialectRegistry &registry) {
-  registry.insert<mesh::MeshDialect, tensor::TensorDialect>();
+  registry.insert<shard::ShardDialect, tensor::TensorDialect>();
 }
 
-#define GEN_PASS_DEF_SPMDIZATION
-#include "mlir/Dialect/Mesh/Transforms/Passes.h.inc"
+#define GEN_PASS_DEF_PARTITION
+#include "mlir/Dialect/Shard/Transforms/Passes.h.inc"
 
 using UnshardedToShardedValueMap = DenseMap<Value, Value>;
 
-// Get the types of block arguments for an spmdized block.
+// Get the types of block arguments for an partitioned block.
 // Reads the sharding annotations of the arguments to deduce the sharded types.
 // Types that are not ranked tensors are left unchanged.
 SmallVector<Type>
@@ -563,35 +551,36 @@ shardedBlockArgumentTypes(Block &block,
         Operation *useOp = *rankedTensorArg.getUsers().begin();
         ShardOp shardOp = llvm::dyn_cast<ShardOp>(useOp);
         assert(shardOp);
-        MeshOp mesh = getMesh(shardOp, symbolTableCollection);
-        return cast<Type>(shardShapedType(rankedTensorArg.getType(), mesh,
+        GridOp grid = getGrid(shardOp, symbolTableCollection);
+        return cast<Type>(shardShapedType(rankedTensorArg.getType(), grid,
                                           shardOp.getSharding()));
       });
   return res;
 }
 
-static LogicalResult spmdizeOperation(
-    Operation &op, ArrayRef<Value> spmdizedOperands,
-    ArrayRef<MeshSharding> operandShardings,
-    ArrayRef<MeshSharding> resultShardings, IRMapping &spmdizationMap,
-    SymbolTableCollection &symbolTableCollection, OpBuilder &builder) {
+static LogicalResult
+partitionOperation(Operation &op, ArrayRef<Value> partitionedOperands,
+                   ArrayRef<Sharding> operandShardings,
+                   ArrayRef<Sharding> resultShardings, IRMapping &partitionMap,
+                   SymbolTableCollection &symbolTableCollection,
+                   OpBuilder &builder) {
   ShardingInterface shardingInterface = llvm::dyn_cast<ShardingInterface>(op);
   if (!shardingInterface) {
     // If there is no sharding interface we are conservative and assume that
     // the op should be fully replicated no all devices.
-    spmdizeFullyReplicatedOperation(op, spmdizedOperands, operandShardings,
-                                    resultShardings, spmdizationMap,
-                                    symbolTableCollection, builder);
+    partitionFullyReplicatedOperation(op, partitionedOperands, operandShardings,
+                                      resultShardings, partitionMap,
+                                      symbolTableCollection, builder);
   } else {
-    if (failed(shardingInterface.spmdize(spmdizedOperands, operandShardings,
-                                         resultShardings, spmdizationMap,
-                                         symbolTableCollection, builder))) {
+    if (failed(shardingInterface.partition(
+            partitionedOperands, operandShardings, resultShardings,
+            partitionMap, symbolTableCollection, builder))) {
       return failure();
     }
   }
 
-  assert(llvm::all_of(op.getResults(), [&spmdizationMap](OpResult result) {
-    return spmdizationMap.contains(result);
+  assert(llvm::all_of(op.getResults(), [&partitionMap](OpResult result) {
+    return partitionMap.contains(result);
   }));
 
   return success();
@@ -599,87 +588,87 @@ static LogicalResult spmdizeOperation(
 
 // Retrieve the sharding annotations for the operands of the given operation.
 // If the type is not a ranked tensor it is not require to have an annotation.
-static std::vector<MeshSharding> getOperandShardings(Operation &op) {
-  std::vector<MeshSharding> res;
+static std::vector<Sharding> getOperandShardings(Operation &op) {
+  std::vector<Sharding> res;
   res.reserve(op.getNumOperands());
   llvm::transform(op.getOperands(), std::back_inserter(res), [](Value operand) {
     TypedValue<RankedTensorType> rankedTensor =
         dyn_cast<TypedValue<RankedTensorType>>(operand);
     if (!rankedTensor || rankedTensor.getType().getRank() == 0) {
-      return MeshSharding();
+      return Sharding();
     }
 
     Operation *definingOp = operand.getDefiningOp();
     assert(definingOp);
     ShardOp shardOp = llvm::cast<ShardOp>(definingOp);
-    return MeshSharding(shardOp.getSharding());
+    return Sharding(shardOp.getSharding());
   });
   return res;
 }
 
 // Retrieve the sharding annotations for the results of the given operation.
 // If the type is not a ranked tensor it is not require to have an annotation.
-static std::vector<MeshSharding> getResultShardings(Operation &op) {
-  std::vector<MeshSharding> res;
+static std::vector<Sharding> getResultShardings(Operation &op) {
+  std::vector<Sharding> res;
   res.reserve(op.getNumResults());
   llvm::transform(
       op.getResults(), std::back_inserter(res), [&op](OpResult result) {
         if (!result.hasOneUse() || result.use_empty()) {
-          return MeshSharding();
+          return Sharding();
         }
         TypedValue<RankedTensorType> rankedTensor =
             dyn_cast<TypedValue<RankedTensorType>>(result);
         if (!rankedTensor) {
-          return MeshSharding();
+          return Sharding();
         }
         Operation *userOp = *result.getUsers().begin();
         ShardOp shardOp = llvm::dyn_cast<ShardOp>(userOp);
         if (shardOp) {
-          return MeshSharding(shardOp.getSharding());
+          return Sharding(shardOp.getSharding());
         }
         if (rankedTensor.getType().getRank() == 0) {
           // This is a 0d tensor result without explicit sharding.
-          // Find mesh symbol from operands, if any.
-          // Shardings without mesh are not always fully supported yet.
+          // Find grid symbol from operands, if any.
+          // Shardings without grid are not always fully supported yet.
           for (auto operand : op.getOperands()) {
             if (auto sharding = operand.getDefiningOp<ShardingOp>()) {
-              return MeshSharding(sharding.getMeshAttr());
+              return Sharding(sharding.getGridAttr());
             }
           }
         }
-        return MeshSharding();
+        return Sharding();
       });
   return res;
 }
 
 static LogicalResult
-spmdizeOperation(ShardOp shardOp, IRMapping &spmdizationMap,
-                 SymbolTableCollection &symbolTableCollection,
-                 OpBuilder &builder) {
-  Value targetSpmdValue;
+partitionOperation(ShardOp shardOp, IRMapping &partitionMap,
+                   SymbolTableCollection &symbolTableCollection,
+                   OpBuilder &builder) {
+  Value targetPartitionValue;
 
   // Check if 2 shard ops are chained. If not there is no need for resharding
   // as the source and target shared the same sharding.
   ShardOp srcShardOp = shardOp.getSrc().getDefiningOp<ShardOp>();
   if (!srcShardOp) {
-    targetSpmdValue = spmdizationMap.lookup(shardOp.getSrc());
+    targetPartitionValue = partitionMap.lookup(shardOp.getSrc());
   } else {
     // Insert resharding.
-    TypedValue<ShapedType> srcSpmdValue =
-        cast<TypedValue<ShapedType>>(spmdizationMap.lookup(srcShardOp));
-    targetSpmdValue = reshard(builder, srcShardOp, shardOp, srcSpmdValue,
-                              symbolTableCollection);
+    TypedValue<ShapedType> srcPartitionValue =
+        cast<TypedValue<ShapedType>>(partitionMap.lookup(srcShardOp));
+    targetPartitionValue = reshard(builder, srcShardOp, shardOp,
+                                   srcPartitionValue, symbolTableCollection);
   }
 
-  assert(!spmdizationMap.contains(shardOp.getResult()));
-  spmdizationMap.map(shardOp.getResult(), targetSpmdValue);
+  assert(!partitionMap.contains(shardOp.getResult()));
+  partitionMap.map(shardOp.getResult(), targetPartitionValue);
   return success();
 }
 
 static LogicalResult
-spmdizeOperation(Operation &op, IRMapping &spmdizationMap,
-                 SymbolTableCollection &symbolTableCollection,
-                 OpBuilder &builder) {
+partitionOperation(Operation &op, IRMapping &partitionMap,
+                   SymbolTableCollection &symbolTableCollection,
+                   OpBuilder &builder) {
   if (isa<ShardingOp>(op)) {
     return success();
   }
@@ -689,30 +678,31 @@ spmdizeOperation(Operation &op, IRMapping &spmdizationMap,
       return op.emitError("expected a shard op as source of get_sharding");
     }
     auto newSharding = builder.clone(*shardOp.getSharding().getDefiningOp());
-    spmdizationMap.map(op.getResult(0), newSharding->getResult(0));
+    partitionMap.map(op.getResult(0), newSharding->getResult(0));
     return success();
   }
 
   ShardOp shardOp = llvm::dyn_cast<ShardOp>(op);
   if (shardOp) {
-    return spmdizeOperation(shardOp, spmdizationMap, symbolTableCollection,
-                            builder);
+    return partitionOperation(shardOp, partitionMap, symbolTableCollection,
+                              builder);
   }
 
-  SmallVector<Value> spmdizedOperands;
-  llvm::transform(op.getOperands(), std::back_inserter(spmdizedOperands),
-                  [&spmdizationMap](Value operand) {
-                    assert(spmdizationMap.contains(operand));
-                    return spmdizationMap.lookup(operand);
+  SmallVector<Value> partitionedOperands;
+  llvm::transform(op.getOperands(), std::back_inserter(partitionedOperands),
+                  [&partitionMap](Value operand) {
+                    assert(partitionMap.contains(operand));
+                    return partitionMap.lookup(operand);
                   });
-  return spmdizeOperation(op, spmdizedOperands, getOperandShardings(op),
-                          getResultShardings(op), spmdizationMap,
-                          symbolTableCollection, builder);
+  return partitionOperation(op, partitionedOperands, getOperandShardings(op),
+                            getResultShardings(op), partitionMap,
+                            symbolTableCollection, builder);
 }
 
-static LogicalResult spmdizeBlock(Block &block, IRMapping &spmdizationMap,
-                                  SymbolTableCollection &symbolTableCollection,
-                                  OpBuilder &builder) {
+static LogicalResult
+partitionBlock(Block &block, IRMapping &partitionMap,
+               SymbolTableCollection &symbolTableCollection,
+               OpBuilder &builder) {
 
   SmallVector<Location> argLocations;
   llvm::transform(block.getArguments(), std::back_inserter(argLocations),
@@ -720,16 +710,16 @@ static LogicalResult spmdizeBlock(Block &block, IRMapping &spmdizationMap,
   Block *newBlock = builder.createBlock(
       block.getParent(), {},
       shardedBlockArgumentTypes(block, symbolTableCollection), argLocations);
-  for (auto [unshardedBlockArg, spmdizedBlockArg] :
+  for (auto [unshardedBlockArg, partitionedBlockArg] :
        llvm::zip(block.getArguments(), newBlock->getArguments())) {
-    spmdizationMap.map(unshardedBlockArg, spmdizedBlockArg);
+    partitionMap.map(unshardedBlockArg, partitionedBlockArg);
   }
 
   OpBuilder::InsertionGuard insertionGuard(builder);
   builder.setInsertionPointToEnd(newBlock);
   for (Operation &op : block.getOperations()) {
-    if (failed(spmdizeOperation(op, spmdizationMap, symbolTableCollection,
-                                builder))) {
+    if (failed(partitionOperation(op, partitionMap, symbolTableCollection,
+                                  builder))) {
       return failure();
     }
   }
@@ -738,8 +728,8 @@ static LogicalResult spmdizeBlock(Block &block, IRMapping &spmdizationMap,
 }
 
 static LogicalResult
-spmdizeFuncOp(FunctionOpInterface op, IRMapping &spmdizationMap,
-              SymbolTableCollection &symbolTableCollection) {
+partitionFuncOp(FunctionOpInterface op, IRMapping &partitionMap,
+                SymbolTableCollection &symbolTableCollection) {
   OpBuilder builder(op.getFunctionBody());
 
   // Snapshot the original blocks to not mess up the iteration when adding new
@@ -753,8 +743,8 @@ spmdizeFuncOp(FunctionOpInterface op, IRMapping &spmdizationMap,
   }
 
   for (Block *block : originalBlocks) {
-    if (failed(spmdizeBlock(*block, spmdizationMap, symbolTableCollection,
-                            builder))) {
+    if (failed(partitionBlock(*block, partitionMap, symbolTableCollection,
+                              builder))) {
       return failure();
     }
   }
@@ -787,22 +777,22 @@ spmdizeFuncOp(FunctionOpInterface op, IRMapping &spmdizationMap,
 
 namespace {
 
-struct Spmdization : public impl::SpmdizationBase<Spmdization> {
+struct Partition : public impl::PartitionBase<Partition> {
   void runOnOperation() override {
-    IRMapping spmdizationMap;
+    IRMapping partitionMap;
     SymbolTableCollection symbolTableCollection;
-    if (failed(spmdizeFuncOp(getOperation(), spmdizationMap,
-                             symbolTableCollection))) {
+    if (failed(partitionFuncOp(getOperation(), partitionMap,
+                               symbolTableCollection))) {
       return signalPassFailure();
     }
   }
 
   void getDependentDialects(DialectRegistry &registry) const override {
     reshardingRegisterDependentDialects(registry);
-    registry.insert<mesh::MeshDialect>();
+    registry.insert<shard::ShardDialect>();
   }
 };
 
 } // namespace
 
-} // namespace mlir::mesh
+} // namespace mlir::shard
diff --git a/mlir/lib/Dialect/Mesh/Transforms/ShardingPropagation.cpp b/mlir/lib/Dialect/Shard/Transforms/ShardingPropagation.cpp
index 0a68376..a647128c 100644
--- a/mlir/lib/Dialect/Mesh/Transforms/ShardingPropagation.cpp
+++ b/mlir/lib/Dialect/Shard/Transforms/ShardingPropagation.cpp
@@ -6,11 +6,11 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "mlir/Dialect/Mesh/Transforms/Passes.h"
+#include "mlir/Dialect/Shard/Transforms/Passes.h"
 
-#include "mlir/Dialect/Mesh/IR/MeshDialect.h"
-#include "mlir/Dialect/Mesh/IR/MeshOps.h"
-#include "mlir/Dialect/Mesh/Interfaces/ShardingInterface.h"
+#include "mlir/Dialect/Shard/IR/ShardDialect.h"
+#include "mlir/Dialect/Shard/IR/ShardOps.h"
+#include "mlir/Dialect/Shard/Interfaces/ShardingInterface.h"
 #include "mlir/IR/Verifier.h"
 #include "mlir/Interfaces/FunctionInterfaces.h"
 #include "llvm/ADT/STLExtras.h"
@@ -21,17 +21,17 @@
 #include <vector>
 
 namespace mlir {
-namespace mesh {
+namespace shard {
 #define GEN_PASS_DEF_SHARDINGPROPAGATION
-#include "mlir/Dialect/Mesh/Transforms/Passes.h.inc"
-} // namespace mesh
+#include "mlir/Dialect/Shard/Transforms/Passes.h.inc"
+} // namespace shard
 } // namespace mlir
 
 #define DEBUG_TYPE "sharding-propagation"
 #define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE << "]: ")
 
 using namespace mlir;
-using namespace mlir::mesh;
+using namespace mlir::shard;
 
 enum class ReshardingRquirementKind {
   NO_RESHARDING = 0,
@@ -68,7 +68,7 @@ static llvm::raw_ostream &operator<<(llvm::raw_ostream &stream,
 
 [[maybe_unused]] static llvm::raw_ostream &operator<<(llvm::raw_ostream &stream,
                                                       const ShardingOption &v) {
-  return stream << "{empty = " << v.empty << ", mesh" << v.mesh
+  return stream << "{empty = " << v.empty << ", grid" << v.grid
                 << ", shardingArray = " << v.shardingArray << "}";
 }
 
@@ -105,15 +105,15 @@ operator<<(llvm::raw_ostream &stream, ReshardingRquirementKind v) {
 // specific shardings. For example, mustShardings = [shard0, None] and
 // optionalShardings = [None, shard1], the result will be [[shard0, shard1],
 // [shard0, None]]
-static SmallVector<std::vector<MeshSharding>>
-getOrderedPossibleShardingAttrs(ArrayRef<MeshSharding> mustShardings,
-                                ArrayRef<MeshSharding> optionalShardings) {
-  SmallVector<std::vector<MeshSharding>> allShardingAttrs;
-  std::vector<MeshSharding> curShardingAttrs;
+static SmallVector<std::vector<Sharding>>
+getOrderedPossibleShardingAttrs(ArrayRef<Sharding> mustShardings,
+                                ArrayRef<Sharding> optionalShardings) {
+  SmallVector<std::vector<Sharding>> allShardingAttrs;
+  std::vector<Sharding> curShardingAttrs;
 
   std::function<void(size_t)> dfsCreateShardingAttrs = [&](size_t i) {
     if (i == mustShardings.size()) {
-      allShardingAttrs.push_back(std::vector<MeshSharding>(curShardingAttrs));
+      allShardingAttrs.push_back(std::vector<Sharding>(curShardingAttrs));
       return;
     }
 
@@ -147,14 +147,14 @@ getOrderedPossibleShardingAttrs(ArrayRef<MeshSharding> mustShardings,
 // 1. No resharding is required (all existing annotations are compatible).
 // 2. No resharding for operands/results that have annotation specifically
 //   targeting this operation. This means
-//   * operands that are the result of `mesh.shard` ops marked with
+//   * operands that are the result of `shard.shard` ops marked with
 //     `annotate_for_users`.
-//   * results that are annotated with `mesh.shard` ops without
+//   * results that are annotated with `shard.shard` ops without
 //     `annotate_for_users`.
 // 3. All other cases. Resharding is required for operands/results with
 //   annotation targeting explicitly this operation.
 ReshardingRquirementKind getReshardingRquirementKind(
-    Operation *op, const std::vector<MeshSharding> &operandAndResultShardings) {
+    Operation *op, const std::vector<Sharding> &operandAndResultShardings) {
   ReshardingRquirementKind res = ReshardingRquirementKind::NO_RESHARDING;
 
   size_t operandsCount = op->getOperands().size();
@@ -213,14 +213,13 @@ ReshardingRquirementKind getReshardingRquirementKind(
 // 3. Resharding of existing explicit sharding annotations for this op.
 static FailureOr<ShardingOption> selectShardingOption(
     ShardingInterface shardingOp,
-    ArrayRef<std::vector<MeshSharding>> possibleOperandShardingAttrs,
-    ArrayRef<std::vector<MeshSharding>> possibleResultShardingAttrs) {
+    ArrayRef<std::vector<Sharding>> possibleOperandShardingAttrs,
+    ArrayRef<std::vector<Sharding>> possibleResultShardingAttrs) {
   SmallVector<std::tuple<ShardingOption, ReshardingRquirementKind>>
       shardingOptionsAndReshardingRequirements;
 
-  for (ArrayRef<MeshSharding> resultShardings : possibleResultShardingAttrs) {
-    for (ArrayRef<MeshSharding> operandShardings :
-         possibleOperandShardingAttrs) {
+  for (ArrayRef<Sharding> resultShardings : possibleResultShardingAttrs) {
+    for (ArrayRef<Sharding> operandShardings : possibleOperandShardingAttrs) {
       FailureOr<ShardingOption> shardingOption =
           shardingOp.getShardingOption(operandShardings, resultShardings);
       if (failed(shardingOption) || shardingOption->empty) {
@@ -231,7 +230,7 @@ static FailureOr<ShardingOption> selectShardingOption(
       // They may be missing some annotations.
       // Whatever is returned by getShardingAnnotations is exactly what the op
       // needs.
-      FailureOr<std::vector<MeshSharding>> operandAndResultShardings =
+      FailureOr<std::vector<Sharding>> operandAndResultShardings =
           shardingOp.getShardingAnnotations(*shardingOption);
       if (failed(operandAndResultShardings)) {
         return failure();
@@ -276,13 +275,13 @@ static FailureOr<ShardingOption> selectShardingOption(
 // For each operation that implements the ShardingInterface, infer the sharding
 // option of the operation from its operands and/or results using the
 // `getShardingOption` method. If the inferred sharding option is not empty, add
-// a `mesh.shard` operation for all remaining operands and results that do not
+// a `shard.shard` operation for all remaining operands and results that do not
 // have sharding annotations.
 static LogicalResult visitOp(Operation *op, OpBuilder &builder) {
   ShardingInterface shardingOp = llvm::dyn_cast<ShardingInterface>(op);
   if (op->hasTrait<OpTrait::IsTerminator>() ||
       (op->hasTrait<OpTrait::ConstantLike>() && !shardingOp) ||
-      llvm::isa<mesh::ShardOp, mesh::ShardingOp, mesh::GetShardingOp>(op))
+      llvm::isa<shard::ShardOp, shard::ShardingOp, shard::GetShardingOp>(op))
     return success();
 
   if (!shardingOp) {
@@ -290,14 +289,13 @@ static LogicalResult visitOp(Operation *op, OpBuilder &builder) {
     return failure();
   }
 
-  // collect MeshSharding from results
-  std::vector<MeshSharding> allowConflictsResultShardings;
+  // collect Sharding from results
+  std::vector<Sharding> allowConflictsResultShardings;
   allowConflictsResultShardings.resize(op->getNumResults());
-  std::vector<MeshSharding> resultMustShardings;
+  std::vector<Sharding> resultMustShardings;
   resultMustShardings.resize(op->getNumResults());
   for (OpResult result : op->getResults()) {
-    FailureOr<std::pair<bool, MeshSharding>> maybeShardAttr =
-        getMeshSharding(result);
+    FailureOr<std::pair<bool, Sharding>> maybeShardAttr = getSharding(result);
     if (failed(maybeShardAttr))
       continue;
     if (!maybeShardAttr->first)
@@ -307,14 +305,14 @@ static LogicalResult visitOp(Operation *op, OpBuilder &builder) {
           maybeShardAttr->second;
   }
 
-  // collect MeshSharding from operands
-  std::vector<MeshSharding> allowConflictsOperandShardings;
+  // collect Sharding from operands
+  std::vector<Sharding> allowConflictsOperandShardings;
   allowConflictsOperandShardings.resize(op->getNumOperands());
-  std::vector<MeshSharding> operandMustShardings;
+  std::vector<Sharding> operandMustShardings;
   operandMustShardings.resize(op->getNumOperands());
   for (OpOperand &opOperand : op->getOpOperands()) {
-    FailureOr<std::pair<bool, MeshSharding>> maybeShardAttr =
-        getMeshSharding(opOperand);
+    FailureOr<std::pair<bool, Sharding>> maybeShardAttr =
+        getSharding(opOperand);
     if (failed(maybeShardAttr))
       continue;
 
@@ -327,10 +325,10 @@ static LogicalResult visitOp(Operation *op, OpBuilder &builder) {
   }
 
   // try to get the sharding option
-  SmallVector<std::vector<MeshSharding>> possibleOperandShardingAttrs =
+  SmallVector<std::vector<Sharding>> possibleOperandShardingAttrs =
       getOrderedPossibleShardingAttrs(operandMustShardings,
                                       allowConflictsOperandShardings);
-  SmallVector<std::vector<MeshSharding>> possibleResultShardingAttrs =
+  SmallVector<std::vector<Sharding>> possibleResultShardingAttrs =
       getOrderedPossibleShardingAttrs(resultMustShardings,
                                       allowConflictsResultShardings);
   FailureOr<ShardingOption> shardingOption = selectShardingOption(
@@ -358,7 +356,7 @@ static LogicalResult visitOp(Operation *op, OpBuilder &builder) {
 // ShardingPropagation
 //===----------------------------------------------------------------------===//
 struct ShardingPropagation
-    : public mesh::impl::ShardingPropagationBase<ShardingPropagation> {
+    : public shard::impl::ShardingPropagationBase<ShardingPropagation> {
 
   using ShardingPropagationBase<ShardingPropagation>::ShardingPropagationBase;
 
diff --git a/mlir/lib/Dialect/Mesh/Transforms/Simplifications.cpp b/mlir/lib/Dialect/Shard/Transforms/Simplifications.cpp
index 1315502..a17671e 100644
--- a/mlir/lib/Dialect/Mesh/Transforms/Simplifications.cpp
+++ b/mlir/lib/Dialect/Shard/Transforms/Simplifications.cpp
@@ -1,4 +1,4 @@
-//===- Simplifications.cpp - Mesh Simplifications ---------------*- C++ -*-===//
+//===- Simplifications.cpp - Shard Simplifications -_------------*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
@@ -6,10 +6,10 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "mlir/Dialect/Mesh/Transforms/Simplifications.h"
+#include "mlir/Dialect/Shard/Transforms/Simplifications.h"
 #include "TransformsDetail.h"
 #include "mlir/Dialect/Arith/IR/Arith.h"
-#include "mlir/Dialect/Mesh/IR/MeshOps.h"
+#include "mlir/Dialect/Shard/IR/ShardOps.h"
 #include "mlir/IR/BuiltinTypeInterfaces.h"
 #include "mlir/IR/PatternMatch.h"
 #include "mlir/IR/SymbolTable.h"
@@ -18,7 +18,7 @@
 #include <numeric>
 
 namespace mlir {
-namespace mesh {
+namespace shard {
 
 void populateSimplificationPatterns(
     RewritePatternSet &patterns, SymbolTableCollection &symbolTableCollection) {
@@ -52,53 +52,53 @@ namespace {
 // DialectFoldInterface, because it needs a SymbolTableCollection to cache the
 // symbol tables.
 // We can't use DialectFoldInterface since the cache may be invalidated by some
-// pass changing the referenced MeshOp ops.
-struct MeshShapeFolder
-    : OpRewritePatternWithSymbolTableCollection<MeshShapeOp> {
+// pass changing the referenced GridOp ops.
+struct GridShapeFolder
+    : OpRewritePatternWithSymbolTableCollection<GridShapeOp> {
   using OpRewritePatternWithSymbolTableCollection::
       OpRewritePatternWithSymbolTableCollection;
-  LogicalResult matchAndRewrite(MeshShapeOp op,
+  LogicalResult matchAndRewrite(GridShapeOp op,
                                 PatternRewriter &rewriter) const override {
     ImplicitLocOpBuilder builder(op->getLoc(), rewriter);
-    MeshOp mesh = symbolTableCollection.lookupNearestSymbolFrom<mesh::MeshOp>(
-        op.getOperation(), op.getMeshAttr());
-    if (!mesh) {
+    GridOp grid = symbolTableCollection.lookupNearestSymbolFrom<shard::GridOp>(
+        op.getOperation(), op.getGridAttr());
+    if (!grid) {
       return failure();
     }
-    ArrayRef<MeshAxis> opMeshAxes = op.getAxes();
-    SmallVector<MeshAxis> opAxesIota;
-    if (opMeshAxes.empty()) {
-      opAxesIota.resize(mesh.getRank());
+    ArrayRef<GridAxis> opGridAxes = op.getAxes();
+    SmallVector<GridAxis> opAxesIota;
+    if (opGridAxes.empty()) {
+      opAxesIota.resize(grid.getRank());
       std::iota(opAxesIota.begin(), opAxesIota.end(), 0);
-      opMeshAxes = opAxesIota;
+      opGridAxes = opAxesIota;
     }
-    if (llvm::all_of(opMeshAxes, [&mesh](MeshAxis axis) {
-          return ShapedType::isDynamic(mesh.getShape()[axis]);
+    if (llvm::all_of(opGridAxes, [&grid](GridAxis axis) {
+          return ShapedType::isDynamic(grid.getShape()[axis]);
         })) {
-      // All mesh dimensions are dynamic. Nothing to fold.
+      // All grid dimensions are dynamic. Nothing to fold.
       return failure();
     }
 
     SmallVector<Value> newResults(op->getResults().size());
-    SmallVector<MeshAxis> newShapeOpMeshAxes;
+    SmallVector<GridAxis> newShapeOpGridAxes;
     SmallVector<size_t> newToOldResultsIndexMap;
 
-    for (size_t i = 0; i < opMeshAxes.size(); ++i) {
-      auto meshAxisSize = mesh.getShape()[opMeshAxes[i]];
-      if (ShapedType::isDynamic(meshAxisSize)) {
+    for (size_t i = 0; i < opGridAxes.size(); ++i) {
+      auto gridAxisSize = grid.getShape()[opGridAxes[i]];
+      if (ShapedType::isDynamic(gridAxisSize)) {
         newToOldResultsIndexMap.push_back(i);
-        newShapeOpMeshAxes.push_back(opMeshAxes[i]);
+        newShapeOpGridAxes.push_back(opGridAxes[i]);
       } else {
-        // Fold static mesh axes.
+        // Fold static grid axes.
         newResults[i] = arith::ConstantOp::create(
-            builder, builder.getIndexAttr(meshAxisSize));
+            builder, builder.getIndexAttr(gridAxisSize));
       }
     }
 
-    // Leave only the dynamic mesh axes to be queried.
-    if (!newShapeOpMeshAxes.empty()) {
-      MeshShapeOp newShapeOp =
-          MeshShapeOp::create(builder, mesh.getSymName(), newShapeOpMeshAxes);
+    // Leave only the dynamic grid axes to be queried.
+    if (!newShapeOpGridAxes.empty()) {
+      GridShapeOp newShapeOp =
+          GridShapeOp::create(builder, grid.getSymName(), newShapeOpGridAxes);
       for (size_t i = 0; i < newShapeOp->getResults().size(); ++i) {
         newResults[newToOldResultsIndexMap[i]] = newShapeOp->getResults()[i];
       }
@@ -113,8 +113,8 @@ struct MeshShapeFolder
 
 void populateFoldingPatterns(RewritePatternSet &patterns,
                              SymbolTableCollection &symbolTableCollection) {
-  patterns.add<MeshShapeFolder>(symbolTableCollection, patterns.getContext());
+  patterns.add<GridShapeFolder>(symbolTableCollection, patterns.getContext());
 }
 
-} // namespace mesh
+} // namespace shard
 } // namespace mlir
diff --git a/mlir/lib/Dialect/Mesh/Transforms/Transforms.cpp b/mlir/lib/Dialect/Shard/Transforms/Transforms.cpp
index 1bde1af..772e66f 100644
--- a/mlir/lib/Dialect/Mesh/Transforms/Transforms.cpp
+++ b/mlir/lib/Dialect/Shard/Transforms/Transforms.cpp
@@ -6,7 +6,7 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "mlir/Dialect/Mesh/Transforms/Transforms.h"
+#include "mlir/Dialect/Shard/Transforms/Transforms.h"
 #include "TransformsDetail.h"
 #include "mlir/Dialect/Affine/IR/AffineOps.h"
 #include "mlir/Dialect/Affine/Utils.h"
@@ -14,8 +14,8 @@
 #include "mlir/Dialect/Arith/Utils/Utils.h"
 #include "mlir/Dialect/ControlFlow/IR/ControlFlow.h"
 #include "mlir/Dialect/ControlFlow/IR/ControlFlowOps.h"
-#include "mlir/Dialect/Mesh/IR/MeshDialect.h"
-#include "mlir/Dialect/Mesh/IR/MeshOps.h"
+#include "mlir/Dialect/Shard/IR/ShardDialect.h"
+#include "mlir/Dialect/Shard/IR/ShardOps.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/Dialect/Utils/StaticValueUtils.h"
 #include "mlir/IR/BuiltinTypes.h"
@@ -28,12 +28,12 @@
 #include <iterator>
 #include <numeric>
 
-namespace mlir::mesh {
+namespace mlir::shard {
 
 namespace {
 
-/// Lower `mesh.process_multi_index` into expression using
-/// `mesh.process_linear_index` and `mesh.mesh_shape`.
+/// Lower `shard.process_multi_index` into expression using
+/// `shard.process_linear_index` and `shard.grid_shape`.
 struct ProcessMultiIndexOpLowering
     : OpRewritePatternWithSymbolTableCollection<ProcessMultiIndexOp> {
   using OpRewritePatternWithSymbolTableCollection::
@@ -41,30 +41,30 @@ struct ProcessMultiIndexOpLowering
 
   LogicalResult matchAndRewrite(ProcessMultiIndexOp op,
                                 PatternRewriter &rewriter) const override {
-    MeshOp mesh = getMesh(op, symbolTableCollection);
-    if (!mesh) {
+    GridOp grid = getGrid(op, symbolTableCollection);
+    if (!grid) {
       return failure();
     }
 
     ImplicitLocOpBuilder builder(op->getLoc(), rewriter);
     builder.setInsertionPointAfter(op.getOperation());
-    Value linearIndex = ProcessLinearIndexOp::create(builder, mesh);
-    ValueRange meshShape = MeshShapeOp::create(builder, mesh).getResults();
+    Value linearIndex = ProcessLinearIndexOp::create(builder, grid);
+    ValueRange gridShape = GridShapeOp::create(builder, grid).getResults();
     SmallVector<Value> completeMultiIndex =
         affine::AffineDelinearizeIndexOp::create(builder, linearIndex,
-                                                 meshShape)
+                                                 gridShape)
             .getMultiIndex();
     SmallVector<Value> multiIndex;
-    ArrayRef<MeshAxis> opMeshAxes = op.getAxes();
-    SmallVector<MeshAxis> opAxesIota;
-    if (opMeshAxes.empty()) {
-      opAxesIota.resize(mesh.getRank());
+    ArrayRef<GridAxis> opGridAxes = op.getAxes();
+    SmallVector<GridAxis> opAxesIota;
+    if (opGridAxes.empty()) {
+      opAxesIota.resize(grid.getRank());
       std::iota(opAxesIota.begin(), opAxesIota.end(), 0);
-      opMeshAxes = opAxesIota;
+      opGridAxes = opAxesIota;
     }
-    llvm::transform(opMeshAxes, std::back_inserter(multiIndex),
-                    [&completeMultiIndex](MeshAxis meshAxis) {
-                      return completeMultiIndex[meshAxis];
+    llvm::transform(opGridAxes, std::back_inserter(multiIndex),
+                    [&completeMultiIndex](GridAxis gridAxis) {
+                      return completeMultiIndex[gridAxis];
                     });
     rewriter.replaceAllUsesWith(op.getResults(), multiIndex);
     return success();
@@ -86,15 +86,15 @@ struct AllSliceOpLowering
     // axis.
     // The slice axis is split into equisized parts with count
     // the number of processes in the collective process group induced by
-    // the mesh axes.
+    // the grid axes.
     // The part for each process is determined by the corresponding
     // linear-index in the process group.
     //
     // There are no collectives that require communication.
     // Each process operates on its local tensor.
 
-    MeshOp mesh = getMesh(op, symbolTableCollection);
-    if (!mesh) {
+    GridOp grid = getGrid(op, symbolTableCollection);
+    if (!grid) {
       return failure();
     }
 
@@ -104,15 +104,15 @@ struct AllSliceOpLowering
     Value zero = arith::ConstantOp::create(builder, builder.getIndexAttr(0));
 
     Operation::result_range processInGroupMultiIndex =
-        ProcessMultiIndexOp::create(builder, mesh.getSymName(),
-                                    op.getMeshAxes())
+        ProcessMultiIndexOp::create(builder, grid.getSymName(),
+                                    op.getGridAxes())
             .getResults();
 
     Operation::result_range processGroupShape =
-        MeshShapeOp::create(builder, mesh.getSymName(), op.getMeshAxes())
+        GridShapeOp::create(builder, grid.getSymName(), op.getGridAxes())
             .getResult();
     Value processGroupSize =
-        createCollectiveProcessGroupSize(mesh, op.getMeshAxes(), builder);
+        createCollectiveProcessGroupSize(grid, op.getGridAxes(), builder);
 
     int64_t sliceAxis = op.getSliceAxis().getSExtValue();
     Value operandSliceAxisSize =
@@ -125,7 +125,7 @@ struct AllSliceOpLowering
     cf::AssertOp::create(builder, isTargetShapeExactlyDivisible,
                          "Slicing a tensor with axis size that is "
                          "not exactly divisible by the "
-                         "mesh process group size is not supported.");
+                         "grid process group size is not supported.");
     Value resultSliceAxisSize =
         arith::DivUIOp::create(builder, operandSliceAxisSize, processGroupSize);
     OpFoldResult processInGroupLinearIndex = affine::linearizeIndex(
@@ -172,7 +172,7 @@ void populateProcessMultiIndexOpLoweringPatterns(
 }
 
 void registerProcessMultiIndexOpLoweringDialects(DialectRegistry &registry) {
-  registry.insert<affine::AffineDialect, mesh::MeshDialect>();
+  registry.insert<affine::AffineDialect, shard::ShardDialect>();
 }
 
 void populateAllSliceOpLoweringPatterns(
@@ -183,7 +183,7 @@ void populateAllSliceOpLoweringPatterns(
 
 void registerAllSliceOpLoweringDialects(DialectRegistry &registry) {
   registry.insert<affine::AffineDialect, arith::ArithDialect,
-                  cf::ControlFlowDialect, mesh::MeshDialect,
+                  cf::ControlFlowDialect, shard::ShardDialect,
                   tensor::TensorDialect>();
 }
 
@@ -199,21 +199,21 @@ void registerAllOpLoweringDialects(DialectRegistry &registry) {
 }
 
 TypedValue<IndexType>
-createCollectiveProcessGroupSize(MeshOp mesh, ArrayRef<MeshAxis> axes,
+createCollectiveProcessGroupSize(GridOp grid, ArrayRef<GridAxis> axes,
                                  ImplicitLocOpBuilder &builder) {
-  Operation::result_range meshShape =
-      mesh::MeshShapeOp::create(builder, mesh, axes).getResults();
+  Operation::result_range gridShape =
+      GridShapeOp::create(builder, grid, axes).getResults();
   return cast<TypedValue<IndexType>>(arith::createProduct(
-      builder, builder.getLoc(), llvm::to_vector_of<Value>(meshShape),
+      builder, builder.getLoc(), llvm::to_vector_of<Value>(gridShape),
       builder.getIndexType()));
 }
 
 TypedValue<IndexType>
-createProcessLinearIndex(StringRef mesh, ValueRange processInGroupMultiIndex,
-                         ArrayRef<MeshAxis> meshAxes,
+createProcessLinearIndex(StringRef grid, ValueRange processInGroupMultiIndex,
+                         ArrayRef<GridAxis> gridAxes,
                          ImplicitLocOpBuilder &builder) {
   Operation::result_range processGroupShape =
-      MeshShapeOp::create(builder, mesh, meshAxes).getResult();
+      GridShapeOp::create(builder, grid, gridAxes).getResult();
   OpFoldResult processInGroupLinearIndex = affine::linearizeIndex(
       llvm::to_vector_of<OpFoldResult>(processInGroupMultiIndex),
       llvm::to_vector_of<OpFoldResult>(processGroupShape), builder);
@@ -225,11 +225,11 @@ createProcessLinearIndex(StringRef mesh, ValueRange processInGroupMultiIndex,
   return cast<TypedValue<IndexType>>(res);
 }
 
-TypedValue<IndexType> createProcessLinearIndex(StringRef mesh,
-                                               ArrayRef<MeshAxis> meshAxes,
+TypedValue<IndexType> createProcessLinearIndex(StringRef grid,
+                                               ArrayRef<GridAxis> gridAxes,
                                                ImplicitLocOpBuilder &builder) {
   return createProcessLinearIndex(
-      mesh, ProcessMultiIndexOp::create(builder, mesh, meshAxes).getResults(),
-      meshAxes, builder);
+      grid, ProcessMultiIndexOp::create(builder, grid, gridAxes).getResults(),
+      gridAxes, builder);
 }
-} // namespace mlir::mesh
+} // namespace mlir::shard
diff --git a/mlir/lib/Dialect/Mesh/Transforms/TransformsDetail.h b/mlir/lib/Dialect/Shard/Transforms/TransformsDetail.h
index 3e3f584..60c9828 100644
--- a/mlir/lib/Dialect/Mesh/Transforms/TransformsDetail.h
+++ b/mlir/lib/Dialect/Shard/Transforms/TransformsDetail.h
@@ -6,14 +6,14 @@
 //
 //===----------------------------------------------------------------------===//
 
-#ifndef MLIR_DIALECT_MESH_TRANSFORMS_TRANSFORMSDETAIL_H
-#define MLIR_DIALECT_MESH_TRANSFORMS_TRANSFORMSDETAIL_H
+#ifndef MLIR_DIALECT_SHARD_TRANSFORMS_TRANSFORMSDETAIL_H
+#define MLIR_DIALECT_SHARD_TRANSFORMS_TRANSFORMSDETAIL_H
 
 #include "mlir/IR/PatternMatch.h"
 #include "mlir/IR/SymbolTable.h"
 
 namespace mlir {
-namespace mesh {
+namespace shard {
 
 template <typename Op>
 struct OpRewritePatternWithSymbolTableCollection : OpRewritePattern<Op> {
@@ -29,7 +29,7 @@ protected:
   SymbolTableCollection &symbolTableCollection;
 };
 
-} // namespace mesh
+} // namespace shard
 } // namespace mlir
 
-#endif // MLIR_DIALECT_MESH_TRANSFORMS_TRANSFORMSDETAIL_H
+#endif // MLIR_DIALECT_SHARD_TRANSFORMS_TRANSFORMSDETAIL_H
diff --git a/mlir/lib/Dialect/SparseTensor/Transforms/SparseBufferRewriting.cpp b/mlir/lib/Dialect/SparseTensor/Transforms/SparseBufferRewriting.cpp
index a52872d..3b4140e 100644
--- a/mlir/lib/Dialect/SparseTensor/Transforms/SparseBufferRewriting.cpp
+++ b/mlir/lib/Dialect/SparseTensor/Transforms/SparseBufferRewriting.cpp
@@ -931,10 +931,9 @@ createQuickSort(OpBuilder &builder, ModuleOp module, func::FuncOp func,
   FlatSymbolRefAttr partitionFunc = getMangledSortHelperFunc(
       builder, func, {IndexType::get(context)}, kPartitionFuncNamePrefix, xPerm,
       ny, args.drop_back(nTrailingP), createPartitionFunc);
-  Value p = builder
-                .create<func::CallOp>(loc, partitionFunc,
-                                      TypeRange{IndexType::get(context)},
-                                      args.drop_back(nTrailingP))
+  Value p = func::CallOp::create(builder, loc, partitionFunc,
+                                 TypeRange{IndexType::get(context)},
+                                 args.drop_back(nTrailingP))
                 .getResult(0);
 
   Value lenLow = arith::SubIOp::create(builder, loc, p, lo);
@@ -1028,9 +1027,8 @@ static void createSortStableFunc(OpBuilder &builder, ModuleOp module,
   FlatSymbolRefAttr searchFunc = getMangledSortHelperFunc(
       builder, func, {IndexType::get(context)}, kBinarySearchFuncNamePrefix,
       xPerm, ny, operands, createBinarySearchFunc);
-  Value p = builder
-                .create<func::CallOp>(loc, searchFunc, TypeRange{c1.getType()},
-                                      operands)
+  Value p = func::CallOp::create(builder, loc, searchFunc,
+                                 TypeRange{c1.getType()}, operands)
                 .getResult(0);
 
   // Move the value at data[i] to a temporary location.
diff --git a/mlir/lib/Dialect/SparseTensor/Transforms/SparseGPUCodegen.cpp b/mlir/lib/Dialect/SparseTensor/Transforms/SparseGPUCodegen.cpp
index a317abd..0bd1d34 100644
--- a/mlir/lib/Dialect/SparseTensor/Transforms/SparseGPUCodegen.cpp
+++ b/mlir/lib/Dialect/SparseTensor/Transforms/SparseGPUCodegen.cpp
@@ -98,10 +98,10 @@ static Value genLaunchGPUFunc(OpBuilder &builder, gpu::GPUFuncOp gpuFunc,
   Value numT = constantIndex(builder, loc, numThreads);
   gpu::KernelDim3 gridSize = {one, one, one};
   gpu::KernelDim3 blckSize = {numT, one, one};
-  return builder
-      .create<gpu::LaunchFuncOp>(loc, gpuFunc, gridSize, blckSize,
-                                 /*dynSharedMemSz*/ none, args,
-                                 builder.getType<gpu::AsyncTokenType>(), tokens)
+  return gpu::LaunchFuncOp::create(builder, loc, gpuFunc, gridSize, blckSize,
+                                   /*dynSharedMemSz*/ none, args,
+                                   builder.getType<gpu::AsyncTokenType>(),
+                                   tokens)
       .getAsyncToken();
 }
 
@@ -1168,7 +1168,7 @@ struct ForallRewriter : public OpRewritePattern<scf::ParallelOp> {
   using OpRewritePattern<scf::ParallelOp>::OpRewritePattern;
 
   ForallRewriter(MLIRContext *context, unsigned nT)
-      : OpRewritePattern(context), numThreads(nT){};
+      : OpRewritePattern(context), numThreads(nT) {};
 
   LogicalResult matchAndRewrite(scf::ParallelOp forallOp,
                                 PatternRewriter &rewriter) const override {
diff --git a/mlir/lib/Dialect/SparseTensor/Transforms/SparseIterationToScf.cpp b/mlir/lib/Dialect/SparseTensor/Transforms/SparseIterationToScf.cpp
index dfb1274..9cd4896 100644
--- a/mlir/lib/Dialect/SparseTensor/Transforms/SparseIterationToScf.cpp
+++ b/mlir/lib/Dialect/SparseTensor/Transforms/SparseIterationToScf.cpp
@@ -443,8 +443,8 @@ mlir::SparseIterationTypeConverter::SparseIterationTypeConverter() {
 
   addSourceMaterialization([](OpBuilder &builder, IterSpaceType spTp,
                               ValueRange inputs, Location loc) -> Value {
-    return builder
-        .create<UnrealizedConversionCastOp>(loc, TypeRange(spTp), inputs)
+    return UnrealizedConversionCastOp::create(builder, loc, TypeRange(spTp),
+                                              inputs)
         .getResult(0);
   });
 }
diff --git a/mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorCodegen.cpp b/mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorCodegen.cpp
index 70795e2..7a26cd3 100644
--- a/mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorCodegen.cpp
+++ b/mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorCodegen.cpp
@@ -412,13 +412,13 @@ static Value genSliceToSize(OpBuilder &builder, Location loc, Value mem,
   if (memTp.getRank() > 1)
     return mem;
   // Truncate linear memrefs to given size.
-  return builder
-      .create<memref::SubViewOp>(
-          loc, MemRefType::get({ShapedType::kDynamic}, memTp.getElementType()),
-          mem, ValueRange{}, ValueRange{sz}, ValueRange{},
-          ArrayRef<int64_t>{0},                    // static offset
-          ArrayRef<int64_t>{ShapedType::kDynamic}, // dynamic size
-          ArrayRef<int64_t>{1})                    // static stride
+  return memref::SubViewOp::create(
+             builder, loc,
+             MemRefType::get({ShapedType::kDynamic}, memTp.getElementType()),
+             mem, ValueRange{}, ValueRange{sz}, ValueRange{},
+             ArrayRef<int64_t>{0},                    // static offset
+             ArrayRef<int64_t>{ShapedType::kDynamic}, // dynamic size
+             ArrayRef<int64_t>{1})                    // static stride
       .getResult();
 }
 
@@ -449,7 +449,7 @@ class SparseInsertGenerator
 public:
   SparseInsertGenerator(TensorType rtp, TypeRange retTypes, ValueRange params,
                         bool genCall)
-      : FuncCallOrInlineGenerator(retTypes, params, genCall), rtp(rtp){};
+      : FuncCallOrInlineGenerator(retTypes, params, genCall), rtp(rtp) {};
 
   /// Generates code along an insertion path without the need for a "cursor".
   /// This current insertion strategy comes at the expense of some testing
diff --git a/mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorRewriting.cpp b/mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorRewriting.cpp
index b444ac5..79f4e7f 100644
--- a/mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorRewriting.cpp
+++ b/mlir/lib/Dialect/SparseTensor/Transforms/SparseTensorRewriting.cpp
@@ -904,9 +904,8 @@ public:
         dstTp->withoutDimToLvl(),
         !srcTp->isAllOrdered() || !srcTp->isIdentity() || !dstTp->isIdentity());
     SmallVector<Value> dynSizes;
-    Value buffer = rewriter
-                       .create<AllocTensorOp>(loc, bufferTp, dynSizes, Value(),
-                                              nnz, Attribute())
+    Value buffer = AllocTensorOp::create(rewriter, loc, bufferTp, dynSizes,
+                                         Value(), nnz, Attribute())
                        .getResult();
 
     // Convert src coordinates to dst coordinates by first collapsing it to 1D
@@ -1013,9 +1012,8 @@ public:
         !srcTp.isAllOrdered() || !srcTp.isIdentity() || !dstTp.isIdentity());
 
     Value buffer =
-        rewriter
-            .create<AllocTensorOp>(loc, bufferTp, dstDynSizes, Value(),
-                                   /*sizeHint=*/nnz, Attribute())
+        AllocTensorOp::create(rewriter, loc, bufferTp, dstDynSizes, Value(),
+                              /*sizeHint=*/nnz, Attribute())
             .getResult();
 
     // Implement the sparse2sparse reshape as follows:
diff --git a/mlir/lib/Dialect/Tensor/Extensions/AllExtensions.cpp b/mlir/lib/Dialect/Tensor/Extensions/AllExtensions.cpp
index 0421a6c..0784615 100644
--- a/mlir/lib/Dialect/Tensor/Extensions/AllExtensions.cpp
+++ b/mlir/lib/Dialect/Tensor/Extensions/AllExtensions.cpp
@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Dialect/Tensor/Extensions/AllExtensions.h"
-#include "mlir/Dialect/Tensor/Extensions/MeshShardingExtensions.h"
+#include "mlir/Dialect/Tensor/Extensions/ShardingExtensions.h"
 
 using namespace mlir;
 
diff --git a/mlir/lib/Dialect/Tensor/Extensions/CMakeLists.txt b/mlir/lib/Dialect/Tensor/Extensions/CMakeLists.txt
index dba5933..8f0b7da 100644
--- a/mlir/lib/Dialect/Tensor/Extensions/CMakeLists.txt
+++ b/mlir/lib/Dialect/Tensor/Extensions/CMakeLists.txt
@@ -1,10 +1,10 @@
 set(LLVM_OPTIONAL_SOURCES
   AllExtensions.cpp
-  MeshShardingExtensions.cpp
+  ShardingExtensions.cpp
   )
 
-add_mlir_extension_library(MLIRTensorMeshShardingExtensions
-  MeshShardingExtensions.cpp
+add_mlir_extension_library(MLIRTensorShardingExtensions
+  ShardingExtensions.cpp
 
   ADDITIONAL_HEADER_DIRS
   ${MLIR_MAIN_INCLUDE_DIR}/mlir/Dialect/Tensor/Extensions
@@ -22,5 +22,5 @@ add_mlir_extension_library(MLIRTensorAllExtensions
   ${MLIR_MAIN_INCLUDE_DIR}/mlir/Dialect/Tensor/Extensions
 
   LINK_LIBS PUBLIC
-  MLIRTensorMeshShardingExtensions
+  MLIRTensorShardingExtensions
   )
\ No newline at end of file
diff --git a/mlir/lib/Dialect/Tensor/Extensions/MeshShardingExtensions.cpp b/mlir/lib/Dialect/Tensor/Extensions/ShardingExtensions.cpp
index 7e4a5ac..ca7287c 100644
--- a/mlir/lib/Dialect/Tensor/Extensions/MeshShardingExtensions.cpp
+++ b/mlir/lib/Dialect/Tensor/Extensions/ShardingExtensions.cpp
@@ -6,15 +6,15 @@
 //
 //===----------------------------------------------------------------------===//
 
-#include "mlir/Dialect/Mesh/Interfaces/ShardingInterface.h"
-#include "mlir/Dialect/Mesh/Interfaces/ShardingInterfaceImpl.h"
+#include "mlir/Dialect/Shard/Interfaces/ShardingInterface.h"
+#include "mlir/Dialect/Shard/Interfaces/ShardingInterfaceImpl.h"
 #include "mlir/Dialect/Tensor/IR/ShardingInterfaceImpl.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/IR/DialectRegistry.h"
 
 using namespace mlir;
 using namespace mlir::tensor;
-using namespace mlir::mesh;
+using namespace mlir::shard;
 
 namespace {
 
@@ -40,20 +40,20 @@ struct CreatorOpShardingInterface
         {AffineMap::getMultiDimIdentityMap(type.getRank(), ctx)});
   }
 
-  LogicalResult spmdize(Operation *op, ArrayRef<Value> spmdizedOperands,
-                        ArrayRef<MeshSharding> operandShardings,
-                        ArrayRef<MeshSharding> resultShardings,
-                        IRMapping &spmdizationMap,
-                        SymbolTableCollection &symbolTable,
-                        OpBuilder &builder) const {
+  LogicalResult partition(Operation *op, ArrayRef<Value> partitionedOperands,
+                          ArrayRef<Sharding> operandShardings,
+                          ArrayRef<Sharding> resultShardings,
+                          IRMapping &partitionMap,
+                          SymbolTableCollection &symbolTable,
+                          OpBuilder &builder) const {
     assert(resultShardings.size() == 1);
     auto resType = cast<RankedTensorType>(op->getResult(0).getType());
-    mlir::mesh::MeshOp mesh;
+    mlir::shard::GridOp grid;
     ShapedType shardType;
     if (resType.getRank() > 0) {
-      mesh = mesh::getMesh(op, resultShardings[0].getMeshAttr(), symbolTable);
+      grid = shard::getGrid(op, resultShardings[0].getGridAttr(), symbolTable);
       shardType =
-          cast<ShapedType>(mesh::shardType(resType, mesh, resultShardings[0]));
+          cast<ShapedType>(shard::shardType(resType, grid, resultShardings[0]));
     } else {
       shardType = resType;
     }
@@ -67,7 +67,7 @@ struct CreatorOpShardingInterface
       auto oldType = cast<ShapedType>(resType);
       assert(oldType.getRank() == shardType.getRank());
       int currOldOprndNum = -1;
-      mesh::ShardShapeOp shapeForDevice;
+      shard::ShardShapeOp shapeForDevice;
       ValueRange device;
       Operation *newSharding = nullptr;
       for (auto i = 0; i < oldType.getRank(); ++i) {
@@ -76,23 +76,23 @@ struct CreatorOpShardingInterface
             newSharding =
                 ShardingOp::create(builder, op->getLoc(), resultShardings[0]);
             device =
-                mesh::ProcessMultiIndexOp::create(builder, op->getLoc(), mesh)
+                shard::ProcessMultiIndexOp::create(builder, op->getLoc(), grid)
                     .getResults();
-            shapeForDevice = mesh::ShardShapeOp::create(
-                builder, op->getLoc(), oldType.getShape(), spmdizedOperands,
+            shapeForDevice = shard::ShardShapeOp::create(
+                builder, op->getLoc(), oldType.getShape(), partitionedOperands,
                 newSharding->getResult(0), device);
           }
           newOperands.emplace_back(shapeForDevice.getResult()[i]);
         } else if (oldType.isDynamicDim(i)) {
           assert(shardType.isDynamicDim(i));
-          newOperands.emplace_back(spmdizedOperands[++currOldOprndNum]);
+          newOperands.emplace_back(partitionedOperands[++currOldOprndNum]);
         }
       }
       newOp = OpTy::create(builder, op->getLoc(), shardType, newOperands);
-      spmdizationMap.map(op->getResult(0), newOp->getResult(0));
+      partitionMap.map(op->getResult(0), newOp->getResult(0));
     } else {
       // `clone` will populate the mapping of old to new results.
-      newOp = builder.clone(*op, spmdizationMap);
+      newOp = builder.clone(*op, partitionMap);
     }
     newOp->getResult(0).setType(shardType);
 
diff --git a/mlir/lib/Dialect/Tensor/Transforms/BufferizableOpInterfaceImpl.cpp b/mlir/lib/Dialect/Tensor/Transforms/BufferizableOpInterfaceImpl.cpp
index bc11e56..c3356c1 100644
--- a/mlir/lib/Dialect/Tensor/Transforms/BufferizableOpInterfaceImpl.cpp
+++ b/mlir/lib/Dialect/Tensor/Transforms/BufferizableOpInterfaceImpl.cpp
@@ -784,8 +784,8 @@ struct PadOpInterface
     auto toValue = [&](OpFoldResult ofr) {
       if (auto value = dyn_cast<Value>(ofr))
         return value;
-      return rewriter
-          .create<arith::ConstantIndexOp>(loc, *getConstantIntValue(ofr))
+      return arith::ConstantIndexOp::create(rewriter, loc,
+                                            *getConstantIntValue(ofr))
           .getResult();
     };
 
@@ -919,9 +919,8 @@ struct ReshapeOpInterface
       auto memrefType = MemRefType::get(
           srcType.getShape(), srcType.getElementType(), AffineMap(),
           cast<BaseMemRefType>(srcBuffer->getType()).getMemorySpace());
-      srcBuffer = rewriter
-                      .create<bufferization::ToBufferOp>(
-                          op->getLoc(), memrefType, *tensorAlloc)
+      srcBuffer = bufferization::ToBufferOp::create(rewriter, op->getLoc(),
+                                                    memrefType, *tensorAlloc)
                       .getResult();
     }
 
diff --git a/mlir/lib/Dialect/Tensor/Transforms/IndependenceTransforms.cpp b/mlir/lib/Dialect/Tensor/Transforms/IndependenceTransforms.cpp
index 43d9d70..9fd27d3 100644
--- a/mlir/lib/Dialect/Tensor/Transforms/IndependenceTransforms.cpp
+++ b/mlir/lib/Dialect/Tensor/Transforms/IndependenceTransforms.cpp
@@ -130,8 +130,7 @@ FailureOr<Value> tensor::buildIndependentOp(OpBuilder &b,
   // Create a tensor::ExtractSliceOp.
   SmallVector<OpFoldResult> offsets(newSizes.size(), b.getIndexAttr(0));
   SmallVector<OpFoldResult> strides(newSizes.size(), b.getIndexAttr(1));
-  return b
-      .create<ExtractSliceOp>(loc, newEmptyOp, offsets, emptyOp.getMixedSizes(),
-                              strides)
+  return ExtractSliceOp::create(b, loc, newEmptyOp, offsets,
+                                emptyOp.getMixedSizes(), strides)
       .getResult();
 }
diff --git a/mlir/lib/Dialect/Tensor/Transforms/ReshapePatterns.cpp b/mlir/lib/Dialect/Tensor/Transforms/ReshapePatterns.cpp
index e0af2f7..2ec23e1 100644
--- a/mlir/lib/Dialect/Tensor/Transforms/ReshapePatterns.cpp
+++ b/mlir/lib/Dialect/Tensor/Transforms/ReshapePatterns.cpp
@@ -385,10 +385,9 @@ struct BubbleUpExpandShapeThroughExtractSlice
             return getValueOrCreateConstantIndexOp(rewriter, loc, ofr);
           });
       OpFoldResult collapsedOffset =
-          rewriter
-              .create<affine::AffineLinearizeIndexOp>(loc, offsetVals,
-                                                      reassocGroupSizes,
-                                                      /*disjoint=*/true)
+          affine::AffineLinearizeIndexOp::create(rewriter, loc, offsetVals,
+                                                 reassocGroupSizes,
+                                                 /*disjoint=*/true)
               .getResult();
       collapsedOffsets.push_back(collapsedOffset);
       collapsedSizes.push_back(collapsedSize);
diff --git a/mlir/lib/Dialect/Tosa/CMakeLists.txt b/mlir/lib/Dialect/Tosa/CMakeLists.txt
index b1fac8c..c6a438d 100644
--- a/mlir/lib/Dialect/Tosa/CMakeLists.txt
+++ b/mlir/lib/Dialect/Tosa/CMakeLists.txt
@@ -36,7 +36,7 @@ add_mlir_dialect_library(MLIRTosaShardingInterfaceImpl
 
   LINK_LIBS PUBLIC
   MLIRIR
-  MLIRMeshDialect
+  MLIRShardDialect
   MLIRShardingInterface
   MLIRSupport
   MLIRTosaDialect
diff --git a/mlir/lib/Dialect/Tosa/IR/ShardingInterfaceImpl.cpp b/mlir/lib/Dialect/Tosa/IR/ShardingInterfaceImpl.cpp
index d3a5f44..45994a7 100644
--- a/mlir/lib/Dialect/Tosa/IR/ShardingInterfaceImpl.cpp
+++ b/mlir/lib/Dialect/Tosa/IR/ShardingInterfaceImpl.cpp
@@ -7,9 +7,9 @@
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Dialect/Tosa/IR/ShardingInterfaceImpl.h"
-#include "mlir/Dialect/Mesh/IR/MeshOps.h"
-#include "mlir/Dialect/Mesh/Interfaces/ShardingInterface.h"
-#include "mlir/Dialect/Mesh/Interfaces/ShardingInterfaceImpl.h"
+#include "mlir/Dialect/Shard/IR/ShardOps.h"
+#include "mlir/Dialect/Shard/Interfaces/ShardingInterface.h"
+#include "mlir/Dialect/Shard/Interfaces/ShardingInterfaceImpl.h"
 #include "mlir/Dialect/Tosa/IR/TosaOps.h"
 #include "mlir/IR/AffineMap.h"
 #include "mlir/IR/DialectRegistry.h"
@@ -19,7 +19,7 @@
 
 using namespace mlir;
 using namespace mlir::tosa;
-using namespace mlir::mesh;
+using namespace mlir::shard;
 
 namespace {
 
@@ -87,15 +87,15 @@ struct NegateOpSharding
     return maps;
   }
 
-  LogicalResult spmdize(Operation *op, ArrayRef<Value> spmdizedOperands,
-                        ArrayRef<MeshSharding> operandShardings,
-                        ArrayRef<MeshSharding> resultShardings,
-                        IRMapping &spmdizationMap,
-                        SymbolTableCollection &symbolTable,
-                        OpBuilder &builder) const {
-    spmdizeTriviallyShardableOperation(*op, spmdizedOperands, operandShardings,
-                                       resultShardings, spmdizationMap,
-                                       symbolTable, builder);
+  LogicalResult partition(Operation *op, ArrayRef<Value> partitiondOperands,
+                          ArrayRef<Sharding> operandShardings,
+                          ArrayRef<Sharding> resultShardings,
+                          IRMapping &partitionMap,
+                          SymbolTableCollection &symbolTable,
+                          OpBuilder &builder) const {
+    partitionTriviallyShardableOperation(*op, partitiondOperands,
+                                         operandShardings, resultShardings,
+                                         partitionMap, symbolTable, builder);
     return success();
   }
 };
diff --git a/mlir/lib/Dialect/Tosa/IR/TosaCanonicalizations.cpp b/mlir/lib/Dialect/Tosa/IR/TosaCanonicalizations.cpp
index 1ad2c80..6d2cbb5 100644
--- a/mlir/lib/Dialect/Tosa/IR/TosaCanonicalizations.cpp
+++ b/mlir/lib/Dialect/Tosa/IR/TosaCanonicalizations.cpp
@@ -707,9 +707,8 @@ struct ConcatSliceOptimization : public OpRewritePattern<tosa::SliceOp> {
         auto size_op =
             getTosaConstShape(rewriter, sliceOp.getLoc(), sliceSizes);
         replaceWithSlice =
-            rewriter
-                .create<tosa::SliceOp>(sliceOp.getLoc(), sliceOp.getType(),
-                                       input, start_op, size_op)
+            tosa::SliceOp::create(rewriter, sliceOp.getLoc(), sliceOp.getType(),
+                                  input, start_op, size_op)
                 .getResult();
         break;
       }
diff --git a/mlir/lib/Dialect/Tosa/IR/TosaOps.cpp b/mlir/lib/Dialect/Tosa/IR/TosaOps.cpp
index 648e508a9..3cafb19 100644
--- a/mlir/lib/Dialect/Tosa/IR/TosaOps.cpp
+++ b/mlir/lib/Dialect/Tosa/IR/TosaOps.cpp
@@ -13,8 +13,8 @@
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Dialect/Tosa/IR/TosaOps.h"
-#include "mlir/Dialect/Mesh/Interfaces/ShardingInterface.h"
 #include "mlir/Dialect/Quant/IR/Quant.h"
+#include "mlir/Dialect/Shard/Interfaces/ShardingInterface.h"
 #include "mlir/Dialect/Tensor/IR/Tensor.h"
 #include "mlir/Dialect/Tosa/Utils/QuantUtils.h"
 #include "mlir/Dialect/Tosa/Utils/ShapeUtils.h"
@@ -166,7 +166,7 @@ void TosaDialect::initialize() {
       >();
   addInterfaces<TosaDialectBytecodeInterface, TosaInlinerInterface>();
   declarePromisedInterfaces<
-      mesh::ShardingInterface, ClampOp, SigmoidOp, TanhOp, AddOp,
+      shard::ShardingInterface, ClampOp, SigmoidOp, TanhOp, AddOp,
       ArithmeticRightShiftOp, BitwiseAndOp, BitwiseOrOp, BitwiseXorOp, IntDivOp,
       LogicalAndOp, LogicalLeftShiftOp, LogicalRightShiftOp, LogicalOrOp,
       LogicalXorOp, MaximumOp, MinimumOp, MulOp, PowOp, SubOp, AbsOp,
@@ -3647,6 +3647,22 @@ std::optional<SmallVector<int64_t, 4>> ApplyScaleOp::getShapeForUnroll() {
   return std::nullopt;
 }
 
+static void printInitializationList(OpAsmPrinter &parser,
+                                    Block::BlockArgListType blocksArgs,
+                                    ValueRange initializers,
+                                    StringRef prefix = "") {
+  assert(blocksArgs.size() == initializers.size() &&
+         "expected same length of arguments and initializers");
+  if (initializers.empty())
+    return;
+
+  parser << prefix << '(';
+  llvm::interleaveComma(
+      llvm::zip(blocksArgs, initializers), parser,
+      [&](auto it) { parser << std::get<0>(it) << " = " << std::get<1>(it); });
+  parser << ")";
+}
+
 // parse and print of IfOp refer to the implementation of SCF dialect.
 ParseResult IfOp::parse(OpAsmParser &parser, OperationState &result) {
   // Create the regions for 'then'.
@@ -3654,16 +3670,64 @@ ParseResult IfOp::parse(OpAsmParser &parser, OperationState &result) {
   Region *thenRegion = result.addRegion();
   Region *elseRegion = result.addRegion();
 
-  auto &builder = parser.getBuilder();
   OpAsmParser::UnresolvedOperand cond;
-  // Create a i1 tensor type for the boolean condition.
-  Type i1Type = RankedTensorType::get({}, builder.getIntegerType(1));
-  if (parser.parseOperand(cond) ||
-      parser.resolveOperand(cond, i1Type, result.operands))
+
+  if (parser.parseOperand(cond))
     return failure();
-  // Parse optional results type list.
-  if (parser.parseOptionalArrowTypeList(result.types))
+
+  SmallVector<OpAsmParser::Argument, 4> regionArgs;
+  SmallVector<OpAsmParser::UnresolvedOperand, 4> operands;
+
+  // Parse the optional block arguments
+  OptionalParseResult listResult =
+      parser.parseOptionalAssignmentList(regionArgs, operands);
+  if (listResult.has_value() && failed(listResult.value()))
     return failure();
+
+  // Parse a colon.
+  if (failed(parser.parseColon()))
+    return parser.emitError(parser.getCurrentLocation(),
+                            "expected type for condition operand");
+
+  // Parse the type of the condition operand
+  Type condType;
+  if (failed(parser.parseType(condType)))
+    return parser.emitError(parser.getCurrentLocation(),
+                            "expected type for condition operand");
+
+  // Resolve operand with provided type
+  if (failed(parser.resolveOperand(cond, condType, result.operands)))
+    return failure();
+
+  // Parse optional block arg types
+  if (listResult.has_value()) {
+    FunctionType functionType;
+
+    if (failed(parser.parseType(functionType)))
+      return parser.emitError(parser.getCurrentLocation())
+             << "expected list of types for block arguments "
+             << "followed by arrow type and list of return types";
+
+    result.addTypes(functionType.getResults());
+
+    if (functionType.getNumInputs() != operands.size()) {
+      return parser.emitError(parser.getCurrentLocation())
+             << "expected as many input types as operands "
+             << "(expected " << operands.size() << " got "
+             << functionType.getNumInputs() << ")";
+    }
+
+    // Resolve input operands.
+    if (failed(parser.resolveOperands(operands, functionType.getInputs(),
+                                      parser.getCurrentLocation(),
+                                      result.operands)))
+      return failure();
+  } else {
+    // Parse optional results type list.
+    if (parser.parseOptionalArrowTypeList(result.types))
+      return failure();
+  }
+
   // Parse the 'then' region.
   if (parser.parseRegion(*thenRegion, /*arguments=*/{}, /*argTypes=*/{}))
     return failure();
@@ -3681,26 +3745,28 @@ ParseResult IfOp::parse(OpAsmParser &parser, OperationState &result) {
 }
 
 void IfOp::print(OpAsmPrinter &p) {
-  bool printBlockTerminators = false;
-
   p << " " << getCondition();
-  if (!getResults().empty()) {
-    p << " -> (" << getResultTypes() << ")";
-    // Print yield explicitly if the op defines values.
-    printBlockTerminators = true;
+
+  printInitializationList(p, getThenGraph().front().getArguments(),
+                          getInputList(), " ");
+  p << " : ";
+  p << getCondition().getType();
+
+  if (!getInputList().empty()) {
+    p << " (";
+    llvm::interleaveComma(getInputList().getTypes(), p);
+    p << ")";
   }
-  p << ' ';
-  p.printRegion(getThenGraph(),
-                /*printEntryBlockArgs=*/false,
-                /*printBlockTerminators=*/printBlockTerminators);
+  p.printArrowTypeList(getResultTypes());
+  p << " ";
+
+  p.printRegion(getThenGraph());
 
   // Print the 'else' regions if it exists and has a block.
   auto &elseRegion = getElseGraph();
   if (!elseRegion.empty()) {
     p << " else ";
-    p.printRegion(elseRegion,
-                  /*printEntryBlockArgs=*/false,
-                  /*printBlockTerminators=*/printBlockTerminators);
+    p.printRegion(elseRegion);
   }
 
   p.printOptionalAttrDict((*this)->getAttrs());
@@ -3909,22 +3975,6 @@ ParseResult WhileOp::parse(OpAsmParser &parser, OperationState &result) {
                  parser.parseOptionalAttrDictWithKeyword(result.attributes));
 }
 
-static void printInitializationList(OpAsmPrinter &parser,
-                                    Block::BlockArgListType blocksArgs,
-                                    ValueRange initializers,
-                                    StringRef prefix = "") {
-  assert(blocksArgs.size() == initializers.size() &&
-         "expected same length of arguments and initializers");
-  if (initializers.empty())
-    return;
-
-  parser << prefix << '(';
-  llvm::interleaveComma(
-      llvm::zip(blocksArgs, initializers), parser,
-      [&](auto it) { parser << std::get<0>(it) << " = " << std::get<1>(it); });
-  parser << ")";
-}
-
 void WhileOp::print(OpAsmPrinter &parser) {
   printInitializationList(parser, getCondGraph().front().getArguments(),
                           getInputList(), " ");
diff --git a/mlir/lib/Dialect/Tosa/Transforms/TosaDecomposeDepthwise.cpp b/mlir/lib/Dialect/Tosa/Transforms/TosaDecomposeDepthwise.cpp
index 9474299..0bec0da 100644
--- a/mlir/lib/Dialect/Tosa/Transforms/TosaDecomposeDepthwise.cpp
+++ b/mlir/lib/Dialect/Tosa/Transforms/TosaDecomposeDepthwise.cpp
@@ -81,9 +81,8 @@ struct DepthwiseConv2DIsMul : public OpRewritePattern<tosa::DepthwiseConv2DOp> {
         dyn_cast<RankedTensorType>(input.getType()).getElementType());
     auto revisedInputShapeValue =
         getTosaConstShape(rewriter, op.getLoc(), revisedInputShape);
-    input = rewriter
-                .create<tosa::ReshapeOp>(op.getLoc(), inputType, input,
-                                         revisedInputShapeValue)
+    input = tosa::ReshapeOp::create(rewriter, op.getLoc(), inputType, input,
+                                    revisedInputShapeValue)
                 .getResult();
 
     Type resultETy = resultType.getElementType();
@@ -162,9 +161,8 @@ struct DepthwiseConv2DIsMul : public OpRewritePattern<tosa::DepthwiseConv2DOp> {
         shiftType, rewriter.getIntegerAttr(shiftElementType, 0));
     Value constZero =
         tosa::ConstOp::create(rewriter, op.getLoc(), shiftType, shiftZeroAttr);
-    Value mulValue = rewriter
-                         .create<tosa::MulOp>(op.getLoc(), mulShapeType, input,
-                                              weight, constZero)
+    Value mulValue = tosa::MulOp::create(rewriter, op.getLoc(), mulShapeType,
+                                         input, weight, constZero)
                          .getResult();
 
     // Reshape output to [N, H, W, C * M].
diff --git a/mlir/lib/Dialect/Tosa/Transforms/TosaValidation.cpp b/mlir/lib/Dialect/Tosa/Transforms/TosaValidation.cpp
index 32b5fb6..8ec7765 100644
--- a/mlir/lib/Dialect/Tosa/Transforms/TosaValidation.cpp
+++ b/mlir/lib/Dialect/Tosa/Transforms/TosaValidation.cpp
@@ -1248,16 +1248,14 @@ bool checkErrorIfCondIf(Operation *op) {
   // })
   //
   // Simplified:
-  // %0 = tosa.cond_if %arg2 {
-  //   tosa.yield %arg0
+  // %0 = tosa.cond_if %arg2 (%arg3 = %arg0, %arg4 = %arg1) {
+  //   ^bb0(%arg3, %arg4):
+  //   tosa.yield %arg3
   // } else {
-  //   tosa.yield %arg1
+  //   ^bb0(%arg3, %arg4):
+  //   tosa.yield %arg4
   // }
-  //
-  // Unfortunately, the simplified syntax does not encapsulate values
-  // used in then/else regions (see 'simplified' example above), so it
-  // must be rewritten to use the generic syntax in order to be conformant
-  // to the specification.
+
   return failed(checkIsolatedRegion(op, ifOp.getThenGraph(), "then")) ||
          failed(checkIsolatedRegion(op, ifOp.getElseGraph(), "else"));
 }
diff --git a/mlir/lib/Dialect/Transform/Interfaces/TransformInterfaces.cpp b/mlir/lib/Dialect/Transform/Interfaces/TransformInterfaces.cpp
index 4662836..e297f7c 100644
--- a/mlir/lib/Dialect/Transform/Interfaces/TransformInterfaces.cpp
+++ b/mlir/lib/Dialect/Transform/Interfaces/TransformInterfaces.cpp
@@ -16,15 +16,21 @@
 #include "llvm/ADT/ScopeExit.h"
 #include "llvm/ADT/iterator.h"
 #include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 #include "llvm/Support/ErrorHandling.h"
 #include "llvm/Support/InterleavedRange.h"
 
 #define DEBUG_TYPE "transform-dialect"
 #define DEBUG_TYPE_FULL "transform-dialect-full"
 #define DEBUG_PRINT_AFTER_ALL "transform-dialect-print-top-level-after-all"
-#define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE "] ")
-#define LDBG(X) LLVM_DEBUG(DBGS() << (X))
-#define FULL_LDBG(X) DEBUG_WITH_TYPE(DEBUG_TYPE_FULL, (DBGS() << (X)))
+#ifndef NDEBUG
+#define FULL_LDBG(X)                                                           \
+  DEBUGLOG_WITH_STREAM_AND_TYPE(llvm::dbgs(), DEBUG_TYPE_FULL)
+#else
+#define FULL_LDBG(X)                                                           \
+  for (bool _c = false; _c; _c = false)                                        \
+  ::llvm::nulls()
+#endif
 
 using namespace mlir;
 
@@ -486,24 +492,20 @@ void transform::TransformState::recordOpHandleInvalidationOne(
       newlyInvalidated.count(otherHandle))
     return;
 
-  FULL_LDBG("--recordOpHandleInvalidationOne\n");
-  DEBUG_WITH_TYPE(DEBUG_TYPE_FULL, {
-    (DBGS() << "--ancestors: "
-            << llvm::interleaved(llvm::make_pointee_range(potentialAncestors))
-            << "\n");
-  });
+  FULL_LDBG() << "--recordOpHandleInvalidationOne";
+  FULL_LDBG() << "--ancestors: "
+              << llvm::interleaved(
+                     llvm::make_pointee_range(potentialAncestors));
 
   Operation *owner = consumingHandle.getOwner();
   unsigned operandNo = consumingHandle.getOperandNumber();
   for (Operation *ancestor : potentialAncestors) {
     // clang-format off
-    DEBUG_WITH_TYPE(DEBUG_TYPE_FULL,
-      { (DBGS() << "----handle one ancestor: " << *ancestor << "\n"); });
-    DEBUG_WITH_TYPE(DEBUG_TYPE_FULL,
-      { (DBGS() << "----of payload with name: "
-                << payloadOp->getName().getIdentifier() << "\n"); });
-    DEBUG_WITH_TYPE(DEBUG_TYPE_FULL,
-      { (DBGS() << "----of payload: " << *payloadOp << "\n"); });
+          FULL_LDBG() << "----handle one ancestor: " << *ancestor;;
+
+      FULL_LDBG() << "----of payload with name: "
+                << payloadOp->getName().getIdentifier();
+      FULL_LDBG() << "----of payload: " << *payloadOp;
     // clang-format on
     if (!ancestor->isAncestor(payloadOp))
       continue;
@@ -609,10 +611,8 @@ void transform::TransformState::recordOpHandleInvalidation(
     transform::TransformState::InvalidatedHandleMap &newlyInvalidated) const {
 
   if (potentialAncestors.empty()) {
-    DEBUG_WITH_TYPE(DEBUG_TYPE_FULL, {
-      (DBGS() << "----recording invalidation for empty handle: " << handle.get()
-              << "\n");
-    });
+    FULL_LDBG() << "----recording invalidation for empty handle: "
+                << handle.get();
 
     Operation *owner = handle.getOwner();
     unsigned operandNo = handle.getOperandNumber();
@@ -709,7 +709,7 @@ void transform::TransformState::recordValueHandleInvalidation(
 LogicalResult transform::TransformState::checkAndRecordHandleInvalidationImpl(
     transform::TransformOpInterface transform,
     transform::TransformState::InvalidatedHandleMap &newlyInvalidated) const {
-  FULL_LDBG("--Start checkAndRecordHandleInvalidation\n");
+  FULL_LDBG() << "--Start checkAndRecordHandleInvalidation";
   auto memoryEffectsIface =
       cast<MemoryEffectOpInterface>(transform.getOperation());
   SmallVector<MemoryEffects::EffectInstance> effects;
@@ -717,9 +717,7 @@ LogicalResult transform::TransformState::checkAndRecordHandleInvalidationImpl(
       transform::TransformMappingResource::get(), effects);
 
   for (OpOperand &target : transform->getOpOperands()) {
-    DEBUG_WITH_TYPE(DEBUG_TYPE_FULL, {
-      (DBGS() << "----iterate on handle: " << target.get() << "\n");
-    });
+    FULL_LDBG() << "----iterate on handle: " << target.get();
     // If the operand uses an invalidated handle, report it. If the operation
     // allows handles to point to repeated payload operations, only report
     // pre-existing invalidation errors. Otherwise, also report invalidations
@@ -727,14 +725,14 @@ LogicalResult transform::TransformState::checkAndRecordHandleInvalidationImpl(
     auto it = invalidatedHandles.find(target.get());
     auto nit = newlyInvalidated.find(target.get());
     if (it != invalidatedHandles.end()) {
-      FULL_LDBG("--End checkAndRecordHandleInvalidation, found already "
-                "invalidated -> FAILURE\n");
+      FULL_LDBG() << "--End checkAndRecordHandleInvalidation, found already "
+                     "invalidated -> FAILURE";
       return it->getSecond()(transform->getLoc()), failure();
     }
     if (!transform.allowsRepeatedHandleOperands() &&
         nit != newlyInvalidated.end()) {
-      FULL_LDBG("--End checkAndRecordHandleInvalidation, found newly "
-                "invalidated (by this op) -> FAILURE\n");
+      FULL_LDBG() << "--End checkAndRecordHandleInvalidation, found newly "
+                     "invalidated (by this op) -> FAILURE";
       return nit->getSecond()(transform->getLoc()), failure();
     }
 
@@ -745,27 +743,28 @@ LogicalResult transform::TransformState::checkAndRecordHandleInvalidationImpl(
              effect.getValue() == target.get();
     };
     if (llvm::any_of(effects, consumesTarget)) {
-      FULL_LDBG("----found consume effect\n");
+      FULL_LDBG() << "----found consume effect";
       if (llvm::isa<transform::TransformHandleTypeInterface>(
               target.get().getType())) {
-        FULL_LDBG("----recordOpHandleInvalidation\n");
+        FULL_LDBG() << "----recordOpHandleInvalidation";
         SmallVector<Operation *> payloadOps =
             llvm::to_vector(getPayloadOps(target.get()));
         recordOpHandleInvalidation(target, payloadOps, nullptr,
                                    newlyInvalidated);
       } else if (llvm::isa<transform::TransformValueHandleTypeInterface>(
                      target.get().getType())) {
-        FULL_LDBG("----recordValueHandleInvalidation\n");
+        FULL_LDBG() << "----recordValueHandleInvalidation";
         recordValueHandleInvalidation(target, newlyInvalidated);
       } else {
-        FULL_LDBG("----not a TransformHandle -> SKIP AND DROP ON THE FLOOR\n");
+        FULL_LDBG()
+            << "----not a TransformHandle -> SKIP AND DROP ON THE FLOOR";
       }
     } else {
-      FULL_LDBG("----no consume effect -> SKIP\n");
+      FULL_LDBG() << "----no consume effect -> SKIP";
     }
   }
 
-  FULL_LDBG("--End checkAndRecordHandleInvalidation -> SUCCESS\n");
+  FULL_LDBG() << "--End checkAndRecordHandleInvalidation -> SUCCESS";
   return success();
 }
 
@@ -818,18 +817,14 @@ void transform::TransformState::compactOpHandles() {
 
 DiagnosedSilenceableFailure
 transform::TransformState::applyTransform(TransformOpInterface transform) {
-  LLVM_DEBUG({
-    DBGS() << "applying: ";
-    transform->print(llvm::dbgs(), OpPrintingFlags().skipRegions());
-    llvm::dbgs() << "\n";
-  });
-  DEBUG_WITH_TYPE(DEBUG_TYPE_FULL,
-                  DBGS() << "Top-level payload before application:\n"
-                         << *getTopLevel() << "\n");
+  LDBG() << "applying: "
+         << OpWithFlags(transform, OpPrintingFlags().skipRegions());
+  FULL_LDBG() << "Top-level payload before application:\n" << *getTopLevel();
   auto printOnFailureRAII = llvm::make_scope_exit([this] {
     (void)this;
-    LLVM_DEBUG(DBGS() << "Failing Top-level payload:\n"; getTopLevel()->print(
-        llvm::dbgs(), mlir::OpPrintingFlags().printGenericOpForm()););
+    LDBG() << "Failing Top-level payload:\n"
+           << OpWithFlags(getTopLevel(),
+                          OpPrintingFlags().printGenericOpForm());
   });
 
   // Set current transform op.
@@ -837,47 +832,45 @@ transform::TransformState::applyTransform(TransformOpInterface transform) {
 
   // Expensive checks to detect invalid transform IR.
   if (options.getExpensiveChecksEnabled()) {
-    FULL_LDBG("ExpensiveChecksEnabled\n");
+    FULL_LDBG() << "ExpensiveChecksEnabled";
     if (failed(checkAndRecordHandleInvalidation(transform)))
       return DiagnosedSilenceableFailure::definiteFailure();
 
     for (OpOperand &operand : transform->getOpOperands()) {
-      DEBUG_WITH_TYPE(DEBUG_TYPE_FULL, {
-        (DBGS() << "iterate on handle: " << operand.get() << "\n");
-      });
+      FULL_LDBG() << "iterate on handle: " << operand.get();
       if (!isHandleConsumed(operand.get(), transform)) {
-        FULL_LDBG("--handle not consumed -> SKIP\n");
+        FULL_LDBG() << "--handle not consumed -> SKIP";
         continue;
       }
       if (transform.allowsRepeatedHandleOperands()) {
-        FULL_LDBG("--op allows repeated handles -> SKIP\n");
+        FULL_LDBG() << "--op allows repeated handles -> SKIP";
         continue;
       }
-      FULL_LDBG("--handle is consumed\n");
+      FULL_LDBG() << "--handle is consumed";
 
       Type operandType = operand.get().getType();
       if (llvm::isa<TransformHandleTypeInterface>(operandType)) {
-        FULL_LDBG("--checkRepeatedConsumptionInOperand for Operation*\n");
+        FULL_LDBG() << "--checkRepeatedConsumptionInOperand for Operation*";
         DiagnosedSilenceableFailure check =
             checkRepeatedConsumptionInOperand<Operation *>(
                 getPayloadOpsView(operand.get()), transform,
                 operand.getOperandNumber());
         if (!check.succeeded()) {
-          FULL_LDBG("----FAILED\n");
+          FULL_LDBG() << "----FAILED";
           return check;
         }
       } else if (llvm::isa<TransformValueHandleTypeInterface>(operandType)) {
-        FULL_LDBG("--checkRepeatedConsumptionInOperand For Value\n");
+        FULL_LDBG() << "--checkRepeatedConsumptionInOperand For Value";
         DiagnosedSilenceableFailure check =
             checkRepeatedConsumptionInOperand<Value>(
                 getPayloadValuesView(operand.get()), transform,
                 operand.getOperandNumber());
         if (!check.succeeded()) {
-          FULL_LDBG("----FAILED\n");
+          FULL_LDBG() << "----FAILED";
           return check;
         }
       } else {
-        FULL_LDBG("--not a TransformHandle -> SKIP AND DROP ON THE FLOOR\n");
+        FULL_LDBG() << "--not a TransformHandle -> SKIP AND DROP ON THE FLOOR";
       }
     }
   }
@@ -999,8 +992,7 @@ transform::TransformState::applyTransform(TransformOpInterface transform) {
 
   printOnFailureRAII.release();
   DEBUG_WITH_TYPE(DEBUG_PRINT_AFTER_ALL, {
-    DBGS() << "Top-level payload:\n";
-    getTopLevel()->print(llvm::dbgs());
+    LDBG() << "Top-level payload:\n" << *getTopLevel();
   });
   return result;
 }
@@ -1277,7 +1269,7 @@ void transform::TrackingListener::notifyMatchFailure(
   LLVM_DEBUG({
     Diagnostic diag(loc, DiagnosticSeverity::Remark);
     reasonCallback(diag);
-    DBGS() << "Match Failure : " << diag.str() << "\n";
+    LDBG() << "Match Failure : " << diag.str();
   });
 }
 
diff --git a/mlir/lib/Dialect/Vector/IR/VectorOps.cpp b/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
index 4e9f93b..bce358d 100644
--- a/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
+++ b/mlir/lib/Dialect/Vector/IR/VectorOps.cpp
@@ -372,9 +372,8 @@ SmallVector<Value> vector::getAsValues(OpBuilder &builder, Location loc,
   llvm::transform(foldResults, std::back_inserter(values),
                   [&](OpFoldResult foldResult) {
                     if (auto attr = dyn_cast<Attribute>(foldResult))
-                      return builder
-                          .create<arith::ConstantIndexOp>(
-                              loc, cast<IntegerAttr>(attr).getInt())
+                      return arith::ConstantIndexOp::create(
+                                 builder, loc, cast<IntegerAttr>(attr).getInt())
                           .getResult();
 
                     return cast<Value>(foldResult);
diff --git a/mlir/lib/Dialect/Vector/Transforms/LowerVectorGather.cpp b/mlir/lib/Dialect/Vector/Transforms/LowerVectorGather.cpp
index 2484670..e062f55 100644
--- a/mlir/lib/Dialect/Vector/Transforms/LowerVectorGather.cpp
+++ b/mlir/lib/Dialect/Vector/Transforms/LowerVectorGather.cpp
@@ -248,11 +248,10 @@ struct Gather1DToConditionalLoads : OpRewritePattern<vector::GatherOp> {
         scf::YieldOp::create(b, loc, result);
       };
 
-      result =
-          rewriter
-              .create<scf::IfOp>(loc, condition, /*thenBuilder=*/loadBuilder,
+      result = scf::IfOp::create(rewriter, loc, condition,
+                                 /*thenBuilder=*/loadBuilder,
                                  /*elseBuilder=*/passThruBuilder)
-              .getResult(0);
+                   .getResult(0);
     }
 
     rewriter.replaceOp(op, result);
diff --git a/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp b/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp
index e910932..4baeb11 100644
--- a/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp
+++ b/mlir/lib/Dialect/Vector/Transforms/LowerVectorTransfer.cpp
@@ -142,8 +142,8 @@ struct TransferReadPermutationLowering
 
     // Transpose result of transfer_read.
     SmallVector<int64_t> transposePerm(permutation.begin(), permutation.end());
-    return rewriter
-        .create<vector::TransposeOp>(op.getLoc(), newRead, transposePerm)
+    return vector::TransposeOp::create(rewriter, op.getLoc(), newRead,
+                                       transposePerm)
         .getResult();
   }
 };
@@ -371,8 +371,8 @@ struct TransferOpReduceRank
         rewriter, op.getLoc(), newReadType, op.getBase(), op.getIndices(),
         AffineMapAttr::get(newMap), op.getPadding(), op.getMask(),
         newInBoundsAttr);
-    return rewriter
-        .create<vector::BroadcastOp>(op.getLoc(), originalVecType, newRead)
+    return vector::BroadcastOp::create(rewriter, op.getLoc(), originalVecType,
+                                       newRead)
         .getVector();
   }
 };
diff --git a/mlir/lib/Dialect/Vector/Transforms/VectorDistribute.cpp b/mlir/lib/Dialect/Vector/Transforms/VectorDistribute.cpp
index 58e94ea..bb0f339 100644
--- a/mlir/lib/Dialect/Vector/Transforms/VectorDistribute.cpp
+++ b/mlir/lib/Dialect/Vector/Transforms/VectorDistribute.cpp
@@ -451,10 +451,9 @@ struct WarpOpTransferWrite : public WarpDistributionPattern {
     }
     SmallVector<Value> delinearized;
     if (map.getNumResults() > 1) {
-      delinearized = rewriter
-                         .create<mlir::affine::AffineDelinearizeIndexOp>(
-                             newWarpOp.getLoc(), newWarpOp.getLaneid(),
-                             delinearizedIdSizes)
+      delinearized = mlir::affine::AffineDelinearizeIndexOp::create(
+                         rewriter, newWarpOp.getLoc(), newWarpOp.getLaneid(),
+                         delinearizedIdSizes)
                          .getResults();
     } else {
       // If there is only one map result, we can elide the delinearization
@@ -1538,19 +1537,18 @@ struct WarpOpInsertScalar : public WarpDistributionPattern {
         arith::CmpIOp::create(rewriter, loc, arith::CmpIPredicate::eq,
                               newWarpOp.getLaneid(), insertingLane);
     Value newResult =
-        rewriter
-            .create<scf::IfOp>(
-                loc, isInsertingLane,
-                /*thenBuilder=*/
-                [&](OpBuilder &builder, Location loc) {
-                  Value newInsert = vector::InsertOp::create(
-                      builder, loc, newSource, distributedVec, newPos);
-                  scf::YieldOp::create(builder, loc, newInsert);
-                },
-                /*elseBuilder=*/
-                [&](OpBuilder &builder, Location loc) {
-                  scf::YieldOp::create(builder, loc, distributedVec);
-                })
+        scf::IfOp::create(
+            rewriter, loc, isInsertingLane,
+            /*thenBuilder=*/
+            [&](OpBuilder &builder, Location loc) {
+              Value newInsert = vector::InsertOp::create(
+                  builder, loc, newSource, distributedVec, newPos);
+              scf::YieldOp::create(builder, loc, newInsert);
+            },
+            /*elseBuilder=*/
+            [&](OpBuilder &builder, Location loc) {
+              scf::YieldOp::create(builder, loc, distributedVec);
+            })
             .getResult(0);
     rewriter.replaceAllUsesWith(newWarpOp->getResult(operandNumber), newResult);
     return success();
@@ -1661,10 +1659,9 @@ struct WarpOpInsert : public WarpDistributionPattern {
       auto nonInsertingBuilder = [&](OpBuilder &builder, Location loc) {
         scf::YieldOp::create(builder, loc, distributedDest);
       };
-      newResult = rewriter
-                      .create<scf::IfOp>(loc, isInsertingLane,
-                                         /*thenBuilder=*/insertingBuilder,
-                                         /*elseBuilder=*/nonInsertingBuilder)
+      newResult = scf::IfOp::create(rewriter, loc, isInsertingLane,
+                                    /*thenBuilder=*/insertingBuilder,
+                                    /*elseBuilder=*/nonInsertingBuilder)
                       .getResult(0);
     }
 
diff --git a/mlir/lib/Dialect/Vector/Transforms/VectorDropLeadUnitDim.cpp b/mlir/lib/Dialect/Vector/Transforms/VectorDropLeadUnitDim.cpp
index 73388a5..9889d7f2 100644
--- a/mlir/lib/Dialect/Vector/Transforms/VectorDropLeadUnitDim.cpp
+++ b/mlir/lib/Dialect/Vector/Transforms/VectorDropLeadUnitDim.cpp
@@ -466,9 +466,9 @@ mlir::vector::castAwayContractionLeadingOneDim(vector::ContractionOp contractOp,
     newOp = mlir::vector::maskOperation(rewriter, newOp, newMask);
   }
 
-  return rewriter
-      .create<vector::BroadcastOp>(loc, contractOp->getResultTypes()[0],
-                                   newOp->getResults()[0])
+  return vector::BroadcastOp::create(rewriter, loc,
+                                     contractOp->getResultTypes()[0],
+                                     newOp->getResults()[0])
       .getResult();
 }
 
diff --git a/mlir/lib/Dialect/Vector/Transforms/VectorEmulateNarrowType.cpp b/mlir/lib/Dialect/Vector/Transforms/VectorEmulateNarrowType.cpp
index e6bb96f..f78e579 100644
--- a/mlir/lib/Dialect/Vector/Transforms/VectorEmulateNarrowType.cpp
+++ b/mlir/lib/Dialect/Vector/Transforms/VectorEmulateNarrowType.cpp
@@ -32,7 +32,7 @@
 #include "mlir/IR/Value.h"
 #include "mlir/Transforms/DialectConversion.h"
 #include "llvm/ADT/SmallVector.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 #include "llvm/Support/MathExtras.h"
 #include "llvm/Support/raw_ostream.h"
 #include <cstdint>
@@ -41,9 +41,6 @@
 using namespace mlir;
 
 #define DEBUG_TYPE "vector-narrow-type-emulation"
-#define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE "]: ")
-#define DBGSNL() (llvm::dbgs() << "\n")
-#define LDBG(X) LLVM_DEBUG(DBGS() << X << "\n")
 
 using VectorValue = TypedValue<VectorType>;
 using MemRefValue = TypedValue<MemRefType>;
@@ -135,17 +132,16 @@ static FailureOr<Operation *> getCompressedMaskOp(OpBuilder &rewriter,
                 return vector::CreateMaskOp::create(rewriter, loc, newMaskType,
                                                     newMaskOperands);
               })
-          .Case<vector::ConstantMaskOp>(
-              [&](auto constantMaskOp) -> std::optional<Operation *> {
-                // Take the shape of mask, compress its trailing dimension:
-                SmallVector<int64_t> maskDimSizes(
-                    constantMaskOp.getMaskDimSizes());
-                int64_t &maskIndex = maskDimSizes.back();
-                maskIndex = llvm::divideCeil(numFrontPadElems + maskIndex,
-                                             numSrcElemsPerDest);
-                return vector::ConstantMaskOp::create(
-                    rewriter, loc, newMaskType, maskDimSizes);
-              })
+          .Case<vector::ConstantMaskOp>([&](auto constantMaskOp)
+                                            -> std::optional<Operation *> {
+            // Take the shape of mask, compress its trailing dimension:
+            SmallVector<int64_t> maskDimSizes(constantMaskOp.getMaskDimSizes());
+            int64_t &maskIndex = maskDimSizes.back();
+            maskIndex = llvm::divideCeil(numFrontPadElems + maskIndex,
+                                         numSrcElemsPerDest);
+            return vector::ConstantMaskOp::create(rewriter, loc, newMaskType,
+                                                  maskDimSizes);
+          })
           .Case<arith::ConstantOp>([&](auto constantOp)
                                        -> std::optional<Operation *> {
             // TODO: Support multiple dimensions.
@@ -232,9 +228,8 @@ static Value staticallyExtractSubvector(OpBuilder &rewriter, Location loc,
 
   auto resultVectorType =
       VectorType::get({numElemsToExtract}, vectorType.getElementType());
-  return rewriter
-      .create<vector::ExtractStridedSliceOp>(loc, resultVectorType, src,
-                                             offsets, sizes, strides)
+  return vector::ExtractStridedSliceOp::create(rewriter, loc, resultVectorType,
+                                               src, offsets, sizes, strides)
       ->getResult(0);
 }
 
@@ -1526,11 +1521,11 @@ BitCastBitsEnumerator::BitCastBitsEnumerator(VectorType sourceVectorType,
          "requires -D non-scalable vector type");
   int64_t sourceBitWidth = sourceVectorType.getElementTypeBitWidth();
   int64_t mostMinorSourceDim = sourceVectorType.getShape().back();
-  LDBG("sourceVectorType: " << sourceVectorType);
+  LDBG() << "sourceVectorType: " << sourceVectorType;
 
   int64_t targetBitWidth = targetVectorType.getElementTypeBitWidth();
   int64_t mostMinorTargetDim = targetVectorType.getShape().back();
-  LDBG("targetVectorType: " << targetVectorType);
+  LDBG() << "targetVectorType: " << targetVectorType;
 
   int64_t bitwidth = targetBitWidth * mostMinorTargetDim;
   (void)mostMinorSourceDim;
@@ -1555,7 +1550,7 @@ BitCastBitsEnumerator::BitCastBitsEnumerator(VectorType sourceVectorType,
 BitCastRewriter::BitCastRewriter(VectorType sourceVectorType,
                                  VectorType targetVectorType)
     : enumerator(BitCastBitsEnumerator(sourceVectorType, targetVectorType)) {
-  LDBG("\n" << enumerator.sourceElementRanges);
+  LDBG() << "\n" << enumerator.sourceElementRanges;
 }
 
 /// Verify that the precondition type meets the common preconditions for any
diff --git a/mlir/lib/Dialect/Vector/Transforms/VectorTransferOpTransforms.cpp b/mlir/lib/Dialect/Vector/Transforms/VectorTransferOpTransforms.cpp
index 2676d25..48d680c 100644
--- a/mlir/lib/Dialect/Vector/Transforms/VectorTransferOpTransforms.cpp
+++ b/mlir/lib/Dialect/Vector/Transforms/VectorTransferOpTransforms.cpp
@@ -330,8 +330,8 @@ createMaskDropNonScalableUnitDims(PatternRewriter &rewriter, Location loc,
     }
     reducedOperands.push_back(operand);
   }
-  return rewriter
-      .create<vector::CreateMaskOp>(loc, reducedType, reducedOperands)
+  return vector::CreateMaskOp::create(rewriter, loc, reducedType,
+                                      reducedOperands)
       .getResult();
 }
 
diff --git a/mlir/lib/Dialect/Vector/Transforms/VectorTransferSplitRewritePatterns.cpp b/mlir/lib/Dialect/Vector/Transforms/VectorTransferSplitRewritePatterns.cpp
index 05b0074..5e12dc4 100644
--- a/mlir/lib/Dialect/Vector/Transforms/VectorTransferSplitRewritePatterns.cpp
+++ b/mlir/lib/Dialect/Vector/Transforms/VectorTransferSplitRewritePatterns.cpp
@@ -348,24 +348,23 @@ getLocationToWriteFullVec(RewriterBase &b, vector::TransferWriteOp xferOp,
   Location loc = xferOp.getLoc();
   Value zero = arith::ConstantIndexOp::create(b, loc, 0);
   Value memref = xferOp.getBase();
-  return b
-      .create<scf::IfOp>(
-          loc, inBoundsCond,
-          [&](OpBuilder &b, Location loc) {
-            Value res =
-                castToCompatibleMemRefType(b, memref, compatibleMemRefType);
-            scf::ValueVector viewAndIndices{res};
-            llvm::append_range(viewAndIndices, xferOp.getIndices());
-            scf::YieldOp::create(b, loc, viewAndIndices);
-          },
-          [&](OpBuilder &b, Location loc) {
-            Value casted =
-                castToCompatibleMemRefType(b, alloc, compatibleMemRefType);
-            scf::ValueVector viewAndIndices{casted};
-            viewAndIndices.insert(viewAndIndices.end(),
-                                  xferOp.getTransferRank(), zero);
-            scf::YieldOp::create(b, loc, viewAndIndices);
-          })
+  return scf::IfOp::create(
+             b, loc, inBoundsCond,
+             [&](OpBuilder &b, Location loc) {
+               Value res =
+                   castToCompatibleMemRefType(b, memref, compatibleMemRefType);
+               scf::ValueVector viewAndIndices{res};
+               llvm::append_range(viewAndIndices, xferOp.getIndices());
+               scf::YieldOp::create(b, loc, viewAndIndices);
+             },
+             [&](OpBuilder &b, Location loc) {
+               Value casted =
+                   castToCompatibleMemRefType(b, alloc, compatibleMemRefType);
+               scf::ValueVector viewAndIndices{casted};
+               viewAndIndices.insert(viewAndIndices.end(),
+                                     xferOp.getTransferRank(), zero);
+               scf::YieldOp::create(b, loc, viewAndIndices);
+             })
       ->getResults();
 }
 
diff --git a/mlir/lib/Dialect/Vector/Transforms/VectorTransforms.cpp b/mlir/lib/Dialect/Vector/Transforms/VectorTransforms.cpp
index 73ca327..8de87fe 100644
--- a/mlir/lib/Dialect/Vector/Transforms/VectorTransforms.cpp
+++ b/mlir/lib/Dialect/Vector/Transforms/VectorTransforms.cpp
@@ -410,9 +410,8 @@ FailureOr<Value> combineContractAndBroadcast(vector::ContractionOp contractOp,
           oldMaskType.getScalableDims().drop_front(unusedDimsBitVector.count());
       VectorType maskOpType =
           VectorType::get(newShape, rewriter.getI1Type(), newShapeScalableDims);
-      mask = rewriter
-                 .create<vector::ShapeCastOp>(contractOp.getLoc(), maskOpType,
-                                              maskingOp.getMask())
+      mask = vector::ShapeCastOp::create(rewriter, contractOp.getLoc(),
+                                         maskOpType, maskingOp.getMask())
                  .getResult();
     }
 
diff --git a/mlir/lib/Dialect/Vector/Transforms/VectorUnroll.cpp b/mlir/lib/Dialect/Vector/Transforms/VectorUnroll.cpp
index 062c51f..501abec 100644
--- a/mlir/lib/Dialect/Vector/Transforms/VectorUnroll.cpp
+++ b/mlir/lib/Dialect/Vector/Transforms/VectorUnroll.cpp
@@ -16,13 +16,11 @@
 #include "mlir/Interfaces/VectorInterfaces.h"
 #include "llvm/ADT/MapVector.h"
 #include "llvm/ADT/STLExtras.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 #include "llvm/Support/InterleavedRange.h"
 #include <optional>
 
 #define DEBUG_TYPE "vector-unroll"
-#define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE "]: ")
-#define LDBG(X) LLVM_DEBUG(DBGS() << X << "\n")
 
 using namespace mlir;
 using namespace mlir::vector;
@@ -90,10 +88,9 @@ static Operation *cloneOpWithOperandsAndTypes(OpBuilder &builder, Location loc,
 /// std::nullopt if the op shouldn't be or cannot be unrolled.
 static std::optional<SmallVector<int64_t>>
 getTargetShape(const vector::UnrollVectorOptions &options, Operation *op) {
-  LDBG("");
-  LDBG("Get unroll shape for op " << op->getName().getStringRef());
+  LDBG() << "Get unroll shape for op " << op->getName().getStringRef();
   if (options.filterConstraint && failed(options.filterConstraint(op))) {
-    LDBG("--no filter constraint -> BAIL");
+    LDBG() << "--no filter constraint -> BAIL";
     return std::nullopt;
   }
   assert(options.nativeShape &&
@@ -101,33 +98,33 @@ getTargetShape(const vector::UnrollVectorOptions &options, Operation *op) {
          "shape call back function to be set");
   auto unrollableVectorOp = dyn_cast<VectorUnrollOpInterface>(op);
   if (!unrollableVectorOp) {
-    LDBG("--not an unrollable op -> BAIL");
+    LDBG() << "--not an unrollable op -> BAIL";
     return std::nullopt;
   }
   auto maybeUnrollShape = unrollableVectorOp.getShapeForUnroll();
   if (!maybeUnrollShape) {
-    LDBG("--could not get shape of op " << *op << " -> BAIL");
+    LDBG() << "--could not get shape of op " << *op << " -> BAIL";
     return std::nullopt;
   }
-  LDBG("--vector op shape: " << llvm::interleaved(*maybeUnrollShape));
+  LDBG() << "--vector op shape: " << llvm::interleaved(*maybeUnrollShape);
 
   std::optional<SmallVector<int64_t>> targetShape = options.nativeShape(op);
   if (!targetShape) {
-    LDBG("--no unrolling target shape defined " << *op << "-> SKIP");
+    LDBG() << "--no unrolling target shape defined " << *op << "-> SKIP";
     return std::nullopt;
   }
-  LDBG("--target shape: " << llvm::interleaved(*targetShape));
+  LDBG() << "--target shape: " << llvm::interleaved(*targetShape);
 
   auto maybeShapeRatio = computeShapeRatio(*maybeUnrollShape, *targetShape);
   if (!maybeShapeRatio) {
-    LDBG("--could not compute integral shape ratio -> BAIL");
+    LDBG() << "--could not compute integral shape ratio -> BAIL";
     return std::nullopt;
   }
   if (llvm::all_of(*maybeShapeRatio, [](int64_t v) { return v == 1; })) {
-    LDBG("--no unrolling needed -> SKIP");
+    LDBG() << "--no unrolling needed -> SKIP";
     return std::nullopt;
   }
-  LDBG("--found an integral shape ratio to unroll to -> SUCCESS");
+  LDBG() << "--found an integral shape ratio to unroll to -> SUCCESS";
   return targetShape;
 }
 
diff --git a/mlir/lib/Dialect/Vector/Utils/VectorUtils.cpp b/mlir/lib/Dialect/Vector/Utils/VectorUtils.cpp
index c045063..10ed2bc 100644
--- a/mlir/lib/Dialect/Vector/Utils/VectorUtils.cpp
+++ b/mlir/lib/Dialect/Vector/Utils/VectorUtils.cpp
@@ -27,13 +27,11 @@
 #include "mlir/Support/LLVM.h"
 
 #include "llvm/ADT/DenseSet.h"
+#include "llvm/Support/DebugLog.h"
 #include "llvm/Support/InterleavedRange.h"
 
 #define DEBUG_TYPE "vector-utils"
 
-#define DBGS() (llvm::dbgs() << '[' << DEBUG_TYPE << "] ")
-#define LDBG(X) LLVM_DEBUG(DBGS() << X << "\n")
-
 using namespace mlir;
 
 /// Helper function that creates a memref::DimOp or tensor::DimOp depending on
@@ -369,14 +367,14 @@ Value vector::createReadOrMaskedRead(OpBuilder &builder, Location loc,
 LogicalResult
 vector::isValidMaskedInputVector(ArrayRef<int64_t> shape,
                                  ArrayRef<int64_t> inputVectorSizes) {
-  LDBG("Iteration space static sizes:" << llvm::interleaved(shape));
+  LDBG() << "Iteration space static sizes:" << llvm::interleaved(shape);
 
   if (inputVectorSizes.size() != shape.size()) {
-    LDBG("Input vector sizes don't match the number of loops");
+    LDBG() << "Input vector sizes don't match the number of loops";
     return failure();
   }
   if (ShapedType::isDynamicShape(inputVectorSizes)) {
-    LDBG("Input vector sizes can't have dynamic dimensions");
+    LDBG() << "Input vector sizes can't have dynamic dimensions";
     return failure();
   }
   if (!llvm::all_of(llvm::zip(shape, inputVectorSizes),
@@ -386,8 +384,9 @@ vector::isValidMaskedInputVector(ArrayRef<int64_t> shape,
                       return ShapedType::isDynamic(staticSize) ||
                              staticSize <= inputSize;
                     })) {
-    LDBG("Input vector sizes must be greater than or equal to iteration space "
-         "static sizes");
+    LDBG() << "Input vector sizes must be greater than or equal to iteration "
+              "space "
+              "static sizes";
     return failure();
   }
   return success();
diff --git a/mlir/lib/Dialect/XeGPU/Transforms/XeGPUBlocking.cpp b/mlir/lib/Dialect/XeGPU/Transforms/XeGPUBlocking.cpp
index 4656f11..d82c541 100644
--- a/mlir/lib/Dialect/XeGPU/Transforms/XeGPUBlocking.cpp
+++ b/mlir/lib/Dialect/XeGPU/Transforms/XeGPUBlocking.cpp
@@ -17,6 +17,7 @@
 #include "mlir/Transforms/DialectConversion.h"
 #include "mlir/Transforms/GreedyPatternRewriteDriver.h"
 #include "llvm/ADT/STLExtras.h"
+#include "llvm/Support/DebugLog.h"
 
 namespace mlir {
 namespace xegpu {
@@ -26,8 +27,6 @@ namespace xegpu {
 } // namespace mlir
 
 #define DEBUG_TYPE "xegpu-blocking"
-#define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE "]: ")
-#define LDBG(X) LLVM_DEBUG(DBGS() << X << "\n")
 
 using namespace mlir;
 
@@ -53,7 +52,7 @@ resolveUnrealizedConversionCastOp(UnrealizedConversionCastOp castOp) {
   // We only interest in the case where all inputs and outputs have the
   // identical VectorTypes
   if (!hasIdenticalVectorTypes(inputs) || !hasIdenticalVectorTypes(outputs)) {
-    LDBG("skip unrealized conversion cast op not emulating pack/unpack.");
+    LDBG() << "skip unrealized conversion cast op not emulating pack/unpack.";
     return;
   }
 
@@ -149,7 +148,7 @@ XeGPUBlockingPass::getTileShape(const T &operandOrResult) const {
     if (auto type = dyn_cast<ShapedType>(value.getType()))
       return llvm::to_vector(type.getShape());
   }
-  LDBG("failed to getTileShape for: " << value);
+  LDBG() << "failed to getTileShape for: " << value;
   return std::nullopt;
 }
 
@@ -214,7 +213,7 @@ bool XeGPUBlockingPass::needsUnroll(Operation *op) const {
         return layout && layout.isWgLayout();
       });
   if (hasWgLayoutOperands || hasWgLayoutResults) {
-    LDBG("skip unrolling for op with workgroup level layout: " << *op);
+    LDBG() << "skip unrolling for op with workgroup level layout: " << *op;
     return false;
   }
 
diff --git a/mlir/lib/Dialect/XeGPU/Transforms/XeGPUUnroll.cpp b/mlir/lib/Dialect/XeGPU/Transforms/XeGPUUnroll.cpp
index a6208b4..ec8fad4 100644
--- a/mlir/lib/Dialect/XeGPU/Transforms/XeGPUUnroll.cpp
+++ b/mlir/lib/Dialect/XeGPU/Transforms/XeGPUUnroll.cpp
@@ -17,7 +17,7 @@
 #include "mlir/Dialect/XeGPU/Transforms/Transforms.h"
 #include "mlir/Dialect/XeGPU/Utils/XeGPUUtils.h"
 #include "llvm/ADT/STLExtras.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 
 namespace mlir {
 namespace xegpu {
@@ -27,8 +27,6 @@ namespace xegpu {
 } // namespace mlir
 
 #define DEBUG_TYPE "xegpu-unroll"
-#define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE "]: ")
-#define LDBG(X) LLVM_DEBUG(DBGS() << X << "\n")
 
 using namespace mlir;
 
@@ -44,11 +42,10 @@ protected:
   /// Return the target shape for the given `op`. Return std::nullopt if the
   /// op shouldn't be or cannot be unrolled.
   std::optional<SmallVector<int64_t>> getTargetShape(Operation *op) const {
-    LDBG("");
-    LDBG("Get unroll shape for: " << *op);
+    LDBG() << "Get unroll shape for: " << *op;
 
     if (options.filterConstraint && failed(options.filterConstraint(op))) {
-      LDBG("--no filter constraint -> BAIL");
+      LDBG() << "--no filter constraint -> BAIL";
       return std::nullopt;
     }
 
diff --git a/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp b/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp
index 229a289..850f70c 100644
--- a/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp
+++ b/mlir/lib/Dialect/XeGPU/Transforms/XeGPUWgToSgDistribute.cpp
@@ -207,7 +207,7 @@ struct WgToSgCreateNdOp : public OpConversionPattern<xegpu::CreateNdDescOp> {
       // Subtract startOfRange from the original subgroup id to get the adjusted
       // sg id
       Value startOfRangeVal =
-          rewriter.create<arith::ConstantIndexOp>(loc, startOfRange);
+          arith::ConstantIndexOp::create(rewriter, loc, startOfRange);
       adjustedSgId =
           rewriter.createOrFold<index::SubOp>(loc, linearSgId, startOfRangeVal);
     }
@@ -431,8 +431,8 @@ struct WgToSgVectorBroadcastOp
 
     SmallVector<Value> newBroadcastOps;
     for (auto operand : adaptor.getOperands().front()) {
-      auto newBroadcast = rewriter.create<vector::BroadcastOp>(
-          op.getLoc(), newResultType, operand);
+      auto newBroadcast = vector::BroadcastOp::create(rewriter, op.getLoc(),
+                                                      newResultType, operand);
       xegpu::setLayoutAttr(newBroadcast->getResult(0),
                            layout.dropSgLayoutAndData());
       newBroadcastOps.push_back(newBroadcast.getResult());
@@ -563,8 +563,8 @@ struct WgToSgConvertLayoutOp
     if (input && target) {
       // keep the ConvertLayoutOp for rest fields, e.g., inst_data.
       for (auto [i, src] : llvm::enumerate(adaptor.getSource())) {
-        auto newOp = rewriter.create<xegpu::ConvertLayoutOp>(
-            op.getLoc(), src.getType(), src, input, target);
+        auto newOp = xegpu::ConvertLayoutOp::create(
+            rewriter, op.getLoc(), src.getType(), src, input, target);
         newOps[i] = newOp;
       }
     }
diff --git a/mlir/lib/IR/PatternLoggingListener.cpp b/mlir/lib/IR/PatternLoggingListener.cpp
index ce2123a..0db13ab 100644
--- a/mlir/lib/IR/PatternLoggingListener.cpp
+++ b/mlir/lib/IR/PatternLoggingListener.cpp
@@ -1,50 +1,48 @@
 #include "mlir/IR/PatternMatch.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 
 #define DEBUG_TYPE "pattern-logging-listener"
-#define DBGS() (llvm::dbgs() << "[" << DEBUG_TYPE << "] ")
-#define LDBG(X) LLVM_DEBUG(DBGS() << X << "\n")
 
 using namespace mlir;
 
 void RewriterBase::PatternLoggingListener::notifyOperationInserted(
     Operation *op, InsertPoint previous) {
-  LDBG(patternName << " | notifyOperationInserted"
-                   << " | " << op->getName());
+  LDBG() << patternName << " | notifyOperationInserted"
+         << " | " << op->getName();
   ForwardingListener::notifyOperationInserted(op, previous);
 }
 
 void RewriterBase::PatternLoggingListener::notifyOperationModified(
     Operation *op) {
-  LDBG(patternName << " | notifyOperationModified"
-                   << " | " << op->getName());
+  LDBG() << patternName << " | notifyOperationModified"
+         << " | " << op->getName();
   ForwardingListener::notifyOperationModified(op);
 }
 
 void RewriterBase::PatternLoggingListener::notifyOperationReplaced(
     Operation *op, Operation *newOp) {
-  LDBG(patternName << " | notifyOperationReplaced (with op)"
-                   << " | " << op->getName() << " | " << newOp->getName());
+  LDBG() << patternName << " | notifyOperationReplaced (with op)"
+         << " | " << op->getName() << " | " << newOp->getName();
   ForwardingListener::notifyOperationReplaced(op, newOp);
 }
 
 void RewriterBase::PatternLoggingListener::notifyOperationReplaced(
     Operation *op, ValueRange replacement) {
-  LDBG(patternName << " | notifyOperationReplaced (with values)"
-                   << " | " << op->getName());
+  LDBG() << patternName << " | notifyOperationReplaced (with values)"
+         << " | " << op->getName();
   ForwardingListener::notifyOperationReplaced(op, replacement);
 }
 
 void RewriterBase::PatternLoggingListener::notifyOperationErased(
     Operation *op) {
-  LDBG(patternName << " | notifyOperationErased"
-                   << " | " << op->getName());
+  LDBG() << patternName << " | notifyOperationErased"
+         << " | " << op->getName();
   ForwardingListener::notifyOperationErased(op);
 }
 
 void RewriterBase::PatternLoggingListener::notifyPatternBegin(
     const Pattern &pattern, Operation *op) {
-  LDBG(patternName << " | notifyPatternBegin"
-                   << " | " << op->getName());
+  LDBG() << patternName << " | notifyPatternBegin"
+         << " | " << op->getName();
   ForwardingListener::notifyPatternBegin(pattern, op);
 }
diff --git a/mlir/lib/IR/PatternMatch.cpp b/mlir/lib/IR/PatternMatch.cpp
index 5c98417..9332f55 100644
--- a/mlir/lib/IR/PatternMatch.cpp
+++ b/mlir/lib/IR/PatternMatch.cpp
@@ -156,6 +156,11 @@ void RewriterBase::eraseOp(Operation *op) {
   assert(op->use_empty() && "expected 'op' to have no uses");
   auto *rewriteListener = dyn_cast_if_present<Listener>(listener);
 
+  // If the current insertion point is before the erased operation, we adjust
+  // the insertion point to be after the operation.
+  if (getInsertionPoint() == op->getIterator())
+    setInsertionPointAfter(op);
+
   // Fast path: If no listener is attached, the op can be dropped in one go.
   if (!rewriteListener) {
     op->erase();
@@ -320,6 +325,11 @@ void RewriterBase::inlineBlockBefore(Block *source, Block *dest,
       moveOpBefore(&source->front(), dest, before);
   }
 
+  // If the current insertion point is within the source block, adjust the
+  // insertion point to the destination block.
+  if (getInsertionBlock() == source)
+    setInsertionPoint(dest, getInsertionPoint());
+
   // Erase the source block.
   assert(source->empty() && "expected 'source' to be empty");
   eraseBlock(source);
diff --git a/mlir/lib/Support/TypeID.cpp b/mlir/lib/Support/TypeID.cpp
index 01ad910..304253c 100644
--- a/mlir/lib/Support/TypeID.cpp
+++ b/mlir/lib/Support/TypeID.cpp
@@ -27,9 +27,6 @@ namespace {
 struct ImplicitTypeIDRegistry {
   /// Lookup or insert a TypeID for the given type name.
   TypeID lookupOrInsert(StringRef typeName) {
-    LLVM_DEBUG(llvm::dbgs() << "ImplicitTypeIDRegistry::lookupOrInsert("
-                            << typeName << ")\n");
-
     // Perform a heuristic check to see if this type is in an anonymous
     // namespace. String equality is not valid for anonymous types, so we try to
     // abort whenever we see them.
diff --git a/mlir/lib/Target/LLVM/CMakeLists.txt b/mlir/lib/Target/LLVM/CMakeLists.txt
index 7c6fc37..f6e44c6 100644
--- a/mlir/lib/Target/LLVM/CMakeLists.txt
+++ b/mlir/lib/Target/LLVM/CMakeLists.txt
@@ -8,6 +8,7 @@ add_mlir_library(MLIRTargetLLVM
   intrinsics_gen
 
   LINK_COMPONENTS
+  BitWriter
   Core
   IPO
   IRReader
diff --git a/mlir/lib/Target/LLVM/NVVM/Target.cpp b/mlir/lib/Target/LLVM/NVVM/Target.cpp
index b4d53d4..55c8a64 100644
--- a/mlir/lib/Target/LLVM/NVVM/Target.cpp
+++ b/mlir/lib/Target/LLVM/NVVM/Target.cpp
@@ -16,7 +16,9 @@
 #include "mlir/Dialect/GPU/IR/CompilationInterfaces.h"
 #include "mlir/Dialect/GPU/IR/GPUDialect.h"
 #include "mlir/Dialect/LLVMIR/NVVMDialect.h"
+#include "mlir/IR/BuiltinDialect.h"
 #include "mlir/IR/BuiltinTypes.h"
+#include "mlir/IR/DialectResourceBlobManager.h"
 #include "mlir/Target/LLVM/NVVM/Utils.h"
 #include "mlir/Target/LLVMIR/Dialect/GPU/GPUToLLVMIRTranslation.h"
 #include "mlir/Target/LLVMIR/Dialect/LLVMIR/LLVMToLLVMIRTranslation.h"
diff --git a/mlir/lib/Target/LLVMIR/ModuleImport.cpp b/mlir/lib/Target/LLVMIR/ModuleImport.cpp
index 94db7f8..58e3c44 100644
--- a/mlir/lib/Target/LLVMIR/ModuleImport.cpp
+++ b/mlir/lib/Target/LLVMIR/ModuleImport.cpp
@@ -142,6 +142,7 @@ static LogicalResult convertInstructionImpl(OpBuilder &odsBuilder,
   // TODO: Implement the `convertInstruction` hooks in the
   // `LLVMDialectLLVMIRImportInterface` and move the following include there.
 #include "mlir/Dialect/LLVMIR/LLVMOpFromLLVMIRConversions.inc"
+
   return failure();
 }
 
@@ -1626,12 +1627,11 @@ FailureOr<Value> ModuleImport::convertConstant(llvm::Constant *constant) {
   // Convert dso_local_equivalent.
   if (auto *dsoLocalEquivalent = dyn_cast<llvm::DSOLocalEquivalent>(constant)) {
     Type type = convertType(dsoLocalEquivalent->getType());
-    return builder
-        .create<DSOLocalEquivalentOp>(
-            loc, type,
-            FlatSymbolRefAttr::get(
-                builder.getContext(),
-                dsoLocalEquivalent->getGlobalValue()->getName()))
+    return DSOLocalEquivalentOp::create(
+               builder, loc, type,
+               FlatSymbolRefAttr::get(
+                   builder.getContext(),
+                   dsoLocalEquivalent->getGlobalValue()->getName()))
         .getResult();
   }
 
@@ -1736,9 +1736,9 @@ FailureOr<Value> ModuleImport::convertConstant(llvm::Constant *constant) {
         FlatSymbolRefAttr::get(context, blockAddr->getFunction()->getName());
     auto blockTag =
         BlockTagAttr::get(context, blockAddr->getBasicBlock()->getNumber());
-    return builder
-        .create<BlockAddressOp>(loc, convertType(blockAddr->getType()),
-                                BlockAddressAttr::get(context, fnSym, blockTag))
+    return BlockAddressOp::create(
+               builder, loc, convertType(blockAddr->getType()),
+               BlockAddressAttr::get(context, fnSym, blockTag))
         .getRes();
   }
 
@@ -2228,17 +2228,16 @@ LogicalResult ModuleImport::convertInstruction(llvm::Instruction *inst) {
         if (!resultTy)
           return failure();
         ArrayAttr operandAttrs = convertAsmInlineOperandAttrs(*callInst);
-        return builder
-            .create<InlineAsmOp>(
-                loc, resultTy, *operands,
-                builder.getStringAttr(asmI->getAsmString()),
-                builder.getStringAttr(asmI->getConstraintString()),
-                asmI->hasSideEffects(), asmI->isAlignStack(),
-                convertTailCallKindFromLLVM(callInst->getTailCallKind()),
-                AsmDialectAttr::get(
-                    mlirModule.getContext(),
-                    convertAsmDialectFromLLVM(asmI->getDialect())),
-                operandAttrs)
+        return InlineAsmOp::create(
+                   builder, loc, resultTy, *operands,
+                   builder.getStringAttr(asmI->getAsmString()),
+                   builder.getStringAttr(asmI->getConstraintString()),
+                   asmI->hasSideEffects(), asmI->isAlignStack(),
+                   convertTailCallKindFromLLVM(callInst->getTailCallKind()),
+                   AsmDialectAttr::get(
+                       mlirModule.getContext(),
+                       convertAsmDialectFromLLVM(asmI->getDialect())),
+                   operandAttrs)
             .getOperation();
       }
       bool isIncompatibleCall;
diff --git a/mlir/lib/Transforms/RemoveDeadValues.cpp b/mlir/lib/Transforms/RemoveDeadValues.cpp
index ddd5f2b..4ccb83f 100644
--- a/mlir/lib/Transforms/RemoveDeadValues.cpp
+++ b/mlir/lib/Transforms/RemoveDeadValues.cpp
@@ -36,6 +36,7 @@
 #include "mlir/IR/Builders.h"
 #include "mlir/IR/BuiltinAttributes.h"
 #include "mlir/IR/Dialect.h"
+#include "mlir/IR/Operation.h"
 #include "mlir/IR/OperationSupport.h"
 #include "mlir/IR/SymbolTable.h"
 #include "mlir/IR/Value.h"
@@ -51,6 +52,7 @@
 #include "mlir/Transforms/Passes.h"
 #include "llvm/ADT/STLExtras.h"
 #include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 #include <cassert>
 #include <cstddef>
 #include <memory>
@@ -58,8 +60,6 @@
 #include <vector>
 
 #define DEBUG_TYPE "remove-dead-values"
-#define DBGS() (llvm::dbgs() << '[' << DEBUG_TYPE << "] ")
-#define LDBG(X) LLVM_DEBUG(DBGS() << X << "\n")
 
 namespace mlir {
 #define GEN_PASS_DEF_REMOVEDEADVALUES
@@ -119,21 +119,21 @@ static bool hasLive(ValueRange values, const DenseSet<Value> &nonLiveSet,
                     RunLivenessAnalysis &la) {
   for (Value value : values) {
     if (nonLiveSet.contains(value)) {
-      LDBG("Value " << value << " is already marked non-live (dead)");
+      LDBG() << "Value " << value << " is already marked non-live (dead)";
       continue;
     }
 
     const Liveness *liveness = la.getLiveness(value);
     if (!liveness) {
-      LDBG("Value " << value
-                    << " has no liveness info, conservatively considered live");
+      LDBG() << "Value " << value
+             << " has no liveness info, conservatively considered live";
       return true;
     }
     if (liveness->isLive) {
-      LDBG("Value " << value << " is live according to liveness analysis");
+      LDBG() << "Value " << value << " is live according to liveness analysis";
       return true;
     } else {
-      LDBG("Value " << value << " is dead according to liveness analysis");
+      LDBG() << "Value " << value << " is dead according to liveness analysis";
     }
   }
   return false;
@@ -148,8 +148,8 @@ static BitVector markLives(ValueRange values, const DenseSet<Value> &nonLiveSet,
   for (auto [index, value] : llvm::enumerate(values)) {
     if (nonLiveSet.contains(value)) {
       lives.reset(index);
-      LDBG("Value " << value << " is already marked non-live (dead) at index "
-                    << index);
+      LDBG() << "Value " << value
+             << " is already marked non-live (dead) at index " << index;
       continue;
     }
 
@@ -161,17 +161,17 @@ static BitVector markLives(ValueRange values, const DenseSet<Value> &nonLiveSet,
     // (because they weren't erased) and also their liveness is null because
     // liveness analysis ran before their creation.
     if (!liveness) {
-      LDBG("Value " << value << " at index " << index
-                    << " has no liveness info, conservatively considered live");
+      LDBG() << "Value " << value << " at index " << index
+             << " has no liveness info, conservatively considered live";
       continue;
     }
     if (!liveness->isLive) {
       lives.reset(index);
-      LDBG("Value " << value << " at index " << index
-                    << " is dead according to liveness analysis");
+      LDBG() << "Value " << value << " at index " << index
+             << " is dead according to liveness analysis";
     } else {
-      LDBG("Value " << value << " at index " << index
-                    << " is live according to liveness analysis");
+      LDBG() << "Value " << value << " at index " << index
+             << " is live according to liveness analysis";
     }
   }
 
@@ -187,8 +187,8 @@ static void collectNonLiveValues(DenseSet<Value> &nonLiveSet, ValueRange range,
     if (!nonLive[index])
       continue;
     nonLiveSet.insert(result);
-    LDBG("Marking value " << result << " as non-live (dead) at index "
-                          << index);
+    LDBG() << "Marking value " << result << " as non-live (dead) at index "
+           << index;
   }
 }
 
@@ -258,16 +258,18 @@ static SmallVector<OpOperand *> operandsToOpOperands(OperandRange operands) {
 static void processSimpleOp(Operation *op, RunLivenessAnalysis &la,
                             DenseSet<Value> &nonLiveSet,
                             RDVFinalCleanupList &cl) {
-  LDBG("Processing simple op: " << *op);
+  LDBG() << "Processing simple op: " << *op;
   if (!isMemoryEffectFree(op) || hasLive(op->getResults(), nonLiveSet, la)) {
-    LDBG("Simple op is not memory effect free or has live results, skipping: "
-         << *op);
+    LDBG()
+        << "Simple op is not memory effect free or has live results, skipping: "
+        << *op;
     return;
   }
 
-  LDBG("Simple op has all dead results and is memory effect free, scheduling "
-       "for removal: "
-       << *op);
+  LDBG()
+      << "Simple op has all dead results and is memory effect free, scheduling "
+         "for removal: "
+      << *op;
   cl.operations.push_back(op);
   collectNonLiveValues(nonLiveSet, op->getResults(),
                        BitVector(op->getNumResults(), true));
@@ -286,10 +288,10 @@ static void processSimpleOp(Operation *op, RunLivenessAnalysis &la,
 static void processFuncOp(FunctionOpInterface funcOp, Operation *module,
                           RunLivenessAnalysis &la, DenseSet<Value> &nonLiveSet,
                           RDVFinalCleanupList &cl) {
-  LDBG("Processing function op: " << funcOp.getOperation()->getName());
+  LDBG() << "Processing function op: " << funcOp.getOperation()->getName();
   if (funcOp.isPublic() || funcOp.isExternal()) {
-    LDBG("Function is public or external, skipping: "
-         << funcOp.getOperation()->getName());
+    LDBG() << "Function is public or external, skipping: "
+           << funcOp.getOperation()->getName();
     return;
   }
 
@@ -409,9 +411,8 @@ static void processRegionBranchOp(RegionBranchOpInterface regionBranchOp,
                                   RunLivenessAnalysis &la,
                                   DenseSet<Value> &nonLiveSet,
                                   RDVFinalCleanupList &cl) {
-  LLVM_DEBUG(DBGS() << "Processing region branch op: "; regionBranchOp->print(
-      llvm::dbgs(), OpPrintingFlags().skipRegions());
-             llvm::dbgs() << "\n");
+  LDBG() << "Processing region branch op: "
+         << OpWithFlags(regionBranchOp, OpPrintingFlags().skipRegions());
   // Mark live results of `regionBranchOp` in `liveResults`.
   auto markLiveResults = [&](BitVector &liveResults) {
     liveResults = markLives(regionBranchOp->getResults(), nonLiveSet, la);
@@ -697,7 +698,7 @@ static void processRegionBranchOp(RegionBranchOpInterface regionBranchOp,
 static void processBranchOp(BranchOpInterface branchOp, RunLivenessAnalysis &la,
                             DenseSet<Value> &nonLiveSet,
                             RDVFinalCleanupList &cl) {
-  LDBG("Processing branch op: " << *branchOp);
+  LDBG() << "Processing branch op: " << *branchOp;
   unsigned numSuccessors = branchOp->getNumSuccessors();
 
   for (unsigned succIdx = 0; succIdx < numSuccessors; ++succIdx) {
diff --git a/mlir/lib/Transforms/Utils/DialectConversion.cpp b/mlir/lib/Transforms/Utils/DialectConversion.cpp
index d224f73..08803e0 100644
--- a/mlir/lib/Transforms/Utils/DialectConversion.cpp
+++ b/mlir/lib/Transforms/Utils/DialectConversion.cpp
@@ -14,8 +14,10 @@
 #include "mlir/IR/Dominance.h"
 #include "mlir/IR/IRMapping.h"
 #include "mlir/IR/Iterators.h"
+#include "mlir/IR/Operation.h"
 #include "mlir/Interfaces/FunctionInterfaces.h"
 #include "mlir/Rewrite/PatternApplicator.h"
+#include "llvm/ADT/ScopeExit.h"
 #include "llvm/ADT/SmallPtrSet.h"
 #include "llvm/Support/Debug.h"
 #include "llvm/Support/FormatVariadic.h"
@@ -130,11 +132,6 @@ struct ConversionValueMapping {
   ///   value.
   ValueVector lookupOrDefault(Value from, TypeRange desiredTypes = {}) const;
 
-  /// Lookup the given value within the map, or return an empty vector if the
-  /// value is not mapped. If it is mapped, this follows the same behavior
-  /// as `lookupOrDefault`.
-  ValueVector lookupOrNull(Value from, TypeRange desiredTypes = {}) const;
-
   template <typename T>
   struct IsValueVector : std::is_same<std::decay_t<T>, ValueVector> {};
 
@@ -237,15 +234,6 @@ ConversionValueMapping::lookupOrDefault(Value from,
   return !desiredValue.empty() ? std::move(desiredValue) : std::move(current);
 }
 
-ValueVector ConversionValueMapping::lookupOrNull(Value from,
-                                                 TypeRange desiredTypes) const {
-  ValueVector result = lookupOrDefault(from, desiredTypes);
-  if (result == ValueVector{from} ||
-      (!desiredTypes.empty() && TypeRange(ValueRange(result)) != desiredTypes))
-    return {};
-  return result;
-}
-
 //===----------------------------------------------------------------------===//
 // Rewriter and Translation State
 //===----------------------------------------------------------------------===//
@@ -521,9 +509,11 @@ private:
 class MoveBlockRewrite : public BlockRewrite {
 public:
   MoveBlockRewrite(ConversionPatternRewriterImpl &rewriterImpl, Block *block,
-                   Region *region, Block *insertBeforeBlock)
-      : BlockRewrite(Kind::MoveBlock, rewriterImpl, block), region(region),
-        insertBeforeBlock(insertBeforeBlock) {}
+                   Region *previousRegion, Region::iterator previousIt)
+      : BlockRewrite(Kind::MoveBlock, rewriterImpl, block),
+        region(previousRegion),
+        insertBeforeBlock(previousIt == previousRegion->end() ? nullptr
+                                                              : &*previousIt) {}
 
   static bool classof(const IRRewrite *rewrite) {
     return rewrite->getKind() == Kind::MoveBlock;
@@ -630,9 +620,12 @@ protected:
 class MoveOperationRewrite : public OperationRewrite {
 public:
   MoveOperationRewrite(ConversionPatternRewriterImpl &rewriterImpl,
-                       Operation *op, Block *block, Operation *insertBeforeOp)
-      : OperationRewrite(Kind::MoveOperation, rewriterImpl, op), block(block),
-        insertBeforeOp(insertBeforeOp) {}
+                       Operation *op, OpBuilder::InsertPoint previous)
+      : OperationRewrite(Kind::MoveOperation, rewriterImpl, op),
+        block(previous.getBlock()),
+        insertBeforeOp(previous.getPoint() == previous.getBlock()->end()
+                           ? nullptr
+                           : &*previous.getPoint()) {}
 
   static bool classof(const IRRewrite *rewrite) {
     return rewrite->getKind() == Kind::MoveOperation;
@@ -926,6 +919,23 @@ struct ConversionPatternRewriterImpl : public RewriterBase::Listener {
   /// Return "true" if the given operation was replaced or erased.
   bool wasOpReplaced(Operation *op) const;
 
+  /// Lookup the most recently mapped values with the desired types in the
+  /// mapping.
+  ///
+  /// Special cases:
+  /// - If the desired type range is empty, simply return the most recently
+  ///   mapped values.
+  /// - If there is no mapping to the desired types, also return the most
+  ///   recently mapped values.
+  /// - If there is no mapping for the given values at all, return the given
+  ///   value.
+  ValueVector lookupOrDefault(Value from, TypeRange desiredTypes = {}) const;
+
+  /// Lookup the given value within the map, or return an empty vector if the
+  /// value is not mapped. If it is mapped, this follows the same behavior
+  /// as `lookupOrDefault`.
+  ValueVector lookupOrNull(Value from, TypeRange desiredTypes = {}) const;
+
   //===--------------------------------------------------------------------===//
   // IR Rewrites / Type Conversion
   //===--------------------------------------------------------------------===//
@@ -1248,6 +1258,22 @@ void ConversionPatternRewriterImpl::applyRewrites() {
 // State Management
 //===----------------------------------------------------------------------===//
 
+ValueVector
+ConversionPatternRewriterImpl::lookupOrDefault(Value from,
+                                               TypeRange desiredTypes) const {
+  return mapping.lookupOrDefault(from, desiredTypes);
+}
+
+ValueVector
+ConversionPatternRewriterImpl::lookupOrNull(Value from,
+                                            TypeRange desiredTypes) const {
+  ValueVector result = lookupOrDefault(from, desiredTypes);
+  if (result == ValueVector{from} ||
+      (!desiredTypes.empty() && TypeRange(ValueRange(result)) != desiredTypes))
+    return {};
+  return result;
+}
+
 RewriterState ConversionPatternRewriterImpl::getCurrentState() {
   return RewriterState(rewrites.size(), ignoredOps.size(), replacedOps.size());
 }
@@ -1295,7 +1321,7 @@ LogicalResult ConversionPatternRewriterImpl::remapValues(
       // The current pattern does not have a type converter. I.e., it does not
       // distinguish between legal and illegal types. For each operand, simply
       // pass through the most recently mapped values.
-      remapped.push_back(mapping.lookupOrDefault(operand));
+      remapped.push_back(lookupOrDefault(operand));
       continue;
     }
 
@@ -1314,7 +1340,7 @@ LogicalResult ConversionPatternRewriterImpl::remapValues(
       continue;
     }
 
-    ValueVector repl = mapping.lookupOrDefault(operand, legalTypes);
+    ValueVector repl = lookupOrDefault(operand, legalTypes);
     if (!repl.empty() && TypeRange(ValueRange(repl)) == legalTypes) {
       // Mapped values have the correct type or there is an existing
       // materialization. Or the operand is not mapped at all and has the
@@ -1324,7 +1350,7 @@ LogicalResult ConversionPatternRewriterImpl::remapValues(
     }
 
     // Create a materialization for the most recently mapped values.
-    repl = mapping.lookupOrDefault(operand);
+    repl = lookupOrDefault(operand);
     ValueRange castValues = buildUnresolvedMaterialization(
         MaterializationKind::Target, computeInsertPoint(repl), operandLoc,
         /*valuesToMap=*/repl, /*inputs=*/repl, /*outputTypes=*/legalTypes,
@@ -1519,7 +1545,7 @@ Value ConversionPatternRewriterImpl::findOrBuildReplacementValue(
   // Try to find a replacement value with the same type in the conversion value
   // mapping. This includes cached materializations. We try to reuse those
   // instead of generating duplicate IR.
-  ValueVector repl = mapping.lookupOrNull(value, value.getType());
+  ValueVector repl = lookupOrNull(value, value.getType());
   if (!repl.empty())
     return repl.front();
 
@@ -1535,7 +1561,7 @@ Value ConversionPatternRewriterImpl::findOrBuildReplacementValue(
   // No replacement value was found. Get the latest replacement value
   // (regardless of the type) and build a source materialization to the
   // original type.
-  repl = mapping.lookupOrNull(value);
+  repl = lookupOrNull(value);
   if (repl.empty()) {
     // No replacement value is registered in the mapping. This means that the
     // value is dropped and no longer needed. (If the value were still needed,
@@ -1568,23 +1594,30 @@ Value ConversionPatternRewriterImpl::findOrBuildReplacementValue(
 
 void ConversionPatternRewriterImpl::notifyOperationInserted(
     Operation *op, OpBuilder::InsertPoint previous) {
+  // If no previous insertion point is provided, the op used to be detached.
+  bool wasDetached = !previous.isSet();
   LLVM_DEBUG({
-    logger.startLine() << "** Insert  : '" << op->getName() << "'(" << op
-                       << ")\n";
+    logger.startLine() << "** Insert  : '" << op->getName() << "' (" << op
+                       << ")";
+    if (wasDetached)
+      logger.getOStream() << " (was detached)";
+    logger.getOStream() << "\n";
   });
   assert(!wasOpReplaced(op->getParentOp()) &&
          "attempting to insert into a block within a replaced/erased op");
 
-  if (!previous.isSet()) {
-    // This is a newly created op.
+  if (wasDetached) {
+    // If the op was detached, it is most likely a newly created op.
+    // TODO: If the same op is inserted multiple times from a detached state,
+    // the rollback mechanism may erase the same op multiple times. This is a
+    // bug in the rollback-based dialect conversion driver.
     appendRewrite<CreateOperationRewrite>(op);
     patternNewOps.insert(op);
     return;
   }
-  Operation *prevOp = previous.getPoint() == previous.getBlock()->end()
-                          ? nullptr
-                          : &*previous.getPoint();
-  appendRewrite<MoveOperationRewrite>(op, previous.getBlock(), prevOp);
+
+  // The op was moved from one place to another.
+  appendRewrite<MoveOperationRewrite>(op, previous);
 }
 
 void ConversionPatternRewriterImpl::replaceOp(
@@ -1649,29 +1682,40 @@ void ConversionPatternRewriterImpl::eraseBlock(Block *block) {
 
 void ConversionPatternRewriterImpl::notifyBlockInserted(
     Block *block, Region *previous, Region::iterator previousIt) {
-  assert(!wasOpReplaced(block->getParentOp()) &&
-         "attempting to insert into a region within a replaced/erased op");
+  // If no previous insertion point is provided, the block used to be detached.
+  bool wasDetached = !previous;
+  Operation *newParentOp = block->getParentOp();
   LLVM_DEBUG(
       {
-        Operation *parent = block->getParentOp();
+        Operation *parent = newParentOp;
         if (parent) {
           logger.startLine() << "** Insert Block into : '" << parent->getName()
-                             << "'(" << parent << ")\n";
+                             << "' (" << parent << ")";
         } else {
           logger.startLine()
-              << "** Insert Block into detached Region (nullptr parent op)'\n";
+              << "** Insert Block into detached Region (nullptr parent op)";
         }
+        if (wasDetached)
+          logger.getOStream() << " (was detached)";
+        logger.getOStream() << "\n";
       });
+  assert(!wasOpReplaced(newParentOp) &&
+         "attempting to insert into a region within a replaced/erased op");
+  (void)newParentOp;
 
   patternInsertedBlocks.insert(block);
 
-  if (!previous) {
-    // This is a newly created block.
+  if (wasDetached) {
+    // If the block was detached, it is most likely a newly created block.
+    // TODO: If the same block is inserted multiple times from a detached state,
+    // the rollback mechanism may erase the same block multiple times. This is a
+    // bug in the rollback-based dialect conversion driver.
     appendRewrite<CreateBlockRewrite>(block);
     return;
   }
-  Block *prevBlock = previousIt == previous->end() ? nullptr : &*previousIt;
-  appendRewrite<MoveBlockRewrite>(block, previous, prevBlock);
+
+  // The block was moved from one place to another.
+  appendRewrite<MoveBlockRewrite>(block, previous, previousIt);
 }
 
 void ConversionPatternRewriterImpl::inlineBlockBefore(Block *source,
@@ -1716,6 +1760,12 @@ void ConversionPatternRewriter::replaceOp(Operation *op, ValueRange newValues) {
     impl->logger.startLine()
         << "** Replace : '" << op->getName() << "'(" << op << ")\n";
   });
+
+  // If the current insertion point is before the erased operation, we adjust
+  // the insertion point to be after the operation.
+  if (getInsertionPoint() == op->getIterator())
+    setInsertionPointAfter(op);
+
   SmallVector<SmallVector<Value>> newVals =
       llvm::map_to_vector(newValues, [](Value v) -> SmallVector<Value> {
         return v ? SmallVector<Value>{v} : SmallVector<Value>();
@@ -1731,6 +1781,12 @@ void ConversionPatternRewriter::replaceOpWithMultiple(
     impl->logger.startLine()
         << "** Replace : '" << op->getName() << "'(" << op << ")\n";
   });
+
+  // If the current insertion point is before the erased operation, we adjust
+  // the insertion point to be after the operation.
+  if (getInsertionPoint() == op->getIterator())
+    setInsertionPointAfter(op);
+
   impl->replaceOp(op, std::move(newValues));
 }
 
@@ -1739,6 +1795,12 @@ void ConversionPatternRewriter::eraseOp(Operation *op) {
     impl->logger.startLine()
         << "** Erase   : '" << op->getName() << "'(" << op << ")\n";
   });
+
+  // If the current insertion point is before the erased operation, we adjust
+  // the insertion point to be after the operation.
+  if (getInsertionPoint() == op->getIterator())
+    setInsertionPointAfter(op);
+
   SmallVector<SmallVector<Value>> nullRepls(op->getNumResults(), {});
   impl->replaceOp(op, std::move(nullRepls));
 }
@@ -1845,6 +1907,11 @@ void ConversionPatternRewriter::inlineBlockBefore(Block *source, Block *dest,
       moveOpBefore(&source->front(), dest, before);
   }
 
+  // If the current insertion point is within the source block, adjust the
+  // insertion point to the destination block.
+  if (getInsertionBlock() == source)
+    setInsertionPoint(dest, getInsertionPoint());
+
   // Erase the source block.
   eraseBlock(source);
 }
@@ -1976,6 +2043,7 @@ private:
   /// Legalize the resultant IR after successfully applying the given pattern.
   LogicalResult legalizePatternResult(Operation *op, const Pattern &pattern,
                                       ConversionPatternRewriter &rewriter,
+                                      const RewriterState &curState,
                                       const SetVector<Operation *> &newOps,
                                       const SetVector<Operation *> &modifiedOps,
                                       const SetVector<Block *> &insertedBlocks);
@@ -2092,8 +2160,9 @@ OperationLegalizer::legalize(Operation *op,
 
     // If the operation has no regions, just print it here.
     if (!isIgnored && op->getNumRegions() == 0) {
-      op->print(logger.startLine(), OpPrintingFlags().printGenericOpForm());
-      logger.getOStream() << "\n\n";
+      logger.startLine() << OpWithFlags(op,
+                                        OpPrintingFlags().printGenericOpForm())
+                         << "\n";
     }
   });
 
@@ -2172,23 +2241,39 @@ OperationLegalizer::legalizeWithFold(Operation *op,
     rewriterImpl.logger.startLine() << "* Fold {\n";
     rewriterImpl.logger.indent();
   });
-  (void)rewriterImpl;
+
+  // Clear pattern state, so that the next pattern application starts with a
+  // clean slate. (The op/block sets are populated by listener notifications.)
+  auto cleanup = llvm::make_scope_exit([&]() {
+    rewriterImpl.patternNewOps.clear();
+    rewriterImpl.patternModifiedOps.clear();
+    rewriterImpl.patternInsertedBlocks.clear();
+  });
+
+  // Upon failure, undo all changes made by the folder.
+  RewriterState curState = rewriterImpl.getCurrentState();
 
   // Try to fold the operation.
   StringRef opName = op->getName().getStringRef();
   SmallVector<Value, 2> replacementValues;
   SmallVector<Operation *, 2> newOps;
   rewriter.setInsertionPoint(op);
+  rewriter.startOpModification(op);
   if (failed(rewriter.tryFold(op, replacementValues, &newOps))) {
     LLVM_DEBUG(logFailure(rewriterImpl.logger, "unable to fold"));
+    rewriter.cancelOpModification(op);
     return failure();
   }
+  rewriter.finalizeOpModification(op);
 
   // An empty list of replacement values indicates that the fold was in-place.
   // As the operation changed, a new legalization needs to be attempted.
   if (replacementValues.empty())
     return legalize(op, rewriter);
 
+  // Insert a replacement for 'op' with the folded replacement values.
+  rewriter.replaceOp(op, replacementValues);
+
   // Recursively legalize any new constant operations.
   for (Operation *newOp : newOps) {
     if (failed(legalize(newOp, rewriter))) {
@@ -2201,16 +2286,12 @@ OperationLegalizer::legalizeWithFold(Operation *op,
             "op '" + opName +
             "' folder rollback of IR modifications requested");
       }
-      // Legalization failed: erase all materialized constants.
-      for (Operation *op : newOps)
-        rewriter.eraseOp(op);
+      rewriterImpl.resetState(
+          curState, std::string(op->getName().getStringRef()) + " folder");
       return failure();
     }
   }
 
-  // Insert a replacement for 'op' with the folded replacement values.
-  rewriter.replaceOp(op, replacementValues);
-
   LLVM_DEBUG(logSuccess(rewriterImpl.logger, ""));
   return success();
 }
@@ -2220,6 +2301,32 @@ OperationLegalizer::legalizeWithPattern(Operation *op,
                                         ConversionPatternRewriter &rewriter) {
   auto &rewriterImpl = rewriter.getImpl();
 
+#if MLIR_ENABLE_EXPENSIVE_PATTERN_API_CHECKS
+  Operation *checkOp;
+  std::optional<OperationFingerPrint> topLevelFingerPrint;
+  if (!rewriterImpl.config.allowPatternRollback) {
+    // The op may be getting erased, so we have to check the parent op.
+    // (In rare cases, a pattern may even erase the parent op, which will cause
+    // a crash here. Expensive checks are "best effort".) Skip the check if the
+    // op does not have a parent op.
+    if ((checkOp = op->getParentOp())) {
+      if (!op->getContext()->isMultithreadingEnabled()) {
+        topLevelFingerPrint = OperationFingerPrint(checkOp);
+      } else {
+        // Another thread may be modifying a sibling operation. Therefore, the
+        // fingerprinting mechanism of the parent op works only in
+        // single-threaded mode.
+        LLVM_DEBUG({
+          rewriterImpl.logger.startLine()
+              << "WARNING: Multi-threadeding is enabled. Some dialect "
+                 "conversion expensive checks are skipped in multithreading "
+                 "mode!\n";
+        });
+      }
+    }
+  }
+#endif // MLIR_ENABLE_EXPENSIVE_PATTERN_API_CHECKS
+
   // Functor that returns if the given pattern may be applied.
   auto canApply = [&](const Pattern &pattern) {
     bool canApply = canApplyPattern(op, pattern, rewriter);
@@ -2232,6 +2339,17 @@ OperationLegalizer::legalizeWithPattern(Operation *op,
   RewriterState curState = rewriterImpl.getCurrentState();
   auto onFailure = [&](const Pattern &pattern) {
     assert(rewriterImpl.pendingRootUpdates.empty() && "dangling root updates");
+#if MLIR_ENABLE_EXPENSIVE_PATTERN_API_CHECKS
+    if (!rewriterImpl.config.allowPatternRollback) {
+      // Returning "failure" after modifying IR is not allowed.
+      if (checkOp) {
+        OperationFingerPrint fingerPrintAfterPattern(checkOp);
+        if (fingerPrintAfterPattern != *topLevelFingerPrint)
+          llvm::report_fatal_error("pattern '" + pattern.getDebugName() +
+                                   "' returned failure but IR did change");
+      }
+    }
+#endif // MLIR_ENABLE_EXPENSIVE_PATTERN_API_CHECKS
     rewriterImpl.patternNewOps.clear();
     rewriterImpl.patternModifiedOps.clear();
     rewriterImpl.patternInsertedBlocks.clear();
@@ -2260,7 +2378,7 @@ OperationLegalizer::legalizeWithPattern(Operation *op,
         moveAndReset(rewriterImpl.patternModifiedOps);
     SetVector<Block *> insertedBlocks =
         moveAndReset(rewriterImpl.patternInsertedBlocks);
-    auto result = legalizePatternResult(op, pattern, rewriter, newOps,
+    auto result = legalizePatternResult(op, pattern, rewriter, curState, newOps,
                                         modifiedOps, insertedBlocks);
     appliedPatterns.erase(&pattern);
     if (failed(result)) {
@@ -2303,7 +2421,7 @@ bool OperationLegalizer::canApplyPattern(Operation *op, const Pattern &pattern,
 
 LogicalResult OperationLegalizer::legalizePatternResult(
     Operation *op, const Pattern &pattern, ConversionPatternRewriter &rewriter,
-    const SetVector<Operation *> &newOps,
+    const RewriterState &curState, const SetVector<Operation *> &newOps,
     const SetVector<Operation *> &modifiedOps,
     const SetVector<Block *> &insertedBlocks) {
   auto &impl = rewriter.getImpl();
@@ -2319,7 +2437,8 @@ LogicalResult OperationLegalizer::legalizePatternResult(
     return hasRewrite<ModifyOperationRewrite>(newRewrites, op);
   };
   if (!replacedRoot() && !updatedRootInPlace())
-    llvm::report_fatal_error("expected pattern to replace the root operation");
+    llvm::report_fatal_error(
+        "expected pattern to replace the root operation or modify it in place");
 #endif // MLIR_ENABLE_EXPENSIVE_PATTERN_API_CHECKS
 
   // Legalize each of the actions registered during application.
diff --git a/mlir/lib/Transforms/Utils/Inliner.cpp b/mlir/lib/Transforms/Utils/Inliner.cpp
index b639e87f..26c965c 100644
--- a/mlir/lib/Transforms/Utils/Inliner.cpp
+++ b/mlir/lib/Transforms/Utils/Inliner.cpp
@@ -21,7 +21,7 @@
 #include "mlir/Transforms/InliningUtils.h"
 #include "llvm/ADT/SCCIterator.h"
 #include "llvm/ADT/STLExtras.h"
-#include "llvm/Support/Debug.h"
+#include "llvm/Support/DebugLog.h"
 
 #define DEBUG_TYPE "inlining"
 
@@ -348,13 +348,11 @@ static void collectCallOps(iterator_range<Region::iterator> blocks,
 // InlinerInterfaceImpl
 //===----------------------------------------------------------------------===//
 
-#ifndef NDEBUG
 static std::string getNodeName(CallOpInterface op) {
   if (llvm::dyn_cast_if_present<SymbolRefAttr>(op.getCallableForCallee()))
     return debugString(op);
   return "_unnamed_callee_";
 }
-#endif
 
 /// Return true if the specified `inlineHistoryID`  indicates an inline history
 /// that already includes `node`.
@@ -614,10 +612,10 @@ Inliner::Impl::inlineCallsInSCC(InlinerInterfaceImpl &inlinerIface,
   std::vector<InlineHistoryT> callHistory(calls.size(), InlineHistoryT{});
 
   LLVM_DEBUG({
-    llvm::dbgs() << "* Inliner: Initial calls in SCC are: {\n";
+    LDBG() << "* Inliner: Initial calls in SCC are: {";
     for (unsigned i = 0, e = calls.size(); i < e; ++i)
-      llvm::dbgs() << "  " << i << ". " << calls[i].call << ",\n";
-    llvm::dbgs() << "}\n";
+      LDBG() << "  " << i << ". " << calls[i].call << ",";
+    LDBG() << "}";
   });
 
   // Try to inline each of the call operations. Don't cache the end iterator
@@ -635,9 +633,9 @@ Inliner::Impl::inlineCallsInSCC(InlinerInterfaceImpl &inlinerIface,
     CallOpInterface call = it.call;
     LLVM_DEBUG({
       if (doInline)
-        llvm::dbgs() << "* Inlining call: " << i << ". " << call << "\n";
+        LDBG() << "* Inlining call: " << i << ". " << call;
       else
-        llvm::dbgs() << "* Not inlining call: " << i << ". " << call << "\n";
+        LDBG() << "* Not inlining call: " << i << ". " << call;
     });
     if (!doInline)
       continue;
@@ -654,7 +652,7 @@ Inliner::Impl::inlineCallsInSCC(InlinerInterfaceImpl &inlinerIface,
                    cast<CallableOpInterface>(targetRegion->getParentOp()),
                    targetRegion, /*shouldCloneInlinedRegion=*/!inlineInPlace);
     if (failed(inlineResult)) {
-      LLVM_DEBUG(llvm::dbgs() << "** Failed to inline\n");
+      LDBG() << "** Failed to inline";
       continue;
     }
     inlinedAnyCalls = true;
@@ -667,19 +665,16 @@ Inliner::Impl::inlineCallsInSCC(InlinerInterfaceImpl &inlinerIface,
     auto historyToString = [](InlineHistoryT h) {
       return h.has_value() ? std::to_string(*h) : "root";
     };
-    (void)historyToString;
-    LLVM_DEBUG(llvm::dbgs()
-               << "* new inlineHistory entry: " << newInlineHistoryID << ". ["
-               << getNodeName(call) << ", " << historyToString(inlineHistoryID)
-               << "]\n");
+    LDBG() << "* new inlineHistory entry: " << newInlineHistoryID << ". ["
+           << getNodeName(call) << ", " << historyToString(inlineHistoryID)
+           << "]";
 
     for (unsigned k = prevSize; k != calls.size(); ++k) {
       callHistory.push_back(newInlineHistoryID);
-      LLVM_DEBUG(llvm::dbgs() << "* new call " << k << " {" << calls[i].call
-                              << "}\n   with historyID = " << newInlineHistoryID
-                              << ", added due to inlining of\n  call {" << call
-                              << "}\n with historyID = "
-                              << historyToString(inlineHistoryID) << "\n");
+      LDBG() << "* new call " << k << " {" << calls[k].call
+             << "}\n   with historyID = " << newInlineHistoryID
+             << ", added due to inlining of\n  call {" << call
+             << "}\n with historyID = " << historyToString(inlineHistoryID);
     }
 
     // If the inlining was successful, Merge the new uses into the source node.
diff --git a/mlir/test/Conversion/MeshToMPI/convert-mesh-to-mpi.mlir b/mlir/test/Conversion/ShardToMPI/convert-shard-to-mpi.mlir
index d54d003..5e20b5a 100644
--- a/mlir/test/Conversion/MeshToMPI/convert-mesh-to-mpi.mlir
+++ b/mlir/test/Conversion/ShardToMPI/convert-shard-to-mpi.mlir
@@ -1,14 +1,14 @@
-// RUN: mlir-opt %s -convert-mesh-to-mpi -canonicalize -split-input-file | FileCheck %s
+// RUN: mlir-opt %s -convert-shard-to-mpi -canonicalize -split-input-file | FileCheck %s
 
 // -----
-// CHECK: mesh.mesh @mesh0
-mesh.mesh @mesh0(shape = 3x4x5)
+// CHECK: shard.grid @grid0
+shard.grid @grid0(shape = 3x4x5)
 func.func @process_multi_index() -> (index, index, index) {
   // CHECK: mpi.comm_rank
   // CHECK-DAG: %[[v4:.*]] = arith.remsi
   // CHECK-DAG: %[[v0:.*]] = arith.remsi
   // CHECK-DAG: %[[v1:.*]] = arith.remsi
-  %0:3 = mesh.process_multi_index on @mesh0 axes = [] : index, index, index
+  %0:3 = shard.process_multi_index on @grid0 axes = [] : index, index, index
   // CHECK: return %[[v1]], %[[v0]], %[[v4]] : index, index, index
   return %0#0, %0#1, %0#2 : index, index, index
 }
@@ -17,7 +17,7 @@ func.func @process_multi_index() -> (index, index, index) {
 func.func @process_linear_index() -> index {
   // CHECK: %[[RES:.*]], %[[rank:.*]] = mpi.comm_rank
   // CHECK: %[[cast:.*]] = arith.index_cast %[[rank]] : i32 to index
-  %0 = mesh.process_linear_index on @mesh0 : index
+  %0 = shard.process_linear_index on @grid0 : index
   // CHECK: return %[[cast]] : index
   return %0 : index
 }
@@ -29,7 +29,7 @@ func.func @neighbors_dim0(%arg0 : tensor<120x120x120xi8>) -> (index, index) {
   %c4 = arith.constant 4 : index
   // CHECK-DAG: [[up:%.*]] = arith.constant 44 : index
   // CHECK-DAG: [[down:%.*]] = arith.constant 4 : index
-  %idx:2 = mesh.neighbors_linear_indices on @mesh0[%c1, %c0, %c4] split_axes = [0] : index, index
+  %idx:2 = shard.neighbors_linear_indices on @grid0[%c1, %c0, %c4] split_axes = [0] : index, index
   // CHECK: return [[down]], [[up]] : index, index
   return %idx#0, %idx#1 : index, index
 }
@@ -41,7 +41,7 @@ func.func @neighbors_dim1(%arg0 : tensor<120x120x120xi8>) -> (index, index) {
   %c4 = arith.constant 4 : index
   // CHECK-DAG: [[up:%.*]] = arith.constant 29 : index
   // CHECK-DAG: [[down:%.*]] = arith.constant -1 : index
-  %idx:2 = mesh.neighbors_linear_indices on @mesh0[%c1, %c0, %c4] split_axes = [1] : index, index
+  %idx:2 = shard.neighbors_linear_indices on @grid0[%c1, %c0, %c4] split_axes = [1] : index, index
   // CHECK: return [[down]], [[up]] : index, index
   return %idx#0, %idx#1 : index, index
 }
@@ -53,20 +53,20 @@ func.func @neighbors_dim2(%arg0 : tensor<120x120x120xi8>) -> (index, index) {
   %c4 = arith.constant 4 : index
   // CHECK-DAG: [[up:%.*]] = arith.constant -1 : index
   // CHECK-DAG: [[down:%.*]] = arith.constant 23 : index
-  %idx:2 = mesh.neighbors_linear_indices on @mesh0[%c1, %c0, %c4] split_axes = [2] : index, index
+  %idx:2 = shard.neighbors_linear_indices on @grid0[%c1, %c0, %c4] split_axes = [2] : index, index
   // CHECK: return [[down]], [[up]] : index, index
   return %idx#0, %idx#1 : index, index
 }
 
 // -----
-// CHECK: mesh.mesh @mesh0
+// CHECK: shard.grid @grid0
 module attributes { mpi.dlti = #dlti.map<"MPI:comm_world_rank" = 24> } {
-  mesh.mesh @mesh0(shape = 3x4x5)
+  shard.grid @grid0(shape = 3x4x5)
   func.func @process_multi_index() -> (index, index, index) {
     // CHECK-DAG: %[[c4:.*]] = arith.constant 4 : index
     // CHECK-DAG: %[[c0:.*]] = arith.constant 0 : index
     // CHECK-DAG: %[[c1:.*]] = arith.constant 1 : index
-    %0:3 = mesh.process_multi_index on @mesh0 axes = [] : index, index, index
+    %0:3 = shard.process_multi_index on @grid0 axes = [] : index, index, index
     // CHECK: return %[[c1]], %[[c0]], %[[c4]] : index, index, index
     return %0#0, %0#1, %0#2 : index, index, index
   }
@@ -74,7 +74,7 @@ module attributes { mpi.dlti = #dlti.map<"MPI:comm_world_rank" = 24> } {
   // CHECK-LABEL: func @process_linear_index
   func.func @process_linear_index() -> index {
     // CHECK: %[[c24:.*]] = arith.constant 24 : index
-    %0 = mesh.process_linear_index on @mesh0 : index
+    %0 = shard.process_linear_index on @grid0 : index
     // CHECK: return %[[c24]] : index
     return %0 : index
   }
@@ -82,7 +82,7 @@ module attributes { mpi.dlti = #dlti.map<"MPI:comm_world_rank" = 24> } {
 
 // -----
 module attributes { mpi.dlti = #dlti.map<"MPI:comm_world_rank" = 7> } {
-  mesh.mesh @mesh0(shape = 3x4x5)
+  shard.grid @grid0(shape = 3x4x5)
   // CHECK-LABEL: func.func @allreduce_tensor(
   func.func @allreduce_tensor(
     // CHECK-SAME: [[varg0:%.*]]: tensor<3x4xf32>
@@ -97,7 +97,7 @@ module attributes { mpi.dlti = #dlti.map<"MPI:comm_world_rank" = 7> } {
     // CHECK: [[vcollapse_shape:%.*]] = memref.collapse_shape [[valloc]] {{\[\[}}0, 1]] : memref<3x4xf32> into memref<12xf32>
     // CHECK: mpi.allreduce([[vcollapse_shape]], [[vcollapse_shape]], MPI_MAX, [[vnewcomm]]) : memref<12xf32>, memref<12xf32>
     // CHECK: [[v2:%.*]] = bufferization.to_tensor [[valloc]] restrict : memref<3x4xf32> to tensor<3x4xf32>
-    %0 = mesh.all_reduce %arg0 on @mesh0 mesh_axes = [0, 1] reduction = max : tensor<3x4xf32> -> tensor<3x4xf32>
+    %0 = shard.all_reduce %arg0 on @grid0 grid_axes = [0, 1] reduction = max : tensor<3x4xf32> -> tensor<3x4xf32>
     // CHECK: return [[v2]] : tensor<3x4xf32>
     return %0 : tensor<3x4xf32>
   }
@@ -114,7 +114,7 @@ module attributes { mpi.dlti = #dlti.map<"MPI:comm_world_rank" = 7> } {
     // CHECK: [[vnewcomm:%.*]] = mpi.comm_split([[v0]], [[vc2_i32]], [[vc4_i32]]) : !mpi.comm
     // CHECK: [[vcollapse_shape:%.*]] = memref.collapse_shape [[valloc]] {{\[\[}}0, 1]] : memref<3x4xf32> into memref<12xf32>
     // CHECK: mpi.allreduce([[vcollapse_shape]], [[vcollapse_shape]], MPI_MAX, [[vnewcomm]]) : memref<12xf32>, memref<12xf32>
-    %0 = mesh.all_reduce %arg0 on @mesh0 mesh_axes = [0, 1] reduction = max : memref<3x4xf32> -> memref<3x4xf32>
+    %0 = shard.all_reduce %arg0 on @grid0 grid_axes = [0, 1] reduction = max : memref<3x4xf32> -> memref<3x4xf32>
     // CHECK: return [[valloc]] : memref<3x4xf32>
     return %0 : memref<3x4xf32>
   }
@@ -131,14 +131,14 @@ module attributes { mpi.dlti = #dlti.map<"MPI:comm_world_rank" = 7> } {
     // CHECK: [[vnewcomm:%.*]] = mpi.comm_split([[v0]], [[vc2_i32]], [[vc4_i32]]) : !mpi.comm
     // CHECK: [[vcollapse_shape:%.*]] = memref.collapse_shape [[valloc]] {{\[\[}}0, 1]] : memref<3x4xf64> into memref<12xf64>
     // CHECK: mpi.allreduce([[vcollapse_shape]], [[vcollapse_shape]], MPI_MAX, [[vnewcomm]]) : memref<12xf64>, memref<12xf64>
-    %0 = mesh.all_reduce %arg0 on @mesh0 mesh_axes = [0, 1] reduction = max : memref<3x4xf32> -> memref<3x4xf64>
+    %0 = shard.all_reduce %arg0 on @grid0 grid_axes = [0, 1] reduction = max : memref<3x4xf32> -> memref<3x4xf64>
     // CHECK: return [[valloc]] : memref<3x4xf64>
     return %0 : memref<3x4xf64>
   }
 }
 
 // -----
-mesh.mesh @mesh0(shape = 3x4x5)
+shard.grid @grid0(shape = 3x4x5)
 // CHECK-LABEL: func @update_halo_1d_first
 func.func @update_halo_1d_first(
   // CHECK-SAME: [[arg0:%.*]]: memref<120x120x120xi8>
@@ -155,14 +155,14 @@ func.func @update_halo_1d_first(
   // CHECK: mpi.recv(
   // CHECK-SAME: : memref<3x120x120xi8>, i32, i32
   // CHECK: memref.subview [[arg0]][117, 0, 0] [3, 120, 120] [1, 1, 1] : memref<120x120x120xi8> to memref<3x120x120xi8
-  %res = mesh.update_halo %arg0 on @mesh0 split_axes = [[0]] halo_sizes = [2, 3] : memref<120x120x120xi8>
+  %res = shard.update_halo %arg0 on @grid0 split_axes = [[0]] halo_sizes = [2, 3] : memref<120x120x120xi8>
   // CHECK: return [[res:%.*]] : memref<120x120x120xi8>
   return %res : memref<120x120x120xi8>
 }
 
 // -----
 module attributes { mpi.dlti = #dlti.map<"MPI:comm_world_rank" = 1> } {
-  mesh.mesh @mesh0(shape = 4)
+  shard.grid @grid0(shape = 4)
   // CHECK-LABEL: func @update_halo_1d_with_zero
   func.func @update_halo_1d_with_zero (
     // CHECK-SAME: [[varg0:%.*]]: memref<120x120x120xi8>
@@ -179,7 +179,7 @@ module attributes { mpi.dlti = #dlti.map<"MPI:comm_world_rank" = 1> } {
     // CHECK: [[vsubview_0:%.*]] = memref.subview [[varg0]][0, 0, 0] [2, 120, 120] [1, 1, 1] : memref<120x120x120xi8> to memref<2x120x120xi8, strided<[14400, 120, 1]>>
     // CHECK: memref.copy [[valloc]], [[vsubview_0]] : memref<2x120x120xi8> to memref<2x120x120xi8, strided<[14400, 120, 1]>>
     // CHECK: memref.dealloc [[valloc]] : memref<2x120x120xi8>
-    %res = mesh.update_halo %arg0 on @mesh0 split_axes = [[0]] halo_sizes = [2, 0] : memref<120x120x120xi8>
+    %res = shard.update_halo %arg0 on @grid0 split_axes = [[0]] halo_sizes = [2, 0] : memref<120x120x120xi8>
     // CHECK: return [[varg0]] : memref<120x120x120xi8>
     return %res : memref<120x120x120xi8>
   }
@@ -187,7 +187,7 @@ module attributes { mpi.dlti = #dlti.map<"MPI:comm_world_rank" = 1> } {
 
 // -----
 module attributes { mpi.dlti = #dlti.map<"MPI:comm_world_rank" = 24> } {
-  mesh.mesh @mesh0(shape = 3x4x5)
+  shard.grid @grid0(shape = 3x4x5)
   // CHECK-LABEL: func @update_halo_3d
   func.func @update_halo_3d(
     // CHECK-SAME: [[varg0:%.*]]: memref<120x120x120xi8>
@@ -236,7 +236,7 @@ module attributes { mpi.dlti = #dlti.map<"MPI:comm_world_rank" = 24> } {
     // CHECK: memref.copy [[vsubview_11]], [[valloc_10]] : memref<2x120x120xi8, strided<[14400, 120, 1], offset: 14400>> to memref<2x120x120xi8>
     // CHECK: mpi.send([[valloc_10]], [[vc91_i32]], [[vc23_i32]], [[v2]]) : memref<2x120x120xi8>, i32, i32
     // CHECK: memref.dealloc [[valloc_10]] : memref<2x120x120xi8>
-    %res = mesh.update_halo %arg0 on @mesh0 split_axes = [[2], [1], [0]] halo_sizes = [1, 2, 3, 4, 5, 6] : memref<120x120x120xi8>
+    %res = shard.update_halo %arg0 on @grid0 split_axes = [[2], [1], [0]] halo_sizes = [1, 2, 3, 4, 5, 6] : memref<120x120x120xi8>
     // CHECK: return [[varg0]] : memref<120x120x120xi8>
     return %res : memref<120x120x120xi8>
   }
@@ -291,18 +291,18 @@ module attributes { mpi.dlti = #dlti.map<"MPI:comm_world_rank" = 24> } {
     // CHECK: mpi.send([[valloc_10]], [[vc91_i32]], [[vc23_i32]], [[v3]]) : memref<2x120x120xi8>, i32, i32
     // CHECK: memref.dealloc [[valloc_10]] : memref<2x120x120xi8>
     // CHECK: [[v4:%.*]] = bufferization.to_tensor [[v0]] restrict writable : memref<120x120x120xi8> to tensor<120x120x120xi8>
-    %res = mesh.update_halo %arg0 on @mesh0 split_axes = [[2], [1], [0]] halo_sizes = [1, 2, 3, 4, 5, 6] : tensor<120x120x120xi8>
+    %res = shard.update_halo %arg0 on @grid0 split_axes = [[2], [1], [0]] halo_sizes = [1, 2, 3, 4, 5, 6] : tensor<120x120x120xi8>
     // CHECK: return [[v4]] : tensor<120x120x120xi8>
     return %res : tensor<120x120x120xi8>
   }
 }
 
 // -----
-mesh.mesh @mesh0(shape = 2x2x4)
+shard.grid @grid0(shape = 2x2x4)
 // CHECK-LABEL: func.func @return_sharding(
 // CHECK-SAME: [[varg0:%.*]]: tensor<2x4xf32>) -> (tensor<2x4xf32>, tensor<?x?xi16>, tensor<?x?xi64>, tensor<?x?xi64>) {
-func.func @return_sharding(%arg0: tensor<2x4xf32>) -> (tensor<2x4xf32>, !mesh.sharding) {
-  %sharding = mesh.sharding @mesh0 split_axes = [[0, 1], [2]] : !mesh.sharding
+func.func @return_sharding(%arg0: tensor<2x4xf32>) -> (tensor<2x4xf32>, !shard.sharding) {
+  %sharding = shard.sharding @grid0 split_axes = [[0, 1], [2]] : !shard.sharding
   // CHECK: [[vcst:%.*]] = arith.constant dense<2> : tensor<1xi16>
   // CHECK: [[vcst_0:%.*]] = arith.constant dense<[0, 1]> : tensor<2xi16>
   // CHECK: [[vcm1_i16:%.*]] = arith.constant -1 : i16
@@ -316,13 +316,13 @@ func.func @return_sharding(%arg0: tensor<2x4xf32>) -> (tensor<2x4xf32>, !mesh.sh
   // CHECK: [[vcast_2:%.*]] = tensor.cast [[v2]] : tensor<0x0xi64> to tensor<?x?xi64>
   // CHECK: [[vcast_3:%.*]] = tensor.cast [[v3]] : tensor<0x0xi64> to tensor<?x?xi64>
   // CHECK: return [[varg0]], [[vcast]], [[vcast_2]], [[vcast_3]] : tensor<2x4xf32>, tensor<?x?xi16>, tensor<?x?xi64>, tensor<?x?xi64>
-  return %arg0, %sharding : tensor<2x4xf32>, !mesh.sharding
+  return %arg0, %sharding : tensor<2x4xf32>, !shard.sharding
 }
 
 // CHECK-LABEL: func.func @return_sharding_halos(
 // CHECK-SAME: [[varg0:%.*]]: tensor<6x8xf32>) -> (tensor<6x8xf32>, tensor<?x?xi16>, tensor<?x?xi64>, tensor<?x?xi64>) {
-func.func @return_sharding_halos(%arg0: tensor<6x8xf32>) -> (tensor<6x8xf32>, !mesh.sharding) {
-  %sharding = mesh.sharding @mesh0 split_axes = [[0, 1], [2]] halo_sizes = [0, 4, 3, 1] : !mesh.sharding
+func.func @return_sharding_halos(%arg0: tensor<6x8xf32>) -> (tensor<6x8xf32>, !shard.sharding) {
+  %sharding = shard.sharding @grid0 split_axes = [[0, 1], [2]] halo_sizes = [0, 4, 3, 1] : !shard.sharding
   // CHECK: [[vcst:%.*]] = arith.constant dense<{{\[\[}}0, 4], [3, 1]]> : tensor<2x2xi64>
   // CHECK: [[vcst_0:%.*]] = arith.constant dense<2> : tensor<1xi16>
   // CHECK: [[vcst_1:%.*]] = arith.constant dense<[0, 1]> : tensor<2xi16>
@@ -336,13 +336,13 @@ func.func @return_sharding_halos(%arg0: tensor<6x8xf32>) -> (tensor<6x8xf32>, !m
   // CHECK: [[vcast_3:%.*]] = tensor.cast [[vcst]] : tensor<2x2xi64> to tensor<?x?xi64>
   // CHECK: [[vcast_4:%.*]] = tensor.cast [[v2]] : tensor<0x0xi64> to tensor<?x?xi64>
   // CHECK: return [[varg0]], [[vcast]], [[vcast_3]], [[vcast_4]] : tensor<6x8xf32>, tensor<?x?xi16>, tensor<?x?xi64>, tensor<?x?xi64>
-  return %arg0, %sharding : tensor<6x8xf32>, !mesh.sharding
+  return %arg0, %sharding : tensor<6x8xf32>, !shard.sharding
 }
 
 // CHECK-LABEL: func.func @return_sharding_offs(
 // CHECK-SAME: [[varg0:%.*]]: tensor<?x?xf32>) -> (tensor<?x?xf32>, tensor<?x?xi16>, tensor<?x?xi64>, tensor<?x?xi64>) {
-func.func @return_sharding_offs(%arg0: tensor<?x?xf32>) -> (tensor<?x?xf32>, !mesh.sharding) {
-  %sharding = mesh.sharding @mesh0 split_axes = [[0, 1], [2]] sharded_dims_offsets = [0, 3, 5, 7, 8, 0, 0, 5, 10, 16] : !mesh.sharding
+func.func @return_sharding_offs(%arg0: tensor<?x?xf32>) -> (tensor<?x?xf32>, !shard.sharding) {
+  %sharding = shard.sharding @grid0 split_axes = [[0, 1], [2]] sharded_dims_offsets = [0, 3, 5, 7, 8, 0, 0, 5, 10, 16] : !shard.sharding
   // CHECK: [[vcst:%.*]] = arith.constant dense<[0, 0, 5, 10, 16]> : tensor<5xi64>
   // CHECK: [[vcst_0:%.*]] = arith.constant dense<[0, 3, 5, 7, 8]> : tensor<5xi64>
   // CHECK: [[vcm9223372036854775808_i64:%.*]] = arith.constant -9223372036854775808 : i64
@@ -362,5 +362,5 @@ func.func @return_sharding_offs(%arg0: tensor<?x?xf32>) -> (tensor<?x?xf32>, !me
   // CHECK: [[vcast_6:%.*]] = tensor.cast [[v2]] : tensor<0x0xi64> to tensor<?x?xi64>
   // CHECK: [[vcast_7:%.*]] = tensor.cast [[vinserted_slice_5]] : tensor<2x5xi64> to tensor<?x?xi64>
   // CHECK: return [[varg0]], [[vcast]], [[vcast_6]], [[vcast_7]] : tensor<?x?xf32>, tensor<?x?xi16>, tensor<?x?xi64>, tensor<?x?xi64>
-  return %arg0, %sharding : tensor<?x?xf32>, !mesh.sharding
+  return %arg0, %sharding : tensor<?x?xf32>, !shard.sharding
 }
diff --git a/mlir/test/Conversion/MeshToMPI/convert-shardshape-to-mpi.mlir b/mlir/test/Conversion/ShardToMPI/convert-shardshape-to-mpi.mlir
index 156bbfb..9729d2b 100644
--- a/mlir/test/Conversion/MeshToMPI/convert-shardshape-to-mpi.mlir
+++ b/mlir/test/Conversion/ShardToMPI/convert-shardshape-to-mpi.mlir
@@ -1,21 +1,21 @@
-// RUN: mlir-opt %s --convert-mesh-to-mpi -canonicalize | FileCheck %s
+// RUN: mlir-opt %s --convert-shard-to-mpi -canonicalize | FileCheck %s
 
 module attributes { mpi.dlti = #dlti.map<"MPI:comm_world_rank" = 24> } {
 
-  // CHECK: mesh.mesh @mesh0
-  mesh.mesh @mesh0(shape = 3x4x5)
+  // CHECK: shard.grid @grid0
+  shard.grid @grid0(shape = 3x4x5)
   
-  // Notice: comm_world_rank/linear index 24 is multiindex [1, 0, 4] in @mesh0
+  // Notice: comm_world_rank/linear index 24 is multiindex [1, 0, 4] in @grid0
 
   // all shards are equal
   // CHECK-LABEL: func.func @shard_shape_equal() -> (index, index, index) {
   func.func @shard_shape_equal() -> (index, index, index) {
-    %sharding = mesh.sharding @mesh0 split_axes = [[0], [1], [2]] : !mesh.sharding
-    %0:3 = mesh.process_multi_index on @mesh0 : index, index, index
+    %sharding = shard.sharding @grid0 split_axes = [[0], [1], [2]] : !shard.sharding
+    %0:3 = shard.process_multi_index on @grid0 : index, index, index
     %c9 = arith.constant 9 : index
     %c12 = arith.constant 12 : index
     // CHECK: [[vc3:%.*]] = arith.constant 3 : index
-    %1:3 = mesh.shard_shape dims = [%c9, %c12, 15] sharding = %sharding device = [%0#0, %0#1, %0#2] : index, index, index
+    %1:3 = shard.shard_shape dims = [%c9, %c12, 15] sharding = %sharding device = [%0#0, %0#1, %0#2] : index, index, index
     // CHECK: return [[vc3]], [[vc3]], [[vc3]] : index, index, index
     return %1#0, %1#1, %1#2 : index, index, index
   }
@@ -23,13 +23,13 @@ module attributes { mpi.dlti = #dlti.map<"MPI:comm_world_rank" = 24> } {
   // last shard in last dim gets an extra element
   // CHECK-LABEL: func.func @shard_shape_odd_1() -> (index, index, index) {
   func.func @shard_shape_odd_1() -> (index, index, index) {
-    %sharding = mesh.sharding @mesh0 split_axes = [[0], [1], [2]] : !mesh.sharding
-    %0:3 = mesh.process_multi_index on @mesh0 : index, index, index
+    %sharding = shard.sharding @grid0 split_axes = [[0], [1], [2]] : !shard.sharding
+    %0:3 = shard.process_multi_index on @grid0 : index, index, index
     %c9 = arith.constant 9 : index
     %c12 = arith.constant 12 : index
     // CHECK-DAG: [[vc3:%.*]] = arith.constant 3 : index
     // CHECK-DAG: [[vc4:%.*]] = arith.constant 4 : index
-    %1:3 = mesh.shard_shape dims = [%c9, %c12, 16] sharding = %sharding device = [%0#0, %0#1, %0#2] : index, index, index
+    %1:3 = shard.shard_shape dims = [%c9, %c12, 16] sharding = %sharding device = [%0#0, %0#1, %0#2] : index, index, index
     // CHECK: return [[vc3]], [[vc3]], [[vc4]] : index, index, index
     return %1#0, %1#1, %1#2 : index, index, index
   }
@@ -37,11 +37,11 @@ module attributes { mpi.dlti = #dlti.map<"MPI:comm_world_rank" = 24> } {
   // In the second dimension the shard sizes are now [3 4 4 4]
   // CHECK-LABEL: func.func @shard_shape_odd_2() -> (index, index, index) {
   func.func @shard_shape_odd_2() -> (index, index, index) {
-    %sharding = mesh.sharding @mesh0 split_axes = [[0], [1], [2]] : !mesh.sharding
-    %0:3 = mesh.process_multi_index on @mesh0 : index, index, index
+    %sharding = shard.sharding @grid0 split_axes = [[0], [1], [2]] : !shard.sharding
+    %0:3 = shard.process_multi_index on @grid0 : index, index, index
     %c9 = arith.constant 9 : index
     // CHECK: [[vc3:%.*]] = arith.constant 3 : index
-    %1:3 = mesh.shard_shape dims = [%c9, 15, 15] sharding = %sharding device = [%0#0, %0#1, %0#2] : index, index, index
+    %1:3 = shard.shard_shape dims = [%c9, 15, 15] sharding = %sharding device = [%0#0, %0#1, %0#2] : index, index, index
     // CHECK: return [[vc3]], [[vc3]], [[vc3]] : index, index, index
     return %1#0, %1#1, %1#2 : index, index, index
   }
@@ -49,11 +49,11 @@ module attributes { mpi.dlti = #dlti.map<"MPI:comm_world_rank" = 24> } {
   // In the first dimension the shard sizes are now [3 4 4]
   // CHECK-LABEL: func.func @shard_shape_odd_3() -> (index, index, index) {
   func.func @shard_shape_odd_3() -> (index, index, index) {
-    %sharding = mesh.sharding @mesh0 split_axes = [[0], [1], [2]] : !mesh.sharding
-    %0:3 = mesh.process_multi_index on @mesh0 : index, index, index
+    %sharding = shard.sharding @grid0 split_axes = [[0], [1], [2]] : !shard.sharding
+    %0:3 = shard.process_multi_index on @grid0 : index, index, index
     // CHECK-DAG: [[vc3:%.*]] = arith.constant 3 : index
     // CHECK-DAG: [[vc4:%.*]] = arith.constant 4 : index
-    %1:3 = mesh.shard_shape dims = [11, 12, 15] sharding = %sharding device = [%0#0, %0#1, %0#2] : index, index, index
+    %1:3 = shard.shard_shape dims = [11, 12, 15] sharding = %sharding device = [%0#0, %0#1, %0#2] : index, index, index
     // CHECK: return [[vc4]], [[vc3]], [[vc3]] : index, index, index
     return %1#0, %1#1, %1#2 : index, index, index
   }
@@ -61,14 +61,14 @@ module attributes { mpi.dlti = #dlti.map<"MPI:comm_world_rank" = 24> } {
   // extract from sharded_dims_offsets
   // CHECK-LABEL: func.func @shard_shape_sharded_dims_offs() -> (index, index, index) {
   func.func @shard_shape_sharded_dims_offs() -> (index, index, index) {
-    %sharding = mesh.sharding @mesh0 split_axes = [[0], [1], [2]]
-        sharded_dims_offsets = [0, 1, 4, 9, 0, 2, 6, 12, 12, 0, 3, 6, 9, 12, 15]: !mesh.sharding
-    %0:3 = mesh.process_multi_index on @mesh0 : index, index, index
+    %sharding = shard.sharding @grid0 split_axes = [[0], [1], [2]]
+        sharded_dims_offsets = [0, 1, 4, 9, 0, 2, 6, 12, 12, 0, 3, 6, 9, 12, 15]: !shard.sharding
+    %0:3 = shard.process_multi_index on @grid0 : index, index, index
     %c9 = arith.constant 9 : index
     %c12 = arith.constant 12 : index
     // CHECK: [[vc3:%.*]] = arith.constant 3 : index
     // CHECK: [[vc2:%.*]] = arith.constant 2 : index
-    %1:3 = mesh.shard_shape dims = [%c9, %c12, 15] sharding = %sharding device = [%0#0, %0#1, %0#2] : index, index, index
+    %1:3 = shard.shard_shape dims = [%c9, %c12, 15] sharding = %sharding device = [%0#0, %0#1, %0#2] : index, index, index
     // CHECK: return [[vc3]], [[vc2]], [[vc3]] : index, index, index
     return %1#0, %1#1, %1#2 : index, index, index
   }
diff --git a/mlir/test/Conversion/TosaToSCF/tosa-to-scf.mlir b/mlir/test/Conversion/TosaToSCF/tosa-to-scf.mlir
index fa7a91c..b6f2383 100644
--- a/mlir/test/Conversion/TosaToSCF/tosa-to-scf.mlir
+++ b/mlir/test/Conversion/TosaToSCF/tosa-to-scf.mlir
@@ -36,7 +36,7 @@ func.func @while_test(%arg0 : tensor<i32>) -> (tensor<i32>) {
 func.func @if_test(%arg0 : tensor<f32>, %arg1 : tensor<f32>, %arg2 : tensor<i1>) -> (tensor<f32>) {
   // CHECK: [[EX:%.+]] = tensor.extract [[ARG2]]
   // CHECK: [[IF:%.+]] = scf.if [[EX]] -> (tensor<f32>) {
-  %0 = tosa.cond_if %arg2 -> (tensor<f32>) {
+  %0 = tosa.cond_if %arg2 : tensor<i1> -> tensor<f32> {
 
   // CHECK:   scf.yield [[ARG0]]
     tosa.yield %arg0 : tensor<f32>
diff --git a/mlir/test/Dialect/Arith/mesh-spmdize.mlir b/mlir/test/Dialect/Arith/mesh-spmdize.mlir
deleted file mode 100644
index 6b55dd5..0000000
--- a/mlir/test/Dialect/Arith/mesh-spmdize.mlir
+++ /dev/null
@@ -1,17 +0,0 @@
-// RUN: mlir-opt \
-// RUN:   --pass-pipeline="builtin.module(func.func(mesh-spmdization))" \
-// RUN:   %s | FileCheck %s
-
-mesh.mesh @mesh4x4(shape = 4x4)
-
-// CHECK-LABEL: func @test_spmdize_constant
-// CHECK-NEXT: [[vcst:%.*]] = arith.constant dense<0.000000e+00> :
-// tensor<256x1024xf32> CHECK-NEXT: [[vc434_i32:%.*]] = arith.constant 434 :
-// i32 CHECK-NEXT: return [[vcst]] : tensor<256x1024xf32>
-func.func @test_spmdize_constant() ->(tensor<1024x1024xf32>)attributes{llvm.emit_c_interface} {
-  %cst = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
-  %sharding_1 = mesh.sharding @mesh4x4 split_axes = [[0]] : !mesh.sharding
-  %sharding_annotated_1 = mesh.shard %cst to %sharding_1 : tensor<1024x1024xf32>
-  %ci = arith.constant 434 : i32
-  return %sharding_annotated_1 : tensor<1024x1024xf32>
-}
diff --git a/mlir/test/Dialect/Arith/shard-partition.mlir b/mlir/test/Dialect/Arith/shard-partition.mlir
new file mode 100644
index 0000000..be89427
--- /dev/null
+++ b/mlir/test/Dialect/Arith/shard-partition.mlir
@@ -0,0 +1,17 @@
+// RUN: mlir-opt \
+// RUN:   --pass-pipeline="builtin.module(func.func(shard-partition))" \
+// RUN:   %s | FileCheck %s
+
+shard.grid @grid4x4(shape = 4x4)
+
+// CHECK-LABEL: func @test_partition_constant
+// CHECK-NEXT: [[vcst:%.*]] = arith.constant dense<0.000000e+00> :
+// tensor<256x1024xf32> CHECK-NEXT: [[vc434_i32:%.*]] = arith.constant 434 :
+// i32 CHECK-NEXT: return [[vcst]] : tensor<256x1024xf32>
+func.func @test_partition_constant() ->(tensor<1024x1024xf32>)attributes{llvm.emit_c_interface} {
+  %cst = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
+  %sharding_1 = shard.sharding @grid4x4 split_axes = [[0]] : !shard.sharding
+  %sharded_1 = shard.shard %cst to %sharding_1 : tensor<1024x1024xf32>
+  %ci = arith.constant 434 : i32
+  return %sharded_1 : tensor<1024x1024xf32>
+}
diff --git a/mlir/test/Dialect/Arith/sharding-propagation.mlir b/mlir/test/Dialect/Arith/sharding-propagation.mlir
index 19eb340..762620d 100644
--- a/mlir/test/Dialect/Arith/sharding-propagation.mlir
+++ b/mlir/test/Dialect/Arith/sharding-propagation.mlir
@@ -1,54 +1,54 @@
 // RUN: mlir-opt --pass-pipeline="builtin.module(func.func(sharding-propagation))" %s | FileCheck %s
 
-mesh.mesh @mesh4x4(shape = 4x4)
+shard.grid @grid4x4(shape = 4x4)
 
 // CHECK-LABEL: func.func @test_shard_constant() -> tensor<1024x1024xf32> attributes {llvm.emit_c_interface} {
 // CHECK-NEXT: [[vcst:%.*]] = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
-// CHECK-NEXT: [[vsharding:%.*]] = mesh.sharding @mesh4x4 split_axes = {{\[\[}}0]] : !mesh.sharding
-// CHECK-NEXT: [[vsharding_annotated:%.*]] = mesh.shard [[vcst]] to [[vsharding]] : tensor<1024x1024xf32>
+// CHECK-NEXT: [[vsharding:%.*]] = shard.sharding @grid4x4 split_axes = {{\[\[}}0]] : !shard.sharding
+// CHECK-NEXT: [[vsharded:%.*]] = shard.shard [[vcst]] to [[vsharding]] : tensor<1024x1024xf32>
 // CHECK-NEXT: [[vcst_0:%.*]] = arith.constant 4.340000e+01 : f32
 // CHECK-NEXT: [[v0:%.*]] = tensor.empty() : tensor<1024x1024xf32>
-// CHECK-NEXT: [[vsharding_1:%.*]] = mesh.sharding @mesh4x4 split_axes = {{\[\[}}0]] : !mesh.sharding
-// CHECK-NEXT: [[vsharding_annotated_2:%.*]] = mesh.shard [[v0]] to [[vsharding_1]] : tensor<1024x1024xf32>
-// CHECK-NEXT: [[vsharding_3:%.*]] = mesh.sharding @mesh4x4 split_axes = {{\[\[}}0]] : !mesh.sharding
-// CHECK-NEXT: [[vsharding_annotated_4:%.*]] = mesh.shard [[vsharding_annotated]] to [[vsharding_3]] annotate_for_users : tensor<1024x1024xf32>
-// CHECK-NEXT: [[vsharding_5:%.*]] = mesh.sharding @mesh4x4 split_axes = {{\[\[}}0]] : !mesh.sharding
-// CHECK-NEXT: [[vsharding_annotated_6:%.*]] = mesh.shard [[vsharding_annotated_2]] to [[vsharding_5]] annotate_for_users : tensor<1024x1024xf32>
-// CHECK-NEXT: [[v1:%.*]] = linalg.add ins([[vsharding_annotated_4]], [[vcst_0]] : tensor<1024x1024xf32>, f32) outs([[vsharding_annotated_6]] : tensor<1024x1024xf32>) -> tensor<1024x1024xf32>
-// CHECK-NEXT: [[vsharding_7:%.*]] = mesh.sharding @mesh4x4 split_axes = {{\[\[}}0]] : !mesh.sharding
-// CHECK-NEXT: [[vsharding_annotated_8:%.*]] = mesh.shard [[v1]] to [[vsharding_7]] : tensor<1024x1024xf32>
-// CHECK-NEXT: return [[vsharding_annotated_8]] : tensor<1024x1024xf32>
+// CHECK-NEXT: [[vsharding_1:%.*]] = shard.sharding @grid4x4 split_axes = {{\[\[}}0]] : !shard.sharding
+// CHECK-NEXT: [[vsharded_2:%.*]] = shard.shard [[v0]] to [[vsharding_1]] : tensor<1024x1024xf32>
+// CHECK-NEXT: [[vsharding_3:%.*]] = shard.sharding @grid4x4 split_axes = {{\[\[}}0]] : !shard.sharding
+// CHECK-NEXT: [[vsharded_4:%.*]] = shard.shard [[vsharded]] to [[vsharding_3]] annotate_for_users : tensor<1024x1024xf32>
+// CHECK-NEXT: [[vsharding_5:%.*]] = shard.sharding @grid4x4 split_axes = {{\[\[}}0]] : !shard.sharding
+// CHECK-NEXT: [[vsharded_6:%.*]] = shard.shard [[vsharded_2]] to [[vsharding_5]] annotate_for_users : tensor<1024x1024xf32>
+// CHECK-NEXT: [[v1:%.*]] = linalg.add ins([[vsharded_4]], [[vcst_0]] : tensor<1024x1024xf32>, f32) outs([[vsharded_6]] : tensor<1024x1024xf32>) -> tensor<1024x1024xf32>
+// CHECK-NEXT: [[vsharding_7:%.*]] = shard.sharding @grid4x4 split_axes = {{\[\[}}0]] : !shard.sharding
+// CHECK-NEXT: [[vsharded_8:%.*]] = shard.shard [[v1]] to [[vsharding_7]] : tensor<1024x1024xf32>
+// CHECK-NEXT: return [[vsharded_8]] : tensor<1024x1024xf32>
 func.func @test_shard_constant() -> (tensor<1024x1024xf32>) attributes {llvm.emit_c_interface} {
     %cst_1 = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
-    %sharding_1 = mesh.sharding @mesh4x4 split_axes = [[0]] : !mesh.sharding
-    %sharding_annotated_1 = mesh.shard %cst_1 to %sharding_1 : tensor<1024x1024xf32>
+    %sharding_1 = shard.sharding @grid4x4 split_axes = [[0]] : !shard.sharding
+    %sharded_1 = shard.shard %cst_1 to %sharding_1 : tensor<1024x1024xf32>
     %ci = arith.constant 43.4e+00 : f32
     %o1 = tensor.empty() : tensor<1024x1024xf32>
-    %res = linalg.add ins(%sharding_annotated_1, %ci : tensor<1024x1024xf32>, f32) outs(%o1 : tensor<1024x1024xf32>) -> tensor<1024x1024xf32>
+    %res = linalg.add ins(%sharded_1, %ci : tensor<1024x1024xf32>, f32) outs(%o1 : tensor<1024x1024xf32>) -> tensor<1024x1024xf32>
     return %res : tensor<1024x1024xf32>
 }
 
 // CHECK-LABEL: func.func @test_shard_constant_back() -> tensor<1024x1024xf32> attributes {llvm.emit_c_interface} {
 // CHECK-NEXT: [[vcst:%.*]] = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
-// CHECK-NEXT: [[vsharding:%.*]] = mesh.sharding @mesh4x4 split_axes = {{\[\[}}0]] : !mesh.sharding
-// CHECK-NEXT: [[vsharding_annotated:%.*]] = mesh.shard [[vcst]] to [[vsharding]] : tensor<1024x1024xf32>
+// CHECK-NEXT: [[vsharding:%.*]] = shard.sharding @grid4x4 split_axes = {{\[\[}}0]] : !shard.sharding
+// CHECK-NEXT: [[vsharded:%.*]] = shard.shard [[vcst]] to [[vsharding]] : tensor<1024x1024xf32>
 // CHECK-NEXT: [[vcst_0:%.*]] = arith.constant 4.340000e+01 : f32
 // CHECK-NEXT: [[v0:%.*]] = tensor.empty() : tensor<1024x1024xf32>
-// CHECK-NEXT: [[vsharding_1:%.*]] = mesh.sharding @mesh4x4 split_axes = {{\[\[}}0]] : !mesh.sharding
-// CHECK-NEXT: [[vsharding_annotated_2:%.*]] = mesh.shard [[v0]] to [[vsharding_1]] : tensor<1024x1024xf32>
-// CHECK-NEXT: [[vsharding_3:%.*]] = mesh.sharding @mesh4x4 split_axes = {{\[\[}}0]] : !mesh.sharding
-// CHECK-NEXT: [[vsharding_annotated_4:%.*]] = mesh.shard [[vsharding_annotated]] to [[vsharding_3]] annotate_for_users : tensor<1024x1024xf32>
-// CHECK-NEXT: [[vsharding_5:%.*]] = mesh.sharding @mesh4x4 split_axes = {{\[\[}}0]] : !mesh.sharding
-// CHECK-NEXT: [[vsharding_annotated_6:%.*]] = mesh.shard [[vsharding_annotated_2]] to [[vsharding_5]] annotate_for_users : tensor<1024x1024xf32>
-// CHECK-NEXT: [[v1:%.*]] = linalg.add ins([[vsharding_annotated_4]], [[vcst_0]] : tensor<1024x1024xf32>, f32) outs([[vsharding_annotated_6]] : tensor<1024x1024xf32>) -> tensor<1024x1024xf32>
-// CHECK-NEXT: [[vsharding_7:%.*]] = mesh.sharding @mesh4x4 split_axes = {{\[\[}}0]] : !mesh.sharding
-// CHECK-NEXT: [[vsharding_annotated_8:%.*]] = mesh.shard [[v1]] to [[vsharding_7]] : tensor<1024x1024xf32>
+// CHECK-NEXT: [[vsharding_1:%.*]] = shard.sharding @grid4x4 split_axes = {{\[\[}}0]] : !shard.sharding
+// CHECK-NEXT: [[vsharded_2:%.*]] = shard.shard [[v0]] to [[vsharding_1]] : tensor<1024x1024xf32>
+// CHECK-NEXT: [[vsharding_3:%.*]] = shard.sharding @grid4x4 split_axes = {{\[\[}}0]] : !shard.sharding
+// CHECK-NEXT: [[vsharded_4:%.*]] = shard.shard [[vsharded]] to [[vsharding_3]] annotate_for_users : tensor<1024x1024xf32>
+// CHECK-NEXT: [[vsharding_5:%.*]] = shard.sharding @grid4x4 split_axes = {{\[\[}}0]] : !shard.sharding
+// CHECK-NEXT: [[vsharded_6:%.*]] = shard.shard [[vsharded_2]] to [[vsharding_5]] annotate_for_users : tensor<1024x1024xf32>
+// CHECK-NEXT: [[v1:%.*]] = linalg.add ins([[vsharded_4]], [[vcst_0]] : tensor<1024x1024xf32>, f32) outs([[vsharded_6]] : tensor<1024x1024xf32>) -> tensor<1024x1024xf32>
+// CHECK-NEXT: [[vsharding_7:%.*]] = shard.sharding @grid4x4 split_axes = {{\[\[}}0]] : !shard.sharding
+// CHECK-NEXT: [[vsharded_8:%.*]] = shard.shard [[v1]] to [[vsharding_7]] : tensor<1024x1024xf32>
 func.func @test_shard_constant_back() -> (tensor<1024x1024xf32>) attributes {llvm.emit_c_interface} {
     %cst_1 = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
     %ci = arith.constant 43.4e+00 : f32
     %o1 = tensor.empty() : tensor<1024x1024xf32>
     %res = linalg.add ins(%cst_1, %ci : tensor<1024x1024xf32>, f32) outs(%o1 : tensor<1024x1024xf32>) -> tensor<1024x1024xf32>
-    %sharding_1 = mesh.sharding @mesh4x4 split_axes = [[0]] : !mesh.sharding
-    %sharding_annotated_1 = mesh.shard %res to %sharding_1 : tensor<1024x1024xf32>
-    return %sharding_annotated_1 : tensor<1024x1024xf32>
+    %sharding_1 = shard.sharding @grid4x4 split_axes = [[0]] : !shard.sharding
+    %sharded_1 = shard.shard %res to %sharding_1 : tensor<1024x1024xf32>
+    return %sharded_1 : tensor<1024x1024xf32>
 }
diff --git a/mlir/test/Dialect/EmitC/wrap_emitc_func_in_class.mlir b/mlir/test/Dialect/EmitC/wrap_emitc_func_in_class.mlir
index c67a0c1..029fa78 100644
--- a/mlir/test/Dialect/EmitC/wrap_emitc_func_in_class.mlir
+++ b/mlir/test/Dialect/EmitC/wrap_emitc_func_in_class.mlir
@@ -1,4 +1,4 @@
-// RUN: mlir-opt --wrap-emitc-func-in-class='named-attribute=emitc.name_hint' %s | FileCheck %s
+// RUN: mlir-opt --wrap-emitc-func-in-class %s | FileCheck %s
 
 module attributes { } {
   emitc.func @model(%arg0: !emitc.array<1xf32> {emitc.name_hint = "another_feature"}, 
diff --git a/mlir/test/Dialect/Linalg/mesh-sharding-propagation.mlir b/mlir/test/Dialect/Linalg/mesh-sharding-propagation.mlir
deleted file mode 100644
index 5297eeb..0000000
--- a/mlir/test/Dialect/Linalg/mesh-sharding-propagation.mlir
+++ /dev/null
@@ -1,42 +0,0 @@
-// RUN: mlir-opt \
-// RUN:   --verify-each \
-// RUN:   --pass-pipeline="builtin.module(func.func(sharding-propagation))" \
-// RUN:   %s | FileCheck %s
-
-mesh.mesh @mesh_2(shape = 2)
-
-// CHECK-LABEL: func @matmul_shard_prallel_axis
-func.func @matmul_shard_prallel_axis(
-  // CHECK-SAME: %[[IN1:[A-Za-z0-9_]+]]: tensor<2x3xf32>,
-  %arg0 : tensor<2x3xf32>,
-  // CHECK-SAME: %[[IN2:[A-Za-z0-9_]+]]: tensor<3x2xf32>,
-  %arg1 : tensor<3x2xf32>,
-  // CHECK-SAME: %[[DPS_OUT:[A-Za-z0-9_]+]]: tensor<2x2xf32>
-  %out_dps: tensor<2x2xf32>
-) -> tensor<2x2xf32> {
-  // CHECK: %[[SIN1_ANNOTATED_0:.*]] = mesh.sharding @mesh_2 split_axes = {{\[}}[0]] : !mesh.sharding
-  // CHECK-NEXT: %[[IN1_ANNOTATED_0:.*]] = mesh.shard %[[IN1]] to %[[SIN1_ANNOTATED_0]] : tensor<2x3xf32>
-  // CHECK: %[[SIN1_ANNOTATED_1:.*]] = mesh.sharding @mesh_2 split_axes = {{\[}}[0]] : !mesh.sharding
-  // CHECK-NEXT: %[[IN1_ANNOTATED_1:.*]] = mesh.shard %[[IN1_ANNOTATED_0]] to %[[SIN1_ANNOTATED_1]] annotate_for_users : tensor<2x3xf32>
-  // CHECK: %[[SIN2_ANNOTATED:.*]] = mesh.sharding @mesh_2 split_axes = {{\[}}[]] : !mesh.sharding
-  // CHECK-NEXT: %[[IN2_ANNOTATED:.*]] = mesh.shard %[[IN2]] to %[[SIN2_ANNOTATED]] annotate_for_users : tensor<3x2xf32>
-  // CHECK: %[[SDPS_OUT_ANNOTATED:.*]] = mesh.sharding @mesh_2 split_axes = {{\[}}[0]] : !mesh.sharding
-  // CHECK-NEXT: %[[DPS_OUT_ANNOTATED:.*]] = mesh.shard %[[DPS_OUT]] to %[[SDPS_OUT_ANNOTATED]] annotate_for_users : tensor<2x2xf32>
-  %sarg0_sharded = mesh.sharding @mesh_2 split_axes = [[0]] : !mesh.sharding
-  %arg0_sharded = mesh.shard %arg0 to %sarg0_sharded : tensor<2x3xf32>
-
-  // CHECK: %[[RES:.*]] = linalg.matmul ins(%[[IN1_ANNOTATED_1]], %[[IN2_ANNOTATED]] : tensor<2x3xf32>, tensor<3x2xf32>)
-  // CHECK-SAME:  outs(%[[DPS_OUT_ANNOTATED]] : tensor<2x2xf32>) -> tensor<2x2xf32>
-  %res = linalg.matmul ins(%arg0_sharded, %arg1 : tensor<2x3xf32>, tensor<3x2xf32>)
-    outs(%out_dps : tensor<2x2xf32>) -> tensor<2x2xf32>
-
-  // CHECK: %[[SRES_ANNOTATED_0:.*]] = mesh.sharding @mesh_2 split_axes = {{\[}}[0]] : !mesh.sharding
-  // CHECK-NEXT: %[[RES_ANNOTATED_0:.*]] = mesh.shard %[[RES]] to %[[SRES_ANNOTATED_0]] : tensor<2x2xf32>
-  // CHECK: %[[SRES_ANNOTATED_1:.*]] = mesh.sharding @mesh_2 split_axes = {{\[}}[]] : !mesh.sharding
-  // CHECK-NEXT: %[[RES_ANNOTATED_1:.*]] = mesh.shard %[[RES_ANNOTATED_0]] to %[[SRES_ANNOTATED_1]] annotate_for_users : tensor<2x2xf32>
-  %sres_sharded = mesh.sharding @mesh_2 split_axes = [[]] : !mesh.sharding
-  %res_sharded = mesh.shard %res to %sres_sharded annotate_for_users : tensor<2x2xf32>
-
-  // CHECK: return %[[RES_ANNOTATED_1]] : tensor<2x2xf32>
-  return %res_sharded : tensor<2x2xf32>
-}
diff --git a/mlir/test/Dialect/Linalg/mesh-spmdization.mlir b/mlir/test/Dialect/Linalg/shard-partition.mlir
index ce12b29..aee9707 100644
--- a/mlir/test/Dialect/Linalg/mesh-spmdization.mlir
+++ b/mlir/test/Dialect/Linalg/shard-partition.mlir
@@ -1,15 +1,15 @@
 // RUN: mlir-opt \
-// RUN:  --pass-pipeline="builtin.module(func.func(mesh-spmdization,test-single-fold))" \
+// RUN:  --pass-pipeline="builtin.module(func.func(shard-partition,test-single-fold))" \
 // RUN:  --split-input-file \
 // RUN:  %s | FileCheck %s
 
 // CHECK: #[[$MAP_IDENTITY_1D:.*]] = affine_map<(d0) -> (d0)>
 #map_identity_1d = affine_map<(d0) -> (d0)>
 
-mesh.mesh @mesh_1d(shape = 2)
+shard.grid @grid_1d(shape = 2)
 
-// CHECK-LABEL: func @elementwise_static_1d_mesh_static_1d_tensor
-func.func @elementwise_static_1d_mesh_static_1d_tensor(
+// CHECK-LABEL: func @elementwise_static_1d_grid_static_1d_tensor
+func.func @elementwise_static_1d_grid_static_1d_tensor(
   // CHECK-SAME: %[[IN1:[A-Za-z0-9_]+]]: tensor<1xi8>,
   %in1: tensor<2xi8>,
   // CHECK-SAME: %[[IN2:[A-Za-z0-9_]+]]: tensor<1xi8>,
@@ -18,13 +18,13 @@ func.func @elementwise_static_1d_mesh_static_1d_tensor(
   %dps_out: tensor<2xi8>
 // CHECK-SAME: -> tensor<1xi8> {
 ) -> tensor<2xi8> {
-  %sharding = mesh.sharding @mesh_1d split_axes = [[0]]  : !mesh.sharding
-  %in1_sharded1 = mesh.shard %in1 to %sharding  : tensor<2xi8>
-  %in1_sharded2 = mesh.shard %in1_sharded1 to %sharding annotate_for_users : tensor<2xi8>
-  %in2_sharded1 = mesh.shard %in2 to %sharding : tensor<2xi8>
-  %in2_sharded2 = mesh.shard %in2_sharded1 to %sharding annotate_for_users : tensor<2xi8>
-  %dps_out_sharded1 = mesh.shard %dps_out to %sharding : tensor<2xi8>
-  %dps_out_shared2 = mesh.shard %dps_out_sharded1 to %sharding annotate_for_users : tensor<2xi8>
+  %sharding = shard.sharding @grid_1d split_axes = [[0]]  : !shard.sharding
+  %in1_sharded1 = shard.shard %in1 to %sharding  : tensor<2xi8>
+  %in1_sharded2 = shard.shard %in1_sharded1 to %sharding annotate_for_users : tensor<2xi8>
+  %in2_sharded1 = shard.shard %in2 to %sharding : tensor<2xi8>
+  %in2_sharded2 = shard.shard %in2_sharded1 to %sharding annotate_for_users : tensor<2xi8>
+  %dps_out_sharded1 = shard.shard %dps_out to %sharding : tensor<2xi8>
+  %dps_out_shared2 = shard.shard %dps_out_sharded1 to %sharding annotate_for_users : tensor<2xi8>
   // CHECK: %[[RES:.*]] = linalg.generic {
   // CHECK-SAME: indexing_maps = [#[[$MAP_IDENTITY_1D]], #[[$MAP_IDENTITY_1D]], #[[$MAP_IDENTITY_1D]]],
   // CHECK-SAME: iterator_types = ["parallel"]}
@@ -39,18 +39,18 @@ func.func @elementwise_static_1d_mesh_static_1d_tensor(
       %res_scalar = arith.muli %in1_scalar, %in2_scalar : i8
       linalg.yield %res_scalar : i8
     } -> tensor<2xi8>
-  %res_sharded1 = mesh.shard %res to %sharding : tensor<2xi8>
-  %res_shared2 = mesh.shard %res_sharded1 to %sharding annotate_for_users : tensor<2xi8>
+  %res_sharded1 = shard.shard %res to %sharding : tensor<2xi8>
+  %res_shared2 = shard.shard %res_sharded1 to %sharding annotate_for_users : tensor<2xi8>
   // CHECK: return %[[RES]] : tensor<1xi8>
   return %res_shared2 : tensor<2xi8>
 }
 
 // -----
 
-mesh.mesh @mesh_1d(shape = 4)
+shard.grid @grid_1d(shape = 4)
 
-// CHECK-LABEL: func @matmul_1d_mesh_static_tensors_parallel_iterator_sharding
-func.func @matmul_1d_mesh_static_tensors_parallel_iterator_sharding(
+// CHECK-LABEL: func @matmul_1d_grid_static_tensors_parallel_iterator_sharding
+func.func @matmul_1d_grid_static_tensors_parallel_iterator_sharding(
   // CHECK-SAME: %[[IN1:[A-Za-z0-9_]+]]: tensor<1x3xi8>,
   %in1: tensor<4x3xi8>,
 // CHECK-SAME: %[[IN2:[A-Za-z0-9_]+]]: tensor<3x8xi8>,
@@ -59,32 +59,32 @@ func.func @matmul_1d_mesh_static_tensors_parallel_iterator_sharding(
   %dps_out: tensor<4x8xi8>
 // CHECK-SAME: -> tensor<1x8xi8> {
 ) -> tensor<4x8xi8> {
-  %sharding = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %in1_shared1 = mesh.shard %in1 to %sharding : tensor<4x3xi8>
-  %in1_shared2 = mesh.shard %in1_shared1 to %sharding annotate_for_users : tensor<4x3xi8>
-  %sharding2 = mesh.sharding @mesh_1d split_axes = [[]] : !mesh.sharding
-  %in2_shared1 = mesh.shard %in2 to %sharding2 : tensor<3x8xi8>
-  %in2_shared2 = mesh.shard %in2_shared1 to %sharding2 annotate_for_users : tensor<3x8xi8>
-  %dps_out_shared1 = mesh.shard %dps_out to %sharding : tensor<4x8xi8>
-  %dps_out_shared2 = mesh.shard %dps_out_shared1 to %sharding annotate_for_users : tensor<4x8xi8>
+  %sharding = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %in1_shared1 = shard.shard %in1 to %sharding : tensor<4x3xi8>
+  %in1_shared2 = shard.shard %in1_shared1 to %sharding annotate_for_users : tensor<4x3xi8>
+  %sharding2 = shard.sharding @grid_1d split_axes = [[]] : !shard.sharding
+  %in2_shared1 = shard.shard %in2 to %sharding2 : tensor<3x8xi8>
+  %in2_shared2 = shard.shard %in2_shared1 to %sharding2 annotate_for_users : tensor<3x8xi8>
+  %dps_out_shared1 = shard.shard %dps_out to %sharding : tensor<4x8xi8>
+  %dps_out_shared2 = shard.shard %dps_out_shared1 to %sharding annotate_for_users : tensor<4x8xi8>
   // CHECK: %[[RES:.*]] = linalg.matmul
   // CHECK-SAME: ins(%[[IN1]], %[[IN2]] : tensor<1x3xi8>, tensor<3x8xi8>)
   // CHECK-SAME: outs(%[[DPS_OUT]] : tensor<1x8xi8>)
   // CHECK-SAME: -> tensor<1x8xi8>
   %res = linalg.matmul ins(%in1_shared2, %in2_shared2 : tensor<4x3xi8>, tensor<3x8xi8>)
       outs(%dps_out_shared2 : tensor<4x8xi8>) -> tensor<4x8xi8>
-  %res_shared1 = mesh.shard %res to %sharding : tensor<4x8xi8>
-  %res_shared2 = mesh.shard %res_shared1 to %sharding annotate_for_users : tensor<4x8xi8>
+  %res_shared1 = shard.shard %res to %sharding : tensor<4x8xi8>
+  %res_shared2 = shard.shard %res_shared1 to %sharding annotate_for_users : tensor<4x8xi8>
   // CHECK: return %[[RES]] : tensor<1x8xi8>
   return %res_shared2 : tensor<4x8xi8>
 }
 
 // -----
 
-mesh.mesh @mesh_1d(shape = 3)
+shard.grid @grid_1d(shape = 3)
 
-// CHECK-LABEL: func @matmul_1d_mesh_static_tensors_reduction_iterator_sharding
-func.func @matmul_1d_mesh_static_tensors_reduction_iterator_sharding(
+// CHECK-LABEL: func @matmul_1d_grid_static_tensors_reduction_iterator_sharding
+func.func @matmul_1d_grid_static_tensors_reduction_iterator_sharding(
   // CHECK-SAME: %[[IN1:[A-Za-z0-9_]+]]: tensor<4x2xi8>,
   %in1: tensor<4x6xi8>,
 // CHECK-SAME: %[[IN2:[A-Za-z0-9_]+]]: tensor<2x8xi8>,
@@ -93,19 +93,19 @@ func.func @matmul_1d_mesh_static_tensors_reduction_iterator_sharding(
   %dps_out: tensor<4x8xi8>
 // CHECK-SAME: -> tensor<4x8xi8> {
 ) -> tensor<4x8xi8> {
-  %sharding = mesh.sharding @mesh_1d split_axes = [[], [0]] : !mesh.sharding
-  %in1_shared1 = mesh.shard %in1 to %sharding : tensor<4x6xi8>
-  %in1_shared2 = mesh.shard %in1_shared1 to %sharding annotate_for_users : tensor<4x6xi8>
-  %sharding2 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %in2_shared1 = mesh.shard %in2 to %sharding2 : tensor<6x8xi8>
-  %in2_shared2 = mesh.shard %in2_shared1 to %sharding2 annotate_for_users : tensor<6x8xi8>
-  %sharding3 = mesh.sharding @mesh_1d split_axes = [[]] : !mesh.sharding
-  %dps_out_shared1 = mesh.shard %dps_out to %sharding3 : tensor<4x8xi8>
-  %dps_out_shared2 = mesh.shard %dps_out_shared1 to %sharding3 annotate_for_users : tensor<4x8xi8>
+  %sharding = shard.sharding @grid_1d split_axes = [[], [0]] : !shard.sharding
+  %in1_shared1 = shard.shard %in1 to %sharding : tensor<4x6xi8>
+  %in1_shared2 = shard.shard %in1_shared1 to %sharding annotate_for_users : tensor<4x6xi8>
+  %sharding2 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %in2_shared1 = shard.shard %in2 to %sharding2 : tensor<6x8xi8>
+  %in2_shared2 = shard.shard %in2_shared1 to %sharding2 annotate_for_users : tensor<6x8xi8>
+  %sharding3 = shard.sharding @grid_1d split_axes = [[]] : !shard.sharding
+  %dps_out_shared1 = shard.shard %dps_out to %sharding3 : tensor<4x8xi8>
+  %dps_out_shared2 = shard.shard %dps_out_shared1 to %sharding3 annotate_for_users : tensor<4x8xi8>
   // CHECK-DAG:  %[[C0:.*]] = arith.constant 0 : index
   // CHECK-DAG:  %[[C0_I8:.*]] = arith.constant 0 : i8
-  // CHECK-DAG:  %[[PROCESS_IDX:.*]] = mesh.process_multi_index on @mesh_1d axes = [0] : index
-  // CHECK-DAG:  %[[MESH_SIZE:.*]] = mesh.mesh_shape @mesh_1d axes = [0] : index
+  // CHECK-DAG:  %[[PROCESS_IDX:.*]] = shard.process_multi_index on @grid_1d axes = [0] : index
+  // CHECK-DAG:  %[[SHARD_SIZE:.*]] = shard.grid_shape @grid_1d axes = [0] : index
   // CHECK:      %[[DPS_INIT_OPERAND_CONDITION:.*]] = arith.cmpi eq, %[[PROCESS_IDX]], %[[C0]] : index
   // CHECK:      %[[DPS_INIT_OPERAND:.*]] = scf.if %[[DPS_INIT_OPERAND_CONDITION]] -> (tensor<4x8xi8>) {
   // CHECK:        scf.yield %[[DPS_OUT]] : tensor<4x8xi8>
@@ -117,21 +117,21 @@ func.func @matmul_1d_mesh_static_tensors_reduction_iterator_sharding(
   // CHECK:      }
   // CHECK:      %[[SHARDED_MATMUL:.*]] = linalg.matmul ins(%[[IN1]], %[[IN2]] : tensor<4x2xi8>, tensor<2x8xi8>)
   // CHECK-SAME:     outs(%[[DPS_INIT_OPERAND]] : tensor<4x8xi8>) -> tensor<4x8xi8>
-  // CHECK:      %[[ALL_REDUCED:.*]] = mesh.all_reduce %[[SHARDED_MATMUL]] on @mesh_1d mesh_axes = [0] : tensor<4x8xi8> -> tensor<4x8xi8>
+  // CHECK:      %[[ALL_REDUCED:.*]] = shard.all_reduce %[[SHARDED_MATMUL]] on @grid_1d grid_axes = [0] : tensor<4x8xi8> -> tensor<4x8xi8>
   %res = linalg.matmul ins(%in1_shared2, %in2_shared2 : tensor<4x6xi8>, tensor<6x8xi8>)
       outs(%dps_out_shared2 : tensor<4x8xi8>) -> tensor<4x8xi8>
-  %res_shared1 = mesh.shard %res to %sharding3 : tensor<4x8xi8>
-  %res_shared2 = mesh.shard %res_shared1 to %sharding3 annotate_for_users : tensor<4x8xi8>
+  %res_shared1 = shard.shard %res to %sharding3 : tensor<4x8xi8>
+  %res_shared2 = shard.shard %res_shared1 to %sharding3 annotate_for_users : tensor<4x8xi8>
   // CHECK:      return %[[ALL_REDUCED]] : tensor<4x8xi8>
   return %res_shared2 : tensor<4x8xi8>
 }
 
 // -----
 
-mesh.mesh @mesh_1d(shape = 4)
+shard.grid @grid_1d(shape = 4)
 
-// CHECK-LABEL: func @matmul_1d_mesh_static_tensors_parallel_iterator_unsplit_last_axis
-func.func @matmul_1d_mesh_static_tensors_parallel_iterator_unsplit_last_axis(
+// CHECK-LABEL: func @matmul_1d_grid_static_tensors_parallel_iterator_unsplit_last_axis
+func.func @matmul_1d_grid_static_tensors_parallel_iterator_unsplit_last_axis(
   // CHECK-SAME: %[[IN1:[A-Za-z0-9_]+]]: tensor<4x6xi8>,
   %in1: tensor<4x6xi8>,
   // CHECK-SAME: %[[IN2:[A-Za-z0-9_]+]]: tensor<6x8xi8>,
@@ -140,25 +140,25 @@ func.func @matmul_1d_mesh_static_tensors_parallel_iterator_unsplit_last_axis(
   %dps_out: tensor<4x8xi8>
   // CHECK-SAME: -> tensor<4x8xi8> {
 ) -> tensor<4x8xi8> {
-  %sharding1 = mesh.sharding @mesh_1d split_axes = [[], []] : !mesh.sharding
-  %in1_replicated1 = mesh.shard %in1 to %sharding1 : tensor<4x6xi8>
-  %in1_replicated2 = mesh.shard %in1_replicated1 to %sharding1 annotate_for_users : tensor<4x6xi8>
-  // CHECK: %[[ALL_SLICE1:.*]] = mesh.all_slice %[[IN2]] on @mesh_1d mesh_axes = [0] slice_axis = 1
-  %in2_replicated = mesh.shard %in2 to %sharding1 : tensor<6x8xi8>
-  %sharding2 = mesh.sharding @mesh_1d split_axes = [[], [0]] : !mesh.sharding
-  %in2_sharded = mesh.shard %in2_replicated to %sharding2 annotate_for_users : tensor<6x8xi8>
-  // CHECK: %[[ALL_SLICE2:.*]] = mesh.all_slice %[[DPS_OUT]] on @mesh_1d mesh_axes = [0] slice_axis = 1
-  %dps_out_replicated = mesh.shard %dps_out to %sharding1 : tensor<4x8xi8>
-  %dps_out_sharded = mesh.shard %dps_out_replicated to %sharding2 annotate_for_users : tensor<4x8xi8>
+  %sharding1 = shard.sharding @grid_1d split_axes = [[], []] : !shard.sharding
+  %in1_replicated1 = shard.shard %in1 to %sharding1 : tensor<4x6xi8>
+  %in1_replicated2 = shard.shard %in1_replicated1 to %sharding1 annotate_for_users : tensor<4x6xi8>
+  // CHECK: %[[ALL_SLICE1:.*]] = shard.all_slice %[[IN2]] on @grid_1d grid_axes = [0] slice_axis = 1
+  %in2_replicated = shard.shard %in2 to %sharding1 : tensor<6x8xi8>
+  %sharding2 = shard.sharding @grid_1d split_axes = [[], [0]] : !shard.sharding
+  %in2_sharded = shard.shard %in2_replicated to %sharding2 annotate_for_users : tensor<6x8xi8>
+  // CHECK: %[[ALL_SLICE2:.*]] = shard.all_slice %[[DPS_OUT]] on @grid_1d grid_axes = [0] slice_axis = 1
+  %dps_out_replicated = shard.shard %dps_out to %sharding1 : tensor<4x8xi8>
+  %dps_out_sharded = shard.shard %dps_out_replicated to %sharding2 annotate_for_users : tensor<4x8xi8>
   // CHECK: %[[MATMUL_RES:.*]] = linalg.matmul
   // CHECK-SAME: ins(%[[IN1]], %[[ALL_SLICE1]] : tensor<4x6xi8>, tensor<6x2xi8>)
   // CHECK-SAME: outs(%[[ALL_SLICE2]] : tensor<4x2xi8>)
   // CHECK-SAME: -> tensor<4x2xi8>
   %res = linalg.matmul ins(%in1_replicated2, %in2_sharded : tensor<4x6xi8>, tensor<6x8xi8>)
       outs(%dps_out_sharded : tensor<4x8xi8>) -> tensor<4x8xi8>
-  // CHECK: %[[ALL_GATHER:.*]] = mesh.all_gather %[[MATMUL_RES]] on @mesh_1d mesh_axes = [0] gather_axis = 1 : tensor<4x2xi8> -> tensor<4x8xi8>
-  %res_sharded = mesh.shard %res to %sharding2 : tensor<4x8xi8>
-  %res_replicated = mesh.shard %res_sharded to %sharding1 annotate_for_users : tensor<4x8xi8>
+  // CHECK: %[[ALL_GATHER:.*]] = shard.all_gather %[[MATMUL_RES]] on @grid_1d grid_axes = [0] gather_axis = 1 : tensor<4x2xi8> -> tensor<4x8xi8>
+  %res_sharded = shard.shard %res to %sharding2 : tensor<4x8xi8>
+  %res_replicated = shard.shard %res_sharded to %sharding1 annotate_for_users : tensor<4x8xi8>
   // CHECK: return %[[ALL_GATHER]] : tensor<4x8xi8>
   return %res_replicated : tensor<4x8xi8>
 }
diff --git a/mlir/test/Dialect/Linalg/sharding-propagation.mlir b/mlir/test/Dialect/Linalg/sharding-propagation.mlir
new file mode 100644
index 0000000..e0ecefc
--- /dev/null
+++ b/mlir/test/Dialect/Linalg/sharding-propagation.mlir
@@ -0,0 +1,42 @@
+// RUN: mlir-opt \
+// RUN:   --verify-each \
+// RUN:   --pass-pipeline="builtin.module(func.func(sharding-propagation))" \
+// RUN:   %s | FileCheck %s
+
+shard.grid @grid_2(shape = 2)
+
+// CHECK-LABEL: func @matmul_shard_prallel_axis
+func.func @matmul_shard_prallel_axis(
+  // CHECK-SAME: %[[IN1:[A-Za-z0-9_]+]]: tensor<2x3xf32>,
+  %arg0 : tensor<2x3xf32>,
+  // CHECK-SAME: %[[IN2:[A-Za-z0-9_]+]]: tensor<3x2xf32>,
+  %arg1 : tensor<3x2xf32>,
+  // CHECK-SAME: %[[DPS_OUT:[A-Za-z0-9_]+]]: tensor<2x2xf32>
+  %out_dps: tensor<2x2xf32>
+) -> tensor<2x2xf32> {
+  // CHECK: %[[SIN1_ANNOTATED_0:.*]] = shard.sharding @grid_2 split_axes = {{\[}}[0]] : !shard.sharding
+  // CHECK-NEXT: %[[IN1_ANNOTATED_0:.*]] = shard.shard %[[IN1]] to %[[SIN1_ANNOTATED_0]] : tensor<2x3xf32>
+  // CHECK: %[[SIN1_ANNOTATED_1:.*]] = shard.sharding @grid_2 split_axes = {{\[}}[0]] : !shard.sharding
+  // CHECK-NEXT: %[[IN1_ANNOTATED_1:.*]] = shard.shard %[[IN1_ANNOTATED_0]] to %[[SIN1_ANNOTATED_1]] annotate_for_users : tensor<2x3xf32>
+  // CHECK: %[[SIN2_ANNOTATED:.*]] = shard.sharding @grid_2 split_axes = {{\[}}[]] : !shard.sharding
+  // CHECK-NEXT: %[[IN2_ANNOTATED:.*]] = shard.shard %[[IN2]] to %[[SIN2_ANNOTATED]] annotate_for_users : tensor<3x2xf32>
+  // CHECK: %[[SDPS_OUT_ANNOTATED:.*]] = shard.sharding @grid_2 split_axes = {{\[}}[0]] : !shard.sharding
+  // CHECK-NEXT: %[[DPS_OUT_ANNOTATED:.*]] = shard.shard %[[DPS_OUT]] to %[[SDPS_OUT_ANNOTATED]] annotate_for_users : tensor<2x2xf32>
+  %sarg0_sharded = shard.sharding @grid_2 split_axes = [[0]] : !shard.sharding
+  %arg0_sharded = shard.shard %arg0 to %sarg0_sharded : tensor<2x3xf32>
+
+  // CHECK: %[[RES:.*]] = linalg.matmul ins(%[[IN1_ANNOTATED_1]], %[[IN2_ANNOTATED]] : tensor<2x3xf32>, tensor<3x2xf32>)
+  // CHECK-SAME:  outs(%[[DPS_OUT_ANNOTATED]] : tensor<2x2xf32>) -> tensor<2x2xf32>
+  %res = linalg.matmul ins(%arg0_sharded, %arg1 : tensor<2x3xf32>, tensor<3x2xf32>)
+    outs(%out_dps : tensor<2x2xf32>) -> tensor<2x2xf32>
+
+  // CHECK: %[[SRES_ANNOTATED_0:.*]] = shard.sharding @grid_2 split_axes = {{\[}}[0]] : !shard.sharding
+  // CHECK-NEXT: %[[RES_ANNOTATED_0:.*]] = shard.shard %[[RES]] to %[[SRES_ANNOTATED_0]] : tensor<2x2xf32>
+  // CHECK: %[[SRES_ANNOTATED_1:.*]] = shard.sharding @grid_2 split_axes = {{\[}}[]] : !shard.sharding
+  // CHECK-NEXT: %[[RES_ANNOTATED_1:.*]] = shard.shard %[[RES_ANNOTATED_0]] to %[[SRES_ANNOTATED_1]] annotate_for_users : tensor<2x2xf32>
+  %sres_sharded = shard.sharding @grid_2 split_axes = [[]] : !shard.sharding
+  %res_sharded = shard.shard %res to %sres_sharded annotate_for_users : tensor<2x2xf32>
+
+  // CHECK: return %[[RES_ANNOTATED_1]] : tensor<2x2xf32>
+  return %res_sharded : tensor<2x2xf32>
+}
diff --git a/mlir/test/Dialect/Linalg/vectorization/linalg-ops.mlir b/mlir/test/Dialect/Linalg/vectorization/linalg-ops.mlir
index 98e8f50..d41d861 100644
--- a/mlir/test/Dialect/Linalg/vectorization/linalg-ops.mlir
+++ b/mlir/test/Dialect/Linalg/vectorization/linalg-ops.mlir
@@ -941,20 +941,17 @@ module attributes {transform.with_named_sequence} {
 
 // CHECK-LABEL: func @test_vectorize_dynamic_shapes_unpack
 // CHECK-SAME:      %[[ARG_0:.*]]: tensor<?x?xf32>,
+// CHECK-SAME:      %[[ARG_1:.*]]: tensor<?x?x16x2xf32>
 func.func @test_vectorize_dynamic_shapes_unpack(%arg0: tensor<?x?xf32>, %arg1: tensor<?x?x16x2xf32>) -> tensor<?x?xf32> {
 // CHECK: %[[C0:.*]] = arith.constant 0
-// CHECK: %[[DIM:.*]] = tensor.dim %arg0, %[[C0]] : tensor<?x?xf32>
-// CHECK: %[[C1:.*]] = arith.constant 1 : index
-// CHECK: %[[DIM0:.*]] = tensor.dim %arg0, %[[C1]] : tensor<?x?xf32>
-// CHECK: %[[CST:.*]] = arith.constant 0.000000e+00
 // CHECK: %[[C01:.*]] = arith.constant 0
 // CHECK: %[[C02:.*]] = arith.constant 0
-// CHECK: %[[DIM4:.*]] = tensor.dim %arg1, %[[C02]] : tensor<?x?x16x2xf32>
-// CHECK: %[[CNST14:.*]] = arith.constant 1
-// CHECK: %[[DIM6:.*]] = tensor.dim %arg1, %[[CNST14]] : tensor<?x?x16x2xf32>
+// CHECK: %[[DIM_0:.*]] = tensor.dim %[[ARG_1]], %[[C02]] : tensor<?x?x16x2xf32>
+// CHECK: %[[C1:.*]] = arith.constant 1
+// CHECK: %[[DIM6:.*]] = tensor.dim %[[ARG_1]], %[[C1]] : tensor<?x?x16x2xf32>
 // CHECK: %[[CNST16:.*]] = arith.constant 16 : index
 // CHECK: %[[CNST2:.*]] = arith.constant 2 : index
-// CHECK: %[[readMsk0:.*]] = vector.create_mask %[[DIM4]], %[[DIM6]], %[[CNST16]], %[[CNST2]] : vector<2x1x16x2xi1>
+// CHECK: %[[readMsk0:.*]] = vector.create_mask %[[DIM_0]], %[[DIM6]], %[[CNST16]], %[[CNST2]] : vector<2x1x16x2xi1>
 // CHECK: %[[read0:.*]] = vector.mask %[[readMsk0]] {{.*}} vector.transfer_read %{{.*}} : tensor<?x?x16x2xf32>, vector<2x1x16x2xf32> } : vector<2x1x16x2xi1> -> vector<2x1x16x2xf32>
 // CHECK: %[[trans0:.*]] = vector.transpose %[[read0]], [0, 3, 1, 2] : vector<2x1x16x2xf32> to vector<2x2x1x16xf32>
 // CHECK: %[[sc0:.*]] = vector.shape_cast %[[trans0]] : vector<2x2x1x16xf32> to vector<4x16xf32>
diff --git a/mlir/test/Dialect/Mesh/canonicalization.mlir b/mlir/test/Dialect/Mesh/canonicalization.mlir
deleted file mode 100644
index aff07bb..0000000
--- a/mlir/test/Dialect/Mesh/canonicalization.mlir
+++ /dev/null
@@ -1,248 +0,0 @@
-// RUN: mlir-opt --canonicalize %s | FileCheck %s
-
-mesh.mesh @mesh0(shape = 2x4)
-
-// CHECK-LABEL: func @all_reduce_empty_mesh_axes
-func.func @all_reduce_empty_mesh_axes(
-// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
-    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
-// CHECK-NOT: mesh.all_reduce
-  %0 = mesh.all_reduce %arg0 on @mesh0
-    mesh_axes = []
-    : tensor<4xf32> -> tensor<4xf32>
-// CHECK: return %[[ARG]]
-  return %0 : tensor<4xf32>
-}
-
-// CHECK-LABEL: func @all_reduce_empty_mesh_axes_different_return_type
-func.func @all_reduce_empty_mesh_axes_different_return_type(
-    %arg0 : tensor<4xf32>) -> tensor<4xf64> {
-// CHECK: mesh.all_reduce
-  %0 = mesh.all_reduce %arg0 on @mesh0
-// CHECK-NOT: mesh_axes
-    mesh_axes = []
-    : tensor<4xf32> -> tensor<4xf64>
-  return %0 : tensor<4xf64>
-}
-
-// CHECK-LABEL: func @all_reduce_default_reduction
-func.func @all_reduce_default_reduction(
-    %arg0 : tensor<4xf32>) -> tensor<4xf64> {
-  %0 = mesh.all_reduce %arg0 on @mesh0
-    mesh_axes = [0]
-// CHECK-NOT: reduction
-    reduction = sum
-    : tensor<4xf32> -> tensor<4xf64>
-  return %0 : tensor<4xf64>
-}
-
-// CHECK-LABEL: func @all_to_all_empty_mesh_axes
-func.func @all_to_all_empty_mesh_axes(
-// CHECK-SAME: %[[ARG:.*]]: tensor<8xf32>
-    %arg0 : tensor<8xf32>) -> tensor<8xf32> {
-// CHECK-NOT: mesh.all_to_all
-  %0 = mesh.all_to_all %arg0 on @mesh0
-    mesh_axes = []
-    split_axis = 0
-    concat_axis = 0
-    : tensor<8xf32> -> tensor<8xf32>
-// CHECK: return %[[ARG]]
-  return %0 : tensor<8xf32>
-}
-
-// CHECK-LABEL: func @all_gather_empty_mesh_axes
-func.func @all_gather_empty_mesh_axes(
-// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
-    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
-// CHECK-NOT: mesh.all_gather
-  %0 = mesh.all_gather %arg0 on @mesh0
-    mesh_axes = []
-    gather_axis = 0
-    : tensor<4xf32> -> tensor<4xf32>
-// CHECK: return %[[ARG]]
-  return %0 : tensor<4xf32>
-}
-
-// CHECK-LABEL: func @all_slice_empty_mesh_axes
-func.func @all_slice_empty_mesh_axes(
-// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
-    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
-// CHECK-NOT: mesh.scatter
-  %0 = mesh.all_slice %arg0 on @mesh0
-    mesh_axes = []
-    slice_axis = 0
-    : tensor<4xf32> -> tensor<4xf32>
-// CHECK: return %[[ARG]]
-  return %0 : tensor<4xf32>
-}
-
-// CHECK-LABEL: func @broadcast_empty_mesh_axes
-func.func @broadcast_empty_mesh_axes(
-// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
-    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
-// CHECK-NOT: mesh.broadcast
-  %0 = mesh.broadcast %arg0 on @mesh0
-    mesh_axes = []
-    root = []
-    : (tensor<4xf32>) -> tensor<4xf32>
-// CHECK: return %[[ARG]]
-  return %0 : tensor<4xf32>
-}
-
-// CHECK-LABEL: func @gather_empty_mesh_axes
-func.func @gather_empty_mesh_axes(
-// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
-    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
-// CHECK-NOT: mesh.gather
-  %0 = mesh.gather %arg0 on @mesh0
-    mesh_axes = []
-    gather_axis = 0
-    root = []
-    : (tensor<4xf32>) -> tensor<4xf32>
-// CHECK: return %[[ARG]]
-  return %0 : tensor<4xf32>
-}
-
-// CHECK-LABEL: func @receive_empty_mesh_axes
-func.func @receive_empty_mesh_axes(
-// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
-    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
-// CHECK-NOT: mesh.recv
-  %0 = mesh.recv %arg0 on @mesh0
-    mesh_axes = []
-    : (tensor<4xf32>) -> tensor<4xf32>
-// CHECK: return %[[ARG]]
-  return %0 : tensor<4xf32>
-}
-
-// CHECK-LABEL: func @reduce_empty_mesh_axes
-func.func @reduce_empty_mesh_axes(
-// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
-    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
-// CHECK-NOT: mesh.reduce
-  %0 = mesh.reduce %arg0 on @mesh0
-    mesh_axes = []
-    root = []
-    : (tensor<4xf32>) -> tensor<4xf32>
-// CHECK: return %[[ARG]]
-  return %0 : tensor<4xf32>
-}
-
-// CHECK-LABEL: func @reduce_scatter_empty_mesh_axes
-func.func @reduce_scatter_empty_mesh_axes(
-// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
-    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
-// CHECK-NOT: mesh.reduce_scatter
-  %0 = mesh.reduce_scatter %arg0 on @mesh0
-    mesh_axes = []
-    scatter_axis = 0
-    : tensor<4xf32> -> tensor<4xf32>
-// CHECK: return %[[ARG]]
-  return %0 : tensor<4xf32>
-}
-
-// CHECK-LABEL: func @reduce_scatter_empty_mesh_axes_different_return_type
-func.func @reduce_scatter_empty_mesh_axes_different_return_type(
-    %arg0 : tensor<4xf32>) -> tensor<4xf64> {
-// CHECK: mesh.reduce_scatter
-  %0 = mesh.reduce_scatter %arg0 on @mesh0
-// CHECK-NOT: mesh_axes
-    mesh_axes = []
-    scatter_axis = 0
-    : tensor<4xf32> -> tensor<4xf64>
-  return %0 : tensor<4xf64>
-}
-
-// CHECK-LABEL: func @reduce_scatter_default_reduction
-func.func @reduce_scatter_default_reduction(
-    %arg0 : tensor<4xf32>) -> tensor<2xf64> {
-  %0 = mesh.reduce_scatter %arg0 on @mesh0
-    mesh_axes = [0]
-// CHECK-NOT: reduction
-    reduction = sum
-    scatter_axis = 0
-    : tensor<4xf32> -> tensor<2xf64>
-  return %0 : tensor<2xf64>
-}
-
-// CHECK-LABEL: func @scatter_empty_mesh_axes
-func.func @scatter_empty_mesh_axes(
-// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
-    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
-// CHECK-NOT: mesh.scatter
-  %0 = mesh.scatter %arg0 on @mesh0
-    mesh_axes = []
-    scatter_axis = 0
-    root = []
-    : (tensor<4xf32>) -> tensor<4xf32>
-// CHECK: return %[[ARG]]
-  return %0 : tensor<4xf32>
-}
-
-// CHECK-LABEL: func @send_empty_mesh_axes
-func.func @send_empty_mesh_axes(
-// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
-    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
-// CHECK-NOT: mesh.send
-  %0 = mesh.send %arg0 on @mesh0
-    mesh_axes = []
-    destination = []
-    : (tensor<4xf32>) -> tensor<4xf32>
-// CHECK: return %[[ARG]]
-  return %0 : tensor<4xf32>
-}
-
-mesh.mesh @mesh4x4(shape = 4x4)
-// CHECK-LABEL: func @test_halo_sizes
-func.func @test_halo_sizes() -> !mesh.sharding {
-  %c2_i64 = arith.constant 2 : i64
-  // CHECK mesh.sharding @mesh4x4 split_axes = [[0], [1]] halo_sizes = [1, 2, 2, 22] : !mesh.sharding
-  %sharding = mesh.sharding @mesh4x4 split_axes = [[0], [1]] halo_sizes = [1, %c2_i64, %c2_i64, 22] : !mesh.sharding
-  return %sharding : !mesh.sharding
-}
-
-// CHECK-LABEL: func @test_shard_offs
-func.func @test_shard_offs() -> !mesh.sharding {
-  %c2_i64 = arith.constant 2 : i64
-  // CHECK mesh.sharding @mesh4x4 split_axes = [[0], [1]] sharded_dims_offsets = [0, 1, 2, 3, 4, 0, 2, 3, 4, 22] : !mesh.sharding
-  %sharding = mesh.sharding @mesh4x4 split_axes = [[0], [1]] sharded_dims_offsets = [0, 1, %c2_i64, 3, 4, 0, %c2_i64, 3, 4, 22] : !mesh.sharding
-  return %sharding : !mesh.sharding
-}
-
-// CHECK-LABEL: func @test_duplicate_shardops
-func.func @test_duplicate_shardops() -> (tensor<1024x1024xf32>, tensor<1024x1024xf32>) attributes {llvm.emit_c_interface} {
-  // CHECK-NEXT: [[vcst:%.*]] = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
-  %cst_1 = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
-  // CHECK-NEXT: [[vsharding:%.*]] = mesh.sharding @mesh4x4 split_axes = {{\[\[}}0, 1]] : !mesh.sharding
-  %sharding_1 = mesh.sharding @mesh4x4 split_axes = [[0, 1]] : !mesh.sharding
-  %cst_2 = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
-  %sharding_2 = mesh.sharding @mesh4x4 split_axes = [[0, 1]] : !mesh.sharding
-  %sharding_annotated_2 = mesh.shard %cst_2 to %sharding_2 : tensor<1024x1024xf32>
-  %cst_3 = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
-  %sharding_3 = mesh.sharding @mesh4x4 split_axes = [[0, 1]] : !mesh.sharding
-  %sharding_annotated_3 = mesh.shard %cst_3 to %sharding_3 : tensor<1024x1024xf32>
-  // CHECK-NEXT: [[vsharding_annotated:%.*]] = mesh.shard [[vcst]] to [[vsharding]] : tensor<1024x1024xf32>
-  %sharding_annotated_1 = mesh.shard %cst_1 to %sharding_1 : tensor<1024x1024xf32>
-  // CHECK-NEXT: return [[vsharding_annotated]], [[vsharding_annotated]] : tensor<1024x1024xf32>, tensor<1024x1024xf32>
-  return %sharding_annotated_1, %sharding_annotated_2 : tensor<1024x1024xf32>, tensor<1024x1024xf32>
-}
-
-// CHECK-LABEL: func @test_duplicate_shardops_diff
-func.func @test_duplicate_shardops_diff() -> (tensor<1024x1024xf32>, tensor<1024x1024xf32>) attributes {llvm.emit_c_interface} {
-  // CHECK-NEXT: [[vcst:%.*]] = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
-  %cst_1 = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
-  // CHECK-NEXT: [[vsharding:%.*]] = mesh.sharding @mesh4x4 split_axes = {{\[\[}}0]] : !mesh.sharding
-  %sharding_1 = mesh.sharding @mesh4x4 split_axes = [[0]] : !mesh.sharding
-  %cst_2 = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
-  // CHECK-NEXT: [[vsharding_0:%.*]] = mesh.sharding @mesh4x4 split_axes = {{\[\[}}0, 1]] : !mesh.sharding
-  %sharding_2 = mesh.sharding @mesh4x4 split_axes = [[0, 1]] : !mesh.sharding
-  // CHECK-NEXT: [[vsharding_annotated:%.*]] = mesh.shard [[vcst]] to [[vsharding_0]] : tensor<1024x1024xf32>
-  %sharding_annotated_2 = mesh.shard %cst_2 to %sharding_2 : tensor<1024x1024xf32>
-  %cst_3 = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
-  %sharding_3 = mesh.sharding @mesh4x4 split_axes = [[0]] : !mesh.sharding
-  %sharding_annotated_3 = mesh.shard %cst_3 to %sharding_3 : tensor<1024x1024xf32>
-  // CHECK-NEXT: [[vsharding_annotated_1:%.*]] = mesh.shard [[vsharding_annotated]] to [[vsharding]] : tensor<1024x1024xf32>
-  %sharding_annotated_1 = mesh.shard %cst_1 to %sharding_1 : tensor<1024x1024xf32>
-  // CHECK-NEXT: return [[vsharding_annotated_1]], [[vsharding_annotated]] : tensor<1024x1024xf32>, tensor<1024x1024xf32>
-  return %sharding_annotated_1, %sharding_annotated_2 : tensor<1024x1024xf32>, tensor<1024x1024xf32>
-}
diff --git a/mlir/test/Dialect/Mesh/folding.mlir b/mlir/test/Dialect/Mesh/folding.mlir
deleted file mode 100644
index 369f316d..0000000
--- a/mlir/test/Dialect/Mesh/folding.mlir
+++ /dev/null
@@ -1,22 +0,0 @@
-// RUN: mlir-opt -test-mesh-simplifications %s | FileCheck %s
-
-mesh.mesh @mesh0(shape = 4x?x2)
-mesh.mesh @mesh1(shape = 2x3)
-
-// CHECK-LABEL: func.func @mesh_shape_op_folding
-func.func @mesh_shape_op_folding() -> (index, index) {
-  // CHECK: %[[AXIS_2_SIZE:.*]] = arith.constant 2 : index
-  // CHECK: %[[AXIS_1_SIZE:.*]] = mesh.mesh_shape @mesh0 axes = [1] : index
-  %0:2 = mesh.mesh_shape @mesh0 axes = [2, 1] : index, index
-  // CHECK: return %[[AXIS_2_SIZE]], %[[AXIS_1_SIZE]]
-  return %0#0, %0#1 : index, index
-}
-
-// CHECK-LABEL: func.func @mesh_shape_op_folding_all_axes_static_mesh
-func.func @mesh_shape_op_folding_all_axes_static_mesh() -> (index, index) {
-  // CHECK: %[[AXIS_0_SIZE:.*]] = arith.constant 2 : index
-  // CHECK: %[[AXIS_1_SIZE:.*]] = arith.constant 3 : index
-  %0:2 = mesh.mesh_shape @mesh1 : index, index
-  // CHECK: return %[[AXIS_0_SIZE]], %[[AXIS_1_SIZE]]
-  return %0#0, %0#1 : index, index
-}
diff --git a/mlir/test/Dialect/Mesh/forward-sharding-propagation.mlir b/mlir/test/Dialect/Mesh/forward-sharding-propagation.mlir
deleted file mode 100644
index 6ab711b..0000000
--- a/mlir/test/Dialect/Mesh/forward-sharding-propagation.mlir
+++ /dev/null
@@ -1,49 +0,0 @@
-// RUN: mlir-opt --pass-pipeline="builtin.module(func.func(sharding-propagation{traversal=forward}))" %s | FileCheck %s
-
-#map = affine_map<(d0, d1) -> (d0, d1)>
-module attributes {dlti.map = #dlti.map<"MPI:Implementation" = "mpich", "MPI:comm_world_rank" = 0 : i32>} {
-  mesh.mesh @mesh(shape = 1) {sym_visibility = "private"}
-  func.func @test_forward() -> (tensor<6x6xi32>, tensor<6x6xi32>, tensor<i32>) attributes {llvm.emit_c_interface} {
-    %c1_i32 = arith.constant 1 : i32
-    // CHECK: [[v3:%.*]] = tensor.empty() : tensor<6x6xi32>
-    %0 = tensor.empty() : tensor<6x6xi32>
-    // CHECK: [[v1:%.*]] = linalg.fill ins
-    // CHECK: [[vsharding_0:%.*]] = mesh.sharding @mesh split_axes = {{\[\[}}0]] : !mesh.sharding
-    // CHECK: [[vsharding_annotated_1:%.*]] = mesh.shard [[v1]] to [[vsharding_0]] : tensor<6x6xi32>
-    %1 = linalg.fill ins(%c1_i32 : i32) outs(%0 : tensor<6x6xi32>) -> tensor<6x6xi32>
-    %sharding = mesh.sharding @mesh split_axes = [[0]] : !mesh.sharding
-    %sharding_annotated = mesh.shard %1 to %sharding : tensor<6x6xi32>
-    // CHECK: [[v2:%.*]] = tensor.empty() : tensor<6x6xi32>
-    // CHECK: [[vsharding_2:%.*]] = mesh.sharding @mesh split_axes = {{\[\[}}0]] : !mesh.sharding
-    // CHECK: [[vsharding_annotated_3:%.*]] = mesh.shard [[vsharding_annotated_1]] to [[vsharding_2]] annotate_for_users : tensor<6x6xi32>
-    %3 = tensor.empty() : tensor<6x6xi32>
-    // CHECK: [[vsharding_4:%.*]] = mesh.sharding @mesh split_axes = {{\[\[}}0]] : !mesh.sharding
-    // CHECK: [[vsharding_annotated_5:%.*]] = mesh.shard [[v2]] to [[vsharding_4]] annotate_for_users : tensor<6x6xi32>
-    // CHECK: [[v3:%.*]] = linalg.generic {indexing_maps = [#map, #map, #map], iterator_types = ["parallel", "parallel"]}
-    // CHECK-SAME: ins([[vsharding_annotated_3]], [[vsharding_annotated_3]] : tensor<6x6xi32>, tensor<6x6xi32>) outs([[vsharding_annotated_5]] : tensor<6x6xi32>) {
-    // CHECK: [[vsharding_6:%.*]] = mesh.sharding @mesh split_axes = {{\[\[}}0]] : !mesh.sharding
-    // CHECK: [[vsharding_annotated_7:%.*]] = mesh.shard [[v3]] to [[vsharding_6]] : tensor<6x6xi32>
-    %4 = linalg.generic {indexing_maps = [#map, #map, #map], iterator_types = ["parallel", "parallel"]} ins(%sharding_annotated, %sharding_annotated
-        : tensor<6x6xi32>, tensor<6x6xi32>) outs(%3 : tensor<6x6xi32>) {
-    ^bb0(%in: i32, %in_2: i32, %out: i32):
-      %9 = arith.addi %in, %in_2 : i32
-      linalg.yield %9 : i32
-    } -> tensor<6x6xi32>
-    %c0_i32 = arith.constant 0 : i32
-    %6 = tensor.empty() : tensor<i32>
-    %7 = linalg.fill ins(%c0_i32 : i32) outs(%6 : tensor<i32>) -> tensor<i32>
-    // CHECK: [[vreduced:%.*]] = linalg.reduce ins
-    // CHECK: [[vsharding_12:%.*]] = mesh.sharding @mesh split_axes = [] : !mesh.sharding
-    // CHECK: [[vsharding_annotated_13:%.*]] = mesh.shard [[vreduced]] to [[vsharding_12]] : tensor<i32>
-    %reduced = linalg.reduce ins(%4 : tensor<6x6xi32>) outs(%7 : tensor<i32>) dimensions = [0, 1] 
-      (%in: i32, %init: i32) {
-        %9 = arith.addi %in, %init : i32
-        linalg.yield %9 : i32
-      }
-    // CHECK: [[vsharding_14:%.*]] = mesh.sharding @mesh split_axes = {{\[\[}}]] : !mesh.sharding
-    %sharding_0 = mesh.sharding @mesh split_axes = [[]] : !mesh.sharding
-    // CHECK: [[vsharding_annotated_15:%.*]] = mesh.shard [[vsharding_annotated_13]] to [[vsharding_14]] annotate_for_users : tensor<i32>
-    %sharding_annotated_1 = mesh.shard %reduced to %sharding_0 annotate_for_users : tensor<i32>
-    return %sharding_annotated, %4, %sharding_annotated_1 : tensor<6x6xi32>, tensor<6x6xi32>, tensor<i32>
-  }
-}
diff --git a/mlir/test/Dialect/Mesh/inlining.mlir b/mlir/test/Dialect/Mesh/inlining.mlir
deleted file mode 100644
index c41a709..0000000
--- a/mlir/test/Dialect/Mesh/inlining.mlir
+++ /dev/null
@@ -1,15 +0,0 @@
-// RUN: mlir-opt -inline %s | FileCheck %s
-
-mesh.mesh @mesh0(shape = 4x?x2)
-
-func.func private @mesh_to_inline() -> (index, index) {
-  %0:2 = mesh.mesh_shape @mesh0 axes = [2, 1] : index, index
-  return %0#0, %0#1 : index, index
-}
-// CHECK-LABEL: func.func @main
-func.func @main() -> (index, index) {
-  // CHECK-NEXT: %[[AXIS_SIZE:.*]]:2 = mesh.mesh_shape @mesh0 axes = [2, 1] : index
-  %0:2 = func.call @mesh_to_inline() : () -> (index, index)
-  // CHECK-NEXT: return %[[AXIS_SIZE]]#0, %[[AXIS_SIZE]]#1
-  return %0#0, %0#1 : index, index
-}
diff --git a/mlir/test/Dialect/Mesh/process-multi-index-op-lowering.mlir b/mlir/test/Dialect/Mesh/process-multi-index-op-lowering.mlir
deleted file mode 100644
index e23cfd7..0000000
--- a/mlir/test/Dialect/Mesh/process-multi-index-op-lowering.mlir
+++ /dev/null
@@ -1,23 +0,0 @@
-// RUN: mlir-opt -test-mesh-process-multi-index-op-lowering %s | FileCheck %s
-
-mesh.mesh @mesh2d(shape = ?x?)
-
-// CHECK-LABEL: func.func @multi_index_2d_mesh
-func.func @multi_index_2d_mesh() -> (index, index) {
-  // CHECK: %[[LINEAR_IDX:.*]] = mesh.process_linear_index on @mesh2d : index
-  // CHECK: %[[MESH_SHAPE:.*]]:2 = mesh.mesh_shape @mesh2d : index, index
-  // CHECK: %[[MULTI_IDX:.*]]:2 = affine.delinearize_index %[[LINEAR_IDX]] into (%[[MESH_SHAPE]]#0, %[[MESH_SHAPE]]#1) : index, index
-  %0:2 = mesh.process_multi_index on @mesh2d : index, index
-  // CHECK: return %[[MULTI_IDX]]#0, %[[MULTI_IDX]]#1 : index, index
-  return %0#0, %0#1 : index, index
-}
-
-// CHECK-LABEL: func.func @multi_index_2d_mesh_single_inner_axis
-func.func @multi_index_2d_mesh_single_inner_axis() -> index {
-  // CHECK: %[[LINEAR_IDX:.*]] = mesh.process_linear_index on @mesh2d : index
-  // CHECK: %[[MESH_SHAPE:.*]]:2 = mesh.mesh_shape @mesh2d : index, index
-  // CHECK: %[[MULTI_IDX:.*]]:2 = affine.delinearize_index %[[LINEAR_IDX]] into (%[[MESH_SHAPE]]#0, %[[MESH_SHAPE]]#1) : index, index
-  %0 = mesh.process_multi_index on @mesh2d axes = [0] : index
-  // CHECK: return %[[MULTI_IDX]]#0 : index
-  return %0 : index
-}
diff --git a/mlir/test/Dialect/Mesh/resharding-spmdization.mlir b/mlir/test/Dialect/Mesh/resharding-spmdization.mlir
deleted file mode 100644
index 5e62c92..0000000
--- a/mlir/test/Dialect/Mesh/resharding-spmdization.mlir
+++ /dev/null
@@ -1,168 +0,0 @@
-// RUN: mlir-opt -test-mesh-resharding-spmdization %s | FileCheck %s
-
-mesh.mesh @mesh_1d(shape = 2)
-mesh.mesh @mesh_1d_dynamic(shape = ?)
-
-// CHECK-LABEL: func @same_source_and_target_sharding
-func.func @same_source_and_target_sharding(
-  // CHECK-SAME: %[[ARG:.*]]: tensor<2xf32>
-  %arg0: tensor<2xf32>
-) -> tensor<2xf32> {
-  %s0 = mesh.sharding @mesh_1d split_axes = [[]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0 : tensor<2xf32>
-  %s1 = mesh.sharding @mesh_1d split_axes = [[]] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1 annotate_for_users : tensor<2xf32>
-  // CHECK: return %[[ARG]]
-  return %1 : tensor<2xf32>
-}
-
-// CHECK-LABEL: func @identical_source_and_target_sharding
-func.func @identical_source_and_target_sharding(
-  // CHECK-SAME: %[[ARG:.*]]: tensor<2xf32>
-  %arg0: tensor<2xf32>
-) -> tensor<2xf32> {
-  %s0 = mesh.sharding @mesh_1d split_axes = [[]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0 : tensor<2xf32>
-  %1 = mesh.shard %0 to %s0 annotate_for_users : tensor<2xf32>
-  // CHECK: return %[[ARG]]
-  return %1 : tensor<2xf32>
-}
-
-// CHECK-LABEL: func @split_replicated_tensor_axis
-func.func @split_replicated_tensor_axis(
-  // CHECK-SAME: %[[ARG:.*]]: tensor<3x14xf32>
-  %arg0: tensor<3x14xf32>
-) -> tensor<3x14xf32> {
-  // CHECK: %[[ALL_SLICE:.*]] = mesh.all_slice %[[ARG]] on @mesh_1d mesh_axes = [0] slice_axis = 1
-  // CHECK-SAME: tensor<3x14xf32> -> tensor<3x7xf32>
-  // CHECK: %[[RESULT:.*]] = builtin.unrealized_conversion_cast %[[ALL_SLICE]] : tensor<3x7xf32> to tensor<3x14xf32>
-  %s0 = mesh.sharding @mesh_1d split_axes = [[]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0 : tensor<3x14xf32>
-  %s1 = mesh.sharding @mesh_1d split_axes = [[], [0]] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1 annotate_for_users : tensor<3x14xf32>
-  // CHECK: return %[[RESULT]] : tensor<3x14xf32>
-  return %1 : tensor<3x14xf32>
-}
-
-// CHECK-LABEL: func @split_replicated_tensor_axis_dynamic
-func.func @split_replicated_tensor_axis_dynamic(
-  // CHECK-SAME: %[[ARG:.*]]: tensor<?x3x?xf32>
-  %arg0: tensor<?x3x?xf32>
-) -> tensor<?x3x?xf32> {
-  // CHECK: %[[RESULT:.*]] = mesh.all_slice %[[ARG]] on @mesh_1d_dynamic mesh_axes = [0] slice_axis = 0
-  // CHECK-SAME: tensor<?x3x?xf32> -> tensor<?x3x?xf32>
-  %s0 = mesh.sharding @mesh_1d_dynamic split_axes = [[], [], []] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0 : tensor<?x3x?xf32>
-  %s1 = mesh.sharding @mesh_1d_dynamic split_axes = [[0]] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1 annotate_for_users : tensor<?x3x?xf32>
-  // CHECK: return %[[RESULT]] : tensor<?x3x?xf32>
-  return %1 : tensor<?x3x?xf32>
-}
-
-// CHECK-LABEL: func @move_split_axis
-func.func @move_split_axis(
-  // CHECK-SAME: %[[ARG:.*]]: tensor<10x14xf32>
-  %arg0: tensor<10x14xf32>
-) -> tensor<10x14xf32> {
-  // CHECK: %[[SOURCE_SHARD:.*]] = builtin.unrealized_conversion_cast %[[ARG]] : tensor<10x14xf32> to tensor<5x14xf32>
-  // CHECK: %[[TARGET_SHARD:.*]] = mesh.all_to_all %[[SOURCE_SHARD]] on @mesh_1d mesh_axes = [0] split_axis = 1 concat_axis = 0 : tensor<5x14xf32> -> tensor<10x7xf32>
-  // CHECK: %[[RES:.*]] = builtin.unrealized_conversion_cast %[[TARGET_SHARD]] : tensor<10x7xf32> to tensor<10x14xf32>
-  %s0 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0 : tensor<10x14xf32>
-  %s1 = mesh.sharding @mesh_1d split_axes = [[], [0]] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1 annotate_for_users : tensor<10x14xf32>
-  // CHECK: return %[[RES]] : tensor<10x14xf32>
-  return %1 : tensor<10x14xf32>
-}
-
-// CHECK-LABEL: func @move_split_axis_dynamic_mesh
-func.func @move_split_axis_dynamic_mesh(
-  // CHECK-SAME: %[[ARG:.*]]: tensor<10x14xf32>
-  %arg0: tensor<10x14xf32>
-) -> tensor<10x14xf32> {
-  // CHECK: %[[SOURCE_SHARD:.*]] = builtin.unrealized_conversion_cast %[[ARG]] : tensor<10x14xf32> to tensor<?x14xf32>
-  // CHECK: %[[ALL_TO_ALL:.*]] = mesh.all_to_all %[[SOURCE_SHARD]] on @mesh_1d_dynamic mesh_axes = [0] split_axis = 1 concat_axis = 0 : tensor<?x14xf32> -> tensor<?x?xf32>
-  // CHECK: %[[TARGET_SHARD:.*]] = tensor.cast %[[ALL_TO_ALL]] : tensor<?x?xf32> to tensor<10x?xf32>
-  // CHECK: %[[RES:.*]] = builtin.unrealized_conversion_cast %[[TARGET_SHARD]] : tensor<10x?xf32> to tensor<10x14xf32>
-  %s0 = mesh.sharding @mesh_1d_dynamic split_axes = [[0]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0 : tensor<10x14xf32>
-  %s1 = mesh.sharding @mesh_1d_dynamic split_axes = [[], [0]] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1 annotate_for_users : tensor<10x14xf32>
-  // CHECK: return %[[RES]] : tensor<10x14xf32>
-  return %1 : tensor<10x14xf32>
-}
-
-// CHECK-LABEL: func @move_split_dynamic_axis
-func.func @move_split_dynamic_axis(
-  // CHECK-SAME: %[[ARG:.*]]: tensor<?x14xf32>
-  %arg0: tensor<?x14xf32>
-) -> tensor<?x14xf32> {
-  // CHECK: %[[TARGET_SHARD:.*]] = mesh.all_to_all %[[ARG]] on @mesh_1d mesh_axes = [0] split_axis = 1 concat_axis = 0 : tensor<?x14xf32> -> tensor<?x7xf32>
-  // CHECK: %[[RES:.*]] = builtin.unrealized_conversion_cast %[[TARGET_SHARD]] : tensor<?x7xf32> to tensor<?x14xf32>
-  %s0 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0 : tensor<?x14xf32>
-  %s1 = mesh.sharding @mesh_1d split_axes = [[], [0]] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1 annotate_for_users : tensor<?x14xf32>
-  // CHECK: return %[[RES]] : tensor<?x14xf32>
-  return %1 : tensor<?x14xf32>
-}
-
-// CHECK-LABEL: func @unshard_static_axis
-func.func @unshard_static_axis(
-  // CHECK-SAME: %[[ARG:.*]]: tensor<10x14xf32>
-  %arg0: tensor<10x14xf32>
-) -> tensor<10x14xf32> {
-  // CHECK: %[[SOURCE_SHARD:.*]] = builtin.unrealized_conversion_cast %[[ARG]] : tensor<10x14xf32> to tensor<5x14xf32>
-  // CHECK: %[[ALL_GATHER:.*]] = mesh.all_gather %[[SOURCE_SHARD]] on @mesh_1d mesh_axes = [0] gather_axis = 0 : tensor<5x14xf32> -> tensor<10x14xf32>
-  %s0 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0 : tensor<10x14xf32>
-  %s1 = mesh.sharding @mesh_1d split_axes = [[]] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1 annotate_for_users : tensor<10x14xf32>
-  // CHECK: return %[[ALL_GATHER]] : tensor<10x14xf32>
-  return %1 : tensor<10x14xf32>
-}
-
-// CHECK-LABEL: func @unshard_static_last_axis
-func.func @unshard_static_last_axis(
-  // CHECK-SAME: %[[ARG:.*]]: tensor<10x14xf32>
-  %arg0: tensor<10x14xf32>
-) -> tensor<10x14xf32> {
-  // CHECK: %[[SOURCE_SHARD:.*]] = builtin.unrealized_conversion_cast %[[ARG]] : tensor<10x14xf32> to tensor<10x7xf32>
-  // CHECK: %[[ALL_GATHER:.*]] = mesh.all_gather %[[SOURCE_SHARD]] on @mesh_1d mesh_axes = [0] gather_axis = 1 : tensor<10x7xf32> -> tensor<10x14xf32>
-  %s0 = mesh.sharding @mesh_1d split_axes = [[], [0]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0 : tensor<10x14xf32>
-  %s1 = mesh.sharding @mesh_1d split_axes = [[], []] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1 annotate_for_users : tensor<10x14xf32>
-  // CHECK: return %[[ALL_GATHER]] : tensor<10x14xf32>
-  return %1 : tensor<10x14xf32>
-}
-
-// CHECK-LABEL: func @unshard_dynamic_axis
-func.func @unshard_dynamic_axis(
-  // CHECK-SAME: %[[ARG:.*]]: tensor<?x14xf32>
-  %arg0: tensor<?x14xf32>
-) -> tensor<?x14xf32> {
-  // CHECK: %[[ALL_GATHER:.*]] = mesh.all_gather %[[ARG]] on @mesh_1d mesh_axes = [0] gather_axis = 0 : tensor<?x14xf32> -> tensor<?x14xf32>
-  %s0 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0 : tensor<?x14xf32>
-  %s1 = mesh.sharding @mesh_1d split_axes = [[]] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1 annotate_for_users : tensor<?x14xf32>
-  // CHECK: return %[[ALL_GATHER]] : tensor<?x14xf32>
-  return %1 : tensor<?x14xf32>
-}
-
-// CHECK-LABEL: func @unshard_static_axis_on_dynamic_mesh_axis
-func.func @unshard_static_axis_on_dynamic_mesh_axis(
-// CHECK-SAME: %[[ARG:.*]]: tensor<10x14xf32>  
-  %arg0: tensor<10x14xf32>
-) -> tensor<10x14xf32> {
-  // CHECK: %[[SOURCE_SHARD:.*]] = builtin.unrealized_conversion_cast %[[ARG]] : tensor<10x14xf32> to tensor<?x14xf32>
-  // CHECK: %[[ALL_GATHER:.*]] = mesh.all_gather %[[SOURCE_SHARD]] on @mesh_1d_dynamic mesh_axes = [0] gather_axis = 0 : tensor<?x14xf32> -> tensor<?x14xf32>
-  // CHECK: %[[RES:.*]] = tensor.cast %[[ALL_GATHER]] : tensor<?x14xf32> to tensor<10x14xf32>
-  %s0 = mesh.sharding @mesh_1d_dynamic split_axes = [[0]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0 : tensor<10x14xf32>
-  %s1 = mesh.sharding @mesh_1d_dynamic split_axes = [[]] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1 annotate_for_users : tensor<10x14xf32>
-  // CHECK: return %[[RES]] : tensor<10x14xf32>
-  return %1 : tensor<10x14xf32>
-}
diff --git a/mlir/test/Dialect/Mesh/sharding-propagation.mlir b/mlir/test/Dialect/Mesh/sharding-propagation.mlir
deleted file mode 100644
index 0881d994..0000000
--- a/mlir/test/Dialect/Mesh/sharding-propagation.mlir
+++ /dev/null
@@ -1,301 +0,0 @@
-// RUN: mlir-opt --pass-pipeline="builtin.module(func.func(sharding-propagation,cse))" %s | FileCheck %s
-
-mesh.mesh @mesh_2(shape = 2)
-mesh.mesh @mesh_1d(shape = ?)
-mesh.mesh @mesh_2d(shape = 2x4)
-mesh.mesh @mesh_3d(shape = ?x?x?)
-
-// CHECK-LABEL: func.func @element_wise_empty_sharding_info
-func.func @element_wise_empty_sharding_info(%arg0: tensor<8x16xf32>) -> tensor<8x16xf32> {
-  // CHECK-NEXT: tosa.sigmoid
-  %0 = tosa.sigmoid %arg0 : (tensor<8x16xf32>) -> tensor<8x16xf32>
-  // CHECK-NEXT: return
-  return %0 : tensor<8x16xf32>
-}
-
-// CHECK-LABEL: func.func @element_wise_on_def
-// CHECK-SAME:    %[[ARG:.*]]: tensor<8x16xf32>
-func.func @element_wise_on_def(%arg0: tensor<8x16xf32>) -> tensor<8x16xf32> {
-  // CHECK-NEXT:  %[[S0:.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}0], [1]] : !mesh.sharding
-  // CHECK-NEXT:  %[[V0:.*]] = mesh.shard %[[ARG]] to %[[S0]] annotate_for_users  : tensor<8x16xf32>
-  // CHECK-NEXT:  %[[V1:.*]] = tosa.sigmoid %[[V0]]
-  %0 = tosa.sigmoid %arg0 : (tensor<8x16xf32>) -> tensor<8x16xf32>
-  // CHECK-NEXT:  %[[V2:.*]] = mesh.shard %[[V1]] to %[[S0]]  : tensor<8x16xf32>
-  %s1 = mesh.sharding @mesh_2d split_axes = [[0], [1]] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1  : tensor<8x16xf32>
-  // CHECK-NEXT:  return %[[V2]]
-  return %1 : tensor<8x16xf32>
-}
-
-// CHECK-LABEL: func.func @element_wise_on_use
-// CHECK-SAME:    %[[ARG:.*]]: tensor<8x16xf32>
-func.func @element_wise_on_use(%arg0: tensor<8x16xf32>) -> tensor<8x16xf32> {
-  // CHECK-NEXT:  %[[S0:.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}0], [1]] : !mesh.sharding
-  // CHECK-NEXT:  %[[V0:.*]] = mesh.shard %[[ARG]] to %[[S0]] annotate_for_users  : tensor<8x16xf32>
-  %s0 = mesh.sharding @mesh_2d split_axes = [[0], [1]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0 annotate_for_users  : tensor<8x16xf32>
-  // CHECK-NEXT:  %[[V1:.*]] = tosa.sigmoid %[[V0]]
-  %1 = tosa.sigmoid %0 : (tensor<8x16xf32>) -> tensor<8x16xf32>
-  // CHECK-NEXT:  %[[V2:.*]] = mesh.shard %[[V1]] to %[[S0]]  : tensor<8x16xf32>
-  // CHECK-NEXT:  return %[[V2]]
-  return %1 : tensor<8x16xf32>
-}
-
-// CHECK-LABEL: func.func @element_wise_on_graph_output
-// CHECK-SAME:    %[[ARG:.*]]: tensor<8x16xf32>
-func.func @element_wise_on_graph_output(%arg0: tensor<8x16xf32>) -> tensor<8x16xf32> {
-  // CHECK-NEXT:  %[[S0:.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}0], [1]] : !mesh.sharding
-  // CHECK-NEXT:  %[[V0:.*]] = mesh.shard %[[ARG]] to %[[S0]] annotate_for_users  : tensor<8x16xf32>
-  // CHECK-NEXT:  %[[V1:.*]] = tosa.sigmoid %[[V0]]
-  %0 = tosa.sigmoid %arg0 : (tensor<8x16xf32>) -> tensor<8x16xf32>
-  // CHECK-NEXT:  %[[V2:.*]] = mesh.shard %[[V1]] to %[[S0]]  : tensor<8x16xf32>
-  // CHECK-NEXT:  %[[V3:.*]] = mesh.shard %[[V2]] to %[[S0]] annotate_for_users  : tensor<8x16xf32>
-  %s1 = mesh.sharding @mesh_2d split_axes = [[0], [1]] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1 annotate_for_users  : tensor<8x16xf32>
-  // CHECK-NEXT:  return %[[V3]]
-  return %1 : tensor<8x16xf32>
-}
-
-// CHECK-LABEL: func.func @element_wise_on_graph_input
-// CHECK-SAME:    %[[ARG:.*]]: tensor<8x16xf32>
-func.func @element_wise_on_graph_input(%arg0: tensor<8x16xf32>) -> tensor<8x16xf32> {
-  // CHECK-NEXT:  %[[S0:.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}0], [1]] : !mesh.sharding
-  // CHECK-NEXT:  %[[V0:.*]] = mesh.shard %[[ARG]] to %[[S0]]  : tensor<8x16xf32>
-  // CHECK-NEXT:  %[[V1:.*]] = mesh.shard %[[V0]] to %[[S0]] annotate_for_users  : tensor<8x16xf32>
-  %s0 = mesh.sharding @mesh_2d split_axes = [[0], [1]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0  : tensor<8x16xf32>
-  // CHECK-NEXT:  %[[V2:.*]] = tosa.sigmoid %[[V1]]
-  %1 = tosa.sigmoid %0 : (tensor<8x16xf32>) -> tensor<8x16xf32>
-  // CHECK-NEXT:  %[[V3:.*]] = mesh.shard %[[V2]] to %[[S0]]  : tensor<8x16xf32>
-  // CHECK-NEXT:  return %[[V3]]
-  return %1 : tensor<8x16xf32>
-}
-
-// CHECK-LABEL: func.func @arrow_structure
-// CHECK-SAME:    %[[ARG:.*]]: tensor<8x16xf32>
-func.func @arrow_structure(%arg0: tensor<8x16xf32>, %arg1: tensor<1xf32>, %arg2: tensor<1xf32>) -> (tensor<8x16xf32>, tensor<8x16xf32>) {
-  // CHECK-NEXT:  %[[S1:.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}0], [1]] : !mesh.sharding
-  // CHECK-NEXT:  %[[V1:.*]] = mesh.shard %[[ARG]] to %[[S1]] annotate_for_users  : tensor<8x16xf32>
-  // CHECK-NEXT:  %[[V2:.*]] = tosa.tanh %[[V1]]
-  // CHECK-NEXT:  %[[V3:.*]] = mesh.shard %[[V2]] to %[[S1]]  : tensor<8x16xf32>
-  %0 = tosa.tanh %arg0 : (tensor<8x16xf32>) -> tensor<8x16xf32>
-  // CHECK-NEXT:  %[[V4:.*]] = mesh.shard %[[V3]] to %[[S1]] annotate_for_users  : tensor<8x16xf32>
-  // CHECK-NEXT:  %[[V5:.*]] = tosa.abs %[[V4]]
-  // CHECK-NEXT:  %[[V6:.*]] = mesh.shard %[[V5]] to %[[S1]]  : tensor<8x16xf32>
-  %1 = tosa.abs %0: (tensor<8x16xf32>) -> tensor<8x16xf32>
-  // CHECK-NEXT:  %[[S3:.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}]] : !mesh.sharding
-  // CHECK-NEXT:  %[[ZP1:.*]] = mesh.shard %arg1 to %[[S3]] annotate_for_users : tensor<1xf32>
-  // CHECK-NEXT:  %[[ZP2:.*]] = mesh.shard %arg2 to %[[S3]] annotate_for_users : tensor<1xf32>
-  // CHECK-NEXT:  %[[V7:.*]] = tosa.negate %[[V4]], %[[ZP1]], %[[ZP2]]
-  // CHECK-NEXT:  %[[V8:.*]] = mesh.shard %[[V7]] to %[[S1]]  : tensor<8x16xf32>
-  %2 = tosa.negate %0, %arg1, %arg2 : (tensor<8x16xf32>, tensor<1xf32>, tensor<1xf32>) -> tensor<8x16xf32>
-  %s3 = mesh.sharding @mesh_2d split_axes = [[0], [1]] : !mesh.sharding
-  %3 = mesh.shard %2 to %s3  : tensor<8x16xf32>
-  // CHECK-NEXT: return %[[V6]], %[[V8]]
-  return %1, %3 : tensor<8x16xf32>, tensor<8x16xf32>
-}
-
-// CHECK-LABEL: func.func @matmul_on_def_shard_batch_and_m
-// CHECK-SAME:     %[[ARG0:.*]]: tensor<2x16x8xf32>, %[[ARG1:.*]]: tensor<2x8x32xf32>, %[[ARG2:.*]]: tensor<1xf32>
-func.func @matmul_on_def_shard_batch_and_m(%arg0: tensor<2x16x8xf32>, %arg1: tensor<2x8x32xf32>, %arg2: tensor<1xf32>) -> tensor<2x16x32xf32> {
-  // CHECK-NEXT:  %[[S0:.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}0], [1]] : !mesh.sharding
-  // CHECK-NEXT:  %[[V0:.*]] = mesh.shard %[[ARG0]] to %[[S0]] annotate_for_users  : tensor<2x16x8xf32>
-  // CHECK-NEXT:  %[[S1:.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}0]] : !mesh.sharding
-  // CHECK-NEXT:  %[[V1:.*]] = mesh.shard %[[ARG1]] to %[[S1]] annotate_for_users  : tensor<2x8x32xf32>
-  // CHECK-NEXT:  %[[S2:.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}]] : !mesh.sharding
-  // CHECK-NEXT:  %[[ZP:.*]] = mesh.shard %[[ARG2]] to %[[S2]] annotate_for_users  : tensor<1xf32>
-  // CHECK-NEXT:  %[[V2:.*]] = tosa.matmul %[[V0]], %[[V1]], %[[ZP]], %[[ZP]]
-  %0 = tosa.matmul %arg0, %arg1, %arg2, %arg2 : (tensor<2x16x8xf32>, tensor<2x8x32xf32>, tensor<1xf32>, tensor<1xf32>)  -> tensor<2x16x32xf32>
-  // CHECK-NEXT:  %[[V3:.*]] = mesh.shard %[[V2]] to %[[S0]]  : tensor<2x16x32xf32>
-  %s1 = mesh.sharding @mesh_2d split_axes = [[0], [1]] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1  : tensor<2x16x32xf32>
-  // CHECK-NEXT:  return %[[V3]]
-  return %1 : tensor<2x16x32xf32>
-}
-
-// CHECK-LABEL: func.func @matmul_on_def_shard_m_and_n
-// CHECK-SAME:     [[varg0:%.*]]: tensor<2x16x8xf32>, [[varg1:%.*]]: tensor<2x8x32xf32>, [[varg2:%.*]]: tensor<1xf32>
-func.func @matmul_on_def_shard_m_and_n(%arg0: tensor<2x16x8xf32>, %arg1: tensor<2x8x32xf32>, %arg2: tensor<1xf32>) -> tensor<2x16x32xf32> {
-  // CHECK: [[vsharding:%.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}], [0]] : !mesh.sharding
-  // CHECK: [[vsharded:%.*]] = mesh.shard [[varg0]] to [[vsharding]] annotate_for_users : tensor<2x16x8xf32>
-  // CHECK: [[vsharding_0:%.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}], [], [1]] : !mesh.sharding
-  // CHECK: [[vsharded_1:%.*]] = mesh.shard [[varg1]] to [[vsharding_0]] annotate_for_users : tensor<2x8x32xf32>
-  // CHECK: [[vsharding_2:%.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}]] : !mesh.sharding
-  // CHECK: [[vsharded_3:%.*]] = mesh.shard [[varg2]] to [[vsharding_2]] annotate_for_users : tensor<1xf32>
-  // CHECK: [[v0:%.*]] = tosa.matmul
-  %0 = tosa.matmul %arg0, %arg1, %arg2, %arg2 : (tensor<2x16x8xf32>, tensor<2x8x32xf32>, tensor<1xf32>, tensor<1xf32>)  -> tensor<2x16x32xf32>
-  // CHECK: [[vsharding_4:%.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}], [0], [1]] : !mesh.sharding
-  // CHECK: [[vsharded_5:%.*]] = mesh.shard [[v0]] to [[vsharding_4]] : tensor<2x16x32xf32>
-  %s1 = mesh.sharding @mesh_2d split_axes = [[], [0], [1]] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1  : tensor<2x16x32xf32>
-  // CHECK-NEXT:  return [[vsharded_5]]
-  return %1 : tensor<2x16x32xf32>
-}
-
-// CHECK-LABEL: func.func @matmul_on_use_shard_m_and_k
-// CHECK-SAME:     [[varg0:%.*]]: tensor<2x16x8xf32>, [[varg1:%.*]]: tensor<2x8x32xf32>, [[varg2:%.*]]: tensor<1xf32>
-func.func @matmul_on_use_shard_m_and_k(%arg0: tensor<2x16x8xf32>, %arg1: tensor<2x8x32xf32>, %arg2: tensor<1xf32>) -> tensor<2x16x32xf32> {
-  // CHECK: [[vsharding:%.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}], [0], [1]] : !mesh.sharding
-  %s0 = mesh.sharding @mesh_2d split_axes = [[], [0], [1]] : !mesh.sharding
-  // CHECK: [[vsharded:%.*]] = mesh.shard [[varg0]] to [[vsharding]] : tensor<2x16x8xf32>
-  %arg0_s = mesh.shard %arg0 to %s0 : tensor<2x16x8xf32>
-  // CHECK: [[vsharded_0:%.*]] = mesh.shard [[vsharded]] to [[vsharding]] annotate_for_users : tensor<2x16x8xf32>
-  // CHECK: [[vsharding_1:%.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}], [1]] : !mesh.sharding
-  // CHECK: [[vsharded_2:%.*]] = mesh.shard [[varg1]] to [[vsharding_1]] annotate_for_users : tensor<2x8x32xf32>
-  // CHECK: [[vsharding_3:%.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}]] : !mesh.sharding
-  // CHECK: [[vsharded_4:%.*]] = mesh.shard [[varg2]] to [[vsharding_3]] annotate_for_users : tensor<1xf32>
-  // CHECK: [[v0:%.*]] = tosa.matmul
-  // CHECK: [[vsharding_5:%.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}], [0]] : !mesh.sharding
-  // CHECK: [[vsharded_6:%.*]] = mesh.shard [[v0]] to [[vsharding_5]] : tensor<2x16x32xf32>
-  %0 = tosa.matmul %arg0_s, %arg1, %arg2, %arg2 : (tensor<2x16x8xf32>, tensor<2x8x32xf32>, tensor<1xf32>, tensor<1xf32>)  -> tensor<2x16x32xf32>
-  // CHECK: return [[vsharded_6]]
-  return %0 : tensor<2x16x32xf32>
-}
-
-// CHECK-LABEL: func.func @matmul_on_use_shard_m_and_duplicted_k
-// CHECK-SAME:     %[[ARG0:.*]]: tensor<2x16x8xf32>, %[[ARG1:.*]]: tensor<2x8x32xf32>, %[[ARG2:.*]]: tensor<1xf32>
-func.func @matmul_on_use_shard_m_and_duplicted_k(%arg0: tensor<2x16x8xf32>, %arg1: tensor<2x8x32xf32>, %arg2: tensor<1xf32>) -> tensor<2x16x32xf32> {
-  // CHECK-NEXT:  %[[S0:.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}], [1], [0]] : !mesh.sharding
-  // CHECK-NEXT:  %[[V0:.*]] = mesh.shard %[[ARG0]] to %[[S0]] annotate_for_users  : tensor<2x16x8xf32>
-  %s0 = mesh.sharding @mesh_2d split_axes = [[], [1], [0]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0 annotate_for_users  : tensor<2x16x8xf32>
-  // CHECK-NEXT:  %[[S1:.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}], [0]] : !mesh.sharding
-  // CHECK-NEXT:  %[[V1:.*]] = mesh.shard %[[ARG1]] to %[[S1]] annotate_for_users  : tensor<2x8x32xf32>
-  %s1 = mesh.sharding @mesh_2d split_axes = [[], [0]] : !mesh.sharding
-  %1 = mesh.shard %arg1 to %s1 annotate_for_users  : tensor<2x8x32xf32>
-  // CHECK-NEXT:  %[[S2:.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}]] : !mesh.sharding
-  // CHECK-NEXT:  %[[ZP:.*]] = mesh.shard %[[ARG2]] to %[[S2]] annotate_for_users  : tensor<1xf32>
-  // CHECK-NEXT:  %[[V2:.*]] = tosa.matmul %[[V0]], %[[V1]], %[[ZP]], %[[ZP]]
-  %2 = tosa.matmul %0, %1, %arg2, %arg2 : (tensor<2x16x8xf32>, tensor<2x8x32xf32>, tensor<1xf32>, tensor<1xf32>)  -> tensor<2x16x32xf32>
-  // CHECK-NEXT:  %[[S3:.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}], [1]] : !mesh.sharding
-  // CHECK-NEXT:  %[[V3:.*]] = mesh.shard %[[V2]] to %[[S3]]  : tensor<2x16x32xf32>
-  // CHECK-NEXT:  return %[[V3]]
-  return %2 : tensor<2x16x32xf32>
-}
-
-// CHECK-LABEL: func.func @resolve_conflicting_annotations
-func.func @resolve_conflicting_annotations(
-  // CHECK-SAME: %[[IN1:.*]]: tensor<2x3xf32>,
-  %arg0: tensor<2x3xf32>,
-  // CHECK-SAME: %[[IN2:.*]]: tensor<3x2xf32>,
-  %arg1: tensor<3x2xf32>,
-  // CHECK-SAME: %[[OUT_DPS:.*]]: tensor<2x2xf32>
-  %out_dps: tensor<2x2xf32>
-// CHECK-SAME: ) -> tensor<2x2xf32> {
-) -> tensor<2x2xf32> {
-  // CHECK: %[[SIN1_SHARDED1:.*]] = mesh.sharding @mesh_2 split_axes = {{\[\[}}0]] : !mesh.sharding
-  // CHECK-NEXT:  %[[IN1_SHARDED1:.*]] = mesh.shard %[[IN1]] to %[[SIN1_SHARDED1]]  : tensor<2x3xf32>
-  // CHECK: %[[SIN2_SHARDED:.*]] = mesh.sharding @mesh_2 split_axes = {{\[\[}}]] : !mesh.sharding
-  // CHECK-NEXT:  %[[IN1_SHARDED2:.*]] = mesh.shard %[[IN1_SHARDED1]] to %[[SIN2_SHARDED]] annotate_for_users  : tensor<2x3xf32>
-  // CHECK-NEXT:  %[[IN2_SHARDED:.*]] = mesh.shard %[[IN2]] to %[[SIN2_SHARDED]] annotate_for_users  : tensor<3x2xf32>
-  // CHECK-NEXT:  %[[OUT_DPS_SHARDED:.*]] = mesh.shard %[[OUT_DPS]] to %[[SIN2_SHARDED]] annotate_for_users  : tensor<2x2xf32>
-  %sarg0_sharded = mesh.sharding @mesh_2 split_axes = [[0]] : !mesh.sharding
-  %arg0_sharded = mesh.shard %arg0 to %sarg0_sharded  : tensor<2x3xf32>
-  // CHECK: %[[MATMUL:.*]] = linalg.matmul ins(%[[IN1_SHARDED2]], %[[IN2_SHARDED]] : tensor<2x3xf32>, tensor<3x2xf32>)
-  // CHECK-SAME: outs(%[[OUT_DPS_SHARDED]] : tensor<2x2xf32>) -> tensor<2x2xf32>
-  %res = linalg.matmul ins(%arg0_sharded, %arg1 : tensor<2x3xf32>, tensor<3x2xf32>)
-    outs(%out_dps : tensor<2x2xf32>) -> tensor<2x2xf32>
-  // CHECK-NEXT: %[[RES:.*]] = mesh.shard %[[MATMUL]] to %[[SIN2_SHARDED]] : tensor<2x2xf32>
-  %sres_sharded = mesh.sharding @mesh_2 split_axes = [[]] : !mesh.sharding
-  %res_sharded = mesh.shard %res to %sres_sharded  : tensor<2x2xf32>
-  // CHECK: return %[[RES]] : tensor<2x2xf32>
-  return %res_sharded : tensor<2x2xf32>
-}
-
-// https://arxiv.org/abs/2211.05102 Figure 2(a)
-// The sharding propagation results in unnecessary reshards,
-//   an optimization pass should be able to remove them.
-// CHECK-LABEL: func.func @mlp_1d_weight_stationary
-// CHECK-SAME:     [[varg0:%.*]]: tensor<2x4x8xf32>, [[varg1:%.*]]: tensor<2x8x32xf32>, [[varg2:%.*]]: tensor<2x32x8xf32>, [[varg3:%.*]]: tensor<1xf32>
-func.func @mlp_1d_weight_stationary(%arg0: tensor<2x4x8xf32>, %arg1: tensor<2x8x32xf32>, %arg2: tensor<2x32x8xf32>, %arg3: tensor<1xf32>) -> tensor<2x4x8xf32> {
-  %s0 = mesh.sharding @mesh_1d split_axes = [[], [], [0, 1, 2]] : !mesh.sharding
-  %sharded0 = mesh.shard %arg0 to %s0 : tensor<2x4x8xf32>
-  %sharded1 = mesh.shard %arg1 to %s0 : tensor<2x8x32xf32>
-  // CHECK: [[vsharding:%.*]] = mesh.sharding @mesh_1d split_axes = {{\[\[}}], [], [0, 1, 2]] : !mesh.sharding
-  // CHECK: [[vsharded:%.*]] = mesh.shard [[varg0]] to [[vsharding]] : tensor<2x4x8xf32>
-  // CHECK: [[vsharded_0:%.*]] = mesh.shard [[varg1]] to [[vsharding]] : tensor<2x8x32xf32>
-  // CHECK: [[vsharded_1:%.*]] = mesh.shard [[vsharded]] to [[vsharding]] annotate_for_users : tensor<2x4x8xf32>
-  // CHECK: [[vsharding_2:%.*]] = mesh.sharding @mesh_1d split_axes = {{\[\[}}], [0, 1, 2]] : !mesh.sharding
-  // CHECK: [[vsharded_3:%.*]] = mesh.shard [[vsharded_0]] to [[vsharding_2]] annotate_for_users : tensor<2x8x32xf32>
-  // CHECK: [[vsharding_4:%.*]] = mesh.sharding @mesh_1d split_axes = {{\[\[}}]] : !mesh.sharding
-  // CHECK: [[vsharded_5:%.*]] = mesh.shard [[varg3]] to [[vsharding_4]] annotate_for_users : tensor<1xf32>
-  // CHECK: [[v0:%.*]] = tosa.matmul
-  %1 = tosa.matmul %sharded0, %sharded1, %arg3, %arg3 : (tensor<2x4x8xf32>, tensor<2x8x32xf32>, tensor<1xf32>, tensor<1xf32>)  -> tensor<2x4x32xf32>
-  // CHECK: [[vsharded_6:%.*]] = mesh.shard [[v0]] to [[vsharding_4]] : tensor<2x4x32xf32>
-  // CHECK: [[vsharded_7:%.*]] = mesh.shard [[vsharded_6]] to [[vsharding_4]] annotate_for_users : tensor<2x4x32xf32>
-  // CHECK: [[v1:%.*]] = tosa.sigmoid [[vsharded_7]] : (tensor<2x4x32xf32>) -> tensor<2x4x32xf32>
-  // CHECK: [[vsharded_8:%.*]] = mesh.shard [[v1]] to [[vsharding_4]] : tensor<2x4x32xf32>
-  %2 = tosa.sigmoid %1 : (tensor<2x4x32xf32>) -> tensor<2x4x32xf32>
-  %sharding = mesh.sharding @mesh_1d split_axes = [[], [0, 1, 2]] : !mesh.sharding
-  // CHECK: [[vsharded_9:%.*]] = mesh.shard [[varg2]] to [[vsharding_2]] : tensor<2x32x8xf32>
-  %sharded2 = mesh.shard %arg2 to %sharding  : tensor<2x32x8xf32>
-  // CHECK: [[vsharded_10:%.*]] = mesh.shard [[vsharded_8]] to [[vsharding_4]] annotate_for_users : tensor<2x4x32xf32>
-  // CHECK: [[vsharded_11:%.*]] = mesh.shard [[vsharded_9]] to [[vsharding]] annotate_for_users : tensor<2x32x8xf32>
-  // CHECK: [[v2:%.*]] = tosa.matmul
-  %3 = tosa.matmul %2, %sharded2, %arg3, %arg3 : (tensor<2x4x32xf32>, tensor<2x32x8xf32>, tensor<1xf32>, tensor<1xf32>)  -> tensor<2x4x8xf32>
-  // CHECK: [[vsharded_12:%.*]] = mesh.shard [[v2]] to [[vsharding]] : tensor<2x4x8xf32>
-  %s4 = mesh.sharding @mesh_1d split_axes = [[], [], [0, 1, 2]] : !mesh.sharding
-  %4 = mesh.shard %3 to %s4  : tensor<2x4x8xf32>
-  // CHECK: return [[vsharded_12]]
-  return %4 : tensor<2x4x8xf32>
-}
-
-// https://arxiv.org/abs/2211.05102 Figure 2(b)
-// The sharding propagation results in unnecessary reshards,
-//   an optimization pass should be able to remove them.
-// CHECK-LABEL: func.func @mlp_2d_weight_stationary
-// CHECK-SAME:     [[varg0:%.*]]: tensor<2x4x8xf32>, [[varg1:%.*]]: tensor<2x8x32xf32>, [[varg2:%.*]]: tensor<2x32x8xf32>, [[varg3:%.*]]: tensor<1xf32>
-func.func @mlp_2d_weight_stationary(%arg0: tensor<2x4x8xf32>, %arg1: tensor<2x8x32xf32>, %arg2: tensor<2x32x8xf32>, %arg3: tensor<1xf32>) -> tensor<2x4x8xf32> {
-    // CHECK: [[vsharding:%.*]] = mesh.sharding @mesh_3d split_axes = {{\[\[}}], [], [0, 1, 2]] : !mesh.sharding
-  %s0 = mesh.sharding @mesh_3d split_axes = [[], [], [0, 1, 2]] : !mesh.sharding
-    // CHECK: [[vsharded:%.*]] = mesh.shard [[varg0]] to [[vsharding]] : tensor<2x4x8xf32>
-  %arg0_s = mesh.shard %arg0 to %s0  : tensor<2x4x8xf32>
-    // CHECK: [[vsharding_0:%.*]] = mesh.sharding @mesh_3d split_axes = {{\[\[}}], [0], [1, 2]] : !mesh.sharding
-  %s1 = mesh.sharding @mesh_3d split_axes = [[], [0], [1, 2]] : !mesh.sharding
-    // CHECK: [[vsharded_1:%.*]] = mesh.shard [[varg1]] to [[vsharding_0]] : tensor<2x8x32xf32>
-  %arg1_s = mesh.shard %arg1 to %s1  : tensor<2x8x32xf32>
-    // CHECK: [[vsharding_2:%.*]] = mesh.sharding @mesh_3d split_axes = {{\[\[}}]] : !mesh.sharding
-    // CHECK: [[vsharded_3:%.*]] = mesh.shard [[vsharded]] to [[vsharding_2]] annotate_for_users : tensor<2x4x8xf32>
-    // CHECK: [[vsharded_4:%.*]] = mesh.shard [[vsharded_1]] to [[vsharding]] annotate_for_users : tensor<2x8x32xf32>
-    // CHECK: [[vsharded_5:%.*]] = mesh.shard [[varg3]] to [[vsharding_2]] annotate_for_users : tensor<1xf32>
-    // CHECK: [[v0:%.*]] = tosa.matmul
-  %1 = tosa.matmul %arg0_s, %arg1_s, %arg3, %arg3 : (tensor<2x4x8xf32>, tensor<2x8x32xf32>, tensor<1xf32>, tensor<1xf32>)  -> tensor<2x4x32xf32>
-    // CHECK: [[vsharded_6:%.*]] = mesh.shard [[v0]] to [[vsharding]] : tensor<2x4x32xf32>
-  %2 = mesh.shard %1 to %s0  : tensor<2x4x32xf32>
-    // CHECK: [[vsharded_7:%.*]] = mesh.shard [[vsharded_6]] to [[vsharding]] annotate_for_users : tensor<2x4x32xf32>
-    // CHECK: [[v1:%.*]] = tosa.sigmoid 
-    // CHECK: [[vsharded_8:%.*]] = mesh.shard [[v1]] to [[vsharding]] : tensor<2x4x32xf32>
-  %3 = tosa.sigmoid %2 : (tensor<2x4x32xf32>) -> tensor<2x4x32xf32>
-    // CHECK: [[vsharding_9:%.*]] = mesh.sharding @mesh_3d split_axes = {{\[\[}}], [1, 2], [0]] : !mesh.sharding
-  %s2 = mesh.sharding @mesh_3d split_axes = [[], [1, 2], [0]] : !mesh.sharding
-    // CHECK: [[vsharded_10:%.*]] = mesh.shard [[varg2]] to [[vsharding_9]] : tensor<2x32x8xf32>
-  %arg2_s = mesh.shard %arg2 to %s2  : tensor<2x32x8xf32>
-    // CHECK: [[vsharded_11:%.*]] = mesh.shard [[vsharded_8]] to [[vsharding_2]] annotate_for_users : tensor<2x4x32xf32>
-    // CHECK: [[vsharded_12:%.*]] = mesh.shard [[vsharded_10]] to [[vsharding]] annotate_for_users : tensor<2x32x8xf32>
-    // CHECK: [[v2:%.*]] = tosa.matmul
-  %4 = tosa.matmul %3, %arg2_s, %arg3, %arg3 : (tensor<2x4x32xf32>, tensor<2x32x8xf32>, tensor<1xf32>, tensor<1xf32>)  -> tensor<2x4x8xf32>
-    // CHECK: [[vsharded_13:%.*]] = mesh.shard [[v2]] to [[vsharding]] : tensor<2x4x8xf32>
-  %5 = mesh.shard %4 to %s0  : tensor<2x4x8xf32>
-    // CHECK: [[vsharded_14:%.*]] = mesh.shard [[vsharded_13]] to [[vsharding]] annotate_for_users : tensor<2x4x8xf32>
-  %6 = mesh.shard %5 to %s0 annotate_for_users  : tensor<2x4x8xf32>
-    // CHECK: return [[vsharded_14]]
-  return %6 : tensor<2x4x8xf32>
-}
-
-// CHECK-LABEL: func.func @elementwise_duplicated_chain
-// CHECK-SAME:    %[[ARG:.*]]: tensor<8x16xf32>
-func.func @elementwise_duplicated_chain(%arg0: tensor<8x16xf32>) -> tensor<8x16xf32> {
-  // CHECK-NEXT:  %[[S0:.*]] = mesh.sharding @mesh_2d split_axes = {{\[\[}}]] : !mesh.sharding
-  // CHECK-NEXT:  %[[V0:.*]] = mesh.shard %[[ARG]] to %[[S0]] annotate_for_users  : tensor<8x16xf32>
-  // CHECK-NEXT:  %[[V1:.*]] = tosa.sigmoid %[[V0]]
-  %0 = tosa.sigmoid %arg0 : (tensor<8x16xf32>) -> tensor<8x16xf32>
-  // CHECK-NEXT:  %[[V2:.*]] = mesh.shard %[[V1]] to %[[S0]] : tensor<8x16xf32>
-  // CHECK-NEXT:  %[[V3:.*]] = mesh.shard %[[V2]] to %[[S0]] annotate_for_users : tensor<8x16xf32>
-  // CHECK-NEXT:  %[[V4:.*]] = tosa.sigmoid %[[V3]]
-  %1 = tosa.sigmoid %0 : (tensor<8x16xf32>) -> tensor<8x16xf32>
-  // CHECK-NEXT:  %[[V5:.*]] = mesh.shard %[[V4]] to %[[S0]]  : tensor<8x16xf32>
-  %s0 = mesh.sharding @mesh_2d split_axes = [[]] : !mesh.sharding
-  %2 = mesh.shard %1 to %s0 : tensor<8x16xf32>
-  // CHECK-NEXT:  return %[[V5]]
-  return %2 : tensor<8x16xf32>
-}
diff --git a/mlir/test/Dialect/Mesh/spmdization.mlir b/mlir/test/Dialect/Mesh/spmdization.mlir
deleted file mode 100644
index 701898c..0000000
--- a/mlir/test/Dialect/Mesh/spmdization.mlir
+++ /dev/null
@@ -1,317 +0,0 @@
-// RUN: mlir-opt \
-// RUN:   --pass-pipeline="builtin.module(func.func(mesh-spmdization,test-single-fold))" \
-// RUN:   %s | FileCheck %s
-
-mesh.mesh @mesh_1d(shape = 2)
-
-// CHECK-LABEL: func @return_sharding
-func.func @return_sharding(
-  // CHECK-SAME: [[ARG:%.*]]: tensor<1xf32>
-  %arg0: tensor<2xf32>
-// CHECK-SAME: ) -> (tensor<1xf32>, !mesh.sharding) {
-) -> (tensor<2xf32>, !mesh.sharding) {
-  %ssharding_annotated = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %sharding_annotated = mesh.shard %arg0 to %ssharding_annotated  : tensor<2xf32>
-  // CHECK-NEXT: [[vsharding:%.*]] = mesh.sharding @mesh_1d split_axes = {{\[\[}}0]] : !mesh.sharding
-  %r = mesh.get_sharding %sharding_annotated : tensor<2xf32> -> !mesh.sharding
-  // CHECK-NEXT: return [[ARG]], [[vsharding]] : tensor<1xf32>, !mesh.sharding
-  return %sharding_annotated, %r : tensor<2xf32>, !mesh.sharding
-}
-
-// CHECK-LABEL: func @full_replication
-func.func @full_replication(
-  // CHECK-SAME: %[[ARG:.*]]: tensor<2xi8>
-  %arg0: tensor<2xi8>
-// CHECK-SAME: -> tensor<2xi8> {
-) -> tensor<2xi8> {
-  %s0 = mesh.sharding @mesh_1d split_axes = [[]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0  : tensor<2xi8>
-  %s1 = mesh.sharding @mesh_1d split_axes = [[]] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1  annotate_for_users : tensor<2xi8>
-  // CHECK: return %[[ARG]] : tensor<2xi8>
-  return %1 : tensor<2xi8>
-}
-
-// CHECK-LABEL: func @sharding_triplet
-func.func @sharding_triplet(
-  // CHECK-SAME: %[[ARG:.*]]: tensor<1xf32>
-  %arg0: tensor<2xf32>
-// CHECK-SAME: ) -> tensor<2xf32> {
-) -> tensor<2xf32> {
-  // CHECK: %[[ALL_GATHER:.*]] = mesh.all_gather %[[ARG]] on @mesh_1d mesh_axes = [0] gather_axis = 0 : tensor<1xf32> -> tensor<2xf32>
-  %ssharding_annotated = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %sharding_annotated = mesh.shard %arg0 to %ssharding_annotated  : tensor<2xf32>
-  %ssharding_annotated_0 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %sharding_annotated_0 = mesh.shard %sharding_annotated to %ssharding_annotated_0  annotate_for_users : tensor<2xf32>
-  %ssharding_annotated_1 = mesh.sharding @mesh_1d split_axes = [[]] : !mesh.sharding
-  %sharding_annotated_1 = mesh.shard %sharding_annotated_0 to %ssharding_annotated_1  : tensor<2xf32>
-  // CHECK: return %[[ALL_GATHER]] : tensor<2xf32>
-  return %sharding_annotated_1 : tensor<2xf32>
-}
-
-
-// CHECK-LABEL: func @move_split_axis
-func.func @move_split_axis(
-  // CHECK-SAME: %[[ARG:.*]]: tensor<1x2xi8>
-  %arg0: tensor<2x2xi8>
-// CHECK-SAME: -> tensor<2x1xi8> {
-) -> tensor<2x2xi8> {
-  // CHECK: %[[ALL_TO_ALL:.*]] = mesh.all_to_all %[[ARG]] on @mesh_1d
-  // CHECK-SAME: mesh_axes = [0] split_axis = 1 concat_axis = 0 : tensor<1x2xi8> -> tensor<2x1xi8>
-  %s0 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0  : tensor<2x2xi8>
-  %s1 = mesh.sharding @mesh_1d split_axes = [[], [0]] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1  annotate_for_users : tensor<2x2xi8>
-  // CHECK: return %[[ALL_TO_ALL]] : tensor<2x1xi8>
-  return %1 : tensor<2x2xi8>
-}
-
-// CHECK-LABEL: func @non_tensor_value
-func.func @non_tensor_value(
-  // CHECK-SAME: %[[ARG:.*]]: i8
-  %arg0: i8
-// CHECK-SAME: -> i8 {
-) -> i8 {
-  // CHECK: %[[RES:.*]] = arith.addi %[[ARG]], %[[ARG]] : i8
-  %0 = arith.addi %arg0, %arg0 : i8
-  // CHECK: return %[[RES]] : i8
-  return %0 : i8
-}
-
-// CHECK-LABEL: func @unary_elementwise
-func.func @unary_elementwise(
-  // CHECK-SAME: %[[ARG:.*]]: tensor<1xi8>
-  %arg0: tensor<2xi8>
-// CHECK-SAME: -> tensor<1xi8> {
-) -> tensor<2xi8> {
-  %s0 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0  : tensor<2xi8>
-  %s1 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1  annotate_for_users : tensor<2xi8>
-  // CHECK: %[[RES:.*]] = tosa.abs %[[ARG]] : (tensor<1xi8>) -> tensor<1xi8>
-  %2 = tosa.abs %1 : (tensor<2xi8>) -> tensor<2xi8>
-  %s3 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %3 = mesh.shard %2 to %s3  : tensor<2xi8>
-  %s4 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %4 = mesh.shard %3 to %s4  annotate_for_users : tensor<2xi8>
-  // CHECK: return %[[RES]] : tensor<1xi8>
-  return %4 : tensor<2xi8>
-}
-
-// full replication -> shard axis -> abs -> shard axis -> full replication
-// CHECK-LABEL: func @unary_elementwise_with_resharding
-func.func @unary_elementwise_with_resharding(
-  // CHECK-SAME: %[[ARG:.*]]: tensor<2xi8>
-  %arg0: tensor<2xi8>
-// CHECK-SAME: -> tensor<2xi8> {
-) -> tensor<2xi8> {
-  // CHECK: %[[SLICE:.*]] = mesh.all_slice %[[ARG]] on @mesh_1d mesh_axes = [0] slice_axis = 0
-  // CHECK-SAME: tensor<2xi8> -> tensor<1xi8>
-  %s0 = mesh.sharding @mesh_1d split_axes = [[]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0  : tensor<2xi8>
-  %s1 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1  annotate_for_users : tensor<2xi8>
-  // CHECK: %[[ABS:.*]] = tosa.abs %[[SLICE]] : (tensor<1xi8>) -> tensor<1xi8>
-  %2 = tosa.abs %1 : (tensor<2xi8>) -> tensor<2xi8>
-  // CHECK: %[[RES:.*]] = mesh.all_gather %[[ABS]] on @mesh_1d
-  // CHECK-SAME: mesh_axes = [0] gather_axis = 0 : tensor<1xi8> -> tensor<2xi8>
-  %s3 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %3 = mesh.shard %2 to %s3  : tensor<2xi8>
-  %s4 = mesh.sharding @mesh_1d split_axes = [[]] : !mesh.sharding
-  %4 = mesh.shard %3 to %s4  annotate_for_users : tensor<2xi8>
-  // CHECK: return %[[RES]] : tensor<2xi8>
-  return %4 : tensor<2xi8>
-}
-
-// CHECK-LABEL: func @binary_elementwise
-func.func @binary_elementwise(
-  // CHECK-SAME: %[[ARG0:.*]]: tensor<1xi8>,
-  %arg0: tensor<2xi8>,
-  // CHECK-SAME: %[[ARG1:.*]]: tensor<1xi8>
-  %arg1: tensor<2xi8>
-// CHECK-SAME: -> tensor<1xi8> {
-) -> tensor<2xi8> {
-  %sarg0_sharded = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %arg0_sharded = mesh.shard %arg0 to %sarg0_sharded  : tensor<2xi8>
-  %sop_arg0 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %op_arg0 = mesh.shard %arg0_sharded to %sop_arg0  annotate_for_users : tensor<2xi8>
-  %sarg1_sharded = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %arg1_sharded = mesh.shard %arg1 to %sarg1_sharded  : tensor<2xi8>
-  %sop_arg1 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %op_arg1 = mesh.shard %arg1_sharded to %sop_arg1  annotate_for_users : tensor<2xi8>
-  // CHECK: %[[RES:.*]] = tosa.add %[[ARG0]], %[[ARG1]] : (tensor<1xi8>, tensor<1xi8>) -> tensor<1xi8>
-  %op_res = tosa.add %op_arg0, %op_arg1 : (tensor<2xi8>, tensor<2xi8>) -> tensor<2xi8>
-  %sop_res_sharded = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %op_res_sharded = mesh.shard %op_res to %sop_res_sharded  : tensor<2xi8>
-  %sres = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %res = mesh.shard %op_res_sharded to %sres  annotate_for_users : tensor<2xi8>
-  // CHECK: return %[[RES]] : tensor<1xi8>
-  return %res : tensor<2xi8>
-}
-
-// reshard
-// abs
-// reshard
-// abs
-// reshard
-// CHECK-LABEL: func @multiple_chained_ops
-func.func @multiple_chained_ops(
-  // CHECK-SAME: %[[ARG:.*]]: tensor<2xi8>
-  %arg0: tensor<2xi8>
-// CHECK-SAME: -> tensor<1xi8> {
-) -> tensor<2xi8> {
-  // CHECK: %[[RESHARD1:.*]] = mesh.all_slice %[[ARG]] on @mesh_1d mesh_axes = [0] slice_axis = 0
-  // CHECK-SAME: tensor<2xi8> -> tensor<1xi8>
-  %s0 = mesh.sharding @mesh_1d split_axes = [[]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0  : tensor<2xi8>
-  %s1 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1  annotate_for_users : tensor<2xi8>
-  // CHECK: %[[ABS1:.*]] = tosa.abs %[[RESHARD1]] : (tensor<1xi8>) -> tensor<1xi8>
-  %2 = tosa.abs %1 : (tensor<2xi8>) -> tensor<2xi8>
-  // CHECK: %[[RESHARD2:.*]] = mesh.all_gather %[[ABS1]] on @mesh_1d
-  // CHECK-SAME: mesh_axes = [0] gather_axis = 0 : tensor<1xi8> -> tensor<2xi8>
-  %s3 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %3 = mesh.shard %2 to %s3  : tensor<2xi8>
-  %s4 = mesh.sharding @mesh_1d split_axes = [[]] : !mesh.sharding
-  %4 = mesh.shard %3 to %s4  annotate_for_users : tensor<2xi8>
-  // CHECK: %[[ABS2:.*]] = tosa.abs %[[RESHARD2]] : (tensor<2xi8>) -> tensor<2xi8>
-  %5 = tosa.abs %4 : (tensor<2xi8>) -> tensor<2xi8>
-  // CHECK: %[[RESHARD3:.*]] = mesh.all_slice %[[ABS2]] on @mesh_1d mesh_axes = [0] slice_axis = 0 :
-  // CHECK-SAME: tensor<2xi8> -> tensor<1xi8>
-  %s6 = mesh.sharding @mesh_1d split_axes = [[]] : !mesh.sharding
-  %6 = mesh.shard %5 to %s6  : tensor<2xi8>
-  %s7 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %7 = mesh.shard %6 to %s7  annotate_for_users : tensor<2xi8>
-  // CHECK: return %[[RESHARD3]] : tensor<1xi8>
-  return %7 : tensor<2xi8>
-}
-
-// CHECK-LABEL: func @incomplete_sharding
-func.func @incomplete_sharding(
-  // CHECK-SAME: %[[ARG:.*]]: tensor<4x16xf32>
-  %arg0: tensor<8x16xf32>
-// CHECK-SAME: -> tensor<4x16xf32> {
-) -> tensor<8x16xf32> {
-  %s0 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0  annotate_for_users : tensor<8x16xf32>
-  // CHECK: %[[RES:.*]] = tosa.sigmoid %[[ARG]] : (tensor<4x16xf32>) -> tensor<4x16xf32>
-  %1 = tosa.sigmoid %0 : (tensor<8x16xf32>) -> tensor<8x16xf32>
-  %s2 = mesh.sharding @mesh_1d split_axes = [[0]] : !mesh.sharding
-  %2 = mesh.shard %1 to %s2  : tensor<8x16xf32>
-  // CHECK: return %[[RES]] : tensor<4x16xf32>
-  return %2 : tensor<8x16xf32>
-}
-
-mesh.mesh @mesh_1d_4(shape = 4)
-
-// CHECK-LABEL: func @ew_chain_with_halo
-func.func @ew_chain_with_halo(
-  // CHECK-SAME: %[[IN1:[A-Za-z0-9_]+]]: tensor<5x16xf32>
-  %arg0: tensor<8x16xf32>,
-  // CHECK-SAME: %[[IN2:[A-Za-z0-9_]+]]: tensor<1xf32>
-  %arg1: tensor<1xf32>,
-  // CHECK-SAME: %[[IN3:[A-Za-z0-9_]+]]: tensor<1xf32>
-  %arg2: tensor<1xf32>)
-  // CHECK-SAME: -> tensor<5x16xf32>
-   -> tensor<8x16xf32> {
-  %ssharding_annotated = mesh.sharding @mesh_1d_4 split_axes = [[0]] halo_sizes = [2, 1] : !mesh.sharding
-  %sharding_annotated = mesh.shard %arg0 to %ssharding_annotated  annotate_for_users : tensor<8x16xf32>
-  // CHECK: %[[TMP1:.*]] = tosa.tanh %[[IN1]] : (tensor<5x16xf32>) -> tensor<5x16xf32>
-  %0 = tosa.tanh %sharding_annotated : (tensor<8x16xf32>) -> tensor<8x16xf32>
-  %ssharding_annotated_0 = mesh.sharding @mesh_1d_4 split_axes = [[0]] halo_sizes = [2, 1] : !mesh.sharding
-  %sharding_annotated_0 = mesh.shard %0 to %ssharding_annotated_0  : tensor<8x16xf32>
-  %ssharding_annotated_1 = mesh.sharding @mesh_1d_4 split_axes = [[0]] halo_sizes = [2, 1] : !mesh.sharding
-  %sharding_annotated_1 = mesh.shard %sharding_annotated_0 to %ssharding_annotated_1  annotate_for_users : tensor<8x16xf32>
-  // CHECK-NEXT: %[[TMP2:.*]] = tosa.abs %[[TMP1]] : (tensor<5x16xf32>) -> tensor<5x16xf32>
-  %1 = tosa.abs %sharding_annotated_1 : (tensor<8x16xf32>) -> tensor<8x16xf32>
-  %ssharding_annotated_2 = mesh.sharding @mesh_1d_4 split_axes = [[0]] halo_sizes = [2, 1] : !mesh.sharding
-  %sharding_annotated_2 = mesh.shard %1 to %ssharding_annotated_2  : tensor<8x16xf32>
-  %ssharding_annotated_4 = mesh.sharding @mesh_1d_4 split_axes = [[0]] halo_sizes = [2, 1] : !mesh.sharding
-  %sharding_annotated_4 = mesh.shard %sharding_annotated_2 to %ssharding_annotated_4  annotate_for_users : tensor<8x16xf32>
-  // CHECK-NEXT: %[[TMP3:.*]] = tosa.negate %[[TMP2]], %[[IN2]], %[[IN3]] : (tensor<5x16xf32>, tensor<1xf32>, tensor<1xf32>) -> tensor<5x16xf32>
-  %sharding_1 = mesh.sharding @mesh_1d_4 split_axes = [[]] : !mesh.sharding
-  %zero_point_1 = mesh.shard %arg1 to %sharding_1 annotate_for_users : tensor<1xf32>
-  %zero_point_2 = mesh.shard %arg2 to %sharding_1 annotate_for_users : tensor<1xf32>
-  %2 = tosa.negate %sharding_annotated_4, %zero_point_1, %zero_point_2 : (tensor<8x16xf32>, tensor<1xf32>, tensor<1xf32>) -> tensor<8x16xf32>
-  %ssharding_annotated_5 = mesh.sharding @mesh_1d_4 split_axes = [[0]] halo_sizes = [2, 1] : !mesh.sharding
-  %sharding_annotated_5 = mesh.shard %2 to %ssharding_annotated_5  : tensor<8x16xf32>
-  %ssharding_annotated_6 = mesh.sharding @mesh_1d_4 split_axes = [[0]] halo_sizes = [2, 1] : !mesh.sharding
-  %sharding_annotated_6 = mesh.shard %sharding_annotated_5 to %ssharding_annotated_6  annotate_for_users : tensor<8x16xf32>
-  // CHECK-NEXT: return %[[TMP3]] : tensor<5x16xf32>
-  return %sharding_annotated_6 : tensor<8x16xf32>
-}
-
-// CHECK-LABEL: func @test_shard_update_halo
-// CHECK-SAME: %[[IN1:[A-Za-z0-9_]+]]: tensor<300x1200xi64>
-func.func @test_shard_update_halo(%arg0: tensor<1200x1200xi64>) -> tensor<1200x1200xi64> {
-  %sharding = mesh.sharding @mesh_1d_4 split_axes = [[0]] : !mesh.sharding
-  // CHECK: %[[T:.*]] = tensor.empty() : tensor<304x1200xi64>
-  // CHECK: %[[inserted_slice:.*]] = tensor.insert_slice %[[IN1]] into %[[T]][2, 0] [300, 1200] [1, 1] : tensor<300x1200xi64> into tensor<304x1200xi64>
-  // CHECK: %[[UH:.*]] = mesh.update_halo %[[inserted_slice]] on @mesh_1d_4 split_axes = {{\[\[0]]}} halo_sizes = [2, 2] : tensor<304x1200xi64>
-  %sharding_annotated = mesh.shard %arg0 to %sharding : tensor<1200x1200xi64>
-  %sharding_0 = mesh.sharding @mesh_1d_4 split_axes = [[0]] halo_sizes = [2, 2] : !mesh.sharding
-  %sharding_annotated_1 = mesh.shard %sharding_annotated to %sharding_0 : tensor<1200x1200xi64>
-  %sharding_annotated_3 = mesh.shard %sharding_annotated_1 to %sharding_0 annotate_for_users : tensor<1200x1200xi64>
-  // CHECK: return %[[UH]] : tensor<304x1200xi64>
-  return %sharding_annotated_3 : tensor<1200x1200xi64>
-}
-
-mesh.mesh @mesh4x4(shape = 4x4)
-// CHECK-LABEL: func @test_shard_update_halo2d
-// CHECK-SAME: %[[IN1:[A-Za-z0-9_]+]]: tensor<300x300xi64>
-func.func @test_shard_update_halo2d(%arg0: tensor<1200x1200xi64>) -> tensor<1200x1200xi64> {
-  %sharding = mesh.sharding @mesh4x4 split_axes = [[0], [1]] : !mesh.sharding
-  // CHECK: %[[T:.*]] = tensor.empty() : tensor<303x307xi64>
-  // CHECK: %[[inserted_slice:.*]] = tensor.insert_slice %[[IN1]] into %[[T]][1, 3] [300, 300] [1, 1] : tensor<300x300xi64> into tensor<303x307xi64>
-  // CHECK: %[[UH:.*]] = mesh.update_halo %[[inserted_slice]] on @mesh4x4 split_axes = {{\[\[}}0], [1]] halo_sizes = [1, 2, 3, 4] : tensor<303x307xi64>
-  %sharding_annotated = mesh.shard %arg0 to %sharding : tensor<1200x1200xi64>
-  %sharding_0 = mesh.sharding @mesh4x4 split_axes = [[0], [1]] halo_sizes = [1, 2, 3, 4] : !mesh.sharding
-  %sharding_annotated_1 = mesh.shard %sharding_annotated to %sharding_0 : tensor<1200x1200xi64>
-  %sharding_annotated_3 = mesh.shard %sharding_annotated_1 to %sharding_0 annotate_for_users : tensor<1200x1200xi64>
-  // CHECK: return %[[UH]] : tensor<303x307xi64>
-  return %sharding_annotated_3 : tensor<1200x1200xi64>
-}
-
-mesh.mesh @mesh(shape = 2)
-// CHECK-LABEL: func.func @test_reduce_0d(
-// CHECK-SAME: %[[ARG0:[A-Za-z0-9_]+]]: tensor<3x6xi32>
-func.func @test_reduce_0d(%arg0: tensor<6x6xi32>) -> (tensor<i32>) {
-  %sharding = mesh.sharding @mesh split_axes = [[0]] : !mesh.sharding
-  %sharded = mesh.shard %arg0 to %sharding annotate_for_users : tensor<6x6xi32>
-  %4 = tensor.empty() : tensor<i32>
-  %sharding_out = mesh.sharding @mesh split_axes = [[]] : !mesh.sharding
-  %sharded_out = mesh.shard %4 to %sharding_out : tensor<i32>
-  %sharded_in = mesh.shard %sharded to %sharding annotate_for_users : tensor<6x6xi32>
-  // CHECK: %[[reduced:.*]] = linalg.reduce ins(%arg0 : tensor<3x6xi32>)
-  %reduced = linalg.reduce ins(%sharded_in : tensor<6x6xi32>) outs(%sharded_out : tensor<i32>) dimensions = [0, 1] 
-    (%in: i32, %init: i32) {
-      %6 = arith.addi %in, %init : i32
-      linalg.yield %6 : i32
-    }
-  // CHECK: %[[all_reduce:.*]] = mesh.all_reduce %[[reduced]] on @mesh mesh_axes = [0] : tensor<i32> -> tensor<i32>
-  %sharded_red = mesh.shard %reduced to %sharding_out : tensor<i32>
-  %sharded_ret = mesh.shard %sharded_red to %sharding_out annotate_for_users : tensor<i32>
-  // CHECK: return %[[all_reduce]] : tensor<i32>
-  return %sharded_ret : tensor<i32>
-}
-
-// CHECK-LABEL: func.func @test_reduce_1d(
-// CHECK-SAME: %[[ARG0:[A-Za-z0-9_]+]]: tensor<3x6xi32>
-func.func @test_reduce_1d(%arg0: tensor<6x6xi32>) -> (tensor<6xi32>) {
-  %sharding = mesh.sharding @mesh split_axes = [[0]] : !mesh.sharding
-  %sharded = mesh.shard %arg0 to %sharding annotate_for_users : tensor<6x6xi32>
-  %4 = tensor.empty() : tensor<6xi32>
-  %sharded_out = mesh.shard %4 to %sharding : tensor<6xi32>
-  %sharded_in = mesh.shard %sharded to %sharding annotate_for_users : tensor<6x6xi32>
-  // CHECK: %[[reduced:.*]] = linalg.reduce ins(%arg0 : tensor<3x6xi32>)
-  %reduced = linalg.reduce ins(%sharded_in : tensor<6x6xi32>) outs(%sharded_out : tensor<6xi32>) dimensions = [1] 
-    (%in: i32, %init: i32) {
-      %6 = arith.addi %in, %init : i32
-      linalg.yield %6 : i32
-    }
-  // CHECK-NOT: mesh.all_reduce
-  %sharded_red = mesh.shard %reduced to %sharding : tensor<6xi32>
-  %sharded_ret = mesh.shard %sharded_red to %sharding annotate_for_users : tensor<6xi32>
-  // CHECK: return %[[reduced]] : tensor<3xi32>
-  return %sharded_ret : tensor<6xi32>
-}
diff --git a/mlir/test/Dialect/Mesh/all-scatter-op-lowering.mlir b/mlir/test/Dialect/Shard/all-scatter-op-lowering.mlir
index 4f54607a..bc91121 100644
--- a/mlir/test/Dialect/Mesh/all-scatter-op-lowering.mlir
+++ b/mlir/test/Dialect/Shard/all-scatter-op-lowering.mlir
@@ -1,43 +1,43 @@
-// RUN: mlir-opt --split-input-file --test-mesh-all-slice-op-lowering --test-mesh-simplifications --cse %s | FileCheck %s
+// RUN: mlir-opt --split-input-file --test-grid-all-slice-op-lowering --test-grid-simplifications --cse %s | FileCheck %s
 
-mesh.mesh @mesh_1d(shape = ?)
+shard.grid @grid_1d(shape = ?)
 
-// CHECK-LABEL: func.func @all_slice_op_lowering_of_dynamic_1d_tensor_on_dynamic_1d_mesh
-func.func @all_slice_op_lowering_of_dynamic_1d_tensor_on_dynamic_1d_mesh(
+// CHECK-LABEL: func.func @all_slice_op_lowering_of_dynamic_1d_tensor_on_dynamic_1d_grid
+func.func @all_slice_op_lowering_of_dynamic_1d_tensor_on_dynamic_1d_grid(
   // CHECK: %[[ARG:.*]]: tensor<?xf16>
   %arg0: tensor<?xf16>
 // CHECK-SAME: -> tensor<?xf16> {
 ) -> tensor<?xf16> {
   // CHECK-DAG: %[[C0:.*]] = arith.constant 0 : index
-  // CHECK-DAG: %[[PROC_IDX:.*]] = mesh.process_multi_index on @mesh_1d axes = [0] : index
-  // CHECK-DAG: %[[MESH_SIZE:.*]] = mesh.mesh_shape @mesh_1d axes = [0] : index
+  // CHECK-DAG: %[[PROC_IDX:.*]] = shard.process_multi_index on @grid_1d axes = [0] : index
+  // CHECK-DAG: %[[SHARD_SIZE:.*]] = shard.grid_shape @grid_1d axes = [0] : index
   // CHECK: %[[TENSOR_AXIS_SIZE:.*]] = tensor.dim %[[ARG]], %c0 : tensor<?xf16>
-  // CHECK: %[[AXIS_SIZE_CHECK_REMINDER:.*]] = arith.remui %[[TENSOR_AXIS_SIZE]], %[[MESH_SIZE]] : index
+  // CHECK: %[[AXIS_SIZE_CHECK_REMINDER:.*]] = arith.remui %[[TENSOR_AXIS_SIZE]], %[[SHARD_SIZE]] : index
   // CHECK: %[[AXIS_SIZE_CHECK:.*]] = arith.cmpi eq, %[[AXIS_SIZE_CHECK_REMINDER]], %[[C0]] : index
   // CHECK: cf.assert %[[AXIS_SIZE_CHECK]]
-  // CHECK: %[[RESULT_AXIS_SIZE:.*]] = arith.divui %[[TENSOR_AXIS_SIZE]], %[[MESH_SIZE]] : index
+  // CHECK: %[[RESULT_AXIS_SIZE:.*]] = arith.divui %[[TENSOR_AXIS_SIZE]], %[[SHARD_SIZE]] : index
   // CHECK: %[[SLICE_OFFSET:.*]] = arith.muli %[[PROC_IDX]], %[[RESULT_AXIS_SIZE]] : index
   // CHECK: %[[RESULT:.*]] = tensor.extract_slice %[[ARG]][%[[SLICE_OFFSET]]] [%[[RESULT_AXIS_SIZE]]] [1] : tensor<?xf16> to tensor<?xf16>
-  %0 = mesh.all_slice %arg0 on @mesh_1d mesh_axes = [0] slice_axis = 0 : tensor<?xf16> -> tensor<?xf16>
+  %0 = shard.all_slice %arg0 on @grid_1d grid_axes = [0] slice_axis = 0 : tensor<?xf16> -> tensor<?xf16>
   // CHECK: return %[[RESULT]] : tensor<?xf16>
   return %0 : tensor<?xf16>
 }
 
 // -----
 
-mesh.mesh @mesh_1d(shape = 2)
+shard.grid @grid_1d(shape = 2)
 
-// CHECK-LABEL: func.func @all_slice_op_lowering_of_static_1d_tensor_on_static_1d_mesh
-func.func @all_slice_op_lowering_of_static_1d_tensor_on_static_1d_mesh(
+// CHECK-LABEL: func.func @all_slice_op_lowering_of_static_1d_tensor_on_static_1d_grid
+func.func @all_slice_op_lowering_of_static_1d_tensor_on_static_1d_grid(
   // CHECK: %[[ARG:.*]]: tensor<2xf16>
   %arg0: tensor<2xf16>
 // CHECK-SAME: -> tensor<1xf16> {
 ) -> tensor<1xf16> {
   // CHECK-DAG: %[[C1:.*]] = arith.constant 1 : index
-  // CHECK: %[[PROC_IDX:.*]] = mesh.process_multi_index on @mesh_1d axes = [0] : index
+  // CHECK: %[[PROC_IDX:.*]] = shard.process_multi_index on @grid_1d axes = [0] : index
   // CHECK: %[[SLICE:.*]] = tensor.extract_slice %[[ARG]][%[[PROC_IDX]]] [%[[C1]]] [1] : tensor<2xf16> to tensor<?xf16>
   // CHECK: %[[RESULT:.*]] = tensor.cast %[[SLICE]] : tensor<?xf16> to tensor<1xf16>
-  %0 = mesh.all_slice %arg0 on @mesh_1d mesh_axes = [0] slice_axis = 0 : tensor<2xf16> -> tensor<1xf16>
+  %0 = shard.all_slice %arg0 on @grid_1d grid_axes = [0] slice_axis = 0 : tensor<2xf16> -> tensor<1xf16>
   // CHECK: return %[[RESULT]] : tensor<1xf16>
   return %0 : tensor<1xf16>
 }
@@ -46,18 +46,18 @@ func.func @all_slice_op_lowering_of_static_1d_tensor_on_static_1d_mesh(
 
 // CHECK: #map = affine_map<()[s0, s1, s2] -> (s0 * s1 + s2)>
 
-mesh.mesh @mesh_4d(shape = ?x?x?x?)
+shard.grid @grid_4d(shape = ?x?x?x?)
 
-// CHECK-LABEL: func.func @all_slice_op_lowering_of_dynamic_2d_tensor_on_dynamic_4d_mesh
-func.func @all_slice_op_lowering_of_dynamic_2d_tensor_on_dynamic_4d_mesh(
+// CHECK-LABEL: func.func @all_slice_op_lowering_of_dynamic_2d_tensor_on_dynamic_4d_grid
+func.func @all_slice_op_lowering_of_dynamic_2d_tensor_on_dynamic_4d_grid(
   // CHECK: %[[ARG:.*]]: tensor<?x?xf16>
   %arg0 : tensor<?x?xf16>
 // CHECK-SAME: -> tensor<?x?xf16> {
 ) -> tensor<?x?xf16> {
   // CHECK-DAG: %[[C1:.*]] = arith.constant 1 : index
   // CHECK-DAG: %[[C0:.*]] = arith.constant 0 : index
-  // CHECK-DAG: %[[IN_GROUP_PROC_MULTI_IDX:.*]]:2 = mesh.process_multi_index on @mesh_4d axes = [3, 1] : index, index
-  // CHECK-DAG: %[[PROC_GROUP_SHAPE:.*]]:2 = mesh.mesh_shape @mesh_4d axes = [3, 1] : index, index
+  // CHECK-DAG: %[[IN_GROUP_PROC_MULTI_IDX:.*]]:2 = shard.process_multi_index on @grid_4d axes = [3, 1] : index, index
+  // CHECK-DAG: %[[PROC_GROUP_SHAPE:.*]]:2 = shard.grid_shape @grid_4d axes = [3, 1] : index, index
   // CHECK: %[[PROC_GROUP_SIZE:.*]] = arith.muli %[[PROC_GROUP_SHAPE]]#0, %[[PROC_GROUP_SHAPE]]#1 : index
   // CHECK: %[[SCATTER_AXIS_SIZE:.*]] = tensor.dim %[[ARG]], %[[C1]] : tensor<?x?xf16>
   // CHECK: %[[AXIS_SIZE_CHECK_REMINDER:.*]] = arith.remui %[[SCATTER_AXIS_SIZE]], %[[PROC_GROUP_SIZE]] : index
@@ -68,7 +68,7 @@ func.func @all_slice_op_lowering_of_dynamic_2d_tensor_on_dynamic_4d_mesh(
   // CHECK: %[[AXIS_0_SIZE:.*]] = tensor.dim %[[ARG]], %[[C0]] : tensor<?x?xf16>
   // CHECK: %[[SCATTER_AXIS_OFFSET:.*]] = arith.muli %[[PROC_IN_GROUP_LINEAR_IDX]], %[[RESULT_SCATTER_AXIS_SIZE]] : index
   // CHECK: %[[RESULT:.*]] = tensor.extract_slice %[[ARG]][0, %[[SCATTER_AXIS_OFFSET]]] [%[[AXIS_0_SIZE]], %[[RESULT_SCATTER_AXIS_SIZE]]] [1, 1] : tensor<?x?xf16> to tensor<?x?xf16>
-  %0 = mesh.all_slice %arg0 on @mesh_4d mesh_axes = [3, 1] slice_axis = 1 : tensor<?x?xf16> -> tensor<?x?xf16>
+  %0 = shard.all_slice %arg0 on @grid_4d grid_axes = [3, 1] slice_axis = 1 : tensor<?x?xf16> -> tensor<?x?xf16>
   // CHECK: return %[[RESULT]] : tensor<?x?xf16>
   return %0 : tensor<?x?xf16>
 }
diff --git a/mlir/test/Dialect/Mesh/backward-sharding-propagation.mlir b/mlir/test/Dialect/Shard/backward-sharding-propagation.mlir
index 4223d01..8894c4a 100644
--- a/mlir/test/Dialect/Mesh/backward-sharding-propagation.mlir
+++ b/mlir/test/Dialect/Shard/backward-sharding-propagation.mlir
@@ -2,17 +2,17 @@
 
 #map = affine_map<(d0, d1) -> (d0, d1)>
 module {
-  mesh.mesh @mesh(shape = 1) {sym_visibility = "private"}
+  shard.grid @grid(shape = 1) {sym_visibility = "private"}
   func.func @test_forward() -> tensor<6x6xi32> {
     %c1_i32 = arith.constant 1 : i32
     // CHECK: tensor.empty()
     %0 = tensor.empty() : tensor<6x6xi32>
-    %sharding = mesh.sharding @mesh split_axes = [[0]] : !mesh.sharding
-    // CHECK-COUNT-2: mesh.shard
-    %sharding_annotated = mesh.shard %0 to %sharding : tensor<6x6xi32>
-    %1 = linalg.fill ins(%c1_i32 : i32) outs(%sharding_annotated : tensor<6x6xi32>) -> tensor<6x6xi32>
+    %sharding = shard.sharding @grid split_axes = [[0]] : !shard.sharding
+    // CHECK-COUNT-2: shard.shard
+    %sharded = shard.shard %0 to %sharding : tensor<6x6xi32>
+    %1 = linalg.fill ins(%c1_i32 : i32) outs(%sharded : tensor<6x6xi32>) -> tensor<6x6xi32>
     // CHECK: tensor.empty()
-    // CHECK-NOT: mesh.shard @
+    // CHECK-NOT: shard.shard @
     %2 = tensor.empty() : tensor<6x6xi32>
     %3 = linalg.generic {indexing_maps = [#map, #map, #map], iterator_types = ["parallel", "parallel"]} ins(%1, %1
         : tensor<6x6xi32>, tensor<6x6xi32>) outs(%2 : tensor<6x6xi32>) {
diff --git a/mlir/test/Dialect/Shard/canonicalization.mlir b/mlir/test/Dialect/Shard/canonicalization.mlir
new file mode 100644
index 0000000..ed40dfb
--- /dev/null
+++ b/mlir/test/Dialect/Shard/canonicalization.mlir
@@ -0,0 +1,248 @@
+// RUN: mlir-opt --canonicalize %s | FileCheck %s
+
+shard.grid @grid0(shape = 2x4)
+
+// CHECK-LABEL: func @all_reduce_empty_grid_axes
+func.func @all_reduce_empty_grid_axes(
+// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
+    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
+// CHECK-NOT: shard.all_reduce
+  %0 = shard.all_reduce %arg0 on @grid0
+    grid_axes = []
+    : tensor<4xf32> -> tensor<4xf32>
+// CHECK: return %[[ARG]]
+  return %0 : tensor<4xf32>
+}
+
+// CHECK-LABEL: func @all_reduce_empty_grid_axes_different_return_type
+func.func @all_reduce_empty_grid_axes_different_return_type(
+    %arg0 : tensor<4xf32>) -> tensor<4xf64> {
+// CHECK: shard.all_reduce
+  %0 = shard.all_reduce %arg0 on @grid0
+// CHECK-NOT: grid_axes
+    grid_axes = []
+    : tensor<4xf32> -> tensor<4xf64>
+  return %0 : tensor<4xf64>
+}
+
+// CHECK-LABEL: func @all_reduce_default_reduction
+func.func @all_reduce_default_reduction(
+    %arg0 : tensor<4xf32>) -> tensor<4xf64> {
+  %0 = shard.all_reduce %arg0 on @grid0
+    grid_axes = [0]
+// CHECK-NOT: reduction
+    reduction = sum
+    : tensor<4xf32> -> tensor<4xf64>
+  return %0 : tensor<4xf64>
+}
+
+// CHECK-LABEL: func @all_to_all_empty_grid_axes
+func.func @all_to_all_empty_grid_axes(
+// CHECK-SAME: %[[ARG:.*]]: tensor<8xf32>
+    %arg0 : tensor<8xf32>) -> tensor<8xf32> {
+// CHECK-NOT: shard.all_to_all
+  %0 = shard.all_to_all %arg0 on @grid0
+    grid_axes = []
+    split_axis = 0
+    concat_axis = 0
+    : tensor<8xf32> -> tensor<8xf32>
+// CHECK: return %[[ARG]]
+  return %0 : tensor<8xf32>
+}
+
+// CHECK-LABEL: func @all_gather_empty_grid_axes
+func.func @all_gather_empty_grid_axes(
+// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
+    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
+// CHECK-NOT: shard.all_gather
+  %0 = shard.all_gather %arg0 on @grid0
+    grid_axes = []
+    gather_axis = 0
+    : tensor<4xf32> -> tensor<4xf32>
+// CHECK: return %[[ARG]]
+  return %0 : tensor<4xf32>
+}
+
+// CHECK-LABEL: func @all_slice_empty_grid_axes
+func.func @all_slice_empty_grid_axes(
+// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
+    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
+// CHECK-NOT: shard.scatter
+  %0 = shard.all_slice %arg0 on @grid0
+    grid_axes = []
+    slice_axis = 0
+    : tensor<4xf32> -> tensor<4xf32>
+// CHECK: return %[[ARG]]
+  return %0 : tensor<4xf32>
+}
+
+// CHECK-LABEL: func @broadcast_empty_grid_axes
+func.func @broadcast_empty_grid_axes(
+// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
+    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
+// CHECK-NOT: shard.broadcast
+  %0 = shard.broadcast %arg0 on @grid0
+    grid_axes = []
+    root = []
+    : (tensor<4xf32>) -> tensor<4xf32>
+// CHECK: return %[[ARG]]
+  return %0 : tensor<4xf32>
+}
+
+// CHECK-LABEL: func @gather_empty_grid_axes
+func.func @gather_empty_grid_axes(
+// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
+    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
+// CHECK-NOT: shard.gather
+  %0 = shard.gather %arg0 on @grid0
+    grid_axes = []
+    gather_axis = 0
+    root = []
+    : (tensor<4xf32>) -> tensor<4xf32>
+// CHECK: return %[[ARG]]
+  return %0 : tensor<4xf32>
+}
+
+// CHECK-LABEL: func @receive_empty_grid_axes
+func.func @receive_empty_grid_axes(
+// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
+    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
+// CHECK-NOT: shard.recv
+  %0 = shard.recv %arg0 on @grid0
+    grid_axes = []
+    : (tensor<4xf32>) -> tensor<4xf32>
+// CHECK: return %[[ARG]]
+  return %0 : tensor<4xf32>
+}
+
+// CHECK-LABEL: func @reduce_empty_grid_axes
+func.func @reduce_empty_grid_axes(
+// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
+    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
+// CHECK-NOT: shard.reduce
+  %0 = shard.reduce %arg0 on @grid0
+    grid_axes = []
+    root = []
+    : (tensor<4xf32>) -> tensor<4xf32>
+// CHECK: return %[[ARG]]
+  return %0 : tensor<4xf32>
+}
+
+// CHECK-LABEL: func @reduce_scatter_empty_grid_axes
+func.func @reduce_scatter_empty_grid_axes(
+// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
+    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
+// CHECK-NOT: shard.reduce_scatter
+  %0 = shard.reduce_scatter %arg0 on @grid0
+    grid_axes = []
+    scatter_axis = 0
+    : tensor<4xf32> -> tensor<4xf32>
+// CHECK: return %[[ARG]]
+  return %0 : tensor<4xf32>
+}
+
+// CHECK-LABEL: func @reduce_scatter_empty_grid_axes_different_return_type
+func.func @reduce_scatter_empty_grid_axes_different_return_type(
+    %arg0 : tensor<4xf32>) -> tensor<4xf64> {
+// CHECK: shard.reduce_scatter
+  %0 = shard.reduce_scatter %arg0 on @grid0
+// CHECK-NOT: grid_axes
+    grid_axes = []
+    scatter_axis = 0
+    : tensor<4xf32> -> tensor<4xf64>
+  return %0 : tensor<4xf64>
+}
+
+// CHECK-LABEL: func @reduce_scatter_default_reduction
+func.func @reduce_scatter_default_reduction(
+    %arg0 : tensor<4xf32>) -> tensor<2xf64> {
+  %0 = shard.reduce_scatter %arg0 on @grid0
+    grid_axes = [0]
+// CHECK-NOT: reduction
+    reduction = sum
+    scatter_axis = 0
+    : tensor<4xf32> -> tensor<2xf64>
+  return %0 : tensor<2xf64>
+}
+
+// CHECK-LABEL: func @scatter_empty_grid_axes
+func.func @scatter_empty_grid_axes(
+// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
+    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
+// CHECK-NOT: shard.scatter
+  %0 = shard.scatter %arg0 on @grid0
+    grid_axes = []
+    scatter_axis = 0
+    root = []
+    : (tensor<4xf32>) -> tensor<4xf32>
+// CHECK: return %[[ARG]]
+  return %0 : tensor<4xf32>
+}
+
+// CHECK-LABEL: func @send_empty_grid_axes
+func.func @send_empty_grid_axes(
+// CHECK-SAME: %[[ARG:.*]]: tensor<4xf32>
+    %arg0 : tensor<4xf32>) -> tensor<4xf32> {
+// CHECK-NOT: shard.send
+  %0 = shard.send %arg0 on @grid0
+    grid_axes = []
+    destination = []
+    : (tensor<4xf32>) -> tensor<4xf32>
+// CHECK: return %[[ARG]]
+  return %0 : tensor<4xf32>
+}
+
+shard.grid @grid4x4(shape = 4x4)
+// CHECK-LABEL: func @test_halo_sizes
+func.func @test_halo_sizes() -> !shard.sharding {
+  %c2_i64 = arith.constant 2 : i64
+  // CHECK shard.sharding @grid4x4 split_axes = [[0], [1]] halo_sizes = [1, 2, 2, 22] : !shard.sharding
+  %sharding = shard.sharding @grid4x4 split_axes = [[0], [1]] halo_sizes = [1, %c2_i64, %c2_i64, 22] : !shard.sharding
+  return %sharding : !shard.sharding
+}
+
+// CHECK-LABEL: func @test_shard_offs
+func.func @test_shard_offs() -> !shard.sharding {
+  %c2_i64 = arith.constant 2 : i64
+  // CHECK shard.sharding @grid4x4 split_axes = [[0], [1]] sharded_dims_offsets = [0, 1, 2, 3, 4, 0, 2, 3, 4, 22] : !shard.sharding
+  %sharding = shard.sharding @grid4x4 split_axes = [[0], [1]] sharded_dims_offsets = [0, 1, %c2_i64, 3, 4, 0, %c2_i64, 3, 4, 22] : !shard.sharding
+  return %sharding : !shard.sharding
+}
+
+// CHECK-LABEL: func @test_duplicate_shardops
+func.func @test_duplicate_shardops() -> (tensor<1024x1024xf32>, tensor<1024x1024xf32>) attributes {llvm.emit_c_interface} {
+  // CHECK-NEXT: [[vcst:%.*]] = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
+  %cst_1 = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
+  // CHECK-NEXT: [[vsharding:%.*]] = shard.sharding @grid4x4 split_axes = {{\[\[}}0, 1]] : !shard.sharding
+  %sharding_1 = shard.sharding @grid4x4 split_axes = [[0, 1]] : !shard.sharding
+  %cst_2 = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
+  %sharding_2 = shard.sharding @grid4x4 split_axes = [[0, 1]] : !shard.sharding
+  %sharded_2 = shard.shard %cst_2 to %sharding_2 : tensor<1024x1024xf32>
+  %cst_3 = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
+  %sharding_3 = shard.sharding @grid4x4 split_axes = [[0, 1]] : !shard.sharding
+  %sharded_3 = shard.shard %cst_3 to %sharding_3 : tensor<1024x1024xf32>
+  // CHECK-NEXT: [[vsharded:%.*]] = shard.shard [[vcst]] to [[vsharding]] : tensor<1024x1024xf32>
+  %sharded_1 = shard.shard %cst_1 to %sharding_1 : tensor<1024x1024xf32>
+  // CHECK-NEXT: return [[vsharded]], [[vsharded]] : tensor<1024x1024xf32>, tensor<1024x1024xf32>
+  return %sharded_1, %sharded_2 : tensor<1024x1024xf32>, tensor<1024x1024xf32>
+}
+
+// CHECK-LABEL: func @test_duplicate_shardops_diff
+func.func @test_duplicate_shardops_diff() -> (tensor<1024x1024xf32>, tensor<1024x1024xf32>) attributes {llvm.emit_c_interface} {
+  // CHECK-NEXT: [[vcst:%.*]] = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
+  %cst_1 = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
+  // CHECK-NEXT: [[vsharding:%.*]] = shard.sharding @grid4x4 split_axes = {{\[\[}}0]] : !shard.sharding
+  %sharding_1 = shard.sharding @grid4x4 split_axes = [[0]] : !shard.sharding
+  %cst_2 = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
+  // CHECK-NEXT: [[vsharding_0:%.*]] = shard.sharding @grid4x4 split_axes = {{\[\[}}0, 1]] : !shard.sharding
+  %sharding_2 = shard.sharding @grid4x4 split_axes = [[0, 1]] : !shard.sharding
+  // CHECK-NEXT: [[vsharded:%.*]] = shard.shard [[vcst]] to [[vsharding_0]] : tensor<1024x1024xf32>
+  %sharded_2 = shard.shard %cst_2 to %sharding_2 : tensor<1024x1024xf32>
+  %cst_3 = arith.constant dense<0.000000e+00> : tensor<1024x1024xf32>
+  %sharding_3 = shard.sharding @grid4x4 split_axes = [[0]] : !shard.sharding
+  %sharded_3 = shard.shard %cst_3 to %sharding_3 : tensor<1024x1024xf32>
+  // CHECK-NEXT: [[vsharded_1:%.*]] = shard.shard [[vsharded]] to [[vsharding]] : tensor<1024x1024xf32>
+  %sharded_1 = shard.shard %cst_1 to %sharding_1 : tensor<1024x1024xf32>
+  // CHECK-NEXT: return [[vsharded_1]], [[vsharded]] : tensor<1024x1024xf32>, tensor<1024x1024xf32>
+  return %sharded_1, %sharded_2 : tensor<1024x1024xf32>, tensor<1024x1024xf32>
+}
diff --git a/mlir/test/Dialect/Shard/folding.mlir b/mlir/test/Dialect/Shard/folding.mlir
new file mode 100644
index 0000000..5a0f35b
--- /dev/null
+++ b/mlir/test/Dialect/Shard/folding.mlir
@@ -0,0 +1,22 @@
+// RUN: mlir-opt -test-grid-simplifications %s | FileCheck %s
+
+shard.grid @grid0(shape = 4x?x2)
+shard.grid @grid1(shape = 2x3)
+
+// CHECK-LABEL: func.func @grid_shape_op_folding
+func.func @grid_shape_op_folding() -> (index, index) {
+  // CHECK: %[[AXIS_2_SIZE:.*]] = arith.constant 2 : index
+  // CHECK: %[[AXIS_1_SIZE:.*]] = shard.grid_shape @grid0 axes = [1] : index
+  %0:2 = shard.grid_shape @grid0 axes = [2, 1] : index, index
+  // CHECK: return %[[AXIS_2_SIZE]], %[[AXIS_1_SIZE]]
+  return %0#0, %0#1 : index, index
+}
+
+// CHECK-LABEL: func.func @grid_shape_op_folding_all_axes_static_grid
+func.func @grid_shape_op_folding_all_axes_static_grid() -> (index, index) {
+  // CHECK: %[[AXIS_0_SIZE:.*]] = arith.constant 2 : index
+  // CHECK: %[[AXIS_1_SIZE:.*]] = arith.constant 3 : index
+  %0:2 = shard.grid_shape @grid1 : index, index
+  // CHECK: return %[[AXIS_0_SIZE]], %[[AXIS_1_SIZE]]
+  return %0#0, %0#1 : index, index
+}
diff --git a/mlir/test/Dialect/Mesh/forward-backward-sharding-propagation.mlir b/mlir/test/Dialect/Shard/forward-backward-sharding-propagation.mlir
index dd2eee2..0d8d997 100644
--- a/mlir/test/Dialect/Mesh/forward-backward-sharding-propagation.mlir
+++ b/mlir/test/Dialect/Shard/forward-backward-sharding-propagation.mlir
@@ -2,25 +2,25 @@
 
 #map = affine_map<(d0, d1) -> (d0, d1)>
 module {
-  mesh.mesh @mesh(shape = 1) {sym_visibility = "private"}
+  shard.grid @grid(shape = 1) {sym_visibility = "private"}
   func.func @test_forward() -> tensor<6x6xi32> {
     %c1_i32 = arith.constant 1 : i32
     // CHECK: tensor.empty()
     %0 = tensor.empty() : tensor<6x6xi32>
-    // CHECK-COUNT-3: mesh.sharding @mesh split_axes = {{\[\[0}}]]
-    %sharding_row = mesh.sharding @mesh split_axes = [[0]] : !mesh.sharding
-    %annotated_row = mesh.shard %0 to %sharding_row : tensor<6x6xi32>
+    // CHECK-COUNT-3: shard.sharding @grid split_axes = {{\[\[0}}]]
+    %sharding_row = shard.sharding @grid split_axes = [[0]] : !shard.sharding
+    %annotated_row = shard.shard %0 to %sharding_row : tensor<6x6xi32>
     %1 = linalg.fill ins(%c1_i32 : i32) outs(%annotated_row : tensor<6x6xi32>) -> tensor<6x6xi32>
     %2 = tensor.empty() : tensor<6x6xi32>
-    // CHECK-COUNT-4: mesh.sharding @mesh split_axes = {{\[\[1}}]]
+    // CHECK-COUNT-4: shard.sharding @grid split_axes = {{\[\[1}}]]
     %3 = linalg.generic {indexing_maps = [#map, #map, #map], iterator_types = ["parallel", "parallel"]} ins(%2, %1
         : tensor<6x6xi32>, tensor<6x6xi32>) outs(%2 : tensor<6x6xi32>) {
     ^bb0(%in: i32, %in_2: i32, %out: i32):
       %9 = arith.addi %in, %in_2 : i32
       linalg.yield %9 : i32
     } -> tensor<6x6xi32>
-    %sharding_col = mesh.sharding @mesh split_axes = [[1]] : !mesh.sharding
-    %annotated_col = mesh.shard %3 to %sharding_col : tensor<6x6xi32>
+    %sharding_col = shard.sharding @grid split_axes = [[1]] : !shard.sharding
+    %annotated_col = shard.shard %3 to %sharding_col : tensor<6x6xi32>
     // CHECK: return
     return %annotated_col : tensor<6x6xi32>
   }
diff --git a/mlir/test/Dialect/Shard/forward-sharding-propagation.mlir b/mlir/test/Dialect/Shard/forward-sharding-propagation.mlir
new file mode 100644
index 0000000..3cda9ea
--- /dev/null
+++ b/mlir/test/Dialect/Shard/forward-sharding-propagation.mlir
@@ -0,0 +1,49 @@
+// RUN: mlir-opt --pass-pipeline="builtin.module(func.func(sharding-propagation{traversal=forward}))" %s | FileCheck %s
+
+#map = affine_map<(d0, d1) -> (d0, d1)>
+module attributes {dlti.map = #dlti.map<"MPI:Implementation" = "mpich", "MPI:comm_world_rank" = 0 : i32>} {
+  shard.grid @grid(shape = 1) {sym_visibility = "private"}
+  func.func @test_forward() -> (tensor<6x6xi32>, tensor<6x6xi32>, tensor<i32>) attributes {llvm.emit_c_interface} {
+    %c1_i32 = arith.constant 1 : i32
+    // CHECK: [[v3:%.*]] = tensor.empty() : tensor<6x6xi32>
+    %0 = tensor.empty() : tensor<6x6xi32>
+    // CHECK: [[v1:%.*]] = linalg.fill ins
+    // CHECK: [[vsharding_0:%.*]] = shard.sharding @grid split_axes = {{\[\[}}0]] : !shard.sharding
+    // CHECK: [[vsharded_1:%.*]] = shard.shard [[v1]] to [[vsharding_0]] : tensor<6x6xi32>
+    %1 = linalg.fill ins(%c1_i32 : i32) outs(%0 : tensor<6x6xi32>) -> tensor<6x6xi32>
+    %sharding = shard.sharding @grid split_axes = [[0]] : !shard.sharding
+    %sharded = shard.shard %1 to %sharding : tensor<6x6xi32>
+    // CHECK: [[v2:%.*]] = tensor.empty() : tensor<6x6xi32>
+    // CHECK: [[vsharding_2:%.*]] = shard.sharding @grid split_axes = {{\[\[}}0]] : !shard.sharding
+    // CHECK: [[vsharded_3:%.*]] = shard.shard [[vsharded_1]] to [[vsharding_2]] annotate_for_users : tensor<6x6xi32>
+    %3 = tensor.empty() : tensor<6x6xi32>
+    // CHECK: [[vsharding_4:%.*]] = shard.sharding @grid split_axes = {{\[\[}}0]] : !shard.sharding
+    // CHECK: [[vsharded_5:%.*]] = shard.shard [[v2]] to [[vsharding_4]] annotate_for_users : tensor<6x6xi32>
+    // CHECK: [[v3:%.*]] = linalg.generic {indexing_maps = [#map, #map, #map], iterator_types = ["parallel", "parallel"]}
+    // CHECK-SAME: ins([[vsharded_3]], [[vsharded_3]] : tensor<6x6xi32>, tensor<6x6xi32>) outs([[vsharded_5]] : tensor<6x6xi32>) {
+    // CHECK: [[vsharding_6:%.*]] = shard.sharding @grid split_axes = {{\[\[}}0]] : !shard.sharding
+    // CHECK: [[vsharded_7:%.*]] = shard.shard [[v3]] to [[vsharding_6]] : tensor<6x6xi32>
+    %4 = linalg.generic {indexing_maps = [#map, #map, #map], iterator_types = ["parallel", "parallel"]} ins(%sharded, %sharded
+        : tensor<6x6xi32>, tensor<6x6xi32>) outs(%3 : tensor<6x6xi32>) {
+    ^bb0(%in: i32, %in_2: i32, %out: i32):
+      %9 = arith.addi %in, %in_2 : i32
+      linalg.yield %9 : i32
+    } -> tensor<6x6xi32>
+    %c0_i32 = arith.constant 0 : i32
+    %6 = tensor.empty() : tensor<i32>
+    %7 = linalg.fill ins(%c0_i32 : i32) outs(%6 : tensor<i32>) -> tensor<i32>
+    // CHECK: [[vreduced:%.*]] = linalg.reduce ins
+    // CHECK: [[vsharding_12:%.*]] = shard.sharding @grid split_axes = [] : !shard.sharding
+    // CHECK: [[vsharded_13:%.*]] = shard.shard [[vreduced]] to [[vsharding_12]] : tensor<i32>
+    %reduced = linalg.reduce ins(%4 : tensor<6x6xi32>) outs(%7 : tensor<i32>) dimensions = [0, 1] 
+      (%in: i32, %init: i32) {
+        %9 = arith.addi %in, %init : i32
+        linalg.yield %9 : i32
+      }
+    // CHECK: [[vsharding_14:%.*]] = shard.sharding @grid split_axes = {{\[\[}}]] : !shard.sharding
+    %sharding_0 = shard.sharding @grid split_axes = [[]] : !shard.sharding
+    // CHECK: [[vsharded_15:%.*]] = shard.shard [[vsharded_13]] to [[vsharding_14]] annotate_for_users : tensor<i32>
+    %sharded_1 = shard.shard %reduced to %sharding_0 annotate_for_users : tensor<i32>
+    return %sharded, %4, %sharded_1 : tensor<6x6xi32>, tensor<6x6xi32>, tensor<i32>
+  }
+}
diff --git a/mlir/test/Dialect/Shard/inlining.mlir b/mlir/test/Dialect/Shard/inlining.mlir
new file mode 100644
index 0000000..ce664b3
--- /dev/null
+++ b/mlir/test/Dialect/Shard/inlining.mlir
@@ -0,0 +1,15 @@
+// RUN: mlir-opt -inline %s | FileCheck %s
+
+shard.grid @grid0(shape = 4x?x2)
+
+func.func private @grid_to_inline() -> (index, index) {
+  %0:2 = shard.grid_shape @grid0 axes = [2, 1] : index, index
+  return %0#0, %0#1 : index, index
+}
+// CHECK-LABEL: func.func @main
+func.func @main() -> (index, index) {
+  // CHECK-NEXT: %[[AXIS_SIZE:.*]]:2 = shard.grid_shape @grid0 axes = [2, 1] : index
+  %0:2 = func.call @grid_to_inline() : () -> (index, index)
+  // CHECK-NEXT: return %[[AXIS_SIZE]]#0, %[[AXIS_SIZE]]#1
+  return %0#0, %0#1 : index, index
+}
diff --git a/mlir/test/Dialect/Mesh/invalid.mlir b/mlir/test/Dialect/Shard/invalid.mlir
index 2656332..6acac97 100644
--- a/mlir/test/Dialect/Mesh/invalid.mlir
+++ b/mlir/test/Dialect/Shard/invalid.mlir
@@ -1,55 +1,55 @@
 // RUN: mlir-opt -split-input-file -verify-diagnostics %s
 
-// expected-error@+1 {{rank of mesh is expected to be a positive integer}}
-mesh.mesh @mesh0(shape = [])
+// expected-error@+1 {{rank of grid is expected to be a positive integer}}
+shard.grid @grid0(shape = [])
 
 // -----
 
-// expected-error@+1 {{custom op 'mesh.mesh' Failed parsing dimension list. Did you mean an empty list? It must be denoted by "[]".}}
-mesh.mesh @mesh0(shape = -1)
+// expected-error@+1 {{custom op 'shard.grid' Failed parsing dimension list. Did you mean an empty list? It must be denoted by "[]".}}
+shard.grid @grid0(shape = -1)
 
 // -----
 
-mesh.mesh @mesh0(shape = 2x4)
+shard.grid @grid0(shape = 2x4)
 
-func.func @mesh_axis_duplicated_different_subarray(
+func.func @grid_axis_duplicated_different_subarray(
     %arg0 : tensor<4x8xf32>) -> tensor<4x8xf32> {
-  // expected-error@+1 {{mesh axis duplicated}}
-  %s = mesh.sharding @mesh0 split_axes = [[0], [0]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s : tensor<4x8xf32>
+  // expected-error@+1 {{grid axis duplicated}}
+  %s = shard.sharding @grid0 split_axes = [[0], [0]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s : tensor<4x8xf32>
   return %0 : tensor<4x8xf32>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 2x4)
+shard.grid @grid0(shape = 2x4)
 
-func.func @mesh_axis_duplicated_same_subarray(
+func.func @grid_axis_duplicated_same_subarray(
     %arg0 : tensor<4x8xf32>) -> tensor<4x8xf32> {
-  // expected-error@+1 {{mesh axis duplicated}}
-  %s = mesh.sharding @mesh0 split_axes = [[0, 0]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s : tensor<4x8xf32>
+  // expected-error@+1 {{grid axis duplicated}}
+  %s = shard.sharding @grid0 split_axes = [[0, 0]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s : tensor<4x8xf32>
   return %0 : tensor<4x8xf32>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 2x4)
+shard.grid @grid0(shape = 2x4)
 
-func.func @mesh_axis_negtive_in_split_part(
+func.func @grid_axis_negtive_in_split_part(
     %arg0 : tensor<4x8xf32>) -> tensor<4x8xf32> {
-  // expected-error@+1 {{mesh axis is expected to be non-negative}}
-  %s = mesh.sharding @mesh0 split_axes = [[-1]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s : tensor<4x8xf32>
+  // expected-error@+1 {{grid axis is expected to be non-negative}}
+  %s = shard.sharding @grid0 split_axes = [[-1]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s : tensor<4x8xf32>
   return %0 : tensor<4x8xf32>
 }
 
 // -----
 
 func.func @sharding_attribute_invalid_nested_symbol(%arg0 : tensor<4x8xf32>) {
-  // expected-error@+1 {{custom op 'mesh.sharding' invalid kind of attribute specified}}
-  %s = mesh.sharding @a::@b split_axes = [[0]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s : tensor<4x8xf32>
+  // expected-error@+1 {{custom op 'shard.sharding' invalid kind of attribute specified}}
+  %s = shard.sharding @a::@b split_axes = [[0]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s : tensor<4x8xf32>
   return
 }
 
@@ -57,8 +57,8 @@ func.func @sharding_attribute_invalid_nested_symbol(%arg0 : tensor<4x8xf32>) {
 
 func.func @sharding_attribute_invalid_halo(%arg0 : tensor<4x8xf32>) {
   // expected-error@+1 {{halo sizes must be specified for all split axes}}
-  %s = mesh.sharding @mesh0 split_axes = [[0], [1]] halo_sizes = [1, 2] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s : tensor<4x8xf32>
+  %s = shard.sharding @grid0 split_axes = [[0], [1]] halo_sizes = [1, 2] : !shard.sharding
+  %0 = shard.shard %arg0 to %s : tensor<4x8xf32>
   return
 }
 
@@ -66,292 +66,292 @@ func.func @sharding_attribute_invalid_halo(%arg0 : tensor<4x8xf32>) {
 
 func.func @sharding_attribute_invalid_sizes(%arg0 : tensor<4x8xf32>) {
   // expected-error@+1 {{halo sizes and shard offsets are mutually exclusive}}
-  %s = mesh.sharding @mesh0 split_axes = [[0]] halo_sizes = [1, 2] sharded_dims_offsets = [0, 2, 2] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s : tensor<4x8xf32>
+  %s = shard.sharding @grid0 split_axes = [[0]] halo_sizes = [1, 2] sharded_dims_offsets = [0, 2, 2] : !shard.sharding
+  %0 = shard.shard %arg0 to %s : tensor<4x8xf32>
   return
 }
 
 // -----
 
-mesh.mesh @mesh_dyn(shape = ?x?)
-func.func @sharding_dyn_mesh_and_sizes(%arg0 : tensor<4x8xf32>) {
-  // expected-error@+1 {{sharded dims offsets are not allowed for devices meshes with dynamic shape}}
-  %s = mesh.sharding @mesh_dyn split_axes = [[0]] sharded_dims_offsets = [0, 2, 2] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s : tensor<4x8xf32>
+shard.grid @grid_dyn(shape = ?x?)
+func.func @sharding_dyn_grid_and_sizes(%arg0 : tensor<4x8xf32>) {
+  // expected-error@+1 {{sharded dims offsets are not allowed for device grids with dynamic shape}}
+  %s = shard.sharding @grid_dyn split_axes = [[0]] sharded_dims_offsets = [0, 2, 2] : !shard.sharding
+  %0 = shard.shard %arg0 to %s : tensor<4x8xf32>
   return
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 2x4)
+shard.grid @grid0(shape = 2x4)
 func.func @sharding_sizes_count(%arg0 : tensor<4x8xf32>) {
   // expected-error@+1 {{sharded dims offsets has wrong size}}
-  %s = mesh.sharding @mesh0 split_axes = [[0], [1]] sharded_dims_offsets = [0, 2, 4, 0, 2, 4, 6] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s : tensor<4x8xf32>
+  %s = shard.sharding @grid0 split_axes = [[0], [1]] sharded_dims_offsets = [0, 2, 4, 0, 2, 4, 6] : !shard.sharding
+  %0 = shard.shard %arg0 to %s : tensor<4x8xf32>
   return
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 4)
+shard.grid @grid0(shape = 4)
 func.func @sharding_sizes_decreasing(%arg0 : tensor<4x8xf32>) {
   // expected-error@+1 {{sharded dims offsets must be non-decreasing}}
-  %s = mesh.sharding @mesh0 split_axes = [[0]] sharded_dims_offsets = [0, 2, 3, 2] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s : tensor<4x8xf32>
+  %s = shard.sharding @grid0 split_axes = [[0]] sharded_dims_offsets = [0, 2, 3, 2] : !shard.sharding
+  %0 = shard.shard %arg0 to %s : tensor<4x8xf32>
   return
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 2x4)
+shard.grid @grid0(shape = 2x4)
 
-func.func @mesh_shape_mesh_axis_out_of_bounds() -> (index, index) {
-  // expected-error@+1 {{0-based mesh axis index 2 is out of bounds. The referenced mesh "mesh0" is of rank 2.}}
-  %0:2 = mesh.mesh_shape @mesh0 axes = [0, 2] : index, index
+func.func @grid_shape_grid_axis_out_of_bounds() -> (index, index) {
+  // expected-error@+1 {{0-based grid axis index 2 is out of bounds. The referenced grid "grid0" is of rank 2.}}
+  %0:2 = shard.grid_shape @grid0 axes = [0, 2] : index, index
   return %0#0, %0#1 : index, index
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 1x2x3)
+shard.grid @grid0(shape = 1x2x3)
 
-func.func @mesh_shape_duplicate_mesh_axis() -> (index, index, index) {
-  // expected-error@+1 {{Mesh axes contains duplicate elements.}}
-  %0:3 = mesh.mesh_shape @mesh0 axes = [0, 2, 0] : index, index, index
+func.func @grid_shape_duplicate_grid_axis() -> (index, index, index) {
+  // expected-error@+1 {{Grid axes contains duplicate elements.}}
+  %0:3 = shard.grid_shape @grid0 axes = [0, 2, 0] : index, index, index
   return %0#0, %0#1, %0#2 : index, index, index
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 2x4)
+shard.grid @grid0(shape = 2x4)
 
-func.func @mesh_shape_wrong_number_of_results() -> (index, index) {
+func.func @grid_shape_wrong_number_of_results() -> (index, index) {
   // expected-error@+1 {{Unexpected number of results 2. Expected 1.}}
-  %0:2 = mesh.mesh_shape @mesh0 axes = [0] : index, index
+  %0:2 = shard.grid_shape @grid0 axes = [0] : index, index
   return %0#0, %0#1 : index, index
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 1x2x3)
+shard.grid @grid0(shape = 1x2x3)
 
-func.func @mesh_shape_wrong_number_of_results_empty_mesh_axes() -> (index, index) {
+func.func @grid_shape_wrong_number_of_results_empty_grid_axes() -> (index, index) {
   // expected-error@+1 {{Unexpected number of results 2. Expected 3.}}
-  %0:2 = mesh.mesh_shape @mesh0 : index, index
+  %0:2 = shard.grid_shape @grid0 : index, index
   return %0#0, %0#1 : index, index
 }
 
 // -----
 
-func.func @mesh_shape_invalid_mesh_name() -> (index) {
-  // expected-error@+1 {{Undefined required mesh symbol "this_mesh_symbol_does_not_exist".}}
-  %0 = mesh.mesh_shape @this_mesh_symbol_does_not_exist : index
+func.func @grid_shape_invalid_grid_name() -> (index) {
+  // expected-error@+1 {{Undefined required grid symbol "this_grid_symbol_does_not_exist".}}
+  %0 = shard.grid_shape @this_grid_symbol_does_not_exist : index
   return %0#0 : index
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 2x4)
+shard.grid @grid0(shape = 2x4)
 
-func.func @process_multi_index_mesh_axis_out_of_bounds() -> (index, index) {
-  // expected-error@+1 {{0-based mesh axis index 2 is out of bounds. The referenced mesh "mesh0" is of rank 2.}}
-  %0:2 = mesh.process_multi_index on @mesh0 axes = [0, 2] : index, index
+func.func @process_multi_index_grid_axis_out_of_bounds() -> (index, index) {
+  // expected-error@+1 {{0-based grid axis index 2 is out of bounds. The referenced grid "grid0" is of rank 2.}}
+  %0:2 = shard.process_multi_index on @grid0 axes = [0, 2] : index, index
   return %0#0, %0#1 : index, index
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 1x2x3)
+shard.grid @grid0(shape = 1x2x3)
 
-func.func @process_multi_index_duplicate_mesh_axis() -> (index, index, index) {
-  // expected-error@+1 {{Mesh axes contains duplicate elements.}}
-  %0:3 = mesh.process_multi_index on @mesh0 axes = [0, 2, 0] : index, index, index
+func.func @process_multi_index_duplicate_grid_axis() -> (index, index, index) {
+  // expected-error@+1 {{Grid axes contains duplicate elements.}}
+  %0:3 = shard.process_multi_index on @grid0 axes = [0, 2, 0] : index, index, index
   return %0#0, %0#1, %0#2 : index, index, index
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 2x4)
+shard.grid @grid0(shape = 2x4)
 
 func.func @process_multi_index_wrong_number_of_results() -> (index, index) {
   // expected-error@+1 {{Unexpected number of results 2. Expected 1.}}
-  %0:2 = mesh.process_multi_index on @mesh0 axes = [0] : index, index
+  %0:2 = shard.process_multi_index on @grid0 axes = [0] : index, index
   return %0#0, %0#1 : index, index
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 1x2x3)
+shard.grid @grid0(shape = 1x2x3)
 
-func.func @process_multi_index_wrong_number_of_results_empty_mesh_axes() -> (index, index) {
+func.func @process_multi_index_wrong_number_of_results_empty_grid_axes() -> (index, index) {
   // expected-error@+1 {{Unexpected number of results 2. Expected 3.}}
-  %0:2 = mesh.process_multi_index on @mesh0 : index, index
+  %0:2 = shard.process_multi_index on @grid0 : index, index
   return %0#0, %0#1 : index, index
 }
 
 // -----
 
-func.func @process_multi_index_invalid_mesh_name() -> (index) {
-  // expected-error@+1 {{Undefined required mesh symbol "this_mesh_symbol_does_not_exist".}}
-  %0 = mesh.process_multi_index on @this_mesh_symbol_does_not_exist : index
+func.func @process_multi_index_invalid_grid_name() -> (index) {
+  // expected-error@+1 {{Undefined required grid symbol "this_grid_symbol_does_not_exist".}}
+  %0 = shard.process_multi_index on @this_grid_symbol_does_not_exist : index
   return %0 : index
 }
 
 // -----
 
-func.func @process_linear_index_invalid_mesh_name() -> (index) {
-  // expected-error@+1 {{Undefined required mesh symbol "this_mesh_symbol_does_not_exist".}}
-  %0 = mesh.process_linear_index on @this_mesh_symbol_does_not_exist : index
+func.func @process_linear_index_invalid_grid_name() -> (index) {
+  // expected-error@+1 {{Undefined required grid symbol "this_grid_symbol_does_not_exist".}}
+  %0 = shard.process_linear_index on @this_grid_symbol_does_not_exist : index
   return %0 : index
 }
 
 // -----
 
-func.func @all_reduce_invalid_mesh_symbol(
+func.func @all_reduce_invalid_grid_symbol(
     %arg0 : tensor<4xf32>) -> tensor<4xf64> {
-  // expected-error@+1 {{Undefined required mesh symbol "this_mesh_symbol_does_not_exist".}}
-  %0 = mesh.all_reduce %arg0 on @this_mesh_symbol_does_not_exist reduction = sum
+  // expected-error@+1 {{Undefined required grid symbol "this_grid_symbol_does_not_exist".}}
+  %0 = shard.all_reduce %arg0 on @this_grid_symbol_does_not_exist reduction = sum
     : tensor<4xf32> -> tensor<4xf64>
   return %0 : tensor<4xf64>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 2x4)
+shard.grid @grid0(shape = 2x4)
 
-func.func @all_reduce_invalid_mesh_axis(
+func.func @all_reduce_invalid_grid_axis(
     %arg0 : tensor<4xf32>) -> tensor<4xf64> {
-  // expected-error@+1 {{0-based mesh axis index 2 is out of bounds. The referenced mesh "mesh0" is of rank 2.}}
-  %0 = mesh.all_reduce %arg0 on @mesh0 mesh_axes = [2] reduction = sum
+  // expected-error@+1 {{0-based grid axis index 2 is out of bounds. The referenced grid "grid0" is of rank 2.}}
+  %0 = shard.all_reduce %arg0 on @grid0 grid_axes = [2] reduction = sum
     : tensor<4xf32> -> tensor<4xf64>
   return %0 : tensor<4xf64>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 2x4)
+shard.grid @grid0(shape = 2x4)
 
-func.func @all_reduce_duplicate_mesh_axis(
+func.func @all_reduce_duplicate_grid_axis(
     %arg0 : tensor<4xf32>) -> tensor<4xf64> {
-  // expected-error@+1 {{Mesh axes contains duplicate elements.}}
-  %0 = mesh.all_reduce %arg0 on @mesh0 mesh_axes = [0, 1, 0] reduction = sum
+  // expected-error@+1 {{Grid axes contains duplicate elements.}}
+  %0 = shard.all_reduce %arg0 on @grid0 grid_axes = [0, 1, 0] reduction = sum
     : tensor<4xf32> -> tensor<4xf64>
   return %0 : tensor<4xf64>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 2x4)
+shard.grid @grid0(shape = 2x4)
 
 func.func @all_reduce_invalid_tensor_dimension_size(
     %arg0 : tensor<4xf32>) -> tensor<5xf64> {
-  // expected-error@+1 {{'mesh.all_reduce' op requires the same shape for all operands and results}}
-  %0 = mesh.all_reduce %arg0 on @mesh0 : tensor<4xf32> -> tensor<5xf64>
+  // expected-error@+1 {{'shard.all_reduce' op requires the same shape for all operands and results}}
+  %0 = shard.all_reduce %arg0 on @grid0 : tensor<4xf32> -> tensor<5xf64>
   return %0 : tensor<5xf64>
 }
 
 // -----
 
-func.func @all_gather_invalid_mesh_symbol(
+func.func @all_gather_invalid_grid_symbol(
     %arg0 : tensor<4xf32>) -> tensor<4xf32> {
-  // expected-error@+1 {{Undefined required mesh symbol "this_mesh_symbol_does_not_exist".}}
-  %0 = mesh.all_gather %arg0 on @this_mesh_symbol_does_not_exist gather_axis = 0
+  // expected-error@+1 {{Undefined required grid symbol "this_grid_symbol_does_not_exist".}}
+  %0 = shard.all_gather %arg0 on @this_grid_symbol_does_not_exist gather_axis = 0
     : tensor<4xf32> -> tensor<4xf32>
   return %0 : tensor<4xf32>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 2x4)
+shard.grid @grid0(shape = 2x4)
 
-func.func @all_gather_invalid_mesh_axis(
+func.func @all_gather_invalid_grid_axis(
     %arg0 : tensor<4xf32>) -> tensor<4xf32> {
-  // expected-error@+1 {{0-based mesh axis index 2 is out of bounds. The referenced mesh "mesh0" is of rank 2.}}
-  %0 = mesh.all_gather %arg0 on @mesh0 mesh_axes = [2] gather_axis = 0
+  // expected-error@+1 {{0-based grid axis index 2 is out of bounds. The referenced grid "grid0" is of rank 2.}}
+  %0 = shard.all_gather %arg0 on @grid0 grid_axes = [2] gather_axis = 0
     : tensor<4xf32> -> tensor<4xf32>
   return %0 : tensor<4xf32>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 2x4)
+shard.grid @grid0(shape = 2x4)
 
-func.func @all_reduce_duplicate_mesh_axis(
+func.func @all_reduce_duplicate_grid_axis(
     %arg0 : tensor<4xf32>) -> tensor<4xf32> {
-  // expected-error@+1 {{Mesh axes contains duplicate elements.}}
-  %0 = mesh.all_gather %arg0 on @mesh0 mesh_axes = [2, 2] gather_axis = 0
+  // expected-error@+1 {{Grid axes contains duplicate elements.}}
+  %0 = shard.all_gather %arg0 on @grid0 grid_axes = [2, 2] gather_axis = 0
     : tensor<4xf32> -> tensor<4xf32>
   return %0 : tensor<4xf32>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 1)
+shard.grid @grid0(shape = 1)
 
 func.func @all_gather_invalid_non_gather_axis_dimension_size(
     %arg0 : tensor<3x4xf32>) -> tensor<3x5xf32> {
   // expected-error@+1 {{Dimension size mismatch for result axis 1. Expected 4, but got 5.}}
-  %0 = mesh.all_gather %arg0 on @mesh0 mesh_axes = [0] gather_axis = 0
+  %0 = shard.all_gather %arg0 on @grid0 grid_axes = [0] gather_axis = 0
     : tensor<3x4xf32> -> tensor<3x5xf32>
   return %0 : tensor<3x5xf32>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 1x2)
+shard.grid @grid0(shape = 1x2)
 
 func.func @all_gather_invalid_gather_axis_dimension_size(
     %arg0 : tensor<3x4xf32>) -> tensor<3x5xf32> {
   // expected-error@+1 {{Dimension size mismatch for result axis 1. Expected 8, but got 5.}}
-  %0 = mesh.all_gather %arg0 on @mesh0 mesh_axes = [1] gather_axis = 1
+  %0 = shard.all_gather %arg0 on @grid0 grid_axes = [1] gather_axis = 1
     : tensor<3x4xf32> -> tensor<3x5xf32>
   return %0 : tensor<3x5xf32>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 1)
+shard.grid @grid0(shape = 1)
 
 func.func @all_gather_invalid_gather_axis_dynamic_dimension(
     %arg0 : tensor<?xf32>) -> tensor<3xf32> {
   // expected-error@+1 {{Dimension size mismatch for result axis 0. Expected dynamic, but got 3.}}
-  %0 = mesh.all_gather %arg0 on @mesh0 gather_axis = 0
+  %0 = shard.all_gather %arg0 on @grid0 gather_axis = 0
     : tensor<?xf32> -> tensor<3xf32>
   return %0 : tensor<3xf32>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 1)
+shard.grid @grid0(shape = 1)
 
 func.func @all_gather_invalid_gather_axis(
     %arg0 : tensor<3xf32>) -> tensor<3xf32> {
   // expected-error@+1 {{Gather axis 1 is out of bounds [0, 1).}}
-  %0 = mesh.all_gather %arg0 on @mesh0 mesh_axes = [0] gather_axis = 1
+  %0 = shard.all_gather %arg0 on @grid0 grid_axes = [0] gather_axis = 1
     : tensor<3xf32> -> tensor<3xf32>
   return %0 : tensor<3xf32>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 1)
+shard.grid @grid0(shape = 1)
 
 func.func @all_gather_invalid_negative_gather_axis(
     %arg0 : tensor<3xf32>) -> tensor<3xf32> {
   // expected-error@+1 {{Gather axis -1 is out of bounds [0, 1).}}
-  %0 = mesh.all_gather %arg0 on @mesh0 mesh_axes = [0] gather_axis = -1
+  %0 = shard.all_gather %arg0 on @grid0 grid_axes = [0] gather_axis = -1
     : tensor<3xf32> -> tensor<3xf32>
   return %0 : tensor<3xf32>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 3)
+shard.grid @grid0(shape = 3)
 
-func.func @all_slice_duplicate_mesh_axis(
+func.func @all_slice_duplicate_grid_axis(
     %arg0 : tensor<?xf32>) -> tensor<?xf32> {
-  // expected-error@+1 {{Mesh axes contains duplicate elements.}}
-  %0 = mesh.all_slice %arg0 on @mesh0 mesh_axes = [0, 0]
+  // expected-error@+1 {{Grid axes contains duplicate elements.}}
+  %0 = shard.all_slice %arg0 on @grid0 grid_axes = [0, 0]
     slice_axis = 0
     : tensor<?xf32> -> tensor<?xf32>
   return %0 : tensor<?xf32>
@@ -359,12 +359,12 @@ func.func @all_slice_duplicate_mesh_axis(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3)
+shard.grid @grid0(shape = 3)
 
 func.func @all_slice_invalid_dynamic_dimension(
     %arg0 : tensor<?xf32>) -> tensor<2xf32> {
   // expected-error@+1 {{Dimension size mismatch for result axis 0. Expected dynamic, but got 2.}}
-  %0 = mesh.all_slice %arg0 on @mesh0
+  %0 = shard.all_slice %arg0 on @grid0
     slice_axis = 0
     : tensor<?xf32> -> tensor<2xf32>
   return %0 : tensor<2xf32>
@@ -372,12 +372,12 @@ func.func @all_slice_invalid_dynamic_dimension(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3)
+shard.grid @grid0(shape = 3)
 
 func.func @all_slice_invalid_static_dimension_size(
     %arg0 : tensor<3xf32>) -> tensor<2xf32> {
   // expected-error@+1 {{Dimension size mismatch for result axis 0. Expected 1, but got 2.}}
-  %0 = mesh.all_slice %arg0 on @mesh0 mesh_axes = [0]
+  %0 = shard.all_slice %arg0 on @grid0 grid_axes = [0]
     slice_axis = 0
     : tensor<3xf32> -> tensor<2xf32>
   return %0 : tensor<2xf32>
@@ -385,12 +385,12 @@ func.func @all_slice_invalid_static_dimension_size(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3)
+shard.grid @grid0(shape = 3)
 
 func.func @all_slice_invalid_operand_static_dimension_size(
     %arg0 : tensor<4xf32>) -> tensor<?xf32> {
   // expected-error@+1 {{Operand dimension size 4 is not divisible by collective device group size 3 for tensor axis 0.}}
-  %0 = mesh.all_slice %arg0 on @mesh0 mesh_axes = [0]
+  %0 = shard.all_slice %arg0 on @grid0 grid_axes = [0]
     slice_axis = 0
     : tensor<4xf32> -> tensor<?xf32>
   return %0 : tensor<?xf32>
@@ -398,10 +398,10 @@ func.func @all_slice_invalid_operand_static_dimension_size(
 
 // -----
 
-func.func @all_to_all_invalid_mesh_symbol(
+func.func @all_to_all_invalid_grid_symbol(
     %arg0 : tensor<3x6xi8>) -> tensor<3x6xi8> {
-  // expected-error@+1 {{Undefined required mesh symbol "this_mesh_symbol_does_not_exist".}}
-  %0 = mesh.all_to_all %arg0 on @this_mesh_symbol_does_not_exist
+  // expected-error@+1 {{Undefined required grid symbol "this_grid_symbol_does_not_exist".}}
+  %0 = shard.all_to_all %arg0 on @this_grid_symbol_does_not_exist
     split_axis = 1 concat_axis = 0
     : tensor<3x6xi8> -> tensor<3x6xi8>
   return %0 : tensor<3x6xi8>
@@ -409,12 +409,12 @@ func.func @all_to_all_invalid_mesh_symbol(
 
 // -----
 
-mesh.mesh @mesh0(shape = 1)
+shard.grid @grid0(shape = 1)
 
-func.func @all_to_all_duplicate_mesh_axis(
+func.func @all_to_all_duplicate_grid_axis(
     %arg0 : tensor<3x6xi8>) -> tensor<3x6xi8> {
-  // expected-error@+1 {{Mesh axes contains duplicate elements.}}
-  %0 = mesh.all_to_all %arg0 on @mesh0 mesh_axes = [0, 0]
+  // expected-error@+1 {{Grid axes contains duplicate elements.}}
+  %0 = shard.all_to_all %arg0 on @grid0 grid_axes = [0, 0]
     split_axis = 0 concat_axis = 0
     : tensor<3x6xi8> -> tensor<3x6xi8>
   return %0 : tensor<3x6xi8>
@@ -422,12 +422,12 @@ func.func @all_to_all_duplicate_mesh_axis(
 
 // -----
 
-mesh.mesh @mesh0(shape = ?x1)
+shard.grid @grid0(shape = ?x1)
 
 func.func @all_to_all_invalid_non_dynamic_result_dimension_induced_by_dynamic_device_group(
     %arg0 : tensor<3x6xi8>) -> tensor<3x6xi8> {
   // expected-error@+1 {{Dimension size mismatch for result axis 1. Expected dynamic, but got 6.}}
-  %0 = mesh.all_to_all %arg0 on @mesh0 mesh_axes = [0]
+  %0 = shard.all_to_all %arg0 on @grid0 grid_axes = [0]
     split_axis = 0 concat_axis = 1
     : tensor<3x6xi8> -> tensor<3x6xi8>
   return %0 : tensor<3x6xi8>
@@ -435,12 +435,12 @@ func.func @all_to_all_invalid_non_dynamic_result_dimension_induced_by_dynamic_de
 
 // -----
 
-mesh.mesh @mesh0(shape = 1x1)
+shard.grid @grid0(shape = 1x1)
 
 func.func @all_to_all_invalid_non_dynamic_result_split_dimension_induced_by_dynamic_operand_dimension(
     %arg0 : tensor<?x6xi8>) -> tensor<3x?xi8> {
   // expected-error@+1 {{Dimension size mismatch for result axis 0. Expected dynamic, but got 3.}}
-  %0 = mesh.all_to_all %arg0 on @mesh0 mesh_axes = [1]
+  %0 = shard.all_to_all %arg0 on @grid0 grid_axes = [1]
     split_axis = 0 concat_axis = 1
     : tensor<?x6xi8> -> tensor<3x?xi8>
   return %0 : tensor<3x?xi8>
@@ -448,12 +448,12 @@ func.func @all_to_all_invalid_non_dynamic_result_split_dimension_induced_by_dyna
 
 // -----
 
-mesh.mesh @mesh0(shape = 1x1)
+shard.grid @grid0(shape = 1x1)
 
 func.func @all_to_all_invalid_non_dynamic_result_concat_dimension_induced_by_dynamic_operand_dimension(
     %arg0 : tensor<3x?xi8>) -> tensor<?x3xi8> {
   // expected-error@+1 {{Dimension size mismatch for result axis 1. Expected dynamic, but got 3.}}
-  %0 = mesh.all_to_all %arg0 on @mesh0 mesh_axes = [1]
+  %0 = shard.all_to_all %arg0 on @grid0 grid_axes = [1]
     split_axis = 0 concat_axis = 1
     : tensor<3x?xi8> -> tensor<?x3xi8>
   return %0 : tensor<?x3xi8>
@@ -461,12 +461,12 @@ func.func @all_to_all_invalid_non_dynamic_result_concat_dimension_induced_by_dyn
 
 // -----
 
-mesh.mesh @mesh0(shape = 3)
+shard.grid @grid0(shape = 3)
 
 func.func @all_to_all_invalid_non_dynamic_result_concat_dimension_size(
     %arg0 : tensor<3x2xi8>) -> tensor<1x7xi8> {
   // expected-error@+1 {{Dimension size mismatch for result axis 1. Expected 6, but got 7.}}
-  %0 = mesh.all_to_all %arg0  on @mesh0 mesh_axes = [0]
+  %0 = shard.all_to_all %arg0  on @grid0 grid_axes = [0]
     split_axis = 0 concat_axis = 1
     : tensor<3x2xi8> -> tensor<1x7xi8>
   return %0 : tensor<1x7xi8>
@@ -474,12 +474,12 @@ func.func @all_to_all_invalid_non_dynamic_result_concat_dimension_size(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3)
+shard.grid @grid0(shape = 3)
 
 func.func @all_to_all_invalid_non_dynamic_result_split_dimension_size(
     %arg0 : tensor<3x2xi8>) -> tensor<2x6xi8> {
   // expected-error@+1 {{Dimension size mismatch for result axis 0. Expected 1, but got 2.}}
-  %0 = mesh.all_to_all %arg0 on @mesh0 mesh_axes = [0]
+  %0 = shard.all_to_all %arg0 on @grid0 grid_axes = [0]
     split_axis = 0 concat_axis = 1
     : tensor<3x2xi8> -> tensor<2x6xi8>
   return %0 : tensor<2x6xi8>
@@ -487,12 +487,12 @@ func.func @all_to_all_invalid_non_dynamic_result_split_dimension_size(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3x?)
+shard.grid @grid0(shape = 3x?)
 
 func.func @broadcast_root_dimension_out_of_bounds(
     %arg0 : tensor<2xi8>) -> tensor<2xi8> {
   // expected-error@+1 {{Out of bounds coordinate 0 for in-group device "root". Got 3, but expected value in the range [0, 2].}}
-  %0 = mesh.broadcast %arg0 on @mesh0 mesh_axes = [0]
+  %0 = shard.broadcast %arg0 on @grid0 grid_axes = [0]
     root = [3]
     : (tensor<2xi8>) -> tensor<2xi8>
   return %0 : tensor<2xi8>
@@ -500,12 +500,12 @@ func.func @broadcast_root_dimension_out_of_bounds(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3x?)
+shard.grid @grid0(shape = 3x?)
 
 func.func @broadcast_root_wrong_number_dimensions(
     %arg0 : tensor<2xi8>) -> tensor<2xi8> {
   // expected-error@+1 {{In-group device "root" has unexpected multi-index size 2. Expected 1.}}
-  %0 = mesh.broadcast %arg0 on @mesh0 mesh_axes = [0]
+  %0 = shard.broadcast %arg0 on @grid0 grid_axes = [0]
     root = [2, 2]
     : (tensor<2xi8>) -> tensor<2xi8>
   return %0 : tensor<2xi8>
@@ -513,12 +513,12 @@ func.func @broadcast_root_wrong_number_dimensions(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3x?)
+shard.grid @grid0(shape = 3x?)
 
 func.func @broadcast_different_input_and_result_type(
     %arg0 : tensor<2xi8>) -> tensor<2xi16> {
-  // expected-error@+1 {{'mesh.broadcast' op failed to verify that all of {input, result} have same element type}}
-  %0 = mesh.broadcast %arg0 on @mesh0 mesh_axes = [0]
+  // expected-error@+1 {{'shard.broadcast' op failed to verify that all of {input, result} have same element type}}
+  %0 = shard.broadcast %arg0 on @grid0 grid_axes = [0]
     root = [2]
     : (tensor<2xi8>) -> tensor<2xi16>
   return %0 : tensor<2xi16>
@@ -526,84 +526,84 @@ func.func @broadcast_different_input_and_result_type(
 
 // -----
 
-mesh.mesh @mesh0(shape = 1)
+shard.grid @grid0(shape = 1)
 
 func.func @gather_wrong_return_element_type(
     %arg0 : tensor<1xf32>) -> tensor<1xi8> {
-  // expected-error@+1 {{'mesh.gather' op failed to verify that all of {input, result} have same element type}}
-  %0 = mesh.gather %arg0 on @mesh0 mesh_axes = [0] gather_axis = 0 root = [0]
+  // expected-error@+1 {{'shard.gather' op failed to verify that all of {input, result} have same element type}}
+  %0 = shard.gather %arg0 on @grid0 grid_axes = [0] gather_axis = 0 root = [0]
     : (tensor<1xf32>) -> tensor<1xi8>
   return %0 : tensor<1xi8>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 1)
+shard.grid @grid0(shape = 1)
 
 func.func @gather_invalid_non_gather_axis_dimension_size(
     %arg0 : tensor<3x4xf32>) -> tensor<3x5xf32> {
   // expected-error@+1 {{Dimension size mismatch for result axis 1. Expected 4, but got 5.}}
-  %0 = mesh.gather %arg0 on @mesh0 mesh_axes = [0] gather_axis = 0 root = [0]
+  %0 = shard.gather %arg0 on @grid0 grid_axes = [0] gather_axis = 0 root = [0]
     : (tensor<3x4xf32>) -> tensor<3x5xf32>
   return %0 : tensor<3x5xf32>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 1x2)
+shard.grid @grid0(shape = 1x2)
 
 func.func @gather_invalid_gather_axis_dimension_size(
     %arg0 : tensor<3x4xf32>) -> tensor<3x5xf32> {
   // expected-error@+1 {{Dimension size mismatch for result axis 1. Expected 8, but got 5.}}
-  %0 = mesh.gather %arg0 on @mesh0 mesh_axes = [1] gather_axis = 1 root = [0]
+  %0 = shard.gather %arg0 on @grid0 grid_axes = [1] gather_axis = 1 root = [0]
     : (tensor<3x4xf32>) -> tensor<3x5xf32>
   return %0 : tensor<3x5xf32>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 1)
+shard.grid @grid0(shape = 1)
 
 func.func @gather_invalid_gather_axis_dynamic_dimension(
     %arg0 : tensor<?xf32>) -> tensor<3xf32> {
   // expected-error@+1 {{Dimension size mismatch for result axis 0. Expected dynamic, but got 3.}}
-  %0 = mesh.gather %arg0 on @mesh0 gather_axis = 0 root = []
+  %0 = shard.gather %arg0 on @grid0 gather_axis = 0 root = []
     : (tensor<?xf32>) -> tensor<3xf32>
   return %0 : tensor<3xf32>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 1)
+shard.grid @grid0(shape = 1)
 
 func.func @gather_invalid_gather_axis(
     %arg0 : tensor<3xf32>) -> tensor<3xf32> {
   // expected-error@+1 {{Gather axis 1 is out of bounds [0, 1).}}
-  %0 = mesh.gather %arg0 on @mesh0 mesh_axes = [0] gather_axis = 1 root = [0]
+  %0 = shard.gather %arg0 on @grid0 grid_axes = [0] gather_axis = 1 root = [0]
     : (tensor<3xf32>) -> tensor<3xf32>
   return %0 : tensor<3xf32>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 1)
+shard.grid @grid0(shape = 1)
 
 func.func @gather_invalid_negative_gather_axis(
     %arg0 : tensor<3xf32>) -> tensor<3xf32> {
   // expected-error@+1 {{Gather axis -1 is out of bounds [0, 1).}}
-  %0 = mesh.gather %arg0 on @mesh0 mesh_axes = [0] gather_axis = -1 root = [0]
+  %0 = shard.gather %arg0 on @grid0 grid_axes = [0] gather_axis = -1 root = [0]
     : (tensor<3xf32>) -> tensor<3xf32>
   return %0 : tensor<3xf32>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 3x?)
+shard.grid @grid0(shape = 3x?)
 
 func.func @gather_root_dimension_out_of_bounds(
     %arg0 : tensor<2xi8>) -> tensor<6xi8> {
   // expected-error@+1 {{Out of bounds coordinate 0 for in-group device "root". Got 3, but expected value in the range [0, 2].}}
-  %0 = mesh.gather %arg0 on @mesh0 mesh_axes = [0] gather_axis = 0
+  %0 = shard.gather %arg0 on @grid0 grid_axes = [0] gather_axis = 0
     root = [3]
     : (tensor<2xi8>) -> tensor<6xi8>
   return %0 : tensor<6xi8>
@@ -611,12 +611,12 @@ func.func @gather_root_dimension_out_of_bounds(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3x?)
+shard.grid @grid0(shape = 3x?)
 
 func.func @gather_root_wrong_number_dimensions(
     %arg0 : tensor<2xi8>) -> tensor<2xi8> {
   // expected-error@+1 {{In-group device "root" has unexpected multi-index size 2. Expected 1.}}
-  %0 = mesh.gather %arg0 on @mesh0 mesh_axes = [0] gather_axis = 0
+  %0 = shard.gather %arg0 on @grid0 grid_axes = [0] gather_axis = 0
     root = [2, 2]
     : (tensor<2xi8>) -> tensor<2xi8>
   return %0 : tensor<2xi8>
@@ -624,12 +624,12 @@ func.func @gather_root_wrong_number_dimensions(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3x?)
+shard.grid @grid0(shape = 3x?)
 
 func.func @receive_source_dimension_out_of_bounds(
     %arg0 : tensor<2xi8>) -> tensor<2xi8> {
   // expected-error@+1 {{Out of bounds coordinate 0 for in-group device "source". Got 3, but expected value in the range [0, 2].}}
-  %0 = mesh.recv %arg0 on @mesh0 mesh_axes = [0]
+  %0 = shard.recv %arg0 on @grid0 grid_axes = [0]
     source = [3]
     : (tensor<2xi8>) -> tensor<2xi8>
   return %0 : tensor<2xi8>
@@ -637,12 +637,12 @@ func.func @receive_source_dimension_out_of_bounds(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3x?)
+shard.grid @grid0(shape = 3x?)
 
 func.func @receive_source_wrong_number_dimensions(
     %arg0 : tensor<2xi8>) -> tensor<2xi8> {
   // expected-error@+1 {{In-group device "source" has unexpected multi-index size 2. Expected 1.}}
-  %0 = mesh.recv %arg0 on @mesh0 mesh_axes = [0]
+  %0 = shard.recv %arg0 on @grid0 grid_axes = [0]
     source = [2, 2]
     : (tensor<2xi8>) -> tensor<2xi8>
   return %0 : tensor<2xi8>
@@ -650,12 +650,12 @@ func.func @receive_source_wrong_number_dimensions(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3x?)
+shard.grid @grid0(shape = 3x?)
 
 func.func @receive_different_input_and_result_type(
     %arg0 : tensor<2xi8>) -> tensor<2xi16> {
-  // expected-error@+1 {{'mesh.recv' op failed to verify that all of {input, result} have same element type}}
-  %0 = mesh.recv %arg0 on @mesh0 mesh_axes = [0]
+  // expected-error@+1 {{'shard.recv' op failed to verify that all of {input, result} have same element type}}
+  %0 = shard.recv %arg0 on @grid0 grid_axes = [0]
     source = [2]
     : (tensor<2xi8>) -> tensor<2xi16>
   return %0 : tensor<2xi16>
@@ -663,12 +663,12 @@ func.func @receive_different_input_and_result_type(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3x?)
+shard.grid @grid0(shape = 3x?)
 
 func.func @reduce_root_dimension_out_of_bounds(
     %arg0 : tensor<2xi8>) -> tensor<2xi8> {
   // expected-error@+1 {{Out of bounds coordinate 0 for in-group device "root". Got 3, but expected value in the range [0, 2].}}
-  %0 = mesh.reduce %arg0 on @mesh0 mesh_axes = [0]
+  %0 = shard.reduce %arg0 on @grid0 grid_axes = [0]
     root = [3]
     : (tensor<2xi8>) -> tensor<2xi8>
   return %0 : tensor<2xi8>
@@ -676,12 +676,12 @@ func.func @reduce_root_dimension_out_of_bounds(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3x?)
+shard.grid @grid0(shape = 3x?)
 
 func.func @reduce_root_wrong_number_dimensions(
     %arg0 : tensor<2xi8>) -> tensor<2xi8> {
   // expected-error@+1 {{In-group device "root" has unexpected multi-index size 2. Expected 1.}}
-  %0 = mesh.reduce %arg0 on @mesh0 mesh_axes = [0]
+  %0 = shard.reduce %arg0 on @grid0 grid_axes = [0]
     root = [2, 2]
     : (tensor<2xi8>) -> tensor<2xi8>
   return %0 : tensor<2xi8>
@@ -689,12 +689,12 @@ func.func @reduce_root_wrong_number_dimensions(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3x?)
+shard.grid @grid0(shape = 3x?)
 
 func.func @reduce_different_input_and_result_shape(
     %arg0 : tensor<2xi8>) -> tensor<3xi16> {
-  // expected-error@+1 {{'mesh.reduce' op failed to verify that all of {input, result} have same shape}}
-  %0 = mesh.reduce %arg0 on @mesh0 mesh_axes = [0]
+  // expected-error@+1 {{'shard.reduce' op failed to verify that all of {input, result} have same shape}}
+  %0 = shard.reduce %arg0 on @grid0 grid_axes = [0]
     root = [2]
     : (tensor<2xi8>) -> tensor<3xi16>
   return %0 : tensor<3xi16>
@@ -702,60 +702,60 @@ func.func @reduce_different_input_and_result_shape(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3)
+shard.grid @grid0(shape = 3)
 
-func.func @reduce_scatter_duplicate_mesh_axis(
+func.func @reduce_scatter_duplicate_grid_axis(
     %arg0 : tensor<?xf32>) -> tensor<?xf64> {
-  // expected-error@+1 {{Mesh axes contains duplicate elements.}}
-  %0 = mesh.reduce_scatter %arg0 on @mesh0 mesh_axes = [0, 0] scatter_axis = 0
+  // expected-error@+1 {{Grid axes contains duplicate elements.}}
+  %0 = shard.reduce_scatter %arg0 on @grid0 grid_axes = [0, 0] scatter_axis = 0
     : tensor<?xf32> -> tensor<?xf64>
   return %0 : tensor<?xf64>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 3)
+shard.grid @grid0(shape = 3)
 
 func.func @reduce_scatter_invalid_dynamic_dimension(
     %arg0 : tensor<?xf32>) -> tensor<2xf64> {
   // expected-error@+1 {{Dimension size mismatch for result axis 0. Expected dynamic, but got 2.}}
-  %0 = mesh.reduce_scatter %arg0 on @mesh0 scatter_axis = 0
+  %0 = shard.reduce_scatter %arg0 on @grid0 scatter_axis = 0
     : tensor<?xf32> -> tensor<2xf64>
   return %0 : tensor<2xf64>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 3)
+shard.grid @grid0(shape = 3)
 
 func.func @reduce_scatter_invalid_static_dimension_size(
     %arg0 : tensor<3xf32>) -> tensor<2xf64> {
   // expected-error@+1 {{Dimension size mismatch for result axis 0. Expected 1, but got 2.}}
-  %0 = mesh.reduce_scatter %arg0 on @mesh0 mesh_axes = [0] scatter_axis = 0
+  %0 = shard.reduce_scatter %arg0 on @grid0 grid_axes = [0] scatter_axis = 0
     : tensor<3xf32> -> tensor<2xf64>
   return %0 : tensor<2xf64>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 3)
+shard.grid @grid0(shape = 3)
 
 func.func @reduce_scatter_invalid_operand_static_dimension_size(
     %arg0 : tensor<4xf32>) -> tensor<?xf64> {
   // expected-error@+1 {{Operand dimension size 4 is not divisible by collective device group size 3 for tensor axis 0.}}
-  %0 = mesh.reduce_scatter %arg0 on @mesh0 mesh_axes = [0] scatter_axis = 0
+  %0 = shard.reduce_scatter %arg0 on @grid0 grid_axes = [0] scatter_axis = 0
     : tensor<4xf32> -> tensor<?xf64>
   return %0 : tensor<?xf64>
 }
 
 // -----
 
-mesh.mesh @mesh0(shape = 3)
+shard.grid @grid0(shape = 3)
 
-func.func @scatter_duplicate_mesh_axis(
+func.func @scatter_duplicate_grid_axis(
     %arg0 : tensor<?xf32>) -> tensor<?xf32> {
-  // expected-error@+1 {{Mesh axes contains duplicate elements.}}
-  %0 = mesh.scatter %arg0 on @mesh0 mesh_axes = [0, 0]
+  // expected-error@+1 {{Grid axes contains duplicate elements.}}
+  %0 = shard.scatter %arg0 on @grid0 grid_axes = [0, 0]
     scatter_axis = 0 root = [0, 0]
     : (tensor<?xf32>) -> tensor<?xf32>
   return %0 : tensor<?xf32>
@@ -763,12 +763,12 @@ func.func @scatter_duplicate_mesh_axis(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3)
+shard.grid @grid0(shape = 3)
 
 func.func @scatter_invalid_dynamic_dimension(
     %arg0 : tensor<?xf32>) -> tensor<2xf32> {
   // expected-error@+1 {{Dimension size mismatch for result axis 0. Expected dynamic, but got 2.}}
-  %0 = mesh.scatter %arg0 on @mesh0
+  %0 = shard.scatter %arg0 on @grid0
     scatter_axis = 0 root = []
     : (tensor<?xf32>) -> tensor<2xf32>
   return %0 : tensor<2xf32>
@@ -776,12 +776,12 @@ func.func @scatter_invalid_dynamic_dimension(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3)
+shard.grid @grid0(shape = 3)
 
 func.func @scatter_invalid_static_dimension_size(
     %arg0 : tensor<3xf32>) -> tensor<2xf32> {
   // expected-error@+1 {{Dimension size mismatch for result axis 0. Expected 1, but got 2.}}
-  %0 = mesh.scatter %arg0 on @mesh0 mesh_axes = [0]
+  %0 = shard.scatter %arg0 on @grid0 grid_axes = [0]
     scatter_axis = 0 root = [1]
     : (tensor<3xf32>) -> tensor<2xf32>
   return %0 : tensor<2xf32>
@@ -789,12 +789,12 @@ func.func @scatter_invalid_static_dimension_size(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3)
+shard.grid @grid0(shape = 3)
 
 func.func @scatter_invalid_operand_static_dimension_size(
     %arg0 : tensor<4xf32>) -> tensor<?xf32> {
   // expected-error@+1 {{Operand dimension size 4 is not divisible by collective device group size 3 for tensor axis 0.}}
-  %0 = mesh.scatter %arg0 on @mesh0 mesh_axes = [0]
+  %0 = shard.scatter %arg0 on @grid0 grid_axes = [0]
     scatter_axis = 0 root = [1]
     : (tensor<4xf32>) -> tensor<?xf32>
   return %0 : tensor<?xf32>
@@ -802,12 +802,12 @@ func.func @scatter_invalid_operand_static_dimension_size(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3x?)
+shard.grid @grid0(shape = 3x?)
 
 func.func @scatter_root_dimension_out_of_bounds(
     %arg0 : tensor<3xi8>) -> tensor<1xi8> {
   // expected-error@+1 {{Out of bounds coordinate 0 for in-group device "root". Got 3, but expected value in the range [0, 2].}}
-  %0 = mesh.scatter %arg0 on @mesh0 mesh_axes = [0]
+  %0 = shard.scatter %arg0 on @grid0 grid_axes = [0]
     scatter_axis = 0 root = [3]
     : (tensor<3xi8>) -> tensor<1xi8>
   return %0 : tensor<1xi8>
@@ -815,12 +815,12 @@ func.func @scatter_root_dimension_out_of_bounds(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3x?)
+shard.grid @grid0(shape = 3x?)
 
 func.func @scatter_root_wrong_number_dimensions(
     %arg0 : tensor<3xi8>) -> tensor<1xi8> {
   // expected-error@+1 {{In-group device "root" has unexpected multi-index size 2. Expected 1.}}
-  %0 = mesh.scatter %arg0 on @mesh0 mesh_axes = [0]
+  %0 = shard.scatter %arg0 on @grid0 grid_axes = [0]
     scatter_axis = 0 root = [2, 2]
     : (tensor<3xi8>) -> tensor<1xi8>
   return %0 : tensor<1xi8>
@@ -828,12 +828,12 @@ func.func @scatter_root_wrong_number_dimensions(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3x?)
+shard.grid @grid0(shape = 3x?)
 
 func.func @send_destination_dimension_out_of_bounds(
     %arg0 : tensor<2xi8>) -> tensor<2xi8> {
   // expected-error@+1 {{Out of bounds coordinate 0 for in-group device "destination". Got 3, but expected value in the range [0, 2].}}
-  %0 = mesh.send %arg0 on @mesh0 mesh_axes = [0]
+  %0 = shard.send %arg0 on @grid0 grid_axes = [0]
     destination = [3]
     : (tensor<2xi8>) -> tensor<2xi8>
   return %0 : tensor<2xi8>
@@ -841,12 +841,12 @@ func.func @send_destination_dimension_out_of_bounds(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3x?)
+shard.grid @grid0(shape = 3x?)
 
 func.func @send_destination_wrong_number_dimensions(
     %arg0 : tensor<2xi8>) -> tensor<2xi8> {
   // expected-error@+1 {{In-group device "destination" has unexpected multi-index size 2. Expected 1.}}
-  %0 = mesh.send %arg0 on @mesh0 mesh_axes = [0]
+  %0 = shard.send %arg0 on @grid0 grid_axes = [0]
     destination = [2, 2]
     : (tensor<2xi8>) -> tensor<2xi8>
   return %0 : tensor<2xi8>
@@ -854,12 +854,12 @@ func.func @send_destination_wrong_number_dimensions(
 
 // -----
 
-mesh.mesh @mesh0(shape = 3x?)
+shard.grid @grid0(shape = 3x?)
 
 func.func @send_different_input_and_result_type(
     %arg0 : tensor<2xi8>) -> tensor<2xi16> {
-  // expected-error@+1 {{'mesh.send' op failed to verify that all of {input, result} have same element type}}
-  %0 = mesh.send %arg0 on @mesh0 mesh_axes = [0]
+  // expected-error@+1 {{'shard.send' op failed to verify that all of {input, result} have same element type}}
+  %0 = shard.send %arg0 on @grid0 grid_axes = [0]
     destination = [2]
     : (tensor<2xi8>) -> tensor<2xi16>
   return %0 : tensor<2xi16>
@@ -867,10 +867,10 @@ func.func @send_different_input_and_result_type(
 
 // -----
 
-func.func @shift_invalid_mesh_symbol(
+func.func @shift_invalid_grid_symbol(
     %arg0 : tensor<4xi8>) -> tensor<4xi8> {
-  // expected-error@+1 {{Undefined required mesh symbol "this_mesh_symbol_does_not_exist".}}
-  %0 = mesh.shift %arg0 on @this_mesh_symbol_does_not_exist
+  // expected-error@+1 {{Undefined required grid symbol "this_grid_symbol_does_not_exist".}}
+  %0 = shard.shift %arg0 on @this_grid_symbol_does_not_exist
     shift_axis = 0 offset = -2
     : tensor<4xi8> -> tensor<4xi8>
   return %0 : tensor<4xi8>
@@ -878,12 +878,12 @@ func.func @shift_invalid_mesh_symbol(
 
 // -----
 
-mesh.mesh @mesh0(shape = 2x4)
+shard.grid @grid0(shape = 2x4)
 
-func.func @shift_invalid_mesh_axis(
+func.func @shift_invalid_grid_axis(
     %arg0 : tensor<4xi8>) -> tensor<4xi8> {
-  // expected-error@+1 {{0-based mesh axis index 2 is out of bounds. The referenced mesh "mesh0" is of rank 2.}}
-  %0 = mesh.shift %arg0 on @mesh0 mesh_axes = [2]
+  // expected-error@+1 {{0-based grid axis index 2 is out of bounds. The referenced grid "grid0" is of rank 2.}}
+  %0 = shard.shift %arg0 on @grid0 grid_axes = [2]
         shift_axis = 2 offset = -2
     : tensor<4xi8> -> tensor<4xi8>
   return %0 : tensor<4xi8>
@@ -891,12 +891,12 @@ func.func @shift_invalid_mesh_axis(
 
 // -----
 
-mesh.mesh @mesh0(shape = 2x4)
+shard.grid @grid0(shape = 2x4)
 
-func.func @shift_duplicate_mesh_axis(
+func.func @shift_duplicate_grid_axis(
     %arg0 : tensor<4xi8>) -> tensor<4xi8> {
-  // expected-error@+1 {{Mesh axes contains duplicate elements.}}
-  %0 = mesh.shift %arg0 on @mesh0 mesh_axes = [0, 1, 0]
+  // expected-error@+1 {{Grid axes contains duplicate elements.}}
+  %0 = shard.shift %arg0 on @grid0 grid_axes = [0, 1, 0]
     shift_axis = 0 offset = -2
     : tensor<4xi8> -> tensor<4xi8>
   return %0 : tensor<4xi8>
@@ -904,12 +904,12 @@ func.func @shift_duplicate_mesh_axis(
 
 // -----
 
-mesh.mesh @mesh0(shape = 2x4)
+shard.grid @grid0(shape = 2x4)
 
 func.func @shift_invalid_tensor_dimension_size(
     %arg0 : tensor<4xi8>) -> tensor<5xi8> {
-  // expected-error@+1 {{'mesh.shift' op requires the same shape for all operands and results}}
-  %0 = mesh.shift %arg0 on @mesh0 mesh_axes = [0]
+  // expected-error@+1 {{'shard.shift' op requires the same shape for all operands and results}}
+  %0 = shard.shift %arg0 on @grid0 grid_axes = [0]
     shift_axis = 0 offset = 2
     : tensor<4xi8> -> tensor<5xi8>
   return %0 : tensor<5xi8>
@@ -917,12 +917,12 @@ func.func @shift_invalid_tensor_dimension_size(
 
 // -----
 
-mesh.mesh @mesh0(shape = 2x4)
+shard.grid @grid0(shape = 2x4)
 
 func.func @shift_invalid_shift_axis(
     %arg0 : tensor<4xi8>) -> tensor<4xi8> {
-  // expected-error@+1 {{Invalid shift axis 1. It must be one of the grouping mesh axes.}}
-  %0 = mesh.shift %arg0 on @mesh0 mesh_axes = [0]
+  // expected-error@+1 {{Invalid shift axis 1. It must be one of the grouping grid axes.}}
+  %0 = shard.shift %arg0 on @grid0 grid_axes = [0]
     shift_axis = 1 offset = 2
     : tensor<4xi8> -> tensor<4xi8>
   return %0 : tensor<4xi8>
diff --git a/mlir/test/Dialect/Mesh/ops.mlir b/mlir/test/Dialect/Shard/ops.mlir
index c354de5..5265dad 100644
--- a/mlir/test/Dialect/Mesh/ops.mlir
+++ b/mlir/test/Dialect/Shard/ops.mlir
@@ -1,176 +1,176 @@
 // RUN: mlir-opt %s | mlir-opt | FileCheck %s
 
-// CHECK: mesh.mesh @mesh0
-mesh.mesh @mesh0(shape = 2x2x4)
+// CHECK: shard.grid @grid0
+shard.grid @grid0(shape = 2x2x4)
 
-// CHECK: mesh.mesh @mesh1(shape = 4x?)
-mesh.mesh @mesh1(shape = 4x?)
+// CHECK: shard.grid @grid1(shape = 4x?)
+shard.grid @grid1(shape = 4x?)
 
-// CHECK: mesh.mesh @mesh2(shape = ?x4)
-mesh.mesh @mesh2(shape = ?x4)
+// CHECK: shard.grid @grid2(shape = ?x4)
+shard.grid @grid2(shape = ?x4)
 
-// CHECK: mesh.mesh @mesh3(shape = ?x?)
-mesh.mesh @mesh3(shape = ?x?)
+// CHECK: shard.grid @grid3(shape = ?x?)
+shard.grid @grid3(shape = ?x?)
 
-mesh.mesh @mesh4(shape = 3)
+shard.grid @grid4(shape = 3)
 
-// CHECK: mesh.mesh @mesh5(shape = ?)
-mesh.mesh @mesh5(shape = ?)
+// CHECK: shard.grid @grid5(shape = ?)
+shard.grid @grid5(shape = ?)
 
-// CHECK-LABEL: func @mesh_shard_op_fully_replicated
+// CHECK-LABEL: func @grid_shard_op_fully_replicated
 // CHECK-SAME: %[[ARG:.*]]: tensor<4x8xf32>
-func.func @mesh_shard_op_fully_replicated(%arg0 : tensor<4x8xf32>) -> tensor<4x8xf32> {
-  // CHECK-NEXT: %[[S:.*]] = mesh.sharding @mesh0 split_axes = {{\[\[}}]] : !mesh.sharding
-  %s = mesh.sharding @mesh0 split_axes = [[]] : !mesh.sharding
-  // CHECK-NEXT: mesh.shard %[[ARG]] to %[[S]] : tensor<4x8xf32>
-  %0 = mesh.shard %arg0 to %s : tensor<4x8xf32>
+func.func @grid_shard_op_fully_replicated(%arg0 : tensor<4x8xf32>) -> tensor<4x8xf32> {
+  // CHECK-NEXT: %[[S:.*]] = shard.sharding @grid0 split_axes = {{\[\[}}]] : !shard.sharding
+  %s = shard.sharding @grid0 split_axes = [[]] : !shard.sharding
+  // CHECK-NEXT: shard.shard %[[ARG]] to %[[S]] : tensor<4x8xf32>
+  %0 = shard.shard %arg0 to %s : tensor<4x8xf32>
   return %0 : tensor<4x8xf32>
 }
 
-// CHECK-LABEL: func @mesh_shard_op_1st_dim
+// CHECK-LABEL: func @grid_shard_op_1st_dim
 // CHECK-SAME: %[[ARG:.*]]: tensor<4x8xf32>
-func.func @mesh_shard_op_1st_dim(%arg0 : tensor<4x8xf32>) -> tensor<4x8xf32> {
-  // CHECK-NEXT: %[[S:.*]] = mesh.sharding @mesh0 split_axes = {{\[\[}}0]] : !mesh.sharding
-  %s = mesh.sharding @mesh0 split_axes = [[0]] : !mesh.sharding
+func.func @grid_shard_op_1st_dim(%arg0 : tensor<4x8xf32>) -> tensor<4x8xf32> {
+  // CHECK-NEXT: %[[S:.*]] = shard.sharding @grid0 split_axes = {{\[\[}}0]] : !shard.sharding
+  %s = shard.sharding @grid0 split_axes = [[0]] : !shard.sharding
 
-  %0 = mesh.shard %arg0 to %s : tensor<4x8xf32>
+  %0 = shard.shard %arg0 to %s : tensor<4x8xf32>
   return %0 : tensor<4x8xf32>
 }
 
-// CHECK-LABEL: func @mesh_shard_op_2nd_dim
+// CHECK-LABEL: func @grid_shard_op_2nd_dim
 // CHECK-SAME: %[[ARG:.*]]: tensor<4x8xf32>
-func.func @mesh_shard_op_2nd_dim(%arg0 : tensor<4x8xf32>) -> tensor<4x8xf32> {
-  // CHECK-NEXT: %[[S:.*]] = mesh.sharding @mesh1 split_axes = {{\[\[}}], [0]] : !mesh.sharding
-  %s = mesh.sharding @mesh1 split_axes = [[], [0]] : !mesh.sharding
-  // CHECK-NEXT: mesh.shard %[[ARG]] to %[[S]] : tensor<4x8xf32>
-  %0 = mesh.shard %arg0 to %s : tensor<4x8xf32>
+func.func @grid_shard_op_2nd_dim(%arg0 : tensor<4x8xf32>) -> tensor<4x8xf32> {
+  // CHECK-NEXT: %[[S:.*]] = shard.sharding @grid1 split_axes = {{\[\[}}], [0]] : !shard.sharding
+  %s = shard.sharding @grid1 split_axes = [[], [0]] : !shard.sharding
+  // CHECK-NEXT: shard.shard %[[ARG]] to %[[S]] : tensor<4x8xf32>
+  %0 = shard.shard %arg0 to %s : tensor<4x8xf32>
   return %0 : tensor<4x8xf32>
 }
 
-// CHECK-LABEL: func @mesh_shard_op_1st_and_3rd_dim
-func.func @mesh_shard_op_1st_and_3rd_dim(
+// CHECK-LABEL: func @grid_shard_op_1st_and_3rd_dim
+func.func @grid_shard_op_1st_and_3rd_dim(
     // CHECK-SAME: %[[ARG:.*]]: tensor<4x8x16xf32>
     %arg0 : tensor<4x8x16xf32>) -> tensor<4x8x16xf32> {
-  // CHECK-NEXT: %[[S:.*]] = mesh.sharding @mesh3 split_axes = {{\[\[}}0], [], [1]] : !mesh.sharding
-  %s = mesh.sharding @mesh3 split_axes = [[0], [], [1]] : !mesh.sharding
-  // CHECK-NEXT: mesh.shard %[[ARG]] to %[[S]] : tensor<4x8x16xf32>
-  %0 = mesh.shard %arg0 to %s : tensor<4x8x16xf32>
+  // CHECK-NEXT: %[[S:.*]] = shard.sharding @grid3 split_axes = {{\[\[}}0], [], [1]] : !shard.sharding
+  %s = shard.sharding @grid3 split_axes = [[0], [], [1]] : !shard.sharding
+  // CHECK-NEXT: shard.shard %[[ARG]] to %[[S]] : tensor<4x8x16xf32>
+  %0 = shard.shard %arg0 to %s : tensor<4x8x16xf32>
   return %0 : tensor<4x8x16xf32>
 }
 
-// CHECK-LABEL: func @mesh_shard_op_two_users
+// CHECK-LABEL: func @grid_shard_op_two_users
 // CHECK-SAME: %[[ARG:.*]]: tensor<4x8xf32>
-func.func @mesh_shard_op_two_users(%arg0 : tensor<4x8xf32>) ->
+func.func @grid_shard_op_two_users(%arg0 : tensor<4x8xf32>) ->
                                   (tensor<4x8xf32>, tensor<4x8xf32>) {
-  // CHECK-NEXT: %[[V0:.*]] = mesh.sharding @mesh0 split_axes = {{\[\[}}0]] : !mesh.sharding
-  %s0 = mesh.sharding @mesh0 split_axes = [[0]] : !mesh.sharding
-  %0 = mesh.shard %arg0 to %s0 : tensor<4x8xf32>
-  // CHECK-DAG: mesh.sharding @mesh0 split_axes = {{\[\[}}1]] : !mesh.sharding
-  %s1 = mesh.sharding @mesh0 split_axes = [[1]] : !mesh.sharding
-  %1 = mesh.shard %0 to %s1 annotate_for_users : tensor<4x8xf32>
-  // CHECK-DAG: mesh.sharding @mesh0 split_axes = {{\[\[}}2]] : !mesh.sharding
-  %s2 = mesh.sharding @mesh0 split_axes = [[2]] : !mesh.sharding
-  %2 = mesh.shard %0 to %s2 annotate_for_users : tensor<4x8xf32>
+  // CHECK-NEXT: %[[V0:.*]] = shard.sharding @grid0 split_axes = {{\[\[}}0]] : !shard.sharding
+  %s0 = shard.sharding @grid0 split_axes = [[0]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0 : tensor<4x8xf32>
+  // CHECK-DAG: shard.sharding @grid0 split_axes = {{\[\[}}1]] : !shard.sharding
+  %s1 = shard.sharding @grid0 split_axes = [[1]] : !shard.sharding
+  %1 = shard.shard %0 to %s1 annotate_for_users : tensor<4x8xf32>
+  // CHECK-DAG: shard.sharding @grid0 split_axes = {{\[\[}}2]] : !shard.sharding
+  %s2 = shard.sharding @grid0 split_axes = [[2]] : !shard.sharding
+  %2 = shard.shard %0 to %s2 annotate_for_users : tensor<4x8xf32>
   return %1, %2 : tensor<4x8xf32>, tensor<4x8xf32>
 }
 
-// CHECK-LABEL: func @mesh_shard_halo_sizes
-func.func @mesh_shard_halo_sizes() -> () {
+// CHECK-LABEL: func @grid_shard_halo_sizes
+func.func @grid_shard_halo_sizes() -> () {
   // CHECK: %[[C3:.*]] = arith.constant 3 : i64
   %c3 = arith.constant 3 : i64
-  // CHECK: mesh.sharding @mesh4 split_axes = {{\[\[}}0]] halo_sizes = [1, 4] : !mesh.sharding
-  %sharding1 = mesh.sharding @mesh4 split_axes = [[0]] halo_sizes = [1, 4] : !mesh.sharding
-  // CHECK: mesh.sharding @mesh4 split_axes = {{\[\[}}0]] halo_sizes = [4, %[[C3]]] : !mesh.sharding
-  %sharding2 = mesh.sharding @mesh4 split_axes = [[0]] halo_sizes = [4, %c3] : !mesh.sharding
+  // CHECK: shard.sharding @grid4 split_axes = {{\[\[}}0]] halo_sizes = [1, 4] : !shard.sharding
+  %sharding1 = shard.sharding @grid4 split_axes = [[0]] halo_sizes = [1, 4] : !shard.sharding
+  // CHECK: shard.sharding @grid4 split_axes = {{\[\[}}0]] halo_sizes = [4, %[[C3]]] : !shard.sharding
+  %sharding2 = shard.sharding @grid4 split_axes = [[0]] halo_sizes = [4, %c3] : !shard.sharding
   return
 }
 
-// CHECK-LABEL: func @mesh_shard_dims_sizes
-func.func @mesh_shard_dims_sizes() -> () {
+// CHECK-LABEL: func @grid_shard_dims_sizes
+func.func @grid_shard_dims_sizes() -> () {
   // CHECK: %[[C3:.*]] = arith.constant 3 : i64
   %c3 = arith.constant 3 : i64
-  // CHECK: mesh.sharding @mesh4 split_axes = {{\[\[}}0]] sharded_dims_offsets = [0, 1, 4, 6] : !mesh.sharding
-  %sharding1 = mesh.sharding @mesh4 split_axes = [[0]] sharded_dims_offsets = [0, 1, 4, 6] : !mesh.sharding
-  // CHECK: mesh.sharding @mesh4 split_axes = {{\[\[}}0]] sharded_dims_offsets = [0, 2, %[[C3]], 5] : !mesh.sharding
-  %sharding2 = mesh.sharding @mesh4 split_axes = [[0]] sharded_dims_offsets = [0, 2, %c3, 5] : !mesh.sharding
+  // CHECK: shard.sharding @grid4 split_axes = {{\[\[}}0]] sharded_dims_offsets = [0, 1, 4, 6] : !shard.sharding
+  %sharding1 = shard.sharding @grid4 split_axes = [[0]] sharded_dims_offsets = [0, 1, 4, 6] : !shard.sharding
+  // CHECK: shard.sharding @grid4 split_axes = {{\[\[}}0]] sharded_dims_offsets = [0, 2, %[[C3]], 5] : !shard.sharding
+  %sharding2 = shard.sharding @grid4 split_axes = [[0]] sharded_dims_offsets = [0, 2, %c3, 5] : !shard.sharding
   return
 }
 
-// CHECK-LABEL: func @mesh_shard_shape
-func.func @mesh_shard_shape() {
+// CHECK-LABEL: func @grid_shard_shape
+func.func @grid_shard_shape() {
   // CHECK: %[[C3:.*]] = arith.constant 3 : index
   %c3 = arith.constant 3 : index
-  // CHECK-NEXT: %[[S:.*]] = mesh.sharding @mesh0 split_axes = {{\[\[}}]] : !mesh.sharding
-  %s = mesh.sharding @mesh0 split_axes = [[]] : !mesh.sharding
-  // CHECK-NEXT: mesh.shard_shape dims = [8, %[[C3]]
+  // CHECK-NEXT: %[[S:.*]] = shard.sharding @grid0 split_axes = {{\[\[}}]] : !shard.sharding
+  %s = shard.sharding @grid0 split_axes = [[]] : !shard.sharding
+  // CHECK-NEXT: shard.shard_shape dims = [8, %[[C3]]
   // CHECK-SAME: ] sharding = %[[S]] device = [%[[C3]]
   // CHECK-SAME: ] : index, index
-  %shp:2 = mesh.shard_shape dims = [8, %c3] sharding = %s device = [%c3] : index, index
-  // CHECK-NEXT: mesh.shard_shape dims = [8, 4] sharding = %[[S]] device = [3] : index, index
-  %shp1:2 = mesh.shard_shape dims = [8, 4] sharding = %s device = [3] : index, index
+  %shp:2 = shard.shard_shape dims = [8, %c3] sharding = %s device = [%c3] : index, index
+  // CHECK-NEXT: shard.shard_shape dims = [8, 4] sharding = %[[S]] device = [3] : index, index
+  %shp1:2 = shard.shard_shape dims = [8, 4] sharding = %s device = [3] : index, index
   return
 }
 
-// CHECK-LABEL: func @mesh_get_sharding
+// CHECK-LABEL: func @grid_get_sharding
 // CHECK-SAME: %[[ARG:.*]]: tensor<4x8xf32>
-func.func @mesh_get_sharding(%arg0 : tensor<4x8xf32>) -> !mesh.sharding {
-  // CHECK-NEXT: mesh.get_sharding %[[ARG]] : tensor<4x8xf32> -> !mesh.sharding
-  %0 = mesh.get_sharding %arg0 : tensor<4x8xf32> -> !mesh.sharding
-  return %0 : !mesh.sharding
+func.func @grid_get_sharding(%arg0 : tensor<4x8xf32>) -> !shard.sharding {
+  // CHECK-NEXT: shard.get_sharding %[[ARG]] : tensor<4x8xf32> -> !shard.sharding
+  %0 = shard.get_sharding %arg0 : tensor<4x8xf32> -> !shard.sharding
+  return %0 : !shard.sharding
 }
 
-// CHECK-LABEL: func @mesh_shape
-func.func @mesh_shape() -> (index, index) {
-  // CHECK: %[[RES:.*]]:2 = mesh.mesh_shape @mesh0 axes = [0, 1] : index, index
-  %0:2 = mesh.mesh_shape @mesh0 axes = [0, 1] : index, index
+// CHECK-LABEL: func @grid_shape
+func.func @grid_shape() -> (index, index) {
+  // CHECK: %[[RES:.*]]:2 = shard.grid_shape @grid0 axes = [0, 1] : index, index
+  %0:2 = shard.grid_shape @grid0 axes = [0, 1] : index, index
   // CHECK: return %[[RES]]#0, %[[RES]]#1 : index, index
   return %0#0, %0#1 : index, index
 }
 
-// CHECK-LABEL: func @mesh_shape_default_axes
-func.func @mesh_shape_default_axes() -> (index, index, index) {
-  // CHECK: %[[RES:.*]]:3 = mesh.mesh_shape @mesh0 : index, index, index
-  %0:3 = mesh.mesh_shape @mesh0 : index, index, index
+// CHECK-LABEL: func @grid_shape_default_axes
+func.func @grid_shape_default_axes() -> (index, index, index) {
+  // CHECK: %[[RES:.*]]:3 = shard.grid_shape @grid0 : index, index, index
+  %0:3 = shard.grid_shape @grid0 : index, index, index
   // CHECK: return %[[RES]]#0, %[[RES]]#1, %[[RES]]#2 : index, index, index
   return %0#0, %0#1, %0#2 : index, index, index
 }
 
-// CHECK-LABEL: func @mesh_shape_empty_axes
-func.func @mesh_shape_empty_axes() -> (index, index, index) {
-  // CHECK: %[[RES:.*]]:3 = mesh.mesh_shape @mesh0 : index, index, index
-  %0:3 = mesh.mesh_shape @mesh0 axes = [] : index, index, index
+// CHECK-LABEL: func @grid_shape_empty_axes
+func.func @grid_shape_empty_axes() -> (index, index, index) {
+  // CHECK: %[[RES:.*]]:3 = shard.grid_shape @grid0 : index, index, index
+  %0:3 = shard.grid_shape @grid0 axes = [] : index, index, index
   // CHECK: return %[[RES]]#0, %[[RES]]#1, %[[RES]]#2 : index, index, index
   return %0#0, %0#1, %0#2 : index, index, index
 }
 
 // CHECK-LABEL: func @process_multi_index
 func.func @process_multi_index() -> (index, index) {
-  // CHECK: %[[RES:.*]]:2 = mesh.process_multi_index on @mesh0 axes = [0, 1] : index, index
-  %0:2 = mesh.process_multi_index on @mesh0 axes = [0, 1] : index, index
+  // CHECK: %[[RES:.*]]:2 = shard.process_multi_index on @grid0 axes = [0, 1] : index, index
+  %0:2 = shard.process_multi_index on @grid0 axes = [0, 1] : index, index
   // CHECK: return %[[RES]]#0, %[[RES]]#1 : index, index
   return %0#0, %0#1 : index, index
 }
 
 // CHECK-LABEL: func @process_multi_index_default_axes
 func.func @process_multi_index_default_axes() -> (index, index, index) {
-  // CHECK: %[[RES:.*]]:3 = mesh.process_multi_index on @mesh0 : index, index, index
-  %0:3 = mesh.process_multi_index on @mesh0 : index, index, index
+  // CHECK: %[[RES:.*]]:3 = shard.process_multi_index on @grid0 : index, index, index
+  %0:3 = shard.process_multi_index on @grid0 : index, index, index
   // CHECK: return %[[RES]]#0, %[[RES]]#1, %[[RES]]#2 : index, index, index
   return %0#0, %0#1, %0#2 : index, index, index
 }
 
 // CHECK-LABEL: func @process_multi_index_empty_axes
 func.func @process_multi_index_empty_axes() -> (index, index, index) {
-  // CHECK: %[[RES:.*]]:3 = mesh.process_multi_index on @mesh0 : index, index, index
-  %0:3 = mesh.process_multi_index on @mesh0 axes = [] : index, index, index
+  // CHECK: %[[RES:.*]]:3 = shard.process_multi_index on @grid0 : index, index, index
+  %0:3 = shard.process_multi_index on @grid0 axes = [] : index, index, index
   // CHECK: return %[[RES]]#0, %[[RES]]#1, %[[RES]]#2 : index, index, index
   return %0#0, %0#1, %0#2 : index, index, index
 }
 
 // CHECK-LABEL: func @process_linear_index
 func.func @process_linear_index() -> index {
-  // CHECK: %[[RES:.*]] = mesh.process_linear_index on @mesh0 : index
-  %0 = mesh.process_linear_index on @mesh0 : index
+  // CHECK: %[[RES:.*]] = shard.process_linear_index on @grid0 : index
+  %0 = shard.process_linear_index on @grid0 : index
   // CHECK: return %[[RES]] : index
   return %0 : index
 }
@@ -179,9 +179,9 @@ func.func @process_linear_index() -> index {
 func.func @all_reduce(
     // CHECK-SAME: %[[ARG:.*]]: tensor<3x4xf32>
     %arg0 : tensor<3x4xf32>) -> tensor<3x4xf64> {
-  // CHECK-NEXT: mesh.all_reduce %[[ARG]] on @mesh0 mesh_axes = [1, 0] reduction = max
+  // CHECK-NEXT: shard.all_reduce %[[ARG]] on @grid0 grid_axes = [1, 0] reduction = max
   // CHECK-SAME: : tensor<3x4xf32> -> tensor<3x4xf64>
-  %0 = mesh.all_reduce %arg0 on @mesh0 mesh_axes = [1, 0] reduction = max
+  %0 = shard.all_reduce %arg0 on @grid0 grid_axes = [1, 0] reduction = max
     : tensor<3x4xf32> -> tensor<3x4xf64>
   return %0 : tensor<3x4xf64>
 }
@@ -190,9 +190,9 @@ func.func @all_reduce(
 func.func @all_gather(
     // CHECK-SAME: %[[ARG:.*]]: tensor<3x4xf32>
     %arg0 : tensor<3x4xf32>) -> tensor<3x16xf32> {
-  // CHECK-NEXT: mesh.all_gather %[[ARG]] on @mesh0 mesh_axes = [2] gather_axis = 1
+  // CHECK-NEXT: shard.all_gather %[[ARG]] on @grid0 grid_axes = [2] gather_axis = 1
   // CHECK-SAME: : tensor<3x4xf32> -> tensor<3x16xf32>
-  %0 = mesh.all_gather %arg0 on @mesh0 mesh_axes = [2] gather_axis = 1
+  %0 = shard.all_gather %arg0 on @grid0 grid_axes = [2] gather_axis = 1
     : tensor<3x4xf32> -> tensor<3x16xf32>
   return %0 : tensor<3x16xf32>
 }
@@ -201,20 +201,20 @@ func.func @all_gather(
 func.func @all_gather_dynamic_dims_in_tensor(
     // CHECK-SAME: %[[ARG:.*]]: tensor<?x?xf32>
     %arg0 : tensor<?x?xf32>) -> tensor<?x?xf32> {
-  // CHECK-NEXT: mesh.all_gather %[[ARG]] on @mesh0 mesh_axes = [2] gather_axis = 1
+  // CHECK-NEXT: shard.all_gather %[[ARG]] on @grid0 grid_axes = [2] gather_axis = 1
   // CHECK-SAME: : tensor<?x?xf32> -> tensor<?x?xf32>
-  %0 = mesh.all_gather %arg0 on @mesh0 mesh_axes = [2] gather_axis = 1
+  %0 = shard.all_gather %arg0 on @grid0 grid_axes = [2] gather_axis = 1
     : tensor<?x?xf32> -> tensor<?x?xf32>
   return %0 : tensor<?x?xf32>
 }
 
-// CHECK-LABEL: func @all_gather_dynamic_dims_in_mesh
-func.func @all_gather_dynamic_dims_in_mesh(
+// CHECK-LABEL: func @all_gather_dynamic_dims_in_grid
+func.func @all_gather_dynamic_dims_in_grid(
     // CHECK-SAME: %[[ARG:.*]]: tensor<5x6xf32>
     %arg0 : tensor<5x6xf32>) -> tensor<5x?xf32> {
-  // CHECK-NEXT: mesh.all_gather %[[ARG]] on @mesh3 mesh_axes = [1] gather_axis = 1
+  // CHECK-NEXT: shard.all_gather %[[ARG]] on @grid3 grid_axes = [1] gather_axis = 1
   // CHECK-SAME: : tensor<5x6xf32> -> tensor<5x?xf32>
-  %0 = mesh.all_gather %arg0 on @mesh3 mesh_axes = [1] gather_axis = 1
+  %0 = shard.all_gather %arg0 on @grid3 grid_axes = [1] gather_axis = 1
     : tensor<5x6xf32> -> tensor<5x?xf32>
   return %0 : tensor<5x?xf32>
 }
@@ -223,10 +223,10 @@ func.func @all_gather_dynamic_dims_in_mesh(
 func.func @all_slice_static_dimensions(
     // CHECK-SAME: %[[ARG:.*]]: tensor<3x4xf32>
     %arg0 : tensor<3x4xf32>) -> tensor<3x1xf32> {
-  // CHECK-NEXT: mesh.all_slice %[[ARG]]
-  // CHECK-SAME: on @mesh0 mesh_axes = [2] slice_axis = 1
+  // CHECK-NEXT: shard.all_slice %[[ARG]]
+  // CHECK-SAME: on @grid0 grid_axes = [2] slice_axis = 1
   // CHECK-SAME: : tensor<3x4xf32> -> tensor<3x1xf32>
-  %0 = mesh.all_slice %arg0 on @mesh0 mesh_axes = [2] slice_axis = 1
+  %0 = shard.all_slice %arg0 on @grid0 grid_axes = [2] slice_axis = 1
     : tensor<3x4xf32> -> tensor<3x1xf32>
   return %0 : tensor<3x1xf32>
 }
@@ -235,10 +235,10 @@ func.func @all_slice_static_dimensions(
 func.func @all_slice_dynamic_dimensions(
     // CHECK-SAME: %[[ARG:.*]]: tensor<?xf32>
     %arg0 : tensor<?xf32>) -> tensor<?xf32> {
-  // CHECK-NEXT: mesh.all_slice %[[ARG]]
-  // CHECK-SAME: on @mesh3 mesh_axes = [0, 1] slice_axis = 0
+  // CHECK-NEXT: shard.all_slice %[[ARG]]
+  // CHECK-SAME: on @grid3 grid_axes = [0, 1] slice_axis = 0
   // CHECK-SAME: : tensor<?xf32> -> tensor<?xf32>
-  %0 = mesh.all_slice %arg0 on @mesh3 mesh_axes = [0, 1] slice_axis = 0
+  %0 = shard.all_slice %arg0 on @grid3 grid_axes = [0, 1] slice_axis = 0
     : tensor<?xf32> -> tensor<?xf32>
   return %0 : tensor<?xf32>
 }
@@ -247,10 +247,10 @@ func.func @all_slice_dynamic_dimensions(
 func.func @all_to_all(
     // CHECK-SAME: %[[ARG:.*]]: tensor<3x6xi8>
     %arg0 : tensor<3x6xi8>) -> tensor<3x6xi8> {
-  // CHECK-NEXT: mesh.all_to_all %[[ARG]]
-  // CHECK-SAME: on @mesh4 split_axis = 1 concat_axis = 0
+  // CHECK-NEXT: shard.all_to_all %[[ARG]]
+  // CHECK-SAME: on @grid4 split_axis = 1 concat_axis = 0
   // CHECK-SAME: : tensor<3x6xi8> -> tensor<3x6xi8>
-  %0 = mesh.all_to_all %arg0 on @mesh4
+  %0 = shard.all_to_all %arg0 on @grid4
     split_axis = 1 concat_axis = 0
     : tensor<3x6xi8> -> tensor<3x6xi8>
   return %0 : tensor<3x6xi8>
@@ -260,10 +260,10 @@ func.func @all_to_all(
 func.func @all_to_all_dynamic_dims_in_result(
     // CHECK-SAME: %[[ARG:.*]]: tensor<3x6xi8>
     %arg0 : tensor<3x6xi8>) -> tensor<3x?xi8> {
-  // CHECK-NEXT: mesh.all_to_all %[[ARG]]
-  // CHECK-SAME: on @mesh4 split_axis = 1 concat_axis = 0
+  // CHECK-NEXT: shard.all_to_all %[[ARG]]
+  // CHECK-SAME: on @grid4 split_axis = 1 concat_axis = 0
   // CHECK-SAME: : tensor<3x6xi8> -> tensor<3x?xi8>
-  %0 = mesh.all_to_all %arg0 on @mesh4
+  %0 = shard.all_to_all %arg0 on @grid4
     split_axis = 1 concat_axis = 0
     : tensor<3x6xi8> -> tensor<3x?xi8>
   return %0 : tensor<3x?xi8>
@@ -273,10 +273,10 @@ func.func @all_to_all_dynamic_dims_in_result(
 func.func @all_to_all_same_split_concat_dim_with_dynamic_device_group_size(
     // CHECK-SAME: %[[ARG:.*]]: tensor<3xi8>
     %arg0 : tensor<3xi8>) -> tensor<3xi8> {
-  // CHECK-NEXT: mesh.all_to_all %[[ARG]]
-  // CHECK-SAME: @mesh4 split_axis = 0 concat_axis = 0
+  // CHECK-NEXT: shard.all_to_all %[[ARG]]
+  // CHECK-SAME: @grid4 split_axis = 0 concat_axis = 0
   // CHECK-SAME: : tensor<3xi8> -> tensor<3xi8>
-  %0 = mesh.all_to_all %arg0 on @mesh4
+  %0 = shard.all_to_all %arg0 on @grid4
     split_axis = 0 concat_axis = 0
     : tensor<3xi8> -> tensor<3xi8>
   return %0 : tensor<3xi8>
@@ -286,10 +286,10 @@ func.func @all_to_all_same_split_concat_dim_with_dynamic_device_group_size(
 func.func @all_to_all_non_divisible_split_axis_size(
     // CHECK-SAME: %[[ARG:.*]]: tensor<2x3xi8>
     %arg0 : tensor<2x3xi8>) -> tensor<?x12xi8> {
-  // CHECK-NEXT: mesh.all_to_all %[[ARG]]
-  // CHECK-SAME: @mesh0 mesh_axes = [0, 1] split_axis = 0 concat_axis = 1
+  // CHECK-NEXT: shard.all_to_all %[[ARG]]
+  // CHECK-SAME: @grid0 grid_axes = [0, 1] split_axis = 0 concat_axis = 1
   // CHECK-SAME: : tensor<2x3xi8> -> tensor<?x12xi8>
-  %0 = mesh.all_to_all %arg0 on @mesh0 mesh_axes = [0, 1]
+  %0 = shard.all_to_all %arg0 on @grid0 grid_axes = [0, 1]
     split_axis = 0 concat_axis = 1
     : tensor<2x3xi8> -> tensor<?x12xi8>
   return %0 : tensor<?x12xi8>
@@ -299,11 +299,11 @@ func.func @all_to_all_non_divisible_split_axis_size(
 func.func @broadcast_static_root(
     // CHECK-SAME: %[[ARG:.*]]: tensor<3x6xi8>
     %arg0 : tensor<3x6xi8>) -> tensor<3x6xi8> {
-  // CHECK-NEXT: mesh.broadcast %[[ARG]]
-  // CHECK-SAME: on @mesh0 mesh_axes = [0, 2]
+  // CHECK-NEXT: shard.broadcast %[[ARG]]
+  // CHECK-SAME: on @grid0 grid_axes = [0, 2]
   // CHECK-SAME: root = [0, 1]
   // CHECK-SAME: : (tensor<3x6xi8>) -> tensor<3x6xi8>
-  %0 = mesh.broadcast %arg0 on @mesh0 mesh_axes = [0, 2]
+  %0 = shard.broadcast %arg0 on @grid0 grid_axes = [0, 2]
     root = [0, 1]
     : (tensor<3x6xi8>) -> tensor<3x6xi8>
   return %0 : tensor<3x6xi8>
@@ -316,11 +316,11 @@ func.func @broadcast_dynamic_root(
     // CHECK-SAME: %[[ARG1:.*]]: index
     %arg1 : index
     ) -> tensor<3x6xi8> {
-  // CHECK-NEXT: mesh.broadcast %[[ARG0]]
-  // CHECK-SAME: on @mesh0 mesh_axes = [0, 2]
+  // CHECK-NEXT: shard.broadcast %[[ARG0]]
+  // CHECK-SAME: on @grid0 grid_axes = [0, 2]
   // CHECK-SAME: root = [1, %[[ARG1]]]
   // CHECK-SAME: : (tensor<3x6xi8>, index) -> tensor<3x6xi8>
-  %0 = mesh.broadcast %arg0 on @mesh0 mesh_axes = [0, 2]
+  %0 = shard.broadcast %arg0 on @grid0 grid_axes = [0, 2]
     root = [1, %arg1]
     : (tensor<3x6xi8>, index) -> tensor<3x6xi8>
   return %0 : tensor<3x6xi8>
@@ -330,12 +330,12 @@ func.func @broadcast_dynamic_root(
 func.func @gather_static_root(
     // CHECK-SAME: %[[ARG:.*]]: tensor<3x6xi8>
     %arg0 : tensor<3x6xi8>) -> tensor<24x6xi8> {
-  // CHECK-NEXT: mesh.gather %[[ARG]]
-  // CHECK-SAME: on @mesh0 mesh_axes = [0, 2]
+  // CHECK-NEXT: shard.gather %[[ARG]]
+  // CHECK-SAME: on @grid0 grid_axes = [0, 2]
   // CHECK-SAME: gather_axis = 0
   // CHECK-SAME: root = [0, 1]
   // CHECK-SAME: : (tensor<3x6xi8>) -> tensor<24x6xi8>
-  %0 = mesh.gather %arg0 on @mesh0 mesh_axes = [0, 2]
+  %0 = shard.gather %arg0 on @grid0 grid_axes = [0, 2]
     gather_axis = 0
     root = [0, 1]
     : (tensor<3x6xi8>) -> tensor<24x6xi8>
@@ -349,12 +349,12 @@ func.func @gather_dynamic_root(
     // CHECK-SAME: %[[ARG1:.*]]: index
     %arg1 : index
     ) -> tensor<24x6xi8> {
-  // CHECK-NEXT: mesh.gather %[[ARG0]]
-  // CHECK-SAME: on @mesh0 mesh_axes = [0, 2]
+  // CHECK-NEXT: shard.gather %[[ARG0]]
+  // CHECK-SAME: on @grid0 grid_axes = [0, 2]
   // CHECK-SAME: gather_axis = 0
   // CHECK-SAME: root = [1, %[[ARG1]]]
   // CHECK-SAME: : (tensor<3x6xi8>, index) -> tensor<24x6xi8>
-  %0 = mesh.gather %arg0 on @mesh0 mesh_axes = [0, 2]
+  %0 = shard.gather %arg0 on @grid0 grid_axes = [0, 2]
     gather_axis = 0
     root = [1, %arg1]
     : (tensor<3x6xi8>, index) -> tensor<24x6xi8>
@@ -365,11 +365,11 @@ func.func @gather_dynamic_root(
 func.func @receive_static_source(
     // CHECK-SAME: %[[ARG:.*]]: tensor<2xi8>
     %arg0 : tensor<2xi8>) -> tensor<2xi8> {
-  // CHECK-NEXT: mesh.recv %[[ARG]]
-  // CHECK-SAME: on @mesh0 mesh_axes = [0, 2]
+  // CHECK-NEXT: shard.recv %[[ARG]]
+  // CHECK-SAME: on @grid0 grid_axes = [0, 2]
   // CHECK-SAME: source = [0, 1]
   // CHECK-SAME: : (tensor<2xi8>) -> tensor<2xi8>
-  %0 = mesh.recv %arg0 on @mesh0 mesh_axes = [0, 2]
+  %0 = shard.recv %arg0 on @grid0 grid_axes = [0, 2]
     source = [0, 1]
     : (tensor<2xi8>) -> tensor<2xi8>
   return %0 : tensor<2xi8>
@@ -382,11 +382,11 @@ func.func @receive_dynamic_source(
     // CHECK-SAME: %[[ARG1:.*]]: index
     %arg1 : index
     ) -> tensor<2xi8> {
-  // CHECK-NEXT: mesh.recv %[[ARG0]]
-  // CHECK-SAME: on @mesh0 mesh_axes = [0, 2]
+  // CHECK-NEXT: shard.recv %[[ARG0]]
+  // CHECK-SAME: on @grid0 grid_axes = [0, 2]
   // CHECK-SAME: source = [1, %[[ARG1]]]
   // CHECK-SAME: : (tensor<2xi8>, index) -> tensor<2xi8>
-  %0 = mesh.recv %arg0 on @mesh0 mesh_axes = [0, 2]
+  %0 = shard.recv %arg0 on @grid0 grid_axes = [0, 2]
     source = [1, %arg1]
     : (tensor<2xi8>, index) -> tensor<2xi8>
   return %0 : tensor<2xi8>
@@ -396,9 +396,9 @@ func.func @receive_dynamic_source(
 func.func @receive_no_source(
     // CHECK-SAME: %[[ARG:.*]]: tensor<2xi8>
     %arg0 : tensor<2xi8>) -> tensor<2xi8> {
-  // CHECK-NEXT: mesh.recv %[[ARG]]
+  // CHECK-NEXT: shard.recv %[[ARG]]
   // CHECK-NOT: source
-  %0 = mesh.recv %arg0 on @mesh0 mesh_axes = [0, 2]
+  %0 = shard.recv %arg0 on @grid0 grid_axes = [0, 2]
     : (tensor<2xi8>) -> tensor<2xi8>
   return %0 : tensor<2xi8>
 }
@@ -407,11 +407,11 @@ func.func @receive_no_source(
 func.func @reduce_static_root(
     // CHECK-SAME: %[[ARG:.*]]: tensor<2xi8>
     %arg0 : tensor<2xi8>) -> tensor<2xi8> {
-  // CHECK-NEXT: mesh.reduce %[[ARG]]
-  // CHECK-SAME: on @mesh0 mesh_axes = [0, 2]
+  // CHECK-NEXT: shard.reduce %[[ARG]]
+  // CHECK-SAME: on @grid0 grid_axes = [0, 2]
   // CHECK-SAME: root = [0, 1]
   // CHECK-SAME: : (tensor<2xi8>) -> tensor<2xi8>
-  %0 = mesh.reduce %arg0 on @mesh0 mesh_axes = [0, 2]
+  %0 = shard.reduce %arg0 on @grid0 grid_axes = [0, 2]
     root = [0, 1]
     : (tensor<2xi8>) -> tensor<2xi8>
   return %0 : tensor<2xi8>
@@ -424,11 +424,11 @@ func.func @reduce_dynamic_root(
     // CHECK-SAME: %[[ARG1:.*]]: index
     %arg1 : index
     ) -> tensor<2xi8> {
-  // CHECK-NEXT: mesh.reduce %[[ARG0]]
-  // CHECK-SAME: on @mesh0 mesh_axes = [0, 2]
+  // CHECK-NEXT: shard.reduce %[[ARG0]]
+  // CHECK-SAME: on @grid0 grid_axes = [0, 2]
   // CHECK-SAME: root = [1, %[[ARG1]]]
   // CHECK-SAME: : (tensor<2xi8>, index) -> tensor<2xi8>
-  %0 = mesh.reduce %arg0 on @mesh0 mesh_axes = [0, 2]
+  %0 = shard.reduce %arg0 on @grid0 grid_axes = [0, 2]
     root = [1, %arg1]
     : (tensor<2xi8>, index) -> tensor<2xi8>
   return %0 : tensor<2xi8>
@@ -438,11 +438,11 @@ func.func @reduce_dynamic_root(
 func.func @reduce_different_return_element_type(
     // CHECK-SAME: %[[ARG:.*]]: tensor<2xi8>
     %arg0 : tensor<2xi8>) -> tensor<2xi16> {
-  // CHECK-NEXT: mesh.reduce %[[ARG]]
-  // CHECK-SAME: on @mesh0 mesh_axes = [0, 2]
+  // CHECK-NEXT: shard.reduce %[[ARG]]
+  // CHECK-SAME: on @grid0 grid_axes = [0, 2]
   // CHECK-SAME: root = [0, 1]
   // CHECK-SAME: : (tensor<2xi8>) -> tensor<2xi16>
-  %0 = mesh.reduce %arg0 on @mesh0 mesh_axes = [0, 2]
+  %0 = shard.reduce %arg0 on @grid0 grid_axes = [0, 2]
     root = [0, 1]
     : (tensor<2xi8>) -> tensor<2xi16>
   return %0 : tensor<2xi16>
@@ -452,10 +452,10 @@ func.func @reduce_different_return_element_type(
 func.func @reduce_scatter_static_dimensions(
     // CHECK-SAME: %[[ARG:.*]]: tensor<3x4xf32>
     %arg0 : tensor<3x4xf32>) -> tensor<3x1xf64> {
-  // CHECK-NEXT: mesh.reduce_scatter %[[ARG]]
-  // CHECK-SAME: on @mesh0 mesh_axes = [2] reduction = max scatter_axis = 1
+  // CHECK-NEXT: shard.reduce_scatter %[[ARG]]
+  // CHECK-SAME: on @grid0 grid_axes = [2] reduction = max scatter_axis = 1
   // CHECK-SAME: : tensor<3x4xf32> -> tensor<3x1xf64>
-  %0 = mesh.reduce_scatter %arg0 on @mesh0 mesh_axes = [2]
+  %0 = shard.reduce_scatter %arg0 on @grid0 grid_axes = [2]
     reduction = max scatter_axis = 1
     : tensor<3x4xf32> -> tensor<3x1xf64>
   return %0 : tensor<3x1xf64>
@@ -465,10 +465,10 @@ func.func @reduce_scatter_static_dimensions(
 func.func @reduce_scatter_dynamic_dimensions(
     // CHECK-SAME: %[[ARG:.*]]: tensor<?xf32>
     %arg0 : tensor<?xf32>) -> tensor<?xf64> {
-  // CHECK-NEXT: mesh.reduce_scatter %[[ARG]]
-  // CHECK-SAME: on @mesh3 mesh_axes = [0, 1] scatter_axis = 0
+  // CHECK-NEXT: shard.reduce_scatter %[[ARG]]
+  // CHECK-SAME: on @grid3 grid_axes = [0, 1] scatter_axis = 0
   // CHECK-SAME: : tensor<?xf32> -> tensor<?xf64>
-  %0 = mesh.reduce_scatter %arg0 on @mesh3 mesh_axes = [0, 1] scatter_axis = 0
+  %0 = shard.reduce_scatter %arg0 on @grid3 grid_axes = [0, 1] scatter_axis = 0
     : tensor<?xf32> -> tensor<?xf64>
   return %0 : tensor<?xf64>
 }
@@ -477,11 +477,11 @@ func.func @reduce_scatter_dynamic_dimensions(
 func.func @scatter_static_dimensions(
     // CHECK-SAME: %[[ARG:.*]]: tensor<3x4xf32>
     %arg0 : tensor<3x4xf32>) -> tensor<3x1xf32> {
-  // CHECK-NEXT: mesh.scatter %[[ARG]]
-  // CHECK-SAME: on @mesh0 mesh_axes = [2]
+  // CHECK-NEXT: shard.scatter %[[ARG]]
+  // CHECK-SAME: on @grid0 grid_axes = [2]
   // CHECK-SAME: scatter_axis = 1 root = [1]
   // CHECK-SAME: : (tensor<3x4xf32>) -> tensor<3x1xf32>
-  %0 = mesh.scatter %arg0 on @mesh0 mesh_axes = [2]
+  %0 = shard.scatter %arg0 on @grid0 grid_axes = [2]
     scatter_axis = 1 root = [1]
     : (tensor<3x4xf32>) -> tensor<3x1xf32>
   return %0 : tensor<3x1xf32>
@@ -491,11 +491,11 @@ func.func @scatter_static_dimensions(
 func.func @scatter_dynamic_dimensions(
     // CHECK-SAME: %[[ARG:.*]]: tensor<?xf32>
     %arg0 : tensor<?xf32>) -> tensor<?xf32> {
-  // CHECK-NEXT: mesh.scatter %[[ARG]]
-  // CHECK-SAME: on @mesh3 mesh_axes = [0, 1]
+  // CHECK-NEXT: shard.scatter %[[ARG]]
+  // CHECK-SAME: on @grid3 grid_axes = [0, 1]
   // CHECK-SAME: scatter_axis = 0 root = [1, 2]
   // CHECK-SAME: : (tensor<?xf32>) -> tensor<?xf32>
-  %0 = mesh.scatter %arg0 on @mesh3 mesh_axes = [0, 1]
+  %0 = shard.scatter %arg0 on @grid3 grid_axes = [0, 1]
     scatter_axis = 0 root = [1, 2]
     : (tensor<?xf32>) -> tensor<?xf32>
   return %0 : tensor<?xf32>
@@ -508,12 +508,12 @@ func.func @scatter_dynamic_root(
     // CHECK-SAME: %[[ARG1:.*]]: index
     %arg1 : index
     ) -> tensor<1xi8> {
-  // CHECK-NEXT: mesh.scatter %[[ARG0]]
-  // CHECK-SAME: on @mesh0 mesh_axes = [0, 2]
+  // CHECK-NEXT: shard.scatter %[[ARG0]]
+  // CHECK-SAME: on @grid0 grid_axes = [0, 2]
   // CHECK-SAME: scatter_axis = 0
   // CHECK-SAME: root = [1, %[[ARG1]]]
   // CHECK-SAME: : (tensor<8xi8>, index) -> tensor<1xi8>
-  %0 = mesh.scatter %arg0 on @mesh0 mesh_axes = [0, 2]
+  %0 = shard.scatter %arg0 on @grid0 grid_axes = [0, 2]
     scatter_axis = 0
     root = [1, %arg1]
     : (tensor<8xi8>, index) -> tensor<1xi8>
@@ -524,11 +524,11 @@ func.func @scatter_dynamic_root(
 func.func @send_static_destination(
     // CHECK-SAME: %[[ARG:.*]]: tensor<2xi8>
     %arg0 : tensor<2xi8>) -> tensor<2xi8> {
-  // CHECK-NEXT: mesh.send %[[ARG]]
-  // CHECK-SAME: on @mesh0 mesh_axes = [0, 2]
+  // CHECK-NEXT: shard.send %[[ARG]]
+  // CHECK-SAME: on @grid0 grid_axes = [0, 2]
   // CHECK-SAME: destination = [0, 1]
   // CHECK-SAME: : (tensor<2xi8>) -> tensor<2xi8>
-  %0 = mesh.send %arg0 on @mesh0 mesh_axes = [0, 2]
+  %0 = shard.send %arg0 on @grid0 grid_axes = [0, 2]
     destination = [0, 1]
     : (tensor<2xi8>) -> tensor<2xi8>
   return %0 : tensor<2xi8>
@@ -541,11 +541,11 @@ func.func @send_dynamic_destination(
     // CHECK-SAME: %[[ARG1:.*]]: index
     %arg1 : index
     ) -> tensor<2xi8> {
-  // CHECK-NEXT: mesh.send %[[ARG0]]
-  // CHECK-SAME: on @mesh0 mesh_axes = [0, 2]
+  // CHECK-NEXT: shard.send %[[ARG0]]
+  // CHECK-SAME: on @grid0 grid_axes = [0, 2]
   // CHECK-SAME: destination = [1, %[[ARG1]]]
   // CHECK-SAME: : (tensor<2xi8>, index) -> tensor<2xi8>
-  %0 = mesh.send %arg0 on @mesh0 mesh_axes = [0, 2]
+  %0 = shard.send %arg0 on @grid0 grid_axes = [0, 2]
     destination = [1, %arg1]
     : (tensor<2xi8>, index) -> tensor<2xi8>
   return %0 : tensor<2xi8>
@@ -555,11 +555,11 @@ func.func @send_dynamic_destination(
 func.func @shift(
     // CHECK-SAME: %[[ARG:.*]]: tensor<2xi8>
     %arg0 : tensor<2xi8>) -> tensor<2xi8> {
-  // CHECK-NEXT: mesh.shift %[[ARG]]
-  // CHECK-SAME: on @mesh0 mesh_axes = [0, 2]
+  // CHECK-NEXT: shard.shift %[[ARG]]
+  // CHECK-SAME: on @grid0 grid_axes = [0, 2]
   // CHECK-SAME: shift_axis = 2 offset = -2 rotate
   // CHECK-SAME: : tensor<2xi8> -> tensor<2xi8>
-  %0 = mesh.shift %arg0 on @mesh0 mesh_axes = [0, 2]
+  %0 = shard.shift %arg0 on @grid0 grid_axes = [0, 2]
     shift_axis = 2 offset = -2 rotate
     : tensor<2xi8> -> tensor<2xi8>
   return %0 : tensor<2xi8>
@@ -570,16 +570,16 @@ func.func @update_halo(
     // CHECK-SAME: %[[ARG:.*]]: memref<12x12xi8>
     %arg0 : memref<12x12xi8>) {
   // CHECK-NEXT: %[[C2:.*]] = arith.constant 2 : i64
-  // CHECK-NEXT: %[[UH1:.*]] = mesh.update_halo %[[ARG]] on @mesh0
+  // CHECK-NEXT: %[[UH1:.*]] = shard.update_halo %[[ARG]] on @grid0
   // CHECK-SAME: split_axes = {{\[\[}}0]]
   // CHECK-SAME: halo_sizes = [2, %c2_i64] : memref<12x12xi8>
   %c2 = arith.constant 2 : i64
-  %uh1 = mesh.update_halo %arg0 on @mesh0 split_axes = [[0]]
+  %uh1 = shard.update_halo %arg0 on @grid0 split_axes = [[0]]
     halo_sizes = [2, %c2] : memref<12x12xi8>
-  // CHECK-NEXT: %[[UH2:.*]] = mesh.update_halo %[[UH1]] on @mesh0
+  // CHECK-NEXT: %[[UH2:.*]] = shard.update_halo %[[UH1]] on @grid0
   // CHECK-SAME: split_axes = {{\[\[}}0], [1]]
   // CHECK-SAME: halo_sizes = [2, 2, %[[C2]], 2] : memref<12x12xi8>
-  %uh2 = mesh.update_halo %uh1 on @mesh0 split_axes = [[0], [1]]
+  %uh2 = shard.update_halo %uh1 on @grid0 split_axes = [[0], [1]]
     halo_sizes = [2, 2, %c2, 2] : memref<12x12xi8>
   return
 }
diff --git a/mlir/test/Dialect/Shard/partition.mlir b/mlir/test/Dialect/Shard/partition.mlir
new file mode 100644
index 0000000..c2572cc
--- /dev/null
+++ b/mlir/test/Dialect/Shard/partition.mlir
@@ -0,0 +1,317 @@
+// RUN: mlir-opt \
+// RUN:   --pass-pipeline="builtin.module(func.func(shard-partition,test-single-fold))" \
+// RUN:   %s | FileCheck %s
+
+shard.grid @grid_1d(shape = 2)
+
+// CHECK-LABEL: func @return_sharding
+func.func @return_sharding(
+  // CHECK-SAME: [[ARG:%.*]]: tensor<1xf32>
+  %arg0: tensor<2xf32>
+// CHECK-SAME: ) -> (tensor<1xf32>, !shard.sharding) {
+) -> (tensor<2xf32>, !shard.sharding) {
+  %ssharded = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %sharded = shard.shard %arg0 to %ssharded  : tensor<2xf32>
+  // CHECK-NEXT: [[vsharding:%.*]] = shard.sharding @grid_1d split_axes = {{\[\[}}0]] : !shard.sharding
+  %r = shard.get_sharding %sharded : tensor<2xf32> -> !shard.sharding
+  // CHECK-NEXT: return [[ARG]], [[vsharding]] : tensor<1xf32>, !shard.sharding
+  return %sharded, %r : tensor<2xf32>, !shard.sharding
+}
+
+// CHECK-LABEL: func @full_replication
+func.func @full_replication(
+  // CHECK-SAME: %[[ARG:.*]]: tensor<2xi8>
+  %arg0: tensor<2xi8>
+// CHECK-SAME: -> tensor<2xi8> {
+) -> tensor<2xi8> {
+  %s0 = shard.sharding @grid_1d split_axes = [[]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0  : tensor<2xi8>
+  %s1 = shard.sharding @grid_1d split_axes = [[]] : !shard.sharding
+  %1 = shard.shard %0 to %s1  annotate_for_users : tensor<2xi8>
+  // CHECK: return %[[ARG]] : tensor<2xi8>
+  return %1 : tensor<2xi8>
+}
+
+// CHECK-LABEL: func @sharding_triplet
+func.func @sharding_triplet(
+  // CHECK-SAME: %[[ARG:.*]]: tensor<1xf32>
+  %arg0: tensor<2xf32>
+// CHECK-SAME: ) -> tensor<2xf32> {
+) -> tensor<2xf32> {
+  // CHECK: %[[ALL_GATHER:.*]] = shard.all_gather %[[ARG]] on @grid_1d grid_axes = [0] gather_axis = 0 : tensor<1xf32> -> tensor<2xf32>
+  %ssharded = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %sharded = shard.shard %arg0 to %ssharded  : tensor<2xf32>
+  %ssharded_0 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %sharded_0 = shard.shard %sharded to %ssharded_0  annotate_for_users : tensor<2xf32>
+  %ssharded_1 = shard.sharding @grid_1d split_axes = [[]] : !shard.sharding
+  %sharded_1 = shard.shard %sharded_0 to %ssharded_1  : tensor<2xf32>
+  // CHECK: return %[[ALL_GATHER]] : tensor<2xf32>
+  return %sharded_1 : tensor<2xf32>
+}
+
+
+// CHECK-LABEL: func @move_split_axis
+func.func @move_split_axis(
+  // CHECK-SAME: %[[ARG:.*]]: tensor<1x2xi8>
+  %arg0: tensor<2x2xi8>
+// CHECK-SAME: -> tensor<2x1xi8> {
+) -> tensor<2x2xi8> {
+  // CHECK: %[[ALL_TO_ALL:.*]] = shard.all_to_all %[[ARG]] on @grid_1d
+  // CHECK-SAME: grid_axes = [0] split_axis = 1 concat_axis = 0 : tensor<1x2xi8> -> tensor<2x1xi8>
+  %s0 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0  : tensor<2x2xi8>
+  %s1 = shard.sharding @grid_1d split_axes = [[], [0]] : !shard.sharding
+  %1 = shard.shard %0 to %s1  annotate_for_users : tensor<2x2xi8>
+  // CHECK: return %[[ALL_TO_ALL]] : tensor<2x1xi8>
+  return %1 : tensor<2x2xi8>
+}
+
+// CHECK-LABEL: func @non_tensor_value
+func.func @non_tensor_value(
+  // CHECK-SAME: %[[ARG:.*]]: i8
+  %arg0: i8
+// CHECK-SAME: -> i8 {
+) -> i8 {
+  // CHECK: %[[RES:.*]] = arith.addi %[[ARG]], %[[ARG]] : i8
+  %0 = arith.addi %arg0, %arg0 : i8
+  // CHECK: return %[[RES]] : i8
+  return %0 : i8
+}
+
+// CHECK-LABEL: func @unary_elementwise
+func.func @unary_elementwise(
+  // CHECK-SAME: %[[ARG:.*]]: tensor<1xi8>
+  %arg0: tensor<2xi8>
+// CHECK-SAME: -> tensor<1xi8> {
+) -> tensor<2xi8> {
+  %s0 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0  : tensor<2xi8>
+  %s1 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %1 = shard.shard %0 to %s1  annotate_for_users : tensor<2xi8>
+  // CHECK: %[[RES:.*]] = tosa.abs %[[ARG]] : (tensor<1xi8>) -> tensor<1xi8>
+  %2 = tosa.abs %1 : (tensor<2xi8>) -> tensor<2xi8>
+  %s3 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %3 = shard.shard %2 to %s3  : tensor<2xi8>
+  %s4 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %4 = shard.shard %3 to %s4  annotate_for_users : tensor<2xi8>
+  // CHECK: return %[[RES]] : tensor<1xi8>
+  return %4 : tensor<2xi8>
+}
+
+// full replication -> shard axis -> abs -> shard axis -> full replication
+// CHECK-LABEL: func @unary_elementwise_with_resharding
+func.func @unary_elementwise_with_resharding(
+  // CHECK-SAME: %[[ARG:.*]]: tensor<2xi8>
+  %arg0: tensor<2xi8>
+// CHECK-SAME: -> tensor<2xi8> {
+) -> tensor<2xi8> {
+  // CHECK: %[[SLICE:.*]] = shard.all_slice %[[ARG]] on @grid_1d grid_axes = [0] slice_axis = 0
+  // CHECK-SAME: tensor<2xi8> -> tensor<1xi8>
+  %s0 = shard.sharding @grid_1d split_axes = [[]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0  : tensor<2xi8>
+  %s1 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %1 = shard.shard %0 to %s1  annotate_for_users : tensor<2xi8>
+  // CHECK: %[[ABS:.*]] = tosa.abs %[[SLICE]] : (tensor<1xi8>) -> tensor<1xi8>
+  %2 = tosa.abs %1 : (tensor<2xi8>) -> tensor<2xi8>
+  // CHECK: %[[RES:.*]] = shard.all_gather %[[ABS]] on @grid_1d
+  // CHECK-SAME: grid_axes = [0] gather_axis = 0 : tensor<1xi8> -> tensor<2xi8>
+  %s3 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %3 = shard.shard %2 to %s3  : tensor<2xi8>
+  %s4 = shard.sharding @grid_1d split_axes = [[]] : !shard.sharding
+  %4 = shard.shard %3 to %s4  annotate_for_users : tensor<2xi8>
+  // CHECK: return %[[RES]] : tensor<2xi8>
+  return %4 : tensor<2xi8>
+}
+
+// CHECK-LABEL: func @binary_elementwise
+func.func @binary_elementwise(
+  // CHECK-SAME: %[[ARG0:.*]]: tensor<1xi8>,
+  %arg0: tensor<2xi8>,
+  // CHECK-SAME: %[[ARG1:.*]]: tensor<1xi8>
+  %arg1: tensor<2xi8>
+// CHECK-SAME: -> tensor<1xi8> {
+) -> tensor<2xi8> {
+  %sarg0_sharded = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %arg0_sharded = shard.shard %arg0 to %sarg0_sharded  : tensor<2xi8>
+  %sop_arg0 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %op_arg0 = shard.shard %arg0_sharded to %sop_arg0  annotate_for_users : tensor<2xi8>
+  %sarg1_sharded = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %arg1_sharded = shard.shard %arg1 to %sarg1_sharded  : tensor<2xi8>
+  %sop_arg1 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %op_arg1 = shard.shard %arg1_sharded to %sop_arg1  annotate_for_users : tensor<2xi8>
+  // CHECK: %[[RES:.*]] = tosa.add %[[ARG0]], %[[ARG1]] : (tensor<1xi8>, tensor<1xi8>) -> tensor<1xi8>
+  %op_res = tosa.add %op_arg0, %op_arg1 : (tensor<2xi8>, tensor<2xi8>) -> tensor<2xi8>
+  %sop_res_sharded = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %op_res_sharded = shard.shard %op_res to %sop_res_sharded  : tensor<2xi8>
+  %sres = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %res = shard.shard %op_res_sharded to %sres  annotate_for_users : tensor<2xi8>
+  // CHECK: return %[[RES]] : tensor<1xi8>
+  return %res : tensor<2xi8>
+}
+
+// reshard
+// abs
+// reshard
+// abs
+// reshard
+// CHECK-LABEL: func @multiple_chained_ops
+func.func @multiple_chained_ops(
+  // CHECK-SAME: %[[ARG:.*]]: tensor<2xi8>
+  %arg0: tensor<2xi8>
+// CHECK-SAME: -> tensor<1xi8> {
+) -> tensor<2xi8> {
+  // CHECK: %[[RESHARD1:.*]] = shard.all_slice %[[ARG]] on @grid_1d grid_axes = [0] slice_axis = 0
+  // CHECK-SAME: tensor<2xi8> -> tensor<1xi8>
+  %s0 = shard.sharding @grid_1d split_axes = [[]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0  : tensor<2xi8>
+  %s1 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %1 = shard.shard %0 to %s1  annotate_for_users : tensor<2xi8>
+  // CHECK: %[[ABS1:.*]] = tosa.abs %[[RESHARD1]] : (tensor<1xi8>) -> tensor<1xi8>
+  %2 = tosa.abs %1 : (tensor<2xi8>) -> tensor<2xi8>
+  // CHECK: %[[RESHARD2:.*]] = shard.all_gather %[[ABS1]] on @grid_1d
+  // CHECK-SAME: grid_axes = [0] gather_axis = 0 : tensor<1xi8> -> tensor<2xi8>
+  %s3 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %3 = shard.shard %2 to %s3  : tensor<2xi8>
+  %s4 = shard.sharding @grid_1d split_axes = [[]] : !shard.sharding
+  %4 = shard.shard %3 to %s4  annotate_for_users : tensor<2xi8>
+  // CHECK: %[[ABS2:.*]] = tosa.abs %[[RESHARD2]] : (tensor<2xi8>) -> tensor<2xi8>
+  %5 = tosa.abs %4 : (tensor<2xi8>) -> tensor<2xi8>
+  // CHECK: %[[RESHARD3:.*]] = shard.all_slice %[[ABS2]] on @grid_1d grid_axes = [0] slice_axis = 0 :
+  // CHECK-SAME: tensor<2xi8> -> tensor<1xi8>
+  %s6 = shard.sharding @grid_1d split_axes = [[]] : !shard.sharding
+  %6 = shard.shard %5 to %s6  : tensor<2xi8>
+  %s7 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %7 = shard.shard %6 to %s7  annotate_for_users : tensor<2xi8>
+  // CHECK: return %[[RESHARD3]] : tensor<1xi8>
+  return %7 : tensor<2xi8>
+}
+
+// CHECK-LABEL: func @incomplete_sharding
+func.func @incomplete_sharding(
+  // CHECK-SAME: %[[ARG:.*]]: tensor<4x16xf32>
+  %arg0: tensor<8x16xf32>
+// CHECK-SAME: -> tensor<4x16xf32> {
+) -> tensor<8x16xf32> {
+  %s0 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0  annotate_for_users : tensor<8x16xf32>
+  // CHECK: %[[RES:.*]] = tosa.sigmoid %[[ARG]] : (tensor<4x16xf32>) -> tensor<4x16xf32>
+  %1 = tosa.sigmoid %0 : (tensor<8x16xf32>) -> tensor<8x16xf32>
+  %s2 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %2 = shard.shard %1 to %s2  : tensor<8x16xf32>
+  // CHECK: return %[[RES]] : tensor<4x16xf32>
+  return %2 : tensor<8x16xf32>
+}
+
+shard.grid @grid_1d_4(shape = 4)
+
+// CHECK-LABEL: func @ew_chain_with_halo
+func.func @ew_chain_with_halo(
+  // CHECK-SAME: %[[IN1:[A-Za-z0-9_]+]]: tensor<5x16xf32>
+  %arg0: tensor<8x16xf32>,
+  // CHECK-SAME: %[[IN2:[A-Za-z0-9_]+]]: tensor<1xf32>
+  %arg1: tensor<1xf32>,
+  // CHECK-SAME: %[[IN3:[A-Za-z0-9_]+]]: tensor<1xf32>
+  %arg2: tensor<1xf32>)
+  // CHECK-SAME: -> tensor<5x16xf32>
+   -> tensor<8x16xf32> {
+  %ssharded = shard.sharding @grid_1d_4 split_axes = [[0]] halo_sizes = [2, 1] : !shard.sharding
+  %sharded = shard.shard %arg0 to %ssharded  annotate_for_users : tensor<8x16xf32>
+  // CHECK: %[[TMP1:.*]] = tosa.tanh %[[IN1]] : (tensor<5x16xf32>) -> tensor<5x16xf32>
+  %0 = tosa.tanh %sharded : (tensor<8x16xf32>) -> tensor<8x16xf32>
+  %ssharded_0 = shard.sharding @grid_1d_4 split_axes = [[0]] halo_sizes = [2, 1] : !shard.sharding
+  %sharded_0 = shard.shard %0 to %ssharded_0  : tensor<8x16xf32>
+  %ssharded_1 = shard.sharding @grid_1d_4 split_axes = [[0]] halo_sizes = [2, 1] : !shard.sharding
+  %sharded_1 = shard.shard %sharded_0 to %ssharded_1  annotate_for_users : tensor<8x16xf32>
+  // CHECK-NEXT: %[[TMP2:.*]] = tosa.abs %[[TMP1]] : (tensor<5x16xf32>) -> tensor<5x16xf32>
+  %1 = tosa.abs %sharded_1 : (tensor<8x16xf32>) -> tensor<8x16xf32>
+  %ssharded_2 = shard.sharding @grid_1d_4 split_axes = [[0]] halo_sizes = [2, 1] : !shard.sharding
+  %sharded_2 = shard.shard %1 to %ssharded_2  : tensor<8x16xf32>
+  %ssharded_4 = shard.sharding @grid_1d_4 split_axes = [[0]] halo_sizes = [2, 1] : !shard.sharding
+  %sharded_4 = shard.shard %sharded_2 to %ssharded_4  annotate_for_users : tensor<8x16xf32>
+  // CHECK-NEXT: %[[TMP3:.*]] = tosa.negate %[[TMP2]], %[[IN2]], %[[IN3]] : (tensor<5x16xf32>, tensor<1xf32>, tensor<1xf32>) -> tensor<5x16xf32>
+  %sharding_1 = shard.sharding @grid_1d_4 split_axes = [[]] : !shard.sharding
+  %zero_point_1 = shard.shard %arg1 to %sharding_1 annotate_for_users : tensor<1xf32>
+  %zero_point_2 = shard.shard %arg2 to %sharding_1 annotate_for_users : tensor<1xf32>
+  %2 = tosa.negate %sharded_4, %zero_point_1, %zero_point_2 : (tensor<8x16xf32>, tensor<1xf32>, tensor<1xf32>) -> tensor<8x16xf32>
+  %ssharded_5 = shard.sharding @grid_1d_4 split_axes = [[0]] halo_sizes = [2, 1] : !shard.sharding
+  %sharded_5 = shard.shard %2 to %ssharded_5  : tensor<8x16xf32>
+  %ssharded_6 = shard.sharding @grid_1d_4 split_axes = [[0]] halo_sizes = [2, 1] : !shard.sharding
+  %sharded_6 = shard.shard %sharded_5 to %ssharded_6  annotate_for_users : tensor<8x16xf32>
+  // CHECK-NEXT: return %[[TMP3]] : tensor<5x16xf32>
+  return %sharded_6 : tensor<8x16xf32>
+}
+
+// CHECK-LABEL: func @test_shard_update_halo
+// CHECK-SAME: %[[IN1:[A-Za-z0-9_]+]]: tensor<300x1200xi64>
+func.func @test_shard_update_halo(%arg0: tensor<1200x1200xi64>) -> tensor<1200x1200xi64> {
+  %sharding = shard.sharding @grid_1d_4 split_axes = [[0]] : !shard.sharding
+  // CHECK: %[[T:.*]] = tensor.empty() : tensor<304x1200xi64>
+  // CHECK: %[[inserted_slice:.*]] = tensor.insert_slice %[[IN1]] into %[[T]][2, 0] [300, 1200] [1, 1] : tensor<300x1200xi64> into tensor<304x1200xi64>
+  // CHECK: %[[UH:.*]] = shard.update_halo %[[inserted_slice]] on @grid_1d_4 split_axes = {{\[\[0]]}} halo_sizes = [2, 2] : tensor<304x1200xi64>
+  %sharded = shard.shard %arg0 to %sharding : tensor<1200x1200xi64>
+  %sharding_0 = shard.sharding @grid_1d_4 split_axes = [[0]] halo_sizes = [2, 2] : !shard.sharding
+  %sharded_1 = shard.shard %sharded to %sharding_0 : tensor<1200x1200xi64>
+  %sharded_3 = shard.shard %sharded_1 to %sharding_0 annotate_for_users : tensor<1200x1200xi64>
+  // CHECK: return %[[UH]] : tensor<304x1200xi64>
+  return %sharded_3 : tensor<1200x1200xi64>
+}
+
+shard.grid @grid4x4(shape = 4x4)
+// CHECK-LABEL: func @test_shard_update_halo2d
+// CHECK-SAME: %[[IN1:[A-Za-z0-9_]+]]: tensor<300x300xi64>
+func.func @test_shard_update_halo2d(%arg0: tensor<1200x1200xi64>) -> tensor<1200x1200xi64> {
+  %sharding = shard.sharding @grid4x4 split_axes = [[0], [1]] : !shard.sharding
+  // CHECK: %[[T:.*]] = tensor.empty() : tensor<303x307xi64>
+  // CHECK: %[[inserted_slice:.*]] = tensor.insert_slice %[[IN1]] into %[[T]][1, 3] [300, 300] [1, 1] : tensor<300x300xi64> into tensor<303x307xi64>
+  // CHECK: %[[UH:.*]] = shard.update_halo %[[inserted_slice]] on @grid4x4 split_axes = {{\[\[}}0], [1]] halo_sizes = [1, 2, 3, 4] : tensor<303x307xi64>
+  %sharded = shard.shard %arg0 to %sharding : tensor<1200x1200xi64>
+  %sharding_0 = shard.sharding @grid4x4 split_axes = [[0], [1]] halo_sizes = [1, 2, 3, 4] : !shard.sharding
+  %sharded_1 = shard.shard %sharded to %sharding_0 : tensor<1200x1200xi64>
+  %sharded_3 = shard.shard %sharded_1 to %sharding_0 annotate_for_users : tensor<1200x1200xi64>
+  // CHECK: return %[[UH]] : tensor<303x307xi64>
+  return %sharded_3 : tensor<1200x1200xi64>
+}
+
+shard.grid @grid(shape = 2)
+// CHECK-LABEL: func.func @test_reduce_0d(
+// CHECK-SAME: %[[ARG0:[A-Za-z0-9_]+]]: tensor<3x6xi32>
+func.func @test_reduce_0d(%arg0: tensor<6x6xi32>) -> (tensor<i32>) {
+  %sharding = shard.sharding @grid split_axes = [[0]] : !shard.sharding
+  %sharded = shard.shard %arg0 to %sharding annotate_for_users : tensor<6x6xi32>
+  %4 = tensor.empty() : tensor<i32>
+  %sharding_out = shard.sharding @grid split_axes = [[]] : !shard.sharding
+  %sharded_out = shard.shard %4 to %sharding_out : tensor<i32>
+  %sharded_in = shard.shard %sharded to %sharding annotate_for_users : tensor<6x6xi32>
+  // CHECK: %[[reduced:.*]] = linalg.reduce ins(%arg0 : tensor<3x6xi32>)
+  %reduced = linalg.reduce ins(%sharded_in : tensor<6x6xi32>) outs(%sharded_out : tensor<i32>) dimensions = [0, 1] 
+    (%in: i32, %init: i32) {
+      %6 = arith.addi %in, %init : i32
+      linalg.yield %6 : i32
+    }
+  // CHECK: %[[all_reduce:.*]] = shard.all_reduce %[[reduced]] on @grid grid_axes = [0] : tensor<i32> -> tensor<i32>
+  %sharded_red = shard.shard %reduced to %sharding_out : tensor<i32>
+  %sharded_ret = shard.shard %sharded_red to %sharding_out annotate_for_users : tensor<i32>
+  // CHECK: return %[[all_reduce]] : tensor<i32>
+  return %sharded_ret : tensor<i32>
+}
+
+// CHECK-LABEL: func.func @test_reduce_1d(
+// CHECK-SAME: %[[ARG0:[A-Za-z0-9_]+]]: tensor<3x6xi32>
+func.func @test_reduce_1d(%arg0: tensor<6x6xi32>) -> (tensor<6xi32>) {
+  %sharding = shard.sharding @grid split_axes = [[0]] : !shard.sharding
+  %sharded = shard.shard %arg0 to %sharding annotate_for_users : tensor<6x6xi32>
+  %4 = tensor.empty() : tensor<6xi32>
+  %sharded_out = shard.shard %4 to %sharding : tensor<6xi32>
+  %sharded_in = shard.shard %sharded to %sharding annotate_for_users : tensor<6x6xi32>
+  // CHECK: %[[reduced:.*]] = linalg.reduce ins(%arg0 : tensor<3x6xi32>)
+  %reduced = linalg.reduce ins(%sharded_in : tensor<6x6xi32>) outs(%sharded_out : tensor<6xi32>) dimensions = [1] 
+    (%in: i32, %init: i32) {
+      %6 = arith.addi %in, %init : i32
+      linalg.yield %6 : i32
+    }
+  // CHECK-NOT: shard.all_reduce
+  %sharded_red = shard.shard %reduced to %sharding : tensor<6xi32>
+  %sharded_ret = shard.shard %sharded_red to %sharding annotate_for_users : tensor<6xi32>
+  // CHECK: return %[[reduced]] : tensor<3xi32>
+  return %sharded_ret : tensor<6xi32>
+}
diff --git a/mlir/test/Dialect/Shard/process-multi-index-op-lowering.mlir b/mlir/test/Dialect/Shard/process-multi-index-op-lowering.mlir
new file mode 100644
index 0000000..33c7a8f
--- /dev/null
+++ b/mlir/test/Dialect/Shard/process-multi-index-op-lowering.mlir
@@ -0,0 +1,23 @@
+// RUN: mlir-opt -test-grid-process-multi-index-op-lowering %s | FileCheck %s
+
+shard.grid @grid2d(shape = ?x?)
+
+// CHECK-LABEL: func.func @multi_index_2d_grid
+func.func @multi_index_2d_grid() -> (index, index) {
+  // CHECK: %[[LINEAR_IDX:.*]] = shard.process_linear_index on @grid2d : index
+  // CHECK: %[[SHARD_SHAPE:.*]]:2 = shard.grid_shape @grid2d : index, index
+  // CHECK: %[[MULTI_IDX:.*]]:2 = affine.delinearize_index %[[LINEAR_IDX]] into (%[[SHARD_SHAPE]]#0, %[[SHARD_SHAPE]]#1) : index, index
+  %0:2 = shard.process_multi_index on @grid2d : index, index
+  // CHECK: return %[[MULTI_IDX]]#0, %[[MULTI_IDX]]#1 : index, index
+  return %0#0, %0#1 : index, index
+}
+
+// CHECK-LABEL: func.func @multi_index_2d_grid_single_inner_axis
+func.func @multi_index_2d_grid_single_inner_axis() -> index {
+  // CHECK: %[[LINEAR_IDX:.*]] = shard.process_linear_index on @grid2d : index
+  // CHECK: %[[SHARD_SHAPE:.*]]:2 = shard.grid_shape @grid2d : index, index
+  // CHECK: %[[MULTI_IDX:.*]]:2 = affine.delinearize_index %[[LINEAR_IDX]] into (%[[SHARD_SHAPE]]#0, %[[SHARD_SHAPE]]#1) : index, index
+  %0 = shard.process_multi_index on @grid2d axes = [0] : index
+  // CHECK: return %[[MULTI_IDX]]#0 : index
+  return %0 : index
+}
diff --git a/mlir/test/Dialect/Shard/resharding-partition.mlir b/mlir/test/Dialect/Shard/resharding-partition.mlir
new file mode 100644
index 0000000..ff9e840
--- /dev/null
+++ b/mlir/test/Dialect/Shard/resharding-partition.mlir
@@ -0,0 +1,168 @@
+// RUN: mlir-opt -test-grid-resharding-partition %s | FileCheck %s
+
+shard.grid @grid_1d(shape = 2)
+shard.grid @grid_1d_dynamic(shape = ?)
+
+// CHECK-LABEL: func @same_source_and_target_sharding
+func.func @same_source_and_target_sharding(
+  // CHECK-SAME: %[[ARG:.*]]: tensor<2xf32>
+  %arg0: tensor<2xf32>
+) -> tensor<2xf32> {
+  %s0 = shard.sharding @grid_1d split_axes = [[]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0 : tensor<2xf32>
+  %s1 = shard.sharding @grid_1d split_axes = [[]] : !shard.sharding
+  %1 = shard.shard %0 to %s1 annotate_for_users : tensor<2xf32>
+  // CHECK: return %[[ARG]]
+  return %1 : tensor<2xf32>
+}
+
+// CHECK-LABEL: func @identical_source_and_target_sharding
+func.func @identical_source_and_target_sharding(
+  // CHECK-SAME: %[[ARG:.*]]: tensor<2xf32>
+  %arg0: tensor<2xf32>
+) -> tensor<2xf32> {
+  %s0 = shard.sharding @grid_1d split_axes = [[]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0 : tensor<2xf32>
+  %1 = shard.shard %0 to %s0 annotate_for_users : tensor<2xf32>
+  // CHECK: return %[[ARG]]
+  return %1 : tensor<2xf32>
+}
+
+// CHECK-LABEL: func @split_replicated_tensor_axis
+func.func @split_replicated_tensor_axis(
+  // CHECK-SAME: %[[ARG:.*]]: tensor<3x14xf32>
+  %arg0: tensor<3x14xf32>
+) -> tensor<3x14xf32> {
+  // CHECK: %[[ALL_SLICE:.*]] = shard.all_slice %[[ARG]] on @grid_1d grid_axes = [0] slice_axis = 1
+  // CHECK-SAME: tensor<3x14xf32> -> tensor<3x7xf32>
+  // CHECK: %[[RESULT:.*]] = builtin.unrealized_conversion_cast %[[ALL_SLICE]] : tensor<3x7xf32> to tensor<3x14xf32>
+  %s0 = shard.sharding @grid_1d split_axes = [[]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0 : tensor<3x14xf32>
+  %s1 = shard.sharding @grid_1d split_axes = [[], [0]] : !shard.sharding
+  %1 = shard.shard %0 to %s1 annotate_for_users : tensor<3x14xf32>
+  // CHECK: return %[[RESULT]] : tensor<3x14xf32>
+  return %1 : tensor<3x14xf32>
+}
+
+// CHECK-LABEL: func @split_replicated_tensor_axis_dynamic
+func.func @split_replicated_tensor_axis_dynamic(
+  // CHECK-SAME: %[[ARG:.*]]: tensor<?x3x?xf32>
+  %arg0: tensor<?x3x?xf32>
+) -> tensor<?x3x?xf32> {
+  // CHECK: %[[RESULT:.*]] = shard.all_slice %[[ARG]] on @grid_1d_dynamic grid_axes = [0] slice_axis = 0
+  // CHECK-SAME: tensor<?x3x?xf32> -> tensor<?x3x?xf32>
+  %s0 = shard.sharding @grid_1d_dynamic split_axes = [[], [], []] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0 : tensor<?x3x?xf32>
+  %s1 = shard.sharding @grid_1d_dynamic split_axes = [[0]] : !shard.sharding
+  %1 = shard.shard %0 to %s1 annotate_for_users : tensor<?x3x?xf32>
+  // CHECK: return %[[RESULT]] : tensor<?x3x?xf32>
+  return %1 : tensor<?x3x?xf32>
+}
+
+// CHECK-LABEL: func @move_split_axis
+func.func @move_split_axis(
+  // CHECK-SAME: %[[ARG:.*]]: tensor<10x14xf32>
+  %arg0: tensor<10x14xf32>
+) -> tensor<10x14xf32> {
+  // CHECK: %[[SOURCE_SHARD:.*]] = builtin.unrealized_conversion_cast %[[ARG]] : tensor<10x14xf32> to tensor<5x14xf32>
+  // CHECK: %[[TARGET_SHARD:.*]] = shard.all_to_all %[[SOURCE_SHARD]] on @grid_1d grid_axes = [0] split_axis = 1 concat_axis = 0 : tensor<5x14xf32> -> tensor<10x7xf32>
+  // CHECK: %[[RES:.*]] = builtin.unrealized_conversion_cast %[[TARGET_SHARD]] : tensor<10x7xf32> to tensor<10x14xf32>
+  %s0 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0 : tensor<10x14xf32>
+  %s1 = shard.sharding @grid_1d split_axes = [[], [0]] : !shard.sharding
+  %1 = shard.shard %0 to %s1 annotate_for_users : tensor<10x14xf32>
+  // CHECK: return %[[RES]] : tensor<10x14xf32>
+  return %1 : tensor<10x14xf32>
+}
+
+// CHECK-LABEL: func @move_split_axis_dynamic_grid
+func.func @move_split_axis_dynamic_grid(
+  // CHECK-SAME: %[[ARG:.*]]: tensor<10x14xf32>
+  %arg0: tensor<10x14xf32>
+) -> tensor<10x14xf32> {
+  // CHECK: %[[SOURCE_SHARD:.*]] = builtin.unrealized_conversion_cast %[[ARG]] : tensor<10x14xf32> to tensor<?x14xf32>
+  // CHECK: %[[ALL_TO_ALL:.*]] = shard.all_to_all %[[SOURCE_SHARD]] on @grid_1d_dynamic grid_axes = [0] split_axis = 1 concat_axis = 0 : tensor<?x14xf32> -> tensor<?x?xf32>
+  // CHECK: %[[TARGET_SHARD:.*]] = tensor.cast %[[ALL_TO_ALL]] : tensor<?x?xf32> to tensor<10x?xf32>
+  // CHECK: %[[RES:.*]] = builtin.unrealized_conversion_cast %[[TARGET_SHARD]] : tensor<10x?xf32> to tensor<10x14xf32>
+  %s0 = shard.sharding @grid_1d_dynamic split_axes = [[0]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0 : tensor<10x14xf32>
+  %s1 = shard.sharding @grid_1d_dynamic split_axes = [[], [0]] : !shard.sharding
+  %1 = shard.shard %0 to %s1 annotate_for_users : tensor<10x14xf32>
+  // CHECK: return %[[RES]] : tensor<10x14xf32>
+  return %1 : tensor<10x14xf32>
+}
+
+// CHECK-LABEL: func @move_split_dynamic_axis
+func.func @move_split_dynamic_axis(
+  // CHECK-SAME: %[[ARG:.*]]: tensor<?x14xf32>
+  %arg0: tensor<?x14xf32>
+) -> tensor<?x14xf32> {
+  // CHECK: %[[TARGET_SHARD:.*]] = shard.all_to_all %[[ARG]] on @grid_1d grid_axes = [0] split_axis = 1 concat_axis = 0 : tensor<?x14xf32> -> tensor<?x7xf32>
+  // CHECK: %[[RES:.*]] = builtin.unrealized_conversion_cast %[[TARGET_SHARD]] : tensor<?x7xf32> to tensor<?x14xf32>
+  %s0 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0 : tensor<?x14xf32>
+  %s1 = shard.sharding @grid_1d split_axes = [[], [0]] : !shard.sharding
+  %1 = shard.shard %0 to %s1 annotate_for_users : tensor<?x14xf32>
+  // CHECK: return %[[RES]] : tensor<?x14xf32>
+  return %1 : tensor<?x14xf32>
+}
+
+// CHECK-LABEL: func @unshard_static_axis
+func.func @unshard_static_axis(
+  // CHECK-SAME: %[[ARG:.*]]: tensor<10x14xf32>
+  %arg0: tensor<10x14xf32>
+) -> tensor<10x14xf32> {
+  // CHECK: %[[SOURCE_SHARD:.*]] = builtin.unrealized_conversion_cast %[[ARG]] : tensor<10x14xf32> to tensor<5x14xf32>
+  // CHECK: %[[ALL_GATHER:.*]] = shard.all_gather %[[SOURCE_SHARD]] on @grid_1d grid_axes = [0] gather_axis = 0 : tensor<5x14xf32> -> tensor<10x14xf32>
+  %s0 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0 : tensor<10x14xf32>
+  %s1 = shard.sharding @grid_1d split_axes = [[]] : !shard.sharding
+  %1 = shard.shard %0 to %s1 annotate_for_users : tensor<10x14xf32>
+  // CHECK: return %[[ALL_GATHER]] : tensor<10x14xf32>
+  return %1 : tensor<10x14xf32>
+}
+
+// CHECK-LABEL: func @unshard_static_last_axis
+func.func @unshard_static_last_axis(
+  // CHECK-SAME: %[[ARG:.*]]: tensor<10x14xf32>
+  %arg0: tensor<10x14xf32>
+) -> tensor<10x14xf32> {
+  // CHECK: %[[SOURCE_SHARD:.*]] = builtin.unrealized_conversion_cast %[[ARG]] : tensor<10x14xf32> to tensor<10x7xf32>
+  // CHECK: %[[ALL_GATHER:.*]] = shard.all_gather %[[SOURCE_SHARD]] on @grid_1d grid_axes = [0] gather_axis = 1 : tensor<10x7xf32> -> tensor<10x14xf32>
+  %s0 = shard.sharding @grid_1d split_axes = [[], [0]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0 : tensor<10x14xf32>
+  %s1 = shard.sharding @grid_1d split_axes = [[], []] : !shard.sharding
+  %1 = shard.shard %0 to %s1 annotate_for_users : tensor<10x14xf32>
+  // CHECK: return %[[ALL_GATHER]] : tensor<10x14xf32>
+  return %1 : tensor<10x14xf32>
+}
+
+// CHECK-LABEL: func @unshard_dynamic_axis
+func.func @unshard_dynamic_axis(
+  // CHECK-SAME: %[[ARG:.*]]: tensor<?x14xf32>
+  %arg0: tensor<?x14xf32>
+) -> tensor<?x14xf32> {
+  // CHECK: %[[ALL_GATHER:.*]] = shard.all_gather %[[ARG]] on @grid_1d grid_axes = [0] gather_axis = 0 : tensor<?x14xf32> -> tensor<?x14xf32>
+  %s0 = shard.sharding @grid_1d split_axes = [[0]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0 : tensor<?x14xf32>
+  %s1 = shard.sharding @grid_1d split_axes = [[]] : !shard.sharding
+  %1 = shard.shard %0 to %s1 annotate_for_users : tensor<?x14xf32>
+  // CHECK: return %[[ALL_GATHER]] : tensor<?x14xf32>
+  return %1 : tensor<?x14xf32>
+}
+
+// CHECK-LABEL: func @unshard_static_axis_on_dynamic_grid_axis
+func.func @unshard_static_axis_on_dynamic_grid_axis(
+// CHECK-SAME: %[[ARG:.*]]: tensor<10x14xf32>  
+  %arg0: tensor<10x14xf32>
+) -> tensor<10x14xf32> {
+  // CHECK: %[[SOURCE_SHARD:.*]] = builtin.unrealized_conversion_cast %[[ARG]] : tensor<10x14xf32> to tensor<?x14xf32>
+  // CHECK: %[[ALL_GATHER:.*]] = shard.all_gather %[[SOURCE_SHARD]] on @grid_1d_dynamic grid_axes = [0] gather_axis = 0 : tensor<?x14xf32> -> tensor<?x14xf32>
+  // CHECK: %[[RES:.*]] = tensor.cast %[[ALL_GATHER]] : tensor<?x14xf32> to tensor<10x14xf32>
+  %s0 = shard.sharding @grid_1d_dynamic split_axes = [[0]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0 : tensor<10x14xf32>
+  %s1 = shard.sharding @grid_1d_dynamic split_axes = [[]] : !shard.sharding
+  %1 = shard.shard %0 to %s1 annotate_for_users : tensor<10x14xf32>
+  // CHECK: return %[[RES]] : tensor<10x14xf32>
+  return %1 : tensor<10x14xf32>
+}
diff --git a/mlir/test/Dialect/Mesh/sharding-propagation-failed.mlir b/mlir/test/Dialect/Shard/sharding-propagation-failed.mlir
index b5eb98d..b5eb98d 100644
--- a/mlir/test/Dialect/Mesh/sharding-propagation-failed.mlir
+++ b/mlir/test/Dialect/Shard/sharding-propagation-failed.mlir
diff --git a/mlir/test/Dialect/Shard/sharding-propagation.mlir b/mlir/test/Dialect/Shard/sharding-propagation.mlir
new file mode 100644
index 0000000..34aaf05
--- /dev/null
+++ b/mlir/test/Dialect/Shard/sharding-propagation.mlir
@@ -0,0 +1,301 @@
+// RUN: mlir-opt --pass-pipeline="builtin.module(func.func(sharding-propagation,cse))" %s | FileCheck %s
+
+shard.grid @grid_2(shape = 2)
+shard.grid @grid_1d(shape = ?)
+shard.grid @grid_2d(shape = 2x4)
+shard.grid @grid_3d(shape = ?x?x?)
+
+// CHECK-LABEL: func.func @element_wise_empty_sharding_info
+func.func @element_wise_empty_sharding_info(%arg0: tensor<8x16xf32>) -> tensor<8x16xf32> {
+  // CHECK-NEXT: tosa.sigmoid
+  %0 = tosa.sigmoid %arg0 : (tensor<8x16xf32>) -> tensor<8x16xf32>
+  // CHECK-NEXT: return
+  return %0 : tensor<8x16xf32>
+}
+
+// CHECK-LABEL: func.func @element_wise_on_def
+// CHECK-SAME:    %[[ARG:.*]]: tensor<8x16xf32>
+func.func @element_wise_on_def(%arg0: tensor<8x16xf32>) -> tensor<8x16xf32> {
+  // CHECK-NEXT:  %[[S0:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}0], [1]] : !shard.sharding
+  // CHECK-NEXT:  %[[V0:.*]] = shard.shard %[[ARG]] to %[[S0]] annotate_for_users  : tensor<8x16xf32>
+  // CHECK-NEXT:  %[[V1:.*]] = tosa.sigmoid %[[V0]]
+  %0 = tosa.sigmoid %arg0 : (tensor<8x16xf32>) -> tensor<8x16xf32>
+  // CHECK-NEXT:  %[[V2:.*]] = shard.shard %[[V1]] to %[[S0]]  : tensor<8x16xf32>
+  %s1 = shard.sharding @grid_2d split_axes = [[0], [1]] : !shard.sharding
+  %1 = shard.shard %0 to %s1  : tensor<8x16xf32>
+  // CHECK-NEXT:  return %[[V2]]
+  return %1 : tensor<8x16xf32>
+}
+
+// CHECK-LABEL: func.func @element_wise_on_use
+// CHECK-SAME:    %[[ARG:.*]]: tensor<8x16xf32>
+func.func @element_wise_on_use(%arg0: tensor<8x16xf32>) -> tensor<8x16xf32> {
+  // CHECK-NEXT:  %[[S0:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}0], [1]] : !shard.sharding
+  // CHECK-NEXT:  %[[V0:.*]] = shard.shard %[[ARG]] to %[[S0]] annotate_for_users  : tensor<8x16xf32>
+  %s0 = shard.sharding @grid_2d split_axes = [[0], [1]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0 annotate_for_users  : tensor<8x16xf32>
+  // CHECK-NEXT:  %[[V1:.*]] = tosa.sigmoid %[[V0]]
+  %1 = tosa.sigmoid %0 : (tensor<8x16xf32>) -> tensor<8x16xf32>
+  // CHECK-NEXT:  %[[V2:.*]] = shard.shard %[[V1]] to %[[S0]]  : tensor<8x16xf32>
+  // CHECK-NEXT:  return %[[V2]]
+  return %1 : tensor<8x16xf32>
+}
+
+// CHECK-LABEL: func.func @element_wise_on_graph_output
+// CHECK-SAME:    %[[ARG:.*]]: tensor<8x16xf32>
+func.func @element_wise_on_graph_output(%arg0: tensor<8x16xf32>) -> tensor<8x16xf32> {
+  // CHECK-NEXT:  %[[S0:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}0], [1]] : !shard.sharding
+  // CHECK-NEXT:  %[[V0:.*]] = shard.shard %[[ARG]] to %[[S0]] annotate_for_users  : tensor<8x16xf32>
+  // CHECK-NEXT:  %[[V1:.*]] = tosa.sigmoid %[[V0]]
+  %0 = tosa.sigmoid %arg0 : (tensor<8x16xf32>) -> tensor<8x16xf32>
+  // CHECK-NEXT:  %[[V2:.*]] = shard.shard %[[V1]] to %[[S0]]  : tensor<8x16xf32>
+  // CHECK-NEXT:  %[[V3:.*]] = shard.shard %[[V2]] to %[[S0]] annotate_for_users  : tensor<8x16xf32>
+  %s1 = shard.sharding @grid_2d split_axes = [[0], [1]] : !shard.sharding
+  %1 = shard.shard %0 to %s1 annotate_for_users  : tensor<8x16xf32>
+  // CHECK-NEXT:  return %[[V3]]
+  return %1 : tensor<8x16xf32>
+}
+
+// CHECK-LABEL: func.func @element_wise_on_graph_input
+// CHECK-SAME:    %[[ARG:.*]]: tensor<8x16xf32>
+func.func @element_wise_on_graph_input(%arg0: tensor<8x16xf32>) -> tensor<8x16xf32> {
+  // CHECK-NEXT:  %[[S0:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}0], [1]] : !shard.sharding
+  // CHECK-NEXT:  %[[V0:.*]] = shard.shard %[[ARG]] to %[[S0]]  : tensor<8x16xf32>
+  // CHECK-NEXT:  %[[V1:.*]] = shard.shard %[[V0]] to %[[S0]] annotate_for_users  : tensor<8x16xf32>
+  %s0 = shard.sharding @grid_2d split_axes = [[0], [1]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0  : tensor<8x16xf32>
+  // CHECK-NEXT:  %[[V2:.*]] = tosa.sigmoid %[[V1]]
+  %1 = tosa.sigmoid %0 : (tensor<8x16xf32>) -> tensor<8x16xf32>
+  // CHECK-NEXT:  %[[V3:.*]] = shard.shard %[[V2]] to %[[S0]]  : tensor<8x16xf32>
+  // CHECK-NEXT:  return %[[V3]]
+  return %1 : tensor<8x16xf32>
+}
+
+// CHECK-LABEL: func.func @arrow_structure
+// CHECK-SAME:    %[[ARG:.*]]: tensor<8x16xf32>
+func.func @arrow_structure(%arg0: tensor<8x16xf32>, %arg1: tensor<1xf32>, %arg2: tensor<1xf32>) -> (tensor<8x16xf32>, tensor<8x16xf32>) {
+  // CHECK-NEXT:  %[[S1:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}0], [1]] : !shard.sharding
+  // CHECK-NEXT:  %[[V1:.*]] = shard.shard %[[ARG]] to %[[S1]] annotate_for_users  : tensor<8x16xf32>
+  // CHECK-NEXT:  %[[V2:.*]] = tosa.tanh %[[V1]]
+  // CHECK-NEXT:  %[[V3:.*]] = shard.shard %[[V2]] to %[[S1]]  : tensor<8x16xf32>
+  %0 = tosa.tanh %arg0 : (tensor<8x16xf32>) -> tensor<8x16xf32>
+  // CHECK-NEXT:  %[[V4:.*]] = shard.shard %[[V3]] to %[[S1]] annotate_for_users  : tensor<8x16xf32>
+  // CHECK-NEXT:  %[[V5:.*]] = tosa.abs %[[V4]]
+  // CHECK-NEXT:  %[[V6:.*]] = shard.shard %[[V5]] to %[[S1]]  : tensor<8x16xf32>
+  %1 = tosa.abs %0: (tensor<8x16xf32>) -> tensor<8x16xf32>
+  // CHECK-NEXT:  %[[S3:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}]] : !shard.sharding
+  // CHECK-NEXT:  %[[ZP1:.*]] = shard.shard %arg1 to %[[S3]] annotate_for_users : tensor<1xf32>
+  // CHECK-NEXT:  %[[ZP2:.*]] = shard.shard %arg2 to %[[S3]] annotate_for_users : tensor<1xf32>
+  // CHECK-NEXT:  %[[V7:.*]] = tosa.negate %[[V4]], %[[ZP1]], %[[ZP2]]
+  // CHECK-NEXT:  %[[V8:.*]] = shard.shard %[[V7]] to %[[S1]]  : tensor<8x16xf32>
+  %2 = tosa.negate %0, %arg1, %arg2 : (tensor<8x16xf32>, tensor<1xf32>, tensor<1xf32>) -> tensor<8x16xf32>
+  %s3 = shard.sharding @grid_2d split_axes = [[0], [1]] : !shard.sharding
+  %3 = shard.shard %2 to %s3  : tensor<8x16xf32>
+  // CHECK-NEXT: return %[[V6]], %[[V8]]
+  return %1, %3 : tensor<8x16xf32>, tensor<8x16xf32>
+}
+
+// CHECK-LABEL: func.func @matmul_on_def_shard_batch_and_m
+// CHECK-SAME:     %[[ARG0:.*]]: tensor<2x16x8xf32>, %[[ARG1:.*]]: tensor<2x8x32xf32>, %[[ARG2:.*]]: tensor<1xf32>
+func.func @matmul_on_def_shard_batch_and_m(%arg0: tensor<2x16x8xf32>, %arg1: tensor<2x8x32xf32>, %arg2: tensor<1xf32>) -> tensor<2x16x32xf32> {
+  // CHECK-NEXT:  %[[S0:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}0], [1]] : !shard.sharding
+  // CHECK-NEXT:  %[[V0:.*]] = shard.shard %[[ARG0]] to %[[S0]] annotate_for_users  : tensor<2x16x8xf32>
+  // CHECK-NEXT:  %[[S1:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}0]] : !shard.sharding
+  // CHECK-NEXT:  %[[V1:.*]] = shard.shard %[[ARG1]] to %[[S1]] annotate_for_users  : tensor<2x8x32xf32>
+  // CHECK-NEXT:  %[[S2:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}]] : !shard.sharding
+  // CHECK-NEXT:  %[[ZP:.*]] = shard.shard %[[ARG2]] to %[[S2]] annotate_for_users  : tensor<1xf32>
+  // CHECK-NEXT:  %[[V2:.*]] = tosa.matmul %[[V0]], %[[V1]], %[[ZP]], %[[ZP]]
+  %0 = tosa.matmul %arg0, %arg1, %arg2, %arg2 : (tensor<2x16x8xf32>, tensor<2x8x32xf32>, tensor<1xf32>, tensor<1xf32>)  -> tensor<2x16x32xf32>
+  // CHECK-NEXT:  %[[V3:.*]] = shard.shard %[[V2]] to %[[S0]]  : tensor<2x16x32xf32>
+  %s1 = shard.sharding @grid_2d split_axes = [[0], [1]] : !shard.sharding
+  %1 = shard.shard %0 to %s1  : tensor<2x16x32xf32>
+  // CHECK-NEXT:  return %[[V3]]
+  return %1 : tensor<2x16x32xf32>
+}
+
+// CHECK-LABEL: func.func @matmul_on_def_shard_m_and_n
+// CHECK-SAME:     [[varg0:%.*]]: tensor<2x16x8xf32>, [[varg1:%.*]]: tensor<2x8x32xf32>, [[varg2:%.*]]: tensor<1xf32>
+func.func @matmul_on_def_shard_m_and_n(%arg0: tensor<2x16x8xf32>, %arg1: tensor<2x8x32xf32>, %arg2: tensor<1xf32>) -> tensor<2x16x32xf32> {
+  // CHECK: [[vsharding:%.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}], [0]] : !shard.sharding
+  // CHECK: [[vsharded:%.*]] = shard.shard [[varg0]] to [[vsharding]] annotate_for_users : tensor<2x16x8xf32>
+  // CHECK: [[vsharding_0:%.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}], [], [1]] : !shard.sharding
+  // CHECK: [[vsharded_1:%.*]] = shard.shard [[varg1]] to [[vsharding_0]] annotate_for_users : tensor<2x8x32xf32>
+  // CHECK: [[vsharding_2:%.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}]] : !shard.sharding
+  // CHECK: [[vsharded_3:%.*]] = shard.shard [[varg2]] to [[vsharding_2]] annotate_for_users : tensor<1xf32>
+  // CHECK: [[v0:%.*]] = tosa.matmul
+  %0 = tosa.matmul %arg0, %arg1, %arg2, %arg2 : (tensor<2x16x8xf32>, tensor<2x8x32xf32>, tensor<1xf32>, tensor<1xf32>)  -> tensor<2x16x32xf32>
+  // CHECK: [[vsharding_4:%.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}], [0], [1]] : !shard.sharding
+  // CHECK: [[vsharded_5:%.*]] = shard.shard [[v0]] to [[vsharding_4]] : tensor<2x16x32xf32>
+  %s1 = shard.sharding @grid_2d split_axes = [[], [0], [1]] : !shard.sharding
+  %1 = shard.shard %0 to %s1  : tensor<2x16x32xf32>
+  // CHECK-NEXT:  return [[vsharded_5]]
+  return %1 : tensor<2x16x32xf32>
+}
+
+// CHECK-LABEL: func.func @matmul_on_use_shard_m_and_k
+// CHECK-SAME:     [[varg0:%.*]]: tensor<2x16x8xf32>, [[varg1:%.*]]: tensor<2x8x32xf32>, [[varg2:%.*]]: tensor<1xf32>
+func.func @matmul_on_use_shard_m_and_k(%arg0: tensor<2x16x8xf32>, %arg1: tensor<2x8x32xf32>, %arg2: tensor<1xf32>) -> tensor<2x16x32xf32> {
+  // CHECK: [[vsharding:%.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}], [0], [1]] : !shard.sharding
+  %s0 = shard.sharding @grid_2d split_axes = [[], [0], [1]] : !shard.sharding
+  // CHECK: [[vsharded:%.*]] = shard.shard [[varg0]] to [[vsharding]] : tensor<2x16x8xf32>
+  %arg0_s = shard.shard %arg0 to %s0 : tensor<2x16x8xf32>
+  // CHECK: [[vsharded_0:%.*]] = shard.shard [[vsharded]] to [[vsharding]] annotate_for_users : tensor<2x16x8xf32>
+  // CHECK: [[vsharding_1:%.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}], [1]] : !shard.sharding
+  // CHECK: [[vsharded_2:%.*]] = shard.shard [[varg1]] to [[vsharding_1]] annotate_for_users : tensor<2x8x32xf32>
+  // CHECK: [[vsharding_3:%.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}]] : !shard.sharding
+  // CHECK: [[vsharded_4:%.*]] = shard.shard [[varg2]] to [[vsharding_3]] annotate_for_users : tensor<1xf32>
+  // CHECK: [[v0:%.*]] = tosa.matmul
+  // CHECK: [[vsharding_5:%.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}], [0]] : !shard.sharding
+  // CHECK: [[vsharded_6:%.*]] = shard.shard [[v0]] to [[vsharding_5]] : tensor<2x16x32xf32>
+  %0 = tosa.matmul %arg0_s, %arg1, %arg2, %arg2 : (tensor<2x16x8xf32>, tensor<2x8x32xf32>, tensor<1xf32>, tensor<1xf32>)  -> tensor<2x16x32xf32>
+  // CHECK: return [[vsharded_6]]
+  return %0 : tensor<2x16x32xf32>
+}
+
+// CHECK-LABEL: func.func @matmul_on_use_shard_m_and_duplicted_k
+// CHECK-SAME:     %[[ARG0:.*]]: tensor<2x16x8xf32>, %[[ARG1:.*]]: tensor<2x8x32xf32>, %[[ARG2:.*]]: tensor<1xf32>
+func.func @matmul_on_use_shard_m_and_duplicted_k(%arg0: tensor<2x16x8xf32>, %arg1: tensor<2x8x32xf32>, %arg2: tensor<1xf32>) -> tensor<2x16x32xf32> {
+  // CHECK-NEXT:  %[[S0:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}], [1], [0]] : !shard.sharding
+  // CHECK-NEXT:  %[[V0:.*]] = shard.shard %[[ARG0]] to %[[S0]] annotate_for_users  : tensor<2x16x8xf32>
+  %s0 = shard.sharding @grid_2d split_axes = [[], [1], [0]] : !shard.sharding
+  %0 = shard.shard %arg0 to %s0 annotate_for_users  : tensor<2x16x8xf32>
+  // CHECK-NEXT:  %[[S1:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}], [0]] : !shard.sharding
+  // CHECK-NEXT:  %[[V1:.*]] = shard.shard %[[ARG1]] to %[[S1]] annotate_for_users  : tensor<2x8x32xf32>
+  %s1 = shard.sharding @grid_2d split_axes = [[], [0]] : !shard.sharding
+  %1 = shard.shard %arg1 to %s1 annotate_for_users  : tensor<2x8x32xf32>
+  // CHECK-NEXT:  %[[S2:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}]] : !shard.sharding
+  // CHECK-NEXT:  %[[ZP:.*]] = shard.shard %[[ARG2]] to %[[S2]] annotate_for_users  : tensor<1xf32>
+  // CHECK-NEXT:  %[[V2:.*]] = tosa.matmul %[[V0]], %[[V1]], %[[ZP]], %[[ZP]]
+  %2 = tosa.matmul %0, %1, %arg2, %arg2 : (tensor<2x16x8xf32>, tensor<2x8x32xf32>, tensor<1xf32>, tensor<1xf32>)  -> tensor<2x16x32xf32>
+  // CHECK-NEXT:  %[[S3:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}], [1]] : !shard.sharding
+  // CHECK-NEXT:  %[[V3:.*]] = shard.shard %[[V2]] to %[[S3]]  : tensor<2x16x32xf32>
+  // CHECK-NEXT:  return %[[V3]]
+  return %2 : tensor<2x16x32xf32>
+}
+
+// CHECK-LABEL: func.func @resolve_conflicting_annotations
+func.func @resolve_conflicting_annotations(
+  // CHECK-SAME: %[[IN1:.*]]: tensor<2x3xf32>,
+  %arg0: tensor<2x3xf32>,
+  // CHECK-SAME: %[[IN2:.*]]: tensor<3x2xf32>,
+  %arg1: tensor<3x2xf32>,
+  // CHECK-SAME: %[[OUT_DPS:.*]]: tensor<2x2xf32>
+  %out_dps: tensor<2x2xf32>
+// CHECK-SAME: ) -> tensor<2x2xf32> {
+) -> tensor<2x2xf32> {
+  // CHECK: %[[SIN1_SHARDED1:.*]] = shard.sharding @grid_2 split_axes = {{\[\[}}0]] : !shard.sharding
+  // CHECK-NEXT:  %[[IN1_SHARDED1:.*]] = shard.shard %[[IN1]] to %[[SIN1_SHARDED1]]  : tensor<2x3xf32>
+  // CHECK: %[[SIN2_SHARDED:.*]] = shard.sharding @grid_2 split_axes = {{\[\[}}]] : !shard.sharding
+  // CHECK-NEXT:  %[[IN1_SHARDED2:.*]] = shard.shard %[[IN1_SHARDED1]] to %[[SIN2_SHARDED]] annotate_for_users  : tensor<2x3xf32>
+  // CHECK-NEXT:  %[[IN2_SHARDED:.*]] = shard.shard %[[IN2]] to %[[SIN2_SHARDED]] annotate_for_users  : tensor<3x2xf32>
+  // CHECK-NEXT:  %[[OUT_DPS_SHARDED:.*]] = shard.shard %[[OUT_DPS]] to %[[SIN2_SHARDED]] annotate_for_users  : tensor<2x2xf32>
+  %sarg0_sharded = shard.sharding @grid_2 split_axes = [[0]] : !shard.sharding
+  %arg0_sharded = shard.shard %arg0 to %sarg0_sharded  : tensor<2x3xf32>
+  // CHECK: %[[MATMUL:.*]] = linalg.matmul ins(%[[IN1_SHARDED2]], %[[IN2_SHARDED]] : tensor<2x3xf32>, tensor<3x2xf32>)
+  // CHECK-SAME: outs(%[[OUT_DPS_SHARDED]] : tensor<2x2xf32>) -> tensor<2x2xf32>
+  %res = linalg.matmul ins(%arg0_sharded, %arg1 : tensor<2x3xf32>, tensor<3x2xf32>)
+    outs(%out_dps : tensor<2x2xf32>) -> tensor<2x2xf32>
+  // CHECK-NEXT: %[[RES:.*]] = shard.shard %[[MATMUL]] to %[[SIN2_SHARDED]] : tensor<2x2xf32>
+  %sres_sharded = shard.sharding @grid_2 split_axes = [[]] : !shard.sharding
+  %res_sharded = shard.shard %res to %sres_sharded  : tensor<2x2xf32>
+  // CHECK: return %[[RES]] : tensor<2x2xf32>
+  return %res_sharded : tensor<2x2xf32>
+}
+
+// https://arxiv.org/abs/2211.05102 Figure 2(a)
+// The sharding propagation results in unnecessary reshards,
+//   an optimization pass should be able to remove them.
+// CHECK-LABEL: func.func @mlp_1d_weight_stationary
+// CHECK-SAME:     [[varg0:%.*]]: tensor<2x4x8xf32>, [[varg1:%.*]]: tensor<2x8x32xf32>, [[varg2:%.*]]: tensor<2x32x8xf32>, [[varg3:%.*]]: tensor<1xf32>
+func.func @mlp_1d_weight_stationary(%arg0: tensor<2x4x8xf32>, %arg1: tensor<2x8x32xf32>, %arg2: tensor<2x32x8xf32>, %arg3: tensor<1xf32>) -> tensor<2x4x8xf32> {
+  %s0 = shard.sharding @grid_1d split_axes = [[], [], [0, 1, 2]] : !shard.sharding
+  %sharded0 = shard.shard %arg0 to %s0 : tensor<2x4x8xf32>
+  %sharded1 = shard.shard %arg1 to %s0 : tensor<2x8x32xf32>
+  // CHECK: [[vsharding:%.*]] = shard.sharding @grid_1d split_axes = {{\[\[}}], [], [0, 1, 2]] : !shard.sharding
+  // CHECK: [[vsharded:%.*]] = shard.shard [[varg0]] to [[vsharding]] : tensor<2x4x8xf32>
+  // CHECK: [[vsharded_0:%.*]] = shard.shard [[varg1]] to [[vsharding]] : tensor<2x8x32xf32>
+  // CHECK: [[vsharded_1:%.*]] = shard.shard [[vsharded]] to [[vsharding]] annotate_for_users : tensor<2x4x8xf32>
+  // CHECK: [[vsharding_2:%.*]] = shard.sharding @grid_1d split_axes = {{\[\[}}], [0, 1, 2]] : !shard.sharding
+  // CHECK: [[vsharded_3:%.*]] = shard.shard [[vsharded_0]] to [[vsharding_2]] annotate_for_users : tensor<2x8x32xf32>
+  // CHECK: [[vsharding_4:%.*]] = shard.sharding @grid_1d split_axes = {{\[\[}}]] : !shard.sharding
+  // CHECK: [[vsharded_5:%.*]] = shard.shard [[varg3]] to [[vsharding_4]] annotate_for_users : tensor<1xf32>
+  // CHECK: [[v0:%.*]] = tosa.matmul
+  %1 = tosa.matmul %sharded0, %sharded1, %arg3, %arg3 : (tensor<2x4x8xf32>, tensor<2x8x32xf32>, tensor<1xf32>, tensor<1xf32>)  -> tensor<2x4x32xf32>
+  // CHECK: [[vsharded_6:%.*]] = shard.shard [[v0]] to [[vsharding_4]] : tensor<2x4x32xf32>
+  // CHECK: [[vsharded_7:%.*]] = shard.shard [[vsharded_6]] to [[vsharding_4]] annotate_for_users : tensor<2x4x32xf32>
+  // CHECK: [[v1:%.*]] = tosa.sigmoid [[vsharded_7]] : (tensor<2x4x32xf32>) -> tensor<2x4x32xf32>
+  // CHECK: [[vsharded_8:%.*]] = shard.shard [[v1]] to [[vsharding_4]] : tensor<2x4x32xf32>
+  %2 = tosa.sigmoid %1 : (tensor<2x4x32xf32>) -> tensor<2x4x32xf32>
+  %sharding = shard.sharding @grid_1d split_axes = [[], [0, 1, 2]] : !shard.sharding
+  // CHECK: [[vsharded_9:%.*]] = shard.shard [[varg2]] to [[vsharding_2]] : tensor<2x32x8xf32>
+  %sharded2 = shard.shard %arg2 to %sharding  : tensor<2x32x8xf32>
+  // CHECK: [[vsharded_10:%.*]] = shard.shard [[vsharded_8]] to [[vsharding_4]] annotate_for_users : tensor<2x4x32xf32>
+  // CHECK: [[vsharded_11:%.*]] = shard.shard [[vsharded_9]] to [[vsharding]] annotate_for_users : tensor<2x32x8xf32>
+  // CHECK: [[v2:%.*]] = tosa.matmul
+  %3 = tosa.matmul %2, %sharded2, %arg3, %arg3 : (tensor<2x4x32xf32>, tensor<2x32x8xf32>, tensor<1xf32>, tensor<1xf32>)  -> tensor<2x4x8xf32>
+  // CHECK: [[vsharded_12:%.*]] = shard.shard [[v2]] to [[vsharding]] : tensor<2x4x8xf32>
+  %s4 = shard.sharding @grid_1d split_axes = [[], [], [0, 1, 2]] : !shard.sharding
+  %4 = shard.shard %3 to %s4  : tensor<2x4x8xf32>
+  // CHECK: return [[vsharded_12]]
+  return %4 : tensor<2x4x8xf32>
+}
+
+// https://arxiv.org/abs/2211.05102 Figure 2(b)
+// The sharding propagation results in unnecessary reshards,
+//   an optimization pass should be able to remove them.
+// CHECK-LABEL: func.func @mlp_2d_weight_stationary
+// CHECK-SAME:     [[varg0:%.*]]: tensor<2x4x8xf32>, [[varg1:%.*]]: tensor<2x8x32xf32>, [[varg2:%.*]]: tensor<2x32x8xf32>, [[varg3:%.*]]: tensor<1xf32>
+func.func @mlp_2d_weight_stationary(%arg0: tensor<2x4x8xf32>, %arg1: tensor<2x8x32xf32>, %arg2: tensor<2x32x8xf32>, %arg3: tensor<1xf32>) -> tensor<2x4x8xf32> {
+    // CHECK: [[vsharding:%.*]] = shard.sharding @grid_3d split_axes = {{\[\[}}], [], [0, 1, 2]] : !shard.sharding
+  %s0 = shard.sharding @grid_3d split_axes = [[], [], [0, 1, 2]] : !shard.sharding
+    // CHECK: [[vsharded:%.*]] = shard.shard [[varg0]] to [[vsharding]] : tensor<2x4x8xf32>
+  %arg0_s = shard.shard %arg0 to %s0  : tensor<2x4x8xf32>
+    // CHECK: [[vsharding_0:%.*]] = shard.sharding @grid_3d split_axes = {{\[\[}}], [0], [1, 2]] : !shard.sharding
+  %s1 = shard.sharding @grid_3d split_axes = [[], [0], [1, 2]] : !shard.sharding
+    // CHECK: [[vsharded_1:%.*]] = shard.shard [[varg1]] to [[vsharding_0]] : tensor<2x8x32xf32>
+  %arg1_s = shard.shard %arg1 to %s1  : tensor<2x8x32xf32>
+    // CHECK: [[vsharding_2:%.*]] = shard.sharding @grid_3d split_axes = {{\[\[}}]] : !shard.sharding
+    // CHECK: [[vsharded_3:%.*]] = shard.shard [[vsharded]] to [[vsharding_2]] annotate_for_users : tensor<2x4x8xf32>
+    // CHECK: [[vsharded_4:%.*]] = shard.shard [[vsharded_1]] to [[vsharding]] annotate_for_users : tensor<2x8x32xf32>
+    // CHECK: [[vsharded_5:%.*]] = shard.shard [[varg3]] to [[vsharding_2]] annotate_for_users : tensor<1xf32>
+    // CHECK: [[v0:%.*]] = tosa.matmul
+  %1 = tosa.matmul %arg0_s, %arg1_s, %arg3, %arg3 : (tensor<2x4x8xf32>, tensor<2x8x32xf32>, tensor<1xf32>, tensor<1xf32>)  -> tensor<2x4x32xf32>
+    // CHECK: [[vsharded_6:%.*]] = shard.shard [[v0]] to [[vsharding]] : tensor<2x4x32xf32>
+  %2 = shard.shard %1 to %s0  : tensor<2x4x32xf32>
+    // CHECK: [[vsharded_7:%.*]] = shard.shard [[vsharded_6]] to [[vsharding]] annotate_for_users : tensor<2x4x32xf32>
+    // CHECK: [[v1:%.*]] = tosa.sigmoid 
+    // CHECK: [[vsharded_8:%.*]] = shard.shard [[v1]] to [[vsharding]] : tensor<2x4x32xf32>
+  %3 = tosa.sigmoid %2 : (tensor<2x4x32xf32>) -> tensor<2x4x32xf32>
+    // CHECK: [[vsharding_9:%.*]] = shard.sharding @grid_3d split_axes = {{\[\[}}], [1, 2], [0]] : !shard.sharding
+  %s2 = shard.sharding @grid_3d split_axes = [[], [1, 2], [0]] : !shard.sharding
+    // CHECK: [[vsharded_10:%.*]] = shard.shard [[varg2]] to [[vsharding_9]] : tensor<2x32x8xf32>
+  %arg2_s = shard.shard %arg2 to %s2  : tensor<2x32x8xf32>
+    // CHECK: [[vsharded_11:%.*]] = shard.shard [[vsharded_8]] to [[vsharding_2]] annotate_for_users : tensor<2x4x32xf32>
+    // CHECK: [[vsharded_12:%.*]] = shard.shard [[vsharded_10]] to [[vsharding]] annotate_for_users : tensor<2x32x8xf32>
+    // CHECK: [[v2:%.*]] = tosa.matmul
+  %4 = tosa.matmul %3, %arg2_s, %arg3, %arg3 : (tensor<2x4x32xf32>, tensor<2x32x8xf32>, tensor<1xf32>, tensor<1xf32>)  -> tensor<2x4x8xf32>
+    // CHECK: [[vsharded_13:%.*]] = shard.shard [[v2]] to [[vsharding]] : tensor<2x4x8xf32>
+  %5 = shard.shard %4 to %s0  : tensor<2x4x8xf32>
+    // CHECK: [[vsharded_14:%.*]] = shard.shard [[vsharded_13]] to [[vsharding]] annotate_for_users : tensor<2x4x8xf32>
+  %6 = shard.shard %5 to %s0 annotate_for_users  : tensor<2x4x8xf32>
+    // CHECK: return [[vsharded_14]]
+  return %6 : tensor<2x4x8xf32>
+}
+
+// CHECK-LABEL: func.func @elementwise_duplicated_chain
+// CHECK-SAME:    %[[ARG:.*]]: tensor<8x16xf32>
+func.func @elementwise_duplicated_chain(%arg0: tensor<8x16xf32>) -> tensor<8x16xf32> {
+  // CHECK-NEXT:  %[[S0:.*]] = shard.sharding @grid_2d split_axes = {{\[\[}}]] : !shard.sharding
+  // CHECK-NEXT:  %[[V0:.*]] = shard.shard %[[ARG]] to %[[S0]] annotate_for_users  : tensor<8x16xf32>
+  // CHECK-NEXT:  %[[V1:.*]] = tosa.sigmoid %[[V0]]
+  %0 = tosa.sigmoid %arg0 : (tensor<8x16xf32>) -> tensor<8x16xf32>
+  // CHECK-NEXT:  %[[V2:.*]] = shard.shard %[[V1]] to %[[S0]] : tensor<8x16xf32>
+  // CHECK-NEXT:  %[[V3:.*]] = shard.shard %[[V2]] to %[[S0]] annotate_for_users : tensor<8x16xf32>
+  // CHECK-NEXT:  %[[V4:.*]] = tosa.sigmoid %[[V3]]
+  %1 = tosa.sigmoid %0 : (tensor<8x16xf32>) -> tensor<8x16xf32>
+  // CHECK-NEXT:  %[[V5:.*]] = shard.shard %[[V4]] to %[[S0]]  : tensor<8x16xf32>
+  %s0 = shard.sharding @grid_2d split_axes = [[]] : !shard.sharding
+  %2 = shard.shard %1 to %s0 : tensor<8x16xf32>
+  // CHECK-NEXT:  return %[[V5]]
+  return %2 : tensor<8x16xf32>
+}
diff --git a/mlir/test/Dialect/Mesh/simplifications.mlir b/mlir/test/Dialect/Shard/simplifications.mlir
index e955f4c..33cd490 100644
--- a/mlir/test/Dialect/Mesh/simplifications.mlir
+++ b/mlir/test/Dialect/Shard/simplifications.mlir
@@ -1,7 +1,7 @@
-// RUN: mlir-opt -test-mesh-simplifications %s | FileCheck %s
+// RUN: mlir-opt -test-grid-simplifications %s | FileCheck %s
 
-mesh.mesh @mesh0(shape = 4x2)
-mesh.mesh @mesh1(shape = 4)
+shard.grid @grid0(shape = 4x2)
+shard.grid @grid1(shape = 4)
 
 // Checks that `all_reduce(x) + all_reduce(y)` gets transformed to
 // `all_reduce(x + y)`.
@@ -11,13 +11,13 @@ func.func @all_reduce_arith_addf_endomorphism(
     %arg0: tensor<5xf32>,
     // CHECK-SAME: %[[ARG1:[A-Za-z0-9_]*]]: tensor<5xf32>
     %arg1: tensor<5xf32>) -> tensor<5xf32> {
-  %0 = mesh.all_reduce %arg0 on @mesh0 mesh_axes = [0]
+  %0 = shard.all_reduce %arg0 on @grid0 grid_axes = [0]
     : tensor<5xf32> -> tensor<5xf32>
-  %1 = mesh.all_reduce %arg1 on @mesh0 mesh_axes = [0]
+  %1 = shard.all_reduce %arg1 on @grid0 grid_axes = [0]
     : tensor<5xf32> -> tensor<5xf32>
   // CHECK: %[[ADD_RES:[A-Za-z0-9_]*]] = arith.addf %[[ARG0]], %[[ARG1]]
   %2 = arith.addf %0, %1 : tensor<5xf32>
-  // CHECK: %[[ALL_REDUCE_RES:[A-Za-z0-9_]*]] = mesh.all_reduce %[[ADD_RES]]
+  // CHECK: %[[ALL_REDUCE_RES:[A-Za-z0-9_]*]] = shard.all_reduce %[[ADD_RES]]
   // CHECK: return %[[ALL_REDUCE_RES]]
   return %2 : tensor<5xf32>
 }
@@ -28,13 +28,13 @@ func.func @all_reduce_arith_addf_endomorphism_multiple_uses_of_result(
     %arg0: tensor<5xf32>,
     // CHECK-SAME: %[[ARG1:[A-Za-z0-9_]*]]: tensor<5xf32>
     %arg1: tensor<5xf32>) -> (tensor<5xf32>, tensor<5xf32>) {
-  %0 = mesh.all_reduce %arg0 on @mesh0 mesh_axes = [0]
+  %0 = shard.all_reduce %arg0 on @grid0 grid_axes = [0]
     : tensor<5xf32> -> tensor<5xf32>
-  %1 = mesh.all_reduce %arg1 on @mesh0 mesh_axes = [0]
+  %1 = shard.all_reduce %arg1 on @grid0 grid_axes = [0]
     : tensor<5xf32> -> tensor<5xf32>
   // CHECK: %[[ADD_RES:[A-Za-z0-9_]*]] = arith.addf %[[ARG0]], %[[ARG1]]
   %2 = arith.addf %0, %1 : tensor<5xf32>
-  // CHECK: %[[ALL_REDUCE_RES:[A-Za-z0-9_]*]] = mesh.all_reduce %[[ADD_RES]]
+  // CHECK: %[[ALL_REDUCE_RES:[A-Za-z0-9_]*]] = shard.all_reduce %[[ADD_RES]]
   // CHECK: return %[[ALL_REDUCE_RES]], %[[ALL_REDUCE_RES]]
   return %2, %2 : tensor<5xf32>, tensor<5xf32>
 }
@@ -46,11 +46,11 @@ func.func @all_reduce_arith_addf_endomorphism_multiple_uses_of_all_reduce_result
     %arg0: tensor<5xf32>,
     // CHECK-SAME: %[[ARG1:[A-Za-z0-9_]*]]: tensor<5xf32>
     %arg1: tensor<5xf32>) -> (tensor<5xf32>, tensor<5xf32>) {
-  // CHECK: %[[ALL_REDUCE_0_RES:[A-Za-z0-9_]*]] = mesh.all_reduce %[[ARG0]]
-  %0 = mesh.all_reduce %arg0 on @mesh0 mesh_axes = [0]
+  // CHECK: %[[ALL_REDUCE_0_RES:[A-Za-z0-9_]*]] = shard.all_reduce %[[ARG0]]
+  %0 = shard.all_reduce %arg0 on @grid0 grid_axes = [0]
     : tensor<5xf32> -> tensor<5xf32>
-  // CHECK: %[[ALL_REDUCE_1_RES:[A-Za-z0-9_]*]] = mesh.all_reduce %[[ARG1]]
-  %1 = mesh.all_reduce %arg1 on @mesh0 mesh_axes = [0]
+  // CHECK: %[[ALL_REDUCE_1_RES:[A-Za-z0-9_]*]] = shard.all_reduce %[[ARG1]]
+  %1 = shard.all_reduce %arg1 on @grid0 grid_axes = [0]
     : tensor<5xf32> -> tensor<5xf32>
   // CHECK: %[[ADD_RES:[A-Za-z0-9_]*]] = arith.addf %[[ALL_REDUCE_0_RES]], %[[ALL_REDUCE_1_RES]]
   %2 = arith.addf %0, %1 : tensor<5xf32>
@@ -58,17 +58,17 @@ func.func @all_reduce_arith_addf_endomorphism_multiple_uses_of_all_reduce_result
   return %0, %2 : tensor<5xf32>, tensor<5xf32>
 }
 
-// CHECK-LABEL: func.func @all_reduce_arith_addf_no_endomorphism_different_mesh
-func.func @all_reduce_arith_addf_no_endomorphism_different_mesh(
+// CHECK-LABEL: func.func @all_reduce_arith_addf_no_endomorphism_different_grid
+func.func @all_reduce_arith_addf_no_endomorphism_different_grid(
     // CHECK-SAME: %[[ARG0:[A-Za-z0-9_]*]]: tensor<5xf32>
     %arg0: tensor<5xf32>,
     // CHECK-SAME: %[[ARG1:[A-Za-z0-9_]*]]: tensor<5xf32>
     %arg1: tensor<5xf32>) -> tensor<5xf32> {
-  // CHECK: %[[ALL_REDUCE0:[A-Za-z0-9_]*]] = mesh.all_reduce %[[ARG0]] on @mesh0
-  %0 = mesh.all_reduce %arg0 on @mesh0 mesh_axes = [0]
+  // CHECK: %[[ALL_REDUCE0:[A-Za-z0-9_]*]] = shard.all_reduce %[[ARG0]] on @grid0
+  %0 = shard.all_reduce %arg0 on @grid0 grid_axes = [0]
     : tensor<5xf32> -> tensor<5xf32>
-  // CHECK: %[[ALL_REDUCE1:[A-Za-z0-9_]*]] = mesh.all_reduce %[[ARG1]] on @mesh1
-  %1 = mesh.all_reduce %arg1 on @mesh1 mesh_axes = [0]
+  // CHECK: %[[ALL_REDUCE1:[A-Za-z0-9_]*]] = shard.all_reduce %[[ARG1]] on @grid1
+  %1 = shard.all_reduce %arg1 on @grid1 grid_axes = [0]
     : tensor<5xf32> -> tensor<5xf32>
   // CHECK: %[[ADD_RES:[A-Za-z0-9_]*]] = arith.addf %[[ALL_REDUCE0]], %[[ALL_REDUCE1]]
   %2 = arith.addf %0, %1 : tensor<5xf32>
@@ -76,17 +76,17 @@ func.func @all_reduce_arith_addf_no_endomorphism_different_mesh(
   return %2 : tensor<5xf32>
 }
 
-// CHECK-LABEL: func.func @all_reduce_arith_addf_no_endomorphism_different_mesh_axes
-func.func @all_reduce_arith_addf_no_endomorphism_different_mesh_axes(
+// CHECK-LABEL: func.func @all_reduce_arith_addf_no_endomorphism_different_grid_axes
+func.func @all_reduce_arith_addf_no_endomorphism_different_grid_axes(
     // CHECK-SAME: %[[ARG0:[A-Za-z0-9_]*]]: tensor<5xf32>
     %arg0: tensor<5xf32>,
     // CHECK-SAME: %[[ARG1:[A-Za-z0-9_]*]]: tensor<5xf32>
     %arg1: tensor<5xf32>) -> tensor<5xf32> {
-  // CHECK: %[[ALL_REDUCE0:[A-Za-z0-9_]*]] = mesh.all_reduce %[[ARG0]] on @mesh0 mesh_axes = [0]
-  %0 = mesh.all_reduce %arg0 on @mesh0 mesh_axes = [0]
+  // CHECK: %[[ALL_REDUCE0:[A-Za-z0-9_]*]] = shard.all_reduce %[[ARG0]] on @grid0 grid_axes = [0]
+  %0 = shard.all_reduce %arg0 on @grid0 grid_axes = [0]
     : tensor<5xf32> -> tensor<5xf32>
-  // CHECK: %[[ALL_REDUCE1:[A-Za-z0-9_]*]] = mesh.all_reduce %[[ARG1]] on @mesh0 mesh_axes = [1]
-  %1 = mesh.all_reduce %arg1 on @mesh0 mesh_axes = [1]
+  // CHECK: %[[ALL_REDUCE1:[A-Za-z0-9_]*]] = shard.all_reduce %[[ARG1]] on @grid0 grid_axes = [1]
+  %1 = shard.all_reduce %arg1 on @grid0 grid_axes = [1]
     : tensor<5xf32> -> tensor<5xf32>
   // CHECK: %[[ADD_RES:[A-Za-z0-9_]*]] = arith.addf %[[ALL_REDUCE0]], %[[ALL_REDUCE1]]
   %2 = arith.addf %0, %1 : tensor<5xf32>
@@ -100,11 +100,11 @@ func.func @all_reduce_arith_addf_no_endomorphism_wrong_reduction_kind(
     %arg0: tensor<5xf32>,
     // CHECK-SAME: %[[ARG1:[A-Za-z0-9_]*]]: tensor<5xf32>
     %arg1: tensor<5xf32>) -> tensor<5xf32> {
-  // CHECK: %[[ALL_REDUCE0:[A-Za-z0-9_]*]] = mesh.all_reduce %[[ARG0]] on @mesh0 mesh_axes = [0] reduction = max
-  %0 = mesh.all_reduce %arg0 on @mesh0 mesh_axes = [0] reduction = max
+  // CHECK: %[[ALL_REDUCE0:[A-Za-z0-9_]*]] = shard.all_reduce %[[ARG0]] on @grid0 grid_axes = [0] reduction = max
+  %0 = shard.all_reduce %arg0 on @grid0 grid_axes = [0] reduction = max
     : tensor<5xf32> -> tensor<5xf32>
-  // CHECK: %[[ALL_REDUCE1:[A-Za-z0-9_]*]] = mesh.all_reduce %[[ARG1]] on @mesh0 mesh_axes = [0]
-  %1 = mesh.all_reduce %arg1 on @mesh0 mesh_axes = [0]
+  // CHECK: %[[ALL_REDUCE1:[A-Za-z0-9_]*]] = shard.all_reduce %[[ARG1]] on @grid0 grid_axes = [0]
+  %1 = shard.all_reduce %arg1 on @grid0 grid_axes = [0]
     : tensor<5xf32> -> tensor<5xf32>
   // CHECK: %[[ADD_RES:[A-Za-z0-9_]*]] = arith.addf %[[ALL_REDUCE0]], %[[ALL_REDUCE1]]
   %2 = arith.addf %0, %1 : tensor<5xf32>
@@ -118,11 +118,11 @@ func.func @all_reduce_arith_addf_no_endomorphism_different_operand_result_elemen
     %arg0: tensor<5xf32>,
     // CHECK-SAME: %[[ARG1:[A-Za-z0-9_]*]]: tensor<5xf32>
     %arg1: tensor<5xf32>) -> tensor<5xf64> {
-  // CHECK: %[[ALL_REDUCE0:[A-Za-z0-9_]*]] = mesh.all_reduce %[[ARG0]] on @mesh0 mesh_axes = [0]
-  %0 = mesh.all_reduce %arg0 on @mesh0 mesh_axes = [0]
+  // CHECK: %[[ALL_REDUCE0:[A-Za-z0-9_]*]] = shard.all_reduce %[[ARG0]] on @grid0 grid_axes = [0]
+  %0 = shard.all_reduce %arg0 on @grid0 grid_axes = [0]
     : tensor<5xf32> -> tensor<5xf64>
-  // CHECK: %[[ALL_REDUCE1:[A-Za-z0-9_]*]] = mesh.all_reduce %[[ARG1]] on @mesh0 mesh_axes = [0]
-  %1 = mesh.all_reduce %arg1 on @mesh0 mesh_axes = [0]
+  // CHECK: %[[ALL_REDUCE1:[A-Za-z0-9_]*]] = shard.all_reduce %[[ARG1]] on @grid0 grid_axes = [0]
+  %1 = shard.all_reduce %arg1 on @grid0 grid_axes = [0]
     : tensor<5xf32> -> tensor<5xf64>
   // CHECK: %[[ADD_RES:[A-Za-z0-9_]*]] = arith.addf %[[ALL_REDUCE0]], %[[ALL_REDUCE1]]
   %2 = arith.addf %0, %1 : tensor<5xf64>
@@ -138,13 +138,13 @@ func.func @all_reduce_arith_minimumf_endomorphism(
     %arg0: tensor<5xf32>,
     // CHECK-SAME: %[[ARG1:[A-Za-z0-9_]*]]: tensor<5xf32>
     %arg1: tensor<5xf32>) -> tensor<5xf32> {
-  %0 = mesh.all_reduce %arg0 on @mesh0 mesh_axes = [0] reduction = min
+  %0 = shard.all_reduce %arg0 on @grid0 grid_axes = [0] reduction = min
     : tensor<5xf32> -> tensor<5xf32>
-  %1 = mesh.all_reduce %arg1 on @mesh0 mesh_axes = [0] reduction = min
+  %1 = shard.all_reduce %arg1 on @grid0 grid_axes = [0] reduction = min
     : tensor<5xf32> -> tensor<5xf32>
   // CHECK: %[[ADD_RES:[A-Za-z0-9_]*]] = arith.minimumf %[[ARG0]], %[[ARG1]]
   %2 = arith.minimumf %0, %1 : tensor<5xf32>
-  // CHECK: %[[ALL_REDUCE_RES:[A-Za-z0-9_]*]] = mesh.all_reduce %[[ADD_RES]] on @mesh0 mesh_axes = [0] reduction = min
+  // CHECK: %[[ALL_REDUCE_RES:[A-Za-z0-9_]*]] = shard.all_reduce %[[ADD_RES]] on @grid0 grid_axes = [0] reduction = min
   // CHECK: return %[[ALL_REDUCE_RES]]
   return %2 : tensor<5xf32>
 }
@@ -155,13 +155,13 @@ func.func @all_reduce_arith_minsi_endomorphism(
     %arg0: tensor<5xi32>,
     // CHECK-SAME: %[[ARG1:[A-Za-z0-9_]*]]: tensor<5xi32>
     %arg1: tensor<5xi32>) -> tensor<5xi32> {
-  %0 = mesh.all_reduce %arg0 on @mesh0 mesh_axes = [0] reduction = min
+  %0 = shard.all_reduce %arg0 on @grid0 grid_axes = [0] reduction = min
     : tensor<5xi32> -> tensor<5xi32>
-  %1 = mesh.all_reduce %arg1 on @mesh0 mesh_axes = [0] reduction = min
+  %1 = shard.all_reduce %arg1 on @grid0 grid_axes = [0] reduction = min
     : tensor<5xi32> -> tensor<5xi32>
   // CHECK: %[[ADD_RES:[A-Za-z0-9_]*]] = arith.minsi %[[ARG0]], %[[ARG1]]
   %2 = arith.minsi %0, %1 : tensor<5xi32>
-  // CHECK: %[[ALL_REDUCE_RES:[A-Za-z0-9_]*]] = mesh.all_reduce %[[ADD_RES]] on @mesh0 mesh_axes = [0] reduction = min
+  // CHECK: %[[ALL_REDUCE_RES:[A-Za-z0-9_]*]] = shard.all_reduce %[[ADD_RES]] on @grid0 grid_axes = [0] reduction = min
   // CHECK: return %[[ALL_REDUCE_RES]]
   return %2 : tensor<5xi32>
 }
diff --git a/mlir/test/Dialect/Tensor/mesh-spmdization.mlir b/mlir/test/Dialect/Tensor/mesh-spmdization.mlir
deleted file mode 100644
index 8598d81..0000000
--- a/mlir/test/Dialect/Tensor/mesh-spmdization.mlir
+++ /dev/null
@@ -1,52 +0,0 @@
-// RUN: mlir-opt \
-// RUN:   --pass-pipeline="builtin.module(func.func(mesh-spmdization,test-single-fold))" \
-// RUN:   %s | FileCheck %s
-
-mesh.mesh @mesh_1d_4(shape = 4)
-
-// CHECK-LABEL: func @tensor_empty_static_sharded_dims_offsets
-func.func @tensor_empty_static_sharded_dims_offsets() -> () {
-  %b = tensor.empty() : tensor<8x16xf32>
-  %sharding = mesh.sharding @mesh_1d_4 split_axes = [[0]] sharded_dims_offsets = [0, 1, 4, 7, 8] : !mesh.sharding
-  %sharded= mesh.shard %b to %sharding : tensor<8x16xf32>
-  // CHECK:  %[[sharding:.*]] = mesh.sharding @mesh_1d_4 split_axes = {{\[\[}}0]] sharded_dims_offsets = [0, 1, 4, 7, 8] : !mesh.sharding
-  // CHECK:  %[[proc_multi_idx:.*]] = mesh.process_multi_index on @mesh_1d_4 : index
-  // CHECK:  %[[V0:.*]]:2 = mesh.shard_shape dims = [8, 16] sharding = %[[sharding]] device = [%[[proc_multi_idx]]
-  // CHECK-SAME: ] : index, index
-  // CHECK:  tensor.empty(%[[V0]]#0) : tensor<?x16xf32>
-
-  return
-}
-
-// CHECK-LABEL: func @tensor_empty_dynamic_sharded_dims_offsets
-// CHECK-SAME: %[[A0:.*]]: index
-func.func @tensor_empty_dynamic_sharded_dims_offsets(%arg0 : index) -> () {
-  %b = tensor.empty(%arg0) : tensor<8x?xf32>
-  %sharding = mesh.sharding @mesh_1d_4 split_axes = [[0]] sharded_dims_offsets = [0, 1, 4, 7, 8] : !mesh.sharding
-  %sharded= mesh.shard %b to %sharding : tensor<8x?xf32>
-  // CHECK:  %[[sharding:.*]] = mesh.sharding @mesh_1d_4 split_axes = {{\[\[}}0]] sharded_dims_offsets = [0, 1, 4, 7, 8] : !mesh.sharding
-  // CHECK:  %[[proc_multi_idx:.*]] = mesh.process_multi_index on @mesh_1d_4 : index
-  // CHECK:  %[[V0:.*]]:2 = mesh.shard_shape dims = [8, %[[A0]]
-  // CHECK-SAME: ] sharding = %[[sharding]] device = [%[[proc_multi_idx]]
-  // CHECK-SAME: ] : index, index
-  // CHECK:  tensor.empty(%[[V0]]#0, %[[A0]]) : tensor<?x?xf32>
-
-  return
-}
-
-// CHECK-LABEL: func @tensor_empty_same_static_dims_sizes
-func.func @tensor_empty_same_static_dims_sizes() -> () {
-  %b = tensor.empty() : tensor<16x16xf32>
-  %sharding = mesh.sharding @mesh_1d_4 split_axes = [[0]] sharded_dims_offsets = [0, 4, 8, 12, 16] : !mesh.sharding
-  %sharded= mesh.shard %b to %sharding : tensor<16x16xf32>
-  // CHECK-NEXT:  tensor.empty() : tensor<4x16xf32>
-
-  return
-}
-
-// CHECK-LABEL: func @tensor_empty_0d
-func.func @tensor_empty_0d() -> () {
-  tensor.empty() : tensor<f32>
-  // CHECK-NEXT:  tensor.empty() : tensor<f32>
-  return
-}
diff --git a/mlir/test/Dialect/Tensor/shard-partition.mlir b/mlir/test/Dialect/Tensor/shard-partition.mlir
new file mode 100644
index 0000000..5918ee1
--- /dev/null
+++ b/mlir/test/Dialect/Tensor/shard-partition.mlir
@@ -0,0 +1,52 @@
+// RUN: mlir-opt \
+// RUN:   --pass-pipeline="builtin.module(func.func(shard-partition,test-single-fold))" \
+// RUN:   %s | FileCheck %s
+
+shard.grid @grid_1d_4(shape = 4)
+
+// CHECK-LABEL: func @tensor_empty_static_sharded_dims_offsets
+func.func @tensor_empty_static_sharded_dims_offsets() -> () {
+  %b = tensor.empty() : tensor<8x16xf32>
+  %sharding = shard.sharding @grid_1d_4 split_axes = [[0]] sharded_dims_offsets = [0, 1, 4, 7, 8] : !shard.sharding
+  %sharded= shard.shard %b to %sharding : tensor<8x16xf32>
+  // CHECK:  %[[sharding:.*]] = shard.sharding @grid_1d_4 split_axes = {{\[\[}}0]] sharded_dims_offsets = [0, 1, 4, 7, 8] : !shard.sharding
+  // CHECK:  %[[proc_multi_idx:.*]] = shard.process_multi_index on @grid_1d_4 : index
+  // CHECK:  %[[V0:.*]]:2 = shard.shard_shape dims = [8, 16] sharding = %[[sharding]] device = [%[[proc_multi_idx]]
+  // CHECK-SAME: ] : index, index
+  // CHECK:  tensor.empty(%[[V0]]#0) : tensor<?x16xf32>
+
+  return
+}
+
+// CHECK-LABEL: func @tensor_empty_dynamic_sharded_dims_offsets
+// CHECK-SAME: %[[A0:.*]]: index
+func.func @tensor_empty_dynamic_sharded_dims_offsets(%arg0 : index) -> () {
+  %b = tensor.empty(%arg0) : tensor<8x?xf32>
+  %sharding = shard.sharding @grid_1d_4 split_axes = [[0]] sharded_dims_offsets = [0, 1, 4, 7, 8] : !shard.sharding
+  %sharded= shard.shard %b to %sharding : tensor<8x?xf32>
+  // CHECK:  %[[sharding:.*]] = shard.sharding @grid_1d_4 split_axes = {{\[\[}}0]] sharded_dims_offsets = [0, 1, 4, 7, 8] : !shard.sharding
+  // CHECK:  %[[proc_multi_idx:.*]] = shard.process_multi_index on @grid_1d_4 : index
+  // CHECK:  %[[V0:.*]]:2 = shard.shard_shape dims = [8, %[[A0]]
+  // CHECK-SAME: ] sharding = %[[sharding]] device = [%[[proc_multi_idx]]
+  // CHECK-SAME: ] : index, index
+  // CHECK:  tensor.empty(%[[V0]]#0, %[[A0]]) : tensor<?x?xf32>
+
+  return
+}
+
+// CHECK-LABEL: func @tensor_empty_same_static_dims_sizes
+func.func @tensor_empty_same_static_dims_sizes() -> () {
+  %b = tensor.empty() : tensor<16x16xf32>
+  %sharding = shard.sharding @grid_1d_4 split_axes = [[0]] sharded_dims_offsets = [0, 4, 8, 12, 16] : !shard.sharding
+  %sharded= shard.shard %b to %sharding : tensor<16x16xf32>
+  // CHECK-NEXT:  tensor.empty() : tensor<4x16xf32>
+
+  return
+}
+
+// CHECK-LABEL: func @tensor_empty_0d
+func.func @tensor_empty_0d() -> () {
+  tensor.empty() : tensor<f32>
+  // CHECK-NEXT:  tensor.empty() : tensor<f32>
+  return
+}
diff --git a/mlir/test/Dialect/Tosa/availability.mlir b/mlir/test/Dialect/Tosa/availability.mlir
index 0176fc2..6398161 100644
--- a/mlir/test/Dialect/Tosa/availability.mlir
+++ b/mlir/test/Dialect/Tosa/availability.mlir
@@ -645,7 +645,7 @@ func.func @test_identity(%arg0: tensor<13x21x3xi32>) -> tensor<13x21x3xi32> {
 func.func @test_cond_if(%arg0: tensor<f32>, %arg1: tensor<f32>, %arg2: tensor<i1>) -> tensor<f32> {
   // CHECK: tosa.cond_if profiles: [ ]
   // CHECK: tosa.cond_if extensions: [ [controlflow] ]
-  %0 = tosa.cond_if %arg2 -> (tensor<f32>) {
+  %0 = tosa.cond_if %arg2 : tensor<i1> -> tensor<f32> {
     %1 = tosa.add %arg0, %arg1 : (tensor<f32>, tensor<f32>) -> tensor<f32>
     tosa.yield %1 : tensor<f32>
   } else {
diff --git a/mlir/test/Dialect/Tosa/controlflow.mlir b/mlir/test/Dialect/Tosa/controlflow.mlir
new file mode 100644
index 0000000..06312c7
--- /dev/null
+++ b/mlir/test/Dialect/Tosa/controlflow.mlir
@@ -0,0 +1,35 @@
+// RUN: mlir-opt -split-input-file %s | FileCheck %s
+
+// -----
+
+func.func @condif_cond_type_check(%arg0: tensor<f32>, %arg1: tensor<f32>, %arg2: tensor<i1>) -> tensor<f32> {
+  // CHECK: tosa.cond_if %[[ARG2:.*]] : tensor<i1> -> tensor<f32> {
+  %0 = tosa.cond_if %arg2 : tensor<i1> -> tensor<f32> {
+    %1 = tosa.add %arg0, %arg1 : (tensor<f32>, tensor<f32>) -> tensor<f32>
+    tosa.yield %1 : tensor<f32>
+  // CHECK:     } else {
+  } else {
+    %1 = tosa.sub %arg0, %arg1 : (tensor<f32>, tensor<f32>) -> tensor<f32>
+    tosa.yield %1 : tensor<f32>
+  }
+  return %0 : tensor<f32>
+}
+ 
+// -----
+
+func.func @condif_block_args_check(%arg0: tensor<f32>, %arg1: tensor<f32>, %arg2: tensor<i1>) -> tensor<f32> {
+  // CHECK: tosa.cond_if %[[ARG2:.*]] (%[[ARG3:.*]] = %[[ARG0:.*]], %[[ARG4:.*]] = %[[ARG1:.*]]) : tensor<i1> (tensor<f32>, tensor<f32>) -> tensor<f32> {
+  // CHECK-NEXT: ^bb0(%[[ARG3]]: tensor<f32>, %[[ARG4]]: tensor<f32>):
+  %0 = tosa.cond_if %arg2 (%arg3 = %arg0, %arg4 = %arg1) : tensor<i1> (tensor<f32>, tensor<f32>) -> tensor<f32> {
+  ^bb0(%arg3: tensor<f32>, %arg4: tensor<f32>):
+    %1 = tosa.add %arg3, %arg4 : (tensor<f32>, tensor<f32>) -> tensor<f32>
+    tosa.yield %1 : tensor<f32>
+  // CHECK:     } else {
+  // CHECK-NEXT: ^bb0(%[[ARG3]]: tensor<f32>, %[[ARG4]]: tensor<f32>):
+  } else {
+  ^bb0(%arg3: tensor<f32>, %arg4: tensor<f32>):
+    %1 = tosa.sub %arg3, %arg4 : (tensor<f32>, tensor<f32>) -> tensor<f32>
+    tosa.yield %1 : tensor<f32>
+  }
+  return %0 : tensor<f32>
+} 
diff --git a/mlir/test/Dialect/Tosa/error_if_check.mlir b/mlir/test/Dialect/Tosa/error_if_check.mlir
index eb25011..fad1bec 100644
--- a/mlir/test/Dialect/Tosa/error_if_check.mlir
+++ b/mlir/test/Dialect/Tosa/error_if_check.mlir
@@ -259,7 +259,7 @@ func.func @test_cond_if_else_not_isolated_from_above(%arg0: tensor<f32>, %arg1:
 
 func.func @test_cond_if_simplified_form_not_isolated_from_above(%arg0: tensor<f32>, %arg1: tensor<f32>, %arg2: tensor<i1>) -> tensor<f32> {
   // expected-error@+1 {{'tosa.cond_if' op is not conformant to the TOSA specification. It requires the 'then' region is isolated from above.}}
-  %0 = tosa.cond_if %arg2 -> (tensor<f32>) {
+  %0 = tosa.cond_if %arg2 : tensor<i1> -> (tensor<f32>) {
     tosa.yield %arg0 : tensor<f32>
   } else {
     tosa.yield %arg1 : tensor<f32>
diff --git a/mlir/test/Dialect/Tosa/invalid.mlir b/mlir/test/Dialect/Tosa/invalid.mlir
index 716362e..b90d6f5 100644
--- a/mlir/test/Dialect/Tosa/invalid.mlir
+++ b/mlir/test/Dialect/Tosa/invalid.mlir
@@ -1125,7 +1125,7 @@ func.func @test_sub_with_unequal_result_ranks(%arg0: tensor<1x21x3xf32>, %arg1:
 // CHECK-LABEL: test_mul_non_scalar_shift_2d
 func.func @test_mul_non_scalar_shift_2d(%arg0: tensor<13x21x3xf32>, %arg1: tensor<13x1x3xf32>) -> tensor<13x21x3xf32> {
   %shift = "tosa.const"() <{values = dense<0> : tensor<1x1xi8>}> : () -> tensor<1x1xi8>
-  // expected-error@+1 {{'tosa.mul' op operand #2 must be tosa-conformant scalar tensor of 8-bit signless integer values, but got 'tensor<1x1xi8>'}}
+  // expected-error@+1 {{'tosa.mul' op operand #2 must be tosa-conformant unranked tensor of 8-bit signless integer values or tosa-conformant scalar tensor of 8-bit signless integer values, but got 'tensor<1x1xi8>'}}
   %0 = tosa.mul %arg0, %arg1, %shift : (tensor<13x21x3xf32>, tensor<13x1x3xf32>, tensor<1x1xi8>) -> tensor<13x21x3xf32>
   return %0 : tensor<13x21x3xf32>
 }
@@ -1134,7 +1134,7 @@ func.func @test_mul_non_scalar_shift_2d(%arg0: tensor<13x21x3xf32>, %arg1: tenso
 // CHECK-LABEL: test_mul_non_scalar_shift_1d
 func.func @test_mul_non_scalar_shift_1d(%arg0: tensor<13x21x3xf32>, %arg1: tensor<13x1x3xf32>) -> tensor<13x21x3xf32> {
   %shift = "tosa.const"() <{values = dense<0> : tensor<2xi8>}> : () -> tensor<2xi8>
-  // expected-error@+1 {{'tosa.mul' op operand #2 must be tosa-conformant scalar tensor of 8-bit signless integer values, but got 'tensor<2xi8>'}}
+  // expected-error@+1 {{'tosa.mul' op operand #2 must be tosa-conformant unranked tensor of 8-bit signless integer values or tosa-conformant scalar tensor of 8-bit signless integer values, but got 'tensor<2xi8>'}}
   %0 = tosa.mul %arg0, %arg1, %shift : (tensor<13x21x3xf32>, tensor<13x1x3xf32>, tensor<2xi8>) -> tensor<13x21x3xf32>
   return %0 : tensor<13x21x3xf32>
 }
diff --git a/mlir/test/Dialect/Tosa/invalid_extension.mlir b/mlir/test/Dialect/Tosa/invalid_extension.mlir
index 5630c33..3154f54 100644
--- a/mlir/test/Dialect/Tosa/invalid_extension.mlir
+++ b/mlir/test/Dialect/Tosa/invalid_extension.mlir
@@ -337,7 +337,7 @@ func.func @test_cast_bf16_i32(%arg0: tensor<13x21x3xbf16>) -> tensor<13x21x3xi32
 // -----
 func.func @test_cond_if(%arg0: tensor<f32>, %arg1: tensor<f32>, %arg2: tensor<i1>) -> tensor<f32> {
   // expected-error@+1 {{'tosa.cond_if' op illegal: requires [controlflow]}}
-  %0 = tosa.cond_if %arg2 -> (tensor<f32>) {
+  %0 = tosa.cond_if %arg2 : tensor<i1> -> tensor<f32> {
     %1 = tosa.add %arg0, %arg1 : (tensor<f32>, tensor<f32>) -> tensor<f32>
     tosa.yield %1 : tensor<f32>
   } else {
diff --git a/mlir/test/Dialect/Tosa/level_check.mlir b/mlir/test/Dialect/Tosa/level_check.mlir
index 0dddf26..cbe0056 100644
--- a/mlir/test/Dialect/Tosa/level_check.mlir
+++ b/mlir/test/Dialect/Tosa/level_check.mlir
@@ -1506,13 +1506,13 @@ func.func @test_while_tensor_list_size(%arg0: tensor<1x1x1x1x1x1x1xf32>, %arg1:
 // -----
 
 func.func @test_cond_if_max_nested_depth(%arg0: tensor<f32>, %arg1: tensor<f32>, %arg2: tensor<i1>, %arg3: tensor<i1>) -> tensor<f32> {
-  %0 = tosa.cond_if %arg2 -> (tensor<f32>) {
-    %1 = tosa.cond_if %arg3 -> (tensor<f32>) {
-      %2 = tosa.cond_if %arg2 -> (tensor<f32>) {
-        %3 = tosa.cond_if %arg3 -> (tensor<f32>) {
-          %4 = tosa.cond_if %arg2 -> (tensor<f32>) {
+  %0 = tosa.cond_if %arg2 : tensor<i1> -> tensor<f32> {
+    %1 = tosa.cond_if %arg3 : tensor<i1>-> tensor<f32> {
+      %2 = tosa.cond_if %arg2 : tensor<i1> -> tensor<f32> {
+        %3 = tosa.cond_if %arg3 : tensor<i1> -> tensor<f32> {
+          %4 = tosa.cond_if %arg2 : tensor<i1>  -> tensor<f32> {
             // expected-error@+1 {{'tosa.cond_if' op failed level check: 6 >= MAX_NESTING}}
-            %5 = tosa.cond_if %arg3 -> (tensor<f32>) {
+            %5 = tosa.cond_if %arg3 : tensor<i1> -> tensor<f32> {
               %res = tosa.sub %arg0, %arg1 : (tensor<f32>, tensor<f32>) -> tensor<f32>
               tosa.yield %res : tensor<f32>
             } else {
diff --git a/mlir/test/Dialect/Tosa/ops.mlir b/mlir/test/Dialect/Tosa/ops.mlir
index ef51197e..30361a8 100644
--- a/mlir/test/Dialect/Tosa/ops.mlir
+++ b/mlir/test/Dialect/Tosa/ops.mlir
@@ -839,7 +839,7 @@ func.func @test_identity(%arg0: tensor<13x21x3xi32>) -> tensor<13x21x3xi32> {
 // -----
 // CHECK-LABEL: cond_if
 func.func @test_cond_if(%arg0: tensor<f32>, %arg1: tensor<f32>, %arg2: tensor<i1>) -> tensor<f32> {
-  %0 = tosa.cond_if %arg2 -> (tensor<f32>) {
+  %0 = tosa.cond_if %arg2 : tensor<i1> -> tensor<f32> {
     %1 = tosa.add %arg0, %arg1 : (tensor<f32>, tensor<f32>) -> tensor<f32>
     tosa.yield %1 : tensor<f32>
   } else {
diff --git a/mlir/test/Dialect/Tosa/tosa-convert-integer-type-to-signless.mlir b/mlir/test/Dialect/Tosa/tosa-convert-integer-type-to-signless.mlir
index 38ac8d8..e957bdd 100644
--- a/mlir/test/Dialect/Tosa/tosa-convert-integer-type-to-signless.mlir
+++ b/mlir/test/Dialect/Tosa/tosa-convert-integer-type-to-signless.mlir
@@ -54,7 +54,7 @@ func.func @test_no_change(%arg0: tensor<13x21x3xi8>) -> tensor<13x21x3xi8> {
 // CHECK-LABEL: test_regions
 // CHECK: %arg0: tensor<i8>, %arg1: tensor<i8>
 func.func @test_regions(%arg0: tensor<ui8>, %arg1: tensor<ui8>, %arg2: tensor<i1>) -> tensor<ui8> {
-  // CHECK: tosa.cond_if %arg2 -> (tensor<i8>)
+  // CHECK: tosa.cond_if %arg2 (%arg3 = %arg0, %arg4 = %arg1) : tensor<i1> (tensor<i8>, tensor<i8>) -> tensor<i8> 
   %0 = "tosa.cond_if"(%arg2, %arg0, %arg1) ({
   ^bb0(%arg3: tensor<ui8>, %arg4: tensor<ui8>):
     // CHECK: %1 = tosa.add %arg0, %arg1 : (tensor<i8>, tensor<i8>) -> tensor<i8>
diff --git a/mlir/test/Dialect/Tosa/tosa-infer-shapes.mlir b/mlir/test/Dialect/Tosa/tosa-infer-shapes.mlir
index 9d43f89..7b8fc24 100644
--- a/mlir/test/Dialect/Tosa/tosa-infer-shapes.mlir
+++ b/mlir/test/Dialect/Tosa/tosa-infer-shapes.mlir
@@ -357,6 +357,17 @@ func.func @test_accepts_unranked_scalar_tensor(%arg0: tensor<1x2x2xf32>, %arg1:
 
 // -----
 
+// CHECK-LABEL: @test_unranked_scalar_i8_tensor
+func.func @test_unranked_scalar_i8_tensor(%arg0: tensor<4xi32>, %arg1: tensor<4xi32>, %arg2: tensor<1xi8>) -> tensor<4xi32> {
+  // CHECK: %[[SHIFT:.*]] = tosa.cast %arg2 : (tensor<1xi8>) -> tensor<1xi8>
+  %shift = tosa.cast %arg2 : (tensor<1xi8>) -> tensor<*xi8>
+  // CHECK: tosa.mul %arg0, %arg1, %[[SHIFT]] : (tensor<4xi32>, tensor<4xi32>, tensor<1xi8>) -> tensor<4xi32>
+  %0 = tosa.mul %arg0, %arg1, %shift : (tensor<4xi32>, tensor<4xi32>, tensor<*xi8>) -> tensor<4xi32>
+  return %0 : tensor<4xi32>
+}
+
+// -----
+
 // CHECK-LABEL: @test_table_static
 func.func @test_table_static(%arg0 : tensor<4x5xi16>, %arg1 : tensor<513xi16>) -> () {
   // CHECK:tosa.table %arg0, %arg1 : (tensor<4x5xi16>, tensor<513xi16>) -> tensor<4x5xi16>
@@ -1166,8 +1177,8 @@ func.func @if_test_simple(%arg0 : tensor<f32>, %arg1 : tensor<f32>, %arg2 : tens
   %b = tosa.log %arg1 : (tensor<f32>) -> tensor<f32>
 
   // CHECK: tosa.cond_if
-  // CHECK: -> (tensor<f32>)
-  %0 = tosa.cond_if %arg2 -> (tensor<f32>) {
+  // CHECK: -> tensor<f32>
+  %0 = tosa.cond_if %arg2 : tensor<i1> -> tensor<f32> {
     tosa.yield %a : tensor<f32>
   } else {
     tosa.yield %b : tensor<f32>
@@ -1180,8 +1191,8 @@ func.func @if_test_simple(%arg0 : tensor<f32>, %arg1 : tensor<f32>, %arg2 : tens
 // CHECK-LABEL: @if_test_dynamic
 func.func @if_test_dynamic(%arg0 : tensor<2xf32>, %arg1 : tensor<3xf32>, %arg2 : tensor<i1>) -> () {
   // CHECK: tosa.cond_if
-  // CHECK: -> (tensor<?xf32>)
-  %0 = tosa.cond_if %arg2 -> (tensor<?xf32>) {
+  // CHECK: -> tensor<?xf32>
+  %0 = tosa.cond_if %arg2 : tensor<i1> -> tensor<?xf32> {
     tosa.yield %arg0 : tensor<2xf32>
   } else {
     tosa.yield %arg1 : tensor<3xf32>
@@ -1194,8 +1205,8 @@ func.func @if_test_dynamic(%arg0 : tensor<2xf32>, %arg1 : tensor<3xf32>, %arg2 :
 // CHECK-LABEL: @if_test_unranked
 func.func @if_test_unranked(%arg0 : tensor<f32>, %arg1 : tensor<3xf32>, %arg2 : tensor<i1>) -> () {
   // CHECK: tosa.cond_if
-  // CHECK: -> (tensor<*xf32>)
-  %0 = tosa.cond_if %arg2 -> (tensor<*xf32>) {
+  // CHECK: -> tensor<*xf32>
+  %0 = tosa.cond_if %arg2 : tensor<i1> -> tensor<*xf32> {
     tosa.yield %arg0 : tensor<f32>
   } else {
     tosa.yield %arg1 : tensor<3xf32>
@@ -1208,8 +1219,8 @@ func.func @if_test_unranked(%arg0 : tensor<f32>, %arg1 : tensor<3xf32>, %arg2 :
 // CHECK-LABEL: @if_test_propagate
 func.func @if_test_propagate(%arg0 : tensor<f32>, %arg1 : tensor<f32>, %arg2 : tensor<i1>) -> () {
   // CHECK: tosa.cond_if
-  // CHECK: -> (tensor<f32>)
-  %0 = tosa.cond_if %arg2 -> (tensor<f32>) {
+  // CHECK: -> tensor<f32>
+  %0 = tosa.cond_if %arg2 : tensor<i1> -> tensor<f32> {
     %1 = tosa.add %arg0, %arg1 : (tensor<f32>, tensor<f32>) -> tensor<f32>
     tosa.yield %1 : tensor<f32>
   } else {
diff --git a/mlir/test/Dialect/Tosa/verifier.mlir b/mlir/test/Dialect/Tosa/verifier.mlir
index b305236..2a937b0 100644
--- a/mlir/test/Dialect/Tosa/verifier.mlir
+++ b/mlir/test/Dialect/Tosa/verifier.mlir
@@ -500,9 +500,39 @@ func.func @test_cond_if_input_list_mismatch_else_block_2(%arg0: tensor<f32>, %ar
 
 // -----
 
+func.func @test_cond_if_input_list_mismatch_else_block_simple(%arg0: tensor<f32>, %arg1: tensor<f32>, %arg2: tensor<i1>) -> tensor<f32> {
+  // expected-error@+1 {{'tosa.cond_if' op require same number of values in 'else_graph' arguments (1) and 'input_list' (2)}}
+  %0 = tosa.cond_if %arg2 (%arg3 = %arg0, %arg4 = %arg1) : tensor<i1> (tensor<f32>, tensor<f32>) -> tensor<f32> {
+  ^bb0(%arg3: tensor<f32>, %arg4: tensor<f32>):
+    %1 = tosa.add %arg3, %arg4 : (tensor<f32>, tensor<f32>) -> tensor<f32>
+    tosa.yield %1 : tensor<f32>
+  } else {
+  ^bb0(%arg3: tensor<f32>):
+    tosa.yield %arg3 : tensor<f32>
+  }
+  return %0 : tensor<f32>
+}
+
+// -----
+
+func.func @test_cond_if_input_list_mismatch_else_block_simple_2(%arg0: tensor<f32>, %arg1: tensor<f32>, %arg2: tensor<i1>) -> tensor<f32> {
+  // expected-error@+1 {{'tosa.cond_if' op require same number of values in 'else_graph' arguments (2) and 'input_list' (1)}}
+  %0 = tosa.cond_if %arg2 (%arg3 = %arg0) : tensor<i1> (tensor<f32>) -> tensor<f32> {
+  ^bb0(%arg3: tensor<f32>):
+    tosa.yield %arg3 : tensor<f32>
+  } else {
+  ^bb0(%arg3: tensor<f32>, %arg4: tensor<f32>):
+    %1 = tosa.sub %arg3, %arg4 : (tensor<f32>, tensor<f32>) -> tensor<f32>
+    tosa.yield %1 : tensor<f32>
+  }
+  return %0 : tensor<f32>
+}
+
+// -----
+
 func.func @test_cond_if_output_list_mismatch_then_block(%arg0: tensor<f32>, %arg1: tensor<f32>, %arg2: tensor<i1>) -> tensor<f32> {
   // expected-error@+1 {{'tosa.cond_if' op require same number of values in 'then_graph' results (2) and 'output_list' (1)}}
-  %0 = tosa.cond_if %arg2 -> (tensor<f32>) {
+  %0 = tosa.cond_if %arg2 : tensor<i1> -> tensor<f32> {
     %1 = tosa.add %arg0, %arg1 : (tensor<f32>, tensor<f32>) -> tensor<f32>
     %2 = tosa.add %1, %arg1 : (tensor<f32>, tensor<f32>) -> tensor<f32>
     tosa.yield %1, %2 : tensor<f32>, tensor<f32>
@@ -517,7 +547,7 @@ func.func @test_cond_if_output_list_mismatch_then_block(%arg0: tensor<f32>, %arg
 
 func.func @test_cond_if_output_list_mismatch_then_block_2(%arg0: tensor<f32>, %arg1: tensor<f32>, %arg2: tensor<i1>) -> tensor<f32> {
   // expected-error@+1 {{'tosa.cond_if' op require same number of values in 'then_graph' results (1) and 'output_list' (2)}}
-  %0, %2 = tosa.cond_if %arg2 -> (tensor<f32>, tensor<f32>) {
+  %0, %2 = tosa.cond_if %arg2 : tensor<i1> -> (tensor<f32>, tensor<f32>) {
     %1 = tosa.add %arg0, %arg1 : (tensor<f32>, tensor<f32>) -> tensor<f32>
     tosa.yield %1 : tensor<f32>
   } else {
@@ -531,7 +561,7 @@ func.func @test_cond_if_output_list_mismatch_then_block_2(%arg0: tensor<f32>, %a
 
 func.func @test_cond_if_output_list_mismatch_else_block(%arg0: tensor<f32>, %arg1: tensor<f32>, %arg2: tensor<i1>) -> tensor<f32> {
   // expected-error@+1 {{'tosa.cond_if' op require same number of values in 'else_graph' results (2) and 'output_list' (1)}}
-  %0 = tosa.cond_if %arg2 -> (tensor<f32>) {
+  %0 = tosa.cond_if %arg2 : tensor<i1> -> tensor<f32> {
     %1 = tosa.add %arg0, %arg1 : (tensor<f32>, tensor<f32>) -> tensor<f32>
     tosa.yield %1 : tensor<f32>
   } else {
@@ -546,7 +576,7 @@ func.func @test_cond_if_output_list_mismatch_else_block(%arg0: tensor<f32>, %arg
 
 func.func @test_cond_if_output_list_mismatch_else_block_2(%arg0: tensor<f32>, %arg1: tensor<f32>, %arg2: tensor<i1>) -> tensor<f32> {
   // expected-error@+1 {{'tosa.cond_if' op require same number of values in 'else_graph' results (1) and 'output_list' (2)}}
-  %0, %2 = tosa.cond_if %arg2 -> (tensor<f32>, tensor<f32>) {
+  %0, %2 = tosa.cond_if %arg2 : tensor<i1> -> (tensor<f32>, tensor<f32>) {
     %1 = tosa.add %arg0, %arg1 : (tensor<f32>, tensor<f32>) -> tensor<f32>
     %2 = tosa.sub %arg0, %arg1 : (tensor<f32>, tensor<f32>) -> tensor<f32>
     tosa.yield %1, %2 : tensor<f32>, tensor<f32>
@@ -574,6 +604,53 @@ func.func @test_cond_if_cond_input_not_size_one(%arg0: tensor<f32>, %arg1: tenso
 
 // -----
 
+// CHECK-LABEL: cond_if_cond_type
+func.func @test_cond_if_cond_type(%arg0: tensor<f32>, %arg1: tensor<f32>, %arg2: tensor<i1>) -> tensor<f32> {
+  // expected-error@+2 {{expected ':'}}
+  // expected-error@+1 {{custom op 'tosa.cond_if' expected type for condition operand}}
+  %0 = tosa.cond_if %arg2 -> (tensor<f32>) {
+    tosa.yield %arg0 : tensor<f32>
+  } else {
+    tosa.yield %arg1 : tensor<f32>
+  }
+  return %0 : tensor<f32>
+}
+
+// -----
+
+func.func @test_cond_if_input_list_type_mismatch_simple(%arg0: tensor<f32>, %arg1: tensor<f32>, %arg2: tensor<i1>) -> tensor<f32> {
+  // expected-error@+1 {{custom op 'tosa.cond_if' expected as many input types as operands (expected 2 got 0)}}
+  %0 = tosa.cond_if %arg2 (%arg3 = %arg0, %arg4 = %arg1) : tensor<i1> () -> tensor<f32> {
+  ^bb0(%arg3: tensor<f32>, %arg4: tensor<f32>):
+    %1 = tosa.add %arg3, %arg4 : (tensor<f32>, tensor<f32>) -> tensor<f32>
+    tosa.yield %1 : tensor<f32>
+  } else {
+  ^bb0(%arg3: tensor<f32>, %arg4: tensor<f32>):
+    %1 = tosa.sub %arg3, %arg4 : (tensor<f32>, tensor<f32>) -> tensor<f32>
+    tosa.yield %1 : tensor<f32>
+  }
+  return %0 : tensor<f32>
+}
+
+// -----
+
+func.func @test_cond_if_incorrect_type_simple(%arg0: tensor<f32>, %arg1: tensor<f32>, %arg2: tensor<i1>) -> tensor<f32> {
+  // expected-error@+2 {{expected non-function type}}
+  // expected-error@+1 {{custom op 'tosa.cond_if' expected list of types for block arguments followed by arrow type and list of return types}}
+  %0 = tosa.cond_if %arg2 (%arg3 = %arg0, %arg4 = %arg1) : tensor<i1> (%arg3) -> tensor<f32> {
+  ^bb0(%arg3: tensor<f32>, %arg4: tensor<f32>):
+    %1 = tosa.add %arg3, %arg4 : (tensor<f32>, tensor<f32>) -> tensor<f32>
+    tosa.yield %1 : tensor<f32>
+  } else {
+  ^bb0(%arg3: tensor<f32>, %arg4: tensor<f32>):
+    %1 = tosa.sub %arg3, %arg4 : (tensor<f32>, tensor<f32>) -> tensor<f32>
+    tosa.yield %1 : tensor<f32>
+  }
+  return %0 : tensor<f32>
+}
+
+// -----
+
 func.func @test_while_loop_input_list_mismatch_body_block_in(%arg0: tensor<10xi32>, %arg1: tensor<i32>) {
   %0 = "tosa.const"() {values = dense<0> : tensor<i32>} : () -> tensor<i32>
   // expected-error@+1 {{'tosa.while_loop' op require same number of values in 'body_graph' arguments (3) and 'input_list' (2)}}
diff --git a/mlir/test/Examples/transform/Ch3/ops.mlir b/mlir/test/Examples/transform/Ch3/ops.mlir
index b2d47cc..707a09f 100644
--- a/mlir/test/Examples/transform/Ch3/ops.mlir
+++ b/mlir/test/Examples/transform/Ch3/ops.mlir
@@ -30,9 +30,30 @@ module attributes {transform.with_named_sequence} {
 // -----
 
 func.func private @orig()
+func.func private @updated()
 
 // CHECK-LABEL: func @test2
 func.func @test2() {
+  // CHECK: call @updated
+  call @orig() : () -> ()
+  return
+}
+
+module attributes {transform.with_named_sequence} {
+  transform.named_sequence @__transform_main(%arg0: !transform.any_op) {
+    %call = transform.structured.match ops{["func.call"]} in %arg0 : (!transform.any_op) -> !transform.my.call_op_interface
+    // CHECK: transform.my.change_call_target %{{.*}}, "updated" : !transform.my.call_op_interface
+    transform.my.change_call_target %call, "updated" : !transform.my.call_op_interface
+    transform.yield
+  }
+}
+
+// -----
+
+func.func private @orig()
+
+// CHECK-LABEL: func @test3
+func.func @test3() {
   // CHECK: "my.mm4"
   call @orig() : () -> ()
   return
diff --git a/mlir/test/Examples/transform/Ch3/sequence.mlir b/mlir/test/Examples/transform/Ch3/sequence.mlir
index 4d28518..877b006 100644
--- a/mlir/test/Examples/transform/Ch3/sequence.mlir
+++ b/mlir/test/Examples/transform/Ch3/sequence.mlir
@@ -101,11 +101,12 @@ module attributes {transform.with_named_sequence} {
     %_1, %outline_target = transform.structured.fuse_into_containing_op %matmul_fused_2 into %loop_third
         : (!transform.any_op, !transform.any_op) -> (!transform.any_op, !transform.any_op)
     %func, %call = transform.loop.outline %outline_target {func_name = "outlined"}
-        : (!transform.any_op) -> (!transform.any_op, !transform.op<"func.call">)
-  
-    // Rewrite the call target.
-    transform.my.change_call_target %call, "microkernel" : !transform.op<"func.call">
-  
+        : (!transform.any_op) -> (!transform.any_op, !transform.any_op)
+    // Cast to our new type.
+    %casted = transform.cast %call : !transform.any_op to !transform.my.call_op_interface
+    // Using our new operation.
+    transform.my.change_call_target %casted, "microkernel" : !transform.my.call_op_interface
+
     transform.yield
   }
 }
diff --git a/mlir/test/Target/LLVMIR/omptarget-debug-reduc-fn-loc.mlir b/mlir/test/Target/LLVMIR/omptarget-debug-reduc-fn-loc.mlir
deleted file mode 100644
index d889ef4..0000000
--- a/mlir/test/Target/LLVMIR/omptarget-debug-reduc-fn-loc.mlir
+++ /dev/null
@@ -1,121 +0,0 @@
-// RUN: mlir-translate -mlir-to-llvmir %s | FileCheck %s
-
-module attributes {dlti.dl_spec = #dlti.dl_spec<#dlti.dl_entry<"dlti.alloca_memory_space", 5 : ui32>>, llvm.target_triple = "amdgcn-amd-amdhsa", omp.is_gpu = true, omp.is_target_device = true} {
-  omp.private {type = private} @_QFEi_private_i32 : i32 loc(#loc1)
-  omp.declare_reduction @add_reduction_i32 : i32 init {
-  ^bb0(%arg0: i32 loc("test.f90":8:7)):
-    %0 = llvm.mlir.constant(0 : i32) : i32 loc(#loc2)
-    omp.yield(%0 : i32) loc(#loc2)
-  } combiner {
-  ^bb0(%arg0: i32 loc("test.f90":8:7), %arg1: i32 loc("test.f90":8:7)):
-    %0 = llvm.add %arg0, %arg1 : i32 loc(#loc2)
-    omp.yield(%0 : i32) loc(#loc2)
-  } loc(#loc2)
-  llvm.func @_QQmain() {
-    %0 = llvm.mlir.constant(1 : i64) : i64 loc(#loc4)
-    %1 = llvm.alloca %0 x i32 {bindc_name = "x"} : (i64) -> !llvm.ptr<5> loc(#loc4)
-    %2 = llvm.addrspacecast %1 : !llvm.ptr<5> to !llvm.ptr loc(#loc4)
-    %3 = llvm.mlir.constant(1 : i64) : i64 loc(#loc1)
-    %4 = llvm.alloca %3 x i32 {bindc_name = "i"} : (i64) -> !llvm.ptr<5> loc(#loc1)
-    %5 = llvm.addrspacecast %4 : !llvm.ptr<5> to !llvm.ptr loc(#loc1)
-    %6 = llvm.mlir.constant(8191 : index) : i64 loc(#loc5)
-    %7 = llvm.mlir.constant(0 : index) : i64 loc(#loc5)
-    %8 = llvm.mlir.constant(1 : index) : i64 loc(#loc5)
-    %9 = llvm.mlir.constant(0 : i32) : i32 loc(#loc5)
-    %10 = llvm.mlir.constant(8192 : index) : i64 loc(#loc5)
-    %11 = llvm.mlir.addressof @_QFEarr : !llvm.ptr<1> loc(#loc6)
-    %12 = llvm.addrspacecast %11 : !llvm.ptr<1> to !llvm.ptr loc(#loc6)
-    llvm.store %9, %2 : i32, !llvm.ptr loc(#loc7)
-    %15 = omp.map.info var_ptr(%2 : !llvm.ptr, i32) map_clauses(tofrom) capture(ByRef) -> !llvm.ptr {name = "x"} loc(#loc4)
-    %16 = omp.map.info var_ptr(%5 : !llvm.ptr, i32) map_clauses(implicit, exit_release_or_enter_alloc) capture(ByCopy) -> !llvm.ptr {name = "i"} loc(#loc7)
-    %17 = omp.map.bounds lower_bound(%7 : i64) upper_bound(%6 : i64) extent(%10 : i64) stride(%8 : i64) start_idx(%8 : i64) loc(#loc7)
-    %18 = omp.map.info var_ptr(%12 : !llvm.ptr, !llvm.array<8192 x i32>) map_clauses(implicit, tofrom) capture(ByRef) bounds(%17) -> !llvm.ptr {name = "arr"} loc(#loc7)
-    omp.target map_entries(%15 -> %arg0, %16 -> %arg1, %18 -> %arg2 : !llvm.ptr, !llvm.ptr, !llvm.ptr) {
-      %19 = llvm.mlir.constant(8192 : i32) : i32 loc(#loc5)
-      %20 = llvm.mlir.constant(1 : i32) : i32 loc(#loc5)
-      %21 = llvm.mlir.constant(8192 : index) : i64 loc(#loc6)
-      omp.teams reduction(@add_reduction_i32 %arg0 -> %arg3 : !llvm.ptr) {
-        omp.parallel private(@_QFEi_private_i32 %arg1 -> %arg4 : !llvm.ptr) {
-          omp.distribute {
-            omp.wsloop reduction(@add_reduction_i32 %arg3 -> %arg5 : !llvm.ptr) {
-              omp.loop_nest (%arg6) : i32 = (%20) to (%19) inclusive step (%20) {
-                llvm.store %arg6, %arg4 : i32, !llvm.ptr loc(#loc2)
-                %22 = llvm.load %arg5 : !llvm.ptr -> i32 loc(#loc8)
-                %23 = llvm.load %arg4 : !llvm.ptr -> i32 loc(#loc8)
-                %34 = llvm.add %22, %23 : i32 loc(#loc8)
-                llvm.store %34, %arg5 : i32, !llvm.ptr loc(#loc8)
-                omp.yield loc(#loc2)
-              } loc(#loc2)
-            } {omp.composite} loc(#loc2)
-          } {omp.composite} loc(#loc2)
-          omp.terminator loc(#loc2)
-        } {omp.composite} loc(#loc2)
-        omp.terminator loc(#loc2)
-      } loc(#loc2)
-      omp.terminator loc(#loc2)
-    } loc(#loc13)
-    llvm.return loc(#loc9)
-  } loc(#loc12)
-  llvm.mlir.global internal @_QFEarr() {addr_space = 1 : i32} : !llvm.array<8192 x i32> {
-    %0 = llvm.mlir.zero : !llvm.array<8192 x i32> loc(#loc6)
-    llvm.return %0 : !llvm.array<8192 x i32> loc(#loc6)
-  } loc(#loc6)
-} loc(#loc)
-
-#loc = loc("test.f90":4:18)
-#loc1 = loc("test.f90":4:18)
-#loc2 = loc("test.f90":8:7)
-#loc3 = loc("test.f90":1:7)
-#loc4 = loc("test.f90":3:18)
-#loc5 = loc(unknown)
-#loc6 = loc("test.f90":5:18)
-#loc7 = loc("test.f90":6:7)
-#loc8 = loc("test.f90":10:7)
-#loc9 = loc("test.f90":16:7)
-
-#di_file = #llvm.di_file<"target7.f90" in "">
-#di_null_type = #llvm.di_null_type
-#di_compile_unit = #llvm.di_compile_unit<id = distinct[0]<>,
- sourceLanguage = DW_LANG_Fortran95, file = #di_file, producer = "flang",
- isOptimized = false, emissionKind = LineTablesOnly>
-#di_subroutine_type = #llvm.di_subroutine_type<
-  callingConvention = DW_CC_program, types = #di_null_type>
-#di_subprogram = #llvm.di_subprogram<id = distinct[1]<>,
-  compileUnit = #di_compile_unit, scope = #di_file, name = "main",
-  file = #di_file, subprogramFlags = "Definition|MainSubprogram",
-  type = #di_subroutine_type>
-#di_subprogram1 = #llvm.di_subprogram<compileUnit = #di_compile_unit,
-  name = "target", file = #di_file, subprogramFlags = "Definition",
-  type = #di_subroutine_type>
-
-
-#loc12 = loc(fused<#di_subprogram>[#loc3])
-#loc13 = loc(fused<#di_subprogram1>[#loc2])
-
-// CHECK-DAG: define internal void @_omp_reduction_shuffle_and_reduce_func
-// CHECK-NOT: !dbg
-// CHECK: }
-// CHECK-DAG: define internal void @_omp_reduction_inter_warp_copy_func
-// CHECK-NOT: !dbg
-// CHECK: }
-// CHECK-DAG: define internal void @"__omp_offloading_{{.*}}__QQmain_l8_omp$reduction$reduction_func.1"
-// CHECK-NOT: !dbg
-// CHECK: }
-// CHECK-DAG: define internal void @_omp_reduction_shuffle_and_reduce_func.2
-// CHECK-NOT: !dbg
-// CHECK: }
-// CHECK-DAG: define internal void @_omp_reduction_inter_warp_copy_func.3
-// CHECK-NOT: !dbg
-// CHECK: }
-// CHECK-DAG: define internal void @_omp_reduction_list_to_global_copy_func
-// CHECK-NOT: !dbg
-// CHECK: }
-// CHECK-DAG: define internal void @_omp_reduction_list_to_global_reduce_func
-// CHECK-NOT: !dbg
-// CHECK: }
-// CHECK-DAG: define internal void @_omp_reduction_global_to_list_copy_func
-// CHECK-NOT: !dbg
-// CHECK: }
-// CHECK-DAG: define internal void @_omp_reduction_global_to_list_reduce_func
-// CHECK-NOT: !dbg
-// CHECK: }
diff --git a/mlir/test/Transforms/compose-subview.mlir b/mlir/test/Transforms/compose-subview.mlir
index 53fbb8a..d6fa442 100644
--- a/mlir/test/Transforms/compose-subview.mlir
+++ b/mlir/test/Transforms/compose-subview.mlir
@@ -1,9 +1,9 @@
 // RUN: mlir-opt %s -test-compose-subview -split-input-file | FileCheck %s
 
 // CHECK-LABEL: func.func @subview_strided(
-// CHECK-SAME:  %[[VAL_0:.*]]: memref<4x1024xf32>) -> memref<1x128xf32, strided<[1024, 1], offset: 3456>> {
+// CHECK-SAME: %[[input:.*]]: memref<4x1024xf32>) -> memref<1x128xf32, strided<[1024, 1], offset: 3456>> {
 func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x128xf32, strided<[1024, 1], offset: 3456>> {
-  // CHECK: %[[VAL_1:.*]] = memref.subview %[[VAL_0]][3, 384] [1, 128] [1, 1] : memref<4x1024xf32> to memref<1x128xf32, strided<[1024, 1], offset: 3456>>
+  // CHECK: {{.*}} = memref.subview %[[input]][3, 384] [1, 128] [1, 1] : memref<4x1024xf32> to memref<1x128xf32, strided<[1024, 1], offset: 3456>>
   %0 = memref.subview %input[2, 256] [2, 256] [1, 1] : memref<4x1024xf32> to memref<2x256xf32, strided<[1024, 1], offset: 2304>>
   %1 = memref.subview %0[1, 128] [1, 128] [1, 1] : memref<2x256xf32, strided<[1024, 1], offset: 2304>> to memref<1x128xf32, strided<[1024, 1], offset: 3456>>
   return %1 : memref<1x128xf32, strided<[1024, 1], offset: 3456>>
@@ -12,9 +12,9 @@ func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x128xf32, stri
 // -----
 
 // CHECK-LABEL: func.func @subview_strided(
-// CHECK-SAME:  %[[VAL_0:.*]]: memref<4x1024xf32>) -> memref<1x10xf32, strided<[1024, 1], offset: 3745>> {
+// CHECK-SAME:  %[[input:.*]]: memref<4x1024xf32>) -> memref<1x10xf32, strided<[1024, 1], offset: 3745>> {
 func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x10xf32, strided<[1024, 1], offset: 3745>> {
-  // CHECK:     %[[VAL_1:.*]] = memref.subview %[[VAL_0]][3, 673] [1, 10] [1, 1] : memref<4x1024xf32> to memref<1x10xf32, strided<[1024, 1], offset: 3745>>
+  // CHECK:     {{.*}} = memref.subview %[[input]][3, 673] [1, 10] [1, 1] : memref<4x1024xf32> to memref<1x10xf32, strided<[1024, 1], offset: 3745>>
   %0 = memref.subview %input[1, 512] [3, 256] [1, 1] : memref<4x1024xf32> to memref<3x256xf32, strided<[1024, 1], offset: 1536>>
   %1 = memref.subview %0[1, 128] [2, 128] [1, 1] : memref<3x256xf32, strided<[1024, 1], offset: 1536>> to memref<2x128xf32, strided<[1024, 1], offset: 2688>>
   %2 = memref.subview %1[1, 33] [1, 10] [1, 1] : memref<2x128xf32, strided<[1024, 1], offset: 2688>> to memref<1x10xf32, strided<[1024, 1], offset: 3745>>
@@ -24,12 +24,12 @@ func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x10xf32, strid
 // -----
 
 // CHECK-LABEL: func.func @subview_strided(
-// CHECK-SAME:  %[[VAL_0:.*]]: memref<4x1024xf32>) -> memref<1x128xf32, strided<[1024, 1], offset: ?>> {
+// CHECK-SAME:  %[[input:.*]]: memref<4x1024xf32>) -> memref<1x128xf32, strided<[1024, 1], offset: ?>> {
 func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x128xf32, strided<[1024, 1], offset: ?>> {
-  // CHECK: %[[VAL_1:.*]] = arith.constant 3 : index
+  // CHECK: %[[C3:.*]] = arith.constant 3 : index
   %cst_1 = arith.constant 1 : index
   %cst_2 = arith.constant 2 : index
-  // CHECK: %[[VAL_2:.*]] = memref.subview %[[VAL_0]]{{\[}}%[[VAL_1]], 384] [1, 128] [1, 1] : memref<4x1024xf32> to memref<1x128xf32, strided<[1024, 1], offset: ?>>
+  // CHECK: {{.*}} = memref.subview %[[input]]{{\[}}%[[C3]], 384] [1, 128] [1, 1] : memref<4x1024xf32> to memref<1x128xf32, strided<[1024, 1], offset: ?>>
   %0 = memref.subview %input[%cst_2, 256] [2, 256] [1, 1] : memref<4x1024xf32> to memref<2x256xf32, strided<[1024, 1], offset: ?>>
   %1 = memref.subview %0[%cst_1, 128] [1, 128] [1, 1] : memref<2x256xf32, strided<[1024, 1], offset: ?>> to memref<1x128xf32, strided<[1024, 1], offset: ?>>
   return %1 : memref<1x128xf32, strided<[1024, 1], offset: ?>>
@@ -38,13 +38,13 @@ func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x128xf32, stri
 // -----
 
 // CHECK-LABEL: func.func @subview_strided(
-// CHECK-SAME:  %[[VAL_0:.*]]: memref<4x1024xf32>) -> memref<1x128xf32, strided<[1024, 1], offset: ?>> {
+// CHECK-SAME:  %[[input:.*]]: memref<4x1024xf32>) -> memref<1x128xf32, strided<[1024, 1], offset: ?>> {
 func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x128xf32, strided<[1024, 1], offset: ?>> {
-  // CHECK: %[[VAL_1:.*]] = arith.constant 3 : index
+  // CHECK: %[[C3:.*]] = arith.constant 3 : index
   %cst_2 = arith.constant 2 : index
-  // CHECK: %[[VAL_2:.*]] = arith.constant 384 : index
+  // CHECK: %[[C384:.*]] = arith.constant 384 : index
   %cst_128 = arith.constant 128 : index
-  // CHECK: %[[VAL_3:.*]] = memref.subview %[[VAL_0]]{{\[}}%[[VAL_1]], %[[VAL_2]]] [1, 128] [1, 1] : memref<4x1024xf32> to memref<1x128xf32, strided<[1024, 1], offset: ?>>
+  // CHECK: {{.*}} = memref.subview %[[input]]{{\[}}%[[C3]], %[[C384]]] [1, 128] [1, 1] : memref<4x1024xf32> to memref<1x128xf32, strided<[1024, 1], offset: ?>>
   %0 = memref.subview %input[%cst_2, 256] [2, 256] [1, 1] : memref<4x1024xf32> to memref<2x256xf32, strided<[1024, 1], offset: ?>>
   %1 = memref.subview %0[1, %cst_128] [1, 128] [1, 1] : memref<2x256xf32, strided<[1024, 1], offset: ?>> to memref<1x128xf32, strided<[1024, 1], offset: ?>>
   return %1 : memref<1x128xf32, strided<[1024, 1], offset: ?>>
@@ -53,9 +53,9 @@ func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x128xf32, stri
 // -----
 
 // CHECK-LABEL: func.func @subview_strided(
-// CHECK-SAME:  %[[VAL_0:.*]]: memref<8x1024xf32>) -> memref<1x64xf32, strided<[4096, 4], offset: 4480>> {
+// CHECK-SAME:  %[[input:.*]]: memref<8x1024xf32>) -> memref<1x64xf32, strided<[4096, 4], offset: 4480>> {
 func.func @subview_strided(%input: memref<8x1024xf32>) -> memref<1x64xf32, strided<[4096, 4], offset: 4480>> {
-  // CHECK:     %[[VAL_1:.*]] = memref.subview %[[VAL_0]][4, 384] [1, 64] [4, 4] : memref<8x1024xf32> to memref<1x64xf32, strided<[4096, 4], offset: 4480>>
+  // CHECK:     {{.*}} = memref.subview %[[input]][4, 384] [1, 64] [4, 4] : memref<8x1024xf32> to memref<1x64xf32, strided<[4096, 4], offset: 4480>>
   %0 = memref.subview %input[2, 256] [2, 256] [2, 2] : memref<8x1024xf32> to memref<2x256xf32, strided<[2048, 2], offset: 2304>>
   %1 = memref.subview %0[1, 64] [1, 64] [2, 2] : memref<2x256xf32, strided<[2048, 2], offset: 2304>> to memref<1x64xf32, strided<[4096, 4], offset: 4480>>
   return %1 : memref<1x64xf32, strided<[4096, 4], offset: 4480>>
@@ -64,9 +64,9 @@ func.func @subview_strided(%input: memref<8x1024xf32>) -> memref<1x64xf32, strid
 // -----
 
 // CHECK-LABEL: func.func @subview_strided(
-// CHECK-SAME:  %[[VAL_0:.*]]: memref<30x30xf32>) -> memref<2x2xf32, strided<[240, 8], offset: 217>> {
+// CHECK-SAME:  %[[input:.*]]: memref<30x30xf32>) -> memref<2x2xf32, strided<[240, 8], offset: 217>> {
 func.func @subview_strided(%input: memref<30x30xf32>) -> memref<2x2xf32, strided<[240, 8], offset: 217>> {
-  // CHECK:     %[[VAL_1:.*]] = memref.subview %[[VAL_0]][7, 7] [2, 2] [8, 8] : memref<30x30xf32> to memref<2x2xf32, strided<[240, 8], offset: 217>>
+  // CHECK:     {{.*}} = memref.subview %[[input]][7, 7] [2, 2] [8, 8] : memref<30x30xf32> to memref<2x2xf32, strided<[240, 8], offset: 217>>
   %0 = memref.subview %input[1, 1] [12, 12] [2, 2] : memref<30x30xf32> to memref<12x12xf32, strided<[60, 2], offset: 31>>
   %1 = memref.subview %0[1, 1] [5, 5] [2, 2] : memref<12x12xf32, strided<[60, 2], offset: 31>> to memref<5x5xf32, strided<[120, 4], offset: 93>>
   %2 = memref.subview %1[1, 1] [2, 2] [2, 2] : memref<5x5xf32, strided<[120, 4], offset: 93>> to memref<2x2xf32, strided<[240, 8], offset: 217>>
@@ -76,13 +76,13 @@ func.func @subview_strided(%input: memref<30x30xf32>) -> memref<2x2xf32, strided
 // -----
 
 // CHECK-LABEL: func.func @subview_strided(
-// CHECK-SAME:  %[[VAL_0:.*]]: memref<4x1024xf32>) -> memref<1x64xf32, strided<[4096, 4], offset: ?>> {
+// CHECK-SAME:  %[[input:.*]]: memref<4x1024xf32>) -> memref<1x64xf32, strided<[4096, 4], offset: ?>> {
 func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x64xf32, strided<[4096, 4], offset: ?>> {
-  // CHECK:     %[[VAL_1:.*]] = arith.constant 4 : index
+  // CHECK:     %[[C4:.*]] = arith.constant 4 : index
   %cst_2 = arith.constant 2 : index
-  // CHECK:     %[[VAL_2:.*]] = arith.constant 384 : index
+  // CHECK:     %[[C384:.*]] = arith.constant 384 : index
   %cst_64 = arith.constant 64 : index
-  // CHECK:     %[[VAL_3:.*]] = memref.subview %[[VAL_0]]{{\[}}%[[VAL_1]], %[[VAL_2]]] [1, 64] [4, 4] : memref<4x1024xf32> to memref<1x64xf32, strided<[4096, 4], offset: ?>>
+  // CHECK:     {{.*}} = memref.subview %[[input]]{{\[}}%[[C4]], %[[C384]]] [1, 64] [4, 4] : memref<4x1024xf32> to memref<1x64xf32, strided<[4096, 4], offset: ?>>
   %0 = memref.subview %input[%cst_2, 256] [2, 256] [2, 2] : memref<4x1024xf32> to memref<2x256xf32, strided<[2048, 2], offset: ?>>
   %1 = memref.subview %0[1, %cst_64] [1, 64] [2, 2] : memref<2x256xf32, strided<[2048, 2], offset: ?>> to memref<1x64xf32, strided<[4096, 4], offset: ?>>
   return %1 : memref<1x64xf32, strided<[4096, 4], offset: ?>>
@@ -91,13 +91,39 @@ func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x64xf32, strid
 // -----
 
 // CHECK-LABEL: func.func @subview_strided(
-// CHECK-SAME:  %[[VAL_0:.*]]: memref<4x1024xf32>) -> memref<1x64xf32, strided<[4096, 4], offset: ?>> {
+// CHECK-SAME:  %[[input:.*]]: memref<4x1024xf32>) -> memref<1x64xf32, strided<[4096, 4], offset: ?>> {
 func.func @subview_strided(%input: memref<4x1024xf32>) -> memref<1x64xf32, strided<[4096, 4], offset: ?>> {
-  // CHECK:     %[[VAL_1:.*]] = arith.constant 4 : index
+  // CHECK:     %[[C4:.*]] = arith.constant 4 : index
   %cst_1 = arith.constant 1 : index
   %cst_2 = arith.constant 2 : index
-  // CHECK:     %[[VAL_2:.*]] = memref.subview %[[VAL_0]]{{\[}}%[[VAL_1]], 384] [1, 64] [4, 4] : memref<4x1024xf32> to memref<1x64xf32, strided<[4096, 4], offset: ?>>
+  // CHECK:     {{.*}} = memref.subview %[[input]]{{\[}}%[[C4]], 384] [1, 64] [4, 4] : memref<4x1024xf32> to memref<1x64xf32, strided<[4096, 4], offset: ?>>
   %0 = memref.subview %input[%cst_2, 256] [2, 256] [2, 2] : memref<4x1024xf32> to memref<2x256xf32, strided<[2048, 2], offset: ?>>
   %1 = memref.subview %0[%cst_1, 64] [1, 64] [2, 2] : memref<2x256xf32, strided<[2048, 2], offset: ?>> to memref<1x64xf32, strided<[4096, 4], offset: ?>>
   return %1 : memref<1x64xf32, strided<[4096, 4], offset: ?>>
 }
+
+// -----
+
+// CHECK-LABEL: func.func @single_dynamic_size_subview(
+// CHECK-SAME:  %[[input:.*]]: memref<256x?xf32>,
+// CHECK-SAME:  %{{.*}}: index,
+// CHECK-SAME:  %[[SIZE_1:.*]]: index) -> memref<8x?xf32> {
+func.func @single_dynamic_size_subview(%input: memref<256x?xf32>, %size0 : index, %size1 : index) -> memref<8x?xf32>{
+  %subview = memref.subview %input[0, 0][8, %size0][1, 1] : memref<256x?xf32> to memref<8x?xf32> 
+  %subview_1 = memref.subview %subview[0, 0][8, %size1][1, 1] : memref<8x?xf32> to memref<8x?xf32>
+  // CHECK:  %{{.*}} = memref.subview %[[input]][0, 0] [8, %[[SIZE_1]]] [1, 1] : memref<256x?xf32> to memref<8x?xf32>
+  return %subview_1 : memref<8x?xf32>
+}
+
+// -----
+
+// CHECK-LABEL: func.func @all_dynamic_size_subview(
+// CHECK-SAME:  %[[input:.*]]: memref<256x?xf32>,
+// CHECK-SAME:  %{{.*}}: index,
+// CHECK-SAME:  %[[SIZE1:.*]]: index) -> memref<?x?xf32> {
+func.func @all_dynamic_size_subview(%input: memref<256x?xf32>, %size0 : index, %size1 : index) -> memref<?x?xf32>{
+  %subview = memref.subview %input[0, 0][%size0, %size0][1, 1] : memref<256x?xf32> to memref<?x?xf32> 
+  %subview_1 = memref.subview %subview[0, 0][%size1, %size1][1, 1] : memref<?x?xf32> to memref<?x?xf32>
+  // CHECK:  {{.*}} = memref.subview %[[input]][0, 0] {{\[}}%[[SIZE1]], %[[SIZE1]]] [1, 1] : memref<256x?xf32> to memref<?x?xf32>
+  return %subview_1 : memref<?x?xf32>
+}
diff --git a/mlir/test/Transforms/test-legalize-type-conversion.mlir b/mlir/test/Transforms/test-legalize-type-conversion.mlir
index db8bd0f..9bffe92 100644
--- a/mlir/test/Transforms/test-legalize-type-conversion.mlir
+++ b/mlir/test/Transforms/test-legalize-type-conversion.mlir
@@ -104,8 +104,8 @@ func.func @test_signature_conversion_no_converter() {
   "test.signature_conversion_no_converter"() ({
   // expected-error@below {{failed to legalize unresolved materialization from ('f64') to ('f32') that remained live after conversion}}
   ^bb0(%arg0: f32):
-    "test.type_consumer"(%arg0) : (f32) -> ()
     // expected-note@below{{see existing live user here}}
+    "test.type_consumer"(%arg0) : (f32) -> ()
     "test.return"(%arg0) : (f32) -> ()
   }) : () -> ()
   return
diff --git a/mlir/test/Transforms/test-legalizer-analysis.mlir b/mlir/test/Transforms/test-legalizer-analysis.mlir
index 19a1310..5b07055 100644
--- a/mlir/test/Transforms/test-legalizer-analysis.mlir
+++ b/mlir/test/Transforms/test-legalizer-analysis.mlir
@@ -1,4 +1,4 @@
-// RUN: mlir-opt -allow-unregistered-dialect -test-legalize-patterns -verify-diagnostics -test-legalize-mode=analysis %s | FileCheck %s
+// RUN: mlir-opt -allow-unregistered-dialect -test-legalize-patterns="test-legalize-mode=analysis" -verify-diagnostics %s | FileCheck %s
 // expected-remark@-2 {{op 'builtin.module' is legalizable}}
 
 // expected-remark@+1 {{op 'func.func' is legalizable}}
diff --git a/mlir/test/Transforms/test-legalizer-full.mlir b/mlir/test/Transforms/test-legalizer-full.mlir
index 5f1148c..dcd0172 100644
--- a/mlir/test/Transforms/test-legalizer-full.mlir
+++ b/mlir/test/Transforms/test-legalizer-full.mlir
@@ -1,4 +1,4 @@
-// RUN: mlir-opt -allow-unregistered-dialect -test-legalize-patterns -test-legalize-mode=full -split-input-file -verify-diagnostics %s | FileCheck %s
+// RUN: mlir-opt -allow-unregistered-dialect -test-legalize-patterns="test-legalize-mode=full" -split-input-file -verify-diagnostics %s | FileCheck %s
 
 // CHECK-LABEL: func @multi_level_mapping
 func.func @multi_level_mapping() {
diff --git a/mlir/test/lib/Conversion/MathToVCIX/TestMathToVCIXConversion.cpp b/mlir/test/lib/Conversion/MathToVCIX/TestMathToVCIXConversion.cpp
index 8a01a0a..016052c 100644
--- a/mlir/test/lib/Conversion/MathToVCIX/TestMathToVCIXConversion.cpp
+++ b/mlir/test/lib/Conversion/MathToVCIX/TestMathToVCIXConversion.cpp
@@ -69,25 +69,25 @@ struct MathCosToVCIX final : OpRewritePattern<math::CosOp> {
     if (legalType.isScalable())
       // Use arbitrary runtime vector length when vector type is scalable.
       // Proper conversion pass should take it from the IR.
-      rvl = rewriter.create<arith::ConstantOp>(loc,
-                                               rewriter.getI64IntegerAttr(9));
+      rvl = arith::ConstantOp::create(rewriter, loc,
+                                      rewriter.getI64IntegerAttr(9));
     Value res;
     if (n == 1) {
-      res = rewriter.create<vcix::BinaryImmOp>(loc, legalType, opcodeAttr, vec,
-                                               immAttr, rvl);
+      res = vcix::BinaryImmOp::create(rewriter, loc, legalType, opcodeAttr, vec,
+                                      immAttr, rvl);
     } else {
       const unsigned eltCount = legalType.getShape()[0];
       Type eltTy = legalType.getElementType();
-      Value zero = rewriter.create<arith::ConstantOp>(
-          loc, eltTy, rewriter.getZeroAttr(eltTy));
-      res = rewriter.create<vector::BroadcastOp>(loc, opType, zero /*dummy*/);
+      Value zero = arith::ConstantOp::create(rewriter, loc, eltTy,
+                                             rewriter.getZeroAttr(eltTy));
+      res = vector::BroadcastOp::create(rewriter, loc, opType, zero /*dummy*/);
       for (unsigned i = 0; i < n; ++i) {
-        Value extracted = rewriter.create<vector::ScalableExtractOp>(
-            loc, legalType, vec, i * eltCount);
-        Value v = rewriter.create<vcix::BinaryImmOp>(loc, legalType, opcodeAttr,
-                                                     extracted, immAttr, rvl);
-        res = rewriter.create<vector::ScalableInsertOp>(loc, v, res,
-                                                        i * eltCount);
+        Value extracted = vector::ScalableExtractOp::create(
+            rewriter, loc, legalType, vec, i * eltCount);
+        Value v = vcix::BinaryImmOp::create(
+            rewriter, loc, legalType, opcodeAttr, extracted, immAttr, rvl);
+        res = vector::ScalableInsertOp::create(rewriter, loc, v, res,
+                                               i * eltCount);
       }
     }
     rewriter.replaceOp(op, res);
@@ -112,25 +112,25 @@ struct MathSinToVCIX final : OpRewritePattern<math::SinOp> {
     if (legalType.isScalable())
       // Use arbitrary runtime vector length when vector type is scalable.
       // Proper conversion pass should take it from the IR.
-      rvl = rewriter.create<arith::ConstantOp>(loc,
-                                               rewriter.getI64IntegerAttr(9));
+      rvl = arith::ConstantOp::create(rewriter, loc,
+                                      rewriter.getI64IntegerAttr(9));
     Value res;
     if (n == 1) {
-      res = rewriter.create<vcix::BinaryOp>(loc, legalType, opcodeAttr, vec,
-                                            vec, rvl);
+      res = vcix::BinaryOp::create(rewriter, loc, legalType, opcodeAttr, vec,
+                                   vec, rvl);
     } else {
       const unsigned eltCount = legalType.getShape()[0];
       Type eltTy = legalType.getElementType();
-      Value zero = rewriter.create<arith::ConstantOp>(
-          loc, eltTy, rewriter.getZeroAttr(eltTy));
-      res = rewriter.create<vector::BroadcastOp>(loc, opType, zero /*dummy*/);
+      Value zero = arith::ConstantOp::create(rewriter, loc, eltTy,
+                                             rewriter.getZeroAttr(eltTy));
+      res = vector::BroadcastOp::create(rewriter, loc, opType, zero /*dummy*/);
       for (unsigned i = 0; i < n; ++i) {
-        Value extracted = rewriter.create<vector::ScalableExtractOp>(
-            loc, legalType, vec, i * eltCount);
-        Value v = rewriter.create<vcix::BinaryOp>(loc, legalType, opcodeAttr,
-                                                  extracted, extracted, rvl);
-        res = rewriter.create<vector::ScalableInsertOp>(loc, v, res,
-                                                        i * eltCount);
+        Value extracted = vector::ScalableExtractOp::create(
+            rewriter, loc, legalType, vec, i * eltCount);
+        Value v = vcix::BinaryOp::create(rewriter, loc, legalType, opcodeAttr,
+                                         extracted, extracted, rvl);
+        res = vector::ScalableInsertOp::create(rewriter, loc, v, res,
+                                               i * eltCount);
       }
     }
     rewriter.replaceOp(op, res);
@@ -152,28 +152,28 @@ struct MathTanToVCIX final : OpRewritePattern<math::TanOp> {
     Location loc = op.getLoc();
     Value vec = op.getOperand();
     Attribute opcodeAttr = rewriter.getI64IntegerAttr(0);
-    Value zero = rewriter.create<arith::ConstantOp>(
-        loc, eltTy, rewriter.getZeroAttr(eltTy));
+    Value zero = arith::ConstantOp::create(rewriter, loc, eltTy,
+                                           rewriter.getZeroAttr(eltTy));
     Value rvl = nullptr;
     if (legalType.isScalable())
       // Use arbitrary runtime vector length when vector type is scalable.
       // Proper conversion pass should take it from the IR.
-      rvl = rewriter.create<arith::ConstantOp>(loc,
-                                               rewriter.getI64IntegerAttr(9));
+      rvl = arith::ConstantOp::create(rewriter, loc,
+                                      rewriter.getI64IntegerAttr(9));
     Value res;
     if (n == 1) {
-      res = rewriter.create<vcix::BinaryOp>(loc, legalType, opcodeAttr, vec,
-                                            zero, rvl);
+      res = vcix::BinaryOp::create(rewriter, loc, legalType, opcodeAttr, vec,
+                                   zero, rvl);
     } else {
       const unsigned eltCount = legalType.getShape()[0];
-      res = rewriter.create<vector::BroadcastOp>(loc, opType, zero /*dummy*/);
+      res = vector::BroadcastOp::create(rewriter, loc, opType, zero /*dummy*/);
       for (unsigned i = 0; i < n; ++i) {
-        Value extracted = rewriter.create<vector::ScalableExtractOp>(
-            loc, legalType, vec, i * eltCount);
-        Value v = rewriter.create<vcix::BinaryOp>(loc, legalType, opcodeAttr,
-                                                  extracted, zero, rvl);
-        res = rewriter.create<vector::ScalableInsertOp>(loc, v, res,
-                                                        i * eltCount);
+        Value extracted = vector::ScalableExtractOp::create(
+            rewriter, loc, legalType, vec, i * eltCount);
+        Value v = vcix::BinaryOp::create(rewriter, loc, legalType, opcodeAttr,
+                                         extracted, zero, rvl);
+        res = vector::ScalableInsertOp::create(rewriter, loc, v, res,
+                                               i * eltCount);
       }
     }
     rewriter.replaceOp(op, res);
@@ -195,30 +195,30 @@ struct MathLogToVCIX final : OpRewritePattern<math::LogOp> {
     Value vec = op.getOperand();
     Attribute opcodeAttr = rewriter.getI64IntegerAttr(0);
     Value rvl = nullptr;
-    Value zeroInt = rewriter.create<arith::ConstantOp>(
-        loc, rewriter.getI32Type(), rewriter.getI32IntegerAttr(0));
+    Value zeroInt = arith::ConstantOp::create(
+        rewriter, loc, rewriter.getI32Type(), rewriter.getI32IntegerAttr(0));
     if (legalType.isScalable())
       // Use arbitrary runtime vector length when vector type is scalable.
       // Proper conversion pass should take it from the IR.
-      rvl = rewriter.create<arith::ConstantOp>(loc,
-                                               rewriter.getI64IntegerAttr(9));
+      rvl = arith::ConstantOp::create(rewriter, loc,
+                                      rewriter.getI64IntegerAttr(9));
     Value res;
     if (n == 1) {
-      res = rewriter.create<vcix::BinaryOp>(loc, legalType, opcodeAttr, vec,
-                                            zeroInt, rvl);
+      res = vcix::BinaryOp::create(rewriter, loc, legalType, opcodeAttr, vec,
+                                   zeroInt, rvl);
     } else {
       const unsigned eltCount = legalType.getShape()[0];
       Type eltTy = legalType.getElementType();
-      Value zero = rewriter.create<arith::ConstantOp>(
-          loc, eltTy, rewriter.getZeroAttr(eltTy));
-      res = rewriter.create<vector::BroadcastOp>(loc, opType, zero /*dummy*/);
+      Value zero = arith::ConstantOp::create(rewriter, loc, eltTy,
+                                             rewriter.getZeroAttr(eltTy));
+      res = vector::BroadcastOp::create(rewriter, loc, opType, zero /*dummy*/);
       for (unsigned i = 0; i < n; ++i) {
-        Value extracted = rewriter.create<vector::ScalableExtractOp>(
-            loc, legalType, vec, i * eltCount);
-        Value v = rewriter.create<vcix::BinaryOp>(loc, legalType, opcodeAttr,
-                                                  extracted, zeroInt, rvl);
-        res = rewriter.create<vector::ScalableInsertOp>(loc, v, res,
-                                                        i * eltCount);
+        Value extracted = vector::ScalableExtractOp::create(
+            rewriter, loc, legalType, vec, i * eltCount);
+        Value v = vcix::BinaryOp::create(rewriter, loc, legalType, opcodeAttr,
+                                         extracted, zeroInt, rvl);
+        res = vector::ScalableInsertOp::create(rewriter, loc, v, res,
+                                               i * eltCount);
       }
     }
     rewriter.replaceOp(op, res);
diff --git a/mlir/test/lib/Dialect/Affine/TestReifyValueBounds.cpp b/mlir/test/lib/Dialect/Affine/TestReifyValueBounds.cpp
index ed5d06d..3569a73 100644
--- a/mlir/test/lib/Dialect/Affine/TestReifyValueBounds.cpp
+++ b/mlir/test/lib/Dialect/Affine/TestReifyValueBounds.cpp
@@ -145,7 +145,7 @@ static LogicalResult testReifyValueBounds(FunctionOpInterface funcOp,
         if (reifiedScalable->map.getNumInputs() == 1) {
           // The only possible input to the bound is vscale.
           vscaleOperand.push_back(std::make_pair(
-              rewriter.create<vector::VectorScaleOp>(loc), std::nullopt));
+              vector::VectorScaleOp::create(rewriter, loc), std::nullopt));
         }
         reified = affine::materializeComputedBound(
             rewriter, loc, reifiedScalable->map, vscaleOperand);
@@ -169,8 +169,9 @@ static LogicalResult testReifyValueBounds(FunctionOpInterface funcOp,
       rewriter.replaceOp(op, val);
       return WalkResult::skip();
     }
-    Value constOp = rewriter.create<arith::ConstantIndexOp>(
-        op->getLoc(), cast<IntegerAttr>(cast<Attribute>(*reified)).getInt());
+    Value constOp = arith::ConstantIndexOp::create(
+        rewriter, op->getLoc(),
+        cast<IntegerAttr>(cast<Attribute>(*reified)).getInt());
     rewriter.replaceOp(op, constOp);
     return WalkResult::skip();
   });
diff --git a/mlir/test/lib/Dialect/Arith/TestEmulateWideInt.cpp b/mlir/test/lib/Dialect/Arith/TestEmulateWideInt.cpp
index 738d4ee59..a792d08 100644
--- a/mlir/test/lib/Dialect/Arith/TestEmulateWideInt.cpp
+++ b/mlir/test/lib/Dialect/Arith/TestEmulateWideInt.cpp
@@ -60,7 +60,7 @@ struct TestEmulateWideIntPass
     // casts (and vice versa) and using it insted of `llvm.bitcast`.
     auto addBitcast = [](OpBuilder &builder, Type type, ValueRange inputs,
                          Location loc) -> Value {
-      auto cast = builder.create<LLVM::BitcastOp>(loc, type, inputs);
+      auto cast = LLVM::BitcastOp::create(builder, loc, type, inputs);
       return cast->getResult(0);
     };
     typeConverter.addSourceMaterialization(addBitcast);
diff --git a/mlir/test/lib/Dialect/CMakeLists.txt b/mlir/test/lib/Dialect/CMakeLists.txt
index eb2f74e..3b7bd9b 100644
--- a/mlir/test/lib/Dialect/CMakeLists.txt
+++ b/mlir/test/lib/Dialect/CMakeLists.txt
@@ -10,7 +10,7 @@ add_subdirectory(Linalg)
 add_subdirectory(LLVM)
 add_subdirectory(Math)
 add_subdirectory(MemRef)
-add_subdirectory(Mesh)
+add_subdirectory(Shard)
 add_subdirectory(NVGPU)
 add_subdirectory(SCF)
 add_subdirectory(Shape)
diff --git a/mlir/test/lib/Dialect/Func/TestDecomposeCallGraphTypes.cpp b/mlir/test/lib/Dialect/Func/TestDecomposeCallGraphTypes.cpp
index d0b62e7..c67bcd9 100644
--- a/mlir/test/lib/Dialect/Func/TestDecomposeCallGraphTypes.cpp
+++ b/mlir/test/lib/Dialect/Func/TestDecomposeCallGraphTypes.cpp
@@ -48,8 +48,8 @@ static SmallVector<Value> buildDecomposeTuple(OpBuilder &builder,
     }
     for (unsigned i = 0, e = tupleType.size(); i < e; ++i) {
       Type elementType = tupleType.getType(i);
-      Value element = builder.create<test::GetTupleElementOp>(
-          loc, elementType, tuple, builder.getI32IntegerAttr(i));
+      Value element = test::GetTupleElementOp::create(
+          builder, loc, elementType, tuple, builder.getI32IntegerAttr(i));
       decompose(element);
     }
   };
@@ -94,7 +94,7 @@ static Value buildMakeTupleOp(OpBuilder &builder, TupleType resultType,
   }
 
   // Assemble the tuple from the elements.
-  return builder.create<test::MakeTupleOp>(loc, resultType, elements);
+  return test::MakeTupleOp::create(builder, loc, resultType, elements);
 }
 
 /// A pass for testing call graph type decomposition.
diff --git a/mlir/test/lib/Dialect/SCF/TestSCFUtils.cpp b/mlir/test/lib/Dialect/SCF/TestSCFUtils.cpp
index 9eade75..9a394d2 100644
--- a/mlir/test/lib/Dialect/SCF/TestSCFUtils.cpp
+++ b/mlir/test/lib/Dialect/SCF/TestSCFUtils.cpp
@@ -56,7 +56,7 @@ struct TestSCFForUtilsPass
           SmallVector<Value> newYieldValues;
           for (auto yieldVal : oldYieldValues) {
             newYieldValues.push_back(
-                b.create<arith::AddFOp>(loc, yieldVal, yieldVal));
+                arith::AddFOp::create(b, loc, yieldVal, yieldVal));
           }
           return newYieldValues;
         };
@@ -160,13 +160,13 @@ struct TestSCFPipeliningPass
                                 Value pred) {
     Location loc = op->getLoc();
     auto ifOp =
-        rewriter.create<scf::IfOp>(loc, op->getResultTypes(), pred, true);
+        scf::IfOp::create(rewriter, loc, op->getResultTypes(), pred, true);
     // True branch.
     rewriter.moveOpBefore(op, &ifOp.getThenRegion().front(),
                           ifOp.getThenRegion().front().begin());
     rewriter.setInsertionPointAfter(op);
     if (op->getNumResults() > 0)
-      rewriter.create<scf::YieldOp>(loc, op->getResults());
+      scf::YieldOp::create(rewriter, loc, op->getResults());
     // False branch.
     rewriter.setInsertionPointToStart(&ifOp.getElseRegion().front());
     SmallVector<Value> elseYieldOperands;
@@ -181,12 +181,12 @@ struct TestSCFPipeliningPass
     } else {
       // Default to assuming constant numeric values.
       for (Type type : op->getResultTypes()) {
-        elseYieldOperands.push_back(rewriter.create<arith::ConstantOp>(
-            loc, rewriter.getZeroAttr(type)));
+        elseYieldOperands.push_back(arith::ConstantOp::create(
+            rewriter, loc, rewriter.getZeroAttr(type)));
       }
     }
     if (op->getNumResults() > 0)
-      rewriter.create<scf::YieldOp>(loc, elseYieldOperands);
+      scf::YieldOp::create(rewriter, loc, elseYieldOperands);
     return ifOp.getOperation();
   }
 
diff --git a/mlir/test/lib/Dialect/SCF/TestWhileOpBuilder.cpp b/mlir/test/lib/Dialect/SCF/TestWhileOpBuilder.cpp
index d3113c0..d3f7f0e6 100644
--- a/mlir/test/lib/Dialect/SCF/TestWhileOpBuilder.cpp
+++ b/mlir/test/lib/Dialect/SCF/TestWhileOpBuilder.cpp
@@ -50,23 +50,23 @@ struct TestSCFWhileOpBuilderPass
       // Create a WhileOp with the same operands and result types.
       TypeRange resultTypes = whileOp->getResultTypes();
       ValueRange operands = whileOp->getOperands();
-      builder.create<WhileOp>(
-          loc, resultTypes, operands, /*beforeBuilder=*/
+      WhileOp::create(
+          builder, loc, resultTypes, operands, /*beforeBuilder=*/
           [&](OpBuilder &b, Location loc, ValueRange args) {
             // Just cast the before args into the right types for condition.
             ImplicitLocOpBuilder builder(loc, b);
             auto castOp =
-                builder.create<UnrealizedConversionCastOp>(resultTypes, args);
-            auto cmp = builder.create<ConstantIntOp>(/*value=*/1, /*width=*/1);
-            builder.create<ConditionOp>(cmp, castOp->getResults());
+                UnrealizedConversionCastOp::create(builder, resultTypes, args);
+            auto cmp = ConstantIntOp::create(builder, /*value=*/1, /*width=*/1);
+            ConditionOp::create(builder, cmp, castOp->getResults());
           },
           /*afterBuilder=*/
           [&](OpBuilder &b, Location loc, ValueRange args) {
             // Just cast the after args into the right types for yield.
             ImplicitLocOpBuilder builder(loc, b);
-            auto castOp = builder.create<UnrealizedConversionCastOp>(
-                operands.getTypes(), args);
-            builder.create<YieldOp>(castOp->getResults());
+            auto castOp = UnrealizedConversionCastOp::create(
+                builder, operands.getTypes(), args);
+            YieldOp::create(builder, castOp->getResults());
           });
     });
   }
diff --git a/mlir/test/lib/Dialect/Mesh/CMakeLists.txt b/mlir/test/lib/Dialect/Shard/CMakeLists.txt
index 7bd0493..f91c547 100644
--- a/mlir/test/lib/Dialect/Mesh/CMakeLists.txt
+++ b/mlir/test/lib/Dialect/Shard/CMakeLists.txt
@@ -1,14 +1,14 @@
 # Exclude tests from libMLIR.so
-add_mlir_library(MLIRMeshTest
+add_mlir_library(MLIRShardTest
   TestOpLowering.cpp
-  TestReshardingSpmdization.cpp
+  TestReshardingPartition.cpp
   TestSimplifications.cpp
 
   EXCLUDE_FROM_LIBMLIR
   )
-mlir_target_link_libraries(MLIRMeshTest PUBLIC
-  MLIRMeshDialect
-  MLIRMeshTransforms
+mlir_target_link_libraries(MLIRShardTest PUBLIC
+  MLIRShardDialect
+  MLIRShardTransforms
   MLIRPass
   MLIRRewrite
   MLIRTransformUtils
diff --git a/mlir/test/lib/Dialect/Mesh/TestOpLowering.cpp b/mlir/test/lib/Dialect/Shard/TestOpLowering.cpp
index dbae93b..43f3b3f 100644
--- a/mlir/test/lib/Dialect/Mesh/TestOpLowering.cpp
+++ b/mlir/test/lib/Dialect/Shard/TestOpLowering.cpp
@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Dialect/Arith/IR/Arith.h"
-#include "mlir/Dialect/Mesh/Transforms/Transforms.h"
+#include "mlir/Dialect/Shard/Transforms/Transforms.h"
 #include "mlir/Dialect/Utils/IndexingUtils.h"
 #include "mlir/IR/SymbolTable.h"
 #include "mlir/Pass/Pass.h"
@@ -24,17 +24,17 @@ struct TestAllSliceOpLoweringPass
   void runOnOperation() override {
     RewritePatternSet patterns(&getContext());
     SymbolTableCollection symbolTableCollection;
-    mesh::populateAllSliceOpLoweringPatterns(patterns, symbolTableCollection);
+    shard::populateAllSliceOpLoweringPatterns(patterns, symbolTableCollection);
     LogicalResult status =
         applyPatternsGreedily(getOperation(), std::move(patterns));
     (void)status;
     assert(succeeded(status) && "applyPatternsGreedily failed.");
   }
   void getDependentDialects(DialectRegistry &registry) const override {
-    mesh::registerAllSliceOpLoweringDialects(registry);
+    shard::registerAllSliceOpLoweringDialects(registry);
   }
   StringRef getArgument() const final {
-    return "test-mesh-all-slice-op-lowering";
+    return "test-grid-all-slice-op-lowering";
   }
   StringRef getDescription() const final {
     return "Test lowering of all-slice.";
@@ -48,21 +48,21 @@ struct TestMultiIndexOpLoweringPass
   void runOnOperation() override {
     RewritePatternSet patterns(&getContext());
     SymbolTableCollection symbolTableCollection;
-    mesh::populateProcessMultiIndexOpLoweringPatterns(patterns,
-                                                      symbolTableCollection);
+    shard::populateProcessMultiIndexOpLoweringPatterns(patterns,
+                                                       symbolTableCollection);
     LogicalResult status =
         applyPatternsGreedily(getOperation(), std::move(patterns));
     (void)status;
     assert(succeeded(status) && "applyPatternsGreedily failed.");
   }
   void getDependentDialects(DialectRegistry &registry) const override {
-    mesh::registerProcessMultiIndexOpLoweringDialects(registry);
+    shard::registerProcessMultiIndexOpLoweringDialects(registry);
   }
   StringRef getArgument() const final {
-    return "test-mesh-process-multi-index-op-lowering";
+    return "test-grid-process-multi-index-op-lowering";
   }
   StringRef getDescription() const final {
-    return "Test lowering of mesh.process_multi_index op.";
+    return "Test lowering of shard.process_multi_index op.";
   }
 };
 
diff --git a/mlir/test/lib/Dialect/Mesh/TestReshardingSpmdization.cpp b/mlir/test/lib/Dialect/Shard/TestReshardingPartition.cpp
index 102e64d..23fdad1 100644
--- a/mlir/test/lib/Dialect/Mesh/TestReshardingSpmdization.cpp
+++ b/mlir/test/lib/Dialect/Shard/TestReshardingPartition.cpp
@@ -7,9 +7,9 @@
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Dialect/Func/IR/FuncOps.h"
-#include "mlir/Dialect/Mesh/IR/MeshOps.h"
-#include "mlir/Dialect/Mesh/Transforms/Spmdization.h"
 #include "mlir/Dialect/SPIRV/IR/SPIRVOps.h"
+#include "mlir/Dialect/Shard/IR/ShardOps.h"
+#include "mlir/Dialect/Shard/Transforms/Partition.h"
 #include "mlir/IR/BuiltinDialect.h"
 #include "mlir/IR/BuiltinOps.h"
 #include "mlir/IR/BuiltinTypeInterfaces.h"
@@ -22,11 +22,11 @@
 #include "mlir/Transforms/GreedyPatternRewriteDriver.h"
 
 using namespace mlir;
-using namespace mlir::mesh;
+using namespace mlir::shard;
 
 namespace {
 
-struct TestMeshReshardingRewritePattern : OpRewritePattern<ShardOp> {
+struct TestReshardingRewritePattern : OpRewritePattern<ShardOp> {
   using OpRewritePattern<ShardOp>::OpRewritePattern;
 
   LogicalResult matchAndRewrite(ShardOp op,
@@ -36,18 +36,18 @@ struct TestMeshReshardingRewritePattern : OpRewritePattern<ShardOp> {
     }
 
     SymbolTableCollection symbolTable;
-    mesh::MeshOp mesh = symbolTable.lookupNearestSymbolFrom<mesh::MeshOp>(
-        op, cast<ShardingOp>(op.getSharding().getDefiningOp()).getMeshAttr());
+    shard::GridOp grid = symbolTable.lookupNearestSymbolFrom<shard::GridOp>(
+        op, cast<ShardingOp>(op.getSharding().getDefiningOp()).getGridAttr());
 
     bool foundUser = false;
     for (auto user : op->getUsers()) {
       if (auto targetShardOp = llvm::dyn_cast<ShardOp>(user)) {
         if (targetShardOp.getAnnotateForUsers() &&
-            mesh == symbolTable.lookupNearestSymbolFrom<mesh::MeshOp>(
+            grid == symbolTable.lookupNearestSymbolFrom<shard::GridOp>(
                         targetShardOp,
                         cast<ShardingOp>(
                             targetShardOp.getSharding().getDefiningOp())
-                            .getMeshAttr())) {
+                            .getGridAttr())) {
           foundUser = true;
           break;
         }
@@ -61,26 +61,25 @@ struct TestMeshReshardingRewritePattern : OpRewritePattern<ShardOp> {
     for (auto user : op->getUsers()) {
       auto targetShardOp = llvm::dyn_cast<ShardOp>(user);
       if (!targetShardOp || !targetShardOp.getAnnotateForUsers() ||
-          symbolTable.lookupNearestSymbolFrom<mesh::MeshOp>(
+          symbolTable.lookupNearestSymbolFrom<shard::GridOp>(
               targetShardOp,
               cast<ShardingOp>(targetShardOp.getSharding().getDefiningOp())
-                  .getMeshAttr()) != mesh) {
+                  .getGridAttr()) != grid) {
         continue;
       }
 
       ImplicitLocOpBuilder builder(op->getLoc(), rewriter);
       ShapedType sourceShardShape =
-          shardShapedType(op.getResult().getType(), mesh, op.getSharding());
+          shardShapedType(op.getResult().getType(), grid, op.getSharding());
       TypedValue<ShapedType> sourceShard = cast<TypedValue<ShapedType>>(
-          builder
-              .create<UnrealizedConversionCastOp>(sourceShardShape, op.getSrc())
+          UnrealizedConversionCastOp::create(builder, sourceShardShape,
+                                             op.getSrc())
               ->getResult(0));
       TypedValue<ShapedType> targetShard =
-          reshard(builder, mesh, op, targetShardOp, sourceShard);
+          reshard(builder, grid, op, targetShardOp, sourceShard);
       Value newTargetUnsharded =
-          builder
-              .create<UnrealizedConversionCastOp>(
-                  targetShardOp.getResult().getType(), targetShard)
+          UnrealizedConversionCastOp::create(
+              builder, targetShardOp.getResult().getType(), targetShard)
               ->getResult(0);
       rewriter.replaceAllUsesWith(targetShardOp.getResult(),
                                   newTargetUnsharded);
@@ -90,13 +89,13 @@ struct TestMeshReshardingRewritePattern : OpRewritePattern<ShardOp> {
   }
 };
 
-struct TestMeshReshardingPass
-    : public PassWrapper<TestMeshReshardingPass, OperationPass<ModuleOp>> {
-  MLIR_DEFINE_EXPLICIT_INTERNAL_INLINE_TYPE_ID(TestMeshReshardingPass)
+struct TestReshardingPass
+    : public PassWrapper<TestReshardingPass, OperationPass<ModuleOp>> {
+  MLIR_DEFINE_EXPLICIT_INTERNAL_INLINE_TYPE_ID(TestReshardingPass)
 
   void runOnOperation() override {
     RewritePatternSet patterns(&getContext());
-    patterns.insert<TestMeshReshardingRewritePattern>(&getContext());
+    patterns.insert<TestReshardingRewritePattern>(&getContext());
     if (failed(applyPatternsGreedily(getOperation().getOperation(),
                                      std::move(patterns)))) {
       return signalPassFailure();
@@ -107,18 +106,18 @@ struct TestMeshReshardingPass
     registry.insert<BuiltinDialect>();
   }
   StringRef getArgument() const final {
-    return "test-mesh-resharding-spmdization";
+    return "test-grid-resharding-partition";
   }
   StringRef getDescription() const final {
-    return "Test Mesh dialect resharding spmdization.";
+    return "Test Shard dialect resharding partition.";
   }
 };
 } // namespace
 
 namespace mlir {
 namespace test {
-void registerTestMeshReshardingSpmdizationPass() {
-  PassRegistration<TestMeshReshardingPass>();
+void registerTestReshardingPartitionPass() {
+  PassRegistration<TestReshardingPass>();
 }
 } // namespace test
 } // namespace mlir
diff --git a/mlir/test/lib/Dialect/Mesh/TestSimplifications.cpp b/mlir/test/lib/Dialect/Shard/TestSimplifications.cpp
index 01e196d..2885215 100644
--- a/mlir/test/lib/Dialect/Mesh/TestSimplifications.cpp
+++ b/mlir/test/lib/Dialect/Shard/TestSimplifications.cpp
@@ -7,8 +7,8 @@
 //===----------------------------------------------------------------------===//
 
 #include "mlir/Dialect/Arith/IR/Arith.h"
-#include "mlir/Dialect/Mesh/IR/MeshDialect.h"
-#include "mlir/Dialect/Mesh/Transforms/Simplifications.h"
+#include "mlir/Dialect/Shard/IR/ShardDialect.h"
+#include "mlir/Dialect/Shard/Transforms/Simplifications.h"
 #include "mlir/IR/SymbolTable.h"
 #include "mlir/Pass/Pass.h"
 #include "mlir/Transforms/GreedyPatternRewriteDriver.h"
@@ -16,23 +16,23 @@
 using namespace mlir;
 
 namespace {
-struct TestMeshSimplificationsPass
-    : public PassWrapper<TestMeshSimplificationsPass, OperationPass<>> {
-  MLIR_DEFINE_EXPLICIT_INTERNAL_INLINE_TYPE_ID(TestMeshSimplificationsPass)
+struct TestShardSimplificationsPass
+    : public PassWrapper<TestShardSimplificationsPass, OperationPass<>> {
+  MLIR_DEFINE_EXPLICIT_INTERNAL_INLINE_TYPE_ID(TestShardSimplificationsPass)
 
   void runOnOperation() override;
   void getDependentDialects(DialectRegistry &registry) const override {
-    registry.insert<arith::ArithDialect, mesh::MeshDialect>();
+    registry.insert<arith::ArithDialect, shard::ShardDialect>();
   }
-  StringRef getArgument() const final { return "test-mesh-simplifications"; }
-  StringRef getDescription() const final { return "Test mesh simplifications"; }
+  StringRef getArgument() const final { return "test-grid-simplifications"; }
+  StringRef getDescription() const final { return "Test grid simplifications"; }
 };
 } // namespace
 
-void TestMeshSimplificationsPass::runOnOperation() {
+void TestShardSimplificationsPass::runOnOperation() {
   RewritePatternSet patterns(&getContext());
   SymbolTableCollection symbolTableCollection;
-  mesh::populateSimplificationPatterns(patterns, symbolTableCollection);
+  shard::populateSimplificationPatterns(patterns, symbolTableCollection);
   [[maybe_unused]] LogicalResult status =
       applyPatternsGreedily(getOperation(), std::move(patterns));
   assert(succeeded(status) && "Rewrite patters application did not converge.");
@@ -40,8 +40,8 @@ void TestMeshSimplificationsPass::runOnOperation() {
 
 namespace mlir {
 namespace test {
-void registerTestMeshSimplificationsPass() {
-  PassRegistration<TestMeshSimplificationsPass>();
+void registerTestShardSimplificationsPass() {
+  PassRegistration<TestShardSimplificationsPass>();
 }
 } // namespace test
 } // namespace mlir
diff --git a/mlir/test/lib/Dialect/Tensor/TestTensorTransforms.cpp b/mlir/test/lib/Dialect/Tensor/TestTensorTransforms.cpp
index 0e191c3..687473e 100644
--- a/mlir/test/lib/Dialect/Tensor/TestTensorTransforms.cpp
+++ b/mlir/test/lib/Dialect/Tensor/TestTensorTransforms.cpp
@@ -192,8 +192,8 @@ struct RewriteExtractSliceFromCollapseShapeBase
     // Create the destination tensor using the above values.
     Type elementType = op.getSourceType().getElementType();
     SmallVector<OpFoldResult> outputShape = reifiedShapes[0];
-    Value dest = rewriter.create<tensor::EmptyOp>(op->getLoc(), outputShape,
-                                                  elementType);
+    Value dest = tensor::EmptyOp::create(rewriter, op->getLoc(), outputShape,
+                                         elementType);
 
     // Calculate the parameters for the tile loop nest.
     FailureOr<tensor::ExtractSliceFromCollapseHelper> params =
@@ -215,8 +215,8 @@ struct RewriteExtractSliceFromCollapseShapeUsingScfFor
                                 PatternRewriter &rewriter) const override {
     Location loc = op.getLoc();
     const unsigned numTiledDims = helper.getIterationSpaceSizes().size();
-    auto zero = rewriter.create<arith::ConstantIndexOp>(loc, 0);
-    auto one = rewriter.create<arith::ConstantIndexOp>(loc, 1);
+    auto zero = arith::ConstantIndexOp::create(rewriter, loc, 0);
+    auto one = arith::ConstantIndexOp::create(rewriter, loc, 1);
     SmallVector<Value> lbs(numTiledDims, zero);
     SmallVector<Value> steps(numTiledDims, one);
 
@@ -228,8 +228,8 @@ struct RewriteExtractSliceFromCollapseShapeUsingScfFor
               helper.emitLoopNestBody(nestedBuilder, loc, outputIvs);
 
           // Insert the slice into the destination.
-          return {nestedBuilder.create<tensor::InsertSliceOp>(
-              loc, tile, iterArgs[0], insertParams)};
+          return {tensor::InsertSliceOp::create(nestedBuilder, loc, tile,
+                                                iterArgs[0], insertParams)};
         });
     rewriter.replaceOp(op, nest.results);
 
@@ -245,8 +245,9 @@ struct RewriteExtractSliceFromCollapseShapeUsingScfForeach
                                 tensor::ExtractSliceFromCollapseHelper &helper,
                                 PatternRewriter &rewriter) const override {
     Location loc = op.getLoc();
-    auto forallOp = rewriter.create<scf::ForallOp>(
-        loc, /*numThreads=*/getAsOpFoldResult(helper.getIterationSpaceSizes()),
+    auto forallOp = scf::ForallOp::create(
+        rewriter, loc,
+        /*numThreads=*/getAsOpFoldResult(helper.getIterationSpaceSizes()),
         /*outputs=*/dest,
         /*mapping=*/std::nullopt,
         [&](OpBuilder &nestedBuilder, Location loc, ValueRange regionArgs) {
@@ -261,10 +262,10 @@ struct RewriteExtractSliceFromCollapseShapeUsingScfForeach
           auto [tile, insertParams] =
               helper.emitLoopNestBody(nestedBuilder, loc, outputIvs);
           // Insert the slice into the destination.
-          auto term = nestedBuilder.create<scf::InParallelOp>(loc);
+          auto term = scf::InParallelOp::create(nestedBuilder, loc);
           nestedBuilder.setInsertionPointToStart(term.getBody());
-          nestedBuilder.create<tensor::ParallelInsertSliceOp>(
-              loc, tile, outputArgs[0], insertParams);
+          tensor::ParallelInsertSliceOp::create(nestedBuilder, loc, tile,
+                                                outputArgs[0], insertParams);
         });
     rewriter.replaceOp(op, forallOp->getResult(0));
     return success();
@@ -355,8 +356,8 @@ static LogicalResult testTrackingListenerReplacements(Operation *rootOp) {
   MLIRContext *context = rootOp->getContext();
   OpBuilder builder(context);
   OwningOpRef<transform::NamedSequenceOp> transformOp =
-      builder.create<transform::NamedSequenceOp>(
-          rootOp->getLoc(),
+      transform::NamedSequenceOp::create(
+          builder, rootOp->getLoc(),
           /*sym_name=*/"test_sequence",
           /*function_type=*/
           TypeAttr::get(FunctionType::get(context, TypeRange{}, TypeRange{})),
diff --git a/mlir/test/lib/Dialect/Test/TestDialect.cpp b/mlir/test/lib/Dialect/Test/TestDialect.cpp
index 1bbf2cc..a4c615b 100644
--- a/mlir/test/lib/Dialect/Test/TestDialect.cpp
+++ b/mlir/test/lib/Dialect/Test/TestDialect.cpp
@@ -346,7 +346,7 @@ TestDialect::~TestDialect() {
 
 Operation *TestDialect::materializeConstant(OpBuilder &builder, Attribute value,
                                             Type type, Location loc) {
-  return builder.create<TestOpConstant>(loc, type, value);
+  return TestOpConstant::create(builder, loc, type, value);
 }
 
 void *TestDialect::getRegisteredInterfaceForOp(TypeID typeID,
diff --git a/mlir/test/lib/Dialect/Test/TestDialectInterfaces.cpp b/mlir/test/lib/Dialect/Test/TestDialectInterfaces.cpp
index 01ae245..1235a5f 100644
--- a/mlir/test/lib/Dialect/Test/TestDialectInterfaces.cpp
+++ b/mlir/test/lib/Dialect/Test/TestDialectInterfaces.cpp
@@ -354,7 +354,7 @@ struct TestInlinerInterface : public DialectInlinerInterface {
         !(input.getType().isSignlessInteger(16) ||
           input.getType().isSignlessInteger(32)))
       return nullptr;
-    return builder.create<TestCastOp>(conversionLoc, resultType, input);
+    return TestCastOp::create(builder, conversionLoc, resultType, input);
   }
 
   Value handleArgument(OpBuilder &builder, Operation *call, Operation *callable,
@@ -362,16 +362,16 @@ struct TestInlinerInterface : public DialectInlinerInterface {
                        DictionaryAttr argumentAttrs) const final {
     if (!argumentAttrs.contains("test.handle_argument"))
       return argument;
-    return builder.create<TestTypeChangerOp>(call->getLoc(), argument.getType(),
-                                             argument);
+    return TestTypeChangerOp::create(builder, call->getLoc(),
+                                     argument.getType(), argument);
   }
 
   Value handleResult(OpBuilder &builder, Operation *call, Operation *callable,
                      Value result, DictionaryAttr resultAttrs) const final {
     if (!resultAttrs.contains("test.handle_result"))
       return result;
-    return builder.create<TestTypeChangerOp>(call->getLoc(), result.getType(),
-                                             result);
+    return TestTypeChangerOp::create(builder, call->getLoc(), result.getType(),
+                                     result);
   }
 
   void processInlinedCallBlocks(
diff --git a/mlir/test/lib/Dialect/Test/TestFromLLVMIRTranslation.cpp b/mlir/test/lib/Dialect/Test/TestFromLLVMIRTranslation.cpp
index dc6413b..b98f6ce 100644
--- a/mlir/test/lib/Dialect/Test/TestFromLLVMIRTranslation.cpp
+++ b/mlir/test/lib/Dialect/Test/TestFromLLVMIRTranslation.cpp
@@ -43,11 +43,11 @@ static LogicalResult convertLoad(OpBuilder &builder, llvm::Instruction *inst,
   if (failed(addr))
     return failure();
   // Create the LoadOp
-  Value loadOp = builder.create<LLVM::LoadOp>(
-      moduleImport.translateLoc(inst->getDebugLoc()),
+  Value loadOp = LLVM::LoadOp::create(
+      builder, moduleImport.translateLoc(inst->getDebugLoc()),
       moduleImport.convertType(inst->getType()), *addr);
-  moduleImport.mapValue(inst) = builder.create<SameOperandElementTypeOp>(
-      loadOp.getLoc(), loadOp.getType(), loadOp, loadOp);
+  moduleImport.mapValue(inst) = SameOperandElementTypeOp::create(
+      builder, loadOp.getLoc(), loadOp.getType(), loadOp, loadOp);
   return success();
 }
 
diff --git a/mlir/test/lib/Dialect/Test/TestOpDefs.cpp b/mlir/test/lib/Dialect/Test/TestOpDefs.cpp
index 3ab4ef2..f79e2cf 100644
--- a/mlir/test/lib/Dialect/Test/TestOpDefs.cpp
+++ b/mlir/test/lib/Dialect/Test/TestOpDefs.cpp
@@ -286,9 +286,9 @@ LogicalResult OpWithResultShapeInterfaceOp::reifyReturnTypeShapes(
         llvm::map_range(llvm::seq<int64_t>(0, rank), [&](int64_t dim) -> Value {
           return builder.createOrFold<tensor::DimOp>(loc, operand, dim);
         }));
-    shapes.push_back(builder.create<tensor::FromElementsOp>(
-        getLoc(), RankedTensorType::get({rank}, builder.getIndexType()),
-        currShape));
+    shapes.push_back(tensor::FromElementsOp::create(
+        builder, getLoc(),
+        RankedTensorType::get({rank}, builder.getIndexType()), currShape));
   }
   return success();
 }
@@ -1302,8 +1302,8 @@ llvm::SmallVector<MemorySlot> TestMultiSlotAlloca::getPromotableSlots() {
 
 Value TestMultiSlotAlloca::getDefaultValue(const MemorySlot &slot,
                                            OpBuilder &builder) {
-  return builder.create<TestOpConstant>(getLoc(), slot.elemType,
-                                        builder.getI32IntegerAttr(42));
+  return TestOpConstant::create(builder, getLoc(), slot.elemType,
+                                builder.getI32IntegerAttr(42));
 }
 
 void TestMultiSlotAlloca::handleBlockArgument(const MemorySlot &slot,
@@ -1335,7 +1335,7 @@ createNewMultiAllocaWithoutSlot(const MemorySlot &slot, OpBuilder &builder,
   OpBuilder::InsertionGuard guard(builder);
   builder.setInsertionPoint(oldOp);
   auto replacement =
-      builder.create<TestMultiSlotAlloca>(oldOp->getLoc(), newTypes);
+      TestMultiSlotAlloca::create(builder, oldOp->getLoc(), newTypes);
   for (auto [oldResult, newResult] :
        llvm::zip_equal(remainingValues, replacement.getResults()))
     oldResult.replaceAllUsesWith(newResult);
@@ -1384,7 +1384,7 @@ DenseMap<Attribute, MemorySlot> TestMultiSlotAlloca::destructure(
   for (Attribute usedIndex : usedIndices) {
     Type elemType = slot.subelementTypes.lookup(usedIndex);
     MemRefType elemPtr = MemRefType::get({}, elemType);
-    auto subAlloca = builder.create<TestMultiSlotAlloca>(getLoc(), elemPtr);
+    auto subAlloca = TestMultiSlotAlloca::create(builder, getLoc(), elemPtr);
     newAllocators.push_back(subAlloca);
     slotMap.try_emplace<MemorySlot>(usedIndex,
                                     {subAlloca.getResult(0), elemType});
@@ -1412,8 +1412,8 @@ TestMultiSlotAlloca::handleDestructuringComplete(
   const auto bufferizedOutType = test::TestMemrefType::get(
       getContext(), outType.getShape(), outType.getElementType(), nullptr);
   // replace op with memref analogy
-  auto dummyMemrefOp = rewriter.create<test::TestDummyMemrefOp>(
-      getLoc(), bufferizedOutType, *buffer);
+  auto dummyMemrefOp = test::TestDummyMemrefOp::create(
+      rewriter, getLoc(), bufferizedOutType, *buffer);
 
   mlir::bufferization::replaceOpWithBufferizedValues(rewriter, getOperation(),
                                                      dummyMemrefOp.getResult());
@@ -1434,7 +1434,7 @@ TestMultiSlotAlloca::handleDestructuringComplete(
 
   // replace op with memref analogy
   auto createMemrefOp =
-      rewriter.create<test::TestCreateMemrefOp>(getLoc(), *bufferizedOutType);
+      test::TestCreateMemrefOp::create(rewriter, getLoc(), *bufferizedOutType);
 
   mlir::bufferization::replaceOpWithBufferizedValues(
       rewriter, getOperation(), createMemrefOp.getResult());
diff --git a/mlir/test/lib/Dialect/Test/TestOps.td b/mlir/test/lib/Dialect/Test/TestOps.td
index ab3f847..a7c6cd6 100644
--- a/mlir/test/lib/Dialect/Test/TestOps.td
+++ b/mlir/test/lib/Dialect/Test/TestOps.td
@@ -2035,7 +2035,7 @@ def IllegalOpWithRegion : TEST_Op<"illegal_op_with_region"> {
        OpBuilder::InsertionGuard g($_builder);
        Block *body = $_builder.createBlock(bodyRegion);
        $_builder.setInsertionPointToEnd(body);
-       $_builder.create<IllegalOpTerminator>($_state.location);
+       IllegalOpTerminator::create($_builder,$_state.location);
     }]>];
 }
 def IllegalOpWithRegionAnchor : TEST_Op<"illegal_op_with_region_anchor">;
@@ -2738,7 +2738,7 @@ def TestLinalgConvOp :
     static void regionBuilder(mlir::ImplicitLocOpBuilder &b, mlir::Block &block,
                               mlir::ArrayRef<mlir::NamedAttribute> attrs,
                               llvm::function_ref<mlir::InFlightDiagnostic()> emitError) {
-      b.create<mlir::linalg::YieldOp>(block.getArguments().back());
+      mlir::linalg::YieldOp::create(b,block.getArguments().back());
     }
 
     static std::function<void(mlir::ImplicitLocOpBuilder &, mlir::Block &,
@@ -2801,7 +2801,7 @@ def TestLinalgFillOp :
     static void regionBuilder(mlir::ImplicitLocOpBuilder &b, mlir::Block &block,
                               mlir::ArrayRef<mlir::NamedAttribute> attrs,
                               llvm::function_ref<mlir::InFlightDiagnostic()> emitError) {
-      b.create<mlir::linalg::YieldOp>(block.getArguments().back());
+      mlir::linalg::YieldOp::create(b,block.getArguments().back());
     }
 
     static std::function<void(mlir::ImplicitLocOpBuilder &, mlir::Block &,
diff --git a/mlir/test/lib/Dialect/Test/TestOpsSyntax.cpp b/mlir/test/lib/Dialect/Test/TestOpsSyntax.cpp
index 6d4e5e3..cc131ad 100644
--- a/mlir/test/lib/Dialect/Test/TestOpsSyntax.cpp
+++ b/mlir/test/lib/Dialect/Test/TestOpsSyntax.cpp
@@ -313,7 +313,7 @@ ParseResult WrappingRegionOp::parse(OpAsmParser &parser,
   SmallVector<Value, 8> returnOperands(wrappedOp->getResults());
   OpBuilder builder(parser.getContext());
   builder.setInsertionPointToEnd(&block);
-  builder.create<TestReturnOp>(wrappedOp->getLoc(), returnOperands);
+  TestReturnOp::create(builder, wrappedOp->getLoc(), returnOperands);
 
   // Get the results type for the wrapping op from the terminator operands.
   Operation &returnOp = body.back().back();
@@ -397,7 +397,7 @@ ParseResult PrettyPrintedRegionOp::parse(OpAsmParser &parser,
       builder.create(opLoc, innerOpName, /*operands=*/{lhs, rhs}, innerOpType);
 
   // Insert a return statement in the block returning the inner-op's result.
-  builder.create<TestReturnOp>(innerOp->getLoc(), innerOp->getResults());
+  TestReturnOp::create(builder, innerOp->getLoc(), innerOp->getResults());
 
   // Populate the op operation-state with result-type and location.
   result.addTypes(opFntype.getResults());
diff --git a/mlir/test/lib/Dialect/Test/TestPatterns.cpp b/mlir/test/lib/Dialect/Test/TestPatterns.cpp
index 1fff57e..eda618f 100644
--- a/mlir/test/lib/Dialect/Test/TestPatterns.cpp
+++ b/mlir/test/lib/Dialect/Test/TestPatterns.cpp
@@ -33,14 +33,14 @@ static Value chooseOperand(Value input1, Value input2, BoolAttr choice) {
 }
 
 static void createOpI(PatternRewriter &rewriter, Location loc, Value input) {
-  rewriter.create<OpI>(loc, input);
+  OpI::create(rewriter, loc, input);
 }
 
 static void handleNoResultOp(PatternRewriter &rewriter,
                              OpSymbolBindingNoResult op) {
   // Turn the no result op to a one-result op.
-  rewriter.create<OpSymbolBindingB>(op.getLoc(), op.getOperand().getType(),
-                                    op.getOperand());
+  OpSymbolBindingB::create(rewriter, op.getLoc(), op.getOperand().getType(),
+                           op.getOperand());
 }
 
 static bool getFirstI32Result(Operation *op, Value &value) {
@@ -120,8 +120,8 @@ public:
       return failure();
     rewriter.setInsertionPointToStart(op->getBlock());
 
-    auto constOp = rewriter.create<arith::ConstantOp>(
-        op.getLoc(), rewriter.getBoolAttr(true));
+    auto constOp = arith::ConstantOp::create(rewriter, op.getLoc(),
+                                             rewriter.getBoolAttr(true));
     rewriter.replaceOpWithNewOp<TestCastOp>(op, rewriter.getI32Type(),
                                             Value(constOp));
     return success();
@@ -844,8 +844,8 @@ struct TestRegionRewriteUndo : public RewritePattern {
                             rewriter.getUnknownLoc());
 
     // Add an explicitly illegal operation to ensure the conversion fails.
-    rewriter.create<ILLegalOpF>(op->getLoc(), rewriter.getIntegerType(32));
-    rewriter.create<TestValidOp>(op->getLoc(), ArrayRef<Value>());
+    ILLegalOpF::create(rewriter, op->getLoc(), rewriter.getIntegerType(32));
+    TestValidOp::create(rewriter, op->getLoc(), ArrayRef<Value>());
 
     // Drop this operation.
     rewriter.eraseOp(op);
@@ -864,7 +864,7 @@ struct TestCreateBlock : public RewritePattern {
     Type i32Type = rewriter.getIntegerType(32);
     Location loc = op->getLoc();
     rewriter.createBlock(&region, region.end(), {i32Type, i32Type}, {loc, loc});
-    rewriter.create<TerminatorOp>(loc);
+    TerminatorOp::create(rewriter, loc);
     rewriter.eraseOp(op);
     return success();
   }
@@ -883,8 +883,8 @@ struct TestCreateIllegalBlock : public RewritePattern {
     Location loc = op->getLoc();
     rewriter.createBlock(&region, region.end(), {i32Type, i32Type}, {loc, loc});
     // Create an illegal op to ensure the conversion fails.
-    rewriter.create<ILLegalOpF>(loc, i32Type);
-    rewriter.create<TerminatorOp>(loc);
+    ILLegalOpF::create(rewriter, loc, i32Type);
+    TerminatorOp::create(rewriter, loc);
     rewriter.eraseOp(op);
     return success();
   }
@@ -939,7 +939,7 @@ struct TestUndoBlockErase : public ConversionPattern {
                   ConversionPatternRewriter &rewriter) const final {
     Block *secondBlock = &*std::next(op->getRegion(0).begin());
     rewriter.setInsertionPointToStart(secondBlock);
-    rewriter.create<ILLegalOpF>(op->getLoc(), rewriter.getF32Type());
+    ILLegalOpF::create(rewriter, op->getLoc(), rewriter.getF32Type());
     rewriter.eraseBlock(secondBlock);
     rewriter.modifyOpInPlace(op, [] {});
     return success();
@@ -1007,9 +1007,8 @@ struct TestPassthroughInvalidOp : public ConversionPattern {
       // This is a 1:N replacement. Insert a test.cast op. (That's what the
       // argument materialization used to do.)
       flattened.push_back(
-          rewriter
-              .create<TestCastOp>(op->getLoc(),
-                                  op->getOperand(it.index()).getType(), range)
+          TestCastOp::create(rewriter, op->getLoc(),
+                             op->getOperand(it.index()).getType(), range)
               .getResult());
     }
     rewriter.replaceOpWithNewOp<TestValidOp>(op, TypeRange(), flattened,
@@ -1114,8 +1113,8 @@ struct TestNonRootReplacement : public RewritePattern {
   LogicalResult matchAndRewrite(Operation *op,
                                 PatternRewriter &rewriter) const final {
     auto resultType = *op->result_type_begin();
-    auto illegalOp = rewriter.create<ILLegalOpF>(op->getLoc(), resultType);
-    auto legalOp = rewriter.create<LegalOpB>(op->getLoc(), resultType);
+    auto illegalOp = ILLegalOpF::create(rewriter, op->getLoc(), resultType);
+    auto legalOp = LegalOpB::create(rewriter, op->getLoc(), resultType);
 
     rewriter.replaceOp(illegalOp, legalOp);
     rewriter.replaceOp(op, illegalOp);
@@ -1181,7 +1180,7 @@ struct TestCreateUnregisteredOp : public OpRewritePattern<ILLegalOpG> {
   LogicalResult matchAndRewrite(ILLegalOpG op,
                                 PatternRewriter &rewriter) const final {
     IntegerAttr attr = rewriter.getI32IntegerAttr(0);
-    Value val = rewriter.create<arith::ConstantOp>(op->getLoc(), attr);
+    Value val = arith::ConstantOp::create(rewriter, op->getLoc(), attr);
     rewriter.replaceOpWithNewOp<LegalOpC>(op, val);
     return success();
   };
@@ -1354,7 +1353,7 @@ struct TestTypeConverter : public TypeConverter {
   /// 1->N type mappings.
   static Value materializeCast(OpBuilder &builder, Type resultType,
                                ValueRange inputs, Location loc) {
-    return builder.create<TestCastOp>(loc, resultType, inputs).getResult();
+    return TestCastOp::create(builder, loc, resultType, inputs).getResult();
   }
 };
 
@@ -1362,6 +1361,10 @@ struct TestLegalizePatternDriver
     : public PassWrapper<TestLegalizePatternDriver, OperationPass<>> {
   MLIR_DEFINE_EXPLICIT_INTERNAL_INLINE_TYPE_ID(TestLegalizePatternDriver)
 
+  TestLegalizePatternDriver() = default;
+  TestLegalizePatternDriver(const TestLegalizePatternDriver &other)
+      : PassWrapper(other) {}
+
   StringRef getArgument() const final { return "test-legalize-patterns"; }
   StringRef getDescription() const final {
     return "Run test dialect legalization patterns";
@@ -1369,8 +1372,6 @@ struct TestLegalizePatternDriver
   /// The mode of conversion to use with the driver.
   enum class ConversionMode { Analysis, Full, Partial };
 
-  TestLegalizePatternDriver(ConversionMode mode) : mode(mode) {}
-
   void getDependentDialects(DialectRegistry &registry) const override {
     registry.insert<func::FuncDialect, test::TestDialect>();
   }
@@ -1499,24 +1500,19 @@ struct TestLegalizePatternDriver
       op->emitRemark() << "op '" << op->getName() << "' is legalizable";
   }
 
-  /// The mode of conversion to use.
-  ConversionMode mode;
+  Option<ConversionMode> mode{
+      *this, "test-legalize-mode",
+      llvm::cl::desc("The legalization mode to use with the test driver"),
+      llvm::cl::init(ConversionMode::Partial),
+      llvm::cl::values(
+          clEnumValN(ConversionMode::Analysis, "analysis",
+                     "Perform an analysis conversion"),
+          clEnumValN(ConversionMode::Full, "full", "Perform a full conversion"),
+          clEnumValN(ConversionMode::Partial, "partial",
+                     "Perform a partial conversion"))};
 };
 } // namespace
 
-static llvm::cl::opt<TestLegalizePatternDriver::ConversionMode>
-    legalizerConversionMode(
-        "test-legalize-mode",
-        llvm::cl::desc("The legalization mode to use with the test driver"),
-        llvm::cl::init(TestLegalizePatternDriver::ConversionMode::Partial),
-        llvm::cl::values(
-            clEnumValN(TestLegalizePatternDriver::ConversionMode::Analysis,
-                       "analysis", "Perform an analysis conversion"),
-            clEnumValN(TestLegalizePatternDriver::ConversionMode::Full, "full",
-                       "Perform a full conversion"),
-            clEnumValN(TestLegalizePatternDriver::ConversionMode::Partial,
-                       "partial", "Perform a partial conversion")));
-
 //===----------------------------------------------------------------------===//
 // ConversionPatternRewriter::getRemappedValue testing. This method is used
 // to get the remapped value of an original value that was replaced using
@@ -1916,15 +1912,15 @@ struct TestTypeConversionDriver
       // Allow casting from F64 back to F32.
       if (!resultType.isF16() && inputs.size() == 1 &&
           inputs[0].getType().isF64())
-        return builder.create<TestCastOp>(loc, resultType, inputs).getResult();
+        return TestCastOp::create(builder, loc, resultType, inputs).getResult();
       // Allow producing an i32 or i64 from nothing.
       if ((resultType.isInteger(32) || resultType.isInteger(64)) &&
           inputs.empty())
-        return builder.create<TestTypeProducerOp>(loc, resultType);
+        return TestTypeProducerOp::create(builder, loc, resultType);
       // Allow producing an i64 from an integer.
       if (isa<IntegerType>(resultType) && inputs.size() == 1 &&
           isa<IntegerType>(inputs[0].getType()))
-        return builder.create<TestCastOp>(loc, resultType, inputs).getResult();
+        return TestCastOp::create(builder, loc, resultType, inputs).getResult();
       // Otherwise, fail.
       return nullptr;
     });
@@ -2007,7 +2003,7 @@ struct TestTargetMaterializationWithNoUses
     });
     converter.addTargetMaterialization(
         [](OpBuilder &builder, Type type, ValueRange inputs, Location loc) {
-          return builder.create<TestCastOp>(loc, type, inputs).getResult();
+          return TestCastOp::create(builder, loc, type, inputs).getResult();
         });
 
     ConversionTarget target(getContext());
@@ -2058,7 +2054,7 @@ struct TestUndoBlocksMerge : public ConversionPattern {
     Operation *branchOp = firstBlock.getTerminator();
     Block *secondBlock = &*(std::next(op->getRegion(0).begin()));
     rewriter.setInsertionPointToStart(secondBlock);
-    rewriter.create<ILLegalOpF>(op->getLoc(), rewriter.getF32Type());
+    ILLegalOpF::create(rewriter, op->getLoc(), rewriter.getF32Type());
     auto succOperands = branchOp->getOperands();
     SmallVector<Value, 2> replacements(succOperands);
     rewriter.eraseOp(branchOp);
@@ -2202,9 +2198,7 @@ void registerPatternsTestPass() {
   PassRegistration<TestStrictPatternDriver>();
   PassRegistration<TestWalkPatternDriver>();
 
-  PassRegistration<TestLegalizePatternDriver>([] {
-    return std::make_unique<TestLegalizePatternDriver>(legalizerConversionMode);
-  });
+  PassRegistration<TestLegalizePatternDriver>();
 
   PassRegistration<TestRemappedValue>();
 
diff --git a/mlir/test/lib/Dialect/Test/TestToLLVMIRTranslation.cpp b/mlir/test/lib/Dialect/Test/TestToLLVMIRTranslation.cpp
index 103817d..7831b27 100644
--- a/mlir/test/lib/Dialect/Test/TestToLLVMIRTranslation.cpp
+++ b/mlir/test/lib/Dialect/Test/TestToLLVMIRTranslation.cpp
@@ -68,8 +68,8 @@ LogicalResult TestDialectLLVMIRTranslationInterface::amendOperation(
 
               if (createSymbol) {
                 OpBuilder builder(op->getRegion(0));
-                builder.create<test::SymbolOp>(
-                    op->getLoc(),
+                test::SymbolOp::create(
+                    builder, op->getLoc(),
                     StringAttr::get(op->getContext(), "sym_from_attr"),
                     /*sym_visibility=*/nullptr);
               }
diff --git a/mlir/test/lib/Dialect/TestIRDLToCpp/TestIRDLToCppDialect.cpp b/mlir/test/lib/Dialect/TestIRDLToCpp/TestIRDLToCppDialect.cpp
index bda614a..9550e4c 100644
--- a/mlir/test/lib/Dialect/TestIRDLToCpp/TestIRDLToCppDialect.cpp
+++ b/mlir/test/lib/Dialect/TestIRDLToCpp/TestIRDLToCppDialect.cpp
@@ -47,9 +47,9 @@ struct TestOpConversion : public OpConversionPattern<test_irdl_to_cpp::BeefOp> {
         op, op->getResultTypes().front());
     rewriter.setInsertionPointAfter(bar);
 
-    rewriter.create<test_irdl_to_cpp::HashOp>(
-        bar.getLoc(), rewriter.getIntegerType(32), adaptor.getLhs(),
-        adaptor.getRhs());
+    test_irdl_to_cpp::HashOp::create(rewriter, bar.getLoc(),
+                                     rewriter.getIntegerType(32),
+                                     adaptor.getLhs(), adaptor.getRhs());
     return success();
   }
 };
diff --git a/mlir/test/lib/Dialect/Tosa/TosaTestPasses.cpp b/mlir/test/lib/Dialect/Tosa/TosaTestPasses.cpp
index 3389a1c..6457487 100644
--- a/mlir/test/lib/Dialect/Tosa/TosaTestPasses.cpp
+++ b/mlir/test/lib/Dialect/Tosa/TosaTestPasses.cpp
@@ -87,9 +87,9 @@ ConvertTosaNegateOp::matchAndRewrite(Operation *op,
     return failure();
 
   auto newConstOp =
-      rewriter.create<tosa::ConstOp>(op->getLoc(), dstQConstType, inputElems);
-  auto newNegateOp = rewriter.create<tosa::NegateOp>(
-      op->getLoc(), dstQConstType, newConstOp.getResult());
+      tosa::ConstOp::create(rewriter, op->getLoc(), dstQConstType, inputElems);
+  auto newNegateOp = tosa::NegateOp::create(
+      rewriter, op->getLoc(), dstQConstType, newConstOp.getResult());
 
   rewriter.replaceOp(op, {newNegateOp.getResult()});
   return success();
@@ -145,8 +145,8 @@ ConvertTosaConv2DOp::matchAndRewrite(Operation *op,
   auto newTosaConv2DOpType =
       RankedTensorType::get(outputType.getShape(), rewriter.getIntegerType(32));
 
-  auto newTosaConv2DOp = rewriter.create<tosa::Conv2DOp>(
-      op->getLoc(), newTosaConv2DOpType, tosaConv2DOp.getInput(),
+  auto newTosaConv2DOp = tosa::Conv2DOp::create(
+      rewriter, op->getLoc(), newTosaConv2DOpType, tosaConv2DOp.getInput(),
       tosaConv2DOp.getWeight(), tosaConv2DOp.getBias(),
       tosaConv2DOp.getPadAttr(), tosaConv2DOp.getStrideAttr(),
       tosaConv2DOp.getDilationAttr(), tosaConv2DOp.getAccTypeAttr());
@@ -178,8 +178,8 @@ ConvertTosaConv2DOp::matchAndRewrite(Operation *op,
       newTosaConv2DOp.getResult().getType().isUnsignedInteger();
   bool outputUnsigned = outputType.isUnsignedInteger();
 
-  auto newTosaRescaleOp = rewriter.create<tosa::RescaleOp>(
-      op->getLoc(), outputType, newTosaConv2DOp.getResult(),
+  auto newTosaRescaleOp = tosa::RescaleOp::create(
+      rewriter, op->getLoc(), outputType, newTosaConv2DOp.getResult(),
       getConstTensorInt<int32_t>(rewriter, op->getLoc(), {multiplier}),
       getConstTensorInt<int8_t>(rewriter, op->getLoc(),
                                 {static_cast<int8_t>(shift)}),
diff --git a/mlir/test/lib/Dialect/Transform/TestTransformDialectExtension.cpp b/mlir/test/lib/Dialect/Transform/TestTransformDialectExtension.cpp
index cdf44c2..97fc699 100644
--- a/mlir/test/lib/Dialect/Transform/TestTransformDialectExtension.cpp
+++ b/mlir/test/lib/Dialect/Transform/TestTransformDialectExtension.cpp
@@ -796,8 +796,8 @@ DiagnosedSilenceableFailure mlir::test::TestProduceInvalidIR::applyToOne(
     transform::TransformState &state) {
   // Provide some IR that does not verify.
   rewriter.setInsertionPointToStart(&target->getRegion(0).front());
-  rewriter.create<TestDummyPayloadOp>(target->getLoc(), TypeRange(),
-                                      ValueRange(), /*failToVerify=*/true);
+  TestDummyPayloadOp::create(rewriter, target->getLoc(), TypeRange(),
+                             ValueRange(), /*failToVerify=*/true);
   return DiagnosedSilenceableFailure::success();
 }
 
@@ -877,7 +877,8 @@ public:
                                        Location loc) -> Value {
       if (inputs.size() != 1)
         return Value();
-      return builder.create<UnrealizedConversionCastOp>(loc, resultType, inputs)
+      return UnrealizedConversionCastOp::create(builder, loc, resultType,
+                                                inputs)
           .getResult(0);
     };
     addSourceMaterialization(unrealizedCastConverter);
diff --git a/mlir/test/lib/Dialect/Vector/TestVectorTransforms.cpp b/mlir/test/lib/Dialect/Vector/TestVectorTransforms.cpp
index a7285ab..f89c944 100644
--- a/mlir/test/lib/Dialect/Vector/TestVectorTransforms.cpp
+++ b/mlir/test/lib/Dialect/Vector/TestVectorTransforms.cpp
@@ -546,8 +546,8 @@ static Value allocateGlobalSharedMemory(Location loc, OpBuilder &builder,
 
   auto ip = builder.saveInsertionPoint();
   builder.setInsertionPoint(moduleOp);
-  auto global = builder.create<memref::GlobalOp>(
-      loc,
+  auto global = memref::GlobalOp::create(
+      builder, loc,
       /*sym_name=*/symbolName,
       /*sym_visibility=*/builder.getStringAttr("private"),
       /*type=*/memrefType,
@@ -560,19 +560,18 @@ static Value allocateGlobalSharedMemory(Location loc, OpBuilder &builder,
   global->moveBefore(&moduleOp.front());
 
   builder.restoreInsertionPoint(ip);
-  return builder.create<memref::GetGlobalOp>(loc, memrefType, symbolName);
+  return memref::GetGlobalOp::create(builder, loc, memrefType, symbolName);
 }
 
 static Value warpReduction(Location loc, OpBuilder &builder, Value input,
                            CombiningKind kind, uint32_t size) {
   // First reduce on a single thread to get per lane reduction value.
-  Value laneVal = builder.create<vector::ReductionOp>(loc, kind, input);
+  Value laneVal = vector::ReductionOp::create(builder, loc, kind, input);
   // Parallel reduction using butterfly shuffles.
   for (uint64_t i = 1; i < size; i <<= 1) {
-    Value shuffled = builder
-                         .create<gpu::ShuffleOp>(loc, laneVal, i,
-                                                 /*width=*/size,
-                                                 /*mode=*/gpu::ShuffleMode::XOR)
+    Value shuffled = gpu::ShuffleOp::create(builder, loc, laneVal, i,
+                                            /*width=*/size,
+                                            /*mode=*/gpu::ShuffleMode::XOR)
                          .getShuffleResult();
     laneVal = makeArithReduction(builder, loc, kind, laneVal, shuffled);
   }
@@ -647,12 +646,11 @@ struct TestVectorDistribution
              "unsupported shuffle type");
       Type i32Type = builder.getIntegerType(32);
       Value srcIdxI32 =
-          builder.create<arith::IndexCastOp>(loc, i32Type, srcIdx);
-      Value warpSzI32 = builder.create<arith::ConstantOp>(
-          loc, builder.getIntegerAttr(i32Type, warpSz));
-      Value result = builder
-                         .create<gpu::ShuffleOp>(loc, val, srcIdxI32, warpSzI32,
-                                                 gpu::ShuffleMode::IDX)
+          arith::IndexCastOp::create(builder, loc, i32Type, srcIdx);
+      Value warpSzI32 = arith::ConstantOp::create(
+          builder, loc, builder.getIntegerAttr(i32Type, warpSz));
+      Value result = gpu::ShuffleOp::create(builder, loc, val, srcIdxI32,
+                                            warpSzI32, gpu::ShuffleMode::IDX)
                          .getResult(0);
       return result;
     };
@@ -680,7 +678,7 @@ struct TestVectorDistribution
     options.warpAllocationFn = allocateGlobalSharedMemory;
     options.warpSyncronizationFn = [](Location loc, OpBuilder &builder,
                                       gpu::WarpExecuteOnLane0Op warpOp) {
-      builder.create<gpu::BarrierOp>(loc);
+      gpu::BarrierOp::create(builder, loc);
     };
     // Test on one pattern in isolation.
     if (warpOpToSCF) {
diff --git a/mlir/test/lib/Dialect/XeGPU/TestXeGPUTransforms.cpp b/mlir/test/lib/Dialect/XeGPU/TestXeGPUTransforms.cpp
index f71fcf7..c6245b6 100644
--- a/mlir/test/lib/Dialect/XeGPU/TestXeGPUTransforms.cpp
+++ b/mlir/test/lib/Dialect/XeGPU/TestXeGPUTransforms.cpp
@@ -20,8 +20,6 @@ using namespace mlir::xegpu;
 namespace {
 
 #define DEBUG_TYPE "test-xegpu-unroll"
-#define DBGS() (llvm::dbgs() << "[" DEBUG_TYPE "]: ")
-#define LDBG(X) LLVM_DEBUG(DBGS() << X << "\n")
 
 struct TestXeGPUUnrollingPatterns
     : public PassWrapper<TestXeGPUUnrollingPatterns,
diff --git a/mlir/test/lib/IR/TestPrintInvalid.cpp b/mlir/test/lib/IR/TestPrintInvalid.cpp
index 8697918..25d1b19 100644
--- a/mlir/test/lib/IR/TestPrintInvalid.cpp
+++ b/mlir/test/lib/IR/TestPrintInvalid.cpp
@@ -34,13 +34,14 @@ struct TestPrintInvalidPass
   void runOnOperation() override {
     Location loc = getOperation().getLoc();
     OpBuilder builder(getOperation().getBodyRegion());
-    auto funcOp = builder.create<func::FuncOp>(
-        loc, "test", FunctionType::get(getOperation().getContext(), {}, {}));
+    auto funcOp = func::FuncOp::create(
+        builder, loc, "test",
+        FunctionType::get(getOperation().getContext(), {}, {}));
     funcOp.addEntryBlock();
     // The created function is invalid because there is no return op.
     llvm::outs() << "Invalid operation:\n" << funcOp << "\n";
     builder.setInsertionPointToEnd(&funcOp.getBody().front());
-    builder.create<func::ReturnOp>(loc);
+    func::ReturnOp::create(builder, loc);
     // Now this function is valid.
     llvm::outs() << "Valid operation:\n" << funcOp << "\n";
     funcOp.erase();
diff --git a/mlir/test/lib/IR/TestSlicing.cpp b/mlir/test/lib/IR/TestSlicing.cpp
index 92fd6de..5a5ac45 100644
--- a/mlir/test/lib/IR/TestSlicing.cpp
+++ b/mlir/test/lib/IR/TestSlicing.cpp
@@ -30,8 +30,8 @@ static LogicalResult createBackwardSliceFunction(Operation *op,
   OpBuilder builder(parentFuncOp);
   Location loc = op->getLoc();
   std::string clonedFuncOpName = parentFuncOp.getName().str() + suffix.str();
-  func::FuncOp clonedFuncOp = builder.create<func::FuncOp>(
-      loc, clonedFuncOpName, parentFuncOp.getFunctionType());
+  func::FuncOp clonedFuncOp = func::FuncOp::create(
+      builder, loc, clonedFuncOpName, parentFuncOp.getFunctionType());
   IRMapping mapper;
   builder.setInsertionPointToEnd(clonedFuncOp.addEntryBlock());
   for (const auto &arg : enumerate(parentFuncOp.getArguments()))
@@ -46,7 +46,7 @@ static LogicalResult createBackwardSliceFunction(Operation *op,
   (void)result;
   for (Operation *slicedOp : slice)
     builder.clone(*slicedOp, mapper);
-  builder.create<func::ReturnOp>(loc);
+  func::ReturnOp::create(builder, loc);
   return success();
 }
 
diff --git a/mlir/test/lib/Pass/TestPassManager.cpp b/mlir/test/lib/Pass/TestPassManager.cpp
index 7afe210..25c8e53 100644
--- a/mlir/test/lib/Pass/TestPassManager.cpp
+++ b/mlir/test/lib/Pass/TestPassManager.cpp
@@ -217,8 +217,8 @@ struct TestInvalidParentPass
   void runOnOperation() final {
     FunctionOpInterface op = getOperation();
     OpBuilder b(op.getFunctionBody());
-    b.create<test::TestCallOp>(op.getLoc(), TypeRange(), "some_unknown_func",
-                               ValueRange());
+    test::TestCallOp::create(b, op.getLoc(), TypeRange(), "some_unknown_func",
+                             ValueRange());
   }
 };
 
diff --git a/mlir/test/lib/Transforms/TestDialectConversion.cpp b/mlir/test/lib/Transforms/TestDialectConversion.cpp
index 8278937..dc0538e 100644
--- a/mlir/test/lib/Transforms/TestDialectConversion.cpp
+++ b/mlir/test/lib/Transforms/TestDialectConversion.cpp
@@ -45,7 +45,7 @@ struct PDLLTypeConverter : public TypeConverter {
   /// Hook for materializing a conversion.
   static Value materializeCast(OpBuilder &builder, Type resultType,
                                ValueRange inputs, Location loc) {
-    return builder.create<UnrealizedConversionCastOp>(loc, resultType, inputs)
+    return UnrealizedConversionCastOp::create(builder, loc, resultType, inputs)
         .getResult(0);
   }
 };
diff --git a/mlir/test/lib/Transforms/TestInliningCallback.cpp b/mlir/test/lib/Transforms/TestInliningCallback.cpp
index c518f3f..2888c3c 100644
--- a/mlir/test/lib/Transforms/TestInliningCallback.cpp
+++ b/mlir/test/lib/Transforms/TestInliningCallback.cpp
@@ -53,8 +53,8 @@ struct InlinerCallback
     mlir::Operation &call = inlineBlock->back();
     builder.setInsertionPointAfter(&call);
 
-    auto executeRegionOp = builder.create<mlir::scf::ExecuteRegionOp>(
-        call.getLoc(), call.getResultTypes());
+    auto executeRegionOp = mlir::scf::ExecuteRegionOp::create(
+        builder, call.getLoc(), call.getResultTypes());
     mlir::Region &region = executeRegionOp.getRegion();
 
     // Move the inlined blocks into the region
@@ -70,8 +70,8 @@ struct InlinerCallback
         if (test::TestReturnOp returnOp =
                 llvm::dyn_cast<test::TestReturnOp>(&op)) {
           mlir::OpBuilder returnBuilder(returnOp);
-          returnBuilder.create<mlir::scf::YieldOp>(returnOp.getLoc(),
-                                                   returnOp.getOperands());
+          mlir::scf::YieldOp::create(returnBuilder, returnOp.getLoc(),
+                                     returnOp.getOperands());
           returnOp.erase();
         }
       }
@@ -79,8 +79,8 @@ struct InlinerCallback
 
     // Add test.return after scf.execute_region
     builder.setInsertionPointAfter(executeRegionOp);
-    builder.create<test::TestReturnOp>(executeRegionOp.getLoc(),
-                                       executeRegionOp.getResults());
+    test::TestReturnOp::create(builder, executeRegionOp.getLoc(),
+                               executeRegionOp.getResults());
   }
 
   void runOnOperation() override {
diff --git a/mlir/test/lib/Transforms/TestMakeIsolatedFromAbove.cpp b/mlir/test/lib/Transforms/TestMakeIsolatedFromAbove.cpp
index 4e0213c..c1fb706 100644
--- a/mlir/test/lib/Transforms/TestMakeIsolatedFromAbove.cpp
+++ b/mlir/test/lib/Transforms/TestMakeIsolatedFromAbove.cpp
@@ -28,7 +28,7 @@ makeIsolatedFromAboveImpl(RewriterBase &rewriter,
   SmallVector<Value> operands = regionOp.getOperands();
   operands.append(capturedValues);
   auto isolatedRegionOp =
-      rewriter.create<test::IsolatedOneRegionOp>(regionOp.getLoc(), operands);
+      test::IsolatedOneRegionOp::create(rewriter, regionOp.getLoc(), operands);
   rewriter.inlineRegionBefore(region, isolatedRegionOp.getRegion(),
                               isolatedRegionOp.getRegion().begin());
   rewriter.eraseOp(regionOp);
diff --git a/mlir/test/lib/Transforms/TestTransformsOps.cpp b/mlir/test/lib/Transforms/TestTransformsOps.cpp
index 9a5632b..ff5838d 100644
--- a/mlir/test/lib/Transforms/TestTransformsOps.cpp
+++ b/mlir/test/lib/Transforms/TestTransformsOps.cpp
@@ -74,8 +74,8 @@ transform::TestMakeComposedFoldedAffineApply::applyToOne(
   if (auto v = dyn_cast<Value>(ofr)) {
     result = v;
   } else {
-    result = rewriter.create<arith::ConstantIndexOp>(
-        loc, getConstantIntValue(ofr).value());
+    result = arith::ConstantIndexOp::create(rewriter, loc,
+                                            getConstantIntValue(ofr).value());
   }
   results.push_back(result.getDefiningOp());
   rewriter.replaceOp(affineApplyOp, result);
diff --git a/mlir/test/mlir-runner/simple.mlir b/mlir/test/mlir-runner/simple.mlir
index 1a03b99..21dabdd 100644
--- a/mlir/test/mlir-runner/simple.mlir
+++ b/mlir/test/mlir-runner/simple.mlir
@@ -15,10 +15,10 @@
 // RUN: ls %t.o
 // RUN: rm %t.o
 
-// RUN: mlir-runner %s -dump-object-file -object-filename=%T/test.o \
+// RUN: mlir-runner %s -dump-object-file -object-filename=%t.o \
 // RUN:   %if target={{s390x-.*}} %{ -argext-abi-check=false %} | FileCheck %s
-// RUN: ls %T/test.o
-// RUN: rm %T/test.o
+// RUN: ls %t.o
+// RUN: rm %t.o
 
 // Declarations of C library functions.
 llvm.func @logbf(f32) -> f32
diff --git a/mlir/test/mlir-tblgen/rewriter-attributes-properties.td b/mlir/test/mlir-tblgen/rewriter-attributes-properties.td
index 40af548..23ab24e 100644
--- a/mlir/test/mlir-tblgen/rewriter-attributes-properties.td
+++ b/mlir/test/mlir-tblgen/rewriter-attributes-properties.td
@@ -44,7 +44,7 @@ def test1 : Pat<(AOp (BOp:$x $y), $_), (AOp $x, $y)>;
 // CHECK: test::AOp::Properties tblgen_props;
 // CHECK: tblgen_values.push_back((*x.getODSResults(0).begin()));
 // CHECK: tblgen_props.y = ::llvm::dyn_cast_if_present<decltype(tblgen_props.y)>(y);
-// CHECK: tblgen_AOp_0 = rewriter.create<test::AOp>(odsLoc, tblgen_types, tblgen_values, tblgen_props);
+// CHECK: tblgen_AOp_0 = test::AOp::create(rewriter, odsLoc, tblgen_types, tblgen_values, tblgen_props);
 
 // Note: These use strings to pick up a non-trivial storage/interface type
 // difference.
diff --git a/mlir/test/mlir-tblgen/rewriter-indexing.td b/mlir/test/mlir-tblgen/rewriter-indexing.td
index 0a94746..9bb6103 100644
--- a/mlir/test/mlir-tblgen/rewriter-indexing.td
+++ b/mlir/test/mlir-tblgen/rewriter-indexing.td
@@ -55,7 +55,7 @@ def test2 : Pat<(COp $attr1, $op1, $attr2, (AOp $op2)),
 // We expect ODSOperand 0 here, the attribute before the operand in BOp
 // definition shouldn't shift the counter.
 // CHECK: op1 = (*castedOp0.getODSOperands(0).begin()).getDefiningOp();
-// CHECK: rewriter.create<test::BOp>((*a.getODSResults(0).begin()).getLoc()
+// CHECK: test::BOp::create(rewriter, (*a.getODSResults(0).begin()).getLoc()
 def test3 : Pat<(BOp $attr, (AOp:$a $input)),
                 (BOp $attr, (AOp $input), (location $a))>;
 
diff --git a/mlir/test/python/ir/array_attributes.py b/mlir/test/python/ir/array_attributes.py
index ef1d835..66f7ec8 100644
--- a/mlir/test/python/ir/array_attributes.py
+++ b/mlir/test/python/ir/array_attributes.py
@@ -31,6 +31,7 @@ def testGetDenseElementsUnsupported():
             # CHECK: unimplemented array format conversion from format:
             print(e)
 
+
 # CHECK-LABEL: TEST: testGetDenseElementsUnSupportedTypeOkIfExplicitTypeProvided
 @run
 def testGetDenseElementsUnSupportedTypeOkIfExplicitTypeProvided():
@@ -41,8 +42,9 @@ def testGetDenseElementsUnSupportedTypeOkIfExplicitTypeProvided():
         # realistic example would be a NumPy extension type like the bfloat16
         # type from the ml_dtypes package, which isn't a dependency of this
         # test.
-        attr = DenseElementsAttr.get(array.view(np.datetime64),
-                                     type=IntegerType.get_signless(64))
+        attr = DenseElementsAttr.get(
+            array.view(np.datetime64), type=IntegerType.get_signless(64)
+        )
         # CHECK: dense<{{\[}}[1, 2, 3], [4, 5, 6]]> : tensor<2x3xi64>
         print(attr)
         # CHECK: {{\[}}[1 2 3]
@@ -135,6 +137,7 @@ def testGetDenseElementsFromListMixedTypes():
 # Splats.
 ################################################################################
 
+
 # CHECK-LABEL: TEST: testGetDenseElementsSplatInt
 @run
 def testGetDenseElementsSplatInt():
@@ -617,3 +620,18 @@ def testGetDenseResourceElementsAttr():
     # CHECK: BACKING MEMORY DELETED
     # CHECK: EXIT FUNCTION
     print("EXIT FUNCTION")
+
+
+# CHECK-LABEL: TEST: testDanglingResource
+print("TEST: testDanglingResource")
+# see https://github.com/llvm/llvm-project/pull/149414, https://github.com/llvm/llvm-project/pull/150137, https://github.com/llvm/llvm-project/pull/150561
+# This error occurs only when there is an alive context with a DenseResourceElementsAttr
+# in the end of the program, so we put it here without an encapsulating function.
+ctx = Context()
+
+with ctx, Location.unknown():
+    DenseResourceElementsAttr.get_from_buffer(
+        memoryview(np.array([1, 2, 3])),
+        "some_resource",
+        RankedTensorType.get((3,), IntegerType.get_signed(32)),
+    )
diff --git a/mlir/tools/mlir-opt/CMakeLists.txt b/mlir/tools/mlir-opt/CMakeLists.txt
index 26d7597..6958fe3 100644
--- a/mlir/tools/mlir-opt/CMakeLists.txt
+++ b/mlir/tools/mlir-opt/CMakeLists.txt
@@ -29,7 +29,7 @@ if(MLIR_INCLUDE_TESTS)
     MLIRTestMathToVCIX
     MLIRMemRefTestPasses
     MLIRTestMemRefToLLVMWithTransforms
-    MLIRMeshTest
+    MLIRShardTest
     MLIRNVGPUTestPasses
     MLIRSCFTestPasses
     MLIRShapeTestPasses
diff --git a/mlir/tools/mlir-opt/mlir-opt.cpp b/mlir/tools/mlir-opt/mlir-opt.cpp
index 143a5e8..2c09753 100644
--- a/mlir/tools/mlir-opt/mlir-opt.cpp
+++ b/mlir/tools/mlir-opt/mlir-opt.cpp
@@ -130,8 +130,8 @@ void registerTestIrdlTestDialectConversionPass();
 void registerTestMemRefDependenceCheck();
 void registerTestMemRefStrideCalculation();
 void registerTestMemRefToLLVMWithTransforms();
-void registerTestMeshReshardingSpmdizationPass();
-void registerTestMeshSimplificationsPass();
+void registerTestReshardingPartitionPass();
+void registerTestShardSimplificationsPass();
 void registerTestMultiBuffering();
 void registerTestNextAccessPass();
 void registerTestNVGPULowerings();
@@ -276,8 +276,8 @@ void registerTestPasses() {
   mlir::test::registerTestMemRefDependenceCheck();
   mlir::test::registerTestMemRefStrideCalculation();
   mlir::test::registerTestMemRefToLLVMWithTransforms();
-  mlir::test::registerTestMeshReshardingSpmdizationPass();
-  mlir::test::registerTestMeshSimplificationsPass();
+  mlir::test::registerTestReshardingPartitionPass();
+  mlir::test::registerTestShardSimplificationsPass();
   mlir::test::registerTestMultiBuffering();
   mlir::test::registerTestNextAccessPass();
   mlir::test::registerTestNVGPULowerings();
diff --git a/mlir/tools/mlir-tblgen/RewriterGen.cpp b/mlir/tools/mlir-tblgen/RewriterGen.cpp
index 975a524..605033d 100644
--- a/mlir/tools/mlir-tblgen/RewriterGen.cpp
+++ b/mlir/tools/mlir-tblgen/RewriterGen.cpp
@@ -632,7 +632,8 @@ void PatternEmitter::emitOpMatch(DagNode tree, StringRef opName, int depth) {
         ++opArgIdx;
         continue;
       }
-      if (auto *operand = llvm::dyn_cast_if_present<NamedTypeConstraint *>(opArg)) {
+      if (auto *operand =
+              llvm::dyn_cast_if_present<NamedTypeConstraint *>(opArg)) {
         if (argTree.isVariadic()) {
           if (!operand->isVariadic()) {
             auto error = formatv("variadic DAG construct can't match op {0}'s "
@@ -1695,7 +1696,7 @@ std::string PatternEmitter::handleOpCreation(DagNode tree, int resultIndex,
 
     // Then create the op.
     os.scope("", "\n}\n").os
-        << formatv("{0} = rewriter.create<{1}>({2}, tblgen_values, {3});",
+        << formatv("{0} = {1}::create(rewriter, {2}, tblgen_values, {3});",
                    valuePackName, resultOp.getQualCppClassName(), locToUse,
                    useProperties ? "tblgen_props" : "tblgen_attrs");
     return resultValue;
@@ -1714,7 +1715,7 @@ std::string PatternEmitter::handleOpCreation(DagNode tree, int resultIndex,
     // aggregate-parameter builders.
     createSeparateLocalVarsForOpArgs(tree, childNodeNames);
 
-    os.scope().os << formatv("{0} = rewriter.create<{1}>({2}", valuePackName,
+    os.scope().os << formatv("{0} = {1}::create(rewriter, {2}", valuePackName,
                              resultOp.getQualCppClassName(), locToUse);
     supplyValuesForOpArgs(tree, childNodeNames, depth);
     os << "\n  );\n}\n";
@@ -1753,7 +1754,7 @@ std::string PatternEmitter::handleOpCreation(DagNode tree, int resultIndex,
                       resultIndex + i);
     }
   }
-  os << formatv("{0} = rewriter.create<{1}>({2}, tblgen_types, "
+  os << formatv("{0} = {1}::create(rewriter, {2}, tblgen_types, "
                 "tblgen_values, {3});\n",
                 valuePackName, resultOp.getQualCppClassName(), locToUse,
                 useProperties ? "tblgen_props" : "tblgen_attrs");
@@ -1772,8 +1773,8 @@ void PatternEmitter::createSeparateLocalVarsForOpArgs(
 
   int valueIndex = 0; // An index for uniquing local variable names.
   for (int argIndex = 0, e = resultOp.getNumArgs(); argIndex < e; ++argIndex) {
-    const auto *operand =
-        llvm::dyn_cast_if_present<NamedTypeConstraint *>(resultOp.getArg(argIndex));
+    const auto *operand = llvm::dyn_cast_if_present<NamedTypeConstraint *>(
+        resultOp.getArg(argIndex));
     // We do not need special handling for attributes or properties.
     if (!operand)
       continue;
@@ -1828,7 +1829,8 @@ void PatternEmitter::supplyValuesForOpArgs(
 
     Argument opArg = resultOp.getArg(argIndex);
     // Handle the case of operand first.
-    if (auto *operand = llvm::dyn_cast_if_present<NamedTypeConstraint *>(opArg)) {
+    if (auto *operand =
+            llvm::dyn_cast_if_present<NamedTypeConstraint *>(opArg)) {
       if (!operand->name.empty())
         os << "/*" << operand->name << "=*/";
       os << childNodeNames.lookup(argIndex);
diff --git a/mlir/tools/mlir-tblgen/SPIRVUtilsGen.cpp b/mlir/tools/mlir-tblgen/SPIRVUtilsGen.cpp
index 7256705..41ffdfc 100644
--- a/mlir/tools/mlir-tblgen/SPIRVUtilsGen.cpp
+++ b/mlir/tools/mlir-tblgen/SPIRVUtilsGen.cpp
@@ -397,10 +397,9 @@ static void emitAvailabilityQueryForBitEnum(const Record &enumDef,
                   avail.getMergeInstanceType(), avail.getQueryFnName(),
                   enumName);
 
-    os << formatv(
-        "  assert(::llvm::popcount(static_cast<{0}>(value)) <= 1"
-        " && \"cannot have more than one bit set\");\n",
-        underlyingType);
+    os << formatv("  assert(::llvm::popcount(static_cast<{0}>(value)) <= 1"
+                  " && \"cannot have more than one bit set\");\n",
+                  underlyingType);
 
     os << "  switch (value) {\n";
     for (const auto &caseSpecPair : classCasePair.getValue()) {
@@ -933,7 +932,8 @@ static void emitOperandDeserialization(const Operator &op, ArrayRef<SMLoc> loc,
   // Process operands/attributes
   for (unsigned i = 0, e = op.getNumArgs(); i < e; ++i) {
     auto argument = op.getArg(i);
-    if (auto *valueArg = llvm::dyn_cast_if_present<NamedTypeConstraint *>(argument)) {
+    if (auto *valueArg =
+            llvm::dyn_cast_if_present<NamedTypeConstraint *>(argument)) {
       if (valueArg->isVariableLength()) {
         if (i != e - 1) {
           PrintFatalError(
@@ -1044,7 +1044,7 @@ static void emitDeserializationFunction(const Record *attrClass,
   emitDecorationDeserialization(op, "  ", valueID, attributes, os);
 
   os << formatv("  Location loc = createFileLineColLoc(opBuilder);\n");
-  os << formatv("  auto {1} = opBuilder.create<{0}>(loc, {2}, {3}, {4}); "
+  os << formatv("  auto {1} = {0}::create(opBuilder, loc, {2}, {3}, {4}); "
                 "(void){1};\n",
                 op.getQualCppClassName(), opVar, resultTypes, operands,
                 attributes);
diff --git a/mlir/tools/tblgen-to-irdl/OpDefinitionsGen.cpp b/mlir/tools/tblgen-to-irdl/OpDefinitionsGen.cpp
index c2ad09f..4343f2d 100644
--- a/mlir/tools/tblgen-to-irdl/OpDefinitionsGen.cpp
+++ b/mlir/tools/tblgen-to-irdl/OpDefinitionsGen.cpp
@@ -52,33 +52,33 @@ Value createPredicate(OpBuilder &builder, tblgen::Pred pred) {
       }
       if (combiner == "PredCombinerAnd") {
         auto op =
-            builder.create<irdl::AllOfOp>(UnknownLoc::get(ctx), constraints);
+            irdl::AllOfOp::create(builder, UnknownLoc::get(ctx), constraints);
         return op.getOutput();
       }
       auto op =
-          builder.create<irdl::AnyOfOp>(UnknownLoc::get(ctx), constraints);
+          irdl::AnyOfOp::create(builder, UnknownLoc::get(ctx), constraints);
       return op.getOutput();
     }
   }
 
   std::string condition = pred.getCondition();
   // Build a CPredOp to match the C constraint built.
-  irdl::CPredOp op = builder.create<irdl::CPredOp>(
-      UnknownLoc::get(ctx), StringAttr::get(ctx, condition));
+  irdl::CPredOp op = irdl::CPredOp::create(builder, UnknownLoc::get(ctx),
+                                           StringAttr::get(ctx, condition));
   return op;
 }
 
 Value typeToConstraint(OpBuilder &builder, Type type) {
   MLIRContext *ctx = builder.getContext();
   auto op =
-      builder.create<irdl::IsOp>(UnknownLoc::get(ctx), TypeAttr::get(type));
+      irdl::IsOp::create(builder, UnknownLoc::get(ctx), TypeAttr::get(type));
   return op.getOutput();
 }
 
 Value baseToConstraint(OpBuilder &builder, StringRef baseClass) {
   MLIRContext *ctx = builder.getContext();
-  auto op = builder.create<irdl::BaseOp>(UnknownLoc::get(ctx),
-                                         StringAttr::get(ctx, baseClass));
+  auto op = irdl::BaseOp::create(builder, UnknownLoc::get(ctx),
+                                 StringAttr::get(ctx, baseClass));
   return op.getOutput();
 }
 
@@ -179,7 +179,7 @@ Value createTypeConstraint(OpBuilder &builder, tblgen::Constraint constraint) {
     return createTypeConstraint(builder, predRec.getValueAsDef("baseType"));
 
   if (predRec.getName() == "AnyType") {
-    auto op = builder.create<irdl::AnyOp>(UnknownLoc::get(ctx));
+    auto op = irdl::AnyOp::create(builder, UnknownLoc::get(ctx));
     return op.getOutput();
   }
 
@@ -190,12 +190,12 @@ Value createTypeConstraint(OpBuilder &builder, tblgen::Constraint constraint) {
       SmallVector<FlatSymbolRefAttr> nested = {
           SymbolRefAttr::get(ctx, combined)};
       auto typeSymbol = SymbolRefAttr::get(ctx, dialect, nested);
-      auto op = builder.create<irdl::BaseOp>(UnknownLoc::get(ctx), typeSymbol);
+      auto op = irdl::BaseOp::create(builder, UnknownLoc::get(ctx), typeSymbol);
       return op.getOutput();
     }
     std::string typeName = ("!" + predRec.getValueAsString("typeName")).str();
-    auto op = builder.create<irdl::BaseOp>(UnknownLoc::get(ctx),
-                                           StringAttr::get(ctx, typeName));
+    auto op = irdl::BaseOp::create(builder, UnknownLoc::get(ctx),
+                                   StringAttr::get(ctx, typeName));
     return op.getOutput();
   }
 
@@ -205,7 +205,7 @@ Value createTypeConstraint(OpBuilder &builder, tblgen::Constraint constraint) {
       constraints.push_back(
           createTypeConstraint(builder, tblgen::Constraint(child)));
     }
-    auto op = builder.create<irdl::AnyOfOp>(UnknownLoc::get(ctx), constraints);
+    auto op = irdl::AnyOfOp::create(builder, UnknownLoc::get(ctx), constraints);
     return op.getOutput();
   }
 
@@ -215,14 +215,14 @@ Value createTypeConstraint(OpBuilder &builder, tblgen::Constraint constraint) {
       constraints.push_back(
           createTypeConstraint(builder, tblgen::Constraint(child)));
     }
-    auto op = builder.create<irdl::AllOfOp>(UnknownLoc::get(ctx), constraints);
+    auto op = irdl::AllOfOp::create(builder, UnknownLoc::get(ctx), constraints);
     return op.getOutput();
   }
 
   // Integer types
   if (predRec.getName() == "AnyInteger") {
-    auto op = builder.create<irdl::BaseOp>(
-        UnknownLoc::get(ctx), StringAttr::get(ctx, "!builtin.integer"));
+    auto op = irdl::BaseOp::create(builder, UnknownLoc::get(ctx),
+                                   StringAttr::get(ctx, "!builtin.integer"));
     return op.getOutput();
   }
 
@@ -235,7 +235,7 @@ Value createTypeConstraint(OpBuilder &builder, tblgen::Constraint constraint) {
                          IntegerType::get(ctx, width, IntegerType::Signed)),
         typeToConstraint(builder,
                          IntegerType::get(ctx, width, IntegerType::Unsigned))};
-    auto op = builder.create<irdl::AnyOfOp>(UnknownLoc::get(ctx), types);
+    auto op = irdl::AnyOfOp::create(builder, UnknownLoc::get(ctx), types);
     return op.getOutput();
   }
 
@@ -253,7 +253,7 @@ Value createTypeConstraint(OpBuilder &builder, tblgen::Constraint constraint) {
     for (const Record *child : predRec.getValueAsListOfDefs("predicateList")) {
       constraints.push_back(createPredicate(builder, tblgen::Pred(child)));
     }
-    auto op = builder.create<irdl::AllOfOp>(UnknownLoc::get(ctx), constraints);
+    auto op = irdl::AllOfOp::create(builder, UnknownLoc::get(ctx), constraints);
     return op.getOutput();
   }
 
@@ -279,7 +279,7 @@ Value createAttrConstraint(OpBuilder &builder, tblgen::Constraint constraint) {
       constraints.push_back(createPredicate(
           builder, tblgen::Pred(child->getValueAsDef("predicate"))));
     }
-    auto op = builder.create<irdl::AllOfOp>(UnknownLoc::get(ctx), constraints);
+    auto op = irdl::AllOfOp::create(builder, UnknownLoc::get(ctx), constraints);
     return op.getOutput();
   }
 
@@ -290,12 +290,12 @@ Value createAttrConstraint(OpBuilder &builder, tblgen::Constraint constraint) {
       constraints.push_back(
           createAttrConstraint(builder, tblgen::Constraint(child)));
     }
-    auto op = builder.create<irdl::AnyOfOp>(UnknownLoc::get(ctx), constraints);
+    auto op = irdl::AnyOfOp::create(builder, UnknownLoc::get(ctx), constraints);
     return op.getOutput();
   }
 
   if (predRec.getName() == "AnyAttr") {
-    auto op = builder.create<irdl::AnyOp>(UnknownLoc::get(ctx));
+    auto op = irdl::AnyOp::create(builder, UnknownLoc::get(ctx));
     return op.getOutput();
   }
 
@@ -317,7 +317,7 @@ Value createAttrConstraint(OpBuilder &builder, tblgen::Constraint constraint) {
 
   if (predRec.getName() == "UnitAttr") {
     auto op =
-        builder.create<irdl::IsOp>(UnknownLoc::get(ctx), UnitAttr::get(ctx));
+        irdl::IsOp::create(builder, UnknownLoc::get(ctx), UnitAttr::get(ctx));
     return op.getOutput();
   }
 
@@ -329,12 +329,12 @@ Value createAttrConstraint(OpBuilder &builder, tblgen::Constraint constraint) {
 
       };
       auto typeSymbol = SymbolRefAttr::get(ctx, dialect, nested);
-      auto op = builder.create<irdl::BaseOp>(UnknownLoc::get(ctx), typeSymbol);
+      auto op = irdl::BaseOp::create(builder, UnknownLoc::get(ctx), typeSymbol);
       return op.getOutput();
     }
     std::string typeName = ("#" + predRec.getValueAsString("attrName")).str();
-    auto op = builder.create<irdl::BaseOp>(UnknownLoc::get(ctx),
-                                           StringAttr::get(ctx, typeName));
+    auto op = irdl::BaseOp::create(builder, UnknownLoc::get(ctx),
+                                   StringAttr::get(ctx, typeName));
     return op.getOutput();
   }
 
@@ -348,15 +348,15 @@ Value createRegionConstraint(OpBuilder &builder, tblgen::Region constraint) {
   if (predRec.getName() == "AnyRegion") {
     ValueRange entryBlockArgs = {};
     auto op =
-        builder.create<irdl::RegionOp>(UnknownLoc::get(ctx), entryBlockArgs);
+        irdl::RegionOp::create(builder, UnknownLoc::get(ctx), entryBlockArgs);
     return op.getResult();
   }
 
   if (predRec.isSubClassOf("SizedRegion")) {
     ValueRange entryBlockArgs = {};
     auto ty = IntegerType::get(ctx, 32);
-    auto op = builder.create<irdl::RegionOp>(
-        UnknownLoc::get(ctx), entryBlockArgs,
+    auto op = irdl::RegionOp::create(
+        builder, UnknownLoc::get(ctx), entryBlockArgs,
         IntegerAttr::get(ty, predRec.getValueAsInt("blocks")));
     return op.getResult();
   }
@@ -388,8 +388,8 @@ irdl::OperationOp createIRDLOperation(OpBuilder &builder,
   MLIRContext *ctx = builder.getContext();
   StringRef opName = getOperatorName(tblgenOp);
 
-  irdl::OperationOp op = builder.create<irdl::OperationOp>(
-      UnknownLoc::get(ctx), StringAttr::get(ctx, opName));
+  irdl::OperationOp op = irdl::OperationOp::create(
+      builder, UnknownLoc::get(ctx), StringAttr::get(ctx, opName));
 
   // Add the block in the region.
   Block &opBlock = op.getBody().emplaceBlock();
@@ -471,19 +471,19 @@ irdl::OperationOp createIRDLOperation(OpBuilder &builder,
 
   // Create the operands and results operations.
   if (!operands.empty())
-    consBuilder.create<irdl::OperandsOp>(UnknownLoc::get(ctx), operands,
-                                         ArrayAttr::get(ctx, operandNames),
-                                         operandVariadicity);
+    irdl::OperandsOp::create(consBuilder, UnknownLoc::get(ctx), operands,
+                             ArrayAttr::get(ctx, operandNames),
+                             operandVariadicity);
   if (!results.empty())
-    consBuilder.create<irdl::ResultsOp>(UnknownLoc::get(ctx), results,
-                                        ArrayAttr::get(ctx, resultNames),
-                                        resultVariadicity);
+    irdl::ResultsOp::create(consBuilder, UnknownLoc::get(ctx), results,
+                            ArrayAttr::get(ctx, resultNames),
+                            resultVariadicity);
   if (!attributes.empty())
-    consBuilder.create<irdl::AttributesOp>(UnknownLoc::get(ctx), attributes,
-                                           ArrayAttr::get(ctx, attrNames));
+    irdl::AttributesOp::create(consBuilder, UnknownLoc::get(ctx), attributes,
+                               ArrayAttr::get(ctx, attrNames));
   if (!regions.empty())
-    consBuilder.create<irdl::RegionsOp>(UnknownLoc::get(ctx), regions,
-                                        ArrayAttr::get(ctx, regionNames));
+    irdl::RegionsOp::create(consBuilder, UnknownLoc::get(ctx), regions,
+                            ArrayAttr::get(ctx, regionNames));
 
   return op;
 }
@@ -493,8 +493,8 @@ irdl::TypeOp createIRDLType(OpBuilder &builder, tblgen::TypeDef &tblgenType) {
   StringRef typeName = getTypeName(tblgenType);
   std::string combined = ("!" + typeName).str();
 
-  irdl::TypeOp op = builder.create<irdl::TypeOp>(
-      UnknownLoc::get(ctx), StringAttr::get(ctx, combined));
+  irdl::TypeOp op = irdl::TypeOp::create(builder, UnknownLoc::get(ctx),
+                                         StringAttr::get(ctx, combined));
 
   op.getBody().emplaceBlock();
 
@@ -507,8 +507,8 @@ irdl::AttributeOp createIRDLAttr(OpBuilder &builder,
   StringRef attrName = getAttrName(tblgenAttr);
   std::string combined = ("#" + attrName).str();
 
-  irdl::AttributeOp op = builder.create<irdl::AttributeOp>(
-      UnknownLoc::get(ctx), StringAttr::get(ctx, combined));
+  irdl::AttributeOp op = irdl::AttributeOp::create(
+      builder, UnknownLoc::get(ctx), StringAttr::get(ctx, combined));
 
   op.getBody().emplaceBlock();
 
@@ -517,8 +517,8 @@ irdl::AttributeOp createIRDLAttr(OpBuilder &builder,
 
 static irdl::DialectOp createIRDLDialect(OpBuilder &builder) {
   MLIRContext *ctx = builder.getContext();
-  return builder.create<irdl::DialectOp>(UnknownLoc::get(ctx),
-                                         StringAttr::get(ctx, selectedDialect));
+  return irdl::DialectOp::create(builder, UnknownLoc::get(ctx),
+                                 StringAttr::get(ctx, selectedDialect));
 }
 
 static bool emitDialectIRDLDefs(const RecordKeeper &records, raw_ostream &os) {
@@ -529,7 +529,7 @@ static bool emitDialectIRDLDefs(const RecordKeeper &records, raw_ostream &os) {
 
   // Create a module op and set it as the insertion point.
   OwningOpRef<ModuleOp> module =
-      builder.create<ModuleOp>(UnknownLoc::get(&ctx));
+      ModuleOp::create(builder, UnknownLoc::get(&ctx));
   builder = builder.atBlockBegin(module->getBody());
   // Create the dialect and insert it.
   irdl::DialectOp dialect = createIRDLDialect(builder);
diff --git a/mlir/unittests/Conversion/PDLToPDLInterp/RootOrderingTest.cpp b/mlir/unittests/Conversion/PDLToPDLInterp/RootOrderingTest.cpp
index f82ece0..020c0fe 100644
--- a/mlir/unittests/Conversion/PDLToPDLInterp/RootOrderingTest.cpp
+++ b/mlir/unittests/Conversion/PDLToPDLInterp/RootOrderingTest.cpp
@@ -41,7 +41,7 @@ protected:
     builder.setInsertionPointToStart(&block);
     for (int i = 0; i < 4; ++i)
       // Ops will be deleted when `block` is destroyed.
-      v[i] = builder.create<ConstantIntOp>(builder.getUnknownLoc(), i, 32);
+      v[i] = ConstantIntOp::create(builder, builder.getUnknownLoc(), i, 32);
   }
 
   /// Checks that optimal branching on graph has the given cost and
diff --git a/mlir/unittests/Dialect/OpenACC/OpenACCOpsTest.cpp b/mlir/unittests/Dialect/OpenACC/OpenACCOpsTest.cpp
index 836efdb..6ac9a87 100644
--- a/mlir/unittests/Dialect/OpenACC/OpenACCOpsTest.cpp
+++ b/mlir/unittests/Dialect/OpenACC/OpenACCOpsTest.cpp
@@ -45,7 +45,7 @@ protected:
 template <typename Op>
 void testAsyncOnly(OpBuilder &b, MLIRContext &context, Location loc,
                    llvm::SmallVector<DeviceType> &dtypes) {
-  OwningOpRef<Op> op = b.create<Op>(loc, TypeRange{}, ValueRange{});
+  OwningOpRef<Op> op = Op::create(b, loc, TypeRange{}, ValueRange{});
   EXPECT_FALSE(op->hasAsyncOnly());
   for (auto d : dtypes)
     EXPECT_FALSE(op->hasAsyncOnly(d));
@@ -82,12 +82,12 @@ void testAsyncOnlyDataEntry(OpBuilder &b, MLIRContext &context, Location loc,
                             llvm::SmallVector<DeviceType> &dtypes) {
   auto memrefTy = MemRefType::get({}, b.getI32Type());
   OwningOpRef<memref::AllocaOp> varPtrOp =
-      b.create<memref::AllocaOp>(loc, memrefTy);
+      memref::AllocaOp::create(b, loc, memrefTy);
 
   TypedValue<PointerLikeType> varPtr =
       cast<TypedValue<PointerLikeType>>(varPtrOp->getResult());
-  OwningOpRef<Op> op = b.create<Op>(loc, varPtr,
-                                    /*structured=*/true, /*implicit=*/true);
+  OwningOpRef<Op> op = Op::create(b, loc, varPtr,
+                                  /*structured=*/true, /*implicit=*/true);
 
   EXPECT_FALSE(op->hasAsyncOnly());
   for (auto d : dtypes)
@@ -128,7 +128,7 @@ TEST_F(OpenACCOpsTest, asyncOnlyTestDataEntry) {
 template <typename Op>
 void testAsyncValue(OpBuilder &b, MLIRContext &context, Location loc,
                     llvm::SmallVector<DeviceType> &dtypes) {
-  OwningOpRef<Op> op = b.create<Op>(loc, TypeRange{}, ValueRange{});
+  OwningOpRef<Op> op = Op::create(b, loc, TypeRange{}, ValueRange{});
 
   mlir::Value empty;
   EXPECT_EQ(op->getAsyncValue(), empty);
@@ -136,7 +136,7 @@ void testAsyncValue(OpBuilder &b, MLIRContext &context, Location loc,
     EXPECT_EQ(op->getAsyncValue(d), empty);
 
   OwningOpRef<arith::ConstantIndexOp> val =
-      b.create<arith::ConstantIndexOp>(loc, 1);
+      arith::ConstantIndexOp::create(b, loc, 1);
   auto dtypeNvidia = DeviceTypeAttr::get(&context, DeviceType::Nvidia);
   op->setAsyncOperandsDeviceTypeAttr(b.getArrayAttr({dtypeNvidia}));
   op->getAsyncOperandsMutable().assign(val->getResult());
@@ -158,12 +158,12 @@ void testAsyncValueDataEntry(OpBuilder &b, MLIRContext &context, Location loc,
                              llvm::SmallVector<DeviceType> &dtypes) {
   auto memrefTy = MemRefType::get({}, b.getI32Type());
   OwningOpRef<memref::AllocaOp> varPtrOp =
-      b.create<memref::AllocaOp>(loc, memrefTy);
+      memref::AllocaOp::create(b, loc, memrefTy);
 
   TypedValue<PointerLikeType> varPtr =
       cast<TypedValue<PointerLikeType>>(varPtrOp->getResult());
-  OwningOpRef<Op> op = b.create<Op>(loc, varPtr,
-                                    /*structured=*/true, /*implicit=*/true);
+  OwningOpRef<Op> op = Op::create(b, loc, varPtr,
+                                  /*structured=*/true, /*implicit=*/true);
 
   mlir::Value empty;
   EXPECT_EQ(op->getAsyncValue(), empty);
@@ -171,7 +171,7 @@ void testAsyncValueDataEntry(OpBuilder &b, MLIRContext &context, Location loc,
     EXPECT_EQ(op->getAsyncValue(d), empty);
 
   OwningOpRef<arith::ConstantIndexOp> val =
-      b.create<arith::ConstantIndexOp>(loc, 1);
+      arith::ConstantIndexOp::create(b, loc, 1);
   auto dtypeNvidia = DeviceTypeAttr::get(&context, DeviceType::Nvidia);
   op->setAsyncOperandsDeviceTypeAttr(b.getArrayAttr({dtypeNvidia}));
   op->getAsyncOperandsMutable().assign(val->getResult());
@@ -197,13 +197,13 @@ template <typename Op>
 void testNumGangsValues(OpBuilder &b, MLIRContext &context, Location loc,
                         llvm::SmallVector<DeviceType> &dtypes,
                         llvm::SmallVector<DeviceType> &dtypesWithoutNone) {
-  OwningOpRef<Op> op = b.create<Op>(loc, TypeRange{}, ValueRange{});
+  OwningOpRef<Op> op = Op::create(b, loc, TypeRange{}, ValueRange{});
   EXPECT_EQ(op->getNumGangsValues().begin(), op->getNumGangsValues().end());
 
   OwningOpRef<arith::ConstantIndexOp> val1 =
-      b.create<arith::ConstantIndexOp>(loc, 1);
+      arith::ConstantIndexOp::create(b, loc, 1);
   OwningOpRef<arith::ConstantIndexOp> val2 =
-      b.create<arith::ConstantIndexOp>(loc, 4);
+      arith::ConstantIndexOp::create(b, loc, 4);
   auto dtypeNone = DeviceTypeAttr::get(&context, DeviceType::None);
   op->getNumGangsMutable().assign(val1->getResult());
   op->setNumGangsDeviceTypeAttr(b.getArrayAttr({dtypeNone}));
@@ -264,7 +264,7 @@ TEST_F(OpenACCOpsTest, numGangsValuesTest) {
 template <typename Op>
 void testVectorLength(OpBuilder &b, MLIRContext &context, Location loc,
                       llvm::SmallVector<DeviceType> &dtypes) {
-  OwningOpRef<Op> op = b.create<Op>(loc, TypeRange{}, ValueRange{});
+  OwningOpRef<Op> op = Op::create(b, loc, TypeRange{}, ValueRange{});
 
   mlir::Value empty;
   EXPECT_EQ(op->getVectorLengthValue(), empty);
@@ -272,7 +272,7 @@ void testVectorLength(OpBuilder &b, MLIRContext &context, Location loc,
     EXPECT_EQ(op->getVectorLengthValue(d), empty);
 
   OwningOpRef<arith::ConstantIndexOp> val =
-      b.create<arith::ConstantIndexOp>(loc, 1);
+      arith::ConstantIndexOp::create(b, loc, 1);
   auto dtypeNvidia = DeviceTypeAttr::get(&context, DeviceType::Nvidia);
   op->setVectorLengthDeviceTypeAttr(b.getArrayAttr({dtypeNvidia}));
   op->getVectorLengthMutable().assign(val->getResult());
@@ -292,7 +292,7 @@ template <typename Op>
 void testWaitOnly(OpBuilder &b, MLIRContext &context, Location loc,
                   llvm::SmallVector<DeviceType> &dtypes,
                   llvm::SmallVector<DeviceType> &dtypesWithoutNone) {
-  OwningOpRef<Op> op = b.create<Op>(loc, TypeRange{}, ValueRange{});
+  OwningOpRef<Op> op = Op::create(b, loc, TypeRange{}, ValueRange{});
   EXPECT_FALSE(op->hasWaitOnly());
   for (auto d : dtypes)
     EXPECT_FALSE(op->hasWaitOnly(d));
@@ -332,15 +332,15 @@ template <typename Op>
 void testWaitValues(OpBuilder &b, MLIRContext &context, Location loc,
                     llvm::SmallVector<DeviceType> &dtypes,
                     llvm::SmallVector<DeviceType> &dtypesWithoutNone) {
-  OwningOpRef<Op> op = b.create<Op>(loc, TypeRange{}, ValueRange{});
+  OwningOpRef<Op> op = Op::create(b, loc, TypeRange{}, ValueRange{});
   EXPECT_EQ(op->getWaitValues().begin(), op->getWaitValues().end());
 
   OwningOpRef<arith::ConstantIndexOp> val1 =
-      b.create<arith::ConstantIndexOp>(loc, 1);
+      arith::ConstantIndexOp::create(b, loc, 1);
   OwningOpRef<arith::ConstantIndexOp> val2 =
-      b.create<arith::ConstantIndexOp>(loc, 4);
+      arith::ConstantIndexOp::create(b, loc, 4);
   OwningOpRef<arith::ConstantIndexOp> val3 =
-      b.create<arith::ConstantIndexOp>(loc, 5);
+      arith::ConstantIndexOp::create(b, loc, 5);
   auto dtypeNone = DeviceTypeAttr::get(&context, DeviceType::None);
   op->getWaitOperandsMutable().assign(val1->getResult());
   op->setWaitOperandsDeviceTypeAttr(b.getArrayAttr({dtypeNone}));
@@ -426,7 +426,7 @@ TEST_F(OpenACCOpsTest, waitValuesTest) {
 }
 
 TEST_F(OpenACCOpsTest, loopOpGangVectorWorkerTest) {
-  OwningOpRef<LoopOp> op = b.create<LoopOp>(loc, TypeRange{}, ValueRange{});
+  OwningOpRef<LoopOp> op = LoopOp::create(b, loc, TypeRange{}, ValueRange{});
   EXPECT_FALSE(op->hasGang());
   EXPECT_FALSE(op->hasVector());
   EXPECT_FALSE(op->hasWorker());
@@ -473,7 +473,7 @@ TEST_F(OpenACCOpsTest, loopOpGangVectorWorkerTest) {
 
 TEST_F(OpenACCOpsTest, routineOpTest) {
   OwningOpRef<RoutineOp> op =
-      b.create<RoutineOp>(loc, TypeRange{}, ValueRange{});
+      RoutineOp::create(b, loc, TypeRange{}, ValueRange{});
 
   EXPECT_FALSE(op->hasSeq());
   EXPECT_FALSE(op->hasVector());
@@ -564,12 +564,12 @@ void testShortDataEntryOpBuilders(OpBuilder &b, MLIRContext &context,
                                   Location loc, DataClause dataClause) {
   auto memrefTy = MemRefType::get({}, b.getI32Type());
   OwningOpRef<memref::AllocaOp> varPtrOp =
-      b.create<memref::AllocaOp>(loc, memrefTy);
+      memref::AllocaOp::create(b, loc, memrefTy);
 
   TypedValue<PointerLikeType> varPtr =
       cast<TypedValue<PointerLikeType>>(varPtrOp->getResult());
-  OwningOpRef<Op> op = b.create<Op>(loc, varPtr,
-                                    /*structured=*/true, /*implicit=*/true);
+  OwningOpRef<Op> op = Op::create(b, loc, varPtr,
+                                  /*structured=*/true, /*implicit=*/true);
 
   EXPECT_EQ(op->getVarPtr(), varPtr);
   EXPECT_EQ(op->getType(), memrefTy);
@@ -579,24 +579,24 @@ void testShortDataEntryOpBuilders(OpBuilder &b, MLIRContext &context,
   EXPECT_TRUE(op->getBounds().empty());
   EXPECT_FALSE(op->getVarPtrPtr());
 
-  OwningOpRef<Op> op2 = b.create<Op>(loc, varPtr,
-                                     /*structured=*/false, /*implicit=*/false);
+  OwningOpRef<Op> op2 = Op::create(b, loc, varPtr,
+                                   /*structured=*/false, /*implicit=*/false);
   EXPECT_FALSE(op2->getImplicit());
   EXPECT_FALSE(op2->getStructured());
 
   OwningOpRef<arith::ConstantIndexOp> extent =
-      b.create<arith::ConstantIndexOp>(loc, 1);
+      arith::ConstantIndexOp::create(b, loc, 1);
   OwningOpRef<DataBoundsOp> bounds =
-      b.create<DataBoundsOp>(loc, extent->getResult());
+      DataBoundsOp::create(b, loc, extent->getResult());
   OwningOpRef<Op> opWithBounds =
-      b.create<Op>(loc, varPtr,
-                   /*structured=*/true, /*implicit=*/true, bounds->getResult());
+      Op::create(b, loc, varPtr,
+                 /*structured=*/true, /*implicit=*/true, bounds->getResult());
   EXPECT_FALSE(opWithBounds->getBounds().empty());
   EXPECT_EQ(opWithBounds->getBounds().back(), bounds->getResult());
 
   OwningOpRef<Op> opWithName =
-      b.create<Op>(loc, varPtr,
-                   /*structured=*/true, /*implicit=*/true, "varName");
+      Op::create(b, loc, varPtr,
+                 /*structured=*/true, /*implicit=*/true, "varName");
   EXPECT_EQ(opWithName->getNameAttr().str(), "varName");
 }
 
@@ -637,17 +637,17 @@ void testShortDataExitOpBuilders(OpBuilder &b, MLIRContext &context,
                                  Location loc, DataClause dataClause) {
   auto memrefTy = MemRefType::get({}, b.getI32Type());
   OwningOpRef<memref::AllocaOp> varPtrOp =
-      b.create<memref::AllocaOp>(loc, memrefTy);
+      memref::AllocaOp::create(b, loc, memrefTy);
   TypedValue<PointerLikeType> varPtr =
       cast<TypedValue<PointerLikeType>>(varPtrOp->getResult());
 
-  OwningOpRef<GetDevicePtrOp> accPtrOp = b.create<GetDevicePtrOp>(
-      loc, varPtr, /*structured=*/true, /*implicit=*/true);
+  OwningOpRef<GetDevicePtrOp> accPtrOp = GetDevicePtrOp::create(
+      b, loc, varPtr, /*structured=*/true, /*implicit=*/true);
   TypedValue<PointerLikeType> accPtr =
       cast<TypedValue<PointerLikeType>>(accPtrOp->getResult());
 
-  OwningOpRef<Op> op = b.create<Op>(loc, accPtr, varPtr,
-                                    /*structured=*/true, /*implicit=*/true);
+  OwningOpRef<Op> op = Op::create(b, loc, accPtr, varPtr,
+                                  /*structured=*/true, /*implicit=*/true);
 
   EXPECT_EQ(op->getVarPtr(), varPtr);
   EXPECT_EQ(op->getAccPtr(), accPtr);
@@ -656,24 +656,24 @@ void testShortDataExitOpBuilders(OpBuilder &b, MLIRContext &context,
   EXPECT_TRUE(op->getStructured());
   EXPECT_TRUE(op->getBounds().empty());
 
-  OwningOpRef<Op> op2 = b.create<Op>(loc, accPtr, varPtr,
-                                     /*structured=*/false, /*implicit=*/false);
+  OwningOpRef<Op> op2 = Op::create(b, loc, accPtr, varPtr,
+                                   /*structured=*/false, /*implicit=*/false);
   EXPECT_FALSE(op2->getImplicit());
   EXPECT_FALSE(op2->getStructured());
 
   OwningOpRef<arith::ConstantIndexOp> extent =
-      b.create<arith::ConstantIndexOp>(loc, 1);
+      arith::ConstantIndexOp::create(b, loc, 1);
   OwningOpRef<DataBoundsOp> bounds =
-      b.create<DataBoundsOp>(loc, extent->getResult());
+      DataBoundsOp::create(b, loc, extent->getResult());
   OwningOpRef<Op> opWithBounds =
-      b.create<Op>(loc, accPtr, varPtr,
-                   /*structured=*/true, /*implicit=*/true, bounds->getResult());
+      Op::create(b, loc, accPtr, varPtr,
+                 /*structured=*/true, /*implicit=*/true, bounds->getResult());
   EXPECT_FALSE(opWithBounds->getBounds().empty());
   EXPECT_EQ(opWithBounds->getBounds().back(), bounds->getResult());
 
   OwningOpRef<Op> opWithName =
-      b.create<Op>(loc, accPtr, varPtr,
-                   /*structured=*/true, /*implicit=*/true, "varName");
+      Op::create(b, loc, accPtr, varPtr,
+                 /*structured=*/true, /*implicit=*/true, "varName");
   EXPECT_EQ(opWithName->getNameAttr().str(), "varName");
 }
 
@@ -689,17 +689,17 @@ void testShortDataExitNoVarPtrOpBuilders(OpBuilder &b, MLIRContext &context,
                                          Location loc, DataClause dataClause) {
   auto memrefTy = MemRefType::get({}, b.getI32Type());
   OwningOpRef<memref::AllocaOp> varPtrOp =
-      b.create<memref::AllocaOp>(loc, memrefTy);
+      memref::AllocaOp::create(b, loc, memrefTy);
   TypedValue<PointerLikeType> varPtr =
       cast<TypedValue<PointerLikeType>>(varPtrOp->getResult());
 
-  OwningOpRef<GetDevicePtrOp> accPtrOp = b.create<GetDevicePtrOp>(
-      loc, varPtr, /*structured=*/true, /*implicit=*/true);
+  OwningOpRef<GetDevicePtrOp> accPtrOp = GetDevicePtrOp::create(
+      b, loc, varPtr, /*structured=*/true, /*implicit=*/true);
   TypedValue<PointerLikeType> accPtr =
       cast<TypedValue<PointerLikeType>>(accPtrOp->getResult());
 
-  OwningOpRef<Op> op = b.create<Op>(loc, accPtr,
-                                    /*structured=*/true, /*implicit=*/true);
+  OwningOpRef<Op> op = Op::create(b, loc, accPtr,
+                                  /*structured=*/true, /*implicit=*/true);
 
   EXPECT_EQ(op->getAccPtr(), accPtr);
   EXPECT_EQ(op->getDataClause(), dataClause);
@@ -707,24 +707,24 @@ void testShortDataExitNoVarPtrOpBuilders(OpBuilder &b, MLIRContext &context,
   EXPECT_TRUE(op->getStructured());
   EXPECT_TRUE(op->getBounds().empty());
 
-  OwningOpRef<Op> op2 = b.create<Op>(loc, accPtr,
-                                     /*structured=*/false, /*implicit=*/false);
+  OwningOpRef<Op> op2 = Op::create(b, loc, accPtr,
+                                   /*structured=*/false, /*implicit=*/false);
   EXPECT_FALSE(op2->getImplicit());
   EXPECT_FALSE(op2->getStructured());
 
   OwningOpRef<arith::ConstantIndexOp> extent =
-      b.create<arith::ConstantIndexOp>(loc, 1);
+      arith::ConstantIndexOp::create(b, loc, 1);
   OwningOpRef<DataBoundsOp> bounds =
-      b.create<DataBoundsOp>(loc, extent->getResult());
+      DataBoundsOp::create(b, loc, extent->getResult());
   OwningOpRef<Op> opWithBounds =
-      b.create<Op>(loc, accPtr,
-                   /*structured=*/true, /*implicit=*/true, bounds->getResult());
+      Op::create(b, loc, accPtr,
+                 /*structured=*/true, /*implicit=*/true, bounds->getResult());
   EXPECT_FALSE(opWithBounds->getBounds().empty());
   EXPECT_EQ(opWithBounds->getBounds().back(), bounds->getResult());
 
   OwningOpRef<Op> opWithName =
-      b.create<Op>(loc, accPtr,
-                   /*structured=*/true, /*implicit=*/true, "varName");
+      Op::create(b, loc, accPtr,
+                 /*structured=*/true, /*implicit=*/true, "varName");
   EXPECT_EQ(opWithName->getNameAttr().str(), "varName");
 }
 
@@ -742,16 +742,16 @@ void testShortDataEntryOpBuildersMappableVar(OpBuilder &b, MLIRContext &context,
   auto int64Ty = b.getI64Type();
   auto memrefTy = MemRefType::get({}, int64Ty);
   OwningOpRef<memref::AllocaOp> varPtrOp =
-      b.create<memref::AllocaOp>(loc, memrefTy);
+      memref::AllocaOp::create(b, loc, memrefTy);
   SmallVector<Value> indices;
   OwningOpRef<memref::LoadOp> loadVarOp =
-      b.create<memref::LoadOp>(loc, int64Ty, varPtrOp->getResult(), indices);
+      memref::LoadOp::create(b, loc, int64Ty, varPtrOp->getResult(), indices);
 
   EXPECT_TRUE(isMappableType(loadVarOp->getResult().getType()));
   TypedValue<MappableType> var =
       cast<TypedValue<MappableType>>(loadVarOp->getResult());
-  OwningOpRef<Op> op = b.create<Op>(loc, var,
-                                    /*structured=*/true, /*implicit=*/true);
+  OwningOpRef<Op> op = Op::create(b, loc, var,
+                                  /*structured=*/true, /*implicit=*/true);
 
   EXPECT_EQ(op->getVar(), var);
   EXPECT_EQ(op->getVarPtr(), nullptr);
diff --git a/mlir/unittests/Dialect/SCF/LoopLikeSCFOpsTest.cpp b/mlir/unittests/Dialect/SCF/LoopLikeSCFOpsTest.cpp
index fecd960..ef23123 100644
--- a/mlir/unittests/Dialect/SCF/LoopLikeSCFOpsTest.cpp
+++ b/mlir/unittests/Dialect/SCF/LoopLikeSCFOpsTest.cpp
@@ -119,45 +119,45 @@ protected:
 
 TEST_F(SCFLoopLikeTest, queryUnidimensionalLooplikes) {
   OwningOpRef<arith::ConstantIndexOp> lb =
-      b.create<arith::ConstantIndexOp>(loc, 0);
+      arith::ConstantIndexOp::create(b, loc, 0);
   OwningOpRef<arith::ConstantIndexOp> ub =
-      b.create<arith::ConstantIndexOp>(loc, 10);
+      arith::ConstantIndexOp::create(b, loc, 10);
   OwningOpRef<arith::ConstantIndexOp> step =
-      b.create<arith::ConstantIndexOp>(loc, 2);
+      arith::ConstantIndexOp::create(b, loc, 2);
 
   OwningOpRef<scf::ForOp> forOp =
-      b.create<scf::ForOp>(loc, lb.get(), ub.get(), step.get());
+      scf::ForOp::create(b, loc, lb.get(), ub.get(), step.get());
   checkUnidimensional(forOp.get());
 
-  OwningOpRef<scf::ForallOp> forallOp = b.create<scf::ForallOp>(
-      loc, ArrayRef<OpFoldResult>(lb->getResult()),
+  OwningOpRef<scf::ForallOp> forallOp = scf::ForallOp::create(
+      b, loc, ArrayRef<OpFoldResult>(lb->getResult()),
       ArrayRef<OpFoldResult>(ub->getResult()),
       ArrayRef<OpFoldResult>(step->getResult()), ValueRange(), std::nullopt);
   checkUnidimensional(forallOp.get());
 
-  OwningOpRef<scf::ParallelOp> parallelOp = b.create<scf::ParallelOp>(
-      loc, ValueRange(lb->getResult()), ValueRange(ub->getResult()),
+  OwningOpRef<scf::ParallelOp> parallelOp = scf::ParallelOp::create(
+      b, loc, ValueRange(lb->getResult()), ValueRange(ub->getResult()),
       ValueRange(step->getResult()), ValueRange());
   checkUnidimensional(parallelOp.get());
 }
 
 TEST_F(SCFLoopLikeTest, queryMultidimensionalLooplikes) {
   OwningOpRef<arith::ConstantIndexOp> lb =
-      b.create<arith::ConstantIndexOp>(loc, 0);
+      arith::ConstantIndexOp::create(b, loc, 0);
   OwningOpRef<arith::ConstantIndexOp> ub =
-      b.create<arith::ConstantIndexOp>(loc, 10);
+      arith::ConstantIndexOp::create(b, loc, 10);
   OwningOpRef<arith::ConstantIndexOp> step =
-      b.create<arith::ConstantIndexOp>(loc, 2);
+      arith::ConstantIndexOp::create(b, loc, 2);
 
-  OwningOpRef<scf::ForallOp> forallOp = b.create<scf::ForallOp>(
-      loc, ArrayRef<OpFoldResult>({lb->getResult(), lb->getResult()}),
+  OwningOpRef<scf::ForallOp> forallOp = scf::ForallOp::create(
+      b, loc, ArrayRef<OpFoldResult>({lb->getResult(), lb->getResult()}),
       ArrayRef<OpFoldResult>({ub->getResult(), ub->getResult()}),
       ArrayRef<OpFoldResult>({step->getResult(), step->getResult()}),
       ValueRange(), std::nullopt);
   checkMultidimensional(forallOp.get());
 
-  OwningOpRef<scf::ParallelOp> parallelOp = b.create<scf::ParallelOp>(
-      loc, ValueRange({lb->getResult(), lb->getResult()}),
+  OwningOpRef<scf::ParallelOp> parallelOp = scf::ParallelOp::create(
+      b, loc, ValueRange({lb->getResult(), lb->getResult()}),
       ValueRange({ub->getResult(), ub->getResult()}),
       ValueRange({step->getResult(), step->getResult()}), ValueRange());
   checkMultidimensional(parallelOp.get());
@@ -165,22 +165,22 @@ TEST_F(SCFLoopLikeTest, queryMultidimensionalLooplikes) {
 
 TEST_F(SCFLoopLikeTest, testForallNormalize) {
   OwningOpRef<arith::ConstantIndexOp> lb =
-      b.create<arith::ConstantIndexOp>(loc, 1);
+      arith::ConstantIndexOp::create(b, loc, 1);
   OwningOpRef<arith::ConstantIndexOp> ub =
-      b.create<arith::ConstantIndexOp>(loc, 10);
+      arith::ConstantIndexOp::create(b, loc, 10);
   OwningOpRef<arith::ConstantIndexOp> step =
-      b.create<arith::ConstantIndexOp>(loc, 3);
+      arith::ConstantIndexOp::create(b, loc, 3);
 
-  scf::ForallOp forallOp = b.create<scf::ForallOp>(
-      loc, ArrayRef<OpFoldResult>({lb->getResult(), lb->getResult()}),
+  scf::ForallOp forallOp = scf::ForallOp::create(
+      b, loc, ArrayRef<OpFoldResult>({lb->getResult(), lb->getResult()}),
       ArrayRef<OpFoldResult>({ub->getResult(), ub->getResult()}),
       ArrayRef<OpFoldResult>({step->getResult(), step->getResult()}),
       ValueRange(), std::nullopt);
   // Create a user of the induction variable. Bitcast is chosen for simplicity
   // since it is unary.
   b.setInsertionPointToStart(forallOp.getBody());
-  b.create<arith::BitcastOp>(UnknownLoc::get(&context), b.getF64Type(),
-                             forallOp.getInductionVar(0));
+  arith::BitcastOp::create(b, UnknownLoc::get(&context), b.getF64Type(),
+                           forallOp.getInductionVar(0));
   IRRewriter rewriter(b);
   FailureOr<scf::ForallOp> maybeNormalizedForallOp =
       normalizeForallOp(rewriter, forallOp);
diff --git a/mlir/unittests/Dialect/SMT/QuantifierTest.cpp b/mlir/unittests/Dialect/SMT/QuantifierTest.cpp
index d7c57f0..5cbc019 100644
--- a/mlir/unittests/Dialect/SMT/QuantifierTest.cpp
+++ b/mlir/unittests/Dialect/SMT/QuantifierTest.cpp
@@ -26,10 +26,10 @@ TEST(QuantifierTest, ExistsBuilderWithPattern) {
   OpBuilder builder(&context);
   auto boolTy = BoolType::get(&context);
 
-  OwningOpRef<ExistsOp> existsOp = builder.create<ExistsOp>(
-      loc, TypeRange{boolTy, boolTy},
+  OwningOpRef<ExistsOp> existsOp = ExistsOp::create(
+      builder, loc, TypeRange{boolTy, boolTy},
       [](OpBuilder &builder, Location loc, ValueRange boundVars) {
-        return builder.create<AndOp>(loc, boundVars);
+        return AndOp::create(builder, loc, boundVars);
       },
       std::nullopt,
       [](OpBuilder &builder, Location loc, ValueRange boundVars) {
@@ -57,10 +57,10 @@ TEST(QuantifierTest, ExistsBuilderNoPattern) {
   OpBuilder builder(&context);
   auto boolTy = BoolType::get(&context);
 
-  OwningOpRef<ExistsOp> existsOp = builder.create<ExistsOp>(
-      loc, TypeRange{boolTy, boolTy},
+  OwningOpRef<ExistsOp> existsOp = ExistsOp::create(
+      builder, loc, TypeRange{boolTy, boolTy},
       [](OpBuilder &builder, Location loc, ValueRange boundVars) {
-        return builder.create<AndOp>(loc, boundVars);
+        return AndOp::create(builder, loc, boundVars);
       },
       ArrayRef<StringRef>{"a", "b"}, nullptr, /*weight=*/0, /*noPattern=*/true);
 
@@ -82,10 +82,10 @@ TEST(QuantifierTest, ExistsBuilderDefault) {
   OpBuilder builder(&context);
   auto boolTy = BoolType::get(&context);
 
-  OwningOpRef<ExistsOp> existsOp = builder.create<ExistsOp>(
-      loc, TypeRange{boolTy, boolTy},
+  OwningOpRef<ExistsOp> existsOp = ExistsOp::create(
+      builder, loc, TypeRange{boolTy, boolTy},
       [](OpBuilder &builder, Location loc, ValueRange boundVars) {
-        return builder.create<AndOp>(loc, boundVars);
+        return AndOp::create(builder, loc, boundVars);
       },
       ArrayRef<StringRef>{"a", "b"});
 
@@ -111,10 +111,10 @@ TEST(QuantifierTest, ForallBuilderWithPattern) {
   OpBuilder builder(&context);
   auto boolTy = BoolType::get(&context);
 
-  OwningOpRef<ForallOp> forallOp = builder.create<ForallOp>(
-      loc, TypeRange{boolTy, boolTy},
+  OwningOpRef<ForallOp> forallOp = ForallOp::create(
+      builder, loc, TypeRange{boolTy, boolTy},
       [](OpBuilder &builder, Location loc, ValueRange boundVars) {
-        return builder.create<AndOp>(loc, boundVars);
+        return AndOp::create(builder, loc, boundVars);
       },
       ArrayRef<StringRef>{"a", "b"},
       [](OpBuilder &builder, Location loc, ValueRange boundVars) {
@@ -142,10 +142,10 @@ TEST(QuantifierTest, ForallBuilderNoPattern) {
   OpBuilder builder(&context);
   auto boolTy = BoolType::get(&context);
 
-  OwningOpRef<ForallOp> forallOp = builder.create<ForallOp>(
-      loc, TypeRange{boolTy, boolTy},
+  OwningOpRef<ForallOp> forallOp = ForallOp::create(
+      builder, loc, TypeRange{boolTy, boolTy},
       [](OpBuilder &builder, Location loc, ValueRange boundVars) {
-        return builder.create<AndOp>(loc, boundVars);
+        return AndOp::create(builder, loc, boundVars);
       },
       ArrayRef<StringRef>{"a", "b"}, nullptr, /*weight=*/0, /*noPattern=*/true);
 
@@ -167,10 +167,10 @@ TEST(QuantifierTest, ForallBuilderDefault) {
   OpBuilder builder(&context);
   auto boolTy = BoolType::get(&context);
 
-  OwningOpRef<ForallOp> forallOp = builder.create<ForallOp>(
-      loc, TypeRange{boolTy, boolTy},
+  OwningOpRef<ForallOp> forallOp = ForallOp::create(
+      builder, loc, TypeRange{boolTy, boolTy},
       [](OpBuilder &builder, Location loc, ValueRange boundVars) {
-        return builder.create<AndOp>(loc, boundVars);
+        return AndOp::create(builder, loc, boundVars);
       },
       std::nullopt);
 
diff --git a/mlir/unittests/Dialect/SPIRV/SerializationTest.cpp b/mlir/unittests/Dialect/SPIRV/SerializationTest.cpp
index ef89c16..af55296 100644
--- a/mlir/unittests/Dialect/SPIRV/SerializationTest.cpp
+++ b/mlir/unittests/Dialect/SPIRV/SerializationTest.cpp
@@ -71,8 +71,8 @@ protected:
   spirv::GlobalVariableOp addGlobalVar(Type type, llvm::StringRef name) {
     OpBuilder builder(module->getRegion());
     auto ptrType = spirv::PointerType::get(type, spirv::StorageClass::Uniform);
-    return builder.create<spirv::GlobalVariableOp>(
-        UnknownLoc::get(&context), TypeAttr::get(ptrType),
+    return spirv::GlobalVariableOp::create(
+        builder, UnknownLoc::get(&context), TypeAttr::get(ptrType),
         builder.getStringAttr(name), nullptr);
   }
 
@@ -82,14 +82,14 @@ protected:
     auto loc = UnknownLoc::get(&context);
 
     if (auto intType = dyn_cast<IntegerType>(type)) {
-      return builder.create<spirv::ConstantOp>(
-          loc, type, builder.getIntegerAttr(type, val));
+      return spirv::ConstantOp::create(builder, loc, type,
+                                       builder.getIntegerAttr(type, val));
     }
     if (auto vectorType = dyn_cast<VectorType>(type)) {
       Type elemType = vectorType.getElementType();
       if (auto intType = dyn_cast<IntegerType>(elemType)) {
-        return builder.create<spirv::ConstantOp>(
-            loc, type,
+        return spirv::ConstantOp::create(
+            builder, loc, type,
             DenseElementsAttr::get(vectorType,
                                    IntegerAttr::get(elemType, val).getValue()));
       }
diff --git a/mlir/unittests/IR/IRMapping.cpp b/mlir/unittests/IR/IRMapping.cpp
index b88009d..983c41a 100644
--- a/mlir/unittests/IR/IRMapping.cpp
+++ b/mlir/unittests/IR/IRMapping.cpp
@@ -26,10 +26,10 @@ TEST(IRMapping, TypedValue) {
   Block block;
   builder.setInsertionPointToEnd(&block);
 
-  Value i64Val = builder.create<test::TestOpConstant>(
-      loc, builder.getI64Type(), builder.getI64IntegerAttr(0));
-  Value f64Val = builder.create<test::TestOpConstant>(
-      loc, builder.getF64Type(), builder.getF64FloatAttr(0.0));
+  Value i64Val = test::TestOpConstant::create(
+      builder, loc, builder.getI64Type(), builder.getI64IntegerAttr(0));
+  Value f64Val = test::TestOpConstant::create(
+      builder, loc, builder.getF64Type(), builder.getF64FloatAttr(0.0));
 
   IRMapping mapping;
   mapping.map(i64Val, f64Val);
diff --git a/mlir/unittests/IR/InterfaceAttachmentTest.cpp b/mlir/unittests/IR/InterfaceAttachmentTest.cpp
index 1b5d3b8..e1e65da 100644
--- a/mlir/unittests/IR/InterfaceAttachmentTest.cpp
+++ b/mlir/unittests/IR/InterfaceAttachmentTest.cpp
@@ -303,7 +303,7 @@ TEST(InterfaceAttachment, Operation) {
 
   // Initially, the operation doesn't have the interface.
   OwningOpRef<ModuleOp> moduleOp =
-      builder.create<ModuleOp>(UnknownLoc::get(&context));
+      ModuleOp::create(builder, UnknownLoc::get(&context));
   ASSERT_FALSE(isa<TestExternalOpInterface>(moduleOp->getOperation()));
 
   // We can attach an external interface and now the operaiton has it.
@@ -317,8 +317,8 @@ TEST(InterfaceAttachment, Operation) {
 
   // Default implementation can be overridden.
   OwningOpRef<UnrealizedConversionCastOp> castOp =
-      builder.create<UnrealizedConversionCastOp>(UnknownLoc::get(&context),
-                                                 TypeRange(), ValueRange());
+      UnrealizedConversionCastOp::create(builder, UnknownLoc::get(&context),
+                                         TypeRange(), ValueRange());
   ASSERT_FALSE(isa<TestExternalOpInterface>(castOp->getOperation()));
   UnrealizedConversionCastOp::attachInterface<TestExternalOpOverridingModel>(
       context);
@@ -368,11 +368,11 @@ TEST(InterfaceAttachment, OperationDelayedContextConstruct) {
   OwningOpRef<ModuleOp> module = ModuleOp::create(UnknownLoc::get(&context));
   OpBuilder builder(module->getBody(), module->getBody()->begin());
   auto opJ =
-      builder.create<test::OpJ>(builder.getUnknownLoc(), builder.getI32Type());
+      test::OpJ::create(builder, builder.getUnknownLoc(), builder.getI32Type());
   auto opH =
-      builder.create<test::OpH>(builder.getUnknownLoc(), opJ.getResult());
+      test::OpH::create(builder, builder.getUnknownLoc(), opJ.getResult());
   auto opI =
-      builder.create<test::OpI>(builder.getUnknownLoc(), opJ.getResult());
+      test::OpI::create(builder, builder.getUnknownLoc(), opJ.getResult());
 
   EXPECT_TRUE(isa<TestExternalOpInterface>(module->getOperation()));
   EXPECT_TRUE(isa<TestExternalOpInterface>(opJ.getOperation()));
@@ -399,11 +399,11 @@ TEST(InterfaceAttachment, OperationDelayedContextAppend) {
   OwningOpRef<ModuleOp> module = ModuleOp::create(UnknownLoc::get(&context));
   OpBuilder builder(module->getBody(), module->getBody()->begin());
   auto opJ =
-      builder.create<test::OpJ>(builder.getUnknownLoc(), builder.getI32Type());
+      test::OpJ::create(builder, builder.getUnknownLoc(), builder.getI32Type());
   auto opH =
-      builder.create<test::OpH>(builder.getUnknownLoc(), opJ.getResult());
+      test::OpH::create(builder, builder.getUnknownLoc(), opJ.getResult());
   auto opI =
-      builder.create<test::OpI>(builder.getUnknownLoc(), opJ.getResult());
+      test::OpI::create(builder, builder.getUnknownLoc(), opJ.getResult());
 
   EXPECT_FALSE(isa<TestExternalOpInterface>(module->getOperation()));
   EXPECT_FALSE(isa<TestExternalOpInterface>(opJ.getOperation()));
diff --git a/mlir/unittests/IR/InterfaceTest.cpp b/mlir/unittests/IR/InterfaceTest.cpp
index 42196b0..235163c 100644
--- a/mlir/unittests/IR/InterfaceTest.cpp
+++ b/mlir/unittests/IR/InterfaceTest.cpp
@@ -27,12 +27,12 @@ TEST(InterfaceTest, OpInterfaceDenseMapKey) {
 
   OwningOpRef<ModuleOp> module = ModuleOp::create(UnknownLoc::get(&context));
   OpBuilder builder(module->getBody(), module->getBody()->begin());
-  auto op1 = builder.create<test::SideEffectOp>(builder.getUnknownLoc(),
-                                                builder.getI32Type());
-  auto op2 = builder.create<test::SideEffectOp>(builder.getUnknownLoc(),
-                                                builder.getI32Type());
-  auto op3 = builder.create<test::SideEffectOp>(builder.getUnknownLoc(),
-                                                builder.getI32Type());
+  auto op1 = test::SideEffectOp::create(builder, builder.getUnknownLoc(),
+                                        builder.getI32Type());
+  auto op2 = test::SideEffectOp::create(builder, builder.getUnknownLoc(),
+                                        builder.getI32Type());
+  auto op3 = test::SideEffectOp::create(builder, builder.getUnknownLoc(),
+                                        builder.getI32Type());
   DenseSet<MemoryEffectOpInterface> opSet;
   opSet.insert(op1);
   opSet.insert(op2);
@@ -64,8 +64,8 @@ TEST(InterfaceTest, TestCustomClassOf) {
   context.loadDialect<test::TestDialect>();
 
   OpBuilder builder(&context);
-  auto op = builder.create<TestOpOptionallyImplementingInterface>(
-      builder.getUnknownLoc(), /*implementsInterface=*/true);
+  auto op = TestOpOptionallyImplementingInterface::create(
+      builder, builder.getUnknownLoc(), /*implementsInterface=*/true);
   EXPECT_TRUE(isa<TestOptionallyImplementedOpInterface>(*op));
   op.setImplementsInterface(false);
   EXPECT_FALSE(isa<TestOptionallyImplementedOpInterface>(*op));
diff --git a/mlir/unittests/IR/OperationSupportTest.cpp b/mlir/unittests/IR/OperationSupportTest.cpp
index 7bc1a04..9f3e7ed 100644
--- a/mlir/unittests/IR/OperationSupportTest.cpp
+++ b/mlir/unittests/IR/OperationSupportTest.cpp
@@ -302,8 +302,8 @@ TEST(OperandStorageTest, PopulateDefaultAttrs) {
   auto req1 = b.getI32IntegerAttr(10);
   auto req2 = b.getI32IntegerAttr(60);
   // Verify default attributes populated post op creation.
-  Operation *op = b.create<test::OpAttrMatch1>(b.getUnknownLoc(), req1, nullptr,
-                                               nullptr, req2);
+  Operation *op = test::OpAttrMatch1::create(b, b.getUnknownLoc(), req1,
+                                             nullptr, nullptr, req2);
   auto opt = op->getInherentAttr("default_valued_attr");
   EXPECT_NE(opt, nullptr) << *op;
 
@@ -343,11 +343,11 @@ TEST(OperationEquivalenceTest, HashWorksWithFlags) {
 
   // Check ignore properties.
   auto req1 = b.getI32IntegerAttr(10);
-  Operation *opWithProperty1 = b.create<test::OpAttrMatch1>(
-      b.getUnknownLoc(), req1, nullptr, nullptr, req1);
+  Operation *opWithProperty1 = test::OpAttrMatch1::create(
+      b, b.getUnknownLoc(), req1, nullptr, nullptr, req1);
   auto req2 = b.getI32IntegerAttr(60);
-  Operation *opWithProperty2 = b.create<test::OpAttrMatch1>(
-      b.getUnknownLoc(), req2, nullptr, nullptr, req2);
+  Operation *opWithProperty2 = test::OpAttrMatch1::create(
+      b, b.getUnknownLoc(), req2, nullptr, nullptr, req2);
   EXPECT_EQ(getHash(opWithProperty1, OperationEquivalence::IgnoreProperties),
             getHash(opWithProperty2, OperationEquivalence::IgnoreProperties));
   EXPECT_NE(getHash(opWithProperty1, OperationEquivalence::None),
diff --git a/mlir/unittests/TableGen/OpBuildGen.cpp b/mlir/unittests/TableGen/OpBuildGen.cpp
index 74e62aa..0943033 100644
--- a/mlir/unittests/TableGen/OpBuildGen.cpp
+++ b/mlir/unittests/TableGen/OpBuildGen.cpp
@@ -36,12 +36,11 @@ protected:
   OpBuildGenTest()
       : ctx(getContext()), builder(&ctx), loc(builder.getUnknownLoc()),
         i32Ty(builder.getI32Type()), f32Ty(builder.getF32Type()),
-        cstI32(builder.create<test::TableGenConstant>(loc, i32Ty)),
-        cstF32(builder.create<test::TableGenConstant>(loc, f32Ty)),
-        noAttrs(), attrStorage{builder.getNamedAttr("attr0",
-                                                    builder.getBoolAttr(true)),
-                               builder.getNamedAttr(
-                                   "attr1", builder.getI32IntegerAttr(33))},
+        cstI32(test::TableGenConstant::create(builder, loc, i32Ty)),
+        cstF32(test::TableGenConstant::create(builder, loc, f32Ty)), noAttrs(),
+        attrStorage{
+            builder.getNamedAttr("attr0", builder.getBoolAttr(true)),
+            builder.getNamedAttr("attr1", builder.getI32IntegerAttr(33))},
         attrs(attrStorage) {}
 
   // Verify that `op` has the given set of result types, operands, and
@@ -123,21 +122,21 @@ protected:
 /// Test basic build methods.
 TEST_F(OpBuildGenTest, BasicBuildMethods) {
   // Test separate args, separate results build method.
-  auto op = builder.create<test::TableGenBuildOp0>(loc, i32Ty, *cstI32);
+  auto op = test::TableGenBuildOp0::create(builder, loc, i32Ty, *cstI32);
   verifyOp(op, {i32Ty}, {*cstI32}, noAttrs);
 
   // Test separate args, collective results build method.
-  op = builder.create<test::TableGenBuildOp0>(loc, TypeRange{i32Ty}, *cstI32);
+  op = test::TableGenBuildOp0::create(builder, loc, TypeRange{i32Ty}, *cstI32);
   verifyOp(op, {i32Ty}, {*cstI32}, noAttrs);
 
   // Test collective args, collective params build method.
-  op = builder.create<test::TableGenBuildOp0>(loc, TypeRange{i32Ty},
-                                              ValueRange{*cstI32});
+  op = test::TableGenBuildOp0::create(builder, loc, TypeRange{i32Ty},
+                                      ValueRange{*cstI32});
   verifyOp(op, {i32Ty}, {*cstI32}, noAttrs);
 
   // Test collective args, collective results, non-empty attributes
-  op = builder.create<test::TableGenBuildOp0>(loc, TypeRange{i32Ty},
-                                              ValueRange{*cstI32}, attrs);
+  op = test::TableGenBuildOp0::create(builder, loc, TypeRange{i32Ty},
+                                      ValueRange{*cstI32}, attrs);
   verifyOp(op, {i32Ty}, {*cstI32}, attrs);
 }
 
@@ -154,25 +153,25 @@ TEST_F(OpBuildGenTest, BasicBuildMethods) {
 /// variadic result.
 TEST_F(OpBuildGenTest, BuildMethodsSingleVariadicArgAndResult) {
   // Test collective args, collective results method, building a unary op.
-  auto op = builder.create<test::TableGenBuildOp1>(loc, TypeRange{i32Ty},
-                                                   ValueRange{*cstI32});
+  auto op = test::TableGenBuildOp1::create(builder, loc, TypeRange{i32Ty},
+                                           ValueRange{*cstI32});
   verifyOp(op, {i32Ty}, {*cstI32}, noAttrs);
 
   // Test collective args, collective results method, building a unary op with
   // named attributes.
-  op = builder.create<test::TableGenBuildOp1>(loc, TypeRange{i32Ty},
-                                              ValueRange{*cstI32}, attrs);
+  op = test::TableGenBuildOp1::create(builder, loc, TypeRange{i32Ty},
+                                      ValueRange{*cstI32}, attrs);
   verifyOp(op, {i32Ty}, {*cstI32}, attrs);
 
   // Test collective args, collective results method, building a binary op.
-  op = builder.create<test::TableGenBuildOp1>(loc, TypeRange{i32Ty, f32Ty},
-                                              ValueRange{*cstI32, *cstF32});
+  op = test::TableGenBuildOp1::create(builder, loc, TypeRange{i32Ty, f32Ty},
+                                      ValueRange{*cstI32, *cstF32});
   verifyOp(op, {i32Ty, f32Ty}, {*cstI32, *cstF32}, noAttrs);
 
   // Test collective args, collective results method, building a binary op with
   // named attributes.
-  op = builder.create<test::TableGenBuildOp1>(
-      loc, TypeRange{i32Ty, f32Ty}, ValueRange{*cstI32, *cstF32}, attrs);
+  op = test::TableGenBuildOp1::create(builder, loc, TypeRange{i32Ty, f32Ty},
+                                      ValueRange{*cstI32, *cstF32}, attrs);
   verifyOp(op, {i32Ty, f32Ty}, {*cstI32, *cstF32}, attrs);
 }
 
@@ -181,22 +180,22 @@ TEST_F(OpBuildGenTest, BuildMethodsSingleVariadicArgAndResult) {
 TEST_F(OpBuildGenTest, BuildMethodsSingleVariadicArgNonVariadicResults) {
   // Test separate arg, separate param build method.
   auto op =
-      builder.create<test::TableGenBuildOp1>(loc, i32Ty, ValueRange{*cstI32});
+      test::TableGenBuildOp1::create(builder, loc, i32Ty, ValueRange{*cstI32});
   verifyOp(op, {i32Ty}, {*cstI32}, noAttrs);
 
   // Test collective params build method, no attributes.
-  op = builder.create<test::TableGenBuildOp1>(loc, TypeRange{i32Ty},
-                                              ValueRange{*cstI32});
+  op = test::TableGenBuildOp1::create(builder, loc, TypeRange{i32Ty},
+                                      ValueRange{*cstI32});
   verifyOp(op, {i32Ty}, {*cstI32}, noAttrs);
 
   // Test collective params build method no attributes, 2 inputs.
-  op = builder.create<test::TableGenBuildOp1>(loc, TypeRange{i32Ty},
-                                              ValueRange{*cstI32, *cstF32});
+  op = test::TableGenBuildOp1::create(builder, loc, TypeRange{i32Ty},
+                                      ValueRange{*cstI32, *cstF32});
   verifyOp(op, {i32Ty}, {*cstI32, *cstF32}, noAttrs);
 
   // Test collective params build method, non-empty attributes.
-  op = builder.create<test::TableGenBuildOp1>(
-      loc, TypeRange{i32Ty}, ValueRange{*cstI32, *cstF32}, attrs);
+  op = test::TableGenBuildOp1::create(builder, loc, TypeRange{i32Ty},
+                                      ValueRange{*cstI32, *cstF32}, attrs);
   verifyOp(op, {i32Ty}, {*cstI32, *cstF32}, attrs);
 }
 
@@ -205,18 +204,18 @@ TEST_F(OpBuildGenTest, BuildMethodsSingleVariadicArgNonVariadicResults) {
 TEST_F(OpBuildGenTest,
        BuildMethodsSingleVariadicArgAndMultipleVariadicResults) {
   // Test separate arg, separate param build method.
-  auto op = builder.create<test::TableGenBuildOp3>(
-      loc, TypeRange{i32Ty}, TypeRange{f32Ty}, ValueRange{*cstI32});
+  auto op = test::TableGenBuildOp3::create(
+      builder, loc, TypeRange{i32Ty}, TypeRange{f32Ty}, ValueRange{*cstI32});
   verifyOp(op, {i32Ty, f32Ty}, {*cstI32}, noAttrs);
 
   // Test collective params build method, no attributes.
-  op = builder.create<test::TableGenBuildOp3>(loc, TypeRange{i32Ty, f32Ty},
-                                              ValueRange{*cstI32});
+  op = test::TableGenBuildOp3::create(builder, loc, TypeRange{i32Ty, f32Ty},
+                                      ValueRange{*cstI32});
   verifyOp(op, {i32Ty, f32Ty}, {*cstI32}, noAttrs);
 
   // Test collective params build method, with attributes.
-  op = builder.create<test::TableGenBuildOp3>(loc, TypeRange{i32Ty, f32Ty},
-                                              ValueRange{*cstI32}, attrs);
+  op = test::TableGenBuildOp3::create(builder, loc, TypeRange{i32Ty, f32Ty},
+                                      ValueRange{*cstI32}, attrs);
   verifyOp(op, {i32Ty, f32Ty}, {*cstI32}, attrs);
 }
 
@@ -227,29 +226,29 @@ TEST_F(OpBuildGenTest,
 // build methods with no result types as they are inferred from the input types.
 TEST_F(OpBuildGenTest, BuildMethodsSameOperandsAndResultTypeSuppression) {
   // Test separate arg, separate param build method.
-  auto op = builder.create<test::TableGenBuildOp4>(
-      loc, i32Ty, ValueRange{*cstI32, *cstI32});
+  auto op = test::TableGenBuildOp4::create(builder, loc, i32Ty,
+                                           ValueRange{*cstI32, *cstI32});
   verifyOp(std::move(op), {i32Ty}, {*cstI32, *cstI32}, noAttrs);
 
   // Test collective params build method.
-  op = builder.create<test::TableGenBuildOp4>(loc, TypeRange{i32Ty},
-                                              ValueRange{*cstI32, *cstI32});
+  op = test::TableGenBuildOp4::create(builder, loc, TypeRange{i32Ty},
+                                      ValueRange{*cstI32, *cstI32});
   verifyOp(std::move(op), {i32Ty}, {*cstI32, *cstI32}, noAttrs);
 
   // Test build method with no result types, default value of attributes.
-  op =
-      builder.create<test::TableGenBuildOp4>(loc, ValueRange{*cstI32, *cstI32});
+  op = test::TableGenBuildOp4::create(builder, loc,
+                                      ValueRange{*cstI32, *cstI32});
   verifyOp(std::move(op), {i32Ty}, {*cstI32, *cstI32}, noAttrs);
 
   // Test build method with no result types and supplied attributes.
-  op = builder.create<test::TableGenBuildOp4>(loc, ValueRange{*cstI32, *cstI32},
-                                              attrs);
+  op = test::TableGenBuildOp4::create(builder, loc,
+                                      ValueRange{*cstI32, *cstI32}, attrs);
   verifyOp(std::move(op), {i32Ty}, {*cstI32, *cstI32}, attrs);
 }
 
 TEST_F(OpBuildGenTest, BuildMethodsRegionsAndInferredType) {
-  auto op = builder.create<test::TableGenBuildOp5>(
-      loc, ValueRange{*cstI32, *cstF32}, /*attributes=*/noAttrs);
+  auto op = test::TableGenBuildOp5::create(
+      builder, loc, ValueRange{*cstI32, *cstF32}, /*attributes=*/noAttrs);
   ASSERT_EQ(op->getNumRegions(), 1u);
   verifyOp(op, {i32Ty}, {*cstI32, *cstF32}, noAttrs);
 }
@@ -266,28 +265,28 @@ TEST_F(OpBuildGenTest, BuildMethodsVariadicProperties) {
   ArrayRef<NamedAttribute> attrs(attrsStorage);
 
   // Test separate arg, separate param build method.
-  auto op = builder.create<test::TableGenBuildOp6>(
-      loc, f32Ty, ValueRange{*cstI32}, ValueRange{*cstI32});
+  auto op = test::TableGenBuildOp6::create(
+      builder, loc, f32Ty, ValueRange{*cstI32}, ValueRange{*cstI32});
   verifyOp(std::move(op), {f32Ty}, {*cstI32}, {*cstI32}, noAttrs);
 
   // Test build method with no result types, default value of attributes.
-  op = builder.create<test::TableGenBuildOp6>(loc, ValueRange{*cstI32},
-                                              ValueRange{*cstI32});
+  op = test::TableGenBuildOp6::create(builder, loc, ValueRange{*cstI32},
+                                      ValueRange{*cstI32});
   verifyOp(std::move(op), {f32Ty}, {*cstI32}, {*cstI32}, noAttrs);
 
   // Test collective params build method.
-  op = builder.create<test::TableGenBuildOp6>(
-      loc, TypeRange{f32Ty}, ValueRange{*cstI32}, ValueRange{*cstI32});
+  op = test::TableGenBuildOp6::create(builder, loc, TypeRange{f32Ty},
+                                      ValueRange{*cstI32}, ValueRange{*cstI32});
   verifyOp(std::move(op), {f32Ty}, {*cstI32}, {*cstI32}, noAttrs);
 
   // Test build method with result types, supplied attributes.
-  op = builder.create<test::TableGenBuildOp6>(
-      loc, TypeRange{f32Ty}, ValueRange{*cstI32, *cstI32}, attrs);
+  op = test::TableGenBuildOp6::create(builder, loc, TypeRange{f32Ty},
+                                      ValueRange{*cstI32, *cstI32}, attrs);
   verifyOp(std::move(op), {f32Ty}, {*cstI32}, {*cstI32}, attrs);
 
   // Test build method with no result types and supplied attributes.
-  op = builder.create<test::TableGenBuildOp6>(loc, ValueRange{*cstI32, *cstI32},
-                                              attrs);
+  op = test::TableGenBuildOp6::create(builder, loc,
+                                      ValueRange{*cstI32, *cstI32}, attrs);
   verifyOp(std::move(op), {f32Ty}, {*cstI32}, {*cstI32}, attrs);
 }
 
@@ -295,14 +294,14 @@ TEST_F(OpBuildGenTest, BuildMethodsInherentDiscardableAttrs) {
   test::TableGenBuildOp7::Properties props;
   props.attr0 = cast<BoolAttr>(attrs[0].getValue());
   ArrayRef<NamedAttribute> discardableAttrs = attrs.drop_front();
-  auto op7 = builder.create<test::TableGenBuildOp7>(
-      loc, TypeRange{}, ValueRange{}, props, discardableAttrs);
+  auto op7 = test::TableGenBuildOp7::create(
+      builder, loc, TypeRange{}, ValueRange{}, props, discardableAttrs);
   verifyOp(op7, {}, {}, attrs);
 
   // Check that the old-style builder where all the attributes go in the same
   // place works.
-  auto op7b = builder.create<test::TableGenBuildOp7>(loc, TypeRange{},
-                                                     ValueRange{}, attrs);
+  auto op7b = test::TableGenBuildOp7::create(builder, loc, TypeRange{},
+                                             ValueRange{}, attrs);
   // Note: this goes before verifyOp() because verifyOp() calls erase(), causing
   // use-after-free.
   ASSERT_EQ(op7b.getProperties().getAttr0(), attrs[0].getValue());
diff --git a/mlir/unittests/Target/LLVM/CMakeLists.txt b/mlir/unittests/Target/LLVM/CMakeLists.txt
index 15835b9..0daac11 100644
--- a/mlir/unittests/Target/LLVM/CMakeLists.txt
+++ b/mlir/unittests/Target/LLVM/CMakeLists.txt
@@ -1,4 +1,4 @@
-set(LLVM_LINK_COMPONENTS nativecodegen)
+set(LLVM_LINK_COMPONENTS nativecodegen BitReader)
 
 get_property(dialect_libs GLOBAL PROPERTY MLIR_DIALECT_LIBS)
 
diff --git a/offload/liboffload/src/Helpers.hpp b/offload/liboffload/src/Helpers.hpp
index 8b85945..62e55e5 100644
--- a/offload/liboffload/src/Helpers.hpp
+++ b/offload/liboffload/src/Helpers.hpp
@@ -75,23 +75,16 @@ public:
   InfoWriter(InfoWriter &) = delete;
   ~InfoWriter() = default;
 
-  template <typename T> llvm::Error write(llvm::Expected<T> &&Val) {
-    if (Val)
-      return getInfo(Size, Target, SizeRet, *Val);
-    return Val.takeError();
+  template <typename T> llvm::Error write(T Val) {
+    return getInfo(Size, Target, SizeRet, Val);
   }
 
-  template <typename T>
-  llvm::Error writeArray(llvm::Expected<T> &&Val, size_t Elems) {
-    if (Val)
-      return getInfoArray(Elems, Size, Target, SizeRet, *Val);
-    return Val.takeError();
+  template <typename T> llvm::Error writeArray(T Val, size_t Elems) {
+    return getInfoArray(Elems, Size, Target, SizeRet, Val);
   }
 
-  llvm::Error writeString(llvm::Expected<llvm::StringRef> &&Val) {
-    if (Val)
-      return getInfoString(Size, Target, SizeRet, *Val);
-    return Val.takeError();
+  llvm::Error writeString(llvm::StringRef Val) {
+    return getInfoString(Size, Target, SizeRet, Val);
   }
 
 private:
diff --git a/offload/liboffload/src/OffloadImpl.cpp b/offload/liboffload/src/OffloadImpl.cpp
index 8f316b8..2444ccd 100644
--- a/offload/liboffload/src/OffloadImpl.cpp
+++ b/offload/liboffload/src/OffloadImpl.cpp
@@ -299,78 +299,62 @@ Error olGetDeviceInfoImplDetail(ol_device_handle_t Device,
     return Plugin::error(ErrorCode::UNIMPLEMENTED, ErrBuffer.c_str());
   };
 
-  // Find the info if it exists under any of the given names
-  auto getInfoString =
-      [&](std::vector<std::string> Names) -> llvm::Expected<const char *> {
-    for (auto &Name : Names) {
-      if (auto Entry = Device->Info.get(Name)) {
-        if (!std::holds_alternative<std::string>((*Entry)->Value))
-          return makeError(ErrorCode::BACKEND_FAILURE,
-                           "plugin returned incorrect type");
-        return std::get<std::string>((*Entry)->Value).c_str();
-      }
-    }
-
-    return makeError(ErrorCode::UNIMPLEMENTED,
-                     "plugin did not provide a response for this information");
-  };
-
-  auto getInfoXyz =
-      [&](std::vector<std::string> Names) -> llvm::Expected<ol_dimensions_t> {
-    for (auto &Name : Names) {
-      if (auto Entry = Device->Info.get(Name)) {
-        auto Node = *Entry;
-        ol_dimensions_t Out{0, 0, 0};
-
-        auto getField = [&](StringRef Name, uint32_t &Dest) {
-          if (auto F = Node->get(Name)) {
-            if (!std::holds_alternative<size_t>((*F)->Value))
-              return makeError(
-                  ErrorCode::BACKEND_FAILURE,
-                  "plugin returned incorrect type for dimensions element");
-            Dest = std::get<size_t>((*F)->Value);
-          } else
-            return makeError(ErrorCode::BACKEND_FAILURE,
-                             "plugin didn't provide all values for dimensions");
-          return Plugin::success();
-        };
-
-        if (auto Res = getField("x", Out.x))
-          return Res;
-        if (auto Res = getField("y", Out.y))
-          return Res;
-        if (auto Res = getField("z", Out.z))
-          return Res;
-
-        return Out;
-      }
-    }
+  // These are not implemented by the plugin interface
+  if (PropName == OL_DEVICE_INFO_PLATFORM)
+    return Info.write<void *>(Device->Platform);
+  if (PropName == OL_DEVICE_INFO_TYPE)
+    return Info.write<ol_device_type_t>(OL_DEVICE_TYPE_GPU);
+  if (PropName >= OL_DEVICE_INFO_LAST)
+    return createOffloadError(ErrorCode::INVALID_ENUMERATION,
+                              "getDeviceInfo enum '%i' is invalid", PropName);
 
+  auto EntryOpt = Device->Info.get(static_cast<DeviceInfo>(PropName));
+  if (!EntryOpt)
     return makeError(ErrorCode::UNIMPLEMENTED,
                      "plugin did not provide a response for this information");
-  };
+  auto Entry = *EntryOpt;
 
   switch (PropName) {
-  case OL_DEVICE_INFO_PLATFORM:
-    return Info.write<void *>(Device->Platform);
-  case OL_DEVICE_INFO_TYPE:
-    return Info.write<ol_device_type_t>(OL_DEVICE_TYPE_GPU);
   case OL_DEVICE_INFO_NAME:
-    return Info.writeString(getInfoString({"Device Name"}));
   case OL_DEVICE_INFO_VENDOR:
-    return Info.writeString(getInfoString({"Vendor Name"}));
-  case OL_DEVICE_INFO_DRIVER_VERSION:
-    return Info.writeString(
-        getInfoString({"CUDA Driver Version", "HSA Runtime Version"}));
-  case OL_DEVICE_INFO_MAX_WORK_GROUP_SIZE:
-    return Info.write(getInfoXyz({"Workgroup Max Size per Dimension" /*AMD*/,
-                                  "Maximum Block Dimensions" /*CUDA*/}));
-  default:
-    return createOffloadError(ErrorCode::INVALID_ENUMERATION,
-                              "getDeviceInfo enum '%i' is invalid", PropName);
+  case OL_DEVICE_INFO_DRIVER_VERSION: {
+    // String values
+    if (!std::holds_alternative<std::string>(Entry->Value))
+      return makeError(ErrorCode::BACKEND_FAILURE,
+                       "plugin returned incorrect type");
+    return Info.writeString(std::get<std::string>(Entry->Value).c_str());
   }
 
-  return Error::success();
+  case OL_DEVICE_INFO_MAX_WORK_GROUP_SIZE: {
+    // {x, y, z} triples
+    ol_dimensions_t Out{0, 0, 0};
+
+    auto getField = [&](StringRef Name, uint32_t &Dest) {
+      if (auto F = Entry->get(Name)) {
+        if (!std::holds_alternative<size_t>((*F)->Value))
+          return makeError(
+              ErrorCode::BACKEND_FAILURE,
+              "plugin returned incorrect type for dimensions element");
+        Dest = std::get<size_t>((*F)->Value);
+      } else
+        return makeError(ErrorCode::BACKEND_FAILURE,
+                         "plugin didn't provide all values for dimensions");
+      return Plugin::success();
+    };
+
+    if (auto Res = getField("x", Out.x))
+      return Res;
+    if (auto Res = getField("y", Out.y))
+      return Res;
+    if (auto Res = getField("z", Out.z))
+      return Res;
+
+    return Info.write(Out);
+  }
+
+  default:
+    llvm_unreachable("Unimplemented device info");
+  }
 }
 
 Error olGetDeviceInfoImplDetailHost(ol_device_handle_t Device,
diff --git a/offload/plugins-nextgen/common/include/JIT.h b/offload/plugins-nextgen/common/include/JIT.h
index 8c53043..d62516d 100644
--- a/offload/plugins-nextgen/common/include/JIT.h
+++ b/offload/plugins-nextgen/common/include/JIT.h
@@ -55,6 +55,10 @@ struct JITEngine {
   process(const __tgt_device_image &Image,
           target::plugin::GenericDeviceTy &Device);
 
+  /// Remove \p Image from the jit engine's cache
+  void erase(const __tgt_device_image &Image,
+             target::plugin::GenericDeviceTy &Device);
+
 private:
   /// Compile the bitcode image \p Image and generate the binary image that can
   /// be loaded to the target device of the triple \p Triple architecture \p
@@ -89,11 +93,13 @@ private:
     /// LLVM Context in which the modules will be constructed.
     LLVMContext Context;
 
-    /// Output images generated from LLVM backend.
-    SmallVector<std::unique_ptr<MemoryBuffer>, 4> JITImages;
+    /// A map of embedded IR images to the buffer used to store JITed code
+    DenseMap<const __tgt_device_image *, std::unique_ptr<MemoryBuffer>>
+        JITImages;
 
     /// A map of embedded IR images to JITed images.
-    DenseMap<const __tgt_device_image *, __tgt_device_image *> TgtImageMap;
+    DenseMap<const __tgt_device_image *, std::unique_ptr<__tgt_device_image>>
+        TgtImageMap;
   };
 
   /// Map from (march) "CPUs" (e.g., sm_80, or gfx90a), which we call compute
diff --git a/offload/plugins-nextgen/common/src/JIT.cpp b/offload/plugins-nextgen/common/src/JIT.cpp
index c82a06e..00720fa 100644
--- a/offload/plugins-nextgen/common/src/JIT.cpp
+++ b/offload/plugins-nextgen/common/src/JIT.cpp
@@ -285,8 +285,8 @@ JITEngine::compile(const __tgt_device_image &Image,
 
   // Check if we JITed this image for the given compute unit kind before.
   ComputeUnitInfo &CUI = ComputeUnitMap[ComputeUnitKind];
-  if (__tgt_device_image *JITedImage = CUI.TgtImageMap.lookup(&Image))
-    return JITedImage;
+  if (CUI.TgtImageMap.contains(&Image))
+    return CUI.TgtImageMap[&Image].get();
 
   auto ObjMBOrErr = getOrCreateObjFile(Image, CUI.Context, ComputeUnitKind);
   if (!ObjMBOrErr)
@@ -296,17 +296,15 @@ JITEngine::compile(const __tgt_device_image &Image,
   if (!ImageMBOrErr)
     return ImageMBOrErr.takeError();
 
-  CUI.JITImages.push_back(std::move(*ImageMBOrErr));
-  __tgt_device_image *&JITedImage = CUI.TgtImageMap[&Image];
-  JITedImage = new __tgt_device_image();
+  CUI.JITImages.insert({&Image, std::move(*ImageMBOrErr)});
+  auto &ImageMB = CUI.JITImages[&Image];
+  CUI.TgtImageMap.insert({&Image, std::make_unique<__tgt_device_image>()});
+  auto &JITedImage = CUI.TgtImageMap[&Image];
   *JITedImage = Image;
-
-  auto &ImageMB = CUI.JITImages.back();
-
   JITedImage->ImageStart = const_cast<char *>(ImageMB->getBufferStart());
   JITedImage->ImageEnd = const_cast<char *>(ImageMB->getBufferEnd());
 
-  return JITedImage;
+  return JITedImage.get();
 }
 
 Expected<const __tgt_device_image *>
@@ -324,3 +322,13 @@ JITEngine::process(const __tgt_device_image &Image,
 
   return &Image;
 }
+
+void JITEngine::erase(const __tgt_device_image &Image,
+                      target::plugin::GenericDeviceTy &Device) {
+  std::lock_guard<std::mutex> Lock(ComputeUnitMapMutex);
+  const std::string &ComputeUnitKind = Device.getComputeUnitKind();
+  ComputeUnitInfo &CUI = ComputeUnitMap[ComputeUnitKind];
+
+  CUI.TgtImageMap.erase(&Image);
+  CUI.JITImages.erase(&Image);
+}
diff --git a/offload/plugins-nextgen/common/src/PluginInterface.cpp b/offload/plugins-nextgen/common/src/PluginInterface.cpp
index 81b9d42..94a050b5 100644
--- a/offload/plugins-nextgen/common/src/PluginInterface.cpp
+++ b/offload/plugins-nextgen/common/src/PluginInterface.cpp
@@ -854,6 +854,9 @@ Error GenericDeviceTy::unloadBinary(DeviceImageTy *Image) {
       return Err;
   }
 
+  if (Image->getTgtImageBitcode())
+    Plugin.getJIT().erase(*Image->getTgtImageBitcode(), Image->getDevice());
+
   return unloadBinaryImpl(Image);
 }
 
diff --git a/offload/test/offloading/fortran/optional-mapped-arguments-3.f90 b/offload/test/offloading/fortran/optional-mapped-arguments-3.f90
new file mode 100644
index 0000000..cf82a85
--- /dev/null
+++ b/offload/test/offloading/fortran/optional-mapped-arguments-3.f90
@@ -0,0 +1,45 @@
+! OpenMP offloading test that checks we do not cause a segfault when mapping
+! optional function arguments (present or otherwise). No results requiring
+! checking other than that the program compiles and runs to completion with no
+! error. This particular variation checks that we're correctly emitting the
+! load/store in both branches mapping the input array.
+! REQUIRES: flang, amdgpu
+
+! RUN: %libomptarget-compile-fortran-run-and-check-generic
+module reproducer_mod
+contains
+   subroutine branching_target_call(dt, switch)
+      implicit none
+      real(4), dimension(:), intent(inout) :: dt
+      logical, intent(in) :: switch
+      integer :: dim, idx
+
+      dim = size(dt)
+      if (switch) then
+!$omp target teams distribute parallel do
+         do idx = 1, dim
+            dt(idx) = 20
+         end do
+      else
+!$omp target teams distribute parallel do
+         do idx = 1, dim
+            dt(idx) = 30
+         end do
+      end if
+   end subroutine branching_target_call
+end module reproducer_mod
+
+program reproducer
+   use reproducer_mod
+   implicit none
+   real(4), dimension(:), allocatable :: dt
+   integer :: n = 21312
+   integer :: i
+
+   allocate (dt(n))
+   call branching_target_call(dt, .FALSE.)
+   call branching_target_call(dt, .TRUE.)
+   print *, "PASSED"
+end program reproducer
+
+! CHECK: PASSED
diff --git a/offload/test/offloading/shared_lib_fp_mapping.c b/offload/test/offloading/shared_lib_fp_mapping.c
index ffb4bc9..c620344 100644
--- a/offload/test/offloading/shared_lib_fp_mapping.c
+++ b/offload/test/offloading/shared_lib_fp_mapping.c
@@ -1,6 +1,7 @@
 // clang-format off
-// RUN: %clang-generic -fPIC -shared %S/../Inputs/declare_indirect_func.c -o %T/libslfm.so  -fopenmp-version=51
-// RUN: %libomptarget-compile-generic -rpath %T -L %T -l slfm -o %t  -fopenmp-version=51
+// RUN: mkdir -p %t.testdir
+// RUN: %clang-generic -fPIC -shared %S/../Inputs/declare_indirect_func.c -o %t.testdir/libslfm.so  -fopenmp-version=51
+// RUN: %libomptarget-compile-generic -rpath %t.testdir -L %t.testdir -l slfm -o %t  -fopenmp-version=51
 // RUN: env LIBOMPTARGET_INFO=32 %t 2>&1 | %fcheck-generic
 // clang-format on
 
diff --git a/openmp/CMakeLists.txt b/openmp/CMakeLists.txt
index c206386..ab34851 100644
--- a/openmp/CMakeLists.txt
+++ b/openmp/CMakeLists.txt
@@ -115,7 +115,7 @@ option(OPENMP_ENABLE_LIBOMPTARGET "Enable building libomptarget for offloading."
 option(OPENMP_ENABLE_LIBOMP_PROFILING "Enable time profiling for libomp." OFF)
 
 # Header install location
-if(${OPENMP_STANDALONE_BUILD})
+if(NOT LLVM_TREE_AVAILABLE)
   set(LIBOMP_HEADERS_INSTALL_PATH "${CMAKE_INSTALL_INCLUDEDIR}")
 else()
   include(GetClangResourceDir)
diff --git a/openmp/runtime/src/CMakeLists.txt b/openmp/runtime/src/CMakeLists.txt
index 698e185..569061c 100644
--- a/openmp/runtime/src/CMakeLists.txt
+++ b/openmp/runtime/src/CMakeLists.txt
@@ -11,7 +11,7 @@
 include(ExtendPath)
 
 # The generated headers will be placed in clang's resource directory if present.
-if(OPENMP_STANDALONE_BUILD OR NOT LLVM_RUNTIMES_BUILD)
+if(NOT LLVM_TREE_AVAILABLE)
   set(LIBOMP_HEADERS_INTDIR ${CMAKE_CURRENT_BINARY_DIR})
 else()
   set(LIBOMP_HEADERS_INTDIR ${LLVM_BINARY_DIR}/${LIBOMP_HEADERS_INSTALL_PATH})
diff --git a/openmp/runtime/test/lit.cfg b/openmp/runtime/test/lit.cfg
index ca32985..4a5aff2 100644
--- a/openmp/runtime/test/lit.cfg
+++ b/openmp/runtime/test/lit.cfg
@@ -217,11 +217,11 @@ if config.has_ompt:
         config.substitutions.append(("%preload-tool", "true ||"))
         config.substitutions.append(("%no-as-needed-flag", "-Wl,--no-as-needed"))
     elif config.operating_system == 'Darwin':
-        config.substitutions.append(("%preload-tool", "env DYLD_INSERT_LIBRARIES=%T/tool.so"))
+        config.substitutions.append(("%preload-tool", "env DYLD_INSERT_LIBRARIES=%t.tool_dir/tool.so"))
         # No such linker flag on Darwin.
         config.substitutions.append(("%no-as-needed-flag", ""))
     else:
-        config.substitutions.append(("%preload-tool", "env LD_PRELOAD=%T/tool.so"))
+        config.substitutions.append(("%preload-tool", "env LD_PRELOAD=%t.tool_dir/tool.so"))
         config.substitutions.append(("%no-as-needed-flag", "-Wl,--no-as-needed"))
 else:
     config.substitutions.append(("FileCheck", config.test_filecheck))
diff --git a/openmp/runtime/test/ompt/loadtool/tool_available/tool_available.c b/openmp/runtime/test/ompt/loadtool/tool_available.c
index ba6ce32..a08581b 100644
--- a/openmp/runtime/test/ompt/loadtool/tool_available/tool_available.c
+++ b/openmp/runtime/test/ompt/loadtool/tool_available.c
@@ -1,5 +1,7 @@
 // The OpenMP standard defines 3 ways of providing ompt_start_tool:
 
+// RUN: mkdir -p %t.tool_dir
+
 // 1. "statically-linking the tool’s definition of ompt_start_tool into an 
 //     OpenMP application"
 
@@ -9,20 +11,20 @@
 // Note: We should compile the tool without -fopenmp as other tools developer
 //      would do. Otherwise this test may pass for the wrong reasons on Darwin.
 
-// RUN: %clang %flags -DTOOL -shared -fPIC %s -o %T/tool.so
+// RUN: %clang %flags -DTOOL -shared -fPIC %s -o %t.tool_dir/tool.so
 
 // 2. "introducing a dynamically-linked library that includes the tool’s 
 //    definition of ompt_start_tool into the application’s address space"
 
 // 2.1 Link with tool during compilation
 
-// RUN: %libomp-compile -DCODE %no-as-needed-flag %T/tool.so && \
+// RUN: %libomp-compile -DCODE %no-as-needed-flag %t.tool_dir/tool.so && \
 // RUN:    env OMP_TOOL_VERBOSE_INIT=stdout %libomp-run | FileCheck %s \
 // RUN:    --check-prefixes CHECK,ADDRSPACE 
 
 // 2.2 Link with tool during compilation, but AFTER the runtime
 
-// RUN: %libomp-compile -DCODE -lomp %no-as-needed-flag %T/tool.so && \
+// RUN: %libomp-compile -DCODE -lomp %no-as-needed-flag %t.tool_dir/tool.so && \
 // RUN:    env OMP_TOOL_VERBOSE_INIT=stdout %libomp-run | FileCheck %s \
 // RUN:    --check-prefixes CHECK,ADDRSPACE
 
@@ -39,36 +41,36 @@
 // 3.1 OMP_TOOL_VERBOSE_INIT not set 
 
 // RUN: %libomp-compile -DCODE && \
-// RUN:    env OMP_TOOL_LIBRARIES=%T/tool.so %libomp-run | FileCheck %s
+// RUN:    env OMP_TOOL_LIBRARIES=%t.tool_dir/tool.so %libomp-run | FileCheck %s
 
 // 3.2 OMP_TOOL_VERBOSE_INIT disabled
 
-// RUN: env OMP_TOOL_LIBRARIES=%T/tool.so OMP_TOOL_VERBOSE_INIT=disabled \
+// RUN: env OMP_TOOL_LIBRARIES=%t.tool_dir/tool.so OMP_TOOL_VERBOSE_INIT=disabled \
 // RUN:    %libomp-run | FileCheck %s
 
 // 3.3 OMP_TOOL_VERBOSE_INIT to stdout
 
-// RUN: %libomp-compile -DCODE && env OMP_TOOL_LIBRARIES=%T/tool.so \
+// RUN: %libomp-compile -DCODE && env OMP_TOOL_LIBRARIES=%t.tool_dir/tool.so \
 // RUN:    OMP_TOOL_VERBOSE_INIT=stdout %libomp-run | \
-// RUN:    FileCheck %s -DPARENTPATH=%T --check-prefixes CHECK,TOOLLIB
+// RUN:    FileCheck %s -DPARENTPATH=%t.tool_dir --check-prefixes CHECK,TOOLLIB
 
 // 3.4 OMP_TOOL_VERBOSE_INIT to stderr, check merged stdout and stderr
 
-// RUN: env OMP_TOOL_LIBRARIES=%T/tool.so OMP_TOOL_VERBOSE_INIT=stderr \
+// RUN: env OMP_TOOL_LIBRARIES=%t.tool_dir/tool.so OMP_TOOL_VERBOSE_INIT=stderr \
 // RUN:    %libomp-run 2>&1 | \
-// RUN:    FileCheck %s -DPARENTPATH=%T --check-prefixes CHECK,TOOLLIB
+// RUN:    FileCheck %s -DPARENTPATH=%t.tool_dir --check-prefixes CHECK,TOOLLIB
 
 // 3.5 OMP_TOOL_VERBOSE_INIT to stderr, check just stderr
 
-// RUN: env OMP_TOOL_LIBRARIES=%T/tool.so OMP_TOOL_VERBOSE_INIT=stderr \
+// RUN: env OMP_TOOL_LIBRARIES=%t.tool_dir/tool.so OMP_TOOL_VERBOSE_INIT=stderr \
 // RUN:    %libomp-run 2>&1 >/dev/null | \
-// RUN:    FileCheck %s -DPARENTPATH=%T --check-prefixes TOOLLIB
+// RUN:    FileCheck %s -DPARENTPATH=%t.tool_dir --check-prefixes TOOLLIB
 
 // 3.6 OMP_TOOL_VERBOSE_INIT to file "init.log"
 
-// RUN: env OMP_TOOL_LIBRARIES=%T/tool.so OMP_TOOL_VERBOSE_INIT=%T/init.log \
-// RUN:    %libomp-run | FileCheck %s && cat %T/init.log | \
-// RUN:    FileCheck %s -DPARENTPATH=%T --check-prefixes TOOLLIB
+// RUN: env OMP_TOOL_LIBRARIES=%t.tool_dir/tool.so OMP_TOOL_VERBOSE_INIT=%t.tool_dir/init.log \
+// RUN:    %libomp-run | FileCheck %s && cat %t.tool_dir/init.log | \
+// RUN:    FileCheck %s -DPARENTPATH=%t.tool_dir --check-prefixes TOOLLIB
 
 
 // REQUIRES: ompt
diff --git a/openmp/runtime/test/ompt/loadtool/tool_available_search/tool_available_search.c b/openmp/runtime/test/ompt/loadtool/tool_available_search.c
index cf1fd33..73e2dea 100644
--- a/openmp/runtime/test/ompt/loadtool/tool_available_search/tool_available_search.c
+++ b/openmp/runtime/test/ompt/loadtool/tool_available_search.c
@@ -1,9 +1,10 @@
-// RUN: %clang %flags -shared -fPIC %s -o %T/first_tool.so
-// RUN: %clang %flags -DTOOL -DSECOND_TOOL -shared -fPIC %s -o %T/second_tool.so
-// RUN: %clang %flags -DTOOL -DTHIRD_TOOL -shared -fPIC %s -o %T/third_tool.so
+// RUN: mkdir -p %t.tool_dir
+// RUN: %clang %flags -shared -fPIC %s -o %t.tool_dir/first_tool.so
+// RUN: %clang %flags -DTOOL -DSECOND_TOOL -shared -fPIC %s -o %t.tool_dir/second_tool.so
+// RUN: %clang %flags -DTOOL -DTHIRD_TOOL -shared -fPIC %s -o %t.tool_dir/third_tool.so
 // RUN: %libomp-compile -DCODE
-// RUN: env OMP_TOOL_LIBRARIES=%T/non_existing_file.so:%T/first_tool.so:%T/second_tool.so:%T/third_tool.so \
-// RUN: OMP_TOOL_VERBOSE_INIT=stdout %libomp-run | FileCheck %s -DPARENTPATH=%T
+// RUN: env OMP_TOOL_LIBRARIES=%t.tool_dir/non_existing_file.so:%t.tool_dir/first_tool.so:%t.tool_dir/second_tool.so:%t.tool_dir/third_tool.so \
+// RUN: OMP_TOOL_VERBOSE_INIT=stdout %libomp-run | FileCheck %s -DPARENTPATH=%t.tool_dir
 
 // REQUIRES: ompt
 // XFAIL: darwin
diff --git a/openmp/runtime/test/ompt/loadtool/tool_not_available/tool_not_available.c b/openmp/runtime/test/ompt/loadtool/tool_not_available.c
index 7e68d8b..df56d31 100644
--- a/openmp/runtime/test/ompt/loadtool/tool_not_available/tool_not_available.c
+++ b/openmp/runtime/test/ompt/loadtool/tool_not_available.c
@@ -1,5 +1,7 @@
 // The OpenMP standard defines 3 ways of providing ompt_start_tool:
 
+// RUN: mkdir -p %t.tool_dir
+
 // 1. "statically-linking the tool’s definition of ompt_start_tool into an
 // OpenMP application"
 
@@ -10,20 +12,20 @@
 // Note: We should compile the tool without -fopenmp as other tools developer
 //      would do. Otherwise this test may pass for the wrong reasons on Darwin.
 
-// RUN: %clang %flags -DTOOL -shared -fPIC %s -o %T/tool.so
+// RUN: %clang %flags -DTOOL -shared -fPIC %s -o %t.tool_dir/tool.so
 
 // 2. "introducing a dynamically-linked library that includes the tool’s 
 //    definition of ompt_start_tool into the application’s address space"
 
 // 2.1 Link with tool during compilation
 
-// RUN: %libomp-compile -DCODE %no-as-needed-flag %T/tool.so && \
+// RUN: %libomp-compile -DCODE %no-as-needed-flag %t.tool_dir/tool.so && \
 // RUN:    env OMP_TOOL_VERBOSE_INIT=stdout %libomp-run | \
 // RUN:    FileCheck %s --check-prefixes CHECK,ADDRSPACE
 
 // 2.2 Link with tool during compilation, but AFTER the runtime
 
-// RUN: %libomp-compile -DCODE -lomp %no-as-needed-flag %T/tool.so && \
+// RUN: %libomp-compile -DCODE -lomp %no-as-needed-flag %t.tool_dir/tool.so && \
 // RUN:    env OMP_TOOL_VERBOSE_INIT=stdout %libomp-run | \
 // RUN:    FileCheck %s --check-prefixes CHECK,ADDRSPACE 
 
@@ -37,9 +39,9 @@
 //    architecture and operating system used by the application in the 
 //    tool-libraries-var ICV"
 
-// RUN: %libomp-compile -DCODE && env OMP_TOOL_LIBRARIES=%T/tool.so \
+// RUN: %libomp-compile -DCODE && env OMP_TOOL_LIBRARIES=%t.tool_dir/tool.so \
 // RUN:    OMP_TOOL_VERBOSE_INIT=stdout %libomp-run | \
-// RUN:    FileCheck %s -DPARENTPATH=%T --check-prefixes CHECK,TOOLLIB
+// RUN:    FileCheck %s -DPARENTPATH=%t.tool_dir --check-prefixes CHECK,TOOLLIB
 
 // REQUIRES: ompt
 
diff --git a/openmp/tools/archer/tests/lit.cfg b/openmp/tools/archer/tests/lit.cfg
index c4bae61..786b2d9 100644
--- a/openmp/tools/archer/tests/lit.cfg
+++ b/openmp/tools/archer/tests/lit.cfg
@@ -130,10 +130,3 @@ config.substitutions.append(("%deflake", os.path.join(os.path.dirname(__file__),
 config.substitutions.append(("FileCheck", config.test_filecheck))
 config.substitutions.append(("%not", config.test_not))
 config.substitutions.append(("%sort-threads", "sort --numeric-sort --stable"))
-if config.operating_system == 'Windows':
-    # No such environment variable on Windows.
-    config.substitutions.append(("%preload-tool", "true ||"))
-elif config.operating_system == 'Darwin':
-    config.substitutions.append(("%preload-tool", "env DYLD_INSERT_LIBRARIES=%T/tool.so"))
-else:
-    config.substitutions.append(("%preload-tool", "env LD_PRELOAD=%T/tool.so"))
diff --git a/runtimes/CMakeLists.txt b/runtimes/CMakeLists.txt
index e4dd4eb..d0d2a53 100644
--- a/runtimes/CMakeLists.txt
+++ b/runtimes/CMakeLists.txt
@@ -92,6 +92,14 @@ include(LLVMCheckCompilerLinkerFlag)
 include(CheckCCompilerFlag)
 include(CheckCXXCompilerFlag)
 
+
+# Determine whether we are in the runtimes/runtimes-bins directory of a
+# bootstrap build.
+set(LLVM_TREE_AVAILABLE OFF)
+if (LLVM_LIBRARY_OUTPUT_INTDIR AND LLVM_RUNTIME_OUTPUT_INTDIR AND PACKAGE_VERSION)
+  set(LLVM_TREE_AVAILABLE ON)
+endif()
+
 # CMake omits default compiler include paths, but in runtimes build, we use
 # -nostdinc and -nostdinc++ and control include paths manually so this behavior
 # is undesirable. Filtering CMAKE_{LANG}_IMPLICIT_INCLUDE_DIRECTORIES to remove
diff --git a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
index 38df25f..d424026 100644
--- a/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/libc/BUILD.bazel
@@ -19,6 +19,7 @@ load(":platforms.bzl", "PLATFORM_CPU_ARM64", "PLATFORM_CPU_X86_64")
 package(
     default_visibility = ["//visibility:public"],
     features = [
+        "parse_headers",
         "-use_header_modules",
         "-layering_check",
     ],
@@ -158,6 +159,7 @@ libc_support_library(
     hdrs = ["include/llvm-libc-types/float128.h"],
     target_compatible_with = select({
         "@platforms//os:linux": [],
+        "@platforms//os:windows": [],
         "//conditions:default": ["@platforms//:incompatible"],
     }),
     deps = [":llvm_libc_macros_float_macros"],
@@ -2184,6 +2186,21 @@ libc_support_library(
 )
 
 libc_support_library(
+    name = "__support_math_acospif16",
+    hdrs = ["src/__support/math/acospif16.h"],
+    deps = [
+        ":__support_fputil_cast",
+        ":__support_fputil_fenv_impl",
+        ":__support_fputil_fp_bits",
+        ":__support_fputil_multiply_add",
+        ":__support_fputil_polyeval",
+        ":__support_fputil_sqrt",
+        ":__support_macros_optimization",
+        ":__support_macros_properties_types",
+    ],
+)
+
+libc_support_library(
     name = "__support_math_asin_utils",
     hdrs = ["src/__support/math/asin_utils.h"],
     deps = [
@@ -2198,6 +2215,52 @@ libc_support_library(
 )
 
 libc_support_library(
+    name = "__support_math_asin",
+    hdrs = ["src/__support/math/asin.h"],
+    deps = [
+        ":__support_math_asin_utils",
+        ":__support_fputil_double_double",
+        ":__support_fputil_dyadic_float",
+        ":__support_fputil_fenv_impl",
+        ":__support_fputil_fp_bits",
+        ":__support_fputil_multiply_add",
+        ":__support_fputil_polyeval",
+        ":__support_fputil_sqrt",
+        ":__support_macros_optimization",
+        ":__support_macros_properties_cpu_features",
+    ],
+)
+
+libc_support_library(
+    name = "__support_math_asinf",
+    hdrs = ["src/__support/math/asinf.h"],
+    deps = [
+        ":__support_math_inv_trigf_utils",
+        ":__support_fputil_fenv_impl",
+        ":__support_fputil_fp_bits",
+        ":__support_fputil_except_value_utils",
+        ":__support_fputil_multiply_add",
+        ":__support_fputil_sqrt",
+        ":__support_macros_config",
+        ":__support_macros_optimization",
+        ":__support_macros_properties_cpu_features",
+    ],
+)
+
+libc_support_library(
+    name = "__support_math_asinf16",
+    hdrs = ["src/__support/math/asinf16.h"],
+    deps = [
+        ":__support_fputil_fenv_impl",
+        ":__support_fputil_fp_bits",
+        ":__support_fputil_polyeval",
+        ":__support_fputil_multiply_add",
+        ":__support_fputil_sqrt",
+        ":__support_macros_optimization",
+    ],
+)
+
+libc_support_library(
     name = "__support_math_erff",
     hdrs = ["src/__support/math/erff.h"],
     deps = [
@@ -2735,30 +2798,31 @@ libc_math_function(
 )
 
 libc_math_function(
+    name = "acospif16",
+    additional_deps = [
+        ":__support_math_acospif16",
+        ":errno",
+    ],
+)
+
+libc_math_function(
+    name = "asin",
+    additional_deps = [
+        ":__support_math_asin",
+    ],
+)
+
+libc_math_function(
     name = "asinf",
     additional_deps = [
-        ":__support_fputil_fma",
-        ":__support_fputil_multiply_add",
-        ":__support_fputil_nearest_integer",
-        ":__support_fputil_polyeval",
-        ":__support_fputil_sqrt",
-        ":__support_macros_optimization",
-        ":__support_macros_properties_cpu_features",
-        ":__support_math_inv_trigf_utils",
+        ":__support_math_asinf",
     ],
 )
 
 libc_math_function(
     name = "asinf16",
     additional_deps = [
-        ":__support_fputil_cast",
-        ":__support_fputil_fma",
-        ":__support_fputil_multiply_add",
-        ":__support_fputil_nearest_integer",
-        ":__support_fputil_polyeval",
-        ":__support_fputil_sqrt",
-        ":__support_macros_optimization",
-        ":__support_math_inv_trigf_utils",
+        ":__support_math_asinf16",
     ],
 )
 
diff --git a/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel b/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel
index 826fb03..178c3ce 100644
--- a/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/mlir/BUILD.bazel
@@ -3057,14 +3057,14 @@ cc_library(
 )
 
 ##---------------------------------------------------------------------------##
-# Mesh Dialect
+# Shard Dialect
 ##---------------------------------------------------------------------------##
 
 td_library(
-    name = "MeshTdFiles",
+    name = "ShardTdFiles",
     srcs = [
-        "include/mlir/Dialect/Mesh/IR/MeshBase.td",
-        "include/mlir/Dialect/Mesh/IR/MeshOps.td",
+        "include/mlir/Dialect/Shard/IR/ShardBase.td",
+        "include/mlir/Dialect/Shard/IR/ShardOps.td",
     ],
     includes = ["include"],
     deps = [
@@ -3076,92 +3076,92 @@ td_library(
 )
 
 gentbl_cc_library(
-    name = "MeshIncGen",
+    name = "ShardIncGen",
     tbl_outs = {
-        "include/mlir/Dialect/Mesh/IR/MeshOps.h.inc": [
+        "include/mlir/Dialect/Shard/IR/ShardOps.h.inc": [
             "-gen-op-decls",
-            "-dialect=mesh",
+            "-dialect=shard",
         ],
-        "include/mlir/Dialect/Mesh/IR/MeshOps.cpp.inc": [
+        "include/mlir/Dialect/Shard/IR/ShardOps.cpp.inc": [
             "-gen-op-defs",
-            "-dialect=mesh",
+            "-dialect=shard",
         ],
-        "include/mlir/Dialect/Mesh/IR/MeshDialect.h.inc": [
+        "include/mlir/Dialect/Shard/IR/ShardDialect.h.inc": [
             "-gen-dialect-decls",
-            "-dialect=mesh",
+            "-dialect=shard",
         ],
-        "include/mlir/Dialect/Mesh/IR/MeshDialect.cpp.inc": [
+        "include/mlir/Dialect/Shard/IR/ShardDialect.cpp.inc": [
             "-gen-dialect-defs",
-            "-dialect=mesh",
+            "-dialect=shard",
         ],
-        "include/mlir/Dialect/Mesh/IR/MeshEnums.h.inc": [
+        "include/mlir/Dialect/Shard/IR/ShardEnums.h.inc": [
             "-gen-enum-decls",
-            "-dialect=mesh",
+            "-dialect=shard",
         ],
-        "include/mlir/Dialect/Mesh/IR/MeshEnums.cpp.inc": [
+        "include/mlir/Dialect/Shard/IR/ShardEnums.cpp.inc": [
             "-gen-enum-defs",
-            "-dialect=mesh",
+            "-dialect=shard",
         ],
-        "include/mlir/Dialect/Mesh/IR/MeshAttributes.h.inc": [
+        "include/mlir/Dialect/Shard/IR/ShardAttributes.h.inc": [
             "-gen-attrdef-decls",
-            "-dialect=mesh",
+            "-dialect=shard",
         ],
-        "include/mlir/Dialect/Mesh/IR/MeshAttributes.cpp.inc": [
+        "include/mlir/Dialect/Shard/IR/ShardAttributes.cpp.inc": [
             "-gen-attrdef-defs",
-            "-dialect=mesh",
+            "-dialect=shard",
         ],
-        "include/mlir/Dialect/Mesh/IR/MeshTypes.h.inc": [
+        "include/mlir/Dialect/Shard/IR/ShardTypes.h.inc": [
             "-gen-typedef-decls",
-            "-typedefs-dialect=mesh",
+            "-typedefs-dialect=shard",
         ],
-        "include/mlir/Dialect/Mesh/IR/MeshTypes.cpp.inc": [
+        "include/mlir/Dialect/Shard/IR/ShardTypes.cpp.inc": [
             "-gen-typedef-defs",
-            "-typedefs-dialect=mesh",
+            "-typedefs-dialect=shard",
         ],
     },
     tblgen = ":mlir-tblgen",
-    td_file = "include/mlir/Dialect/Mesh/IR/MeshOps.td",
+    td_file = "include/mlir/Dialect/Shard/IR/ShardOps.td",
     deps = [
-        ":MeshTdFiles",
         ":ShapeOpsTdFiles",
+        ":ShardTdFiles",
     ],
 )
 
 gentbl_cc_library(
-    name = "MeshShardingInterfaceIncGen",
+    name = "ShardingInterfaceIncGen",
     tbl_outs = {
-        "include/mlir/Dialect/Mesh/Interfaces/ShardingInterface.h.inc": ["-gen-op-interface-decls"],
-        "include/mlir/Dialect/Mesh/Interfaces/ShardingInterface.cpp.inc": ["-gen-op-interface-defs"],
+        "include/mlir/Dialect/Shard/Interfaces/ShardingInterface.h.inc": ["-gen-op-interface-decls"],
+        "include/mlir/Dialect/Shard/Interfaces/ShardingInterface.cpp.inc": ["-gen-op-interface-defs"],
     },
     tblgen = ":mlir-tblgen",
-    td_file = "include/mlir/Dialect/Mesh/Interfaces/ShardingInterface.td",
+    td_file = "include/mlir/Dialect/Shard/Interfaces/ShardingInterface.td",
     deps = [":OpBaseTdFiles"],
 )
 
 cc_library(
-    name = "MeshShardingInterface",
-    srcs = ["lib/Dialect/Mesh/Interfaces/ShardingInterface.cpp"],
+    name = "ShardingInterface",
+    srcs = ["lib/Dialect/Shard/Interfaces/ShardingInterface.cpp"],
     hdrs = [
-        "include/mlir/Dialect/Mesh/Interfaces/ShardingInterface.h",
-        "include/mlir/Dialect/Mesh/Interfaces/ShardingInterfaceImpl.h",
+        "include/mlir/Dialect/Shard/Interfaces/ShardingInterface.h",
+        "include/mlir/Dialect/Shard/Interfaces/ShardingInterfaceImpl.h",
     ],
     includes = ["include"],
     deps = [
         ":DialectUtils",
         ":IR",
-        ":MeshDialect",
-        ":MeshShardingInterfaceIncGen",
+        ":ShardDialect",
+        ":ShardingInterfaceIncGen",
         ":Support",
         "//llvm:Support",
     ],
 )
 
 cc_library(
-    name = "MeshDialect",
-    srcs = ["lib/Dialect/Mesh/IR/MeshOps.cpp"],
+    name = "ShardDialect",
+    srcs = ["lib/Dialect/Shard/IR/ShardOps.cpp"],
     hdrs = [
-        "include/mlir/Dialect/Mesh/IR/MeshDialect.h",
-        "include/mlir/Dialect/Mesh/IR/MeshOps.h",
+        "include/mlir/Dialect/Shard/IR/ShardDialect.h",
+        "include/mlir/Dialect/Shard/IR/ShardOps.h",
     ],
     includes = ["include"],
     deps = [
@@ -3172,7 +3172,7 @@ cc_library(
         ":IR",
         ":InferTypeOpInterface",
         ":InliningUtils",
-        ":MeshIncGen",
+        ":ShardIncGen",
         ":SideEffectInterfaces",
         ":Support",
         ":ViewLikeInterface",
@@ -3181,23 +3181,23 @@ cc_library(
 )
 
 gentbl_cc_library(
-    name = "MeshTransformsPassIncGen",
-    tbl_outs = {"include/mlir/Dialect/Mesh/Transforms/Passes.h.inc": [
+    name = "ShardTransformsPassIncGen",
+    tbl_outs = {"include/mlir/Dialect/Shard/Transforms/Passes.h.inc": [
         "-gen-pass-decls",
-        "-name=Mesh",
+        "-name=Shard",
     ]},
     tblgen = ":mlir-tblgen",
-    td_file = "include/mlir/Dialect/Mesh/Transforms/Passes.td",
+    td_file = "include/mlir/Dialect/Shard/Transforms/Passes.td",
     deps = [":PassBaseTdFiles"],
 )
 
 cc_library(
-    name = "MeshTransforms",
+    name = "ShardTransforms",
     srcs = glob([
-        "lib/Dialect/Mesh/Transforms/*.cpp",
-        "lib/Dialect/Mesh/Transforms/*.h",
+        "lib/Dialect/Shard/Transforms/*.cpp",
+        "lib/Dialect/Shard/Transforms/*.h",
     ]),
-    hdrs = glob(["include/mlir/Dialect/Mesh/Transforms/*.h"]),
+    hdrs = glob(["include/mlir/Dialect/Shard/Transforms/*.h"]),
     includes = ["include"],
     deps = [
         ":AffineDialect",
@@ -3210,10 +3210,10 @@ cc_library(
         ":FuncDialect",
         ":FunctionInterfaces",
         ":IR",
-        ":MeshDialect",
-        ":MeshShardingInterface",
-        ":MeshTransformsPassIncGen",
         ":Pass",
+        ":ShardDialect",
+        ":ShardTransformsPassIncGen",
+        ":ShardingInterface",
         ":Support",
         ":TensorDialect",
         ":Transforms",
@@ -3222,11 +3222,11 @@ cc_library(
 )
 
 cc_library(
-    name = "MeshToMPIConversion",
+    name = "ShardToMPIConversion",
     srcs = glob([
-        "lib/Conversion/MeshToMPI/*.cpp",
+        "lib/Conversion/ShardToMPI/*.cpp",
     ]),
-    hdrs = glob(["include/mlir/Conversion/MeshToMPI/*.h"]),
+    hdrs = glob(["include/mlir/Conversion/ShardToMPI/*.h"]),
     includes = ["include"],
     deps = [
         ":AffineDialect",
@@ -3241,10 +3241,10 @@ cc_library(
         ":LinalgDialect",
         ":MPIDialect",
         ":MemRefDialect",
-        ":MeshDialect",
-        ":MeshTransforms",
         ":Pass",
         ":SCFDialect",
+        ":ShardDialect",
+        ":ShardTransforms",
         ":Support",
         ":TensorDialect",
         ":TransformUtils",
@@ -3989,7 +3989,6 @@ cc_library(
         ":MemRefToEmitC",
         ":MemRefToLLVM",
         ":MemRefToSPIRV",
-        ":MeshToMPIConversion",
         ":NVGPUToNVVM",
         ":NVVMToLLVM",
         ":OpenACCToSCF",
@@ -4003,6 +4002,7 @@ cc_library(
         ":SCFToSPIRV",
         ":SPIRVToLLVM",
         ":ShapeToStandard",
+        ":ShardToMPIConversion",
         ":TensorToLinalg",
         ":TensorToSPIRV",
         ":TosaToArith",
@@ -4523,7 +4523,7 @@ cc_library(
         ":FuncDialect",
         ":IR",
         ":InliningUtils",
-        ":MeshShardingInterface",
+        ":ShardingInterface",
     ],
 )
 
@@ -4622,7 +4622,6 @@ cc_library(
         ":MemRefToEmitC",
         ":MemRefToLLVM",
         ":MemRefTransformOps",
-        ":MeshDialect",
         ":NVGPUTransformOps",
         ":NVVMTarget",
         ":NVVMToLLVM",
@@ -4631,6 +4630,7 @@ cc_library(
         ":ROCDLToLLVMIRTranslation",
         ":SCFToEmitC",
         ":SCFTransformOps",
+        ":ShardDialect",
         ":SparseTensorTransformOps",
         ":TensorExtensions",
         ":TensorTransformOps",
@@ -5119,10 +5119,7 @@ cc_library(
         ":SideEffectInterfaces",
         ":Support",
         ":ViewLikeInterface",
-        "//llvm:AsmParser",
         "//llvm:BinaryFormat",
-        "//llvm:BitReader",
-        "//llvm:BitWriter",
         "//llvm:Core",
         "//llvm:Support",
     ],
@@ -7195,7 +7192,7 @@ cc_library(
     includes = ["include"],
     deps = [
         ":IR",
-        ":MeshShardingInterface",
+        ":ShardingInterface",
         ":TensorDialect",
         "//llvm:Support",
     ],
@@ -9020,8 +9017,6 @@ cc_library(
         ":MemRefToSPIRV",
         ":MemRefTransformOps",
         ":MemRefTransforms",
-        ":MeshDialect",
-        ":MeshTransforms",
         ":NVGPUDialect",
         ":NVGPUPassIncGen",
         ":NVGPUToNVVM",
@@ -9058,6 +9053,8 @@ cc_library(
         ":ShapeToStandard",
         ":ShapeTransforms",
         ":ShapeTransformsPassIncGen",
+        ":ShardDialect",
+        ":ShardTransforms",
         ":SparseTensorDialect",
         ":SparseTensorPipelines",
         ":SparseTensorTransformOps",
@@ -9121,7 +9118,6 @@ cc_binary(
         "//mlir/test:TestMath",
         "//mlir/test:TestMathToVCIX",
         "//mlir/test:TestMemRef",
-        "//mlir/test:TestMesh",
         "//mlir/test:TestNVGPU",
         "//mlir/test:TestPDLL",
         "//mlir/test:TestPass",
@@ -9130,6 +9126,7 @@ cc_binary(
         "//mlir/test:TestSCF",
         "//mlir/test:TestSPIRV",
         "//mlir/test:TestShapeDialect",
+        "//mlir/test:TestShard",
         "//mlir/test:TestTensor",
         "//mlir/test:TestTestDynDialect",
         "//mlir/test:TestTilingInterface",
@@ -9183,7 +9180,6 @@ cc_binary(
         "//mlir/test:TestMathToVCIX",
         "//mlir/test:TestMemRef",
         "//mlir/test:TestMemRefToLLVMWithTransforms",
-        "//mlir/test:TestMesh",
         "//mlir/test:TestNVGPU",
         "//mlir/test:TestPDLL",
         "//mlir/test:TestPass",
@@ -9192,6 +9188,7 @@ cc_binary(
         "//mlir/test:TestSCF",
         "//mlir/test:TestSPIRV",
         "//mlir/test:TestShapeDialect",
+        "//mlir/test:TestShard",
         "//mlir/test:TestTensor",
         "//mlir/test:TestTestDynDialect",
         "//mlir/test:TestTilingInterface",
@@ -10549,9 +10546,9 @@ cc_library(
         ":LinalgStructuredOpsIncGen",
         ":MathDialect",
         ":MemRefDialect",
-        ":MeshShardingInterface",
         ":Parser",
         ":SCFDialect",
+        ":ShardingInterface",
         ":SideEffectInterfaces",
         ":SparseTensorDialect",
         ":SubsetOpInterface",
@@ -10700,14 +10697,14 @@ cc_library(
         ":MathDialect",
         ":MemRefDialect",
         ":MemRefTransforms",
-        ":MeshDialect",
-        ":MeshShardingInterface",
-        ":MeshTransforms",
         ":Pass",
         ":RuntimeVerifiableOpInterface",
         ":SCFDialect",
         ":SCFTransforms",
         ":SCFUtils",
+        ":ShardDialect",
+        ":ShardTransforms",
+        ":ShardingInterface",
         ":SparseTensorDialect",
         ":SubsetOpInterface",
         ":Support",
@@ -10989,6 +10986,7 @@ cc_library(
         ":VectorTransforms",
         ":X86VectorDialect",
         ":X86VectorTransforms",
+        "//llvm:Core",
         "//llvm:Support",
     ],
 )
@@ -11199,10 +11197,10 @@ cc_library(
         ":InferTypeOpInterface",
         ":InliningUtils",
         ":LoopLikeInterface",
-        ":MeshDialect",
-        ":MeshShardingInterface",
         ":Pass",
         ":QuantOps",
+        ":ShardDialect",
+        ":ShardingInterface",
         ":SideEffectInterfaces",
         ":Support",
         ":TensorDialect",
@@ -12142,8 +12140,8 @@ cc_library(
         ":FuncTransforms",
         ":IR",
         ":MemRefDialect",
-        ":MeshShardingInterface",
         ":Pass",
+        ":ShardingInterface",
         ":SideEffectInterfaces",
         ":TensorDialect",
         ":TransformUtils",
diff --git a/utils/bazel/llvm-project-overlay/mlir/test/BUILD.bazel b/utils/bazel/llvm-project-overlay/mlir/test/BUILD.bazel
index 95e3ee4..e7770fc 100644
--- a/utils/bazel/llvm-project-overlay/mlir/test/BUILD.bazel
+++ b/utils/bazel/llvm-project-overlay/mlir/test/BUILD.bazel
@@ -903,8 +903,8 @@ cc_library(
 )
 
 cc_library(
-    name = "TestMesh",
-    srcs = glob(["lib/Dialect/Mesh/**/*.cpp"]),
+    name = "TestShard",
+    srcs = glob(["lib/Dialect/Shard/**/*.cpp"]),
     includes = ["lib/Dialect/Test"],
     deps = [
         ":TestDialect",
@@ -912,8 +912,8 @@ cc_library(
         "//mlir:DialectUtils",
         "//mlir:FuncDialect",
         "//mlir:IR",
-        "//mlir:MeshDialect",
-        "//mlir:MeshTransforms",
+        "//mlir:ShardDialect",
+        "//mlir:ShardTransforms",
         "//mlir:Pass",
         "//mlir:SPIRVDialect",
         "//mlir:Support",
diff --git a/utils/bazel/llvm_configs/llvm-config.h.cmake b/utils/bazel/llvm_configs/llvm-config.h.cmake
index a683229..39136bc 100644
--- a/utils/bazel/llvm_configs/llvm-config.h.cmake
+++ b/utils/bazel/llvm_configs/llvm-config.h.cmake
@@ -101,6 +101,9 @@
 /* Define if LLVM is using tflite */
 #cmakedefine LLVM_HAVE_TFLITE
 
+/* Define if we want to check profile consistency in lit tests */
+#cmakedefine LLVM_ENABLE_PROFCHECK
+
 /* Define to 1 if you have the <sysexits.h> header file. */
 #cmakedefine HAVE_SYSEXITS_H ${HAVE_SYSEXITS_H}